blob: e5387932c266e7e05eff7480857c290bdb7d28bb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Arun Sharma600634972011-07-26 16:09:06 -070026#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <asm/types.h>
28#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000032#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070033#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000034#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000035#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080036#include <linux/sched.h>
Ingo Molnare6017572017-02-01 16:36:40 +010037#include <linux/sched/clock.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020038#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020039#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020040#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040041#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070042#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Tom Herbert7a6ae712015-12-14 11:19:47 -080044/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010089 *
90 * CHECKSUM_NONE:
91 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080092 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010093 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200112 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100134 *
135 * CHECKSUM_PARTIAL:
136 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100146 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100179 * CHECKSUM_UNNECESSARY:
180 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100183 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100187 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
Davide Carattidba00302017-05-18 15:44:40 +0200192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217 */
218
Herbert Xu60476372007-04-09 11:59:39 -0700219/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Tom Herbert77cffe22014-08-27 21:26:46 -0700225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700229#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700242struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800243struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800244struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800245struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248struct nf_conntrack {
249 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700251#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254struct nf_bridge_info {
Reshetova, Elena53869ce2017-06-30 13:07:57 +0300255 refcount_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200260 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200264 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000265 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200269 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200270 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200279 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280};
281#endif
282
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290};
291
292struct sk_buff;
293
Ian Campbell9d4dde52011-12-22 23:39:14 +0000294/* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000300 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000301#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700302#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000303#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000304#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000305#endif
Hans Westgaard Ry5f74f822016-02-03 09:26:57 +0100306extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000316 struct {
317 struct page *p;
318 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700320 __u32 page_offset;
321 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326};
327
Eric Dumazet9e903e02011-10-18 21:00:24 +0000328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
Willem de Bruijnc613c202017-07-31 08:15:47 -0400348static inline bool skb_frag_must_loop(struct page *p)
349{
350#if defined(CONFIG_HIGHMEM)
351 if (PageHighMem(p))
352 return true;
353#endif
354 return false;
355}
356
357/**
358 * skb_frag_foreach_page - loop over pages in a fragment
359 *
360 * @f: skb frag to operate on
361 * @f_off: offset from start of f->page.p
362 * @f_len: length from f_off to loop over
363 * @p: (temp var) current page
364 * @p_off: (temp var) offset from start of current page,
365 * non-zero only on first page.
366 * @p_len: (temp var) length in current page,
367 * < PAGE_SIZE only on first and last page.
368 * @copied: (temp var) length so far, excluding current p_len.
369 *
370 * A fragment can hold a compound page, in which case per-page
371 * operations, notably kmap_atomic, must be called for each
372 * regular page.
373 */
374#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
375 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
376 p_off = (f_off) & (PAGE_SIZE - 1), \
377 p_len = skb_frag_must_loop(p) ? \
378 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
379 copied = 0; \
380 copied < f_len; \
381 copied += p_len, p++, p_off = 0, \
382 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
383
Patrick Ohlyac45f602009-02-12 05:03:37 +0000384#define HAVE_HW_TIME_STAMP
385
386/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800387 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000388 * @hwtstamp: hardware time stamp transformed into duration
389 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000390 *
391 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400392 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000393 *
394 * hwtstamps can only be compared against other hwtstamps from
395 * the same device.
396 *
397 * This structure is attached to packets as part of the
398 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
399 */
400struct skb_shared_hwtstamps {
401 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000402};
403
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000404/* Definitions for tx_flags in struct skb_shared_info */
405enum {
406 /* generate hardware time stamp */
407 SKBTX_HW_TSTAMP = 1 << 0,
408
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400409 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000410 SKBTX_SW_TSTAMP = 1 << 1,
411
412 /* device driver is going to provide hardware time stamp */
413 SKBTX_IN_PROGRESS = 1 << 2,
414
Shirley Maa6686f22011-07-06 12:22:12 +0000415 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000416 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100417
418 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000419 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000420
421 /* This indicates at least one fragment might be overwritten
422 * (as in vmsplice(), sendfile() ...)
423 * If we need to compute a TX checksum, we'll need to copy
424 * all frags to avoid possible bad checksum
425 */
426 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400427
428 /* generate software time stamp when entering packet scheduling */
429 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000430};
431
Willem de Bruijn52267792017-08-03 16:29:39 -0400432#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400433#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400434 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400435#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
436
Shirley Maa6686f22011-07-06 12:22:12 +0000437/*
438 * The callback notifies userspace to release buffers when skb DMA is done in
439 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000440 * The zerocopy_success argument is true if zero copy transmit occurred,
441 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000442 * The ctx field is used to track device context.
443 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000444 */
445struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000446 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000447 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000448 unsigned long desc;
Willem de Bruijn52267792017-08-03 16:29:39 -0400449 u16 zerocopy:1;
450 atomic_t refcnt;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000451};
452
Willem de Bruijn52267792017-08-03 16:29:39 -0400453#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
454
455struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
456
457static inline void sock_zerocopy_get(struct ubuf_info *uarg)
458{
459 atomic_inc(&uarg->refcnt);
460}
461
462void sock_zerocopy_put(struct ubuf_info *uarg);
463void sock_zerocopy_put_abort(struct ubuf_info *uarg);
464
465void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
466
467int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
468 struct msghdr *msg, int len,
469 struct ubuf_info *uarg);
470
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471/* This data is invariant across clones and lives at
472 * the end of the header data, ie. at skb->end.
473 */
474struct skb_shared_info {
Steffen Klassert7f564522017-04-08 20:36:24 +0200475 unsigned short _unused;
Ian Campbell9f42f122012-01-05 07:13:39 +0000476 unsigned char nr_frags;
477 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700478 unsigned short gso_size;
479 /* Warning: this field is not always filled in (UFO)! */
480 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000482 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200483 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400484 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000485 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700486
487 /*
488 * Warning : all fields before dataref are cleared in __alloc_skb()
489 */
490 atomic_t dataref;
491
Johann Baudy69e3c752009-05-18 22:11:22 -0700492 /* Intermediate layers must ensure that destructor_arg
493 * remains valid until skb destructor */
494 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000495
Eric Dumazetfed66382010-07-22 19:09:08 +0000496 /* must be last field, see pskb_expand_head() */
497 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498};
499
500/* We divide dataref into two halves. The higher 16 bits hold references
501 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700502 * the entire skb->data. A clone of a headerless skb holds the length of
503 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 *
505 * All users must obey the rule that the skb->data reference count must be
506 * greater than or equal to the payload reference count.
507 *
508 * Holding a reference to the payload part means that the user does not
509 * care about modifications to the header part of skb->data.
510 */
511#define SKB_DATAREF_SHIFT 16
512#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
513
David S. Millerd179cd12005-08-17 14:57:30 -0700514
515enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700516 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
517 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
518 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700519};
520
Herbert Xu79671682006-06-22 02:40:14 -0700521enum {
522 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700523
524 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700525 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700526
527 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700528 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700529
David S. Millerd9d30ad2017-07-03 07:31:57 -0700530 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800531
David S. Millerd9d30ad2017-07-03 07:31:57 -0700532 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000533
David S. Millerd9d30ad2017-07-03 07:31:57 -0700534 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000535
David S. Millerd9d30ad2017-07-03 07:31:57 -0700536 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000537
David S. Millerd9d30ad2017-07-03 07:31:57 -0700538 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700539
David S. Millerd9d30ad2017-07-03 07:31:57 -0700540 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700541
David S. Millerd9d30ad2017-07-03 07:31:57 -0700542 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700543
David S. Millerd9d30ad2017-07-03 07:31:57 -0700544 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700545
David S. Millerd9d30ad2017-07-03 07:31:57 -0700546 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400547
David S. Millerd9d30ad2017-07-03 07:31:57 -0700548 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400549
David S. Millerd9d30ad2017-07-03 07:31:57 -0700550 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300551
David S. Millerd9d30ad2017-07-03 07:31:57 -0700552 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200553
David S. Millerd9d30ad2017-07-03 07:31:57 -0700554 SKB_GSO_ESP = 1 << 15,
Herbert Xu79671682006-06-22 02:40:14 -0700555};
556
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700557#if BITS_PER_LONG > 32
558#define NET_SKBUFF_DATA_USES_OFFSET 1
559#endif
560
561#ifdef NET_SKBUFF_DATA_USES_OFFSET
562typedef unsigned int sk_buff_data_t;
563#else
564typedef unsigned char *sk_buff_data_t;
565#endif
566
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567/**
568 * struct sk_buff - socket buffer
569 * @next: Next buffer in list
570 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800571 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800572 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700573 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700575 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000576 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700577 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 * @len: Length of actual data
579 * @data_len: Data length
580 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700581 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700582 * @csum: Checksum (must include start/offset pair)
583 * @csum_start: Offset from skb->head where checksumming should start
584 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700585 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700586 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700588 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 * @nohdr: Payload reference only, must not modify header
590 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700591 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700592 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500593 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500594 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500595 * @tc_redirected: packet was redirected by a tc action
596 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800597 * @peeked: this packet has been seen already, so stats have been
598 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700599 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700600 * @protocol: Packet protocol from driver
601 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100602 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800604 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700606 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700607 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700608 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700609 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700610 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700611 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700612 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700613 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100614 * @wifi_acked_valid: wifi_acked was set
615 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000616 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200617 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200618 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300619 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700620 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700621 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000622 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700623 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000624 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000625 * @inner_transport_header: Inner transport layer header (encapsulation)
626 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000627 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700628 * @transport_header: Transport layer header
629 * @network_header: Network layer header
630 * @mac_header: Link layer header
631 * @tail: Tail pointer
632 * @end: End pointer
633 * @head: Head of buffer
634 * @data: Data head pointer
635 * @truesize: Buffer size
636 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 */
638
639struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800640 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800641 struct {
642 /* These two members must be first. */
643 struct sk_buff *next;
644 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000645
Eric Dumazet56b17422014-11-03 08:19:53 -0800646 union {
647 ktime_t tstamp;
Eric Dumazet9a568de2017-05-16 14:00:14 -0700648 u64 skb_mstamp;
Eric Dumazet56b17422014-11-03 08:19:53 -0800649 };
650 };
651 struct rb_node rbnode; /* used in netem & tcp stack */
652 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000653 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
Eric Dumazetc84d9492016-12-08 11:41:55 -0800655 union {
656 struct net_device *dev;
657 /* Some protocols might use this space to store information,
658 * while device pointer would be NULL.
659 * UDP receive path is one user.
660 */
661 unsigned long dev_scratch;
662 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 /*
664 * This is the control buffer. It is free to use for every
665 * layer. Please put your private variables there. If you
666 * want to keep them across layers you have to do a skb_clone()
667 * first. This is owned by whoever has the skb queued ATM.
668 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000669 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
Eric Dumazet7fee2262010-05-11 23:19:48 +0000671 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700672 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000673#ifdef CONFIG_XFRM
674 struct sec_path *sp;
675#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700676#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100677 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700678#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200679#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700680 struct nf_bridge_info *nf_bridge;
681#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700683 data_len;
684 __u16 mac_len,
685 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700686
687 /* Following fields are _not_ copied in __copy_skb_header()
688 * Note that queue_mapping is here mostly to fill a hole.
