blob: 8c0708d2e5e6655c648a5718000fac4103c7f321 [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);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400447 union {
448 struct {
449 unsigned long desc;
450 void *ctx;
451 };
452 struct {
453 u32 id;
454 u16 len;
455 u16 zerocopy:1;
456 u32 bytelen;
457 };
458 };
Willem de Bruijn52267792017-08-03 16:29:39 -0400459 atomic_t refcnt;
Willem de Bruijna91dbff2017-08-03 16:29:43 -0400460
461 struct mmpin {
462 struct user_struct *user;
463 unsigned int num_pg;
464 } mmp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000465};
466
Willem de Bruijn52267792017-08-03 16:29:39 -0400467#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
468
469struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400470struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
471 struct ubuf_info *uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -0400472
473static inline void sock_zerocopy_get(struct ubuf_info *uarg)
474{
475 atomic_inc(&uarg->refcnt);
476}
477
478void sock_zerocopy_put(struct ubuf_info *uarg);
479void sock_zerocopy_put_abort(struct ubuf_info *uarg);
480
481void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
482
483int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
484 struct msghdr *msg, int len,
485 struct ubuf_info *uarg);
486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487/* This data is invariant across clones and lives at
488 * the end of the header data, ie. at skb->end.
489 */
490struct skb_shared_info {
Steffen Klassert7f564522017-04-08 20:36:24 +0200491 unsigned short _unused;
Ian Campbell9f42f122012-01-05 07:13:39 +0000492 unsigned char nr_frags;
493 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700494 unsigned short gso_size;
495 /* Warning: this field is not always filled in (UFO)! */
496 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000498 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200499 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400500 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000501 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700502
503 /*
504 * Warning : all fields before dataref are cleared in __alloc_skb()
505 */
506 atomic_t dataref;
507
Johann Baudy69e3c752009-05-18 22:11:22 -0700508 /* Intermediate layers must ensure that destructor_arg
509 * remains valid until skb destructor */
510 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000511
Eric Dumazetfed66382010-07-22 19:09:08 +0000512 /* must be last field, see pskb_expand_head() */
513 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514};
515
516/* We divide dataref into two halves. The higher 16 bits hold references
517 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700518 * the entire skb->data. A clone of a headerless skb holds the length of
519 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 *
521 * All users must obey the rule that the skb->data reference count must be
522 * greater than or equal to the payload reference count.
523 *
524 * Holding a reference to the payload part means that the user does not
525 * care about modifications to the header part of skb->data.
526 */
527#define SKB_DATAREF_SHIFT 16
528#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
529
David S. Millerd179cd12005-08-17 14:57:30 -0700530
531enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700532 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
533 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
534 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700535};
536
Herbert Xu79671682006-06-22 02:40:14 -0700537enum {
538 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700539
540 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700541 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700542
543 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700544 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700545
David S. Millerd9d30ad2017-07-03 07:31:57 -0700546 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800547
David S. Millerd9d30ad2017-07-03 07:31:57 -0700548 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000549
David S. Millerd9d30ad2017-07-03 07:31:57 -0700550 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000551
David S. Millerd9d30ad2017-07-03 07:31:57 -0700552 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000553
David S. Millerd9d30ad2017-07-03 07:31:57 -0700554 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700555
David S. Millerd9d30ad2017-07-03 07:31:57 -0700556 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700557
David S. Millerd9d30ad2017-07-03 07:31:57 -0700558 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700559
David S. Millerd9d30ad2017-07-03 07:31:57 -0700560 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700561
David S. Millerd9d30ad2017-07-03 07:31:57 -0700562 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400563
David S. Millerd9d30ad2017-07-03 07:31:57 -0700564 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400565
David S. Millerd9d30ad2017-07-03 07:31:57 -0700566 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300567
David S. Millerd9d30ad2017-07-03 07:31:57 -0700568 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200569
David S. Millerd9d30ad2017-07-03 07:31:57 -0700570 SKB_GSO_ESP = 1 << 15,
Herbert Xu79671682006-06-22 02:40:14 -0700571};
572
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700573#if BITS_PER_LONG > 32
574#define NET_SKBUFF_DATA_USES_OFFSET 1
575#endif
576
577#ifdef NET_SKBUFF_DATA_USES_OFFSET
578typedef unsigned int sk_buff_data_t;
579#else
580typedef unsigned char *sk_buff_data_t;
581#endif
582
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583/**
584 * struct sk_buff - socket buffer
585 * @next: Next buffer in list
586 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800587 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800588 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700589 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700591 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000592 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700593 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 * @len: Length of actual data
595 * @data_len: Data length
596 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700597 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700598 * @csum: Checksum (must include start/offset pair)
599 * @csum_start: Offset from skb->head where checksumming should start
600 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700601 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700602 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700604 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 * @nohdr: Payload reference only, must not modify header
606 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700607 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700608 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500609 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500610 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500611 * @tc_redirected: packet was redirected by a tc action
612 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800613 * @peeked: this packet has been seen already, so stats have been
614 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700615 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700616 * @protocol: Packet protocol from driver
617 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100618 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800620 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700622 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700623 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700624 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700625 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700626 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700627 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700628 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700629 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100630 * @wifi_acked_valid: wifi_acked was set
631 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000632 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200633 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200634 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300635 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700636 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700637 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000638 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700639 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000640 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000641 * @inner_transport_header: Inner transport layer header (encapsulation)
642 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000643 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700644 * @transport_header: Transport layer header
645 * @network_header: Network layer header
646 * @mac_header: Link layer header
647 * @tail: Tail pointer
648 * @end: End pointer
649 * @head: Head of buffer
650 * @data: Data head pointer
651 * @truesize: Buffer size
652 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 */
654
655struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800656 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800657 struct {
658 /* These two members must be first. */
659 struct sk_buff *next;
660 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000661
Eric Dumazet56b17422014-11-03 08:19:53 -0800662 union {
663 ktime_t tstamp;
Eric Dumazet9a568de2017-05-16 14:00:14 -0700664 u64 skb_mstamp;
Eric Dumazet56b17422014-11-03 08:19:53 -0800665 };
666 };
667 struct rb_node rbnode; /* used in netem & tcp stack */
668 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000669 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
Eric Dumazetc84d9492016-12-08 11:41:55 -0800671 union {
672 struct net_device *dev;
673 /* Some protocols might use this space to store information,
674 * while device pointer would be NULL.
675 * UDP receive path is one user.
676 */
677 unsigned long dev_scratch;
678 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 /*
680 * This is the control buffer. It is free to use for every
681 * layer. Please put your private variables there. If you
682 * want to keep them across layers you have to do a skb_clone()
683 * first. This is owned by whoever has the skb queued ATM.
684 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000685 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
Eric Dumazet7fee2262010-05-11 23:19:48 +0000687 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700688 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000689#ifdef CONFIG_XFRM
690 struct sec_path *sp;
691#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700692#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100693 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700694#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200695#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700696 struct nf_bridge_info *nf_bridge;
697#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700699 data_len;
700 __u16 mac_len,
701 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700702
703 /* Following fields are _not_ copied in __copy_skb_header()
704 * Note that queue_mapping is here mostly to fill a hole.
705 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200706 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700707 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200708
709/* if you move cloned around you also must adapt those constants */
710#ifdef __BIG_ENDIAN_BITFIELD
711#define CLONED_MASK (1 << 7)
712#else
713#define CLONED_MASK 1
714#endif
715#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
716
717 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700718 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700719 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800720 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800721 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700722 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200723 xmit_more:1,
724 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200725 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726
Eric Dumazetb1937222014-09-28 22:18:47 -0700727 /* fields enclosed in headers_start/headers_end are copied
728 * using a single memcpy() in __copy_skb_header()
729 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700730 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700731 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700732 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200733
734/* if you move pkt_type around you also must adapt those constants */
735#ifdef __BIG_ENDIAN_BITFIELD
736#define PKT_TYPE_MAX (7 << 5)
737#else
738#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100739#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200740#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
741
742 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700743 __u8 pkt_type:3;
744 __u8 pfmemalloc:1;
745 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
Eric Dumazetb1937222014-09-28 22:18:47 -0700747 __u8 nf_trace:1;
748 __u8 ip_summed:2;
749 __u8 ooo_okay:1;
750 __u8 l4_hash:1;
751 __u8 sw_hash:1;
752 __u8 wifi_acked_valid:1;
753 __u8 wifi_acked:1;
754
755 __u8 no_fcs:1;
756 /* Indicates the inner headers are valid in the skbuff. */
757 __u8 encapsulation:1;
758 __u8 encap_hdr_csum:1;
759 __u8 csum_valid:1;
760 __u8 csum_complete_sw:1;
761 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200762 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700763
Julian Anastasov4ff06202017-02-06 23:14:12 +0200764 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700765#ifdef CONFIG_IPV6_NDISC_NODETYPE
766 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700768 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700769 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800770 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200771#ifdef CONFIG_NET_SWITCHDEV
772 __u8 offload_fwd_mark:1;
773#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500774#ifdef CONFIG_NET_CLS_ACT
775 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500776 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500777 __u8 tc_redirected:1;
778 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500779#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700782 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200784
Eric Dumazetb1937222014-09-28 22:18:47 -0700785 union {
786 __wsum csum;
787 struct {
788 __u16 csum_start;
789 __u16 csum_offset;
790 };
791 };
792 __u32 priority;
793 int skb_iif;
794 __u32 hash;
795 __be16 vlan_proto;
796 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800797#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
798 union {
799 unsigned int napi_id;
800 unsigned int sender_cpu;
801 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700802#endif
James Morris984bc162006-06-09 00:29:17 -0700803#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200804 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700805#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700806
Neil Horman3b885782009-10-12 13:26:31 -0700807 union {
808 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000809 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700810 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Tom Herbert8bce6d72014-09-29 20:22:29 -0700812 union {
813 __be16 inner_protocol;
814 __u8 inner_ipproto;
815 };
816
Simon Horman1a37e412013-05-23 21:02:51 +0000817 __u16 inner_transport_header;
818 __u16 inner_network_header;
819 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700820
821 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000822 __u16 transport_header;
823 __u16 network_header;
824 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700825
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700826 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700827 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700828 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700829
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700831 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700832 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700834 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700835 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300836 refcount_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837};
838
839#ifdef __KERNEL__
840/*
841 * Handling routines are only of interest to the kernel
842 */
843#include <linux/slab.h>
844
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845
Mel Gormanc93bdd02012-07-31 16:44:19 -0700846#define SKB_ALLOC_FCLONE 0x01
847#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800848#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700849
850/* Returns true if the skb was allocated from PFMEMALLOC reserves */
851static inline bool skb_pfmemalloc(const struct sk_buff *skb)
852{
853 return unlikely(skb->pfmemalloc);
854}
855
Eric Dumazet7fee2262010-05-11 23:19:48 +0000856/*
857 * skb might have a dst pointer attached, refcounted or not.
858 * _skb_refdst low order bit is set if refcount was _not_ taken
859 */
860#define SKB_DST_NOREF 1UL
861#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
862
Florian Westphala9e419d2017-01-23 18:21:59 +0100863#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000864/**
865 * skb_dst - returns skb dst_entry
866 * @skb: buffer
867 *
868 * Returns skb dst_entry, regardless of reference taken or not.
869 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000870static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
871{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000872 /* If refdst was not refcounted, check we still are in a
873 * rcu_read_lock section
874 */
875 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
876 !rcu_read_lock_held() &&
877 !rcu_read_lock_bh_held());
878 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000879}
880
Eric Dumazet7fee2262010-05-11 23:19:48 +0000881/**
882 * skb_dst_set - sets skb dst
883 * @skb: buffer
884 * @dst: dst entry
885 *
886 * Sets skb dst, assuming a reference was taken on dst and should
887 * be released by skb_dst_drop()
888 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000889static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
890{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000891 skb->_skb_refdst = (unsigned long)dst;
892}
893
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200894/**
895 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
896 * @skb: buffer
897 * @dst: dst entry
898 *
899 * Sets skb dst, assuming a reference was not taken on dst.
900 * If dst entry is cached, we do not take reference and dst_release
901 * will be avoided by refdst_drop. If dst entry is not cached, we take
902 * reference, so that last dst_release can destroy the dst immediately.
