blob: 7c46fd0b8b64a12f0d3f1a01b14cb4695968b611 [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>
18#include <linux/compiler.h>
19#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050020#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080022#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050023#include <linux/socket.h>
Eric Dumazetc1d1b432017-08-31 16:48:22 -070024#include <linux/refcount.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 Ry5f74f82e2016-02-03 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 };
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700459 refcount_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{
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700475 refcount_inc(&uarg->refcnt);
Willem de Bruijn52267792017-08-03 16:29:39 -0400476}
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 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700706 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200707
708/* if you move cloned around you also must adapt those constants */
709#ifdef __BIG_ENDIAN_BITFIELD
710#define CLONED_MASK (1 << 7)
711#else
712#define CLONED_MASK 1
713#endif
714#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
715
716 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700717 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700718 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800719 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800720 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700721 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200722 xmit_more:1,
723 __unused:1; /* one bit hole */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724
Eric Dumazetb1937222014-09-28 22:18:47 -0700725 /* fields enclosed in headers_start/headers_end are copied
726 * using a single memcpy() in __copy_skb_header()
727 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700728 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700729 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700730 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200731
732/* if you move pkt_type around you also must adapt those constants */
733#ifdef __BIG_ENDIAN_BITFIELD
734#define PKT_TYPE_MAX (7 << 5)
735#else
736#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100737#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200738#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
739
740 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700741 __u8 pkt_type:3;
742 __u8 pfmemalloc:1;
743 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744
Eric Dumazetb1937222014-09-28 22:18:47 -0700745 __u8 nf_trace:1;
746 __u8 ip_summed:2;
747 __u8 ooo_okay:1;
748 __u8 l4_hash:1;
749 __u8 sw_hash:1;
750 __u8 wifi_acked_valid:1;
751 __u8 wifi_acked:1;
752
753 __u8 no_fcs:1;
754 /* Indicates the inner headers are valid in the skbuff. */
755 __u8 encapsulation:1;
756 __u8 encap_hdr_csum:1;
757 __u8 csum_valid:1;
758 __u8 csum_complete_sw:1;
759 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200760 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700761
Julian Anastasov4ff06202017-02-06 23:14:12 +0200762 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700763#ifdef CONFIG_IPV6_NDISC_NODETYPE
764 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700766 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700767 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800768 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200769#ifdef CONFIG_NET_SWITCHDEV
770 __u8 offload_fwd_mark:1;
771#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500772#ifdef CONFIG_NET_CLS_ACT
773 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500774 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500775 __u8 tc_redirected:1;
776 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500777#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700780 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200782
Eric Dumazetb1937222014-09-28 22:18:47 -0700783 union {
784 __wsum csum;
785 struct {
786 __u16 csum_start;
787 __u16 csum_offset;
788 };
789 };
790 __u32 priority;
791 int skb_iif;
792 __u32 hash;
793 __be16 vlan_proto;
794 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800795#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
796 union {
797 unsigned int napi_id;
798 unsigned int sender_cpu;
799 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700800#endif
James Morris984bc162006-06-09 00:29:17 -0700801#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200802 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700803#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700804
Neil Horman3b885782009-10-12 13:26:31 -0700805 union {
806 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000807 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700808 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
Tom Herbert8bce6d72014-09-29 20:22:29 -0700810 union {
811 __be16 inner_protocol;
812 __u8 inner_ipproto;
813 };
814
Simon Horman1a37e412013-05-23 21:02:51 +0000815 __u16 inner_transport_header;
816 __u16 inner_network_header;
817 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700818
819 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000820 __u16 transport_header;
821 __u16 network_header;
822 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700823
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700824 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700825 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700826 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700827
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700829 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700830 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700832 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700833 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300834 refcount_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835};
836
837#ifdef __KERNEL__
838/*
839 * Handling routines are only of interest to the kernel
840 */
841#include <linux/slab.h>
842
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843
Mel Gormanc93bdd02012-07-31 16:44:19 -0700844#define SKB_ALLOC_FCLONE 0x01
845#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800846#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700847
848/* Returns true if the skb was allocated from PFMEMALLOC reserves */
849static inline bool skb_pfmemalloc(const struct sk_buff *skb)
850{
851 return unlikely(skb->pfmemalloc);
852}
853
Eric Dumazet7fee2262010-05-11 23:19:48 +0000854/*
855 * skb might have a dst pointer attached, refcounted or not.