689 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200690 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700691 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200692
693/* if you move cloned around you also must adapt those constants */
694#ifdef __BIG_ENDIAN_BITFIELD
695#define CLONED_MASK (1 << 7)
696#else
697#define CLONED_MASK 1
698#endif
699#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
700
701 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700702 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700703 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800704 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800705 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700706 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200707 xmit_more:1,
708 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200709 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710
Eric Dumazetb1937222014-09-28 22:18:47 -0700711 /* fields enclosed in headers_start/headers_end are copied
712 * using a single memcpy() in __copy_skb_header()
713 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700714 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700715 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700716 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200717
718/* if you move pkt_type around you also must adapt those constants */
719#ifdef __BIG_ENDIAN_BITFIELD
720#define PKT_TYPE_MAX (7 << 5)
721#else
722#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100723#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200724#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
725
726 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700727 __u8 pkt_type:3;
728 __u8 pfmemalloc:1;
729 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730
Eric Dumazetb1937222014-09-28 22:18:47 -0700731 __u8 nf_trace:1;
732 __u8 ip_summed:2;
733 __u8 ooo_okay:1;
734 __u8 l4_hash:1;
735 __u8 sw_hash:1;
736 __u8 wifi_acked_valid:1;
737 __u8 wifi_acked:1;
738
739 __u8 no_fcs:1;
740 /* Indicates the inner headers are valid in the skbuff. */
741 __u8 encapsulation:1;
742 __u8 encap_hdr_csum:1;
743 __u8 csum_valid:1;
744 __u8 csum_complete_sw:1;
745 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200746 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700747
Julian Anastasov4ff06202017-02-06 23:14:12 +0200748 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700749#ifdef CONFIG_IPV6_NDISC_NODETYPE
750 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700752 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700753 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800754 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200755#ifdef CONFIG_NET_SWITCHDEV
756 __u8 offload_fwd_mark:1;
757#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500758#ifdef CONFIG_NET_CLS_ACT
759 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500760 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500761 __u8 tc_redirected:1;
762 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500763#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000764
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700766 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200768
Eric Dumazetb1937222014-09-28 22:18:47 -0700769 union {
770 __wsum csum;
771 struct {
772 __u16 csum_start;
773 __u16 csum_offset;
774 };
775 };
776 __u32 priority;
777 int skb_iif;
778 __u32 hash;
779 __be16 vlan_proto;
780 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800781#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
782 union {
783 unsigned int napi_id;
784 unsigned int sender_cpu;
785 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700786#endif
James Morris984bc162006-06-09 00:29:17 -0700787#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200788 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700789#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700790
Neil Horman3b885782009-10-12 13:26:31 -0700791 union {
792 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000793 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700794 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Tom Herbert8bce6d72014-09-29 20:22:29 -0700796 union {
797 __be16 inner_protocol;
798 __u8 inner_ipproto;
799 };
800
Simon Horman1a37e412013-05-23 21:02:51 +0000801 __u16 inner_transport_header;
802 __u16 inner_network_header;
803 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700804
805 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000806 __u16 transport_header;
807 __u16 network_header;
808 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700809
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700810 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700811 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700812 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700815 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700816 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700818 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700819 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300820 refcount_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821};
822
823#ifdef __KERNEL__
824/*
825 * Handling routines are only of interest to the kernel
826 */
827#include <linux/slab.h>
828
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
Mel Gormanc93bdd02012-07-31 16:44:19 -0700830#define SKB_ALLOC_FCLONE 0x01
831#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800832#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700833
834/* Returns true if the skb was allocated from PFMEMALLOC reserves */
835static inline bool skb_pfmemalloc(const struct sk_buff *skb)
836{
837 return unlikely(skb->pfmemalloc);
838}
839
Eric Dumazet7fee2262010-05-11 23:19:48 +0000840/*
841 * skb might have a dst pointer attached, refcounted or not.
842 * _skb_refdst low order bit is set if refcount was _not_ taken
843 */
844#define SKB_DST_NOREF 1UL
845#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
846
Florian Westphala9e419d2017-01-23 18:21:59 +0100847#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000848/**
849 * skb_dst - returns skb dst_entry
850 * @skb: buffer
851 *
852 * Returns skb dst_entry, regardless of reference taken or not.
853 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000854static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
855{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000856 /* If refdst was not refcounted, check we still are in a
857 * rcu_read_lock section
858 */
859 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
860 !rcu_read_lock_held() &&
861 !rcu_read_lock_bh_held());
862 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000863}
864
Eric Dumazet7fee2262010-05-11 23:19:48 +0000865/**
866 * skb_dst_set - sets skb dst
867 * @skb: buffer
868 * @dst: dst entry
869 *
870 * Sets skb dst, assuming a reference was taken on dst and should
871 * be released by skb_dst_drop()
872 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000873static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
874{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000875 skb->_skb_refdst = (unsigned long)dst;
876}
877
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200878/**
879 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
880 * @skb: buffer
881 * @dst: dst entry
882 *
883 * Sets skb dst, assuming a reference was not taken on dst.
884 * If dst entry is cached, we do not take reference and dst_release
885 * will be avoided by refdst_drop. If dst entry is not cached, we take
886 * reference, so that last dst_release can destroy the dst immediately.
887 */
888static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
889{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100890 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
891 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200892}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000893
894/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300895 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000896 * @skb: buffer
897 */
898static inline bool skb_dst_is_noref(const struct sk_buff *skb)
899{
900 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000901}
902
Eric Dumazet511c3f92009-06-02 05:14:27 +0000903static inline struct rtable *skb_rtable(const struct sk_buff *skb)
904{
Eric Dumazetadf30902009-06-02 05:19:30 +0000905 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000906}
907
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400908/* For mangling skb->pkt_type from user space side from applications
909 * such as nft, tc, etc, we only allow a conservative subset of
910 * possible pkt_types to be set.
911*/
912static inline bool skb_pkt_type_ok(u32 ptype)
913{
914 return ptype <= PACKET_OTHERHOST;
915}
916
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200917static inline unsigned int skb_napi_id(const struct sk_buff *skb)
918{
919#ifdef CONFIG_NET_RX_BUSY_POLL
920 return skb->napi_id;
921#else
922 return 0;
923#endif
924}
925
Paolo Abeni3889a8032017-06-12 11:23:41 +0200926/* decrement the reference count and return true if we can free the skb */
927static inline bool skb_unref(struct sk_buff *skb)
928{
929 if (unlikely(!skb))
930 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +0300931 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200932 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +0300933 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200934 return false;
935
936 return true;
937}
938
Paolo Abeni0a463c72017-06-12 11:23:42 +0200939void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700940void kfree_skb(struct sk_buff *skb);
941void kfree_skb_list(struct sk_buff *segs);
942void skb_tx_error(struct sk_buff *skb);
943void consume_skb(struct sk_buff *skb);
Paolo Abeni0a463c72017-06-12 11:23:42 +0200944void consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700945void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000946extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000947
Joe Perches7965bd42013-09-26 14:48:15 -0700948void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
949bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
950 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000951
Joe Perches7965bd42013-09-26 14:48:15 -0700952struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
953 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700954struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700955struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700956static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100957 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700958{
Eric Dumazet564824b2010-10-11 19:05:25 +0000959 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700960}
961
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700962struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
963 unsigned long data_len,
964 int max_page_order,
965 int *errcode,
966 gfp_t gfp_mask);
967
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700968/* Layout of fast clones : [skb1][skb2][fclone_ref] */
969struct sk_buff_fclones {
970 struct sk_buff skb1;
971
972 struct sk_buff skb2;
973
Reshetova, Elena26385952017-06-30 13:07:59 +0300974 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700975};
976
977/**
978 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700979 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700980 * @skb: buffer
981 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900982 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700983 * Some drivers call skb_orphan() in their ndo_start_xmit(),
984 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700985 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700986static inline bool skb_fclone_busy(const struct sock *sk,
987 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700988{
989 const struct sk_buff_fclones *fclones;
990
991 fclones = container_of(skb, struct sk_buff_fclones, skb1);
992
993 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +0300994 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700995 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700996}
997
David S. Millerd179cd12005-08-17 14:57:30 -0700998static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100999 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001000{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001001 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001002}
1003
Joe Perches7965bd42013-09-26 14:48:15 -07001004struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1005int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1006struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1007struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001008struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1009 gfp_t gfp_mask, bool fclone);
1010static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1011 gfp_t gfp_mask)
1012{
1013 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1014}
Eric Dumazet117632e2011-12-03 21:39:53 +00001015
Joe Perches7965bd42013-09-26 14:48:15 -07001016int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1017struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1018 unsigned int headroom);
1019struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1020 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001021int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1022 int offset, int len);
1023int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1024 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001025int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1026int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +00001027#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
Joe Perches7965bd42013-09-26 14:48:15 -07001029int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1030 int getfrag(void *from, char *to, int offset,
1031 int len, int odd, struct sk_buff *skb),
1032 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001033
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001034int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1035 int offset, size_t size);
1036
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001037struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001038 __u32 lower_offset;
1039 __u32 upper_offset;
1040 __u32 frag_idx;
1041 __u32 stepped_offset;
1042 struct sk_buff *root_skb;
1043 struct sk_buff *cur_skb;
1044 __u8 *frag_data;
1045};
1046
Joe Perches7965bd42013-09-26 14:48:15 -07001047void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1048 unsigned int to, struct skb_seq_state *st);
1049unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1050 struct skb_seq_state *st);
1051void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001052
Joe Perches7965bd42013-09-26 14:48:15 -07001053unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001054 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001055
Tom Herbert09323cc2013-12-15 22:16:19 -08001056/*
1057 * Packet hash types specify the type of hash in skb_set_hash.
1058 *
1059 * Hash types refer to the protocol layer addresses which are used to
1060 * construct a packet's hash. The hashes are used to differentiate or identify
1061 * flows of the protocol layer for the hash type. Hash types are either
1062 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1063 *
1064 * Properties of hashes:
1065 *
1066 * 1) Two packets in different flows have different hash values
1067 * 2) Two packets in the same flow should have the same hash value
1068 *
1069 * A hash at a higher layer is considered to be more specific. A driver should
1070 * set the most specific hash possible.
1071 *
1072 * A driver cannot indicate a more specific hash than the layer at which a hash
1073 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1074 *
1075 * A driver may indicate a hash level which is less specific than the
1076 * actual layer the hash was computed on. For instance, a hash computed
1077 * at L4 may be considered an L3 hash. This should only be done if the
1078 * driver can't unambiguously determine that the HW computed the hash at
1079 * the higher layer. Note that the "should" in the second property above
1080 * permits this.
1081 */
1082enum pkt_hash_types {
1083 PKT_HASH_TYPE_NONE, /* Undefined type */
1084 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1085 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1086 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1087};
1088
Tom Herbertbcc83832015-09-01 09:24:24 -07001089static inline void skb_clear_hash(struct sk_buff *skb)
1090{
1091 skb->hash = 0;
1092 skb->sw_hash = 0;
1093 skb->l4_hash = 0;
1094}
1095
1096static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1097{
1098 if (!skb->l4_hash)
1099 skb_clear_hash(skb);
1100}
1101
1102static inline void
1103__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1104{
1105 skb->l4_hash = is_l4;
1106 skb->sw_hash = is_sw;
1107 skb->hash = hash;
1108}
1109
Tom Herbert09323cc2013-12-15 22:16:19 -08001110static inline void
1111skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1112{
Tom Herbertbcc83832015-09-01 09:24:24 -07001113 /* Used by drivers to set hash from HW */
1114 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1115}
1116
1117static inline void
1118__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1119{
1120 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001121}
1122
Tom Herberte5276932015-09-01 09:24:23 -07001123void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001124u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001125u32 skb_get_poff(const struct sk_buff *skb);
1126u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1127 const struct flow_keys *keys, int hlen);
1128__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1129 void *data, int hlen_proto);
1130
1131static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1132 int thoff, u8 ip_proto)
1133{
1134 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1135}
1136
1137void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1138 const struct flow_dissector_key *key,
1139 unsigned int key_count);
1140
1141bool __skb_flow_dissect(const struct sk_buff *skb,
1142 struct flow_dissector *flow_dissector,
1143 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001144 void *data, __be16 proto, int nhoff, int hlen,
1145 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001146
1147static inline bool skb_flow_dissect(const struct sk_buff *skb,
1148 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001149 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001150{
1151 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001152 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001153}
1154
1155static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001156 struct flow_keys *flow,
1157 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001158{
1159 memset(flow, 0, sizeof(*flow));
1160 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001161 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001162}
1163
1164static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1165 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001166 int nhoff, int hlen,
1167 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001168{
1169 memset(flow, 0, sizeof(*flow));
1170 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001171 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001172}
1173
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001174static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001175{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001176 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001177 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001178
Tom Herbert61b905d2014-03-24 15:34:47 -07001179 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001180}
1181
David S. Miller20a17bf2015-09-01 21:19:17 -07001182static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001183{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001184 if (!skb->l4_hash && !skb->sw_hash) {
1185 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001186 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001187
Tom Herbertde4c1f82015-09-01 18:11:04 -07001188 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001189 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001190
1191 return skb->hash;
1192}
1193
Tom Herbert50fb7992015-05-01 11:30:12 -07001194__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1195
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001196static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1197{
Tom Herbert61b905d2014-03-24 15:34:47 -07001198 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001199}
1200
Tom Herbert3df7a742013-12-15 22:16:29 -08001201static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1202{
Tom Herbert61b905d2014-03-24 15:34:47 -07001203 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001204 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001205 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001206};
1207
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001208#ifdef NET_SKBUFF_DATA_USES_OFFSET
1209static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1210{
1211 return skb->head + skb->end;
1212}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001213
1214static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1215{
1216 return skb->end;
1217}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001218#else
1219static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1220{
1221 return skb->end;
1222}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001223
1224static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1225{
1226 return skb->end - skb->head;
1227}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001228#endif
1229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001231#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232
Patrick Ohlyac45f602009-02-12 05:03:37 +00001233static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1234{
1235 return &skb_shinfo(skb)->hwtstamps;
1236}
1237
Willem de Bruijn52267792017-08-03 16:29:39 -04001238static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1239{
1240 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1241
1242 return is_zcopy ? skb_uarg(skb) : NULL;
1243}
1244
1245static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1246{
1247 if (skb && uarg && !skb_zcopy(skb)) {
1248 sock_zerocopy_get(uarg);
1249 skb_shinfo(skb)->destructor_arg = uarg;
1250 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1251 }
1252}
1253
1254/* Release a reference on a zerocopy structure */
1255static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1256{
1257 struct ubuf_info *uarg = skb_zcopy(skb);
1258
1259 if (uarg) {
1260 uarg->zerocopy = uarg->zerocopy && zerocopy;
1261 sock_zerocopy_put(uarg);
1262 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1263 }
1264}
1265
1266/* Abort a zerocopy operation and revert zckey on error in send syscall */
1267static inline void skb_zcopy_abort(struct sk_buff *skb)
1268{
1269 struct ubuf_info *uarg = skb_zcopy(skb);
1270
1271 if (uarg) {
1272 sock_zerocopy_put_abort(uarg);
1273 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1274 }
1275}
1276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277/**
1278 * skb_queue_empty - check if a queue is empty
1279 * @list: queue head
1280 *
1281 * Returns true if the queue is empty, false otherwise.