903 */
904static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
905{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100906 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
907 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200908}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000909
910/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300911 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000912 * @skb: buffer
913 */
914static inline bool skb_dst_is_noref(const struct sk_buff *skb)
915{
916 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000917}
918
Eric Dumazet511c3f92009-06-02 05:14:27 +0000919static inline struct rtable *skb_rtable(const struct sk_buff *skb)
920{
Eric Dumazetadf30902009-06-02 05:19:30 +0000921 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000922}
923
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400924/* For mangling skb->pkt_type from user space side from applications
925 * such as nft, tc, etc, we only allow a conservative subset of
926 * possible pkt_types to be set.
927*/
928static inline bool skb_pkt_type_ok(u32 ptype)
929{
930 return ptype <= PACKET_OTHERHOST;
931}
932
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200933static inline unsigned int skb_napi_id(const struct sk_buff *skb)
934{
935#ifdef CONFIG_NET_RX_BUSY_POLL
936 return skb->napi_id;
937#else
938 return 0;
939#endif
940}
941
Paolo Abeni3889a8032017-06-12 11:23:41 +0200942/* decrement the reference count and return true if we can free the skb */
943static inline bool skb_unref(struct sk_buff *skb)
944{
945 if (unlikely(!skb))
946 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +0300947 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200948 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +0300949 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200950 return false;
951
952 return true;
953}
954
Paolo Abeni0a463c72017-06-12 11:23:42 +0200955void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700956void kfree_skb(struct sk_buff *skb);
957void kfree_skb_list(struct sk_buff *segs);
958void skb_tx_error(struct sk_buff *skb);
959void consume_skb(struct sk_buff *skb);
Paolo Abeni0a463c72017-06-12 11:23:42 +0200960void consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700961void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000962extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000963
Joe Perches7965bd42013-09-26 14:48:15 -0700964void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
965bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
966 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000967
Joe Perches7965bd42013-09-26 14:48:15 -0700968struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
969 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700970struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700971struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700972static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100973 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700974{
Eric Dumazet564824b2010-10-11 19:05:25 +0000975 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700976}
977
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700978struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
979 unsigned long data_len,
980 int max_page_order,
981 int *errcode,
982 gfp_t gfp_mask);
983
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700984/* Layout of fast clones : [skb1][skb2][fclone_ref] */
985struct sk_buff_fclones {
986 struct sk_buff skb1;
987
988 struct sk_buff skb2;
989
Reshetova, Elena26385952017-06-30 13:07:59 +0300990 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700991};
992
993/**
994 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700995 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700996 * @skb: buffer
997 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900998 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700999 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1000 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001001 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001002static inline bool skb_fclone_busy(const struct sock *sk,
1003 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001004{
1005 const struct sk_buff_fclones *fclones;
1006
1007 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1008
1009 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001010 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001011 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001012}
1013
David S. Millerd179cd12005-08-17 14:57:30 -07001014static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001015 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001016{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001017 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001018}
1019
Joe Perches7965bd42013-09-26 14:48:15 -07001020struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1021int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1022struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1023struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001024struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1025 gfp_t gfp_mask, bool fclone);
1026static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1027 gfp_t gfp_mask)
1028{
1029 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1030}
Eric Dumazet117632e2011-12-03 21:39:53 +00001031
Joe Perches7965bd42013-09-26 14:48:15 -07001032int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1033struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1034 unsigned int headroom);
1035struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1036 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001037int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1038 int offset, int len);
1039int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1040 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001041int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1042int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +00001043#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
Joe Perches7965bd42013-09-26 14:48:15 -07001045int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1046 int getfrag(void *from, char *to, int offset,
1047 int len, int odd, struct sk_buff *skb),
1048 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001049
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001050int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1051 int offset, size_t size);
1052
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001053struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001054 __u32 lower_offset;
1055 __u32 upper_offset;
1056 __u32 frag_idx;
1057 __u32 stepped_offset;
1058 struct sk_buff *root_skb;
1059 struct sk_buff *cur_skb;
1060 __u8 *frag_data;
1061};
1062
Joe Perches7965bd42013-09-26 14:48:15 -07001063void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1064 unsigned int to, struct skb_seq_state *st);
1065unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1066 struct skb_seq_state *st);
1067void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001068
Joe Perches7965bd42013-09-26 14:48:15 -07001069unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001070 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001071
Tom Herbert09323cc2013-12-15 22:16:19 -08001072/*
1073 * Packet hash types specify the type of hash in skb_set_hash.
1074 *
1075 * Hash types refer to the protocol layer addresses which are used to
1076 * construct a packet's hash. The hashes are used to differentiate or identify
1077 * flows of the protocol layer for the hash type. Hash types are either
1078 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1079 *
1080 * Properties of hashes:
1081 *
1082 * 1) Two packets in different flows have different hash values
1083 * 2) Two packets in the same flow should have the same hash value
1084 *
1085 * A hash at a higher layer is considered to be more specific. A driver should
1086 * set the most specific hash possible.
1087 *
1088 * A driver cannot indicate a more specific hash than the layer at which a hash
1089 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1090 *
1091 * A driver may indicate a hash level which is less specific than the
1092 * actual layer the hash was computed on. For instance, a hash computed
1093 * at L4 may be considered an L3 hash. This should only be done if the
1094 * driver can't unambiguously determine that the HW computed the hash at
1095 * the higher layer. Note that the "should" in the second property above
1096 * permits this.
1097 */
1098enum pkt_hash_types {
1099 PKT_HASH_TYPE_NONE, /* Undefined type */
1100 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1101 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1102 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1103};
1104
Tom Herbertbcc83832015-09-01 09:24:24 -07001105static inline void skb_clear_hash(struct sk_buff *skb)
1106{
1107 skb->hash = 0;
1108 skb->sw_hash = 0;
1109 skb->l4_hash = 0;
1110}
1111
1112static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1113{
1114 if (!skb->l4_hash)
1115 skb_clear_hash(skb);
1116}
1117
1118static inline void
1119__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1120{
1121 skb->l4_hash = is_l4;
1122 skb->sw_hash = is_sw;
1123 skb->hash = hash;
1124}
1125
Tom Herbert09323cc2013-12-15 22:16:19 -08001126static inline void
1127skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1128{
Tom Herbertbcc83832015-09-01 09:24:24 -07001129 /* Used by drivers to set hash from HW */
1130 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1131}
1132
1133static inline void
1134__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1135{
1136 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001137}
1138
Tom Herberte5276932015-09-01 09:24:23 -07001139void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001140u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001141u32 skb_get_poff(const struct sk_buff *skb);
1142u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1143 const struct flow_keys *keys, int hlen);
1144__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1145 void *data, int hlen_proto);
1146
1147static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1148 int thoff, u8 ip_proto)
1149{
1150 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1151}
1152
1153void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1154 const struct flow_dissector_key *key,
1155 unsigned int key_count);
1156
1157bool __skb_flow_dissect(const struct sk_buff *skb,
1158 struct flow_dissector *flow_dissector,
1159 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001160 void *data, __be16 proto, int nhoff, int hlen,
1161 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001162
1163static inline bool skb_flow_dissect(const struct sk_buff *skb,
1164 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001165 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001166{
1167 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001168 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001169}
1170
1171static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001172 struct flow_keys *flow,
1173 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001174{
1175 memset(flow, 0, sizeof(*flow));
1176 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001177 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001178}
1179
1180static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1181 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001182 int nhoff, int hlen,
1183 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001184{
1185 memset(flow, 0, sizeof(*flow));
1186 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001187 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001188}
1189
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001190static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001191{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001192 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001193 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001194
Tom Herbert61b905d2014-03-24 15:34:47 -07001195 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001196}
1197
David S. Miller20a17bf2015-09-01 21:19:17 -07001198static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001199{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001200 if (!skb->l4_hash && !skb->sw_hash) {
1201 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001202 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001203
Tom Herbertde4c1f82015-09-01 18:11:04 -07001204 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001205 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001206
1207 return skb->hash;
1208}
1209
Tom Herbert50fb7992015-05-01 11:30:12 -07001210__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1211
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001212static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1213{
Tom Herbert61b905d2014-03-24 15:34:47 -07001214 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001215}
1216
Tom Herbert3df7a742013-12-15 22:16:29 -08001217static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1218{
Tom Herbert61b905d2014-03-24 15:34:47 -07001219 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001220 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001221 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001222};
1223
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001224#ifdef NET_SKBUFF_DATA_USES_OFFSET
1225static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1226{
1227 return skb->head + skb->end;
1228}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001229
1230static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1231{
1232 return skb->end;
1233}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001234#else
1235static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1236{
1237 return skb->end;
1238}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001239
1240static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1241{
1242 return skb->end - skb->head;
1243}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001244#endif
1245
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001247#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248
Patrick Ohlyac45f602009-02-12 05:03:37 +00001249static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1250{
1251 return &skb_shinfo(skb)->hwtstamps;
1252}
1253
Willem de Bruijn52267792017-08-03 16:29:39 -04001254static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1255{
1256 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1257
1258 return is_zcopy ? skb_uarg(skb) : NULL;
1259}
1260
1261static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1262{
1263 if (skb && uarg && !skb_zcopy(skb)) {
1264 sock_zerocopy_get(uarg);
1265 skb_shinfo(skb)->destructor_arg = uarg;
1266 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1267 }
1268}
1269
1270/* Release a reference on a zerocopy structure */
1271static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1272{
1273 struct ubuf_info *uarg = skb_zcopy(skb);
1274
1275 if (uarg) {
1276 uarg->zerocopy = uarg->zerocopy && zerocopy;
1277 sock_zerocopy_put(uarg);
1278 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1279 }
1280}
1281
1282/* Abort a zerocopy operation and revert zckey on error in send syscall */
1283static inline void skb_zcopy_abort(struct sk_buff *skb)
1284{
1285 struct ubuf_info *uarg = skb_zcopy(skb);
1286
1287 if (uarg) {
1288 sock_zerocopy_put_abort(uarg);
1289 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1290 }
1291}
1292
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293/**
1294 * skb_queue_empty - check if a queue is empty
1295 * @list: queue head
1296 *
1297 * Returns true if the queue is empty, false otherwise.
1298 */
1299static inline int skb_queue_empty(const struct sk_buff_head *list)
1300{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001301 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302}
1303
1304/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001305 * skb_queue_is_last - check if skb is the last entry in the queue
1306 * @list: queue head
1307 * @skb: buffer
1308 *
1309 * Returns true if @skb is the last buffer on the list.
1310 */
1311static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1312 const struct sk_buff *skb)
1313{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001314 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001315}
1316
1317/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001318 * skb_queue_is_first - check if skb is the first entry in the queue
1319 * @list: queue head
1320 * @skb: buffer
1321 *
1322 * Returns true if @skb is the first buffer on the list.
1323 */
1324static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1325 const struct sk_buff *skb)
1326{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001327 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001328}
1329
1330/**
David S. Miller249c8b42008-09-23 00:44:42 -07001331 * skb_queue_next - return the next packet in the queue
1332 * @list: queue head
1333 * @skb: current buffer
1334 *
1335 * Return the next packet in @list after @skb. It is only valid to
1336 * call this if skb_queue_is_last() evaluates to false.
1337 */
1338static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1339 const struct sk_buff *skb)
1340{
1341 /* This BUG_ON may seem severe, but if we just return then we
1342 * are going to dereference garbage.
1343 */
1344 BUG_ON(skb_queue_is_last(list, skb));
1345 return skb->next;
1346}
1347
1348/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001349 * skb_queue_prev - return the prev packet in the queue
1350 * @list: queue head
1351 * @skb: current buffer
1352 *
1353 * Return the prev packet in @list before @skb. It is only valid to
1354 * call this if skb_queue_is_first() evaluates to false.
1355 */
1356static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1357 const struct sk_buff *skb)
1358{
1359 /* This BUG_ON may seem severe, but if we just return then we
1360 * are going to dereference garbage.
1361 */
1362 BUG_ON(skb_queue_is_first(list, skb));
1363 return skb->prev;
1364}
1365
1366/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 * skb_get - reference buffer
1368 * @skb: buffer to reference
1369 *
1370 * Makes another reference to a socket buffer and returns a pointer
1371 * to the buffer.
