blob: 016b3c4ab99a90a570672c941e3f9f1c143acdd7 [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,
Herbert Xu79671682006-06-22 02:40:14 -0700595};
596
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700597#if BITS_PER_LONG > 32
598#define NET_SKBUFF_DATA_USES_OFFSET 1
599#endif
600
601#ifdef NET_SKBUFF_DATA_USES_OFFSET
602typedef unsigned int sk_buff_data_t;
603#else
604typedef unsigned char *sk_buff_data_t;
605#endif
606
Pedro Tammela161e6132019-03-05 11:35:54 -0300607/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 * struct sk_buff - socket buffer
609 * @next: Next buffer in list
610 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800611 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800612 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700613 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700615 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000616 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700617 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 * @len: Length of actual data
619 * @data_len: Data length
620 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700621 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700622 * @csum: Checksum (must include start/offset pair)
623 * @csum_start: Offset from skb->head where checksumming should start
624 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700625 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700626 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700628 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 * @nohdr: Payload reference only, must not modify header
630 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700631 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700632 * @ipvs_property: skbuff is owned by ipvs
Ido Schimmel875e8932018-12-04 08:15:10 +0000633 * @offload_fwd_mark: Packet was L2-forwarded in hardware
634 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
Willem de Bruijne7246e12017-01-07 17:06:35 -0500635 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500636 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500637 * @tc_redirected: packet was redirected by a tc action
638 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800639 * @peeked: this packet has been seen already, so stats have been
640 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700641 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700642 * @protocol: Packet protocol from driver
643 * @destructor: Destruct function
Eric Dumazete2080072017-10-04 12:59:58 -0700644 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
Florian Westphala9e419d2017-01-23 18:21:59 +0100645 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800647 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700649 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700650 * @queue_mapping: Queue mapping for multiqueue devices
Stefano Brivio8b700862018-07-11 14:39:42 +0200651 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
Florian Westphaldf5042f2018-12-18 17:15:16 +0100652 * @active_extensions: active extensions (skb_ext_id types)
Randy Dunlap553a5672008-04-20 10:51:01 -0700653 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700654 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700655 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700656 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700657 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100658 * @wifi_acked_valid: wifi_acked was set
659 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000660 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200661 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200662 * @dst_pending_confirm: need to confirm neighbour
Stefano Brivioa48d1892018-07-17 11:52:57 +0200663 * @decrypted: Decrypted SKB
Pedro Tammela161e6132019-03-05 11:35:54 -0300664 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700665 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700666 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000667 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700668 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000669 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000670 * @inner_transport_header: Inner transport layer header (encapsulation)
671 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000672 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700673 * @transport_header: Transport layer header
674 * @network_header: Network layer header
675 * @mac_header: Link layer header
676 * @tail: Tail pointer
677 * @end: End pointer
678 * @head: Head of buffer
679 * @data: Data head pointer
680 * @truesize: Buffer size
681 * @users: User count - see {datagram,tcp}.c
Florian Westphaldf5042f2018-12-18 17:15:16 +0100682 * @extensions: allocated extensions, valid if active_extensions is nonzero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 */
684
685struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800686 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800687 struct {
688 /* These two members must be first. */
689 struct sk_buff *next;
690 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000691
Eric Dumazet56b17422014-11-03 08:19:53 -0800692 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700693 struct net_device *dev;
694 /* Some protocols might use this space to store information,
695 * while device pointer would be NULL.
696 * UDP receive path is one user.
697 */
698 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800699 };
700 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000701 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
David Millerd4546c22018-06-24 14:13:49 +0900702 struct list_head list;
Eric Dumazet56b17422014-11-03 08:19:53 -0800703 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000704
705 union {
706 struct sock *sk;
707 int ip_defrag_offset;
708 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
Eric Dumazetc84d9492016-12-08 11:41:55 -0800710 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700711 ktime_t tstamp;
Eric Dumazetd3edd062018-09-21 08:51:50 -0700712 u64 skb_mstamp_ns; /* earliest departure time */
Eric Dumazetc84d9492016-12-08 11:41:55 -0800713 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 /*
715 * This is the control buffer. It is free to use for every
716 * layer. Please put your private variables there. If you
717 * want to keep them across layers you have to do a skb_clone()
718 * first. This is owned by whoever has the skb queued ATM.
719 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000720 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
Eric Dumazete2080072017-10-04 12:59:58 -0700722 union {
723 struct {
724 unsigned long _skb_refdst;
725 void (*destructor)(struct sk_buff *skb);
726 };
727 struct list_head tcp_tsorted_anchor;
728 };
729
Eric Dumazetb1937222014-09-28 22:18:47 -0700730#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100731 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700732#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700734 data_len;
735 __u16 mac_len,
736 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700737
738 /* Following fields are _not_ copied in __copy_skb_header()
739 * Note that queue_mapping is here mostly to fill a hole.
740 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700741 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200742
743/* if you move cloned around you also must adapt those constants */
744#ifdef __BIG_ENDIAN_BITFIELD
745#define CLONED_MASK (1 << 7)
746#else
747#define CLONED_MASK 1
748#endif
749#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
750
751 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700752 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700753 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800754 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800755 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700756 head_frag:1,
Stefano Brivio8b700862018-07-11 14:39:42 +0200757 pfmemalloc:1;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100758#ifdef CONFIG_SKB_EXTENSIONS
759 __u8 active_extensions;
760#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700761 /* fields enclosed in headers_start/headers_end are copied
762 * using a single memcpy() in __copy_skb_header()
763 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700764 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700765 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700766 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200767
768/* if you move pkt_type around you also must adapt those constants */
769#ifdef __BIG_ENDIAN_BITFIELD
770#define PKT_TYPE_MAX (7 << 5)
771#else
772#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100773#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200774#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
775
776 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700777 __u8 pkt_type:3;
Eric Dumazetb1937222014-09-28 22:18:47 -0700778 __u8 ignore_df:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700779 __u8 nf_trace:1;
780 __u8 ip_summed:2;
781 __u8 ooo_okay:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200782
Eric Dumazetb1937222014-09-28 22:18:47 -0700783 __u8 l4_hash:1;
784 __u8 sw_hash:1;
785 __u8 wifi_acked_valid:1;
786 __u8 wifi_acked:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700787 __u8 no_fcs:1;
788 /* Indicates the inner headers are valid in the skbuff. */
789 __u8 encapsulation:1;
790 __u8 encap_hdr_csum:1;
791 __u8 csum_valid:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200792
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100793#ifdef __BIG_ENDIAN_BITFIELD
794#define PKT_VLAN_PRESENT_BIT 7
795#else
796#define PKT_VLAN_PRESENT_BIT 0
797#endif
798#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
799 __u8 __pkt_vlan_present_offset[0];
800 __u8 vlan_present:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700801 __u8 csum_complete_sw:1;
802 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200803 __u8 csum_not_inet:1;
Julian Anastasov4ff06202017-02-06 23:14:12 +0200804 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700805#ifdef CONFIG_IPV6_NDISC_NODETYPE
806 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807#endif
Stefano Brivio8b700862018-07-11 14:39:42 +0200808
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100809 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700810 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800811 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200812#ifdef CONFIG_NET_SWITCHDEV
813 __u8 offload_fwd_mark:1;
Ido Schimmel875e8932018-12-04 08:15:10 +0000814 __u8 offload_l3_fwd_mark:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200815#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500816#ifdef CONFIG_NET_CLS_ACT
817 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500818 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500819 __u8 tc_redirected:1;
820 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500821#endif
Stefano Brivioa48d1892018-07-17 11:52:57 +0200822#ifdef CONFIG_TLS_DEVICE
823 __u8 decrypted:1;
824#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700827 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200829
Eric Dumazetb1937222014-09-28 22:18:47 -0700830 union {
831 __wsum csum;
832 struct {
833 __u16 csum_start;
834 __u16 csum_offset;
835 };
836 };
837 __u32 priority;
838 int skb_iif;
839 __u32 hash;
840 __be16 vlan_proto;
841 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800842#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
843 union {
844 unsigned int napi_id;
845 unsigned int sender_cpu;
846 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700847#endif
James Morris984bc162006-06-09 00:29:17 -0700848#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200849 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700850#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700851
Neil Horman3b885782009-10-12 13:26:31 -0700852 union {
853 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000854 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700855 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
Tom Herbert8bce6d72014-09-29 20:22:29 -0700857 union {
858 __be16 inner_protocol;
859 __u8 inner_ipproto;
860 };
861
Simon Horman1a37e412013-05-23 21:02:51 +0000862 __u16 inner_transport_header;
863 __u16 inner_network_header;
864 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700865
866 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000867 __u16 transport_header;
868 __u16 network_header;
869 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700870
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700871 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700872 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700873 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700874
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700876 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700877 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700879 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700880 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300881 refcount_t users;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100882
883#ifdef CONFIG_SKB_EXTENSIONS
884 /* only useable after checking ->active_extensions != 0 */
885 struct skb_ext *extensions;
886#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887};
888
889#ifdef __KERNEL__
890/*
891 * Handling routines are only of interest to the kernel
892 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
Mel Gormanc93bdd02012-07-31 16:44:19 -0700894#define SKB_ALLOC_FCLONE 0x01
895#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800896#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700897
Pedro Tammela161e6132019-03-05 11:35:54 -0300898/**
899 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
900 * @skb: buffer
901 */
Mel Gormanc93bdd02012-07-31 16:44:19 -0700902static inline bool skb_pfmemalloc(const struct sk_buff *skb)
903{
904 return unlikely(skb->pfmemalloc);
905}
906
Eric Dumazet7fee2262010-05-11 23:19:48 +0000907/*
908 * skb might have a dst pointer attached, refcounted or not.
909 * _skb_refdst low order bit is set if refcount was _not_ taken
910 */
911#define SKB_DST_NOREF 1UL
912#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
913
914/**
915 * skb_dst - returns skb dst_entry
916 * @skb: buffer
917 *
918 * Returns skb dst_entry, regardless of reference taken or not.
919 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000920static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
921{
Pedro Tammela161e6132019-03-05 11:35:54 -0300922 /* If refdst was not refcounted, check we still are in a
Eric Dumazet7fee2262010-05-11 23:19:48 +0000923 * rcu_read_lock section
924 */
925 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
926 !rcu_read_lock_held() &&
927 !rcu_read_lock_bh_held());
928 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000929}
930
Eric Dumazet7fee2262010-05-11 23:19:48 +0000931/**
932 * skb_dst_set - sets skb dst
933 * @skb: buffer
934 * @dst: dst entry
935 *
936 * Sets skb dst, assuming a reference was taken on dst and should
937 * be released by skb_dst_drop()
938 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000939static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
940{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000941 skb->_skb_refdst = (unsigned long)dst;
942}
943
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200944/**
945 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
946 * @skb: buffer
947 * @dst: dst entry
948 *
949 * Sets skb dst, assuming a reference was not taken on dst.
950 * If dst entry is cached, we do not take reference and dst_release
951 * will be avoided by refdst_drop. If dst entry is not cached, we take
952 * reference, so that last dst_release can destroy the dst immediately.
953 */
954static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
955{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100956 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
957 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200958}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000959
960/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300961 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000962 * @skb: buffer
963 */
964static inline bool skb_dst_is_noref(const struct sk_buff *skb)
965{
966 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000967}
968
Pedro Tammela161e6132019-03-05 11:35:54 -0300969/**
970 * skb_rtable - Returns the skb &rtable
971 * @skb: buffer
972 */
Eric Dumazet511c3f92009-06-02 05:14:27 +0000973static inline struct rtable *skb_rtable(const struct sk_buff *skb)
974{
Eric Dumazetadf30902009-06-02 05:19:30 +0000975 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000976}
977
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400978/* For mangling skb->pkt_type from user space side from applications
979 * such as nft, tc, etc, we only allow a conservative subset of
980 * possible pkt_types to be set.
981*/
982static inline bool skb_pkt_type_ok(u32 ptype)
983{
984 return ptype <= PACKET_OTHERHOST;
985}
986
Pedro Tammela161e6132019-03-05 11:35:54 -0300987/**
988 * skb_napi_id - Returns the skb's NAPI id
989 * @skb: buffer
990 */
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200991static inline unsigned int skb_napi_id(const struct sk_buff *skb)
992{
993#ifdef CONFIG_NET_RX_BUSY_POLL
994 return skb->napi_id;
995#else
996 return 0;
997#endif
998}
999
Pedro Tammela161e6132019-03-05 11:35:54 -03001000/**
1001 * skb_unref - decrement the skb's reference count
1002 * @skb: buffer
1003 *
1004 * Returns true if we can free the skb.
1005 */
Paolo Abeni3889a8032017-06-12 11:23:41 +02001006static inline bool skb_unref(struct sk_buff *skb)
1007{
1008 if (unlikely(!skb))
1009 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +03001010 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001011 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +03001012 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001013 return false;
1014
1015 return true;
1016}
1017
Paolo Abeni0a463c72017-06-12 11:23:42 +02001018void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001019void kfree_skb(struct sk_buff *skb);
1020void kfree_skb_list(struct sk_buff *segs);
Willem de Bruijn64131392019-07-07 05:51:55 -04001021void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
Joe Perches7965bd42013-09-26 14:48:15 -07001022void skb_tx_error(struct sk_buff *skb);
1023void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +02001024void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001025void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +00001026extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001027
Joe Perches7965bd42013-09-26 14:48:15 -07001028void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1029bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1030 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001031
Joe Perches7965bd42013-09-26 14:48:15 -07001032struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1033 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001034struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -07001035struct sk_buff *build_skb(void *data, unsigned int frag_size);
Jesper Dangaard Brouerba0509b2019-04-12 17:07:37 +02001036struct sk_buff *build_skb_around(struct sk_buff *skb,
1037 void *data, unsigned int frag_size);
Pedro Tammela161e6132019-03-05 11:35:54 -03001038
1039/**
1040 * alloc_skb - allocate a network buffer
1041 * @size: size to allocate
1042 * @priority: allocation mask
1043 *
1044 * This function is a convenient wrapper around __alloc_skb().