1282 */
1283static inline int skb_queue_empty(const struct sk_buff_head *list)
1284{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001285 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286}
1287
1288/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001289 * skb_queue_is_last - check if skb is the last entry in the queue
1290 * @list: queue head
1291 * @skb: buffer
1292 *
1293 * Returns true if @skb is the last buffer on the list.
1294 */
1295static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1296 const struct sk_buff *skb)
1297{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001298 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001299}
1300
1301/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001302 * skb_queue_is_first - check if skb is the first entry in the queue
1303 * @list: queue head
1304 * @skb: buffer
1305 *
1306 * Returns true if @skb is the first buffer on the list.
1307 */
1308static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1309 const struct sk_buff *skb)
1310{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001311 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001312}
1313
1314/**
David S. Miller249c8b42008-09-23 00:44:42 -07001315 * skb_queue_next - return the next packet in the queue
1316 * @list: queue head
1317 * @skb: current buffer
1318 *
1319 * Return the next packet in @list after @skb. It is only valid to
1320 * call this if skb_queue_is_last() evaluates to false.
1321 */
1322static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1323 const struct sk_buff *skb)
1324{
1325 /* This BUG_ON may seem severe, but if we just return then we
1326 * are going to dereference garbage.
1327 */
1328 BUG_ON(skb_queue_is_last(list, skb));
1329 return skb->next;
1330}
1331
1332/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001333 * skb_queue_prev - return the prev packet in the queue
1334 * @list: queue head
1335 * @skb: current buffer
1336 *
1337 * Return the prev packet in @list before @skb. It is only valid to
1338 * call this if skb_queue_is_first() evaluates to false.
1339 */
1340static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1341 const struct sk_buff *skb)
1342{
1343 /* This BUG_ON may seem severe, but if we just return then we
1344 * are going to dereference garbage.
1345 */
1346 BUG_ON(skb_queue_is_first(list, skb));
1347 return skb->prev;
1348}
1349
1350/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 * skb_get - reference buffer
1352 * @skb: buffer to reference
1353 *
1354 * Makes another reference to a socket buffer and returns a pointer
1355 * to the buffer.
1356 */
1357static inline struct sk_buff *skb_get(struct sk_buff *skb)
1358{
Reshetova, Elena63354792017-06-30 13:07:58 +03001359 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 return skb;
1361}
1362
1363/*
1364 * If users == 1, we are the only owner and are can avoid redundant
1365 * atomic change.
1366 */
1367
1368/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 * skb_cloned - is the buffer a clone
1370 * @skb: buffer to check
1371 *
1372 * Returns true if the buffer was generated with skb_clone() and is
1373 * one of multiple shared copies of the buffer. Cloned buffers are
1374 * shared data so must not be written to under normal circumstances.
1375 */
1376static inline int skb_cloned(const struct sk_buff *skb)
1377{
1378 return skb->cloned &&
1379 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1380}
1381
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001382static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1383{
Mel Gormand0164ad2015-11-06 16:28:21 -08001384 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001385
1386 if (skb_cloned(skb))
1387 return pskb_expand_head(skb, 0, 0, pri);
1388
1389 return 0;
1390}
1391
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392/**
1393 * skb_header_cloned - is the header a clone
1394 * @skb: buffer to check
1395 *
1396 * Returns true if modifying the header part of the buffer requires
1397 * the data to be copied.
1398 */
1399static inline int skb_header_cloned(const struct sk_buff *skb)
1400{
1401 int dataref;
1402
1403 if (!skb->cloned)
1404 return 0;
1405
1406 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1407 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1408 return dataref != 1;
1409}
1410
Eric Dumazet9580bf22016-04-30 10:19:29 -07001411static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1412{
1413 might_sleep_if(gfpflags_allow_blocking(pri));
1414
1415 if (skb_header_cloned(skb))
1416 return pskb_expand_head(skb, 0, 0, pri);
1417
1418 return 0;
1419}
1420
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421/**
1422 * skb_header_release - release reference to header
1423 * @skb: buffer to operate on
1424 *
1425 * Drop a reference to the header part of the buffer. This is done
1426 * by acquiring a payload reference. You must not read from the header
1427 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001428 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 */
1430static inline void skb_header_release(struct sk_buff *skb)
1431{
1432 BUG_ON(skb->nohdr);
1433 skb->nohdr = 1;
1434 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1435}
1436
1437/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001438 * __skb_header_release - release reference to header
1439 * @skb: buffer to operate on
1440 *
1441 * Variant of skb_header_release() assuming skb is private to caller.
1442 * We can avoid one atomic operation.
1443 */
1444static inline void __skb_header_release(struct sk_buff *skb)
1445{
1446 skb->nohdr = 1;
1447 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1448}
1449
1450
1451/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 * skb_shared - is the buffer shared
1453 * @skb: buffer to check
1454 *
1455 * Returns true if more than one person has a reference to this
1456 * buffer.
1457 */
1458static inline int skb_shared(const struct sk_buff *skb)
1459{
Reshetova, Elena63354792017-06-30 13:07:58 +03001460 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461}
1462
1463/**
1464 * skb_share_check - check if buffer is shared and if so clone it
1465 * @skb: buffer to check
1466 * @pri: priority for memory allocation
1467 *
1468 * If the buffer is shared the buffer is cloned and the old copy
1469 * drops a reference. A new clone with a single reference is returned.
1470 * If the buffer is not shared the original buffer is returned. When
1471 * being called from interrupt status or with spinlocks held pri must
1472 * be GFP_ATOMIC.
1473 *
1474 * NULL is returned on a memory allocation failure.
1475 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001476static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477{
Mel Gormand0164ad2015-11-06 16:28:21 -08001478 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 if (skb_shared(skb)) {
1480 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001481
1482 if (likely(nskb))
1483 consume_skb(skb);
1484 else
1485 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 skb = nskb;
1487 }
1488 return skb;
1489}
1490
1491/*
1492 * Copy shared buffers into a new sk_buff. We effectively do COW on
1493 * packets to handle cases where we have a local reader and forward
1494 * and a couple of other messy ones. The normal one is tcpdumping
1495 * a packet thats being forwarded.
1496 */
1497
1498/**
1499 * skb_unshare - make a copy of a shared buffer
1500 * @skb: buffer to check
1501 * @pri: priority for memory allocation
1502 *
1503 * If the socket buffer is a clone then this function creates a new
1504 * copy of the data, drops a reference count on the old copy and returns
1505 * the new copy with the reference count at 1. If the buffer is not a clone
1506 * the original buffer is returned. When called with a spinlock held or
1507 * from interrupt state @pri must be %GFP_ATOMIC
1508 *
1509 * %NULL is returned on a memory allocation failure.
1510 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001511static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001512 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513{
Mel Gormand0164ad2015-11-06 16:28:21 -08001514 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 if (skb_cloned(skb)) {
1516 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001517
1518 /* Free our shared copy */
1519 if (likely(nskb))
1520 consume_skb(skb);
1521 else
1522 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 skb = nskb;
1524 }
1525 return skb;
1526}
1527
1528/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001529 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 * @list_: list to peek at
1531 *
1532 * Peek an &sk_buff. Unlike most other operations you _MUST_
1533 * be careful with this one. A peek leaves the buffer on the
1534 * list and someone else may run off with it. You must hold
1535 * the appropriate locks or have a private queue to do this.
1536 *
1537 * Returns %NULL for an empty list or a pointer to the head element.
1538 * The reference count is not incremented and the reference is therefore
1539 * volatile. Use with caution.
1540 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001541static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542{
Eric Dumazet18d07002012-04-30 16:31:46 +00001543 struct sk_buff *skb = list_->next;
1544
1545 if (skb == (struct sk_buff *)list_)
1546 skb = NULL;
1547 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548}
1549
1550/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001551 * skb_peek_next - peek skb following the given one from a queue
1552 * @skb: skb to start from
1553 * @list_: list to peek at
1554 *
1555 * Returns %NULL when the end of the list is met or a pointer to the
1556 * next element. The reference count is not incremented and the
1557 * reference is therefore volatile. Use with caution.
1558 */
1559static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1560 const struct sk_buff_head *list_)
1561{
1562 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001563
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001564 if (next == (struct sk_buff *)list_)
1565 next = NULL;
1566 return next;
1567}
1568
1569/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001570 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 * @list_: list to peek at
1572 *
1573 * Peek an &sk_buff. Unlike most other operations you _MUST_
1574 * be careful with this one. A peek leaves the buffer on the
1575 * list and someone else may run off with it. You must hold
1576 * the appropriate locks or have a private queue to do this.
1577 *
1578 * Returns %NULL for an empty list or a pointer to the tail element.
1579 * The reference count is not incremented and the reference is therefore
1580 * volatile. Use with caution.
1581 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001582static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583{
Eric Dumazet18d07002012-04-30 16:31:46 +00001584 struct sk_buff *skb = list_->prev;
1585
1586 if (skb == (struct sk_buff *)list_)
1587 skb = NULL;
1588 return skb;
1589
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590}
1591
1592/**
1593 * skb_queue_len - get queue length
1594 * @list_: list to measure
1595 *
1596 * Return the length of an &sk_buff queue.
1597 */
1598static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1599{
1600 return list_->qlen;
1601}
1602
David S. Miller67fed452008-09-21 22:36:24 -07001603/**
1604 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1605 * @list: queue to initialize
1606 *
1607 * This initializes only the list and queue length aspects of
1608 * an sk_buff_head object. This allows to initialize the list
1609 * aspects of an sk_buff_head without reinitializing things like
1610 * the spinlock. It can also be used for on-stack sk_buff_head
1611 * objects where the spinlock is known to not be used.
1612 */
1613static inline void __skb_queue_head_init(struct sk_buff_head *list)
1614{
1615 list->prev = list->next = (struct sk_buff *)list;
1616 list->qlen = 0;
1617}
1618
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001619/*
1620 * This function creates a split out lock class for each invocation;
1621 * this is needed for now since a whole lot of users of the skb-queue
1622 * infrastructure in drivers have different locking usage (in hardirq)
1623 * than the networking core (in softirq only). In the long run either the
1624 * network layer or drivers should need annotation to consolidate the
1625 * main types of usage into 3 classes.
1626 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627static inline void skb_queue_head_init(struct sk_buff_head *list)
1628{
1629 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001630 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631}
1632
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001633static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1634 struct lock_class_key *class)
1635{
1636 skb_queue_head_init(list);
1637 lockdep_set_class(&list->lock, class);
1638}
1639
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001641 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 *
1643 * The "__skb_xxxx()" functions are the non-atomic ones that
1644 * can only be called with interrupts disabled.
1645 */
Joe Perches7965bd42013-09-26 14:48:15 -07001646void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1647 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001648static inline void __skb_insert(struct sk_buff *newsk,
1649 struct sk_buff *prev, struct sk_buff *next,
1650 struct sk_buff_head *list)
1651{
1652 newsk->next = next;
1653 newsk->prev = prev;
1654 next->prev = prev->next = newsk;
1655 list->qlen++;
1656}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657
David S. Miller67fed452008-09-21 22:36:24 -07001658static inline void __skb_queue_splice(const struct sk_buff_head *list,
1659 struct sk_buff *prev,
1660 struct sk_buff *next)
1661{
1662 struct sk_buff *first = list->next;
1663 struct sk_buff *last = list->prev;
1664
1665 first->prev = prev;
1666 prev->next = first;
1667
1668 last->next = next;
1669 next->prev = last;
1670}
1671
1672/**
1673 * skb_queue_splice - join two skb lists, this is designed for stacks
1674 * @list: the new list to add
1675 * @head: the place to add it in the first list
1676 */
1677static inline void skb_queue_splice(const struct sk_buff_head *list,
1678 struct sk_buff_head *head)
1679{
1680 if (!skb_queue_empty(list)) {
1681 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001682 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001683 }
1684}
1685
1686/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001687 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001688 * @list: the new list to add
1689 * @head: the place to add it in the first list
1690 *
1691 * The list at @list is reinitialised
1692 */
1693static inline void skb_queue_splice_init(struct sk_buff_head *list,
1694 struct sk_buff_head *head)
1695{
1696 if (!skb_queue_empty(list)) {
1697 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001698 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001699 __skb_queue_head_init(list);
1700 }
1701}
1702
1703/**
1704 * skb_queue_splice_tail - join two skb lists, each list being a queue
1705 * @list: the new list to add
1706 * @head: the place to add it in the first list
1707 */
1708static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1709 struct sk_buff_head *head)
1710{
1711 if (!skb_queue_empty(list)) {
1712 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001713 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001714 }
1715}
1716
1717/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001718 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001719 * @list: the new list to add
1720 * @head: the place to add it in the first list
1721 *
1722 * Each of the lists is a queue.