1372 */
1373static inline struct sk_buff *skb_get(struct sk_buff *skb)
1374{
Reshetova, Elena63354792017-06-30 13:07:58 +03001375 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 return skb;
1377}
1378
1379/*
1380 * If users == 1, we are the only owner and are can avoid redundant
1381 * atomic change.
1382 */
1383
1384/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 * skb_cloned - is the buffer a clone
1386 * @skb: buffer to check
1387 *
1388 * Returns true if the buffer was generated with skb_clone() and is
1389 * one of multiple shared copies of the buffer. Cloned buffers are
1390 * shared data so must not be written to under normal circumstances.
1391 */
1392static inline int skb_cloned(const struct sk_buff *skb)
1393{
1394 return skb->cloned &&
1395 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1396}
1397
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001398static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1399{
Mel Gormand0164ad2015-11-06 16:28:21 -08001400 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001401
1402 if (skb_cloned(skb))
1403 return pskb_expand_head(skb, 0, 0, pri);
1404
1405 return 0;
1406}
1407
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408/**
1409 * skb_header_cloned - is the header a clone
1410 * @skb: buffer to check
1411 *
1412 * Returns true if modifying the header part of the buffer requires
1413 * the data to be copied.
1414 */
1415static inline int skb_header_cloned(const struct sk_buff *skb)
1416{
1417 int dataref;
1418
1419 if (!skb->cloned)
1420 return 0;
1421
1422 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1423 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1424 return dataref != 1;
1425}
1426
Eric Dumazet9580bf22016-04-30 10:19:29 -07001427static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1428{
1429 might_sleep_if(gfpflags_allow_blocking(pri));
1430
1431 if (skb_header_cloned(skb))
1432 return pskb_expand_head(skb, 0, 0, pri);
1433
1434 return 0;
1435}
1436
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437/**
1438 * skb_header_release - release reference to header
1439 * @skb: buffer to operate on
1440 *
1441 * Drop a reference to the header part of the buffer. This is done
1442 * by acquiring a payload reference. You must not read from the header
1443 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001444 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 */
1446static inline void skb_header_release(struct sk_buff *skb)
1447{
1448 BUG_ON(skb->nohdr);
1449 skb->nohdr = 1;
1450 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1451}
1452
1453/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001454 * __skb_header_release - release reference to header
1455 * @skb: buffer to operate on
1456 *
1457 * Variant of skb_header_release() assuming skb is private to caller.
1458 * We can avoid one atomic operation.
1459 */
1460static inline void __skb_header_release(struct sk_buff *skb)
1461{
1462 skb->nohdr = 1;
1463 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1464}
1465
1466
1467/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 * skb_shared - is the buffer shared
1469 * @skb: buffer to check
1470 *
1471 * Returns true if more than one person has a reference to this
1472 * buffer.
1473 */
1474static inline int skb_shared(const struct sk_buff *skb)
1475{
Reshetova, Elena63354792017-06-30 13:07:58 +03001476 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477}
1478
1479/**
1480 * skb_share_check - check if buffer is shared and if so clone it
1481 * @skb: buffer to check
1482 * @pri: priority for memory allocation
1483 *
1484 * If the buffer is shared the buffer is cloned and the old copy
1485 * drops a reference. A new clone with a single reference is returned.
1486 * If the buffer is not shared the original buffer is returned. When
1487 * being called from interrupt status or with spinlocks held pri must
1488 * be GFP_ATOMIC.
1489 *
1490 * NULL is returned on a memory allocation failure.
1491 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001492static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493{
Mel Gormand0164ad2015-11-06 16:28:21 -08001494 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 if (skb_shared(skb)) {
1496 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001497
1498 if (likely(nskb))
1499 consume_skb(skb);
1500 else
1501 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 skb = nskb;
1503 }
1504 return skb;
1505}
1506
1507/*
1508 * Copy shared buffers into a new sk_buff. We effectively do COW on
1509 * packets to handle cases where we have a local reader and forward
1510 * and a couple of other messy ones. The normal one is tcpdumping
1511 * a packet thats being forwarded.
1512 */
1513
1514/**
1515 * skb_unshare - make a copy of a shared buffer
1516 * @skb: buffer to check
1517 * @pri: priority for memory allocation
1518 *
1519 * If the socket buffer is a clone then this function creates a new
1520 * copy of the data, drops a reference count on the old copy and returns
1521 * the new copy with the reference count at 1. If the buffer is not a clone
1522 * the original buffer is returned. When called with a spinlock held or
1523 * from interrupt state @pri must be %GFP_ATOMIC
1524 *
1525 * %NULL is returned on a memory allocation failure.
1526 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001527static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001528 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529{
Mel Gormand0164ad2015-11-06 16:28:21 -08001530 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 if (skb_cloned(skb)) {
1532 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001533
1534 /* Free our shared copy */
1535 if (likely(nskb))
1536 consume_skb(skb);
1537 else
1538 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 skb = nskb;
1540 }
1541 return skb;
1542}
1543
1544/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001545 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 * @list_: list to peek at
1547 *
1548 * Peek an &sk_buff. Unlike most other operations you _MUST_
1549 * be careful with this one. A peek leaves the buffer on the
1550 * list and someone else may run off with it. You must hold
1551 * the appropriate locks or have a private queue to do this.
1552 *
1553 * Returns %NULL for an empty list or a pointer to the head element.
1554 * The reference count is not incremented and the reference is therefore
1555 * volatile. Use with caution.
1556 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001557static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558{
Eric Dumazet18d07002012-04-30 16:31:46 +00001559 struct sk_buff *skb = list_->next;
1560
1561 if (skb == (struct sk_buff *)list_)
1562 skb = NULL;
1563 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564}
1565
1566/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001567 * skb_peek_next - peek skb following the given one from a queue
1568 * @skb: skb to start from
1569 * @list_: list to peek at
1570 *
1571 * Returns %NULL when the end of the list is met or a pointer to the
1572 * next element. The reference count is not incremented and the
1573 * reference is therefore volatile. Use with caution.
1574 */
1575static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1576 const struct sk_buff_head *list_)
1577{
1578 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001579
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001580 if (next == (struct sk_buff *)list_)
1581 next = NULL;
1582 return next;
1583}
1584
1585/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001586 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 * @list_: list to peek at
1588 *
1589 * Peek an &sk_buff. Unlike most other operations you _MUST_
1590 * be careful with this one. A peek leaves the buffer on the
1591 * list and someone else may run off with it. You must hold
1592 * the appropriate locks or have a private queue to do this.
1593 *
1594 * Returns %NULL for an empty list or a pointer to the tail element.
1595 * The reference count is not incremented and the reference is therefore
1596 * volatile. Use with caution.
1597 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001598static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599{
Eric Dumazet18d07002012-04-30 16:31:46 +00001600 struct sk_buff *skb = list_->prev;
1601
1602 if (skb == (struct sk_buff *)list_)
1603 skb = NULL;
1604 return skb;
1605
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606}
1607
1608/**
1609 * skb_queue_len - get queue length
1610 * @list_: list to measure
1611 *
1612 * Return the length of an &sk_buff queue.
1613 */
1614static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1615{
1616 return list_->qlen;
1617}
1618
David S. Miller67fed452008-09-21 22:36:24 -07001619/**
1620 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1621 * @list: queue to initialize
1622 *
1623 * This initializes only the list and queue length aspects of
1624 * an sk_buff_head object. This allows to initialize the list
1625 * aspects of an sk_buff_head without reinitializing things like
1626 * the spinlock. It can also be used for on-stack sk_buff_head
1627 * objects where the spinlock is known to not be used.
1628 */
1629static inline void __skb_queue_head_init(struct sk_buff_head *list)
1630{
1631 list->prev = list->next = (struct sk_buff *)list;
1632 list->qlen = 0;
1633}
1634
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001635/*
1636 * This function creates a split out lock class for each invocation;
1637 * this is needed for now since a whole lot of users of the skb-queue
1638 * infrastructure in drivers have different locking usage (in hardirq)
1639 * than the networking core (in softirq only). In the long run either the
1640 * network layer or drivers should need annotation to consolidate the
1641 * main types of usage into 3 classes.
1642 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643static inline void skb_queue_head_init(struct sk_buff_head *list)
1644{
1645 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001646 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647}
1648
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001649static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1650 struct lock_class_key *class)
1651{
1652 skb_queue_head_init(list);
1653 lockdep_set_class(&list->lock, class);
1654}
1655
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001657 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 *
1659 * The "__skb_xxxx()" functions are the non-atomic ones that
1660 * can only be called with interrupts disabled.
1661 */
Joe Perches7965bd42013-09-26 14:48:15 -07001662void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1663 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001664static inline void __skb_insert(struct sk_buff *newsk,
1665 struct sk_buff *prev, struct sk_buff *next,
1666 struct sk_buff_head *list)
1667{
1668 newsk->next = next;
1669 newsk->prev = prev;
1670 next->prev = prev->next = newsk;
1671 list->qlen++;
1672}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673
David S. Miller67fed452008-09-21 22:36:24 -07001674static inline void __skb_queue_splice(const struct sk_buff_head *list,
1675 struct sk_buff *prev,
1676 struct sk_buff *next)
1677{
1678 struct sk_buff *first = list->next;
1679 struct sk_buff *last = list->prev;
1680
1681 first->prev = prev;
1682 prev->next = first;
1683
1684 last->next = next;
1685 next->prev = last;
1686}
1687
1688/**
1689 * skb_queue_splice - join two skb lists, this is designed for stacks
1690 * @list: the new list to add
1691 * @head: the place to add it in the first list
1692 */
1693static inline void skb_queue_splice(const 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 }
1700}
1701
1702/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001703 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001704 * @list: the new list to add
1705 * @head: the place to add it in the first list
1706 *
1707 * The list at @list is reinitialised
1708 */
1709static inline void skb_queue_splice_init(struct sk_buff_head *list,
1710 struct sk_buff_head *head)
1711{
1712 if (!skb_queue_empty(list)) {
1713 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001714 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001715 __skb_queue_head_init(list);
1716 }
1717}
1718
1719/**
1720 * skb_queue_splice_tail - join two skb lists, each list being a queue
1721 * @list: the new list to add
1722 * @head: the place to add it in the first list
1723 */
1724static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1725 struct sk_buff_head *head)
1726{
1727 if (!skb_queue_empty(list)) {
1728 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001729 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001730 }
1731}
1732
1733/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001734 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001735 * @list: the new list to add
1736 * @head: the place to add it in the first list
1737 *
1738 * Each of the lists is a queue.
1739 * The list at @list is reinitialised
1740 */
1741static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1742 struct sk_buff_head *head)
1743{
1744 if (!skb_queue_empty(list)) {
1745 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001746 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001747 __skb_queue_head_init(list);
1748 }
1749}
1750
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001752 * __skb_queue_after - queue a buffer at the list head
1753 * @list: list to use
1754 * @prev: place after this buffer
1755 * @newsk: buffer to queue
1756 *
1757 * Queue a buffer int the middle of a list. This function takes no locks
1758 * and you must therefore hold required locks before calling it.
1759 *
1760 * A buffer cannot be placed on two lists at the same time.
1761 */
1762static inline void __skb_queue_after(struct sk_buff_head *list,
1763 struct sk_buff *prev,
1764 struct sk_buff *newsk)
1765{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001766 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001767}
1768
Joe Perches7965bd42013-09-26 14:48:15 -07001769void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1770 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001771
Gerrit Renkerf5572852008-04-14 00:05:28 -07001772static inline void __skb_queue_before(struct sk_buff_head *list,
1773 struct sk_buff *next,
1774 struct sk_buff *newsk)
1775{
1776 __skb_insert(newsk, next->prev, next, list);
1777}
1778
Stephen Hemminger300ce172005-10-30 13:47:34 -08001779/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 * __skb_queue_head - queue a buffer at the list head
1781 * @list: list to use
1782 * @newsk: buffer to queue
1783 *
1784 * Queue a buffer at the start of a list. This function takes no locks
1785 * and you must therefore hold required locks before calling it.
1786 *
1787 * A buffer cannot be placed on two lists at the same time.
1788 */
Joe Perches7965bd42013-09-26 14:48:15 -07001789void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790static inline void __skb_queue_head(struct sk_buff_head *list,
1791 struct sk_buff *newsk)
1792{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001793 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794}
1795
1796/**
1797 * __skb_queue_tail - queue a buffer at the list tail
1798 * @list: list to use
1799 * @newsk: buffer to queue
1800 *
1801 * Queue a buffer at the end of a list. This function takes no locks
1802 * and you must therefore hold required locks before calling it.