856 * _skb_refdst low order bit is set if refcount was _not_ taken
857 */
858#define SKB_DST_NOREF 1UL
859#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
860
Florian Westphala9e419d2017-01-23 18:21:59 +0100861#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000862/**
863 * skb_dst - returns skb dst_entry
864 * @skb: buffer
865 *
866 * Returns skb dst_entry, regardless of reference taken or not.
867 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000868static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
869{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000870 /* If refdst was not refcounted, check we still are in a
871 * rcu_read_lock section
872 */
873 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
874 !rcu_read_lock_held() &&
875 !rcu_read_lock_bh_held());
876 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000877}
878
Eric Dumazet7fee2262010-05-11 23:19:48 +0000879/**
880 * skb_dst_set - sets skb dst
881 * @skb: buffer
882 * @dst: dst entry
883 *
884 * Sets skb dst, assuming a reference was taken on dst and should
885 * be released by skb_dst_drop()
886 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000887static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
888{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000889 skb->_skb_refdst = (unsigned long)dst;
890}
891
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200892/**
893 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
894 * @skb: buffer
895 * @dst: dst entry
896 *
897 * Sets skb dst, assuming a reference was not taken on dst.
898 * If dst entry is cached, we do not take reference and dst_release
899 * will be avoided by refdst_drop. If dst entry is not cached, we take
900 * reference, so that last dst_release can destroy the dst immediately.
901 */
902static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
903{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100904 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
905 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200906}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000907
908/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300909 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000910 * @skb: buffer
911 */
912static inline bool skb_dst_is_noref(const struct sk_buff *skb)
913{
914 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000915}
916
Eric Dumazet511c3f92009-06-02 05:14:27 +0000917static inline struct rtable *skb_rtable(const struct sk_buff *skb)
918{
Eric Dumazetadf30902009-06-02 05:19:30 +0000919 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000920}
921
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400922/* For mangling skb->pkt_type from user space side from applications
923 * such as nft, tc, etc, we only allow a conservative subset of
924 * possible pkt_types to be set.
925*/
926static inline bool skb_pkt_type_ok(u32 ptype)
927{
928 return ptype <= PACKET_OTHERHOST;
929}
930
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200931static inline unsigned int skb_napi_id(const struct sk_buff *skb)
932{
933#ifdef CONFIG_NET_RX_BUSY_POLL
934 return skb->napi_id;
935#else
936 return 0;
937#endif
938}
939
Paolo Abeni3889a8032017-06-12 11:23:41 +0200940/* decrement the reference count and return true if we can free the skb */
941static inline bool skb_unref(struct sk_buff *skb)
942{
943 if (unlikely(!skb))
944 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +0300945 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200946 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +0300947 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200948 return false;
949
950 return true;
951}
952
Paolo Abeni0a463c72017-06-12 11:23:42 +0200953void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700954void kfree_skb(struct sk_buff *skb);
955void kfree_skb_list(struct sk_buff *segs);
956void skb_tx_error(struct sk_buff *skb);
957void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +0200958void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700959void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000960extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000961
Joe Perches7965bd42013-09-26 14:48:15 -0700962void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
963bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
964 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000965
Joe Perches7965bd42013-09-26 14:48:15 -0700966struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
967 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700968struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700969struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700970static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100971 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700972{
Eric Dumazet564824b2010-10-11 19:05:25 +0000973 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700974}
975
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700976struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
977 unsigned long data_len,
978 int max_page_order,
979 int *errcode,
980 gfp_t gfp_mask);
981
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700982/* Layout of fast clones : [skb1][skb2][fclone_ref] */
983struct sk_buff_fclones {
984 struct sk_buff skb1;
985
986 struct sk_buff skb2;
987
Reshetova, Elena26385952017-06-30 13:07:59 +0300988 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700989};
990
991/**
992 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700993 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700994 * @skb: buffer
995 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900996 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700997 * Some drivers call skb_orphan() in their ndo_start_xmit(),
998 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700999 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001000static inline bool skb_fclone_busy(const struct sock *sk,
1001 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001002{
1003 const struct sk_buff_fclones *fclones;
1004
1005 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1006
1007 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001008 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001009 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001010}
1011
David S. Millerd179cd12005-08-17 14:57:30 -07001012static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001013 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001014{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001015 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001016}
1017
Joe Perches7965bd42013-09-26 14:48:15 -07001018struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1019int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1020struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1021struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001022struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1023 gfp_t gfp_mask, bool fclone);
1024static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1025 gfp_t gfp_mask)
1026{
1027 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1028}
Eric Dumazet117632e2011-12-03 21:39:53 +00001029
Joe Perches7965bd42013-09-26 14:48:15 -07001030int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1031struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1032 unsigned int headroom);
1033struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1034 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001035int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1036 int offset, int len);
1037int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1038 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001039int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001040int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1041
1042/**
1043 * skb_pad - zero pad the tail of an skb
1044 * @skb: buffer to pad
1045 * @pad: space to pad
1046 *
1047 * Ensure that a buffer is followed by a padding area that is zero
1048 * filled. Used by network drivers which may DMA or transfer data
1049 * beyond the buffer end onto the wire.