1723 * The list at @list is reinitialised
1724 */
1725static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1726 struct sk_buff_head *head)
1727{
1728 if (!skb_queue_empty(list)) {
1729 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001730 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001731 __skb_queue_head_init(list);
1732 }
1733}
1734
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001736 * __skb_queue_after - queue a buffer at the list head
1737 * @list: list to use
1738 * @prev: place after this buffer
1739 * @newsk: buffer to queue
1740 *
1741 * Queue a buffer int the middle of a list. This function takes no locks
1742 * and you must therefore hold required locks before calling it.
1743 *
1744 * A buffer cannot be placed on two lists at the same time.
1745 */
1746static inline void __skb_queue_after(struct sk_buff_head *list,
1747 struct sk_buff *prev,
1748 struct sk_buff *newsk)
1749{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001750 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001751}
1752
Joe Perches7965bd42013-09-26 14:48:15 -07001753void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1754 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001755
Gerrit Renkerf5572852008-04-14 00:05:28 -07001756static inline void __skb_queue_before(struct sk_buff_head *list,
1757 struct sk_buff *next,
1758 struct sk_buff *newsk)
1759{
1760 __skb_insert(newsk, next->prev, next, list);
1761}
1762
Stephen Hemminger300ce172005-10-30 13:47:34 -08001763/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 * __skb_queue_head - queue a buffer at the list head
1765 * @list: list to use
1766 * @newsk: buffer to queue
1767 *
1768 * Queue a buffer at the start of a list. This function takes no locks
1769 * and you must therefore hold required locks before calling it.
1770 *
1771 * A buffer cannot be placed on two lists at the same time.
1772 */
Joe Perches7965bd42013-09-26 14:48:15 -07001773void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774static inline void __skb_queue_head(struct sk_buff_head *list,
1775 struct sk_buff *newsk)
1776{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001777 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778}
1779
1780/**
1781 * __skb_queue_tail - queue a buffer at the list tail
1782 * @list: list to use
1783 * @newsk: buffer to queue
1784 *
1785 * Queue a buffer at the end of a list. This function takes no locks
1786 * and you must therefore hold required locks before calling it.
1787 *
1788 * A buffer cannot be placed on two lists at the same time.
1789 */
Joe Perches7965bd42013-09-26 14:48:15 -07001790void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791static inline void __skb_queue_tail(struct sk_buff_head *list,
1792 struct sk_buff *newsk)
1793{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001794 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795}
1796
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 * remove sk_buff from list. _Must_ be called atomically, and with
1799 * the list known..
1800 */
Joe Perches7965bd42013-09-26 14:48:15 -07001801void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1803{
1804 struct sk_buff *next, *prev;
1805
1806 list->qlen--;
1807 next = skb->next;
1808 prev = skb->prev;
1809 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 next->prev = prev;
1811 prev->next = next;
1812}
1813
Gerrit Renkerf525c062008-04-14 00:04:12 -07001814/**
1815 * __skb_dequeue - remove from the head of the queue
1816 * @list: list to dequeue from
1817 *
1818 * Remove the head of the list. This function does not take any locks
1819 * so must be used with appropriate locks held only. The head item is
1820 * returned or %NULL if the list is empty.
1821 */
Joe Perches7965bd42013-09-26 14:48:15 -07001822struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001823static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1824{
1825 struct sk_buff *skb = skb_peek(list);
1826 if (skb)
1827 __skb_unlink(skb, list);
1828 return skb;
1829}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830
1831/**
1832 * __skb_dequeue_tail - remove from the tail of the queue
1833 * @list: list to dequeue from
1834 *
1835 * Remove the tail of the list. This function does not take any locks
1836 * so must be used with appropriate locks held only. The tail item is
1837 * returned or %NULL if the list is empty.
1838 */
Joe Perches7965bd42013-09-26 14:48:15 -07001839struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1841{
1842 struct sk_buff *skb = skb_peek_tail(list);
1843 if (skb)
1844 __skb_unlink(skb, list);
1845 return skb;
1846}
1847
1848
David S. Millerbdcc0922012-03-07 20:53:36 -05001849static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850{
1851 return skb->data_len;
1852}
1853
1854static inline unsigned int skb_headlen(const struct sk_buff *skb)
1855{
1856 return skb->len - skb->data_len;
1857}
1858
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001859static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001861 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001863 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001864 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001865 return len;
1866}
1867
1868static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1869{
1870 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871}
1872
Ian Campbell131ea662011-08-19 06:25:00 +00001873/**
1874 * __skb_fill_page_desc - initialise a paged fragment in an skb
1875 * @skb: buffer containing fragment to be initialised
1876 * @i: paged fragment index to initialise
1877 * @page: the page to use for this fragment
1878 * @off: the offset to the data with @page
1879 * @size: the length of the data
1880 *
1881 * Initialises the @i'th fragment of @skb to point to &size bytes at
1882 * offset @off within @page.
1883 *
1884 * Does not take any additional reference on the fragment.
1885 */
1886static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1887 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888{
1889 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1890
Mel Gormanc48a11c2012-07-31 16:44:23 -07001891 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001892 * Propagate page pfmemalloc to the skb if we can. The problem is
1893 * that not all callers have unique ownership of the page but rely
1894 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001895 */
Ian Campbella8605c62011-10-19 23:01:49 +00001896 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001898 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001899
1900 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001901 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001902 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001903}
1904
1905/**
1906 * skb_fill_page_desc - initialise a paged fragment in an skb
1907 * @skb: buffer containing fragment to be initialised
1908 * @i: paged fragment index to initialise
1909 * @page: the page to use for this fragment
1910 * @off: the offset to the data with @page
1911 * @size: the length of the data
1912 *
1913 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001914 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001915 * addition updates @skb such that @i is the last fragment.
1916 *
1917 * Does not take any additional reference on the fragment.
1918 */
1919static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1920 struct page *page, int off, int size)
1921{
1922 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 skb_shinfo(skb)->nr_frags = i + 1;
1924}
1925
Joe Perches7965bd42013-09-26 14:48:15 -07001926void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1927 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001928
Jason Wangf8e617e2013-11-01 14:07:47 +08001929void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1930 unsigned int truesize);
1931
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001933#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1935
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001936#ifdef NET_SKBUFF_DATA_USES_OFFSET
1937static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1938{
1939 return skb->head + skb->tail;
1940}
1941
1942static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1943{
1944 skb->tail = skb->data - skb->head;
1945}
1946
1947static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1948{
1949 skb_reset_tail_pointer(skb);
1950 skb->tail += offset;
1951}
Simon Horman7cc46192013-05-28 20:34:29 +00001952
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001953#else /* NET_SKBUFF_DATA_USES_OFFSET */
1954static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1955{
1956 return skb->tail;
1957}
1958
1959static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1960{
1961 skb->tail = skb->data;
1962}
1963
1964static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1965{
1966 skb->tail = skb->data + offset;
1967}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001968
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001969#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1970
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971/*
1972 * Add data to an sk_buff
1973 */
Johannes Berg4df864c2017-06-16 14:29:21 +02001974void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1975void *skb_put(struct sk_buff *skb, unsigned int len);
1976static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977{
Johannes Berg4df864c2017-06-16 14:29:21 +02001978 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 SKB_LINEAR_ASSERT(skb);
1980 skb->tail += len;
1981 skb->len += len;
1982 return tmp;
1983}
1984
yuan linyude77b962017-06-18 22:48:17 +08001985static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
1986{
1987 void *tmp = __skb_put(skb, len);
1988
1989 memset(tmp, 0, len);
1990 return tmp;
1991}
1992
1993static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
1994 unsigned int len)
1995{
1996 void *tmp = __skb_put(skb, len);
1997
1998 memcpy(tmp, data, len);
1999 return tmp;
2000}
2001
2002static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2003{
2004 *(u8 *)__skb_put(skb, 1) = val;
2005}
2006
Johannes Berg83ad3572017-06-14 22:17:20 +02002007static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002008{
Johannes Berg83ad3572017-06-14 22:17:20 +02002009 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002010
2011 memset(tmp, 0, len);
2012
2013 return tmp;
2014}
2015
Johannes Berg59ae1d12017-06-16 14:29:20 +02002016static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2017 unsigned int len)
2018{
2019 void *tmp = skb_put(skb, len);
2020
2021 memcpy(tmp, data, len);
2022
2023 return tmp;
2024}
2025
Johannes Berg634fef62017-06-16 14:29:24 +02002026static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2027{
2028 *(u8 *)skb_put(skb, 1) = val;
2029}
2030
Johannes Bergd58ff352017-06-16 14:29:23 +02002031void *skb_push(struct sk_buff *skb, unsigned int len);
2032static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033{
2034 skb->data -= len;
2035 skb->len += len;
2036 return skb->data;
2037}
2038
Johannes Bergaf728682017-06-16 14:29:22 +02002039void *skb_pull(struct sk_buff *skb, unsigned int len);
2040static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041{
2042 skb->len -= len;
2043 BUG_ON(skb->len < skb->data_len);
2044 return skb->data += len;
2045}
2046
Johannes Bergaf728682017-06-16 14:29:22 +02002047static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002048{
2049 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2050}
2051
Johannes Bergaf728682017-06-16 14:29:22 +02002052void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053
Johannes Bergaf728682017-06-16 14:29:22 +02002054static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055{
2056 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002057 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 return NULL;
2059 skb->len -= len;
2060 return skb->data += len;
2061}
2062
Johannes Bergaf728682017-06-16 14:29:22 +02002063static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064{
2065 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2066}
2067
2068static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2069{
2070 if (likely(len <= skb_headlen(skb)))
2071 return 1;
2072 if (unlikely(len > skb->len))
2073 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002074 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075}
2076
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002077void skb_condense(struct sk_buff *skb);
2078
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079/**
2080 * skb_headroom - bytes at buffer head
2081 * @skb: buffer to check
2082 *
2083 * Return the number of bytes of free space at the head of an &sk_buff.
2084 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002085static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086{
2087 return skb->data - skb->head;
2088}
2089
2090/**
2091 * skb_tailroom - bytes at buffer end
2092 * @skb: buffer to check
2093 *
2094 * Return the number of bytes of free space at the tail of an sk_buff
2095 */
2096static inline int skb_tailroom(const struct sk_buff *skb)
2097{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002098 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099}
2100
2101/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002102 * skb_availroom - bytes at buffer end
2103 * @skb: buffer to check
2104 *
2105 * Return the number of bytes of free space at the tail of an sk_buff
2106 * allocated by sk_stream_alloc()
2107 */
2108static inline int skb_availroom(const struct sk_buff *skb)
2109{
Eric Dumazet16fad692013-03-14 05:40:32 +00002110 if (skb_is_nonlinear(skb))
2111 return 0;
2112
2113 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002114}
2115
2116/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 * skb_reserve - adjust headroom
2118 * @skb: buffer to alter
2119 * @len: bytes to move
2120 *
2121 * Increase the headroom of an empty &sk_buff by reducing the tail
2122 * room. This is only allowed for an empty buffer.
2123 */
David S. Miller8243126c2006-01-17 02:54:21 -08002124static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125{
2126 skb->data += len;
2127 skb->tail += len;
2128}
2129
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002130/**
2131 * skb_tailroom_reserve - adjust reserved_tailroom
2132 * @skb: buffer to alter
2133 * @mtu: maximum amount of headlen permitted
2134 * @needed_tailroom: minimum amount of reserved_tailroom
2135 *
2136 * Set reserved_tailroom so that headlen can be as large as possible but
2137 * not larger than mtu and tailroom cannot be smaller than
2138 * needed_tailroom.
2139 * The required headroom should already have been reserved before using
2140 * this function.