1045 */
David S. Millerd179cd12005-08-17 14:57:30 -07001046static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001047 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001048{
Eric Dumazet564824b2010-10-11 19:05:25 +00001049 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001050}
1051
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001052struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1053 unsigned long data_len,
1054 int max_page_order,
1055 int *errcode,
1056 gfp_t gfp_mask);
Jakub Kicinskida29e4b2019-06-03 15:16:58 -07001057struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001058
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001059/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1060struct sk_buff_fclones {
1061 struct sk_buff skb1;
1062
1063 struct sk_buff skb2;
1064
Reshetova, Elena26385952017-06-30 13:07:59 +03001065 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001066};
1067
1068/**
1069 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -07001070 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001071 * @skb: buffer
1072 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001073 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001074 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1075 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001076 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001077static inline bool skb_fclone_busy(const struct sock *sk,
1078 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001079{
1080 const struct sk_buff_fclones *fclones;
1081
1082 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1083
1084 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001085 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001086 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001087}
1088
Pedro Tammela161e6132019-03-05 11:35:54 -03001089/**
1090 * alloc_skb_fclone - allocate a network buffer from fclone cache
1091 * @size: size to allocate
1092 * @priority: allocation mask
1093 *
1094 * This function is a convenient wrapper around __alloc_skb().
1095 */
David S. Millerd179cd12005-08-17 14:57:30 -07001096static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001097 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001098{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001099 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001100}
1101
Joe Perches7965bd42013-09-26 14:48:15 -07001102struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Toshiaki Makitab0768a82018-08-03 16:58:09 +09001103void skb_headers_offset_update(struct sk_buff *skb, int off);
Joe Perches7965bd42013-09-26 14:48:15 -07001104int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1105struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
Ilya Lesokhin08303c12018-04-30 10:16:11 +03001106void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
Joe Perches7965bd42013-09-26 14:48:15 -07001107struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001108struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1109 gfp_t gfp_mask, bool fclone);
1110static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1111 gfp_t gfp_mask)
1112{
1113 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1114}
Eric Dumazet117632e2011-12-03 21:39:53 +00001115
Joe Perches7965bd42013-09-26 14:48:15 -07001116int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1117struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1118 unsigned int headroom);
1119struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1120 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001121int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1122 int offset, int len);
1123int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1124 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001125int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001126int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1127
1128/**
1129 * skb_pad - zero pad the tail of an skb
1130 * @skb: buffer to pad
1131 * @pad: space to pad
1132 *
1133 * Ensure that a buffer is followed by a padding area that is zero
1134 * filled. Used by network drivers which may DMA or transfer data
1135 * beyond the buffer end onto the wire.
1136 *
1137 * May return error in out of memory cases. The skb is freed on error.
1138 */
1139static inline int skb_pad(struct sk_buff *skb, int pad)
1140{
1141 return __skb_pad(skb, pad, true);
1142}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001143#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001145int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1146 int offset, size_t size);
1147
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001148struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001149 __u32 lower_offset;
1150 __u32 upper_offset;
1151 __u32 frag_idx;
1152 __u32 stepped_offset;
1153 struct sk_buff *root_skb;
1154 struct sk_buff *cur_skb;
1155 __u8 *frag_data;
1156};
1157
Joe Perches7965bd42013-09-26 14:48:15 -07001158void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1159 unsigned int to, struct skb_seq_state *st);
1160unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1161 struct skb_seq_state *st);
1162void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001163
Joe Perches7965bd42013-09-26 14:48:15 -07001164unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001165 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001166
Tom Herbert09323cc2013-12-15 22:16:19 -08001167/*
1168 * Packet hash types specify the type of hash in skb_set_hash.
1169 *
1170 * Hash types refer to the protocol layer addresses which are used to
1171 * construct a packet's hash. The hashes are used to differentiate or identify
1172 * flows of the protocol layer for the hash type. Hash types are either
1173 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1174 *
1175 * Properties of hashes:
1176 *
1177 * 1) Two packets in different flows have different hash values
1178 * 2) Two packets in the same flow should have the same hash value
1179 *
1180 * A hash at a higher layer is considered to be more specific. A driver should
1181 * set the most specific hash possible.
1182 *
1183 * A driver cannot indicate a more specific hash than the layer at which a hash
1184 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1185 *
1186 * A driver may indicate a hash level which is less specific than the
1187 * actual layer the hash was computed on. For instance, a hash computed
1188 * at L4 may be considered an L3 hash. This should only be done if the
1189 * driver can't unambiguously determine that the HW computed the hash at
1190 * the higher layer. Note that the "should" in the second property above
1191 * permits this.
1192 */
1193enum pkt_hash_types {
1194 PKT_HASH_TYPE_NONE, /* Undefined type */
1195 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1196 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1197 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1198};
1199
Tom Herbertbcc83832015-09-01 09:24:24 -07001200static inline void skb_clear_hash(struct sk_buff *skb)
1201{
1202 skb->hash = 0;
1203 skb->sw_hash = 0;
1204 skb->l4_hash = 0;
1205}
1206
1207static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1208{
1209 if (!skb->l4_hash)
1210 skb_clear_hash(skb);
1211}
1212
1213static inline void
1214__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1215{
1216 skb->l4_hash = is_l4;
1217 skb->sw_hash = is_sw;
1218 skb->hash = hash;
1219}
1220
Tom Herbert09323cc2013-12-15 22:16:19 -08001221static inline void
1222skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1223{
Tom Herbertbcc83832015-09-01 09:24:24 -07001224 /* Used by drivers to set hash from HW */
1225 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1226}
1227
1228static inline void
1229__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1230{
1231 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001232}
1233
Tom Herberte5276932015-09-01 09:24:23 -07001234void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001235u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001236u32 skb_get_poff(const struct sk_buff *skb);
1237u32 __skb_get_poff(const struct sk_buff *skb, void *data,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001238 const struct flow_keys_basic *keys, int hlen);
Tom Herberte5276932015-09-01 09:24:23 -07001239__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1240 void *data, int hlen_proto);
1241
1242static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1243 int thoff, u8 ip_proto)
1244{
1245 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1246}
1247
1248void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1249 const struct flow_dissector_key *key,
1250 unsigned int key_count);
1251
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001252#ifdef CONFIG_NET
Stanislav Fomichev118c8e92019-04-25 14:37:23 -07001253int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1254 union bpf_attr __user *uattr);
Petar Penkovd58e4682018-09-14 07:46:18 -07001255int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1256 struct bpf_prog *prog);
1257
1258int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001259#else
Stanislav Fomichev118c8e92019-04-25 14:37:23 -07001260static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1261 union bpf_attr __user *uattr)
1262{
1263 return -EOPNOTSUPP;
1264}
1265
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001266static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1267 struct bpf_prog *prog)
1268{
1269 return -EOPNOTSUPP;
1270}
1271
1272static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1273{
1274 return -EOPNOTSUPP;
1275}
1276#endif
Petar Penkovd58e4682018-09-14 07:46:18 -07001277
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001278struct bpf_flow_dissector;
1279bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
Stanislav Fomichev086f9562019-07-25 15:52:25 -07001280 __be16 proto, int nhoff, int hlen, unsigned int flags);
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001281
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001282bool __skb_flow_dissect(const struct net *net,
1283 const struct sk_buff *skb,
Tom Herberte5276932015-09-01 09:24:23 -07001284 struct flow_dissector *flow_dissector,
1285 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001286 void *data, __be16 proto, int nhoff, int hlen,
1287 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001288
1289static inline bool skb_flow_dissect(const struct sk_buff *skb,
1290 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001291 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001292{
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001293 return __skb_flow_dissect(NULL, skb, flow_dissector,
1294 target_container, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001295}
1296
1297static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001298 struct flow_keys *flow,
1299 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001300{
1301 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001302 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1303 flow, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001304}
1305
Paolo Abeni72a338b2018-05-04 11:32:59 +02001306static inline bool
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001307skb_flow_dissect_flow_keys_basic(const struct net *net,
1308 const struct sk_buff *skb,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001309 struct flow_keys_basic *flow, void *data,
1310 __be16 proto, int nhoff, int hlen,
1311 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001312{
1313 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001314 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001315 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001316}
1317
Jiri Pirko82828b82019-06-19 09:41:02 +03001318void skb_flow_dissect_meta(const struct sk_buff *skb,
1319 struct flow_dissector *flow_dissector,
1320 void *target_container);
1321
Paul Blakey75a56752019-07-09 10:30:49 +03001322/* Gets a skb connection tracking info, ctinfo map should be a
1323 * a map of mapsize to translate enum ip_conntrack_info states
1324 * to user states.
1325 */
1326void
1327skb_flow_dissect_ct(const struct sk_buff *skb,
1328 struct flow_dissector *flow_dissector,
1329 void *target_container,
1330 u16 *ctinfo_map,
1331 size_t mapsize);
Simon Horman62b32372017-12-04 11:31:48 +01001332void
1333skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1334 struct flow_dissector *flow_dissector,
1335 void *target_container);
1336
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001337static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001338{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001339 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001340 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001341
Tom Herbert61b905d2014-03-24 15:34:47 -07001342 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001343}
1344
David S. Miller20a17bf2015-09-01 21:19:17 -07001345static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001346{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001347 if (!skb->l4_hash && !skb->sw_hash) {
1348 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001349 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001350
Tom Herbertde4c1f82015-09-01 18:11:04 -07001351 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001352 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001353
1354 return skb->hash;
1355}
1356
Eric Dumazet55667442019-10-22 07:57:46 -07001357__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1358 const siphash_key_t *perturb);
Tom Herbert50fb7992015-05-01 11:30:12 -07001359
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001360static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1361{
Tom Herbert61b905d2014-03-24 15:34:47 -07001362 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001363}
1364
Tom Herbert3df7a742013-12-15 22:16:29 -08001365static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1366{
Tom Herbert61b905d2014-03-24 15:34:47 -07001367 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001368 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001369 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001370};
1371
Jakub Kicinski41477662019-08-07 17:03:59 -07001372static inline void skb_copy_decrypted(struct sk_buff *to,
1373 const struct sk_buff *from)
1374{
1375#ifdef CONFIG_TLS_DEVICE
1376 to->decrypted = from->decrypted;
1377#endif
1378}
1379
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001380#ifdef NET_SKBUFF_DATA_USES_OFFSET
1381static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1382{
1383 return skb->head + skb->end;
1384}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001385
1386static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1387{
1388 return skb->end;
1389}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001390#else
1391static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1392{
1393 return skb->end;
1394}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001395
1396static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1397{
1398 return skb->end - skb->head;
1399}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001400#endif
1401
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001403#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404
Patrick Ohlyac45f602009-02-12 05:03:37 +00001405static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1406{
1407 return &skb_shinfo(skb)->hwtstamps;
1408}
1409
Willem de Bruijn52267792017-08-03 16:29:39 -04001410static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1411{
1412 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1413
1414 return is_zcopy ? skb_uarg(skb) : NULL;
1415}
1416
Willem de Bruijn52900d22018-11-30 15:32:40 -05001417static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1418 bool *have_ref)
Willem de Bruijn52267792017-08-03 16:29:39 -04001419{
1420 if (skb && uarg && !skb_zcopy(skb)) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001421 if (unlikely(have_ref && *have_ref))
1422 *have_ref = false;
1423 else
1424 sock_zerocopy_get(uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -04001425 skb_shinfo(skb)->destructor_arg = uarg;
1426 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1427 }
1428}
1429
Willem de Bruijn5cd8d462018-11-20 13:00:18 -05001430static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1431{
1432 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1433 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1434}
1435
1436static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1437{
1438 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1439}
1440
1441static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1442{
1443 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1444}
1445
Willem de Bruijn52267792017-08-03 16:29:39 -04001446/* Release a reference on a zerocopy structure */
1447static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1448{
1449 struct ubuf_info *uarg = skb_zcopy(skb);
1450
1451 if (uarg) {
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001452 if (skb_zcopy_is_nouarg(skb)) {
1453 /* no notification callback */
1454 } else if (uarg->callback == sock_zerocopy_callback) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001455 uarg->zerocopy = uarg->zerocopy && zerocopy;
1456 sock_zerocopy_put(uarg);
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001457 } else {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001458 uarg->callback(uarg, zerocopy);
1459 }
1460
Willem de Bruijn52267792017-08-03 16:29:39 -04001461 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1462 }
1463}
1464
1465/* Abort a zerocopy operation and revert zckey on error in send syscall */
1466static inline void skb_zcopy_abort(struct sk_buff *skb)
1467{
1468 struct ubuf_info *uarg = skb_zcopy(skb);
1469
1470 if (uarg) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001471 sock_zerocopy_put_abort(uarg, false);
Willem de Bruijn52267792017-08-03 16:29:39 -04001472 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1473 }
1474}
1475
David S. Millera8305bf2018-07-29 20:42:53 -07001476static inline void skb_mark_not_on_list(struct sk_buff *skb)
1477{
1478 skb->next = NULL;
1479}
1480
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001481/* Iterate through singly-linked GSO fragments of an skb. */
1482#define skb_list_walk_safe(first, skb, next) \
1483 for ((skb) = (first), (next) = (skb) ? (skb)->next : NULL; (skb); \
1484 (skb) = (next), (next) = (skb) ? (skb)->next : NULL)
1485
David S. Miller992cba72018-07-31 15:27:56 -07001486static inline void skb_list_del_init(struct sk_buff *skb)
1487{
1488 __list_del_entry(&skb->list);
1489 skb_mark_not_on_list(skb);
1490}
1491
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492/**
1493 * skb_queue_empty - check if a queue is empty
1494 * @list: queue head
1495 *
1496 * Returns true if the queue is empty, false otherwise.
1497 */
1498static inline int skb_queue_empty(const struct sk_buff_head *list)
1499{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001500 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501}
1502
1503/**
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001504 * skb_queue_empty_lockless - check if a queue is empty
1505 * @list: queue head
1506 *
1507 * Returns true if the queue is empty, false otherwise.
1508 * This variant can be used in lockless contexts.
1509 */
1510static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1511{
1512 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1513}
1514
1515
1516/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001517 * skb_queue_is_last - check if skb is the last entry in the queue
1518 * @list: queue head
1519 * @skb: buffer
1520 *
1521 * Returns true if @skb is the last buffer on the list.
1522 */
1523static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1524 const struct sk_buff *skb)
1525{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001526 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001527}
1528
1529/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001530 * skb_queue_is_first - check if skb is the first entry in the queue
1531 * @list: queue head
1532 * @skb: buffer
1533 *
1534 * Returns true if @skb is the first buffer on the list.
1535 */
1536static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1537 const struct sk_buff *skb)
1538{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001539 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001540}
1541
1542/**
David S. Miller249c8b42008-09-23 00:44:42 -07001543 * skb_queue_next - return the next packet in the queue
1544 * @list: queue head
1545 * @skb: current buffer
1546 *
1547 * Return the next packet in @list after @skb. It is only valid to
1548 * call this if skb_queue_is_last() evaluates to false.