1803 *
1804 * A buffer cannot be placed on two lists at the same time.
1805 */
Joe Perches7965bd42013-09-26 14:48:15 -07001806void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807static inline void __skb_queue_tail(struct sk_buff_head *list,
1808 struct sk_buff *newsk)
1809{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001810 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811}
1812
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 * remove sk_buff from list. _Must_ be called atomically, and with
1815 * the list known..
1816 */
Joe Perches7965bd42013-09-26 14:48:15 -07001817void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1819{
1820 struct sk_buff *next, *prev;
1821
1822 list->qlen--;
1823 next = skb->next;
1824 prev = skb->prev;
1825 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 next->prev = prev;
1827 prev->next = next;
1828}
1829
Gerrit Renkerf525c062008-04-14 00:04:12 -07001830/**
1831 * __skb_dequeue - remove from the head of the queue
1832 * @list: list to dequeue from
1833 *
1834 * Remove the head of the list. This function does not take any locks
1835 * so must be used with appropriate locks held only. The head item is
1836 * returned or %NULL if the list is empty.
1837 */
Joe Perches7965bd42013-09-26 14:48:15 -07001838struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001839static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1840{
1841 struct sk_buff *skb = skb_peek(list);
1842 if (skb)
1843 __skb_unlink(skb, list);
1844 return skb;
1845}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846
1847/**
1848 * __skb_dequeue_tail - remove from the tail of the queue
1849 * @list: list to dequeue from
1850 *
1851 * Remove the tail of the list. This function does not take any locks
1852 * so must be used with appropriate locks held only. The tail item is
1853 * returned or %NULL if the list is empty.
1854 */
Joe Perches7965bd42013-09-26 14:48:15 -07001855struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1857{
1858 struct sk_buff *skb = skb_peek_tail(list);
1859 if (skb)
1860 __skb_unlink(skb, list);
1861 return skb;
1862}
1863
1864
David S. Millerbdcc0922012-03-07 20:53:36 -05001865static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866{
1867 return skb->data_len;
1868}
1869
1870static inline unsigned int skb_headlen(const struct sk_buff *skb)
1871{
1872 return skb->len - skb->data_len;
1873}
1874
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001875static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001877 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001879 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001880 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001881 return len;
1882}
1883
1884static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1885{
1886 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887}
1888
Ian Campbell131ea662011-08-19 06:25:00 +00001889/**
1890 * __skb_fill_page_desc - initialise a paged fragment in an skb
1891 * @skb: buffer containing fragment to be initialised
1892 * @i: paged fragment index to initialise
1893 * @page: the page to use for this fragment
1894 * @off: the offset to the data with @page
1895 * @size: the length of the data
1896 *
1897 * Initialises the @i'th fragment of @skb to point to &size bytes at
1898 * offset @off within @page.
1899 *
1900 * Does not take any additional reference on the fragment.
1901 */
1902static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1903 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904{
1905 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1906
Mel Gormanc48a11c2012-07-31 16:44:23 -07001907 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001908 * Propagate page pfmemalloc to the skb if we can. The problem is
1909 * that not all callers have unique ownership of the page but rely
1910 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001911 */
Ian Campbella8605c62011-10-19 23:01:49 +00001912 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001914 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001915
1916 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001917 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001918 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001919}
1920
1921/**
1922 * skb_fill_page_desc - initialise a paged fragment in an skb
1923 * @skb: buffer containing fragment to be initialised
1924 * @i: paged fragment index to initialise
1925 * @page: the page to use for this fragment
1926 * @off: the offset to the data with @page
1927 * @size: the length of the data
1928 *
1929 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001930 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001931 * addition updates @skb such that @i is the last fragment.
1932 *
1933 * Does not take any additional reference on the fragment.
1934 */
1935static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1936 struct page *page, int off, int size)
1937{
1938 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 skb_shinfo(skb)->nr_frags = i + 1;
1940}
1941
Joe Perches7965bd42013-09-26 14:48:15 -07001942void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1943 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001944
Jason Wangf8e617e2013-11-01 14:07:47 +08001945void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1946 unsigned int truesize);
1947
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001949#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1951
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001952#ifdef NET_SKBUFF_DATA_USES_OFFSET
1953static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1954{
1955 return skb->head + skb->tail;
1956}
1957
1958static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1959{
1960 skb->tail = skb->data - skb->head;
1961}
1962
1963static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1964{
1965 skb_reset_tail_pointer(skb);
1966 skb->tail += offset;
1967}
Simon Horman7cc46192013-05-28 20:34:29 +00001968
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001969#else /* NET_SKBUFF_DATA_USES_OFFSET */
1970static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1971{
1972 return skb->tail;
1973}
1974
1975static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1976{
1977 skb->tail = skb->data;
1978}
1979
1980static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1981{
1982 skb->tail = skb->data + offset;
1983}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001984
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001985#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1986
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987/*
1988 * Add data to an sk_buff
1989 */
Johannes Berg4df864c2017-06-16 14:29:21 +02001990void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1991void *skb_put(struct sk_buff *skb, unsigned int len);
1992static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993{
Johannes Berg4df864c2017-06-16 14:29:21 +02001994 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 SKB_LINEAR_ASSERT(skb);
1996 skb->tail += len;
1997 skb->len += len;
1998 return tmp;
1999}
2000
yuan linyude77b962017-06-18 22:48:17 +08002001static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2002{
2003 void *tmp = __skb_put(skb, len);
2004
2005 memset(tmp, 0, len);
2006 return tmp;
2007}
2008
2009static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2010 unsigned int len)
2011{
2012 void *tmp = __skb_put(skb, len);
2013
2014 memcpy(tmp, data, len);
2015 return tmp;
2016}
2017
2018static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2019{
2020 *(u8 *)__skb_put(skb, 1) = val;
2021}
2022
Johannes Berg83ad3572017-06-14 22:17:20 +02002023static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002024{
Johannes Berg83ad3572017-06-14 22:17:20 +02002025 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002026
2027 memset(tmp, 0, len);
2028
2029 return tmp;
2030}
2031
Johannes Berg59ae1d12017-06-16 14:29:20 +02002032static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2033 unsigned int len)
2034{
2035 void *tmp = skb_put(skb, len);
2036
2037 memcpy(tmp, data, len);
2038
2039 return tmp;
2040}
2041
Johannes Berg634fef62017-06-16 14:29:24 +02002042static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2043{
2044 *(u8 *)skb_put(skb, 1) = val;
2045}
2046
Johannes Bergd58ff352017-06-16 14:29:23 +02002047void *skb_push(struct sk_buff *skb, unsigned int len);
2048static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049{
2050 skb->data -= len;
2051 skb->len += len;
2052 return skb->data;
2053}
2054
Johannes Bergaf728682017-06-16 14:29:22 +02002055void *skb_pull(struct sk_buff *skb, unsigned int len);
2056static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057{
2058 skb->len -= len;
2059 BUG_ON(skb->len < skb->data_len);
2060 return skb->data += len;
2061}
2062
Johannes Bergaf728682017-06-16 14:29:22 +02002063static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002064{
2065 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2066}
2067
Johannes Bergaf728682017-06-16 14:29:22 +02002068void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069
Johannes Bergaf728682017-06-16 14:29:22 +02002070static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071{
2072 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002073 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 return NULL;
2075 skb->len -= len;
2076 return skb->data += len;
2077}
2078
Johannes Bergaf728682017-06-16 14:29:22 +02002079static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080{
2081 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2082}
2083
2084static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2085{
2086 if (likely(len <= skb_headlen(skb)))
2087 return 1;
2088 if (unlikely(len > skb->len))
2089 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002090 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091}
2092
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002093void skb_condense(struct sk_buff *skb);
2094
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095/**
2096 * skb_headroom - bytes at buffer head
2097 * @skb: buffer to check
2098 *
2099 * Return the number of bytes of free space at the head of an &sk_buff.
2100 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002101static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102{
2103 return skb->data - skb->head;
2104}
2105
2106/**
2107 * skb_tailroom - bytes at buffer end
2108 * @skb: buffer to check
2109 *
2110 * Return the number of bytes of free space at the tail of an sk_buff
2111 */
2112static inline int skb_tailroom(const struct sk_buff *skb)
2113{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002114 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115}
2116
2117/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002118 * skb_availroom - bytes at buffer end
2119 * @skb: buffer to check
2120 *
2121 * Return the number of bytes of free space at the tail of an sk_buff
2122 * allocated by sk_stream_alloc()
2123 */
2124static inline int skb_availroom(const struct sk_buff *skb)
2125{
Eric Dumazet16fad692013-03-14 05:40:32 +00002126 if (skb_is_nonlinear(skb))
2127 return 0;
2128
2129 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002130}
2131
2132/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 * skb_reserve - adjust headroom
2134 * @skb: buffer to alter
2135 * @len: bytes to move
2136 *
2137 * Increase the headroom of an empty &sk_buff by reducing the tail
2138 * room. This is only allowed for an empty buffer.
2139 */
David S. Miller8243126c2006-01-17 02:54:21 -08002140static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141{
2142 skb->data += len;
2143 skb->tail += len;
2144}
2145
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002146/**
2147 * skb_tailroom_reserve - adjust reserved_tailroom
2148 * @skb: buffer to alter
2149 * @mtu: maximum amount of headlen permitted
2150 * @needed_tailroom: minimum amount of reserved_tailroom
2151 *
2152 * Set reserved_tailroom so that headlen can be as large as possible but
2153 * not larger than mtu and tailroom cannot be smaller than
2154 * needed_tailroom.
2155 * The required headroom should already have been reserved before using
2156 * this function.