2141 */
2142static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2143 unsigned int needed_tailroom)
2144{
2145 SKB_LINEAR_ASSERT(skb);
2146 if (mtu < skb_tailroom(skb) - needed_tailroom)
2147 /* use at most mtu */
2148 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2149 else
2150 /* use up to all available space */
2151 skb->reserved_tailroom = needed_tailroom;
2152}
2153
Tom Herbert8bce6d72014-09-29 20:22:29 -07002154#define ENCAP_TYPE_ETHER 0
2155#define ENCAP_TYPE_IPPROTO 1
2156
2157static inline void skb_set_inner_protocol(struct sk_buff *skb,
2158 __be16 protocol)
2159{
2160 skb->inner_protocol = protocol;
2161 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2162}
2163
2164static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2165 __u8 ipproto)
2166{
2167 skb->inner_ipproto = ipproto;
2168 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2169}
2170
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002171static inline void skb_reset_inner_headers(struct sk_buff *skb)
2172{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002173 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002174 skb->inner_network_header = skb->network_header;
2175 skb->inner_transport_header = skb->transport_header;
2176}
2177
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002178static inline void skb_reset_mac_len(struct sk_buff *skb)
2179{
2180 skb->mac_len = skb->network_header - skb->mac_header;
2181}
2182
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002183static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2184 *skb)
2185{
2186 return skb->head + skb->inner_transport_header;
2187}
2188
Tom Herbert55dc5a92015-12-14 11:19:40 -08002189static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2190{
2191 return skb_inner_transport_header(skb) - skb->data;
2192}
2193
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002194static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2195{
2196 skb->inner_transport_header = skb->data - skb->head;
2197}
2198
2199static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2200 const int offset)
2201{
2202 skb_reset_inner_transport_header(skb);
2203 skb->inner_transport_header += offset;
2204}
2205
2206static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2207{
2208 return skb->head + skb->inner_network_header;
2209}
2210
2211static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2212{
2213 skb->inner_network_header = skb->data - skb->head;
2214}
2215
2216static inline void skb_set_inner_network_header(struct sk_buff *skb,
2217 const int offset)
2218{
2219 skb_reset_inner_network_header(skb);
2220 skb->inner_network_header += offset;
2221}
2222
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002223static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2224{
2225 return skb->head + skb->inner_mac_header;
2226}
2227
2228static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2229{
2230 skb->inner_mac_header = skb->data - skb->head;
2231}
2232
2233static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2234 const int offset)
2235{
2236 skb_reset_inner_mac_header(skb);
2237 skb->inner_mac_header += offset;
2238}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002239static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2240{
Cong Wang35d04612013-05-29 15:16:05 +08002241 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002242}
2243
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002244static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2245{
2246 return skb->head + skb->transport_header;
2247}
2248
2249static inline void skb_reset_transport_header(struct sk_buff *skb)
2250{
2251 skb->transport_header = skb->data - skb->head;
2252}
2253
2254static inline void skb_set_transport_header(struct sk_buff *skb,
2255 const int offset)
2256{
2257 skb_reset_transport_header(skb);
2258 skb->transport_header += offset;
2259}
2260
2261static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2262{
2263 return skb->head + skb->network_header;
2264}
2265
2266static inline void skb_reset_network_header(struct sk_buff *skb)
2267{
2268 skb->network_header = skb->data - skb->head;
2269}
2270
2271static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2272{
2273 skb_reset_network_header(skb);
2274 skb->network_header += offset;
2275}
2276
2277static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2278{
2279 return skb->head + skb->mac_header;
2280}
2281
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002282static inline int skb_mac_offset(const struct sk_buff *skb)
2283{
2284 return skb_mac_header(skb) - skb->data;
2285}
2286
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002287static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2288{
2289 return skb->network_header - skb->mac_header;
2290}
2291
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002292static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2293{
Cong Wang35d04612013-05-29 15:16:05 +08002294 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002295}
2296
2297static inline void skb_reset_mac_header(struct sk_buff *skb)
2298{
2299 skb->mac_header = skb->data - skb->head;
2300}
2301
2302static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2303{
2304 skb_reset_mac_header(skb);
2305 skb->mac_header += offset;
2306}
2307
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002308static inline void skb_pop_mac_header(struct sk_buff *skb)
2309{
2310 skb->mac_header = skb->network_header;
2311}
2312
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002313static inline void skb_probe_transport_header(struct sk_buff *skb,
2314 const int offset_hint)
2315{
2316 struct flow_keys keys;
2317
2318 if (skb_transport_header_was_set(skb))
2319 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002320 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002321 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002322 else
2323 skb_set_transport_header(skb, offset_hint);
2324}
2325
Eric Dumazet03606892012-02-23 10:55:02 +00002326static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2327{
2328 if (skb_mac_header_was_set(skb)) {
2329 const unsigned char *old_mac = skb_mac_header(skb);
2330
2331 skb_set_mac_header(skb, -skb->mac_len);
2332 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2333 }
2334}
2335
Michał Mirosław04fb4512010-12-14 15:24:08 +00002336static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2337{
2338 return skb->csum_start - skb_headroom(skb);
2339}
2340
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002341static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2342{
2343 return skb->head + skb->csum_start;
2344}
2345
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002346static inline int skb_transport_offset(const struct sk_buff *skb)
2347{
2348 return skb_transport_header(skb) - skb->data;
2349}
2350
2351static inline u32 skb_network_header_len(const struct sk_buff *skb)
2352{
2353 return skb->transport_header - skb->network_header;
2354}
2355
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002356static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2357{
2358 return skb->inner_transport_header - skb->inner_network_header;
2359}
2360
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002361static inline int skb_network_offset(const struct sk_buff *skb)
2362{
2363 return skb_network_header(skb) - skb->data;
2364}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002365
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002366static inline int skb_inner_network_offset(const struct sk_buff *skb)
2367{
2368 return skb_inner_network_header(skb) - skb->data;
2369}
2370
Changli Gaof9599ce2010-08-04 04:43:44 +00002371static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2372{
2373 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2374}
2375
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376/*
2377 * CPUs often take a performance hit when accessing unaligned memory
2378 * locations. The actual performance hit varies, it can be small if the
2379 * hardware handles it or large if we have to take an exception and fix it
2380 * in software.
2381 *
2382 * Since an ethernet header is 14 bytes network drivers often end up with
2383 * the IP header at an unaligned offset. The IP header can be aligned by
2384 * shifting the start of the packet by 2 bytes. Drivers should do this
2385 * with:
2386 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002387 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388 *
2389 * The downside to this alignment of the IP header is that the DMA is now
2390 * unaligned. On some architectures the cost of an unaligned DMA is high
2391 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002392 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2394 * to be overridden.
2395 */
2396#ifndef NET_IP_ALIGN
2397#define NET_IP_ALIGN 2
2398#endif
2399
Anton Blanchard025be812006-03-31 02:27:06 -08002400/*
2401 * The networking layer reserves some headroom in skb data (via
2402 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2403 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002404 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002405 *
2406 * Unfortunately this headroom changes the DMA alignment of the resulting
2407 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2408 * on some architectures. An architecture can override this value,
2409 * perhaps setting it to a cacheline in size (since that will maintain
2410 * cacheline alignment of the DMA). It must be a power of 2.
2411 *
David S. Millerd6301d32009-02-08 19:24:13 -08002412 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002413 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002414 *
2415 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2416 * to reduce average number of cache lines per packet.
2417 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002418 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002419 */
2420#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002421#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002422#endif
2423
Joe Perches7965bd42013-09-26 14:48:15 -07002424int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002426static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002428 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002429 WARN_ON(1);
2430 return;
2431 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002432 skb->len = len;
2433 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434}
2435
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002436static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2437{
2438 __skb_set_length(skb, len);
2439}
2440
Joe Perches7965bd42013-09-26 14:48:15 -07002441void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442
2443static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2444{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002445 if (skb->data_len)
2446 return ___pskb_trim(skb, len);
2447 __skb_trim(skb, len);
2448 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449}
2450
2451static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2452{
2453 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2454}
2455
2456/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002457 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2458 * @skb: buffer to alter
2459 * @len: new length
2460 *
2461 * This is identical to pskb_trim except that the caller knows that
2462 * the skb is not cloned so we should never get an error due to out-
2463 * of-memory.
2464 */
2465static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2466{
2467 int err = pskb_trim(skb, len);
2468 BUG_ON(err);
2469}
2470
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002471static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2472{
2473 unsigned int diff = len - skb->len;
2474
2475 if (skb_tailroom(skb) < diff) {
2476 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2477 GFP_ATOMIC);
2478 if (ret)
2479 return ret;
2480 }
2481 __skb_set_length(skb, len);
2482 return 0;
2483}
2484
Herbert Xue9fa4f72006-08-13 20:12:58 -07002485/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486 * skb_orphan - orphan a buffer
2487 * @skb: buffer to orphan
2488 *
2489 * If a buffer currently has an owner then we call the owner's
2490 * destructor function and make the @skb unowned. The buffer continues
2491 * to exist but is no longer charged to its former owner.
2492 */
2493static inline void skb_orphan(struct sk_buff *skb)
2494{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002495 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002497 skb->destructor = NULL;
2498 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002499 } else {
2500 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002501 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002502}
2503
2504/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002505 * skb_orphan_frags - orphan the frags contained in a buffer
2506 * @skb: buffer to orphan frags from
2507 * @gfp_mask: allocation mask for replacement pages
2508 *
2509 * For each frag in the SKB which needs a destructor (i.e. has an
2510 * owner) create a copy of that frag and release the original
2511 * page by calling the destructor.
2512 */
2513static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2514{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002515 if (likely(!skb_zcopy(skb)))
2516 return 0;
2517 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2518 return 0;
2519 return skb_copy_ubufs(skb, gfp_mask);
2520}
2521
2522/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2523static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2524{
2525 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002526 return 0;
2527 return skb_copy_ubufs(skb, gfp_mask);
2528}
2529
2530/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002531 * __skb_queue_purge - empty a list
2532 * @list: list to empty
2533 *
2534 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2535 * the list and one reference dropped. This function does not take the
2536 * list lock and the caller must hold the relevant locks to use it.
2537 */
Joe Perches7965bd42013-09-26 14:48:15 -07002538void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002539static inline void __skb_queue_purge(struct sk_buff_head *list)
2540{
2541 struct sk_buff *skb;
2542 while ((skb = __skb_dequeue(list)) != NULL)
2543 kfree_skb(skb);
2544}
2545
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002546void skb_rbtree_purge(struct rb_root *root);
2547
Joe Perches7965bd42013-09-26 14:48:15 -07002548void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549
Joe Perches7965bd42013-09-26 14:48:15 -07002550struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2551 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002552
2553/**
2554 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2555 * @dev: network device to receive on
2556 * @length: length to allocate
2557 *
2558 * Allocate a new &sk_buff and assign it a usage count of one. The
2559 * buffer has unspecified headroom built in. Users should allocate
2560 * the headroom they think they need without accounting for the
2561 * built in space. The built in space is used for optimisations.
2562 *
2563 * %NULL is returned if there is no free memory. Although this function
2564 * allocates memory it can be called from an interrupt.
2565 */
2566static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002567 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002568{
2569 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2570}
2571
Eric Dumazet6f532612012-05-18 05:12:12 +00002572/* legacy helper around __netdev_alloc_skb() */
2573static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2574 gfp_t gfp_mask)
2575{
2576 return __netdev_alloc_skb(NULL, length, gfp_mask);
2577}
2578
2579/* legacy helper around netdev_alloc_skb() */
2580static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2581{
2582 return netdev_alloc_skb(NULL, length);
2583}
2584
2585
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002586static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2587 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002588{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002589 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002590
2591 if (NET_IP_ALIGN && skb)
2592 skb_reserve(skb, NET_IP_ALIGN);
2593 return skb;
2594}
2595
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002596static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2597 unsigned int length)
2598{
2599 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2600}
2601
Alexander Duyck181edb22015-05-06 21:12:03 -07002602static inline void skb_free_frag(void *addr)
2603{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002604 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002605}
2606
Alexander Duyckffde7322014-12-09 19:40:42 -08002607void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002608struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2609 unsigned int length, gfp_t gfp_mask);
2610static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2611 unsigned int length)
2612{
2613 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2614}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002615void napi_consume_skb(struct sk_buff *skb, int budget);
2616
2617void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002618void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002619
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002620/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002621 * __dev_alloc_pages - allocate page for network Rx
2622 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2623 * @order: size of the allocation
2624 *
2625 * Allocate a new page.
2626 *
2627 * %NULL is returned if there is no free memory.
2628*/
2629static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2630 unsigned int order)
2631{
2632 /* This piece of code contains several assumptions.
2633 * 1. This is for device Rx, therefor a cold page is preferred.
2634 * 2. The expectation is the user wants a compound page.
2635 * 3. If requesting a order 0 page it will not be compound
2636 * due to the check to see if order has a value in prep_new_page
2637 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2638 * code in gfp_to_alloc_flags that should be enforcing this.