1050 *
1051 * May return error in out of memory cases. The skb is freed on error.
1052 */
1053static inline int skb_pad(struct sk_buff *skb, int pad)
1054{
1055 return __skb_pad(skb, pad, true);
1056}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001057#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
Joe Perches7965bd42013-09-26 14:48:15 -07001059int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1060 int getfrag(void *from, char *to, int offset,
1061 int len, int odd, struct sk_buff *skb),
1062 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001063
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001064int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1065 int offset, size_t size);
1066
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001067struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001068 __u32 lower_offset;
1069 __u32 upper_offset;
1070 __u32 frag_idx;
1071 __u32 stepped_offset;
1072 struct sk_buff *root_skb;
1073 struct sk_buff *cur_skb;
1074 __u8 *frag_data;
1075};
1076
Joe Perches7965bd42013-09-26 14:48:15 -07001077void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1078 unsigned int to, struct skb_seq_state *st);
1079unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1080 struct skb_seq_state *st);
1081void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001082
Joe Perches7965bd42013-09-26 14:48:15 -07001083unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001084 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001085
Tom Herbert09323cc2013-12-15 22:16:19 -08001086/*
1087 * Packet hash types specify the type of hash in skb_set_hash.
1088 *
1089 * Hash types refer to the protocol layer addresses which are used to
1090 * construct a packet's hash. The hashes are used to differentiate or identify
1091 * flows of the protocol layer for the hash type. Hash types are either
1092 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1093 *
1094 * Properties of hashes:
1095 *
1096 * 1) Two packets in different flows have different hash values
1097 * 2) Two packets in the same flow should have the same hash value
1098 *
1099 * A hash at a higher layer is considered to be more specific. A driver should
1100 * set the most specific hash possible.
1101 *
1102 * A driver cannot indicate a more specific hash than the layer at which a hash
1103 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1104 *
1105 * A driver may indicate a hash level which is less specific than the
1106 * actual layer the hash was computed on. For instance, a hash computed
1107 * at L4 may be considered an L3 hash. This should only be done if the
1108 * driver can't unambiguously determine that the HW computed the hash at
1109 * the higher layer. Note that the "should" in the second property above
1110 * permits this.
1111 */
1112enum pkt_hash_types {
1113 PKT_HASH_TYPE_NONE, /* Undefined type */
1114 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1115 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1116 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1117};
1118
Tom Herbertbcc83832015-09-01 09:24:24 -07001119static inline void skb_clear_hash(struct sk_buff *skb)
1120{
1121 skb->hash = 0;
1122 skb->sw_hash = 0;
1123 skb->l4_hash = 0;
1124}
1125
1126static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1127{
1128 if (!skb->l4_hash)
1129 skb_clear_hash(skb);
1130}
1131
1132static inline void
1133__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1134{
1135 skb->l4_hash = is_l4;
1136 skb->sw_hash = is_sw;
1137 skb->hash = hash;
1138}
1139
Tom Herbert09323cc2013-12-15 22:16:19 -08001140static inline void
1141skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1142{
Tom Herbertbcc83832015-09-01 09:24:24 -07001143 /* Used by drivers to set hash from HW */
1144 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1145}
1146
1147static inline void
1148__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1149{
1150 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001151}
1152
Tom Herberte5276932015-09-01 09:24:23 -07001153void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001154u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001155u32 skb_get_poff(const struct sk_buff *skb);
1156u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1157 const struct flow_keys *keys, int hlen);
1158__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1159 void *data, int hlen_proto);
1160
1161static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1162 int thoff, u8 ip_proto)
1163{
1164 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1165}
1166
1167void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1168 const struct flow_dissector_key *key,
1169 unsigned int key_count);
1170
1171bool __skb_flow_dissect(const struct sk_buff *skb,
1172 struct flow_dissector *flow_dissector,
1173 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001174 void *data, __be16 proto, int nhoff, int hlen,
1175 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001176
1177static inline bool skb_flow_dissect(const struct sk_buff *skb,
1178 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001179 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001180{
1181 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001182 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001183}
1184
1185static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001186 struct flow_keys *flow,
1187 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001188{
1189 memset(flow, 0, sizeof(*flow));
1190 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001191 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001192}
1193
1194static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1195 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001196 int nhoff, int hlen,
1197 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001198{
1199 memset(flow, 0, sizeof(*flow));