1549 */
1550static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1551 const struct sk_buff *skb)
1552{
1553 /* This BUG_ON may seem severe, but if we just return then we
1554 * are going to dereference garbage.
1555 */
1556 BUG_ON(skb_queue_is_last(list, skb));
1557 return skb->next;
1558}
1559
1560/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001561 * skb_queue_prev - return the prev packet in the queue
1562 * @list: queue head
1563 * @skb: current buffer
1564 *
1565 * Return the prev packet in @list before @skb. It is only valid to
1566 * call this if skb_queue_is_first() evaluates to false.
1567 */
1568static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1569 const struct sk_buff *skb)
1570{
1571 /* This BUG_ON may seem severe, but if we just return then we
1572 * are going to dereference garbage.
1573 */
1574 BUG_ON(skb_queue_is_first(list, skb));
1575 return skb->prev;
1576}
1577
1578/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 * skb_get - reference buffer
1580 * @skb: buffer to reference
1581 *
1582 * Makes another reference to a socket buffer and returns a pointer
1583 * to the buffer.
1584 */
1585static inline struct sk_buff *skb_get(struct sk_buff *skb)
1586{
Reshetova, Elena63354792017-06-30 13:07:58 +03001587 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 return skb;
1589}
1590
1591/*
Geert Uytterhoevenf8821f92017-11-30 14:33:56 +01001592 * If users == 1, we are the only owner and can avoid redundant atomic changes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 */
1594
1595/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596 * skb_cloned - is the buffer a clone
1597 * @skb: buffer to check
1598 *
1599 * Returns true if the buffer was generated with skb_clone() and is
1600 * one of multiple shared copies of the buffer. Cloned buffers are
1601 * shared data so must not be written to under normal circumstances.
1602 */
1603static inline int skb_cloned(const struct sk_buff *skb)
1604{
1605 return skb->cloned &&
1606 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1607}
1608
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001609static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1610{
Mel Gormand0164ad2015-11-06 16:28:21 -08001611 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001612
1613 if (skb_cloned(skb))
1614 return pskb_expand_head(skb, 0, 0, pri);
1615
1616 return 0;
1617}
1618
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619/**
1620 * skb_header_cloned - is the header a clone
1621 * @skb: buffer to check
1622 *
1623 * Returns true if modifying the header part of the buffer requires
1624 * the data to be copied.
1625 */
1626static inline int skb_header_cloned(const struct sk_buff *skb)
1627{
1628 int dataref;
1629
1630 if (!skb->cloned)
1631 return 0;
1632
1633 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1634 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1635 return dataref != 1;
1636}
1637
Eric Dumazet9580bf22016-04-30 10:19:29 -07001638static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1639{
1640 might_sleep_if(gfpflags_allow_blocking(pri));
1641
1642 if (skb_header_cloned(skb))
1643 return pskb_expand_head(skb, 0, 0, pri);
1644
1645 return 0;
1646}
1647
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001649 * __skb_header_release - release reference to header
1650 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001651 */
1652static inline void __skb_header_release(struct sk_buff *skb)
1653{
1654 skb->nohdr = 1;
1655 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1656}
1657
1658
1659/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 * skb_shared - is the buffer shared
1661 * @skb: buffer to check
1662 *
1663 * Returns true if more than one person has a reference to this
1664 * buffer.
1665 */
1666static inline int skb_shared(const struct sk_buff *skb)
1667{
Reshetova, Elena63354792017-06-30 13:07:58 +03001668 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669}
1670
1671/**
1672 * skb_share_check - check if buffer is shared and if so clone it
1673 * @skb: buffer to check
1674 * @pri: priority for memory allocation
1675 *
1676 * If the buffer is shared the buffer is cloned and the old copy
1677 * drops a reference. A new clone with a single reference is returned.
1678 * If the buffer is not shared the original buffer is returned. When
1679 * being called from interrupt status or with spinlocks held pri must
1680 * be GFP_ATOMIC.
1681 *
1682 * NULL is returned on a memory allocation failure.
1683 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001684static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685{
Mel Gormand0164ad2015-11-06 16:28:21 -08001686 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 if (skb_shared(skb)) {
1688 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001689
1690 if (likely(nskb))
1691 consume_skb(skb);
1692 else
1693 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 skb = nskb;
1695 }
1696 return skb;
1697}
1698
1699/*
1700 * Copy shared buffers into a new sk_buff. We effectively do COW on
1701 * packets to handle cases where we have a local reader and forward
1702 * and a couple of other messy ones. The normal one is tcpdumping
1703 * a packet thats being forwarded.
1704 */
1705
1706/**
1707 * skb_unshare - make a copy of a shared buffer
1708 * @skb: buffer to check
1709 * @pri: priority for memory allocation
1710 *
1711 * If the socket buffer is a clone then this function creates a new
1712 * copy of the data, drops a reference count on the old copy and returns
1713 * the new copy with the reference count at 1. If the buffer is not a clone
1714 * the original buffer is returned. When called with a spinlock held or
1715 * from interrupt state @pri must be %GFP_ATOMIC
1716 *
1717 * %NULL is returned on a memory allocation failure.
1718 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001719static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001720 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721{
Mel Gormand0164ad2015-11-06 16:28:21 -08001722 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 if (skb_cloned(skb)) {
1724 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001725
1726 /* Free our shared copy */
1727 if (likely(nskb))
1728 consume_skb(skb);
1729 else
1730 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 skb = nskb;
1732 }
1733 return skb;
1734}
1735
1736/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001737 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 * @list_: list to peek at
1739 *
1740 * Peek an &sk_buff. Unlike most other operations you _MUST_
1741 * be careful with this one. A peek leaves the buffer on the
1742 * list and someone else may run off with it. You must hold
1743 * the appropriate locks or have a private queue to do this.
1744 *
1745 * Returns %NULL for an empty list or a pointer to the head element.
1746 * The reference count is not incremented and the reference is therefore
1747 * volatile. Use with caution.
1748 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001749static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
Eric Dumazet18d07002012-04-30 16:31:46 +00001751 struct sk_buff *skb = list_->next;
1752
1753 if (skb == (struct sk_buff *)list_)
1754 skb = NULL;
1755 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756}
1757
1758/**
David S. Miller8b69bd72018-08-11 18:43:38 -07001759 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1760 * @list_: list to peek at
1761 *
1762 * Like skb_peek(), but the caller knows that the list is not empty.
1763 */
1764static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1765{
1766 return list_->next;
1767}
1768
1769/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001770 * skb_peek_next - peek skb following the given one from a queue
1771 * @skb: skb to start from
1772 * @list_: list to peek at
1773 *
1774 * Returns %NULL when the end of the list is met or a pointer to the
1775 * next element. The reference count is not incremented and the
1776 * reference is therefore volatile. Use with caution.
1777 */
1778static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1779 const struct sk_buff_head *list_)
1780{
1781 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001782
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001783 if (next == (struct sk_buff *)list_)
1784 next = NULL;
1785 return next;
1786}
1787
1788/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001789 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 * @list_: list to peek at
1791 *
1792 * Peek an &sk_buff. Unlike most other operations you _MUST_
1793 * be careful with this one. A peek leaves the buffer on the
1794 * list and someone else may run off with it. You must hold
1795 * the appropriate locks or have a private queue to do this.
1796 *
1797 * Returns %NULL for an empty list or a pointer to the tail element.
1798 * The reference count is not incremented and the reference is therefore
1799 * volatile. Use with caution.
1800 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001801static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001803 struct sk_buff *skb = READ_ONCE(list_->prev);
Eric Dumazet18d07002012-04-30 16:31:46 +00001804
1805 if (skb == (struct sk_buff *)list_)
1806 skb = NULL;
1807 return skb;
1808
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809}
1810
1811/**
1812 * skb_queue_len - get queue length
1813 * @list_: list to measure
1814 *
1815 * Return the length of an &sk_buff queue.
1816 */
1817static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1818{
1819 return list_->qlen;
1820}
1821
David S. Miller67fed452008-09-21 22:36:24 -07001822/**
1823 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1824 * @list: queue to initialize
1825 *
1826 * This initializes only the list and queue length aspects of
1827 * an sk_buff_head object. This allows to initialize the list
1828 * aspects of an sk_buff_head without reinitializing things like
1829 * the spinlock. It can also be used for on-stack sk_buff_head
1830 * objects where the spinlock is known to not be used.
1831 */
1832static inline void __skb_queue_head_init(struct sk_buff_head *list)
1833{
1834 list->prev = list->next = (struct sk_buff *)list;
1835 list->qlen = 0;
1836}
1837
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001838/*
1839 * This function creates a split out lock class for each invocation;
1840 * this is needed for now since a whole lot of users of the skb-queue
1841 * infrastructure in drivers have different locking usage (in hardirq)
1842 * than the networking core (in softirq only). In the long run either the
1843 * network layer or drivers should need annotation to consolidate the
1844 * main types of usage into 3 classes.
1845 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846static inline void skb_queue_head_init(struct sk_buff_head *list)
1847{
1848 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001849 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850}
1851
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001852static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1853 struct lock_class_key *class)
1854{
1855 skb_queue_head_init(list);
1856 lockdep_set_class(&list->lock, class);
1857}
1858
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001860 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 *
1862 * The "__skb_xxxx()" functions are the non-atomic ones that
1863 * can only be called with interrupts disabled.
1864 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001865static inline void __skb_insert(struct sk_buff *newsk,
1866 struct sk_buff *prev, struct sk_buff *next,
1867 struct sk_buff_head *list)
1868{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001869 /* See skb_queue_empty_lockless() and skb_peek_tail()
1870 * for the opposite READ_ONCE()
1871 */
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001872 WRITE_ONCE(newsk->next, next);
1873 WRITE_ONCE(newsk->prev, prev);
1874 WRITE_ONCE(next->prev, newsk);
1875 WRITE_ONCE(prev->next, newsk);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001876 list->qlen++;
1877}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878
David S. Miller67fed452008-09-21 22:36:24 -07001879static inline void __skb_queue_splice(const struct sk_buff_head *list,
1880 struct sk_buff *prev,
1881 struct sk_buff *next)
1882{
1883 struct sk_buff *first = list->next;
1884 struct sk_buff *last = list->prev;
1885
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001886 WRITE_ONCE(first->prev, prev);
1887 WRITE_ONCE(prev->next, first);
David S. Miller67fed452008-09-21 22:36:24 -07001888
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001889 WRITE_ONCE(last->next, next);
1890 WRITE_ONCE(next->prev, last);
David S. Miller67fed452008-09-21 22:36:24 -07001891}
1892
1893/**
1894 * skb_queue_splice - join two skb lists, this is designed for stacks
1895 * @list: the new list to add
1896 * @head: the place to add it in the first list
1897 */
1898static inline void skb_queue_splice(const struct sk_buff_head *list,
1899 struct sk_buff_head *head)
1900{
1901 if (!skb_queue_empty(list)) {
1902 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001903 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001904 }
1905}
1906
1907/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001908 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001909 * @list: the new list to add
1910 * @head: the place to add it in the first list
1911 *
1912 * The list at @list is reinitialised
1913 */
1914static inline void skb_queue_splice_init(struct sk_buff_head *list,
1915 struct sk_buff_head *head)
1916{
1917 if (!skb_queue_empty(list)) {
1918 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001919 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001920 __skb_queue_head_init(list);
1921 }
1922}
1923
1924/**
1925 * skb_queue_splice_tail - join two skb lists, each list being a queue
1926 * @list: the new list to add
1927 * @head: the place to add it in the first list
1928 */
1929static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1930 struct sk_buff_head *head)
1931{
1932 if (!skb_queue_empty(list)) {
1933 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001934 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001935 }
1936}
1937
1938/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001939 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001940 * @list: the new list to add
1941 * @head: the place to add it in the first list
1942 *
1943 * Each of the lists is a queue.
1944 * The list at @list is reinitialised
1945 */
1946static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1947 struct sk_buff_head *head)
1948{
1949 if (!skb_queue_empty(list)) {
1950 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001951 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001952 __skb_queue_head_init(list);
1953 }
1954}
1955
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001957 * __skb_queue_after - queue a buffer at the list head
1958 * @list: list to use
1959 * @prev: place after this buffer
1960 * @newsk: buffer to queue
1961 *
1962 * Queue a buffer int the middle of a list. This function takes no locks
1963 * and you must therefore hold required locks before calling it.
1964 *
1965 * A buffer cannot be placed on two lists at the same time.
1966 */
1967static inline void __skb_queue_after(struct sk_buff_head *list,
1968 struct sk_buff *prev,
1969 struct sk_buff *newsk)
1970{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001971 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001972}
1973
Joe Perches7965bd42013-09-26 14:48:15 -07001974void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1975 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001976
Gerrit Renkerf5572852008-04-14 00:05:28 -07001977static inline void __skb_queue_before(struct sk_buff_head *list,
1978 struct sk_buff *next,
1979 struct sk_buff *newsk)
1980{
1981 __skb_insert(newsk, next->prev, next, list);
1982}
1983
Stephen Hemminger300ce172005-10-30 13:47:34 -08001984/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 * __skb_queue_head - queue a buffer at the list head
1986 * @list: list to use
1987 * @newsk: buffer to queue
1988 *
1989 * Queue a buffer at the start of a list. This function takes no locks
1990 * and you must therefore hold required locks before calling it.
1991 *
1992 * A buffer cannot be placed on two lists at the same time.
1993 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994static inline void __skb_queue_head(struct sk_buff_head *list,
1995 struct sk_buff *newsk)
1996{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001997 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08001999void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000
2001/**
2002 * __skb_queue_tail - queue a buffer at the list tail
2003 * @list: list to use
2004 * @newsk: buffer to queue
2005 *
2006 * Queue a buffer at the end of a list. This function takes no locks
2007 * and you must therefore hold required locks before calling it.
2008 *
2009 * A buffer cannot be placed on two lists at the same time.
2010 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011static inline void __skb_queue_tail(struct sk_buff_head *list,
2012 struct sk_buff *newsk)
2013{
Gerrit Renkerf5572852008-04-14 00:05:28 -07002014 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002016void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 * remove sk_buff from list. _Must_ be called atomically, and with
2020 * the list known..
2021 */
Joe Perches7965bd42013-09-26 14:48:15 -07002022void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2024{
2025 struct sk_buff *next, *prev;
2026
2027 list->qlen--;
2028 next = skb->next;
2029 prev = skb->prev;
2030 skb->next = skb->prev = NULL;
Eric Dumazetd7d16a82019-10-23 22:44:48 -07002031 WRITE_ONCE(next->prev, prev);
2032 WRITE_ONCE(prev->next, next);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033}
2034
Gerrit Renkerf525c062008-04-14 00:04:12 -07002035/**
2036 * __skb_dequeue - remove from the head of the queue
2037 * @list: list to dequeue from
2038 *
2039 * Remove the head of the list. This function does not take any locks
2040 * so must be used with appropriate locks held only. The head item is
2041 * returned or %NULL if the list is empty.