2639 */
2640 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2641
2642 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2643}
2644
2645static inline struct page *dev_alloc_pages(unsigned int order)
2646{
Neil Horman95829b32016-05-19 11:30:54 -04002647 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002648}
2649
2650/**
2651 * __dev_alloc_page - allocate a page for network Rx
2652 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2653 *
2654 * Allocate a new page.
2655 *
2656 * %NULL is returned if there is no free memory.
2657 */
2658static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2659{
2660 return __dev_alloc_pages(gfp_mask, 0);
2661}
2662
2663static inline struct page *dev_alloc_page(void)
2664{
Neil Horman95829b32016-05-19 11:30:54 -04002665 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002666}
2667
2668/**
Mel Gorman06140022012-07-31 16:44:24 -07002669 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2670 * @page: The page that was allocated from skb_alloc_page
2671 * @skb: The skb that may need pfmemalloc set
2672 */
2673static inline void skb_propagate_pfmemalloc(struct page *page,
2674 struct sk_buff *skb)
2675{
Michal Hocko2f064f32015-08-21 14:11:51 -07002676 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002677 skb->pfmemalloc = true;
2678}
2679
Eric Dumazet564824b2010-10-11 19:05:25 +00002680/**
Masanari Iidae2278672014-02-18 22:54:36 +09002681 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002682 * @frag: the paged fragment
2683 *
2684 * Returns the &struct page associated with @frag.
2685 */
2686static inline struct page *skb_frag_page(const skb_frag_t *frag)
2687{
Ian Campbella8605c62011-10-19 23:01:49 +00002688 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002689}
2690
2691/**
2692 * __skb_frag_ref - take an addition reference on a paged fragment.
2693 * @frag: the paged fragment
2694 *
2695 * Takes an additional reference on the paged fragment @frag.
2696 */
2697static inline void __skb_frag_ref(skb_frag_t *frag)
2698{
2699 get_page(skb_frag_page(frag));
2700}
2701
2702/**
2703 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2704 * @skb: the buffer
2705 * @f: the fragment offset.
2706 *
2707 * Takes an additional reference on the @f'th paged fragment of @skb.
2708 */
2709static inline void skb_frag_ref(struct sk_buff *skb, int f)
2710{
2711 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2712}
2713
2714/**
2715 * __skb_frag_unref - release a reference on a paged fragment.
2716 * @frag: the paged fragment
2717 *
2718 * Releases a reference on the paged fragment @frag.
2719 */
2720static inline void __skb_frag_unref(skb_frag_t *frag)
2721{
2722 put_page(skb_frag_page(frag));
2723}
2724
2725/**
2726 * skb_frag_unref - release a reference on a paged fragment of an skb.
2727 * @skb: the buffer
2728 * @f: the fragment offset
2729 *
2730 * Releases a reference on the @f'th paged fragment of @skb.
2731 */
2732static inline void skb_frag_unref(struct sk_buff *skb, int f)
2733{
2734 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2735}
2736
2737/**
2738 * skb_frag_address - gets the address of the data contained in a paged fragment
2739 * @frag: the paged fragment buffer
2740 *
2741 * Returns the address of the data within @frag. The page must already
2742 * be mapped.
2743 */
2744static inline void *skb_frag_address(const skb_frag_t *frag)
2745{
2746 return page_address(skb_frag_page(frag)) + frag->page_offset;
2747}
2748
2749/**
2750 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2751 * @frag: the paged fragment buffer
2752 *
2753 * Returns the address of the data within @frag. Checks that the page
2754 * is mapped and returns %NULL otherwise.
2755 */
2756static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2757{
2758 void *ptr = page_address(skb_frag_page(frag));
2759 if (unlikely(!ptr))
2760 return NULL;
2761
2762 return ptr + frag->page_offset;
2763}
2764
2765/**
2766 * __skb_frag_set_page - sets the page contained in a paged fragment
2767 * @frag: the paged fragment
2768 * @page: the page to set
2769 *
2770 * Sets the fragment @frag to contain @page.
2771 */
2772static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2773{
Ian Campbella8605c62011-10-19 23:01:49 +00002774 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002775}
2776
2777/**
2778 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2779 * @skb: the buffer
2780 * @f: the fragment offset
2781 * @page: the page to set
2782 *
2783 * Sets the @f'th fragment of @skb to contain @page.
2784 */
2785static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2786 struct page *page)
2787{
2788 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2789}
2790
Eric Dumazet400dfd32013-10-17 16:27:07 -07002791bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2792
Ian Campbell131ea662011-08-19 06:25:00 +00002793/**
2794 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002795 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002796 * @frag: the paged fragment to map
2797 * @offset: the offset within the fragment (starting at the
2798 * fragment's own offset)
2799 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002800 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002801 *
2802 * Maps the page associated with @frag to @device.
2803 */
2804static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2805 const skb_frag_t *frag,
2806 size_t offset, size_t size,
2807 enum dma_data_direction dir)
2808{
2809 return dma_map_page(dev, skb_frag_page(frag),
2810 frag->page_offset + offset, size, dir);
2811}
2812
Eric Dumazet117632e2011-12-03 21:39:53 +00002813static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2814 gfp_t gfp_mask)
2815{
2816 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2817}
2818
Octavian Purdilabad93e92014-06-12 01:36:26 +03002819
2820static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2821 gfp_t gfp_mask)
2822{
2823 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2824}
2825
2826
Ian Campbell131ea662011-08-19 06:25:00 +00002827/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002828 * skb_clone_writable - is the header of a clone writable
2829 * @skb: buffer to check
2830 * @len: length up to which to write
2831 *
2832 * Returns true if modifying the header part of the cloned buffer
2833 * does not requires the data to be copied.
2834 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002835static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002836{
2837 return !skb_header_cloned(skb) &&
2838 skb_headroom(skb) + len <= skb->hdr_len;
2839}
2840
Daniel Borkmann36976492016-02-19 23:05:25 +01002841static inline int skb_try_make_writable(struct sk_buff *skb,
2842 unsigned int write_len)
2843{
2844 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2845 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2846}
2847
Herbert Xud9cc2042007-09-16 16:21:16 -07002848static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2849 int cloned)
2850{
2851 int delta = 0;
2852
Herbert Xud9cc2042007-09-16 16:21:16 -07002853 if (headroom > skb_headroom(skb))
2854 delta = headroom - skb_headroom(skb);
2855
2856 if (delta || cloned)
2857 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2858 GFP_ATOMIC);
2859 return 0;
2860}
2861
Patrick McHardy334a8132007-06-25 04:35:20 -07002862/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863 * skb_cow - copy header of skb when it is required
2864 * @skb: buffer to cow
2865 * @headroom: needed headroom
2866 *
2867 * If the skb passed lacks sufficient headroom or its data part
2868 * is shared, data is reallocated. If reallocation fails, an error
2869 * is returned and original skb is not changed.
2870 *
2871 * The result is skb with writable area skb->head...skb->tail
2872 * and at least @headroom of space at head.
2873 */
2874static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2875{
Herbert Xud9cc2042007-09-16 16:21:16 -07002876 return __skb_cow(skb, headroom, skb_cloned(skb));
2877}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878
Herbert Xud9cc2042007-09-16 16:21:16 -07002879/**
2880 * skb_cow_head - skb_cow but only making the head writable
2881 * @skb: buffer to cow
2882 * @headroom: needed headroom
2883 *
2884 * This function is identical to skb_cow except that we replace the
2885 * skb_cloned check by skb_header_cloned. It should be used when
2886 * you only need to push on some header and do not need to modify
2887 * the data.
2888 */
2889static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2890{
2891 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002892}
2893
2894/**
2895 * skb_padto - pad an skbuff up to a minimal size
2896 * @skb: buffer to pad
2897 * @len: minimal length
2898 *
2899 * Pads up a buffer to ensure the trailing bytes exist and are
2900 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002901 * is untouched. Otherwise it is extended. Returns zero on
2902 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002903 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002904static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905{
2906 unsigned int size = skb->len;
2907 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002908 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002909 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002910}
2911
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002912/**
2913 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2914 * @skb: buffer to pad
2915 * @len: minimal length
2916 *
2917 * Pads up a buffer to ensure the trailing bytes exist and are
2918 * blanked. If the buffer already contains sufficient data it
2919 * is untouched. Otherwise it is extended. Returns zero on
2920 * success. The skb is freed on error.
2921 */
2922static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2923{
2924 unsigned int size = skb->len;
2925
2926 if (unlikely(size < len)) {
2927 len -= size;
2928 if (skb_pad(skb, len))
2929 return -ENOMEM;
2930 __skb_put(skb, len);
2931 }
2932 return 0;
2933}
2934
Linus Torvalds1da177e2005-04-16 15:20:36 -07002935static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002936 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002937{
2938 const int off = skb->len;
2939
2940 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002941 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002942 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2943 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002944 skb->csum = csum_block_add(skb->csum, csum, off);
2945 return 0;
2946 }
Al Viro15e6cb42016-11-01 22:42:45 -04002947 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002948 return 0;
2949
2950 __skb_trim(skb, off);
2951 return -EFAULT;
2952}
2953
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002954static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2955 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002956{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002957 if (skb_zcopy(skb))
2958 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002959 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002960 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002961
Ian Campbellea2ab692011-08-22 23:44:58 +00002962 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002963 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002964 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002965 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002966}
2967
Herbert Xu364c6ba2006-06-09 16:10:40 -07002968static inline int __skb_linearize(struct sk_buff *skb)
2969{
2970 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2971}
2972
Linus Torvalds1da177e2005-04-16 15:20:36 -07002973/**
2974 * skb_linearize - convert paged skb to linear one
2975 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002976 *
2977 * If there is no free memory -ENOMEM is returned, otherwise zero
2978 * is returned and the old skb data released.
2979 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002980static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002981{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002982 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2983}
2984
2985/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002986 * skb_has_shared_frag - can any frag be overwritten
2987 * @skb: buffer to test
2988 *
2989 * Return true if the skb has at least one frag that might be modified
2990 * by an external entity (as in vmsplice()/sendfile())
2991 */
2992static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2993{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002994 return skb_is_nonlinear(skb) &&
2995 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002996}
2997
2998/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002999 * skb_linearize_cow - make sure skb is linear and writable
3000 * @skb: buffer to process
3001 *
3002 * If there is no free memory -ENOMEM is returned, otherwise zero
3003 * is returned and the old skb data released.
3004 */
3005static inline int skb_linearize_cow(struct sk_buff *skb)
3006{
3007 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3008 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003009}
3010
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003011static __always_inline void
3012__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3013 unsigned int off)
3014{
3015 if (skb->ip_summed == CHECKSUM_COMPLETE)
3016 skb->csum = csum_block_sub(skb->csum,
3017 csum_partial(start, len, 0), off);
3018 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3019 skb_checksum_start_offset(skb) < 0)
3020 skb->ip_summed = CHECKSUM_NONE;
3021}
3022
Linus Torvalds1da177e2005-04-16 15:20:36 -07003023/**
3024 * skb_postpull_rcsum - update checksum for received skb after pull
3025 * @skb: buffer to update
3026 * @start: start of data before pull
3027 * @len: length of data pulled
3028 *
3029 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003030 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3031 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003032 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003034 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003035{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003036 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003037}
3038
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003039static __always_inline void
3040__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3041 unsigned int off)
3042{
3043 if (skb->ip_summed == CHECKSUM_COMPLETE)
3044 skb->csum = csum_block_add(skb->csum,
3045 csum_partial(start, len, 0), off);
3046}
Herbert Xucbb042f2006-03-20 22:43:56 -08003047
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003048/**
3049 * skb_postpush_rcsum - update checksum for received skb after push
3050 * @skb: buffer to update
3051 * @start: start of data after push
3052 * @len: length of data pushed
3053 *
3054 * After doing a push on a received packet, you need to call this to
3055 * update the CHECKSUM_COMPLETE checksum.
3056 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003057static inline void skb_postpush_rcsum(struct sk_buff *skb,
3058 const void *start, unsigned int len)
3059{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003060 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003061}
3062
Johannes Bergaf728682017-06-16 14:29:22 +02003063void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003064
David S. Miller7ce5a272013-12-02 17:26:05 -05003065/**
WANG Cong82a31b92016-06-30 10:15:22 -07003066 * skb_push_rcsum - push skb and update receive checksum
3067 * @skb: buffer to update
3068 * @len: length of data pulled
3069 *
3070 * This function performs an skb_push on the packet and updates
3071 * the CHECKSUM_COMPLETE checksum. It should be used on
3072 * receive path processing instead of skb_push unless you know
3073 * that the checksum difference is zero (e.g., a valid IP header)
3074 * or you are setting ip_summed to CHECKSUM_NONE.
3075 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003076static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003077{
3078 skb_push(skb, len);
3079 skb_postpush_rcsum(skb, skb->data, len);
3080 return skb->data;
3081}
3082
3083/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003084 * pskb_trim_rcsum - trim received skb and update checksum
3085 * @skb: buffer to trim
3086 * @len: new length
3087 *
3088 * This is exactly the same as pskb_trim except that it ensures the
3089 * checksum of received packets are still valid after the operation.