1200 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001201 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001202}
1203
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001204static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001205{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001206 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001207 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001208
Tom Herbert61b905d2014-03-24 15:34:47 -07001209 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001210}
1211
David S. Miller20a17bf2015-09-01 21:19:17 -07001212static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001213{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001214 if (!skb->l4_hash && !skb->sw_hash) {
1215 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001216 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001217
Tom Herbertde4c1f82015-09-01 18:11:04 -07001218 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001219 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001220
1221 return skb->hash;
1222}
1223
Tom Herbert50fb7992015-05-01 11:30:12 -07001224__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1225
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001226static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1227{
Tom Herbert61b905d2014-03-24 15:34:47 -07001228 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001229}
1230
Tom Herbert3df7a742013-12-15 22:16:29 -08001231static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1232{
Tom Herbert61b905d2014-03-24 15:34:47 -07001233 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001234 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001235 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001236};
1237
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001238#ifdef NET_SKBUFF_DATA_USES_OFFSET
1239static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1240{
1241 return skb->head + skb->end;
1242}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001243
1244static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1245{
1246 return skb->end;
1247}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001248#else
1249static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1250{
1251 return skb->end;
1252}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001253
1254static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1255{
1256 return skb->end - skb->head;
1257}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001258#endif
1259
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001261#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262
Patrick Ohlyac45f602009-02-12 05:03:37 +00001263static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1264{
1265 return &skb_shinfo(skb)->hwtstamps;
1266}
1267
Willem de Bruijn52267792017-08-03 16:29:39 -04001268static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1269{
1270 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1271
1272 return is_zcopy ? skb_uarg(skb) : NULL;
1273}
1274
1275static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1276{
1277 if (skb && uarg && !skb_zcopy(skb)) {
1278 sock_zerocopy_get(uarg);
1279 skb_shinfo(skb)->destructor_arg = uarg;
1280 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1281 }
1282}
1283
1284/* Release a reference on a zerocopy structure */
1285static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1286{
1287 struct ubuf_info *uarg = skb_zcopy(skb);
1288
1289 if (uarg) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001290 if (uarg->callback == sock_zerocopy_callback) {
1291 uarg->zerocopy = uarg->zerocopy && zerocopy;
1292 sock_zerocopy_put(uarg);
1293 } else {
1294 uarg->callback(uarg, zerocopy);
1295 }
1296
Willem de Bruijn52267792017-08-03 16:29:39 -04001297 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1298 }
1299}
1300
1301/* Abort a zerocopy operation and revert zckey on error in send syscall */
1302static inline void skb_zcopy_abort(struct sk_buff *skb)
1303{
1304 struct ubuf_info *uarg = skb_zcopy(skb);
1305
1306 if (uarg) {
1307 sock_zerocopy_put_abort(uarg);
1308 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1309 }
1310}
1311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312/**
1313 * skb_queue_empty - check if a queue is empty
1314 * @list: queue head
1315 *
1316 * Returns true if the queue is empty, false otherwise.
1317 */
1318static inline int skb_queue_empty(const struct sk_buff_head *list)
1319{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001320 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321}
1322
1323/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001324 * skb_queue_is_last - check if skb is the last entry in the queue
1325 * @list: queue head
1326 * @skb: buffer
1327 *
1328 * Returns true if @skb is the last buffer on the list.
1329 */
1330static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1331 const struct sk_buff *skb)
1332{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001333 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001334}
1335
1336/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001337 * skb_queue_is_first - check if skb is the first entry in the queue
1338 * @list: queue head
1339 * @skb: buffer
1340 *
1341 * Returns true if @skb is the first buffer on the list.
1342 */
1343static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1344 const struct sk_buff *skb)
1345{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001346 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001347}
1348
1349/**
David S. Miller249c8b42008-09-23 00:44:42 -07001350 * skb_queue_next - return the next packet in the queue
1351 * @list: queue head
1352 * @skb: current buffer
1353 *
1354 * Return the next packet in @list after @skb. It is only valid to
1355 * call this if skb_queue_is_last() evaluates to false.