2042 */
Gerrit Renkerf525c062008-04-14 00:04:12 -07002043static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2044{
2045 struct sk_buff *skb = skb_peek(list);
2046 if (skb)
2047 __skb_unlink(skb, list);
2048 return skb;
2049}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002050struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051
2052/**
2053 * __skb_dequeue_tail - remove from the tail of the queue
2054 * @list: list to dequeue from
2055 *
2056 * Remove the tail of the list. This function does not take any locks
2057 * so must be used with appropriate locks held only. The tail item is
2058 * returned or %NULL if the list is empty.
2059 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2061{
2062 struct sk_buff *skb = skb_peek_tail(list);
2063 if (skb)
2064 __skb_unlink(skb, list);
2065 return skb;
2066}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002067struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068
2069
David S. Millerbdcc0922012-03-07 20:53:36 -05002070static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071{
2072 return skb->data_len;
2073}
2074
2075static inline unsigned int skb_headlen(const struct sk_buff *skb)
2076{
2077 return skb->len - skb->data_len;
2078}
2079
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002080static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002082 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002084 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00002085 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002086 return len;
2087}
2088
2089static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2090{
2091 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092}
2093
Ian Campbell131ea662011-08-19 06:25:00 +00002094/**
2095 * __skb_fill_page_desc - initialise a paged fragment in an skb
2096 * @skb: buffer containing fragment to be initialised
2097 * @i: paged fragment index to initialise
2098 * @page: the page to use for this fragment
2099 * @off: the offset to the data with @page
2100 * @size: the length of the data
2101 *
2102 * Initialises the @i'th fragment of @skb to point to &size bytes at
2103 * offset @off within @page.
2104 *
2105 * Does not take any additional reference on the fragment.
2106 */
2107static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2108 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109{
2110 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2111
Mel Gormanc48a11c2012-07-31 16:44:23 -07002112 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07002113 * Propagate page pfmemalloc to the skb if we can. The problem is
2114 * that not all callers have unique ownership of the page but rely
2115 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07002116 */
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002117 frag->bv_page = page;
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002118 frag->bv_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00002119 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002120
2121 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07002122 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002123 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00002124}
2125
2126/**
2127 * skb_fill_page_desc - initialise a paged fragment in an skb
2128 * @skb: buffer containing fragment to be initialised
2129 * @i: paged fragment index to initialise
2130 * @page: the page to use for this fragment
2131 * @off: the offset to the data with @page
2132 * @size: the length of the data
2133 *
2134 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01002135 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00002136 * addition updates @skb such that @i is the last fragment.
2137 *
2138 * Does not take any additional reference on the fragment.
2139 */
2140static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2141 struct page *page, int off, int size)
2142{
2143 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144 skb_shinfo(skb)->nr_frags = i + 1;
2145}
2146
Joe Perches7965bd42013-09-26 14:48:15 -07002147void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2148 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07002149
Jason Wangf8e617e2013-11-01 14:07:47 +08002150void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2151 unsigned int truesize);
2152
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2154
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002155#ifdef NET_SKBUFF_DATA_USES_OFFSET
2156static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2157{
2158 return skb->head + skb->tail;
2159}
2160
2161static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2162{
2163 skb->tail = skb->data - skb->head;
2164}
2165
2166static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2167{
2168 skb_reset_tail_pointer(skb);
2169 skb->tail += offset;
2170}
Simon Horman7cc46192013-05-28 20:34:29 +00002171
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002172#else /* NET_SKBUFF_DATA_USES_OFFSET */
2173static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2174{
2175 return skb->tail;
2176}
2177
2178static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2179{
2180 skb->tail = skb->data;
2181}
2182
2183static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2184{
2185 skb->tail = skb->data + offset;
2186}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002187
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002188#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2189
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190/*
2191 * Add data to an sk_buff
2192 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002193void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2194void *skb_put(struct sk_buff *skb, unsigned int len);
2195static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196{
Johannes Berg4df864c2017-06-16 14:29:21 +02002197 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 SKB_LINEAR_ASSERT(skb);
2199 skb->tail += len;
2200 skb->len += len;
2201 return tmp;
2202}
2203
yuan linyude77b962017-06-18 22:48:17 +08002204static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2205{
2206 void *tmp = __skb_put(skb, len);
2207
2208 memset(tmp, 0, len);
2209 return tmp;
2210}
2211
2212static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2213 unsigned int len)
2214{
2215 void *tmp = __skb_put(skb, len);
2216
2217 memcpy(tmp, data, len);
2218 return tmp;
2219}
2220
2221static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2222{
2223 *(u8 *)__skb_put(skb, 1) = val;
2224}
2225
Johannes Berg83ad3572017-06-14 22:17:20 +02002226static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002227{
Johannes Berg83ad3572017-06-14 22:17:20 +02002228 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002229
2230 memset(tmp, 0, len);
2231
2232 return tmp;
2233}
2234
Johannes Berg59ae1d12017-06-16 14:29:20 +02002235static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2236 unsigned int len)
2237{
2238 void *tmp = skb_put(skb, len);
2239
2240 memcpy(tmp, data, len);
2241
2242 return tmp;
2243}
2244
Johannes Berg634fef62017-06-16 14:29:24 +02002245static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2246{
2247 *(u8 *)skb_put(skb, 1) = val;
2248}
2249
Johannes Bergd58ff352017-06-16 14:29:23 +02002250void *skb_push(struct sk_buff *skb, unsigned int len);
2251static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252{
2253 skb->data -= len;
2254 skb->len += len;
2255 return skb->data;
2256}
2257
Johannes Bergaf728682017-06-16 14:29:22 +02002258void *skb_pull(struct sk_buff *skb, unsigned int len);
2259static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260{
2261 skb->len -= len;
2262 BUG_ON(skb->len < skb->data_len);
2263 return skb->data += len;
2264}
2265
Johannes Bergaf728682017-06-16 14:29:22 +02002266static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002267{
2268 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2269}
2270
Johannes Bergaf728682017-06-16 14:29:22 +02002271void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002272
Johannes Bergaf728682017-06-16 14:29:22 +02002273static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274{
2275 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002276 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277 return NULL;
2278 skb->len -= len;
2279 return skb->data += len;
2280}
2281
Johannes Bergaf728682017-06-16 14:29:22 +02002282static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283{
2284 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2285}
2286
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002287static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288{
2289 if (likely(len <= skb_headlen(skb)))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002290 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 if (unlikely(len > skb->len))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002292 return false;
Gerrit Renker987c4022008-08-11 18:17:17 -07002293 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294}
2295
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002296void skb_condense(struct sk_buff *skb);
2297
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298/**
2299 * skb_headroom - bytes at buffer head
2300 * @skb: buffer to check
2301 *
2302 * Return the number of bytes of free space at the head of an &sk_buff.
2303 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002304static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305{
2306 return skb->data - skb->head;
2307}
2308
2309/**
2310 * skb_tailroom - bytes at buffer end
2311 * @skb: buffer to check
2312 *
2313 * Return the number of bytes of free space at the tail of an sk_buff
2314 */
2315static inline int skb_tailroom(const struct sk_buff *skb)
2316{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002317 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318}
2319
2320/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002321 * skb_availroom - bytes at buffer end
2322 * @skb: buffer to check
2323 *
2324 * Return the number of bytes of free space at the tail of an sk_buff
2325 * allocated by sk_stream_alloc()
2326 */
2327static inline int skb_availroom(const struct sk_buff *skb)
2328{
Eric Dumazet16fad692013-03-14 05:40:32 +00002329 if (skb_is_nonlinear(skb))
2330 return 0;
2331
2332 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002333}
2334
2335/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336 * skb_reserve - adjust headroom
2337 * @skb: buffer to alter
2338 * @len: bytes to move
2339 *
2340 * Increase the headroom of an empty &sk_buff by reducing the tail
2341 * room. This is only allowed for an empty buffer.
2342 */
David S. Miller8243126c2006-01-17 02:54:21 -08002343static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344{
2345 skb->data += len;
2346 skb->tail += len;
2347}
2348
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002349/**
2350 * skb_tailroom_reserve - adjust reserved_tailroom
2351 * @skb: buffer to alter
2352 * @mtu: maximum amount of headlen permitted
2353 * @needed_tailroom: minimum amount of reserved_tailroom
2354 *
2355 * Set reserved_tailroom so that headlen can be as large as possible but
2356 * not larger than mtu and tailroom cannot be smaller than
2357 * needed_tailroom.
2358 * The required headroom should already have been reserved before using
2359 * this function.
2360 */
2361static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2362 unsigned int needed_tailroom)
2363{
2364 SKB_LINEAR_ASSERT(skb);
2365 if (mtu < skb_tailroom(skb) - needed_tailroom)
2366 /* use at most mtu */
2367 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2368 else
2369 /* use up to all available space */
2370 skb->reserved_tailroom = needed_tailroom;
2371}
2372
Tom Herbert8bce6d72014-09-29 20:22:29 -07002373#define ENCAP_TYPE_ETHER 0
2374#define ENCAP_TYPE_IPPROTO 1
2375
2376static inline void skb_set_inner_protocol(struct sk_buff *skb,
2377 __be16 protocol)
2378{
2379 skb->inner_protocol = protocol;
2380 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2381}
2382
2383static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2384 __u8 ipproto)
2385{
2386 skb->inner_ipproto = ipproto;
2387 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2388}
2389
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002390static inline void skb_reset_inner_headers(struct sk_buff *skb)
2391{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002392 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002393 skb->inner_network_header = skb->network_header;
2394 skb->inner_transport_header = skb->transport_header;
2395}
2396
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002397static inline void skb_reset_mac_len(struct sk_buff *skb)
2398{
2399 skb->mac_len = skb->network_header - skb->mac_header;
2400}
2401
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002402static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2403 *skb)
2404{
2405 return skb->head + skb->inner_transport_header;
2406}
2407
Tom Herbert55dc5a92015-12-14 11:19:40 -08002408static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2409{
2410 return skb_inner_transport_header(skb) - skb->data;
2411}
2412
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002413static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2414{
2415 skb->inner_transport_header = skb->data - skb->head;
2416}
2417
2418static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2419 const int offset)
2420{
2421 skb_reset_inner_transport_header(skb);
2422 skb->inner_transport_header += offset;
2423}
2424
2425static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2426{
2427 return skb->head + skb->inner_network_header;
2428}
2429
2430static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2431{
2432 skb->inner_network_header = skb->data - skb->head;
2433}
2434
2435static inline void skb_set_inner_network_header(struct sk_buff *skb,
2436 const int offset)
2437{
2438 skb_reset_inner_network_header(skb);
2439 skb->inner_network_header += offset;
2440}
2441
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002442static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2443{
2444 return skb->head + skb->inner_mac_header;
2445}
2446
2447static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2448{
2449 skb->inner_mac_header = skb->data - skb->head;
2450}
2451
2452static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2453 const int offset)
2454{
2455 skb_reset_inner_mac_header(skb);
2456 skb->inner_mac_header += offset;
2457}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002458static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2459{
Cong Wang35d04612013-05-29 15:16:05 +08002460 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002461}
2462
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002463static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2464{
2465 return skb->head + skb->transport_header;
2466}
2467
2468static inline void skb_reset_transport_header(struct sk_buff *skb)
2469{
2470 skb->transport_header = skb->data - skb->head;
2471}
2472
2473static inline void skb_set_transport_header(struct sk_buff *skb,
2474 const int offset)
2475{
2476 skb_reset_transport_header(skb);
2477 skb->transport_header += offset;
2478}
2479
2480static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2481{
2482 return skb->head + skb->network_header;
2483}
2484
2485static inline void skb_reset_network_header(struct sk_buff *skb)
2486{
2487 skb->network_header = skb->data - skb->head;
2488}
2489
2490static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2491{
2492 skb_reset_network_header(skb);
2493 skb->network_header += offset;
2494}
2495
2496static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2497{
2498 return skb->head + skb->mac_header;
2499}
2500
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002501static inline int skb_mac_offset(const struct sk_buff *skb)
2502{
2503 return skb_mac_header(skb) - skb->data;
2504}
2505
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002506static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2507{
2508 return skb->network_header - skb->mac_header;
2509}
2510
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002511static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2512{
Cong Wang35d04612013-05-29 15:16:05 +08002513 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002514}
2515
2516static inline void skb_reset_mac_header(struct sk_buff *skb)
2517{
2518 skb->mac_header = skb->data - skb->head;
2519}
2520
2521static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2522{
2523 skb_reset_mac_header(skb);
2524 skb->mac_header += offset;
2525}
2526
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002527static inline void skb_pop_mac_header(struct sk_buff *skb)
2528{
2529 skb->mac_header = skb->network_header;
2530}
2531
Maxim Mikityanskiyd2aa1252019-02-21 12:39:57 +00002532static inline void skb_probe_transport_header(struct sk_buff *skb)
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002533{
Paolo Abeni72a338b2018-05-04 11:32:59 +02002534 struct flow_keys_basic keys;
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002535
2536 if (skb_transport_header_was_set(skb))
2537 return;
Paolo Abeni72a338b2018-05-04 11:32:59 +02002538
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07002539 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2540 NULL, 0, 0, 0, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002541 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002542}
2543
Eric Dumazet03606892012-02-23 10:55:02 +00002544static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2545{
2546 if (skb_mac_header_was_set(skb)) {
2547 const unsigned char *old_mac = skb_mac_header(skb);
2548
2549 skb_set_mac_header(skb, -skb->mac_len);
2550 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2551 }
2552}
2553
Michał Mirosław04fb4512010-12-14 15:24:08 +00002554static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2555{
2556 return skb->csum_start - skb_headroom(skb);
2557}
2558
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002559static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2560{
2561 return skb->head + skb->csum_start;
2562}
2563
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002564static inline int skb_transport_offset(const struct sk_buff *skb)
2565{
2566 return skb_transport_header(skb) - skb->data;
2567}
2568
2569static inline u32 skb_network_header_len(const struct sk_buff *skb)
2570{
2571 return skb->transport_header - skb->network_header;
2572}
2573
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002574static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2575{
2576 return skb->inner_transport_header - skb->inner_network_header;
2577}
2578
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002579static inline int skb_network_offset(const struct sk_buff *skb)
2580{
2581 return skb_network_header(skb) - skb->data;
2582}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002583
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002584static inline int skb_inner_network_offset(const struct sk_buff *skb)
2585{
2586 return skb_inner_network_header(skb) - skb->data;
2587}
2588
Changli Gaof9599ce2010-08-04 04:43:44 +00002589static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2590{
2591 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2592}
2593
Linus Torvalds1da177e2005-04-16 15:20:36 -07002594/*
2595 * CPUs often take a performance hit when accessing unaligned memory
2596 * locations. The actual performance hit varies, it can be small if the
2597 * hardware handles it or large if we have to take an exception and fix it
2598 * in software.