2157 */
2158static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2159 unsigned int needed_tailroom)
2160{
2161 SKB_LINEAR_ASSERT(skb);
2162 if (mtu < skb_tailroom(skb) - needed_tailroom)
2163 /* use at most mtu */
2164 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2165 else
2166 /* use up to all available space */
2167 skb->reserved_tailroom = needed_tailroom;
2168}
2169
Tom Herbert8bce6d72014-09-29 20:22:29 -07002170#define ENCAP_TYPE_ETHER 0
2171#define ENCAP_TYPE_IPPROTO 1
2172
2173static inline void skb_set_inner_protocol(struct sk_buff *skb,
2174 __be16 protocol)
2175{
2176 skb->inner_protocol = protocol;
2177 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2178}
2179
2180static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2181 __u8 ipproto)
2182{
2183 skb->inner_ipproto = ipproto;
2184 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2185}
2186
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002187static inline void skb_reset_inner_headers(struct sk_buff *skb)
2188{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002189 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002190 skb->inner_network_header = skb->network_header;
2191 skb->inner_transport_header = skb->transport_header;
2192}
2193
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002194static inline void skb_reset_mac_len(struct sk_buff *skb)
2195{
2196 skb->mac_len = skb->network_header - skb->mac_header;
2197}
2198
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002199static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2200 *skb)
2201{
2202 return skb->head + skb->inner_transport_header;
2203}
2204
Tom Herbert55dc5a92015-12-14 11:19:40 -08002205static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2206{
2207 return skb_inner_transport_header(skb) - skb->data;
2208}
2209
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002210static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2211{
2212 skb->inner_transport_header = skb->data - skb->head;
2213}
2214
2215static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2216 const int offset)
2217{
2218 skb_reset_inner_transport_header(skb);
2219 skb->inner_transport_header += offset;
2220}
2221
2222static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2223{
2224 return skb->head + skb->inner_network_header;
2225}
2226
2227static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2228{
2229 skb->inner_network_header = skb->data - skb->head;
2230}
2231
2232static inline void skb_set_inner_network_header(struct sk_buff *skb,
2233 const int offset)
2234{
2235 skb_reset_inner_network_header(skb);
2236 skb->inner_network_header += offset;
2237}
2238
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002239static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2240{
2241 return skb->head + skb->inner_mac_header;
2242}
2243
2244static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2245{
2246 skb->inner_mac_header = skb->data - skb->head;
2247}
2248
2249static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2250 const int offset)
2251{
2252 skb_reset_inner_mac_header(skb);
2253 skb->inner_mac_header += offset;
2254}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002255static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2256{
Cong Wang35d04612013-05-29 15:16:05 +08002257 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002258}
2259
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002260static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2261{
2262 return skb->head + skb->transport_header;
2263}
2264
2265static inline void skb_reset_transport_header(struct sk_buff *skb)
2266{
2267 skb->transport_header = skb->data - skb->head;
2268}
2269
2270static inline void skb_set_transport_header(struct sk_buff *skb,
2271 const int offset)
2272{
2273 skb_reset_transport_header(skb);
2274 skb->transport_header += offset;
2275}
2276
2277static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2278{
2279 return skb->head + skb->network_header;
2280}
2281
2282static inline void skb_reset_network_header(struct sk_buff *skb)
2283{
2284 skb->network_header = skb->data - skb->head;
2285}
2286
2287static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2288{
2289 skb_reset_network_header(skb);
2290 skb->network_header += offset;
2291}
2292
2293static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2294{
2295 return skb->head + skb->mac_header;
2296}
2297
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002298static inline int skb_mac_offset(const struct sk_buff *skb)
2299{
2300 return skb_mac_header(skb) - skb->data;
2301}
2302
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002303static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2304{
2305 return skb->network_header - skb->mac_header;
2306}
2307
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002308static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2309{
Cong Wang35d04612013-05-29 15:16:05 +08002310 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002311}
2312
2313static inline void skb_reset_mac_header(struct sk_buff *skb)
2314{
2315 skb->mac_header = skb->data - skb->head;
2316}
2317
2318static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2319{
2320 skb_reset_mac_header(skb);
2321 skb->mac_header += offset;
2322}
2323
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002324static inline void skb_pop_mac_header(struct sk_buff *skb)
2325{
2326 skb->mac_header = skb->network_header;
2327}
2328
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002329static inline void skb_probe_transport_header(struct sk_buff *skb,
2330 const int offset_hint)
2331{
2332 struct flow_keys keys;
2333
2334 if (skb_transport_header_was_set(skb))
2335 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002336 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002337 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002338 else
2339 skb_set_transport_header(skb, offset_hint);
2340}
2341
Eric Dumazet03606892012-02-23 10:55:02 +00002342static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2343{
2344 if (skb_mac_header_was_set(skb)) {
2345 const unsigned char *old_mac = skb_mac_header(skb);
2346
2347 skb_set_mac_header(skb, -skb->mac_len);
2348 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2349 }
2350}
2351
Michał Mirosław04fb4512010-12-14 15:24:08 +00002352static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2353{
2354 return skb->csum_start - skb_headroom(skb);
2355}
2356
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002357static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2358{
2359 return skb->head + skb->csum_start;
2360}
2361
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002362static inline int skb_transport_offset(const struct sk_buff *skb)
2363{
2364 return skb_transport_header(skb) - skb->data;
2365}
2366
2367static inline u32 skb_network_header_len(const struct sk_buff *skb)
2368{
2369 return skb->transport_header - skb->network_header;
2370}
2371
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002372static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2373{
2374 return skb->inner_transport_header - skb->inner_network_header;
2375}
2376
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002377static inline int skb_network_offset(const struct sk_buff *skb)
2378{
2379 return skb_network_header(skb) - skb->data;
2380}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002381
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002382static inline int skb_inner_network_offset(const struct sk_buff *skb)
2383{
2384 return skb_inner_network_header(skb) - skb->data;
2385}
2386
Changli Gaof9599ce2010-08-04 04:43:44 +00002387static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2388{
2389 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2390}
2391
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392/*
2393 * CPUs often take a performance hit when accessing unaligned memory
2394 * locations. The actual performance hit varies, it can be small if the
2395 * hardware handles it or large if we have to take an exception and fix it
2396 * in software.
2397 *
2398 * Since an ethernet header is 14 bytes network drivers often end up with
2399 * the IP header at an unaligned offset. The IP header can be aligned by
2400 * shifting the start of the packet by 2 bytes. Drivers should do this
2401 * with:
2402 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002403 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404 *
2405 * The downside to this alignment of the IP header is that the DMA is now
2406 * unaligned. On some architectures the cost of an unaligned DMA is high
2407 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002408 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2410 * to be overridden.
2411 */
2412#ifndef NET_IP_ALIGN
2413#define NET_IP_ALIGN 2
2414#endif
2415
Anton Blanchard025be812006-03-31 02:27:06 -08002416/*
2417 * The networking layer reserves some headroom in skb data (via
2418 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2419 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002420 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002421 *
2422 * Unfortunately this headroom changes the DMA alignment of the resulting
2423 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2424 * on some architectures. An architecture can override this value,
2425 * perhaps setting it to a cacheline in size (since that will maintain
2426 * cacheline alignment of the DMA). It must be a power of 2.
2427 *
David S. Millerd6301d32009-02-08 19:24:13 -08002428 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002429 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002430 *
2431 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2432 * to reduce average number of cache lines per packet.
2433 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002434 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002435 */
2436#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002437#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002438#endif
2439
Joe Perches7965bd42013-09-26 14:48:15 -07002440int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002442static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002444 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002445 WARN_ON(1);
2446 return;
2447 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002448 skb->len = len;
2449 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450}
2451
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002452static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2453{
2454 __skb_set_length(skb, len);
2455}
2456
Joe Perches7965bd42013-09-26 14:48:15 -07002457void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458
2459static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2460{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002461 if (skb->data_len)
2462 return ___pskb_trim(skb, len);
2463 __skb_trim(skb, len);
2464 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465}
2466
2467static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2468{
2469 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2470}
2471
2472/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002473 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2474 * @skb: buffer to alter
2475 * @len: new length
2476 *
2477 * This is identical to pskb_trim except that the caller knows that
2478 * the skb is not cloned so we should never get an error due to out-
2479 * of-memory.
2480 */
2481static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2482{
2483 int err = pskb_trim(skb, len);
2484 BUG_ON(err);
2485}
2486
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002487static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2488{
2489 unsigned int diff = len - skb->len;
2490
2491 if (skb_tailroom(skb) < diff) {
2492 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2493 GFP_ATOMIC);
2494 if (ret)
2495 return ret;
2496 }
2497 __skb_set_length(skb, len);
2498 return 0;
2499}
2500
Herbert Xue9fa4f72006-08-13 20:12:58 -07002501/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002502 * skb_orphan - orphan a buffer
2503 * @skb: buffer to orphan
2504 *
2505 * If a buffer currently has an owner then we call the owner's
2506 * destructor function and make the @skb unowned. The buffer continues
2507 * to exist but is no longer charged to its former owner.
2508 */
2509static inline void skb_orphan(struct sk_buff *skb)
2510{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002511 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002512 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002513 skb->destructor = NULL;
2514 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002515 } else {
2516 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002517 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518}
2519
2520/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002521 * skb_orphan_frags - orphan the frags contained in a buffer
2522 * @skb: buffer to orphan frags from
2523 * @gfp_mask: allocation mask for replacement pages
2524 *
2525 * For each frag in the SKB which needs a destructor (i.e. has an
2526 * owner) create a copy of that frag and release the original
2527 * page by calling the destructor.
2528 */
2529static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2530{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002531 if (likely(!skb_zcopy(skb)))
2532 return 0;
2533 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2534 return 0;
2535 return skb_copy_ubufs(skb, gfp_mask);
2536}
2537
2538/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2539static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2540{
2541 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002542 return 0;
2543 return skb_copy_ubufs(skb, gfp_mask);
2544}
2545
2546/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002547 * __skb_queue_purge - empty a list
2548 * @list: list to empty
2549 *
2550 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2551 * the list and one reference dropped. This function does not take the
2552 * list lock and the caller must hold the relevant locks to use it.
2553 */
Joe Perches7965bd42013-09-26 14:48:15 -07002554void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555static inline void __skb_queue_purge(struct sk_buff_head *list)
2556{
2557 struct sk_buff *skb;
2558 while ((skb = __skb_dequeue(list)) != NULL)
2559 kfree_skb(skb);
2560}
2561
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002562void skb_rbtree_purge(struct rb_root *root);
2563
Joe Perches7965bd42013-09-26 14:48:15 -07002564void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565
Joe Perches7965bd42013-09-26 14:48:15 -07002566struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2567 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002568
2569/**
2570 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2571 * @dev: network device to receive on
2572 * @length: length to allocate
2573 *
2574 * Allocate a new &sk_buff and assign it a usage count of one. The
2575 * buffer has unspecified headroom built in. Users should allocate
2576 * the headroom they think they need without accounting for the
2577 * built in space. The built in space is used for optimisations.
2578 *
2579 * %NULL is returned if there is no free memory. Although this function
2580 * allocates memory it can be called from an interrupt.
2581 */
2582static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002583 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002584{
2585 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2586}
2587
Eric Dumazet6f532612012-05-18 05:12:12 +00002588/* legacy helper around __netdev_alloc_skb() */
2589static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2590 gfp_t gfp_mask)
2591{
2592 return __netdev_alloc_skb(NULL, length, gfp_mask);
2593}
2594
2595/* legacy helper around netdev_alloc_skb() */
2596static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2597{
2598 return netdev_alloc_skb(NULL, length);
2599}
2600
2601
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002602static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2603 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002604{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002605 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002606
2607 if (NET_IP_ALIGN && skb)
2608 skb_reserve(skb, NET_IP_ALIGN);
2609 return skb;
2610}
2611
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002612static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2613 unsigned int length)
2614{
2615 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2616}
2617
Alexander Duyck181edb22015-05-06 21:12:03 -07002618static inline void skb_free_frag(void *addr)
2619{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002620 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002621}
2622
Alexander Duyckffde7322014-12-09 19:40:42 -08002623void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002624struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2625 unsigned int length, gfp_t gfp_mask);
2626static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2627 unsigned int length)
2628{
2629 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2630}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002631void napi_consume_skb(struct sk_buff *skb, int budget);
2632
2633void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002634void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002635
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002636/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002637 * __dev_alloc_pages - allocate page for network Rx
2638 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2639 * @order: size of the allocation
2640 *
2641 * Allocate a new page.
2642 *
2643 * %NULL is returned if there is no free memory.
2644*/
2645static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2646 unsigned int order)
2647{
2648 /* This piece of code contains several assumptions.
2649 * 1. This is for device Rx, therefor a cold page is preferred.
2650 * 2. The expectation is the user wants a compound page.
2651 * 3. If requesting a order 0 page it will not be compound
2652 * due to the check to see if order has a value in prep_new_page
2653 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2654 * code in gfp_to_alloc_flags that should be enforcing this.
2655 */
2656 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2657
2658 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2659}
2660
2661static inline struct page *dev_alloc_pages(unsigned int order)
2662{
Neil Horman95829b32016-05-19 11:30:54 -04002663 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002664}
2665
2666/**
2667 * __dev_alloc_page - allocate a page for network Rx
2668 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2669 *
2670 * Allocate a new page.
2671 *
2672 * %NULL is returned if there is no free memory.
2673 */
2674static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2675{
2676 return __dev_alloc_pages(gfp_mask, 0);
2677}
2678
2679static inline struct page *dev_alloc_page(void)
2680{
Neil Horman95829b32016-05-19 11:30:54 -04002681 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002682}
2683
2684/**
Mel Gorman06140022012-07-31 16:44:24 -07002685 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2686 * @page: The page that was allocated from skb_alloc_page
2687 * @skb: The skb that may need pfmemalloc set
2688 */
2689static inline void skb_propagate_pfmemalloc(struct page *page,
2690 struct sk_buff *skb)
2691{
Michal Hocko2f064f32015-08-21 14:11:51 -07002692 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002693 skb->pfmemalloc = true;
2694}
2695
Eric Dumazet564824b2010-10-11 19:05:25 +00002696/**
Masanari Iidae2278672014-02-18 22:54:36 +09002697 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002698 * @frag: the paged fragment
2699 *
2700 * Returns the &struct page associated with @frag.
2701 */
2702static inline struct page *skb_frag_page(const skb_frag_t *frag)
2703{
Ian Campbella8605c62011-10-19 23:01:49 +00002704 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002705}
2706
2707/**
2708 * __skb_frag_ref - take an addition reference on a paged fragment.
2709 * @frag: the paged fragment
2710 *
2711 * Takes an additional reference on the paged fragment @frag.