1356 */
1357static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1358 const struct sk_buff *skb)
1359{
1360 /* This BUG_ON may seem severe, but if we just return then we
1361 * are going to dereference garbage.
1362 */
1363 BUG_ON(skb_queue_is_last(list, skb));
1364 return skb->next;
1365}
1366
1367/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001368 * skb_queue_prev - return the prev packet in the queue
1369 * @list: queue head
1370 * @skb: current buffer
1371 *
1372 * Return the prev packet in @list before @skb. It is only valid to
1373 * call this if skb_queue_is_first() evaluates to false.
1374 */
1375static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1376 const struct sk_buff *skb)
1377{
1378 /* This BUG_ON may seem severe, but if we just return then we
1379 * are going to dereference garbage.
1380 */
1381 BUG_ON(skb_queue_is_first(list, skb));
1382 return skb->prev;
1383}
1384
1385/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 * skb_get - reference buffer
1387 * @skb: buffer to reference
1388 *
1389 * Makes another reference to a socket buffer and returns a pointer
1390 * to the buffer.
1391 */
1392static inline struct sk_buff *skb_get(struct sk_buff *skb)
1393{
Reshetova, Elena63354792017-06-30 13:07:58 +03001394 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 return skb;
1396}
1397
1398/*
1399 * If users == 1, we are the only owner and are can avoid redundant
1400 * atomic change.
1401 */
1402
1403/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 * skb_cloned - is the buffer a clone
1405 * @skb: buffer to check
1406 *
1407 * Returns true if the buffer was generated with skb_clone() and is
1408 * one of multiple shared copies of the buffer. Cloned buffers are
1409 * shared data so must not be written to under normal circumstances.
1410 */
1411static inline int skb_cloned(const struct sk_buff *skb)
1412{
1413 return skb->cloned &&
1414 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1415}
1416
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001417static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1418{
Mel Gormand0164ad2015-11-06 16:28:21 -08001419 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001420
1421 if (skb_cloned(skb))
1422 return pskb_expand_head(skb, 0, 0, pri);
1423
1424 return 0;
1425}
1426
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427/**
1428 * skb_header_cloned - is the header a clone
1429 * @skb: buffer to check
1430 *
1431 * Returns true if modifying the header part of the buffer requires
1432 * the data to be copied.
1433 */
1434static inline int skb_header_cloned(const struct sk_buff *skb)
1435{
1436 int dataref;
1437
1438 if (!skb->cloned)
1439 return 0;
1440
1441 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1442 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1443 return dataref != 1;
1444}
1445
Eric Dumazet9580bf22016-04-30 10:19:29 -07001446static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1447{
1448 might_sleep_if(gfpflags_allow_blocking(pri));
1449
1450 if (skb_header_cloned(skb))
1451 return pskb_expand_head(skb, 0, 0, pri);
1452
1453 return 0;
1454}
1455
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456/**
1457 * skb_header_release - release reference to header
1458 * @skb: buffer to operate on
1459 *
1460 * Drop a reference to the header part of the buffer. This is done
1461 * by acquiring a payload reference. You must not read from the header
1462 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001463 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 */
1465static inline void skb_header_release(struct sk_buff *skb)
1466{
1467 BUG_ON(skb->nohdr);
1468 skb->nohdr = 1;
1469 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1470}
1471
1472/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001473 * __skb_header_release - release reference to header
1474 * @skb: buffer to operate on
1475 *
1476 * Variant of skb_header_release() assuming skb is private to caller.
1477 * We can avoid one atomic operation.
1478 */
1479static inline void __skb_header_release(struct sk_buff *skb)
1480{
1481 skb->nohdr = 1;
1482 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1483}
1484
1485
1486/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 * skb_shared - is the buffer shared
1488 * @skb: buffer to check
1489 *
1490 * Returns true if more than one person has a reference to this
1491 * buffer.
1492 */
1493static inline int skb_shared(const struct sk_buff *skb)
1494{
Reshetova, Elena63354792017-06-30 13:07:58 +03001495 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496}
1497
1498/**
1499 * skb_share_check - check if buffer is shared and if so clone it
1500 * @skb: buffer to check
1501 * @pri: priority for memory allocation
1502 *
1503 * If the buffer is shared the buffer is cloned and the old copy
1504 * drops a reference. A new clone with a single reference is returned.