2599 *
2600 * Since an ethernet header is 14 bytes network drivers often end up with
2601 * the IP header at an unaligned offset. The IP header can be aligned by
2602 * shifting the start of the packet by 2 bytes. Drivers should do this
2603 * with:
2604 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002605 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002606 *
2607 * The downside to this alignment of the IP header is that the DMA is now
2608 * unaligned. On some architectures the cost of an unaligned DMA is high
2609 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002610 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2612 * to be overridden.
2613 */
2614#ifndef NET_IP_ALIGN
2615#define NET_IP_ALIGN 2
2616#endif
2617
Anton Blanchard025be812006-03-31 02:27:06 -08002618/*
2619 * The networking layer reserves some headroom in skb data (via
2620 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2621 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002622 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002623 *
2624 * Unfortunately this headroom changes the DMA alignment of the resulting
2625 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2626 * on some architectures. An architecture can override this value,
2627 * perhaps setting it to a cacheline in size (since that will maintain
2628 * cacheline alignment of the DMA). It must be a power of 2.
2629 *
David S. Millerd6301d32009-02-08 19:24:13 -08002630 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002631 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002632 *
2633 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2634 * to reduce average number of cache lines per packet.
2635 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002636 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002637 */
2638#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002639#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002640#endif
2641
Joe Perches7965bd42013-09-26 14:48:15 -07002642int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002644static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002645{
Yangtao Li5e1abdc2018-11-06 10:45:36 -05002646 if (WARN_ON(skb_is_nonlinear(skb)))
Herbert Xu3cc0e872006-06-09 16:13:38 -07002647 return;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002648 skb->len = len;
2649 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002650}
2651
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002652static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2653{
2654 __skb_set_length(skb, len);
2655}
2656
Joe Perches7965bd42013-09-26 14:48:15 -07002657void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658
2659static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2660{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002661 if (skb->data_len)
2662 return ___pskb_trim(skb, len);
2663 __skb_trim(skb, len);
2664 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002665}
2666
2667static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2668{
2669 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2670}
2671
2672/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002673 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2674 * @skb: buffer to alter
2675 * @len: new length
2676 *
2677 * This is identical to pskb_trim except that the caller knows that
2678 * the skb is not cloned so we should never get an error due to out-
2679 * of-memory.
2680 */
2681static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2682{
2683 int err = pskb_trim(skb, len);
2684 BUG_ON(err);
2685}
2686
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002687static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2688{
2689 unsigned int diff = len - skb->len;
2690
2691 if (skb_tailroom(skb) < diff) {
2692 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2693 GFP_ATOMIC);
2694 if (ret)
2695 return ret;
2696 }
2697 __skb_set_length(skb, len);
2698 return 0;
2699}
2700
Herbert Xue9fa4f72006-08-13 20:12:58 -07002701/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002702 * skb_orphan - orphan a buffer
2703 * @skb: buffer to orphan
2704 *
2705 * If a buffer currently has an owner then we call the owner's
2706 * destructor function and make the @skb unowned. The buffer continues
2707 * to exist but is no longer charged to its former owner.
2708 */
2709static inline void skb_orphan(struct sk_buff *skb)
2710{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002711 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002713 skb->destructor = NULL;
2714 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002715 } else {
2716 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002717 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002718}
2719
2720/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002721 * skb_orphan_frags - orphan the frags contained in a buffer
2722 * @skb: buffer to orphan frags from
2723 * @gfp_mask: allocation mask for replacement pages
2724 *
2725 * For each frag in the SKB which needs a destructor (i.e. has an
2726 * owner) create a copy of that frag and release the original
2727 * page by calling the destructor.
2728 */
2729static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2730{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002731 if (likely(!skb_zcopy(skb)))
2732 return 0;
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04002733 if (!skb_zcopy_is_nouarg(skb) &&
2734 skb_uarg(skb)->callback == sock_zerocopy_callback)
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002735 return 0;
2736 return skb_copy_ubufs(skb, gfp_mask);
2737}
2738
2739/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2740static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2741{
2742 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002743 return 0;
2744 return skb_copy_ubufs(skb, gfp_mask);
2745}
2746
2747/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002748 * __skb_queue_purge - empty a list
2749 * @list: list to empty
2750 *
2751 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2752 * the list and one reference dropped. This function does not take the
2753 * list lock and the caller must hold the relevant locks to use it.
2754 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002755static inline void __skb_queue_purge(struct sk_buff_head *list)
2756{
2757 struct sk_buff *skb;
2758 while ((skb = __skb_dequeue(list)) != NULL)
2759 kfree_skb(skb);
2760}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002761void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762
Peter Oskolkov385114d2018-08-02 23:34:38 +00002763unsigned int skb_rbtree_purge(struct rb_root *root);
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002764
Joe Perches7965bd42013-09-26 14:48:15 -07002765void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766
Joe Perches7965bd42013-09-26 14:48:15 -07002767struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2768 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002769
2770/**
2771 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2772 * @dev: network device to receive on
2773 * @length: length to allocate
2774 *
2775 * Allocate a new &sk_buff and assign it a usage count of one. The
2776 * buffer has unspecified headroom built in. Users should allocate
2777 * the headroom they think they need without accounting for the
2778 * built in space. The built in space is used for optimisations.
2779 *
2780 * %NULL is returned if there is no free memory. Although this function
2781 * allocates memory it can be called from an interrupt.
2782 */
2783static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002784 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002785{
2786 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2787}
2788
Eric Dumazet6f532612012-05-18 05:12:12 +00002789/* legacy helper around __netdev_alloc_skb() */
2790static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2791 gfp_t gfp_mask)
2792{
2793 return __netdev_alloc_skb(NULL, length, gfp_mask);
2794}
2795
2796/* legacy helper around netdev_alloc_skb() */
2797static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2798{
2799 return netdev_alloc_skb(NULL, length);
2800}
2801
2802
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002803static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2804 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002805{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002806 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002807
2808 if (NET_IP_ALIGN && skb)
2809 skb_reserve(skb, NET_IP_ALIGN);
2810 return skb;
2811}
2812
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002813static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2814 unsigned int length)
2815{
2816 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2817}
2818
Alexander Duyck181edb22015-05-06 21:12:03 -07002819static inline void skb_free_frag(void *addr)
2820{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002821 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002822}
2823
Alexander Duyckffde7322014-12-09 19:40:42 -08002824void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002825struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2826 unsigned int length, gfp_t gfp_mask);
2827static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2828 unsigned int length)
2829{
2830 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2831}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002832void napi_consume_skb(struct sk_buff *skb, int budget);
2833
2834void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002835void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002836
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002837/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002838 * __dev_alloc_pages - allocate page for network Rx
2839 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2840 * @order: size of the allocation
2841 *
2842 * Allocate a new page.
2843 *
2844 * %NULL is returned if there is no free memory.
2845*/
2846static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2847 unsigned int order)
2848{
2849 /* This piece of code contains several assumptions.
2850 * 1. This is for device Rx, therefor a cold page is preferred.
2851 * 2. The expectation is the user wants a compound page.
2852 * 3. If requesting a order 0 page it will not be compound
2853 * due to the check to see if order has a value in prep_new_page
2854 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2855 * code in gfp_to_alloc_flags that should be enforcing this.
2856 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002857 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002858
2859 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2860}
2861
2862static inline struct page *dev_alloc_pages(unsigned int order)
2863{
Neil Horman95829b32016-05-19 11:30:54 -04002864 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002865}
2866
2867/**
2868 * __dev_alloc_page - allocate a page for network Rx
2869 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2870 *
2871 * Allocate a new page.
2872 *
2873 * %NULL is returned if there is no free memory.
2874 */
2875static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2876{
2877 return __dev_alloc_pages(gfp_mask, 0);
2878}
2879
2880static inline struct page *dev_alloc_page(void)
2881{
Neil Horman95829b32016-05-19 11:30:54 -04002882 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002883}
2884
2885/**
Mel Gorman06140022012-07-31 16:44:24 -07002886 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2887 * @page: The page that was allocated from skb_alloc_page
2888 * @skb: The skb that may need pfmemalloc set
2889 */
2890static inline void skb_propagate_pfmemalloc(struct page *page,
2891 struct sk_buff *skb)
2892{
Michal Hocko2f064f32015-08-21 14:11:51 -07002893 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002894 skb->pfmemalloc = true;
2895}
2896
Eric Dumazet564824b2010-10-11 19:05:25 +00002897/**
Jonathan Lemon7240b602019-07-30 07:40:32 -07002898 * skb_frag_off() - Returns the offset of a skb fragment
2899 * @frag: the paged fragment
2900 */
2901static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2902{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002903 return frag->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002904}
2905
2906/**
2907 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2908 * @frag: skb fragment
2909 * @delta: value to add
2910 */
2911static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2912{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002913 frag->bv_offset += delta;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002914}
2915
2916/**
2917 * skb_frag_off_set() - Sets the offset of a skb fragment
2918 * @frag: skb fragment
2919 * @offset: offset of fragment
2920 */
2921static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2922{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002923 frag->bv_offset = offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002924}
2925
2926/**
2927 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2928 * @fragto: skb fragment where offset is set
2929 * @fragfrom: skb fragment offset is copied from
2930 */
2931static inline void skb_frag_off_copy(skb_frag_t *fragto,
2932 const skb_frag_t *fragfrom)
2933{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002934 fragto->bv_offset = fragfrom->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002935}
2936
2937/**
Masanari Iidae2278672014-02-18 22:54:36 +09002938 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002939 * @frag: the paged fragment
2940 *
2941 * Returns the &struct page associated with @frag.
2942 */
2943static inline struct page *skb_frag_page(const skb_frag_t *frag)
2944{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002945 return frag->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00002946}
2947
2948/**
2949 * __skb_frag_ref - take an addition reference on a paged fragment.
2950 * @frag: the paged fragment
2951 *
2952 * Takes an additional reference on the paged fragment @frag.
2953 */
2954static inline void __skb_frag_ref(skb_frag_t *frag)
2955{
2956 get_page(skb_frag_page(frag));
2957}
2958
2959/**
2960 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2961 * @skb: the buffer
2962 * @f: the fragment offset.
2963 *
2964 * Takes an additional reference on the @f'th paged fragment of @skb.
2965 */
2966static inline void skb_frag_ref(struct sk_buff *skb, int f)
2967{
2968 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2969}
2970
2971/**
2972 * __skb_frag_unref - release a reference on a paged fragment.
2973 * @frag: the paged fragment
2974 *
2975 * Releases a reference on the paged fragment @frag.
2976 */
2977static inline void __skb_frag_unref(skb_frag_t *frag)
2978{
2979 put_page(skb_frag_page(frag));
2980}
2981
2982/**
2983 * skb_frag_unref - release a reference on a paged fragment of an skb.
2984 * @skb: the buffer
2985 * @f: the fragment offset
2986 *
2987 * Releases a reference on the @f'th paged fragment of @skb.
2988 */
2989static inline void skb_frag_unref(struct sk_buff *skb, int f)
2990{
2991 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2992}
2993
2994/**
2995 * skb_frag_address - gets the address of the data contained in a paged fragment
2996 * @frag: the paged fragment buffer
2997 *
2998 * Returns the address of the data within @frag. The page must already
2999 * be mapped.
3000 */
3001static inline void *skb_frag_address(const skb_frag_t *frag)
3002{
Jonathan Lemon7240b602019-07-30 07:40:32 -07003003 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
Ian Campbell131ea662011-08-19 06:25:00 +00003004}
3005
3006/**
3007 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3008 * @frag: the paged fragment buffer
3009 *
3010 * Returns the address of the data within @frag. Checks that the page
3011 * is mapped and returns %NULL otherwise.
3012 */
3013static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3014{
3015 void *ptr = page_address(skb_frag_page(frag));
3016 if (unlikely(!ptr))
3017 return NULL;
3018
Jonathan Lemon7240b602019-07-30 07:40:32 -07003019 return ptr + skb_frag_off(frag);
3020}
3021
3022/**
3023 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3024 * @fragto: skb fragment where page is set
3025 * @fragfrom: skb fragment page is copied from
3026 */
3027static inline void skb_frag_page_copy(skb_frag_t *fragto,
3028 const skb_frag_t *fragfrom)
3029{
3030 fragto->bv_page = fragfrom->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00003031}
3032
3033/**
3034 * __skb_frag_set_page - sets the page contained in a paged fragment
3035 * @frag: the paged fragment
3036 * @page: the page to set
3037 *
3038 * Sets the fragment @frag to contain @page.
3039 */
3040static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3041{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07003042 frag->bv_page = page;
Ian Campbell131ea662011-08-19 06:25:00 +00003043}
3044
3045/**
3046 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3047 * @skb: the buffer
3048 * @f: the fragment offset
3049 * @page: the page to set
3050 *
3051 * Sets the @f'th fragment of @skb to contain @page.
3052 */
3053static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3054 struct page *page)
3055{
3056 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3057}
3058
Eric Dumazet400dfd32013-10-17 16:27:07 -07003059bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3060
Ian Campbell131ea662011-08-19 06:25:00 +00003061/**
3062 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00003063 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00003064 * @frag: the paged fragment to map
3065 * @offset: the offset within the fragment (starting at the
3066 * fragment's own offset)
3067 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03003068 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00003069 *
3070 * Maps the page associated with @frag to @device.
3071 */
3072static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3073 const skb_frag_t *frag,
3074 size_t offset, size_t size,
3075 enum dma_data_direction dir)
3076{
3077 return dma_map_page(dev, skb_frag_page(frag),
Jonathan Lemon7240b602019-07-30 07:40:32 -07003078 skb_frag_off(frag) + offset, size, dir);
Ian Campbell131ea662011-08-19 06:25:00 +00003079}
3080
Eric Dumazet117632e2011-12-03 21:39:53 +00003081static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3082 gfp_t gfp_mask)
3083{
3084 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3085}
3086
Octavian Purdilabad93e92014-06-12 01:36:26 +03003087
3088static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3089 gfp_t gfp_mask)
3090{
3091 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3092}
3093
3094
Ian Campbell131ea662011-08-19 06:25:00 +00003095/**
Patrick McHardy334a8132007-06-25 04:35:20 -07003096 * skb_clone_writable - is the header of a clone writable
3097 * @skb: buffer to check
3098 * @len: length up to which to write
3099 *
3100 * Returns true if modifying the header part of the cloned buffer
3101 * does not requires the data to be copied.