2712 */
2713static inline void __skb_frag_ref(skb_frag_t *frag)
2714{
2715 get_page(skb_frag_page(frag));
2716}
2717
2718/**
2719 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2720 * @skb: the buffer
2721 * @f: the fragment offset.
2722 *
2723 * Takes an additional reference on the @f'th paged fragment of @skb.
2724 */
2725static inline void skb_frag_ref(struct sk_buff *skb, int f)
2726{
2727 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2728}
2729
2730/**
2731 * __skb_frag_unref - release a reference on a paged fragment.
2732 * @frag: the paged fragment
2733 *
2734 * Releases a reference on the paged fragment @frag.
2735 */
2736static inline void __skb_frag_unref(skb_frag_t *frag)
2737{
2738 put_page(skb_frag_page(frag));
2739}
2740
2741/**
2742 * skb_frag_unref - release a reference on a paged fragment of an skb.
2743 * @skb: the buffer
2744 * @f: the fragment offset
2745 *
2746 * Releases a reference on the @f'th paged fragment of @skb.
2747 */
2748static inline void skb_frag_unref(struct sk_buff *skb, int f)
2749{
2750 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2751}
2752
2753/**
2754 * skb_frag_address - gets the address of the data contained in a paged fragment
2755 * @frag: the paged fragment buffer
2756 *
2757 * Returns the address of the data within @frag. The page must already
2758 * be mapped.
2759 */
2760static inline void *skb_frag_address(const skb_frag_t *frag)
2761{
2762 return page_address(skb_frag_page(frag)) + frag->page_offset;
2763}
2764
2765/**
2766 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2767 * @frag: the paged fragment buffer
2768 *
2769 * Returns the address of the data within @frag. Checks that the page
2770 * is mapped and returns %NULL otherwise.
2771 */
2772static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2773{
2774 void *ptr = page_address(skb_frag_page(frag));
2775 if (unlikely(!ptr))
2776 return NULL;
2777
2778 return ptr + frag->page_offset;
2779}
2780
2781/**
2782 * __skb_frag_set_page - sets the page contained in a paged fragment
2783 * @frag: the paged fragment
2784 * @page: the page to set
2785 *
2786 * Sets the fragment @frag to contain @page.
2787 */
2788static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2789{
Ian Campbella8605c62011-10-19 23:01:49 +00002790 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002791}
2792
2793/**
2794 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2795 * @skb: the buffer
2796 * @f: the fragment offset
2797 * @page: the page to set
2798 *
2799 * Sets the @f'th fragment of @skb to contain @page.
2800 */
2801static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2802 struct page *page)
2803{
2804 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2805}
2806
Eric Dumazet400dfd32013-10-17 16:27:07 -07002807bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2808
Ian Campbell131ea662011-08-19 06:25:00 +00002809/**
2810 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002811 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002812 * @frag: the paged fragment to map
2813 * @offset: the offset within the fragment (starting at the
2814 * fragment's own offset)
2815 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002816 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002817 *
2818 * Maps the page associated with @frag to @device.
2819 */
2820static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2821 const skb_frag_t *frag,
2822 size_t offset, size_t size,
2823 enum dma_data_direction dir)
2824{
2825 return dma_map_page(dev, skb_frag_page(frag),
2826 frag->page_offset + offset, size, dir);
2827}
2828
Eric Dumazet117632e2011-12-03 21:39:53 +00002829static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2830 gfp_t gfp_mask)
2831{
2832 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2833}
2834
Octavian Purdilabad93e92014-06-12 01:36:26 +03002835
2836static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2837 gfp_t gfp_mask)
2838{
2839 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2840}
2841
2842
Ian Campbell131ea662011-08-19 06:25:00 +00002843/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002844 * skb_clone_writable - is the header of a clone writable
2845 * @skb: buffer to check
2846 * @len: length up to which to write
2847 *
2848 * Returns true if modifying the header part of the cloned buffer
2849 * does not requires the data to be copied.
2850 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002851static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002852{
2853 return !skb_header_cloned(skb) &&
2854 skb_headroom(skb) + len <= skb->hdr_len;
2855}
2856
Daniel Borkmann36976492016-02-19 23:05:25 +01002857static inline int skb_try_make_writable(struct sk_buff *skb,
2858 unsigned int write_len)
2859{
2860 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2861 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2862}
2863
Herbert Xud9cc2042007-09-16 16:21:16 -07002864static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2865 int cloned)
2866{
2867 int delta = 0;
2868
Herbert Xud9cc2042007-09-16 16:21:16 -07002869 if (headroom > skb_headroom(skb))
2870 delta = headroom - skb_headroom(skb);
2871
2872 if (delta || cloned)
2873 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2874 GFP_ATOMIC);
2875 return 0;
2876}
2877
Patrick McHardy334a8132007-06-25 04:35:20 -07002878/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002879 * skb_cow - copy header of skb when it is required
2880 * @skb: buffer to cow
2881 * @headroom: needed headroom
2882 *
2883 * If the skb passed lacks sufficient headroom or its data part
2884 * is shared, data is reallocated. If reallocation fails, an error
2885 * is returned and original skb is not changed.
2886 *
2887 * The result is skb with writable area skb->head...skb->tail
2888 * and at least @headroom of space at head.
2889 */
2890static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2891{
Herbert Xud9cc2042007-09-16 16:21:16 -07002892 return __skb_cow(skb, headroom, skb_cloned(skb));
2893}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894
Herbert Xud9cc2042007-09-16 16:21:16 -07002895/**
2896 * skb_cow_head - skb_cow but only making the head writable
2897 * @skb: buffer to cow
2898 * @headroom: needed headroom
2899 *
2900 * This function is identical to skb_cow except that we replace the
2901 * skb_cloned check by skb_header_cloned. It should be used when
2902 * you only need to push on some header and do not need to modify
2903 * the data.
2904 */
2905static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2906{
2907 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908}
2909
2910/**
2911 * skb_padto - pad an skbuff up to a minimal size
2912 * @skb: buffer to pad
2913 * @len: minimal length
2914 *
2915 * Pads up a buffer to ensure the trailing bytes exist and are
2916 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002917 * is untouched. Otherwise it is extended. Returns zero on
2918 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002919 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002920static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002921{
2922 unsigned int size = skb->len;
2923 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002924 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002925 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002926}
2927
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002928/**
2929 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2930 * @skb: buffer to pad
2931 * @len: minimal length
2932 *
2933 * Pads up a buffer to ensure the trailing bytes exist and are
2934 * blanked. If the buffer already contains sufficient data it
2935 * is untouched. Otherwise it is extended. Returns zero on
2936 * success. The skb is freed on error.
2937 */
2938static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2939{
2940 unsigned int size = skb->len;
2941
2942 if (unlikely(size < len)) {
2943 len -= size;
2944 if (skb_pad(skb, len))
2945 return -ENOMEM;
2946 __skb_put(skb, len);
2947 }
2948 return 0;
2949}
2950
Linus Torvalds1da177e2005-04-16 15:20:36 -07002951static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002952 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002953{
2954 const int off = skb->len;
2955
2956 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002957 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002958 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2959 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002960 skb->csum = csum_block_add(skb->csum, csum, off);
2961 return 0;
2962 }
Al Viro15e6cb42016-11-01 22:42:45 -04002963 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002964 return 0;
2965
2966 __skb_trim(skb, off);
2967 return -EFAULT;
2968}
2969
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002970static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2971 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002972{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002973 if (skb_zcopy(skb))
2974 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002975 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002976 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002977
Ian Campbellea2ab692011-08-22 23:44:58 +00002978 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002979 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002981 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002982}
2983
Herbert Xu364c6ba2006-06-09 16:10:40 -07002984static inline int __skb_linearize(struct sk_buff *skb)
2985{
2986 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2987}
2988
Linus Torvalds1da177e2005-04-16 15:20:36 -07002989/**
2990 * skb_linearize - convert paged skb to linear one
2991 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002992 *
2993 * If there is no free memory -ENOMEM is returned, otherwise zero
2994 * is returned and the old skb data released.
2995 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002996static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002997{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002998 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2999}
3000
3001/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003002 * skb_has_shared_frag - can any frag be overwritten
3003 * @skb: buffer to test
3004 *
3005 * Return true if the skb has at least one frag that might be modified
3006 * by an external entity (as in vmsplice()/sendfile())
3007 */
3008static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3009{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003010 return skb_is_nonlinear(skb) &&
3011 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003012}
3013
3014/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003015 * skb_linearize_cow - make sure skb is linear and writable
3016 * @skb: buffer to process
3017 *
3018 * If there is no free memory -ENOMEM is returned, otherwise zero
3019 * is returned and the old skb data released.
3020 */
3021static inline int skb_linearize_cow(struct sk_buff *skb)
3022{
3023 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3024 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003025}
3026
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003027static __always_inline void
3028__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3029 unsigned int off)
3030{
3031 if (skb->ip_summed == CHECKSUM_COMPLETE)
3032 skb->csum = csum_block_sub(skb->csum,
3033 csum_partial(start, len, 0), off);
3034 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3035 skb_checksum_start_offset(skb) < 0)
3036 skb->ip_summed = CHECKSUM_NONE;
3037}
3038
Linus Torvalds1da177e2005-04-16 15:20:36 -07003039/**
3040 * skb_postpull_rcsum - update checksum for received skb after pull
3041 * @skb: buffer to update
3042 * @start: start of data before pull
3043 * @len: length of data pulled
3044 *
3045 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003046 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3047 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003048 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003050 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003052 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003053}
3054
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003055static __always_inline void
3056__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3057 unsigned int off)
3058{
3059 if (skb->ip_summed == CHECKSUM_COMPLETE)
3060 skb->csum = csum_block_add(skb->csum,
3061 csum_partial(start, len, 0), off);
3062}
Herbert Xucbb042f2006-03-20 22:43:56 -08003063
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003064/**
3065 * skb_postpush_rcsum - update checksum for received skb after push
3066 * @skb: buffer to update
3067 * @start: start of data after push
3068 * @len: length of data pushed
3069 *
3070 * After doing a push on a received packet, you need to call this to
3071 * update the CHECKSUM_COMPLETE checksum.
3072 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003073static inline void skb_postpush_rcsum(struct sk_buff *skb,
3074 const void *start, unsigned int len)
3075{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003076 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003077}
3078
Johannes Bergaf728682017-06-16 14:29:22 +02003079void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003080
David S. Miller7ce5a272013-12-02 17:26:05 -05003081/**
WANG Cong82a31b92016-06-30 10:15:22 -07003082 * skb_push_rcsum - push skb and update receive checksum
3083 * @skb: buffer to update
3084 * @len: length of data pulled
3085 *
3086 * This function performs an skb_push on the packet and updates
3087 * the CHECKSUM_COMPLETE checksum. It should be used on
3088 * receive path processing instead of skb_push unless you know
3089 * that the checksum difference is zero (e.g., a valid IP header)
3090 * or you are setting ip_summed to CHECKSUM_NONE.
3091 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003092static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003093{
3094 skb_push(skb, len);
3095 skb_postpush_rcsum(skb, skb->data, len);
3096 return skb->data;
3097}
3098
3099/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003100 * pskb_trim_rcsum - trim received skb and update checksum
3101 * @skb: buffer to trim
3102 * @len: new length
3103 *
3104 * This is exactly the same as pskb_trim except that it ensures the
3105 * checksum of received packets are still valid after the operation.