1505 * If the buffer is not shared the original buffer is returned. When
1506 * being called from interrupt status or with spinlocks held pri must
1507 * be GFP_ATOMIC.
1508 *
1509 * NULL is returned on a memory allocation failure.
1510 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001511static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512{
Mel Gormand0164ad2015-11-06 16:28:21 -08001513 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 if (skb_shared(skb)) {
1515 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001516
1517 if (likely(nskb))
1518 consume_skb(skb);
1519 else
1520 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 skb = nskb;
1522 }
1523 return skb;
1524}
1525
1526/*
1527 * Copy shared buffers into a new sk_buff. We effectively do COW on
1528 * packets to handle cases where we have a local reader and forward
1529 * and a couple of other messy ones. The normal one is tcpdumping
1530 * a packet thats being forwarded.
1531 */
1532
1533/**
1534 * skb_unshare - make a copy of a shared buffer
1535 * @skb: buffer to check
1536 * @pri: priority for memory allocation
1537 *
1538 * If the socket buffer is a clone then this function creates a new
1539 * copy of the data, drops a reference count on the old copy and returns
1540 * the new copy with the reference count at 1. If the buffer is not a clone
1541 * the original buffer is returned. When called with a spinlock held or
1542 * from interrupt state @pri must be %GFP_ATOMIC
1543 *
1544 * %NULL is returned on a memory allocation failure.
1545 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001546static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001547 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548{
Mel Gormand0164ad2015-11-06 16:28:21 -08001549 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 if (skb_cloned(skb)) {
1551 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001552
1553 /* Free our shared copy */
1554 if (likely(nskb))
1555 consume_skb(skb);
1556 else
1557 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 skb = nskb;
1559 }
1560 return skb;
1561}
1562
1563/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001564 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 * @list_: list to peek at
1566 *
1567 * Peek an &sk_buff. Unlike most other operations you _MUST_
1568 * be careful with this one. A peek leaves the buffer on the
1569 * list and someone else may run off with it. You must hold
1570 * the appropriate locks or have a private queue to do this.
1571 *
1572 * Returns %NULL for an empty list or a pointer to the head element.
1573 * The reference count is not incremented and the reference is therefore
1574 * volatile. Use with caution.
1575 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001576static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577{
Eric Dumazet18d07002012-04-30 16:31:46 +00001578 struct sk_buff *skb = list_->next;
1579
1580 if (skb == (struct sk_buff *)list_)
1581 skb = NULL;
1582 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583}
1584
1585/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001586 * skb_peek_next - peek skb following the given one from a queue
1587 * @skb: skb to start from
1588 * @list_: list to peek at
1589 *
1590 * Returns %NULL when the end of the list is met or a pointer to the
1591 * next element. The reference count is not incremented and the
1592 * reference is therefore volatile. Use with caution.
1593 */
1594static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1595 const struct sk_buff_head *list_)
1596{
1597 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001598
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001599 if (next == (struct sk_buff *)list_)
1600 next = NULL;
1601 return next;
1602}
1603
1604/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001605 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 * @list_: list to peek at
1607 *
1608 * Peek an &sk_buff. Unlike most other operations you _MUST_
1609 * be careful with this one. A peek leaves the buffer on the
1610 * list and someone else may run off with it. You must hold
1611 * the appropriate locks or have a private queue to do this.
1612 *
1613 * Returns %NULL for an empty list or a pointer to the tail element.
1614 * The reference count is not incremented and the reference is therefore
1615 * volatile. Use with caution.
1616 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001617static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618{
Eric Dumazet18d07002012-04-30 16:31:46 +00001619 struct sk_buff *skb = list_->prev;
1620
1621 if (skb == (struct sk_buff *)list_)
1622 skb = NULL;
1623 return skb;
1624
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625}
1626
1627/**
1628 * skb_queue_len - get queue length
1629 * @list_: list to measure
1630 *
1631 * Return the length of an &sk_buff queue.
1632 */
1633static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1634{
1635 return list_->qlen;
1636}
1637
David S. Miller67fed452008-09-21 22:36:24 -07001638/**
1639 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1640 * @list: queue to initialize
1641 *
1642 * This initializes only the list and queue length aspects of
1643 * an sk_buff_head object. This allows to initialize the list
1644 * aspects of an sk_buff_head without reinitializing things like
1645 * the spinlock. It can also be used for on-stack sk_buff_head
1646 * objects where the spinlock is known to not be used.