3102 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003103static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07003104{
3105 return !skb_header_cloned(skb) &&
3106 skb_headroom(skb) + len <= skb->hdr_len;
3107}
3108
Daniel Borkmann36976492016-02-19 23:05:25 +01003109static inline int skb_try_make_writable(struct sk_buff *skb,
3110 unsigned int write_len)
3111{
3112 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3113 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3114}
3115
Herbert Xud9cc2042007-09-16 16:21:16 -07003116static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3117 int cloned)
3118{
3119 int delta = 0;
3120
Herbert Xud9cc2042007-09-16 16:21:16 -07003121 if (headroom > skb_headroom(skb))
3122 delta = headroom - skb_headroom(skb);
3123
3124 if (delta || cloned)
3125 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3126 GFP_ATOMIC);
3127 return 0;
3128}
3129
Patrick McHardy334a8132007-06-25 04:35:20 -07003130/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07003131 * skb_cow - copy header of skb when it is required
3132 * @skb: buffer to cow
3133 * @headroom: needed headroom
3134 *
3135 * If the skb passed lacks sufficient headroom or its data part
3136 * is shared, data is reallocated. If reallocation fails, an error
3137 * is returned and original skb is not changed.
3138 *
3139 * The result is skb with writable area skb->head...skb->tail
3140 * and at least @headroom of space at head.
3141 */
3142static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3143{
Herbert Xud9cc2042007-09-16 16:21:16 -07003144 return __skb_cow(skb, headroom, skb_cloned(skb));
3145}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003146
Herbert Xud9cc2042007-09-16 16:21:16 -07003147/**
3148 * skb_cow_head - skb_cow but only making the head writable
3149 * @skb: buffer to cow
3150 * @headroom: needed headroom
3151 *
3152 * This function is identical to skb_cow except that we replace the
3153 * skb_cloned check by skb_header_cloned. It should be used when
3154 * you only need to push on some header and do not need to modify
3155 * the data.
3156 */
3157static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3158{
3159 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003160}
3161
3162/**
3163 * skb_padto - pad an skbuff up to a minimal size
3164 * @skb: buffer to pad
3165 * @len: minimal length
3166 *
3167 * Pads up a buffer to ensure the trailing bytes exist and are
3168 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07003169 * is untouched. Otherwise it is extended. Returns zero on
3170 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003171 */
Herbert Xu5b057c62006-06-23 02:06:41 -07003172static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003173{
3174 unsigned int size = skb->len;
3175 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07003176 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07003177 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003178}
3179
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003180/**
Brian Norris4ea7b0c2019-02-11 13:02:25 -08003181 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003182 * @skb: buffer to pad
3183 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07003184 * @free_on_error: free buffer on error
3185 *
3186 * Pads up a buffer to ensure the trailing bytes exist and are
3187 * blanked. If the buffer already contains sufficient data it
3188 * is untouched. Otherwise it is extended. Returns zero on
3189 * success. The skb is freed on error if @free_on_error is true.
3190 */
3191static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3192 bool free_on_error)
3193{
3194 unsigned int size = skb->len;
3195
3196 if (unlikely(size < len)) {
3197 len -= size;
3198 if (__skb_pad(skb, len, free_on_error))
3199 return -ENOMEM;
3200 __skb_put(skb, len);
3201 }
3202 return 0;
3203}
3204
3205/**
3206 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3207 * @skb: buffer to pad
3208 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003209 *
3210 * Pads up a buffer to ensure the trailing bytes exist and are
3211 * blanked. If the buffer already contains sufficient data it
3212 * is untouched. Otherwise it is extended. Returns zero on
3213 * success. The skb is freed on error.
3214 */
3215static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3216{
Florian Fainellicd0a1372017-08-22 15:12:14 -07003217 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003218}
3219
Linus Torvalds1da177e2005-04-16 15:20:36 -07003220static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05003221 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003222{
3223 const int off = skb->len;
3224
3225 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05003226 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04003227 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3228 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003229 skb->csum = csum_block_add(skb->csum, csum, off);
3230 return 0;
3231 }
Al Viro15e6cb42016-11-01 22:42:45 -04003232 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003233 return 0;
3234
3235 __skb_trim(skb, off);
3236 return -EFAULT;
3237}
3238
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003239static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3240 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003241{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003242 if (skb_zcopy(skb))
3243 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003244 if (i) {
Matthew Wilcox (Oracle)d8e18a52019-07-22 20:08:26 -07003245 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003246
Ian Campbellea2ab692011-08-22 23:44:58 +00003247 return page == skb_frag_page(frag) &&
Jonathan Lemon7240b602019-07-30 07:40:32 -07003248 off == skb_frag_off(frag) + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003249 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003250 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003251}
3252
Herbert Xu364c6ba2006-06-09 16:10:40 -07003253static inline int __skb_linearize(struct sk_buff *skb)
3254{
3255 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3256}
3257
Linus Torvalds1da177e2005-04-16 15:20:36 -07003258/**
3259 * skb_linearize - convert paged skb to linear one
3260 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003261 *
3262 * If there is no free memory -ENOMEM is returned, otherwise zero
3263 * is returned and the old skb data released.
3264 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003265static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003266{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003267 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3268}
3269
3270/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003271 * skb_has_shared_frag - can any frag be overwritten
3272 * @skb: buffer to test
3273 *
3274 * Return true if the skb has at least one frag that might be modified
3275 * by an external entity (as in vmsplice()/sendfile())
3276 */
3277static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3278{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003279 return skb_is_nonlinear(skb) &&
3280 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003281}
3282
3283/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003284 * skb_linearize_cow - make sure skb is linear and writable
3285 * @skb: buffer to process
3286 *
3287 * If there is no free memory -ENOMEM is returned, otherwise zero
3288 * is returned and the old skb data released.
3289 */
3290static inline int skb_linearize_cow(struct sk_buff *skb)
3291{
3292 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3293 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003294}
3295
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003296static __always_inline void
3297__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3298 unsigned int off)
3299{
3300 if (skb->ip_summed == CHECKSUM_COMPLETE)
3301 skb->csum = csum_block_sub(skb->csum,
3302 csum_partial(start, len, 0), off);
3303 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3304 skb_checksum_start_offset(skb) < 0)
3305 skb->ip_summed = CHECKSUM_NONE;
3306}
3307
Linus Torvalds1da177e2005-04-16 15:20:36 -07003308/**
3309 * skb_postpull_rcsum - update checksum for received skb after pull
3310 * @skb: buffer to update
3311 * @start: start of data before pull
3312 * @len: length of data pulled
3313 *
3314 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003315 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3316 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003318static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003319 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003321 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322}
3323
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003324static __always_inline void
3325__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3326 unsigned int off)
3327{
3328 if (skb->ip_summed == CHECKSUM_COMPLETE)
3329 skb->csum = csum_block_add(skb->csum,
3330 csum_partial(start, len, 0), off);
3331}
Herbert Xucbb042f2006-03-20 22:43:56 -08003332
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003333/**
3334 * skb_postpush_rcsum - update checksum for received skb after push
3335 * @skb: buffer to update
3336 * @start: start of data after push
3337 * @len: length of data pushed
3338 *
3339 * After doing a push on a received packet, you need to call this to
3340 * update the CHECKSUM_COMPLETE checksum.
3341 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003342static inline void skb_postpush_rcsum(struct sk_buff *skb,
3343 const void *start, unsigned int len)
3344{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003345 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003346}
3347
Johannes Bergaf728682017-06-16 14:29:22 +02003348void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003349
David S. Miller7ce5a272013-12-02 17:26:05 -05003350/**
WANG Cong82a31b92016-06-30 10:15:22 -07003351 * skb_push_rcsum - push skb and update receive checksum
3352 * @skb: buffer to update
3353 * @len: length of data pulled
3354 *
3355 * This function performs an skb_push on the packet and updates
3356 * the CHECKSUM_COMPLETE checksum. It should be used on
3357 * receive path processing instead of skb_push unless you know
3358 * that the checksum difference is zero (e.g., a valid IP header)
3359 * or you are setting ip_summed to CHECKSUM_NONE.
3360 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003361static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003362{
3363 skb_push(skb, len);
3364 skb_postpush_rcsum(skb, skb->data, len);
3365 return skb->data;
3366}
3367
Eric Dumazet88078d92018-04-18 11:43:15 -07003368int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
WANG Cong82a31b92016-06-30 10:15:22 -07003369/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003370 * pskb_trim_rcsum - trim received skb and update checksum
3371 * @skb: buffer to trim
3372 * @len: new length
3373 *
3374 * This is exactly the same as pskb_trim except that it ensures the
3375 * checksum of received packets are still valid after the operation.
Ross Lagerwall6c57f042019-01-17 15:34:38 +00003376 * It can change skb pointers.
David S. Miller7ce5a272013-12-02 17:26:05 -05003377 */
3378
3379static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3380{
3381 if (likely(len >= skb->len))
3382 return 0;
Eric Dumazet88078d92018-04-18 11:43:15 -07003383 return pskb_trim_rcsum_slow(skb, len);
David S. Miller7ce5a272013-12-02 17:26:05 -05003384}
3385
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003386static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3387{
3388 if (skb->ip_summed == CHECKSUM_COMPLETE)
3389 skb->ip_summed = CHECKSUM_NONE;
3390 __skb_trim(skb, len);
3391 return 0;
3392}
3393
3394static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3395{
3396 if (skb->ip_summed == CHECKSUM_COMPLETE)
3397 skb->ip_summed = CHECKSUM_NONE;
3398 return __skb_grow(skb, len);
3399}
3400
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003401#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3402#define skb_rb_first(root) rb_to_skb(rb_first(root))
3403#define skb_rb_last(root) rb_to_skb(rb_last(root))
3404#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3405#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3406
Linus Torvalds1da177e2005-04-16 15:20:36 -07003407#define skb_queue_walk(queue, skb) \
3408 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003409 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003410 skb = skb->next)
3411
James Chapman46f89142007-04-30 00:07:31 -07003412#define skb_queue_walk_safe(queue, skb, tmp) \
3413 for (skb = (queue)->next, tmp = skb->next; \
3414 skb != (struct sk_buff *)(queue); \
3415 skb = tmp, tmp = skb->next)
3416
David S. Miller1164f522008-09-23 00:49:44 -07003417#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003418 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003419 skb = skb->next)
3420
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003421#define skb_rbtree_walk(skb, root) \
3422 for (skb = skb_rb_first(root); skb != NULL; \
3423 skb = skb_rb_next(skb))
3424
3425#define skb_rbtree_walk_from(skb) \
3426 for (; skb != NULL; \
3427 skb = skb_rb_next(skb))
3428
3429#define skb_rbtree_walk_from_safe(skb, tmp) \
3430 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3431 skb = tmp)
3432
David S. Miller1164f522008-09-23 00:49:44 -07003433#define skb_queue_walk_from_safe(queue, skb, tmp) \
3434 for (tmp = skb->next; \
3435 skb != (struct sk_buff *)(queue); \
3436 skb = tmp, tmp = skb->next)
3437
Stephen Hemminger300ce172005-10-30 13:47:34 -08003438#define skb_queue_reverse_walk(queue, skb) \
3439 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003440 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003441 skb = skb->prev)
3442
David S. Miller686a2952011-01-20 22:47:32 -08003443#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3444 for (skb = (queue)->prev, tmp = skb->prev; \
3445 skb != (struct sk_buff *)(queue); \
3446 skb = tmp, tmp = skb->prev)
3447
3448#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3449 for (tmp = skb->prev; \
3450 skb != (struct sk_buff *)(queue); \
3451 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003452
David S. Miller21dc3302010-08-23 00:13:46 -07003453static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003454{
3455 return skb_shinfo(skb)->frag_list != NULL;
3456}
3457
3458static inline void skb_frag_list_init(struct sk_buff *skb)
3459{
3460 skb_shinfo(skb)->frag_list = NULL;
3461}
3462
David S. Milleree039872009-06-09 00:17:13 -07003463#define skb_walk_frags(skb, iter) \
3464 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3465
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003466
3467int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3468 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003469struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3470 struct sk_buff_head *queue,
3471 unsigned int flags,
3472 void (*destructor)(struct sock *sk,
3473 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003474 int *off, int *err,
Paolo Abeni65101ae2017-05-16 11:20:13 +02003475 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003476struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003477 void (*destructor)(struct sock *sk,
3478 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003479 int *off, int *err,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003480 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003481struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003482 void (*destructor)(struct sock *sk,
3483 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003484 int *off, int *err);
Joe Perches7965bd42013-09-26 14:48:15 -07003485struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3486 int *err);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07003487__poll_t datagram_poll(struct file *file, struct socket *sock,
3488 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003489int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3490 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003491static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3492 struct msghdr *msg, int size)
3493{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003494 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003495}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003496int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3497 struct msghdr *msg);
Sagi Grimberg65d69e22018-12-03 17:52:10 -08003498int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3499 struct iov_iter *to, int len,
3500 struct ahash_request *hash);
Al Viro3a654f92014-06-19 14:15:22 -04003501int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3502 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003503int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003504void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003505void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3506static inline void skb_free_datagram_locked(struct sock *sk,
3507 struct sk_buff *skb)
3508{
3509 __skb_free_datagram_locked(sk, skb, 0);
3510}
Joe Perches7965bd42013-09-26 14:48:15 -07003511int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003512int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3513int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3514__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3515 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003516int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003517 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003518 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003519int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3520 int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003521void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003522unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003523int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3524 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003525void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3526int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3527void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Daniel Axtens779b7932018-03-01 17:13:37 +11003528bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
Daniel Axtens2b16f042018-01-31 14:15:33 +11003529bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003530struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003531struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003532int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003533int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003534int skb_vlan_pop(struct sk_buff *skb);
3535int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Davide Carattifa4e0f82019-10-12 13:55:07 +02003536int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
Martin Varghesed04ac222019-12-05 05:57:22 +05303537 int mac_len, bool ethernet);
Martin Varghese040b5cf2019-12-02 10:49:51 +05303538int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3539 bool ethernet);
John Hurleyd27cf5c2019-07-07 15:01:56 +01003540int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
John Hurley2a2ea502019-07-07 15:01:57 +01003541int skb_mpls_dec_ttl(struct sk_buff *skb);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003542struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3543 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003544
Al Viro6ce8e9c2014-04-06 21:25:44 -04003545static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3546{
Al Viro3073f072017-02-17 23:13:25 -05003547 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003548}
3549
Al Viro7eab8d92014-04-06 21:51:23 -04003550static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3551{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003552 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003553}
3554
Daniel Borkmann2817a332013-10-30 11:50:51 +01003555struct skb_checksum_ops {
3556 __wsum (*update)(const void *mem, int len, __wsum wsum);
3557 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3558};
3559
Davide Caratti96178132017-05-18 15:44:37 +02003560extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3561
Daniel Borkmann2817a332013-10-30 11:50:51 +01003562__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3563 __wsum csum, const struct skb_checksum_ops *ops);
3564__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3565 __wsum csum);
3566
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003567static inline void * __must_check
3568__skb_header_pointer(const struct sk_buff *skb, int offset,
3569 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003570{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003571 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003572 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003573
David S. Miller690e36e2014-08-23 12:13:41 -07003574 if (!skb ||
3575 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003576 return NULL;
3577
3578 return buffer;
3579}
3580
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003581static inline void * __must_check
3582skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003583{
3584 return __skb_header_pointer(skb, offset, len, skb->data,
3585 skb_headlen(skb), buffer);
3586}
3587
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003588/**
3589 * skb_needs_linearize - check if we need to linearize a given skb
3590 * depending on the given device features.