3106 */
3107
3108static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3109{
3110 if (likely(len >= skb->len))
3111 return 0;
3112 if (skb->ip_summed == CHECKSUM_COMPLETE)
3113 skb->ip_summed = CHECKSUM_NONE;
3114 return __pskb_trim(skb, len);
3115}
3116
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003117static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3118{
3119 if (skb->ip_summed == CHECKSUM_COMPLETE)
3120 skb->ip_summed = CHECKSUM_NONE;
3121 __skb_trim(skb, len);
3122 return 0;
3123}
3124
3125static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3126{
3127 if (skb->ip_summed == CHECKSUM_COMPLETE)
3128 skb->ip_summed = CHECKSUM_NONE;
3129 return __skb_grow(skb, len);
3130}
3131
Linus Torvalds1da177e2005-04-16 15:20:36 -07003132#define skb_queue_walk(queue, skb) \
3133 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003134 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003135 skb = skb->next)
3136
James Chapman46f89142007-04-30 00:07:31 -07003137#define skb_queue_walk_safe(queue, skb, tmp) \
3138 for (skb = (queue)->next, tmp = skb->next; \
3139 skb != (struct sk_buff *)(queue); \
3140 skb = tmp, tmp = skb->next)
3141
David S. Miller1164f522008-09-23 00:49:44 -07003142#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003143 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003144 skb = skb->next)
3145
3146#define skb_queue_walk_from_safe(queue, skb, tmp) \
3147 for (tmp = skb->next; \
3148 skb != (struct sk_buff *)(queue); \
3149 skb = tmp, tmp = skb->next)
3150
Stephen Hemminger300ce172005-10-30 13:47:34 -08003151#define skb_queue_reverse_walk(queue, skb) \
3152 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003153 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003154 skb = skb->prev)
3155
David S. Miller686a2952011-01-20 22:47:32 -08003156#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3157 for (skb = (queue)->prev, tmp = skb->prev; \
3158 skb != (struct sk_buff *)(queue); \
3159 skb = tmp, tmp = skb->prev)
3160
3161#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3162 for (tmp = skb->prev; \
3163 skb != (struct sk_buff *)(queue); \
3164 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003165
David S. Miller21dc3302010-08-23 00:13:46 -07003166static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003167{
3168 return skb_shinfo(skb)->frag_list != NULL;
3169}
3170
3171static inline void skb_frag_list_init(struct sk_buff *skb)
3172{
3173 skb_shinfo(skb)->frag_list = NULL;
3174}
3175
David S. Milleree039872009-06-09 00:17:13 -07003176#define skb_walk_frags(skb, iter) \
3177 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3178
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003179
3180int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3181 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003182struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3183 struct sk_buff_head *queue,
3184 unsigned int flags,
3185 void (*destructor)(struct sock *sk,
3186 struct sk_buff *skb),
3187 int *peeked, int *off, int *err,
3188 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003189struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003190 void (*destructor)(struct sock *sk,
3191 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003192 int *peeked, int *off, int *err,
3193 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003194struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003195 void (*destructor)(struct sock *sk,
3196 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003197 int *peeked, int *off, int *err);
3198struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3199 int *err);
3200unsigned int datagram_poll(struct file *file, struct socket *sock,
3201 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003202int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3203 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003204static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3205 struct msghdr *msg, int size)
3206{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003207 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003208}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003209int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3210 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003211int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3212 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003213int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003214void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003215void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3216static inline void skb_free_datagram_locked(struct sock *sk,
3217 struct sk_buff *skb)
3218{
3219 __skb_free_datagram_locked(sk, skb, 0);
3220}
Joe Perches7965bd42013-09-26 14:48:15 -07003221int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003222int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3223int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3224__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3225 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003226int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003227 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003228 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003229int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3230 int len);
3231int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003232void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003233unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003234int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3235 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003236void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3237int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3238void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003239unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003240bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003241struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003242struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003243int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003244int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003245int skb_vlan_pop(struct sk_buff *skb);
3246int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003247struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3248 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003249
Al Viro6ce8e9c2014-04-06 21:25:44 -04003250static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3251{
Al Viro3073f072017-02-17 23:13:25 -05003252 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003253}
3254
Al Viro7eab8d92014-04-06 21:51:23 -04003255static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3256{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003257 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003258}
3259
Daniel Borkmann2817a332013-10-30 11:50:51 +01003260struct skb_checksum_ops {
3261 __wsum (*update)(const void *mem, int len, __wsum wsum);
3262 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3263};
3264
Davide Caratti96178132017-05-18 15:44:37 +02003265extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3266
Daniel Borkmann2817a332013-10-30 11:50:51 +01003267__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3268 __wsum csum, const struct skb_checksum_ops *ops);
3269__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3270 __wsum csum);
3271
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003272static inline void * __must_check
3273__skb_header_pointer(const struct sk_buff *skb, int offset,
3274 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003275{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003276 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003277 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003278
David S. Miller690e36e2014-08-23 12:13:41 -07003279 if (!skb ||
3280 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003281 return NULL;
3282
3283 return buffer;
3284}
3285
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003286static inline void * __must_check
3287skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003288{
3289 return __skb_header_pointer(skb, offset, len, skb->data,
3290 skb_headlen(skb), buffer);
3291}
3292
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003293/**
3294 * skb_needs_linearize - check if we need to linearize a given skb
3295 * depending on the given device features.
3296 * @skb: socket buffer to check
3297 * @features: net device features
3298 *
3299 * Returns true if either:
3300 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3301 * 2. skb is fragmented and the device does not support SG.
3302 */
3303static inline bool skb_needs_linearize(struct sk_buff *skb,
3304 netdev_features_t features)
3305{
3306 return skb_is_nonlinear(skb) &&
3307 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3308 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3309}
3310
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003311static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3312 void *to,
3313 const unsigned int len)
3314{
3315 memcpy(to, skb->data, len);
3316}
3317
3318static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3319 const int offset, void *to,
3320 const unsigned int len)
3321{
3322 memcpy(to, skb->data + offset, len);
3323}
3324
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003325static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3326 const void *from,
3327 const unsigned int len)
3328{
3329 memcpy(skb->data, from, len);
3330}
3331
3332static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3333 const int offset,
3334 const void *from,
3335 const unsigned int len)
3336{
3337 memcpy(skb->data + offset, from, len);
3338}
3339
Joe Perches7965bd42013-09-26 14:48:15 -07003340void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003341
Patrick Ohlyac45f602009-02-12 05:03:37 +00003342static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3343{
3344 return skb->tstamp;
3345}
3346
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003347/**
3348 * skb_get_timestamp - get timestamp from a skb
3349 * @skb: skb to get stamp from
3350 * @stamp: pointer to struct timeval to store stamp in
3351 *
3352 * Timestamps are stored in the skb as offsets to a base timestamp.
3353 * This function converts the offset back to a struct timeval and stores
3354 * it in stamp.
3355 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003356static inline void skb_get_timestamp(const struct sk_buff *skb,
3357 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003358{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003359 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003360}
3361
Patrick Ohlyac45f602009-02-12 05:03:37 +00003362static inline void skb_get_timestampns(const struct sk_buff *skb,
3363 struct timespec *stamp)
3364{
3365 *stamp = ktime_to_timespec(skb->tstamp);
3366}
3367
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003368static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003369{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003370 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003371}
3372
Stephen Hemminger164891a2007-04-23 22:26:16 -07003373static inline ktime_t net_timedelta(ktime_t t)
3374{
3375 return ktime_sub(ktime_get_real(), t);
3376}
3377
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003378static inline ktime_t net_invalid_timestamp(void)
3379{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003380 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003381}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003382
Alexander Duyck62bccb82014-09-04 13:31:35 -04003383struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3384
Richard Cochranc1f19b52010-07-17 08:49:36 +00003385#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3386
Joe Perches7965bd42013-09-26 14:48:15 -07003387void skb_clone_tx_timestamp(struct sk_buff *skb);
3388bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003389
3390#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3391
3392static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3393{
3394}
3395
3396static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3397{
3398 return false;
3399}
3400
3401#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3402
3403/**
3404 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3405 *
Richard Cochranda92b192011-10-21 00:49:15 +00003406 * PHY drivers may accept clones of transmitted packets for
3407 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003408 * must call this function to return the skb back to the stack with a
3409 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003410 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003411 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003412 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003413 *
3414 */
3415void skb_complete_tx_timestamp(struct sk_buff *skb,
3416 struct skb_shared_hwtstamps *hwtstamps);
3417
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003418void __skb_tstamp_tx(struct sk_buff *orig_skb,
3419 struct skb_shared_hwtstamps *hwtstamps,
3420 struct sock *sk, int tstype);
3421
Patrick Ohlyac45f602009-02-12 05:03:37 +00003422/**
3423 * skb_tstamp_tx - queue clone of skb with send time stamps
3424 * @orig_skb: the original outgoing packet
3425 * @hwtstamps: hardware time stamps, may be NULL if not available
3426 *
3427 * If the skb has a socket associated, then this function clones the
3428 * skb (thus sharing the actual data and optional structures), stores
3429 * the optional hardware time stamping information (if non NULL) or
3430 * generates a software time stamp (otherwise), then queues the clone
3431 * to the error queue of the socket. Errors are silently ignored.
3432 */
Joe Perches7965bd42013-09-26 14:48:15 -07003433void skb_tstamp_tx(struct sk_buff *orig_skb,
3434 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003435
Richard Cochran4507a712010-07-17 08:48:28 +00003436/**
3437 * skb_tx_timestamp() - Driver hook for transmit timestamping
3438 *
3439 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003440 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003441 *
David S. Miller73409f32013-12-27 13:04:33 -05003442 * Specifically, one should make absolutely sure that this function is
3443 * called before TX completion of this packet can trigger. Otherwise
3444 * the packet could potentially already be freed.
3445 *
Richard Cochran4507a712010-07-17 08:48:28 +00003446 * @skb: A socket buffer.
3447 */
3448static inline void skb_tx_timestamp(struct sk_buff *skb)
3449{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003450 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003451 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3452 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003453}
3454
Johannes Berg6e3e9392011-11-09 10:15:42 +01003455/**
3456 * skb_complete_wifi_ack - deliver skb with wifi status
3457 *
3458 * @skb: the original outgoing packet
3459 * @acked: ack status
3460 *
3461 */
3462void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3463
Joe Perches7965bd42013-09-26 14:48:15 -07003464__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3465__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003466
Herbert Xu60476372007-04-09 11:59:39 -07003467static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3468{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003469 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3470 skb->csum_valid ||
3471 (skb->ip_summed == CHECKSUM_PARTIAL &&
3472 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003473}
3474
Herbert Xufb286bb2005-11-10 13:01:24 -08003475/**
3476 * skb_checksum_complete - Calculate checksum of an entire packet
3477 * @skb: packet to process
3478 *
3479 * This function calculates the checksum over the entire packet plus
3480 * the value of skb->csum. The latter can be used to supply the
3481 * checksum of a pseudo header as used by TCP/UDP. It returns the
3482 * checksum.
3483 *
3484 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3485 * this function can be used to verify that checksum on received
3486 * packets. In that case the function should return zero if the
3487 * checksum is correct. In particular, this function will return zero
3488 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3489 * hardware has already verified the correctness of the checksum.
3490 */
Al Viro4381ca32007-07-15 21:00:11 +01003491static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003492{
Herbert Xu60476372007-04-09 11:59:39 -07003493 return skb_csum_unnecessary(skb) ?
3494 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003495}
3496
Tom Herbert77cffe22014-08-27 21:26:46 -07003497static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3498{
3499 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3500 if (skb->csum_level == 0)
3501 skb->ip_summed = CHECKSUM_NONE;
3502 else
3503 skb->csum_level--;
3504 }
3505}
3506
3507static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3508{
3509 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3510 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3511 skb->csum_level++;
3512 } else if (skb->ip_summed == CHECKSUM_NONE) {
3513 skb->ip_summed = CHECKSUM_UNNECESSARY;
3514 skb->csum_level = 0;
3515 }
3516}
3517
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003518/* Check if we need to perform checksum complete validation.
3519 *
3520 * Returns true if checksum complete is needed, false otherwise
3521 * (either checksum is unnecessary or zero checksum is allowed).
3522 */
3523static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3524 bool zero_okay,
3525 __sum16 check)
3526{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003527 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3528 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003529 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003530 return false;
3531 }
3532
3533 return true;
3534}
3535
3536/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3537 * in checksum_init.
3538 */
3539#define CHECKSUM_BREAK 76
3540
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003541/* Unset checksum-complete
3542 *
3543 * Unset checksum complete can be done when packet is being modified
3544 * (uncompressed for instance) and checksum-complete value is
3545 * invalidated.
3546 */
3547static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3548{
3549 if (skb->ip_summed == CHECKSUM_COMPLETE)
3550 skb->ip_summed = CHECKSUM_NONE;
3551}
3552
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003553/* Validate (init) checksum based on checksum complete.