1647 */
1648static inline void __skb_queue_head_init(struct sk_buff_head *list)
1649{
1650 list->prev = list->next = (struct sk_buff *)list;
1651 list->qlen = 0;
1652}
1653
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001654/*
1655 * This function creates a split out lock class for each invocation;
1656 * this is needed for now since a whole lot of users of the skb-queue
1657 * infrastructure in drivers have different locking usage (in hardirq)
1658 * than the networking core (in softirq only). In the long run either the
1659 * network layer or drivers should need annotation to consolidate the
1660 * main types of usage into 3 classes.
1661 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662static inline void skb_queue_head_init(struct sk_buff_head *list)
1663{
1664 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001665 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666}
1667
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001668static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1669 struct lock_class_key *class)
1670{
1671 skb_queue_head_init(list);
1672 lockdep_set_class(&list->lock, class);
1673}
1674
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001676 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 *
1678 * The "__skb_xxxx()" functions are the non-atomic ones that
1679 * can only be called with interrupts disabled.
1680 */
Joe Perches7965bd42013-09-26 14:48:15 -07001681void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1682 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001683static inline void __skb_insert(struct sk_buff *newsk,
1684 struct sk_buff *prev, struct sk_buff *next,
1685 struct sk_buff_head *list)
1686{
1687 newsk->next = next;
1688 newsk->prev = prev;
1689 next->prev = prev->next = newsk;
1690 list->qlen++;
1691}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692
David S. Miller67fed452008-09-21 22:36:24 -07001693static inline void __skb_queue_splice(const struct sk_buff_head *list,
1694 struct sk_buff *prev,
1695 struct sk_buff *next)
1696{
1697 struct sk_buff *first = list->next;
1698 struct sk_buff *last = list->prev;
1699
1700 first->prev = prev;
1701 prev->next = first;
1702
1703 last->next = next;
1704 next->prev = last;
1705}
1706
1707/**
1708 * skb_queue_splice - join two skb lists, this is designed for stacks
1709 * @list: the new list to add
1710 * @head: the place to add it in the first list
1711 */
1712static inline void skb_queue_splice(const struct sk_buff_head *list,
1713 struct sk_buff_head *head)
1714{
1715 if (!skb_queue_empty(list)) {
1716 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001717 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001718 }
1719}
1720
1721/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001722 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001723 * @list: the new list to add
1724 * @head: the place to add it in the first list
1725 *
1726 * The list at @list is reinitialised
1727 */
1728static inline void skb_queue_splice_init(struct sk_buff_head *list,
1729 struct sk_buff_head *head)
1730{
1731 if (!skb_queue_empty(list)) {
1732 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001733 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001734 __skb_queue_head_init(list);
1735 }
1736}
1737
1738/**
1739 * skb_queue_splice_tail - join two skb lists, each list being a queue
1740 * @list: the new list to add
1741 * @head: the place to add it in the first list
1742 */
1743static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1744 struct sk_buff_head *head)
1745{
1746 if (!skb_queue_empty(list)) {
1747 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001748 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001749 }
1750}
1751
1752/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001753 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001754 * @list: the new list to add
1755 * @head: the place to add it in the first list
1756 *
1757 * Each of the lists is a queue.
1758 * The list at @list is reinitialised
1759 */
1760static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1761 struct sk_buff_head *head)
1762{
1763 if (!skb_queue_empty(list)) {
1764 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001765 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001766 __skb_queue_head_init(list);
1767 }
1768}
1769
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001771 * __skb_queue_after - queue a buffer at the list head
1772 * @list: list to use
1773 * @prev: place after this buffer
1774 * @newsk: buffer to queue
1775 *
1776 * Queue a buffer int the middle of a list. This function takes no locks
1777 * and you must therefore hold required locks before calling it.
1778 *
1779 * A buffer cannot be placed on two lists at the same time.
1780 */
1781static inline void __skb_queue_after(struct sk_buff_head *list,
1782 struct sk_buff *prev,
1783 struct sk_buff *newsk)
1784{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001785 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001786}
1787
Joe Perches7965bd42013-09-26 14:48:15 -07001788void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1789 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001790
Gerrit Renkerf5572852008-04-14 00:05:28 -07001791static inline void __skb_queue_before(struct sk_buff_head *list,
1792 struct sk_buff *next,
1793 struct sk_buff *newsk)
1794{
1795 __skb_insert(newsk, next->prev, next, list);
1796}
1797
Stephen Hemminger300ce172005-10-30 13:47:34 -08001798/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 * __skb_queue_head - queue a buffer at the list head
1800 * @list: list to use
1801 * @newsk: buffer to queue
1802 *
1803 * Queue a buffer at the start of a list. This function takes no locks
1804 * and you must therefore hold required locks before calling it.