3090 */
3091
3092static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3093{
3094 if (likely(len >= skb->len))
3095 return 0;
3096 if (skb->ip_summed == CHECKSUM_COMPLETE)
3097 skb->ip_summed = CHECKSUM_NONE;
3098 return __pskb_trim(skb, len);
3099}
3100
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003101static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3102{
3103 if (skb->ip_summed == CHECKSUM_COMPLETE)
3104 skb->ip_summed = CHECKSUM_NONE;
3105 __skb_trim(skb, len);
3106 return 0;
3107}
3108
3109static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3110{
3111 if (skb->ip_summed == CHECKSUM_COMPLETE)
3112 skb->ip_summed = CHECKSUM_NONE;
3113 return __skb_grow(skb, len);
3114}
3115
Linus Torvalds1da177e2005-04-16 15:20:36 -07003116#define skb_queue_walk(queue, skb) \
3117 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003118 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003119 skb = skb->next)
3120
James Chapman46f89142007-04-30 00:07:31 -07003121#define skb_queue_walk_safe(queue, skb, tmp) \
3122 for (skb = (queue)->next, tmp = skb->next; \
3123 skb != (struct sk_buff *)(queue); \
3124 skb = tmp, tmp = skb->next)
3125
David S. Miller1164f522008-09-23 00:49:44 -07003126#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003127 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003128 skb = skb->next)
3129
3130#define skb_queue_walk_from_safe(queue, skb, tmp) \
3131 for (tmp = skb->next; \
3132 skb != (struct sk_buff *)(queue); \
3133 skb = tmp, tmp = skb->next)
3134
Stephen Hemminger300ce172005-10-30 13:47:34 -08003135#define skb_queue_reverse_walk(queue, skb) \
3136 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003137 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003138 skb = skb->prev)
3139
David S. Miller686a2952011-01-20 22:47:32 -08003140#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3141 for (skb = (queue)->prev, tmp = skb->prev; \
3142 skb != (struct sk_buff *)(queue); \
3143 skb = tmp, tmp = skb->prev)
3144
3145#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3146 for (tmp = skb->prev; \
3147 skb != (struct sk_buff *)(queue); \
3148 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003149
David S. Miller21dc3302010-08-23 00:13:46 -07003150static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003151{
3152 return skb_shinfo(skb)->frag_list != NULL;
3153}
3154
3155static inline void skb_frag_list_init(struct sk_buff *skb)
3156{
3157 skb_shinfo(skb)->frag_list = NULL;
3158}
3159
David S. Milleree039872009-06-09 00:17:13 -07003160#define skb_walk_frags(skb, iter) \
3161 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3162
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003163
3164int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3165 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003166struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3167 struct sk_buff_head *queue,
3168 unsigned int flags,
3169 void (*destructor)(struct sock *sk,
3170 struct sk_buff *skb),
3171 int *peeked, int *off, int *err,
3172 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003173struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003174 void (*destructor)(struct sock *sk,
3175 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003176 int *peeked, int *off, int *err,
3177 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003178struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003179 void (*destructor)(struct sock *sk,
3180 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003181 int *peeked, int *off, int *err);
3182struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3183 int *err);
3184unsigned int datagram_poll(struct file *file, struct socket *sock,
3185 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003186int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3187 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003188static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3189 struct msghdr *msg, int size)
3190{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003191 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003192}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003193int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3194 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003195int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3196 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003197int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003198void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003199void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3200static inline void skb_free_datagram_locked(struct sock *sk,
3201 struct sk_buff *skb)
3202{
3203 __skb_free_datagram_locked(sk, skb, 0);
3204}
Joe Perches7965bd42013-09-26 14:48:15 -07003205int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003206int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3207int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3208__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3209 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003210int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003211 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003212 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003213int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3214 int len);
3215int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003216void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003217unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003218int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3219 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003220void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3221int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3222void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003223unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003224bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003225struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003226struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003227int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003228int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003229int skb_vlan_pop(struct sk_buff *skb);
3230int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003231struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3232 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003233
Al Viro6ce8e9c2014-04-06 21:25:44 -04003234static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3235{
Al Viro3073f072017-02-17 23:13:25 -05003236 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003237}
3238
Al Viro7eab8d92014-04-06 21:51:23 -04003239static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3240{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003241 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003242}
3243
Daniel Borkmann2817a332013-10-30 11:50:51 +01003244struct skb_checksum_ops {
3245 __wsum (*update)(const void *mem, int len, __wsum wsum);
3246 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3247};
3248
Davide Caratti96178132017-05-18 15:44:37 +02003249extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3250
Daniel Borkmann2817a332013-10-30 11:50:51 +01003251__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3252 __wsum csum, const struct skb_checksum_ops *ops);
3253__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3254 __wsum csum);
3255
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003256static inline void * __must_check
3257__skb_header_pointer(const struct sk_buff *skb, int offset,
3258 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003259{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003260 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003261 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003262
David S. Miller690e36e2014-08-23 12:13:41 -07003263 if (!skb ||
3264 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003265 return NULL;
3266
3267 return buffer;
3268}
3269
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003270static inline void * __must_check
3271skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003272{
3273 return __skb_header_pointer(skb, offset, len, skb->data,
3274 skb_headlen(skb), buffer);
3275}
3276
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003277/**
3278 * skb_needs_linearize - check if we need to linearize a given skb
3279 * depending on the given device features.
3280 * @skb: socket buffer to check
3281 * @features: net device features
3282 *
3283 * Returns true if either:
3284 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3285 * 2. skb is fragmented and the device does not support SG.
3286 */
3287static inline bool skb_needs_linearize(struct sk_buff *skb,
3288 netdev_features_t features)
3289{
3290 return skb_is_nonlinear(skb) &&
3291 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3292 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3293}
3294
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003295static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3296 void *to,
3297 const unsigned int len)
3298{
3299 memcpy(to, skb->data, len);
3300}
3301
3302static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3303 const int offset, void *to,
3304 const unsigned int len)
3305{
3306 memcpy(to, skb->data + offset, len);
3307}
3308
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003309static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3310 const void *from,
3311 const unsigned int len)
3312{
3313 memcpy(skb->data, from, len);
3314}
3315
3316static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3317 const int offset,
3318 const void *from,
3319 const unsigned int len)
3320{
3321 memcpy(skb->data + offset, from, len);
3322}
3323
Joe Perches7965bd42013-09-26 14:48:15 -07003324void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003325
Patrick Ohlyac45f602009-02-12 05:03:37 +00003326static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3327{
3328 return skb->tstamp;
3329}
3330
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003331/**
3332 * skb_get_timestamp - get timestamp from a skb
3333 * @skb: skb to get stamp from
3334 * @stamp: pointer to struct timeval to store stamp in
3335 *
3336 * Timestamps are stored in the skb as offsets to a base timestamp.
3337 * This function converts the offset back to a struct timeval and stores
3338 * it in stamp.
3339 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003340static inline void skb_get_timestamp(const struct sk_buff *skb,
3341 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003342{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003343 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003344}
3345
Patrick Ohlyac45f602009-02-12 05:03:37 +00003346static inline void skb_get_timestampns(const struct sk_buff *skb,
3347 struct timespec *stamp)
3348{
3349 *stamp = ktime_to_timespec(skb->tstamp);
3350}
3351
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003352static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003353{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003354 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003355}
3356
Stephen Hemminger164891a2007-04-23 22:26:16 -07003357static inline ktime_t net_timedelta(ktime_t t)
3358{
3359 return ktime_sub(ktime_get_real(), t);
3360}
3361
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003362static inline ktime_t net_invalid_timestamp(void)
3363{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003364 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003365}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003366
Alexander Duyck62bccb82014-09-04 13:31:35 -04003367struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3368
Richard Cochranc1f19b52010-07-17 08:49:36 +00003369#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3370
Joe Perches7965bd42013-09-26 14:48:15 -07003371void skb_clone_tx_timestamp(struct sk_buff *skb);
3372bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003373
3374#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3375
3376static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3377{
3378}
3379
3380static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3381{
3382 return false;
3383}
3384
3385#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3386
3387/**
3388 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3389 *
Richard Cochranda92b192011-10-21 00:49:15 +00003390 * PHY drivers may accept clones of transmitted packets for
3391 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003392 * must call this function to return the skb back to the stack with a
3393 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003394 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003395 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003396 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003397 *
3398 */
3399void skb_complete_tx_timestamp(struct sk_buff *skb,
3400 struct skb_shared_hwtstamps *hwtstamps);
3401
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003402void __skb_tstamp_tx(struct sk_buff *orig_skb,
3403 struct skb_shared_hwtstamps *hwtstamps,
3404 struct sock *sk, int tstype);
3405
Patrick Ohlyac45f602009-02-12 05:03:37 +00003406/**
3407 * skb_tstamp_tx - queue clone of skb with send time stamps
3408 * @orig_skb: the original outgoing packet
3409 * @hwtstamps: hardware time stamps, may be NULL if not available
3410 *
3411 * If the skb has a socket associated, then this function clones the
3412 * skb (thus sharing the actual data and optional structures), stores
3413 * the optional hardware time stamping information (if non NULL) or
3414 * generates a software time stamp (otherwise), then queues the clone
3415 * to the error queue of the socket. Errors are silently ignored.
3416 */
Joe Perches7965bd42013-09-26 14:48:15 -07003417void skb_tstamp_tx(struct sk_buff *orig_skb,
3418 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003419
Richard Cochran4507a712010-07-17 08:48:28 +00003420/**
3421 * skb_tx_timestamp() - Driver hook for transmit timestamping
3422 *
3423 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003424 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003425 *
David S. Miller73409f32013-12-27 13:04:33 -05003426 * Specifically, one should make absolutely sure that this function is
3427 * called before TX completion of this packet can trigger. Otherwise
3428 * the packet could potentially already be freed.
3429 *
Richard Cochran4507a712010-07-17 08:48:28 +00003430 * @skb: A socket buffer.
3431 */
3432static inline void skb_tx_timestamp(struct sk_buff *skb)
3433{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003434 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003435 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3436 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003437}
3438
Johannes Berg6e3e9392011-11-09 10:15:42 +01003439/**
3440 * skb_complete_wifi_ack - deliver skb with wifi status
3441 *
3442 * @skb: the original outgoing packet
3443 * @acked: ack status
3444 *
3445 */
3446void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3447
Joe Perches7965bd42013-09-26 14:48:15 -07003448__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3449__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003450
Herbert Xu60476372007-04-09 11:59:39 -07003451static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3452{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003453 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3454 skb->csum_valid ||
3455 (skb->ip_summed == CHECKSUM_PARTIAL &&
3456 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003457}
3458
Herbert Xufb286bb2005-11-10 13:01:24 -08003459/**
3460 * skb_checksum_complete - Calculate checksum of an entire packet
3461 * @skb: packet to process
3462 *
3463 * This function calculates the checksum over the entire packet plus
3464 * the value of skb->csum. The latter can be used to supply the
3465 * checksum of a pseudo header as used by TCP/UDP. It returns the
3466 * checksum.
3467 *
3468 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3469 * this function can be used to verify that checksum on received
3470 * packets. In that case the function should return zero if the
3471 * checksum is correct. In particular, this function will return zero
3472 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3473 * hardware has already verified the correctness of the checksum.
3474 */
Al Viro4381ca32007-07-15 21:00:11 +01003475static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003476{
Herbert Xu60476372007-04-09 11:59:39 -07003477 return skb_csum_unnecessary(skb) ?
3478 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003479}
3480
Tom Herbert77cffe22014-08-27 21:26:46 -07003481static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3482{
3483 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3484 if (skb->csum_level == 0)
3485 skb->ip_summed = CHECKSUM_NONE;
3486 else
3487 skb->csum_level--;
3488 }
3489}
3490
3491static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3492{
3493 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3494 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3495 skb->csum_level++;
3496 } else if (skb->ip_summed == CHECKSUM_NONE) {
3497 skb->ip_summed = CHECKSUM_UNNECESSARY;
3498 skb->csum_level = 0;
3499 }
3500}
3501
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003502/* Check if we need to perform checksum complete validation.
3503 *
3504 * Returns true if checksum complete is needed, false otherwise
3505 * (either checksum is unnecessary or zero checksum is allowed).
3506 */
3507static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3508 bool zero_okay,
3509 __sum16 check)
3510{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003511 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3512 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003513 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003514 return false;
3515 }
3516
3517 return true;
3518}
3519
3520/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3521 * in checksum_init.
3522 */
3523#define CHECKSUM_BREAK 76
3524
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003525/* Unset checksum-complete
3526 *
3527 * Unset checksum complete can be done when packet is being modified
3528 * (uncompressed for instance) and checksum-complete value is
3529 * invalidated.