3554 *
3555 * Return values:
3556 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3557 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3558 * checksum is stored in skb->csum for use in __skb_checksum_complete
3559 * non-zero: value of invalid checksum
3560 *
3561 */
3562static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3563 bool complete,
3564 __wsum psum)
3565{
3566 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3567 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003568 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003569 return 0;
3570 }
3571 }
3572
3573 skb->csum = psum;
3574
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003575 if (complete || skb->len <= CHECKSUM_BREAK) {
3576 __sum16 csum;
3577
3578 csum = __skb_checksum_complete(skb);
3579 skb->csum_valid = !csum;
3580 return csum;
3581 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003582
3583 return 0;
3584}
3585
3586static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3587{
3588 return 0;
3589}
3590
3591/* Perform checksum validate (init). Note that this is a macro since we only
3592 * want to calculate the pseudo header which is an input function if necessary.
3593 * First we try to validate without any computation (checksum unnecessary) and
3594 * then calculate based on checksum complete calling the function to compute
3595 * pseudo header.
3596 *
3597 * Return values:
3598 * 0: checksum is validated or try to in skb_checksum_complete
3599 * non-zero: value of invalid checksum
3600 */
3601#define __skb_checksum_validate(skb, proto, complete, \
3602 zero_okay, check, compute_pseudo) \
3603({ \
3604 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003605 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003606 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3607 __ret = __skb_checksum_validate_complete(skb, \
3608 complete, compute_pseudo(skb, proto)); \
3609 __ret; \
3610})
3611
3612#define skb_checksum_init(skb, proto, compute_pseudo) \
3613 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3614
3615#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3616 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3617
3618#define skb_checksum_validate(skb, proto, compute_pseudo) \
3619 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3620
3621#define skb_checksum_validate_zero_check(skb, proto, check, \
3622 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003623 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003624
3625#define skb_checksum_simple_validate(skb) \
3626 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3627
Tom Herbertd96535a2014-08-31 15:12:42 -07003628static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3629{
Davide Caratti219f1d792017-05-18 15:44:39 +02003630 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003631}
3632
3633static inline void __skb_checksum_convert(struct sk_buff *skb,
3634 __sum16 check, __wsum pseudo)
3635{
3636 skb->csum = ~pseudo;
3637 skb->ip_summed = CHECKSUM_COMPLETE;
3638}
3639
3640#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3641do { \
3642 if (__skb_checksum_convert_check(skb)) \
3643 __skb_checksum_convert(skb, check, \
3644 compute_pseudo(skb, proto)); \
3645} while (0)
3646
Tom Herbert15e23962015-02-10 16:30:31 -08003647static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3648 u16 start, u16 offset)
3649{
3650 skb->ip_summed = CHECKSUM_PARTIAL;
3651 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3652 skb->csum_offset = offset - start;
3653}
3654
Tom Herbertdcdc8992015-02-02 16:07:34 -08003655/* Update skbuf and packet to reflect the remote checksum offload operation.
3656 * When called, ptr indicates the starting point for skb->csum when
3657 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3658 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3659 */
3660static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003661 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003662{
3663 __wsum delta;
3664
Tom Herbert15e23962015-02-10 16:30:31 -08003665 if (!nopartial) {
3666 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3667 return;
3668 }
3669
Tom Herbertdcdc8992015-02-02 16:07:34 -08003670 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3671 __skb_checksum_complete(skb);
3672 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3673 }
3674
3675 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3676
3677 /* Adjust skb->csum since we changed the packet */
3678 skb->csum = csum_add(skb->csum, delta);
3679}
3680
Florian Westphalcb9c6832017-01-23 18:21:56 +01003681static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3682{
3683#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003684 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003685#else
3686 return NULL;
3687#endif
3688}
3689
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003690#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003691void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003692static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3693{
3694 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003695 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003696}
3697static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3698{
3699 if (nfct)
3700 atomic_inc(&nfct->use);
3701}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003702#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003703#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003704static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3705{
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003706 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003707 kfree(nf_bridge);
3708}
3709static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3710{
3711 if (nf_bridge)
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003712 refcount_inc(&nf_bridge->use);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003713}
3714#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003715static inline void nf_reset(struct sk_buff *skb)
3716{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003717#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003718 nf_conntrack_put(skb_nfct(skb));
3719 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003720#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003721#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003722 nf_bridge_put(skb->nf_bridge);
3723 skb->nf_bridge = NULL;
3724#endif
3725}
3726
Patrick McHardy124dff02013-04-05 20:42:05 +02003727static inline void nf_reset_trace(struct sk_buff *skb)
3728{
Florian Westphal478b3602014-02-15 23:48:45 +01003729#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003730 skb->nf_trace = 0;
3731#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003732}
3733
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003734/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003735static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3736 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003737{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003738#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003739 dst->_nfct = src->_nfct;
3740 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003741#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003742#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003743 dst->nf_bridge = src->nf_bridge;
3744 nf_bridge_get(src->nf_bridge);
3745#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003746#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003747 if (copy)
3748 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003749#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003750}
3751
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003752static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3753{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003754#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003755 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003756#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003757#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003758 nf_bridge_put(dst->nf_bridge);
3759#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003760 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003761}
3762
James Morris984bc162006-06-09 00:29:17 -07003763#ifdef CONFIG_NETWORK_SECMARK
3764static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3765{
3766 to->secmark = from->secmark;
3767}
3768
3769static inline void skb_init_secmark(struct sk_buff *skb)
3770{
3771 skb->secmark = 0;
3772}
3773#else
3774static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3775{ }
3776
3777static inline void skb_init_secmark(struct sk_buff *skb)
3778{ }
3779#endif
3780
Eric W. Biederman574f7192014-04-01 12:20:24 -07003781static inline bool skb_irq_freeable(const struct sk_buff *skb)
3782{
3783 return !skb->destructor &&
3784#if IS_ENABLED(CONFIG_XFRM)
3785 !skb->sp &&
3786#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003787 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003788 !skb->_skb_refdst &&
3789 !skb_has_frag_list(skb);
3790}
3791
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003792static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3793{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003794 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003795}
3796
Stephen Hemminger92477442009-03-21 13:39:26 -07003797static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003798{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003799 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003800}
3801
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003802static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3803{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003804 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003805}
3806
David S. Millerd5a9e242009-01-27 16:22:11 -08003807static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3808{
3809 skb->queue_mapping = rx_queue + 1;
3810}
3811
Stephen Hemminger92477442009-03-21 13:39:26 -07003812static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003813{
3814 return skb->queue_mapping - 1;
3815}
3816
Stephen Hemminger92477442009-03-21 13:39:26 -07003817static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003818{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003819 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003820}
3821
Julian Anastasov4ff06202017-02-06 23:14:12 +02003822static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3823{
3824 skb->dst_pending_confirm = val;
3825}
3826
3827static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3828{
3829 return skb->dst_pending_confirm != 0;
3830}
3831
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003832static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3833{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003834#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003835 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003836#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003837 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003838#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003839}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003840
Pravin B Shelar68c33162013-02-14 14:02:41 +00003841/* Keeps track of mac header offset relative to skb->head.
3842 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3843 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003844 * tunnel skb it points to outer mac header.
3845 * Keeps track of level of encapsulation of network headers.
3846 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003847struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003848 union {
3849 int mac_offset;
3850 int data_offset;
3851 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003852 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003853 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003854 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003855};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003856#define SKB_SGO_CB_OFFSET 32
3857#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003858
3859static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3860{
3861 return (skb_mac_header(inner_skb) - inner_skb->head) -
3862 SKB_GSO_CB(inner_skb)->mac_offset;
3863}
3864
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003865static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3866{
3867 int new_headroom, headroom;
3868 int ret;
3869
3870 headroom = skb_headroom(skb);
3871 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3872 if (ret)
3873 return ret;
3874
3875 new_headroom = skb_headroom(skb);
3876 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3877 return 0;
3878}
3879
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003880static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3881{
3882 /* Do not update partial checksums if remote checksum is enabled. */
3883 if (skb->remcsum_offload)
3884 return;
3885
3886 SKB_GSO_CB(skb)->csum = res;
3887 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3888}
3889
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003890/* Compute the checksum for a gso segment. First compute the checksum value
3891 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3892 * then add in skb->csum (checksum from csum_start to end of packet).
3893 * skb->csum and csum_start are then updated to reflect the checksum of the
3894 * resultant packet starting from the transport header-- the resultant checksum
3895 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3896 * header.
3897 */
3898static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3899{
Alexander Duyck76443452016-02-05 15:27:37 -08003900 unsigned char *csum_start = skb_transport_header(skb);
3901 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3902 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003903
Alexander Duyck76443452016-02-05 15:27:37 -08003904 SKB_GSO_CB(skb)->csum = res;
3905 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003906
Alexander Duyck76443452016-02-05 15:27:37 -08003907 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003908}
3909
David S. Millerbdcc0922012-03-07 20:53:36 -05003910static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003911{
3912 return skb_shinfo(skb)->gso_size;
3913}
3914
Eric Dumazet36a8f392013-09-29 01:21:32 -07003915/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003916static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003917{
3918 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3919}
3920
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003921static inline void skb_gso_reset(struct sk_buff *skb)
3922{
3923 skb_shinfo(skb)->gso_size = 0;
3924 skb_shinfo(skb)->gso_segs = 0;
3925 skb_shinfo(skb)->gso_type = 0;
3926}
3927
Joe Perches7965bd42013-09-26 14:48:15 -07003928void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003929
3930static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3931{
3932 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3933 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003934 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3935
Alexander Duyckb78462e2010-06-02 12:24:37 +00003936 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3937 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003938 __skb_warn_lro_forwarding(skb);
3939 return true;
3940 }
3941 return false;
3942}
3943
Herbert Xu35fc92a2007-03-26 23:22:20 -07003944static inline void skb_forward_csum(struct sk_buff *skb)
3945{
3946 /* Unfortunately we don't support this one. Any brave souls? */
3947 if (skb->ip_summed == CHECKSUM_COMPLETE)
3948 skb->ip_summed = CHECKSUM_NONE;
3949}
3950
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003951/**
3952 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3953 * @skb: skb to check
3954 *
3955 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3956 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3957 * use this helper, to document places where we make this assertion.
3958 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003959static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003960{
3961#ifdef DEBUG
3962 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3963#endif
3964}
3965
Rusty Russellf35d9d82008-02-04 23:49:54 -05003966bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003967
Paul Durranted1f50c2014-01-09 10:02:46 +00003968int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003969struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3970 unsigned int transport_len,
3971 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003972
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003973/**
3974 * skb_head_is_locked - Determine if the skb->head is locked down
3975 * @skb: skb to check
3976 *
3977 * The head on skbs build around a head frag can be removed if they are
3978 * not cloned. This function returns true if the skb head is locked down
3979 * due to either being allocated via kmalloc, or by being a clone with
3980 * multiple references to the head.
3981 */
3982static inline bool skb_head_is_locked(const struct sk_buff *skb)
3983{
3984 return !skb->head_frag || skb_cloned(skb);
3985}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003986
3987/**
3988 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3989 *
3990 * @skb: GSO skb
3991 *
3992 * skb_gso_network_seglen is used to determine the real size of the
3993 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3994 *
3995 * The MAC/L2 header is not accounted for.
3996 */
3997static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3998{
3999 unsigned int hdr_len = skb_transport_header(skb) -
4000 skb_network_header(skb);
4001 return hdr_len + skb_gso_transport_seglen(skb);
4002}
Thomas Grafee122c72015-07-21 10:43:58 +02004003
Edward Cree179bc672016-02-11 20:48:04 +00004004/* Local Checksum Offload.
4005 * Compute outer checksum based on the assumption that the
4006 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00004007 * See Documentation/networking/checksum-offloads.txt for
4008 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004009 * Fill in outer checksum adjustment (e.g. with sum of outer
4010 * pseudo-header) before calling.
4011 * Also ensure that inner checksum is in linear data area.
4012 */
4013static inline __wsum lco_csum(struct sk_buff *skb)
4014{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004015 unsigned char *csum_start = skb_checksum_start(skb);
4016 unsigned char *l4_hdr = skb_transport_header(skb);
4017 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004018
4019 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004020 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4021 skb->csum_offset));
4022
Edward Cree179bc672016-02-11 20:48:04 +00004023 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004024 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004025 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004026 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004027}
4028
Linus Torvalds1da177e2005-04-16 15:20:36 -07004029#endif /* __KERNEL__ */
4030#endif /* _LINUX_SKBUFF_H */