3591 * @skb: socket buffer to check
3592 * @features: net device features
3593 *
3594 * Returns true if either:
3595 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3596 * 2. skb is fragmented and the device does not support SG.
3597 */
3598static inline bool skb_needs_linearize(struct sk_buff *skb,
3599 netdev_features_t features)
3600{
3601 return skb_is_nonlinear(skb) &&
3602 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3603 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3604}
3605
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003606static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3607 void *to,
3608 const unsigned int len)
3609{
3610 memcpy(to, skb->data, len);
3611}
3612
3613static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3614 const int offset, void *to,
3615 const unsigned int len)
3616{
3617 memcpy(to, skb->data + offset, len);
3618}
3619
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003620static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3621 const void *from,
3622 const unsigned int len)
3623{
3624 memcpy(skb->data, from, len);
3625}
3626
3627static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3628 const int offset,
3629 const void *from,
3630 const unsigned int len)
3631{
3632 memcpy(skb->data + offset, from, len);
3633}
3634
Joe Perches7965bd42013-09-26 14:48:15 -07003635void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003636
Patrick Ohlyac45f602009-02-12 05:03:37 +00003637static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3638{
3639 return skb->tstamp;
3640}
3641
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003642/**
3643 * skb_get_timestamp - get timestamp from a skb
3644 * @skb: skb to get stamp from
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003645 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003646 *
3647 * Timestamps are stored in the skb as offsets to a base timestamp.
3648 * This function converts the offset back to a struct timeval and stores
3649 * it in stamp.
3650 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003651static inline void skb_get_timestamp(const struct sk_buff *skb,
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003652 struct __kernel_old_timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003653{
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003654 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003655}
3656
Deepa Dinamani887feae2019-02-02 07:34:50 -08003657static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3658 struct __kernel_sock_timeval *stamp)
3659{
3660 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3661
3662 stamp->tv_sec = ts.tv_sec;
3663 stamp->tv_usec = ts.tv_nsec / 1000;
3664}
3665
Patrick Ohlyac45f602009-02-12 05:03:37 +00003666static inline void skb_get_timestampns(const struct sk_buff *skb,
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003667 struct __kernel_old_timespec *stamp)
Patrick Ohlyac45f602009-02-12 05:03:37 +00003668{
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003669 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3670
3671 stamp->tv_sec = ts.tv_sec;
3672 stamp->tv_nsec = ts.tv_nsec;
Patrick Ohlyac45f602009-02-12 05:03:37 +00003673}
3674
Deepa Dinamani887feae2019-02-02 07:34:50 -08003675static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3676 struct __kernel_timespec *stamp)
3677{
3678 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3679
3680 stamp->tv_sec = ts.tv_sec;
3681 stamp->tv_nsec = ts.tv_nsec;
3682}
3683
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003684static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003685{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003686 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003687}
3688
Stephen Hemminger164891a2007-04-23 22:26:16 -07003689static inline ktime_t net_timedelta(ktime_t t)
3690{
3691 return ktime_sub(ktime_get_real(), t);
3692}
3693
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003694static inline ktime_t net_invalid_timestamp(void)
3695{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003696 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003697}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003698
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003699static inline u8 skb_metadata_len(const struct sk_buff *skb)
3700{
3701 return skb_shinfo(skb)->meta_len;
3702}
3703
3704static inline void *skb_metadata_end(const struct sk_buff *skb)
3705{
3706 return skb_mac_header(skb);
3707}
3708
3709static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3710 const struct sk_buff *skb_b,
3711 u8 meta_len)
3712{
3713 const void *a = skb_metadata_end(skb_a);
3714 const void *b = skb_metadata_end(skb_b);
3715 /* Using more efficient varaiant than plain call to memcmp(). */
3716#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3717 u64 diffs = 0;
3718
3719 switch (meta_len) {
3720#define __it(x, op) (x -= sizeof(u##op))
3721#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3722 case 32: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003723 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003724 case 24: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003725 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003726 case 16: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003727 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003728 case 8: diffs |= __it_diff(a, b, 64);
3729 break;
3730 case 28: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003731 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003732 case 20: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003733 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003734 case 12: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003735 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003736 case 4: diffs |= __it_diff(a, b, 32);
3737 break;
3738 }
3739 return diffs;
3740#else
3741 return memcmp(a - meta_len, b - meta_len, meta_len);
3742#endif
3743}
3744
3745static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3746 const struct sk_buff *skb_b)
3747{
3748 u8 len_a = skb_metadata_len(skb_a);
3749 u8 len_b = skb_metadata_len(skb_b);
3750
3751 if (!(len_a | len_b))
3752 return false;
3753
3754 return len_a != len_b ?
3755 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3756}
3757
3758static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3759{
3760 skb_shinfo(skb)->meta_len = meta_len;
3761}
3762
3763static inline void skb_metadata_clear(struct sk_buff *skb)
3764{
3765 skb_metadata_set(skb, 0);
3766}
3767
Alexander Duyck62bccb82014-09-04 13:31:35 -04003768struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3769
Richard Cochranc1f19b52010-07-17 08:49:36 +00003770#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3771
Joe Perches7965bd42013-09-26 14:48:15 -07003772void skb_clone_tx_timestamp(struct sk_buff *skb);
3773bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003774
3775#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3776
3777static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3778{
3779}
3780
3781static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3782{
3783 return false;
3784}
3785
3786#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3787
3788/**
3789 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3790 *
Richard Cochranda92b192011-10-21 00:49:15 +00003791 * PHY drivers may accept clones of transmitted packets for
3792 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003793 * must call this function to return the skb back to the stack with a
3794 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003795 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003796 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003797 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003798 *
3799 */
3800void skb_complete_tx_timestamp(struct sk_buff *skb,
3801 struct skb_shared_hwtstamps *hwtstamps);
3802
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003803void __skb_tstamp_tx(struct sk_buff *orig_skb,
3804 struct skb_shared_hwtstamps *hwtstamps,
3805 struct sock *sk, int tstype);
3806
Patrick Ohlyac45f602009-02-12 05:03:37 +00003807/**
3808 * skb_tstamp_tx - queue clone of skb with send time stamps
3809 * @orig_skb: the original outgoing packet
3810 * @hwtstamps: hardware time stamps, may be NULL if not available
3811 *
3812 * If the skb has a socket associated, then this function clones the
3813 * skb (thus sharing the actual data and optional structures), stores
3814 * the optional hardware time stamping information (if non NULL) or
3815 * generates a software time stamp (otherwise), then queues the clone
3816 * to the error queue of the socket. Errors are silently ignored.
3817 */
Joe Perches7965bd42013-09-26 14:48:15 -07003818void skb_tstamp_tx(struct sk_buff *orig_skb,
3819 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003820
Richard Cochran4507a712010-07-17 08:48:28 +00003821/**
3822 * skb_tx_timestamp() - Driver hook for transmit timestamping
3823 *
3824 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003825 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003826 *
David S. Miller73409f32013-12-27 13:04:33 -05003827 * Specifically, one should make absolutely sure that this function is
3828 * called before TX completion of this packet can trigger. Otherwise
3829 * the packet could potentially already be freed.
3830 *
Richard Cochran4507a712010-07-17 08:48:28 +00003831 * @skb: A socket buffer.
3832 */
3833static inline void skb_tx_timestamp(struct sk_buff *skb)
3834{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003835 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003836 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3837 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003838}
3839
Johannes Berg6e3e9392011-11-09 10:15:42 +01003840/**
3841 * skb_complete_wifi_ack - deliver skb with wifi status
3842 *
3843 * @skb: the original outgoing packet
3844 * @acked: ack status
3845 *
3846 */
3847void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3848
Joe Perches7965bd42013-09-26 14:48:15 -07003849__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3850__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003851
Herbert Xu60476372007-04-09 11:59:39 -07003852static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3853{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003854 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3855 skb->csum_valid ||
3856 (skb->ip_summed == CHECKSUM_PARTIAL &&
3857 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003858}
3859
Herbert Xufb286bb2005-11-10 13:01:24 -08003860/**
3861 * skb_checksum_complete - Calculate checksum of an entire packet
3862 * @skb: packet to process
3863 *
3864 * This function calculates the checksum over the entire packet plus
3865 * the value of skb->csum. The latter can be used to supply the
3866 * checksum of a pseudo header as used by TCP/UDP. It returns the
3867 * checksum.
3868 *
3869 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3870 * this function can be used to verify that checksum on received
3871 * packets. In that case the function should return zero if the
3872 * checksum is correct. In particular, this function will return zero
3873 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3874 * hardware has already verified the correctness of the checksum.
3875 */
Al Viro4381ca32007-07-15 21:00:11 +01003876static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003877{
Herbert Xu60476372007-04-09 11:59:39 -07003878 return skb_csum_unnecessary(skb) ?
3879 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003880}
3881
Tom Herbert77cffe22014-08-27 21:26:46 -07003882static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3883{
3884 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3885 if (skb->csum_level == 0)
3886 skb->ip_summed = CHECKSUM_NONE;
3887 else
3888 skb->csum_level--;
3889 }
3890}
3891
3892static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3893{
3894 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3895 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3896 skb->csum_level++;
3897 } else if (skb->ip_summed == CHECKSUM_NONE) {
3898 skb->ip_summed = CHECKSUM_UNNECESSARY;
3899 skb->csum_level = 0;
3900 }
3901}
3902
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003903/* Check if we need to perform checksum complete validation.
3904 *
3905 * Returns true if checksum complete is needed, false otherwise
3906 * (either checksum is unnecessary or zero checksum is allowed).
3907 */
3908static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3909 bool zero_okay,
3910 __sum16 check)
3911{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003912 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3913 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003914 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003915 return false;
3916 }
3917
3918 return true;
3919}
3920
David S. Millerda279882018-02-16 15:52:42 -05003921/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003922 * in checksum_init.
3923 */
3924#define CHECKSUM_BREAK 76
3925
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003926/* Unset checksum-complete
3927 *
3928 * Unset checksum complete can be done when packet is being modified
3929 * (uncompressed for instance) and checksum-complete value is
3930 * invalidated.
3931 */
3932static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3933{
3934 if (skb->ip_summed == CHECKSUM_COMPLETE)
3935 skb->ip_summed = CHECKSUM_NONE;
3936}
3937
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003938/* Validate (init) checksum based on checksum complete.
3939 *
3940 * Return values:
3941 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3942 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3943 * checksum is stored in skb->csum for use in __skb_checksum_complete
3944 * non-zero: value of invalid checksum
3945 *
3946 */
3947static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3948 bool complete,
3949 __wsum psum)
3950{
3951 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3952 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003953 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003954 return 0;
3955 }
3956 }
3957
3958 skb->csum = psum;
3959
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003960 if (complete || skb->len <= CHECKSUM_BREAK) {
3961 __sum16 csum;
3962
3963 csum = __skb_checksum_complete(skb);
3964 skb->csum_valid = !csum;
3965 return csum;
3966 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003967
3968 return 0;
3969}
3970
3971static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3972{
3973 return 0;
3974}
3975
3976/* Perform checksum validate (init). Note that this is a macro since we only
3977 * want to calculate the pseudo header which is an input function if necessary.
3978 * First we try to validate without any computation (checksum unnecessary) and
3979 * then calculate based on checksum complete calling the function to compute
3980 * pseudo header.
3981 *
3982 * Return values:
3983 * 0: checksum is validated or try to in skb_checksum_complete
3984 * non-zero: value of invalid checksum
3985 */
3986#define __skb_checksum_validate(skb, proto, complete, \
3987 zero_okay, check, compute_pseudo) \
3988({ \
3989 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003990 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003991 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3992 __ret = __skb_checksum_validate_complete(skb, \
3993 complete, compute_pseudo(skb, proto)); \
3994 __ret; \
3995})
3996
3997#define skb_checksum_init(skb, proto, compute_pseudo) \
3998 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3999
4000#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4001 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4002
4003#define skb_checksum_validate(skb, proto, compute_pseudo) \
4004 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4005
4006#define skb_checksum_validate_zero_check(skb, proto, check, \
4007 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01004008 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004009
4010#define skb_checksum_simple_validate(skb) \
4011 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4012
Tom Herbertd96535a2014-08-31 15:12:42 -07004013static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4014{
Davide Caratti219f1d792017-05-18 15:44:39 +02004015 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07004016}
4017
Li RongQinge4aa33a2019-07-04 17:03:26 +08004018static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
Tom Herbertd96535a2014-08-31 15:12:42 -07004019{
4020 skb->csum = ~pseudo;
4021 skb->ip_summed = CHECKSUM_COMPLETE;
4022}
4023
Li RongQinge4aa33a2019-07-04 17:03:26 +08004024#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
Tom Herbertd96535a2014-08-31 15:12:42 -07004025do { \
4026 if (__skb_checksum_convert_check(skb)) \
Li RongQinge4aa33a2019-07-04 17:03:26 +08004027 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
Tom Herbertd96535a2014-08-31 15:12:42 -07004028} while (0)
4029
Tom Herbert15e23962015-02-10 16:30:31 -08004030static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4031 u16 start, u16 offset)
4032{
4033 skb->ip_summed = CHECKSUM_PARTIAL;
4034 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4035 skb->csum_offset = offset - start;
4036}
4037
Tom Herbertdcdc8992015-02-02 16:07:34 -08004038/* Update skbuf and packet to reflect the remote checksum offload operation.