3530 */
3531static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3532{
3533 if (skb->ip_summed == CHECKSUM_COMPLETE)
3534 skb->ip_summed = CHECKSUM_NONE;
3535}
3536
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003537/* Validate (init) checksum based on checksum complete.
3538 *
3539 * Return values:
3540 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3541 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3542 * checksum is stored in skb->csum for use in __skb_checksum_complete
3543 * non-zero: value of invalid checksum
3544 *
3545 */
3546static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3547 bool complete,
3548 __wsum psum)
3549{
3550 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3551 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003552 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003553 return 0;
3554 }
3555 }
3556
3557 skb->csum = psum;
3558
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003559 if (complete || skb->len <= CHECKSUM_BREAK) {
3560 __sum16 csum;
3561
3562 csum = __skb_checksum_complete(skb);
3563 skb->csum_valid = !csum;
3564 return csum;
3565 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003566
3567 return 0;
3568}
3569
3570static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3571{
3572 return 0;
3573}
3574
3575/* Perform checksum validate (init). Note that this is a macro since we only
3576 * want to calculate the pseudo header which is an input function if necessary.
3577 * First we try to validate without any computation (checksum unnecessary) and
3578 * then calculate based on checksum complete calling the function to compute
3579 * pseudo header.
3580 *
3581 * Return values:
3582 * 0: checksum is validated or try to in skb_checksum_complete
3583 * non-zero: value of invalid checksum
3584 */
3585#define __skb_checksum_validate(skb, proto, complete, \
3586 zero_okay, check, compute_pseudo) \
3587({ \
3588 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003589 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003590 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3591 __ret = __skb_checksum_validate_complete(skb, \
3592 complete, compute_pseudo(skb, proto)); \
3593 __ret; \
3594})
3595
3596#define skb_checksum_init(skb, proto, compute_pseudo) \
3597 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3598
3599#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3600 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3601
3602#define skb_checksum_validate(skb, proto, compute_pseudo) \
3603 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3604
3605#define skb_checksum_validate_zero_check(skb, proto, check, \
3606 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003607 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003608
3609#define skb_checksum_simple_validate(skb) \
3610 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3611
Tom Herbertd96535a2014-08-31 15:12:42 -07003612static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3613{
Davide Caratti219f1d792017-05-18 15:44:39 +02003614 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003615}
3616
3617static inline void __skb_checksum_convert(struct sk_buff *skb,
3618 __sum16 check, __wsum pseudo)
3619{
3620 skb->csum = ~pseudo;
3621 skb->ip_summed = CHECKSUM_COMPLETE;
3622}
3623
3624#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3625do { \
3626 if (__skb_checksum_convert_check(skb)) \
3627 __skb_checksum_convert(skb, check, \
3628 compute_pseudo(skb, proto)); \
3629} while (0)
3630
Tom Herbert15e23962015-02-10 16:30:31 -08003631static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3632 u16 start, u16 offset)
3633{
3634 skb->ip_summed = CHECKSUM_PARTIAL;
3635 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3636 skb->csum_offset = offset - start;
3637}
3638
Tom Herbertdcdc8992015-02-02 16:07:34 -08003639/* Update skbuf and packet to reflect the remote checksum offload operation.
3640 * When called, ptr indicates the starting point for skb->csum when
3641 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3642 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3643 */
3644static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003645 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003646{
3647 __wsum delta;
3648
Tom Herbert15e23962015-02-10 16:30:31 -08003649 if (!nopartial) {
3650 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3651 return;
3652 }
3653
Tom Herbertdcdc8992015-02-02 16:07:34 -08003654 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3655 __skb_checksum_complete(skb);
3656 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3657 }
3658
3659 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3660
3661 /* Adjust skb->csum since we changed the packet */
3662 skb->csum = csum_add(skb->csum, delta);
3663}
3664
Florian Westphalcb9c6832017-01-23 18:21:56 +01003665static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3666{
3667#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003668 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003669#else
3670 return NULL;
3671#endif
3672}
3673
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003674#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003675void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003676static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3677{
3678 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003679 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003680}
3681static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3682{
3683 if (nfct)
3684 atomic_inc(&nfct->use);
3685}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003686#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003687#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003688static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3689{
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003690 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003691 kfree(nf_bridge);
3692}
3693static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3694{
3695 if (nf_bridge)
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003696 refcount_inc(&nf_bridge->use);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003697}
3698#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003699static inline void nf_reset(struct sk_buff *skb)
3700{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003701#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003702 nf_conntrack_put(skb_nfct(skb));
3703 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003704#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003705#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003706 nf_bridge_put(skb->nf_bridge);
3707 skb->nf_bridge = NULL;
3708#endif
3709}
3710
Patrick McHardy124dff02013-04-05 20:42:05 +02003711static inline void nf_reset_trace(struct sk_buff *skb)
3712{
Florian Westphal478b3602014-02-15 23:48:45 +01003713#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003714 skb->nf_trace = 0;
3715#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003716}
3717
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003718/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003719static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3720 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003721{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003722#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003723 dst->_nfct = src->_nfct;
3724 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003725#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003726#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003727 dst->nf_bridge = src->nf_bridge;
3728 nf_bridge_get(src->nf_bridge);
3729#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003730#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003731 if (copy)
3732 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003733#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003734}
3735
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003736static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3737{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003738#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003739 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003740#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003741#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003742 nf_bridge_put(dst->nf_bridge);
3743#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003744 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003745}
3746
James Morris984bc162006-06-09 00:29:17 -07003747#ifdef CONFIG_NETWORK_SECMARK
3748static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3749{
3750 to->secmark = from->secmark;
3751}
3752
3753static inline void skb_init_secmark(struct sk_buff *skb)
3754{
3755 skb->secmark = 0;
3756}
3757#else
3758static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3759{ }
3760
3761static inline void skb_init_secmark(struct sk_buff *skb)
3762{ }
3763#endif
3764
Eric W. Biederman574f7192014-04-01 12:20:24 -07003765static inline bool skb_irq_freeable(const struct sk_buff *skb)
3766{
3767 return !skb->destructor &&
3768#if IS_ENABLED(CONFIG_XFRM)
3769 !skb->sp &&
3770#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003771 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003772 !skb->_skb_refdst &&
3773 !skb_has_frag_list(skb);
3774}
3775
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003776static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3777{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003778 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003779}
3780
Stephen Hemminger92477442009-03-21 13:39:26 -07003781static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003782{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003783 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003784}
3785
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003786static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3787{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003788 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003789}
3790
David S. Millerd5a9e242009-01-27 16:22:11 -08003791static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3792{
3793 skb->queue_mapping = rx_queue + 1;
3794}
3795
Stephen Hemminger92477442009-03-21 13:39:26 -07003796static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003797{
3798 return skb->queue_mapping - 1;
3799}
3800
Stephen Hemminger92477442009-03-21 13:39:26 -07003801static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003802{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003803 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003804}
3805
Julian Anastasov4ff06202017-02-06 23:14:12 +02003806static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3807{
3808 skb->dst_pending_confirm = val;
3809}
3810
3811static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3812{
3813 return skb->dst_pending_confirm != 0;
3814}
3815
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003816static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3817{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003818#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003819 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003820#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003821 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003822#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003823}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003824
Pravin B Shelar68c33162013-02-14 14:02:41 +00003825/* Keeps track of mac header offset relative to skb->head.
3826 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3827 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003828 * tunnel skb it points to outer mac header.
3829 * Keeps track of level of encapsulation of network headers.
3830 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003831struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003832 union {
3833 int mac_offset;
3834 int data_offset;
3835 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003836 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003837 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003838 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003839};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003840#define SKB_SGO_CB_OFFSET 32
3841#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003842
3843static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3844{
3845 return (skb_mac_header(inner_skb) - inner_skb->head) -
3846 SKB_GSO_CB(inner_skb)->mac_offset;
3847}
3848
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003849static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3850{
3851 int new_headroom, headroom;
3852 int ret;
3853
3854 headroom = skb_headroom(skb);
3855 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3856 if (ret)
3857 return ret;
3858
3859 new_headroom = skb_headroom(skb);
3860 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3861 return 0;
3862}
3863
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003864static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3865{
3866 /* Do not update partial checksums if remote checksum is enabled. */
3867 if (skb->remcsum_offload)
3868 return;
3869
3870 SKB_GSO_CB(skb)->csum = res;
3871 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3872}
3873
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003874/* Compute the checksum for a gso segment. First compute the checksum value
3875 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3876 * then add in skb->csum (checksum from csum_start to end of packet).
3877 * skb->csum and csum_start are then updated to reflect the checksum of the
3878 * resultant packet starting from the transport header-- the resultant checksum
3879 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3880 * header.
3881 */
3882static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3883{
Alexander Duyck76443452016-02-05 15:27:37 -08003884 unsigned char *csum_start = skb_transport_header(skb);
3885 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3886 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003887
Alexander Duyck76443452016-02-05 15:27:37 -08003888 SKB_GSO_CB(skb)->csum = res;
3889 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003890
Alexander Duyck76443452016-02-05 15:27:37 -08003891 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003892}
3893
David S. Millerbdcc0922012-03-07 20:53:36 -05003894static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003895{
3896 return skb_shinfo(skb)->gso_size;
3897}
3898
Eric Dumazet36a8f392013-09-29 01:21:32 -07003899/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003900static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003901{
3902 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3903}
3904
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003905static inline void skb_gso_reset(struct sk_buff *skb)
3906{
3907 skb_shinfo(skb)->gso_size = 0;
3908 skb_shinfo(skb)->gso_segs = 0;
3909 skb_shinfo(skb)->gso_type = 0;
3910}
3911
Joe Perches7965bd42013-09-26 14:48:15 -07003912void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003913
3914static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3915{
3916 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3917 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003918 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3919
Alexander Duyckb78462e2010-06-02 12:24:37 +00003920 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3921 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003922 __skb_warn_lro_forwarding(skb);
3923 return true;
3924 }
3925 return false;
3926}
3927
Herbert Xu35fc92a2007-03-26 23:22:20 -07003928static inline void skb_forward_csum(struct sk_buff *skb)
3929{
3930 /* Unfortunately we don't support this one. Any brave souls? */
3931 if (skb->ip_summed == CHECKSUM_COMPLETE)
3932 skb->ip_summed = CHECKSUM_NONE;
3933}
3934
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003935/**
3936 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3937 * @skb: skb to check
3938 *
3939 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3940 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3941 * use this helper, to document places where we make this assertion.
3942 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003943static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003944{
3945#ifdef DEBUG
3946 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3947#endif
3948}
3949
Rusty Russellf35d9d82008-02-04 23:49:54 -05003950bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003951
Paul Durranted1f50c2014-01-09 10:02:46 +00003952int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003953struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3954 unsigned int transport_len,
3955 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003956
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003957/**
3958 * skb_head_is_locked - Determine if the skb->head is locked down
3959 * @skb: skb to check
3960 *
3961 * The head on skbs build around a head frag can be removed if they are
3962 * not cloned. This function returns true if the skb head is locked down
3963 * due to either being allocated via kmalloc, or by being a clone with
3964 * multiple references to the head.
3965 */
3966static inline bool skb_head_is_locked(const struct sk_buff *skb)
3967{
3968 return !skb->head_frag || skb_cloned(skb);
3969}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003970
3971/**
3972 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3973 *
3974 * @skb: GSO skb
3975 *
3976 * skb_gso_network_seglen is used to determine the real size of the
3977 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3978 *
3979 * The MAC/L2 header is not accounted for.
3980 */
3981static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3982{
3983 unsigned int hdr_len = skb_transport_header(skb) -
3984 skb_network_header(skb);
3985 return hdr_len + skb_gso_transport_seglen(skb);
3986}
Thomas Grafee122c72015-07-21 10:43:58 +02003987
Edward Cree179bc672016-02-11 20:48:04 +00003988/* Local Checksum Offload.
3989 * Compute outer checksum based on the assumption that the
3990 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003991 * See Documentation/networking/checksum-offloads.txt for
3992 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003993 * Fill in outer checksum adjustment (e.g. with sum of outer
3994 * pseudo-header) before calling.
3995 * Also ensure that inner checksum is in linear data area.
3996 */
3997static inline __wsum lco_csum(struct sk_buff *skb)
3998{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003999 unsigned char *csum_start = skb_checksum_start(skb);
4000 unsigned char *l4_hdr = skb_transport_header(skb);
4001 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004002
4003 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004004 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4005 skb->csum_offset));
4006
Edward Cree179bc672016-02-11 20:48:04 +00004007 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004008 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004009 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004010 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004011}
4012
Linus Torvalds1da177e2005-04-16 15:20:36 -07004013#endif /* __KERNEL__ */
4014#endif /* _LINUX_SKBUFF_H */