4039 * When called, ptr indicates the starting point for skb->csum when
4040 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4041 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4042 */
4043static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08004044 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08004045{
4046 __wsum delta;
4047
Tom Herbert15e23962015-02-10 16:30:31 -08004048 if (!nopartial) {
4049 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4050 return;
4051 }
4052
Tom Herbertdcdc8992015-02-02 16:07:34 -08004053 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4054 __skb_checksum_complete(skb);
4055 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4056 }
4057
4058 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4059
4060 /* Adjust skb->csum since we changed the packet */
4061 skb->csum = csum_add(skb->csum, delta);
4062}
4063
Florian Westphalcb9c6832017-01-23 18:21:56 +01004064static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4065{
4066#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004067 return (void *)(skb->_nfct & NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01004068#else
4069 return NULL;
4070#endif
4071}
4072
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004073static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004074{
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004075#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4076 return skb->_nfct;
4077#else
4078 return 0UL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004079#endif
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004080}
4081
4082static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4083{
4084#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4085 skb->_nfct = nfct;
4086#endif
4087}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004088
4089#ifdef CONFIG_SKB_EXTENSIONS
4090enum skb_ext_id {
4091#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4092 SKB_EXT_BRIDGE_NF,
4093#endif
Florian Westphal41650792018-12-18 17:15:27 +01004094#ifdef CONFIG_XFRM
4095 SKB_EXT_SEC_PATH,
4096#endif
Paul Blakey95a72332019-09-04 16:56:37 +03004097#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4098 TC_SKB_EXT,
4099#endif
Mat Martineau3ee17bc2020-01-09 07:59:19 -08004100#if IS_ENABLED(CONFIG_MPTCP)
4101 SKB_EXT_MPTCP,
4102#endif
Florian Westphaldf5042f2018-12-18 17:15:16 +01004103 SKB_EXT_NUM, /* must be last */
4104};
4105
4106/**
4107 * struct skb_ext - sk_buff extensions
4108 * @refcnt: 1 on allocation, deallocated on 0
4109 * @offset: offset to add to @data to obtain extension address
4110 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4111 * @data: start of extension data, variable sized
4112 *
4113 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4114 * to use 'u8' types while allowing up to 2kb worth of extension data.
4115 */
4116struct skb_ext {
4117 refcount_t refcnt;
4118 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4119 u8 chunks; /* same */
4120 char data[0] __aligned(8);
4121};
4122
Paolo Abeni8b69a802020-01-09 07:59:24 -08004123struct skb_ext *__skb_ext_alloc(void);
4124void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4125 struct skb_ext *ext);
Florian Westphaldf5042f2018-12-18 17:15:16 +01004126void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4127void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4128void __skb_ext_put(struct skb_ext *ext);
4129
4130static inline void skb_ext_put(struct sk_buff *skb)
4131{
4132 if (skb->active_extensions)
4133 __skb_ext_put(skb->extensions);
4134}
4135
Florian Westphaldf5042f2018-12-18 17:15:16 +01004136static inline void __skb_ext_copy(struct sk_buff *dst,
4137 const struct sk_buff *src)
4138{
4139 dst->active_extensions = src->active_extensions;
4140
4141 if (src->active_extensions) {
4142 struct skb_ext *ext = src->extensions;
4143
4144 refcount_inc(&ext->refcnt);
4145 dst->extensions = ext;
4146 }
4147}
4148
4149static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4150{
4151 skb_ext_put(dst);
4152 __skb_ext_copy(dst, src);
4153}
4154
4155static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4156{
4157 return !!ext->offset[i];
4158}
4159
4160static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4161{
4162 return skb->active_extensions & (1 << id);
4163}
4164
4165static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4166{
4167 if (skb_ext_exist(skb, id))
4168 __skb_ext_del(skb, id);
4169}
4170
4171static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4172{
4173 if (skb_ext_exist(skb, id)) {
4174 struct skb_ext *ext = skb->extensions;
4175
4176 return (void *)ext + (ext->offset[id] << 3);
4177 }
4178
4179 return NULL;
4180}
Florian Westphal174e2382019-09-26 20:37:05 +02004181
4182static inline void skb_ext_reset(struct sk_buff *skb)
4183{
4184 if (unlikely(skb->active_extensions)) {
4185 __skb_ext_put(skb->extensions);
4186 skb->active_extensions = 0;
4187 }
4188}
Florian Westphal677bf082019-11-21 06:56:23 +01004189
4190static inline bool skb_has_extensions(struct sk_buff *skb)
4191{
4192 return unlikely(skb->active_extensions);
4193}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004194#else
4195static inline void skb_ext_put(struct sk_buff *skb) {}
Florian Westphal174e2382019-09-26 20:37:05 +02004196static inline void skb_ext_reset(struct sk_buff *skb) {}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004197static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4198static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4199static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
Florian Westphal677bf082019-11-21 06:56:23 +01004200static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
Florian Westphaldf5042f2018-12-18 17:15:16 +01004201#endif /* CONFIG_SKB_EXTENSIONS */
4202
Florian Westphal895b5c92019-09-29 20:54:03 +02004203static inline void nf_reset_ct(struct sk_buff *skb)
Patrick McHardya193a4a2006-03-20 19:23:05 -08004204{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004205#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004206 nf_conntrack_put(skb_nfct(skb));
4207 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004208#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08004209}
4210
Patrick McHardy124dff02013-04-05 20:42:05 +02004211static inline void nf_reset_trace(struct sk_buff *skb)
4212{
Florian Westphal478b3602014-02-15 23:48:45 +01004213#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00004214 skb->nf_trace = 0;
4215#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004216}
4217
Ye Yin2b5ec1a2017-10-26 16:57:05 +08004218static inline void ipvs_reset(struct sk_buff *skb)
4219{
4220#if IS_ENABLED(CONFIG_IP_VS)
4221 skb->ipvs_property = 0;
4222#endif
4223}
4224
Florian Westphalde8bda12018-12-18 17:15:17 +01004225/* Note: This doesn't put any conntrack info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07004226static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4227 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004228{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004229#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004230 dst->_nfct = src->_nfct;
4231 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004232#endif
Florian Westphal478b3602014-02-15 23:48:45 +01004233#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07004234 if (copy)
4235 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01004236#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004237}
4238
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004239static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4240{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004241#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004242 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004243#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07004244 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004245}
4246
James Morris984bc162006-06-09 00:29:17 -07004247#ifdef CONFIG_NETWORK_SECMARK
4248static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4249{
4250 to->secmark = from->secmark;
4251}
4252
4253static inline void skb_init_secmark(struct sk_buff *skb)
4254{
4255 skb->secmark = 0;
4256}
4257#else
4258static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4259{ }
4260
4261static inline void skb_init_secmark(struct sk_buff *skb)
4262{ }
4263#endif
4264
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004265static inline int secpath_exists(const struct sk_buff *skb)
4266{
4267#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004268 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004269#else
4270 return 0;
4271#endif
4272}
4273
Eric W. Biederman574f7192014-04-01 12:20:24 -07004274static inline bool skb_irq_freeable(const struct sk_buff *skb)
4275{
4276 return !skb->destructor &&
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004277 !secpath_exists(skb) &&
Florian Westphalcb9c6832017-01-23 18:21:56 +01004278 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07004279 !skb->_skb_refdst &&
4280 !skb_has_frag_list(skb);
4281}
4282
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004283static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4284{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004285 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004286}
4287
Stephen Hemminger92477442009-03-21 13:39:26 -07004288static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004289{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004290 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004291}
4292
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004293static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4294{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004295 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004296}
4297
David S. Millerd5a9e242009-01-27 16:22:11 -08004298static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4299{
4300 skb->queue_mapping = rx_queue + 1;
4301}
4302
Stephen Hemminger92477442009-03-21 13:39:26 -07004303static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004304{
4305 return skb->queue_mapping - 1;
4306}
4307
Stephen Hemminger92477442009-03-21 13:39:26 -07004308static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004309{
Eric Dumazeta02cec22010-09-22 20:43:57 +00004310 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08004311}
4312
Julian Anastasov4ff06202017-02-06 23:14:12 +02004313static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4314{
4315 skb->dst_pending_confirm = val;
4316}
4317
4318static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4319{
4320 return skb->dst_pending_confirm != 0;
4321}
4322
Florian Westphal2294be0f2018-12-18 17:15:20 +01004323static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004324{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004325#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004326 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004327#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004328 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004329#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004330}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004331
Pravin B Shelar68c33162013-02-14 14:02:41 +00004332/* Keeps track of mac header offset relative to skb->head.
4333 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4334 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07004335 * tunnel skb it points to outer mac header.
4336 * Keeps track of level of encapsulation of network headers.
4337 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00004338struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04004339 union {
4340 int mac_offset;
4341 int data_offset;
4342 };
Eric Dumazet3347c962013-10-19 11:42:56 -07004343 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08004344 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004345 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00004346};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03004347#define SKB_SGO_CB_OFFSET 32
4348#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00004349
4350static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4351{
4352 return (skb_mac_header(inner_skb) - inner_skb->head) -
4353 SKB_GSO_CB(inner_skb)->mac_offset;
4354}
4355
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00004356static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4357{
4358 int new_headroom, headroom;
4359 int ret;
4360
4361 headroom = skb_headroom(skb);
4362 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4363 if (ret)
4364 return ret;
4365
4366 new_headroom = skb_headroom(skb);
4367 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4368 return 0;
4369}
4370
Alexander Duyck08b64fc2016-02-05 15:27:49 -08004371static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4372{
4373 /* Do not update partial checksums if remote checksum is enabled. */
4374 if (skb->remcsum_offload)
4375 return;
4376
4377 SKB_GSO_CB(skb)->csum = res;
4378 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4379}
4380
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004381/* Compute the checksum for a gso segment. First compute the checksum value
4382 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4383 * then add in skb->csum (checksum from csum_start to end of packet).
4384 * skb->csum and csum_start are then updated to reflect the checksum of the
4385 * resultant packet starting from the transport header-- the resultant checksum
4386 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4387 * header.
4388 */
4389static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4390{
Alexander Duyck76443452016-02-05 15:27:37 -08004391 unsigned char *csum_start = skb_transport_header(skb);
4392 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4393 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004394
Alexander Duyck76443452016-02-05 15:27:37 -08004395 SKB_GSO_CB(skb)->csum = res;
4396 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004397
Alexander Duyck76443452016-02-05 15:27:37 -08004398 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004399}
4400
David S. Millerbdcc0922012-03-07 20:53:36 -05004401static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07004402{
4403 return skb_shinfo(skb)->gso_size;
4404}
4405
Eric Dumazet36a8f392013-09-29 01:21:32 -07004406/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004407static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004408{
4409 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4410}
4411
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004412/* Note: Should be called only if skb_is_gso(skb) is true */
4413static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4414{
4415 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4416}
4417
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004418/* Note: Should be called only if skb_is_gso(skb) is true */
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004419static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4420{
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004421 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004422}
4423
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004424static inline void skb_gso_reset(struct sk_buff *skb)
4425{
4426 skb_shinfo(skb)->gso_size = 0;
4427 skb_shinfo(skb)->gso_segs = 0;
4428 skb_shinfo(skb)->gso_type = 0;
4429}
4430
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004431static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4432 u16 increment)
4433{
4434 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4435 return;
4436 shinfo->gso_size += increment;
4437}
4438
4439static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4440 u16 decrement)
4441{
4442 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4443 return;
4444 shinfo->gso_size -= decrement;
4445}
4446
Joe Perches7965bd42013-09-26 14:48:15 -07004447void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004448
4449static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4450{
4451 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4452 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004453 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4454
Alexander Duyckb78462e2010-06-02 12:24:37 +00004455 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4456 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004457 __skb_warn_lro_forwarding(skb);
4458 return true;
4459 }
4460 return false;
4461}
4462
Herbert Xu35fc92a2007-03-26 23:22:20 -07004463static inline void skb_forward_csum(struct sk_buff *skb)
4464{
4465 /* Unfortunately we don't support this one. Any brave souls? */
4466 if (skb->ip_summed == CHECKSUM_COMPLETE)
4467 skb->ip_summed = CHECKSUM_NONE;
4468}
4469
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004470/**
4471 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4472 * @skb: skb to check
4473 *
4474 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4475 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4476 * use this helper, to document places where we make this assertion.
4477 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004478static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004479{
4480#ifdef DEBUG
4481 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4482#endif
4483}
4484
Rusty Russellf35d9d82008-02-04 23:49:54 -05004485bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004486
Paul Durranted1f50c2014-01-09 10:02:46 +00004487int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004488struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4489 unsigned int transport_len,
4490 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004491
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004492/**
4493 * skb_head_is_locked - Determine if the skb->head is locked down
4494 * @skb: skb to check
4495 *
4496 * The head on skbs build around a head frag can be removed if they are
4497 * not cloned. This function returns true if the skb head is locked down
4498 * due to either being allocated via kmalloc, or by being a clone with
4499 * multiple references to the head.
4500 */
4501static inline bool skb_head_is_locked(const struct sk_buff *skb)
4502{
4503 return !skb->head_frag || skb_cloned(skb);
4504}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004505
Edward Cree179bc672016-02-11 20:48:04 +00004506/* Local Checksum Offload.
4507 * Compute outer checksum based on the assumption that the
4508 * inner checksum will be offloaded later.
Otto Sabartd0dcde62019-01-06 00:29:15 +01004509 * See Documentation/networking/checksum-offloads.rst for
Edward Creee8ae7b02016-02-11 21:03:37 +00004510 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004511 * Fill in outer checksum adjustment (e.g. with sum of outer
4512 * pseudo-header) before calling.
4513 * Also ensure that inner checksum is in linear data area.
4514 */
4515static inline __wsum lco_csum(struct sk_buff *skb)
4516{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004517 unsigned char *csum_start = skb_checksum_start(skb);
4518 unsigned char *l4_hdr = skb_transport_header(skb);
4519 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004520
4521 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004522 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4523 skb->csum_offset));
4524
Edward Cree179bc672016-02-11 20:48:04 +00004525 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004526 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004527 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004528 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004529}
4530
Linus Torvalds1da177e2005-04-16 15:20:36 -07004531#endif /* __KERNEL__ */
4532#endif /* _LINUX_SKBUFF_H */