1805 *
1806 * A buffer cannot be placed on two lists at the same time.
1807 */
Joe Perches7965bd42013-09-26 14:48:15 -07001808void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809static inline void __skb_queue_head(struct sk_buff_head *list,
1810 struct sk_buff *newsk)
1811{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001812 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813}
1814
1815/**
1816 * __skb_queue_tail - queue a buffer at the list tail
1817 * @list: list to use
1818 * @newsk: buffer to queue
1819 *
1820 * Queue a buffer at the end of a list. This function takes no locks
1821 * and you must therefore hold required locks before calling it.
1822 *
1823 * A buffer cannot be placed on two lists at the same time.
1824 */
Joe Perches7965bd42013-09-26 14:48:15 -07001825void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826static inline void __skb_queue_tail(struct sk_buff_head *list,
1827 struct sk_buff *newsk)
1828{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001829 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830}
1831
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 * remove sk_buff from list. _Must_ be called atomically, and with
1834 * the list known..
1835 */
Joe Perches7965bd42013-09-26 14:48:15 -07001836void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1838{
1839 struct sk_buff *next, *prev;
1840
1841 list->qlen--;
1842 next = skb->next;
1843 prev = skb->prev;
1844 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 next->prev = prev;
1846 prev->next = next;
1847}
1848
Gerrit Renkerf525c062008-04-14 00:04:12 -07001849/**
1850 * __skb_dequeue - remove from the head of the queue
1851 * @list: list to dequeue from
1852 *
1853 * Remove the head of the list. This function does not take any locks
1854 * so must be used with appropriate locks held only. The head item is
1855 * returned or %NULL if the list is empty.
1856 */
Joe Perches7965bd42013-09-26 14:48:15 -07001857struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001858static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1859{
1860 struct sk_buff *skb = skb_peek(list);
1861 if (skb)
1862 __skb_unlink(skb, list);
1863 return skb;
1864}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
1866/**
1867 * __skb_dequeue_tail - remove from the tail of the queue
1868 * @list: list to dequeue from
1869 *
1870 * Remove the tail of the list. This function does not take any locks
1871 * so must be used with appropriate locks held only. The tail item is
1872 * returned or %NULL if the list is empty.
1873 */
Joe Perches7965bd42013-09-26 14:48:15 -07001874struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1876{
1877 struct sk_buff *skb = skb_peek_tail(list);
1878 if (skb)
1879 __skb_unlink(skb, list);
1880 return skb;
1881}
1882
1883
David S. Millerbdcc0922012-03-07 20:53:36 -05001884static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885{
1886 return skb->data_len;
1887}
1888
1889static inline unsigned int skb_headlen(const struct sk_buff *skb)
1890{
1891 return skb->len - skb->data_len;
1892}
1893
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001894static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001896 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001898 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001899 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001900 return len;
1901}
1902
1903static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1904{
1905 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906}
1907
Ian Campbell131ea662011-08-19 06:25:00 +00001908/**
1909 * __skb_fill_page_desc - initialise a paged fragment in an skb
1910 * @skb: buffer containing fragment to be initialised
1911 * @i: paged fragment index to initialise
1912 * @page: the page to use for this fragment
1913 * @off: the offset to the data with @page
1914 * @size: the length of the data
1915 *
1916 * Initialises the @i'th fragment of @skb to point to &size bytes at
1917 * offset @off within @page.
1918 *
1919 * Does not take any additional reference on the fragment.
1920 */
1921static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1922 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923{
1924 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1925
Mel Gormanc48a11c2012-07-31 16:44:23 -07001926 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001927 * Propagate page pfmemalloc to the skb if we can. The problem is
1928 * that not all callers have unique ownership of the page but rely
1929 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001930 */
Ian Campbella8605c62011-10-19 23:01:49 +00001931 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001933 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001934
1935 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001936 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001937 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001938}
1939
1940/**
1941 * skb_fill_page_desc - initialise a paged fragment in an skb
1942 * @skb: buffer containing fragment to be initialised
1943 * @i: paged fragment index to initialise
1944 * @page: the page to use for this fragment
1945 * @off: the offset to the data with @page
1946 * @size: the length of the data
1947 *
1948 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001949 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001950 * addition updates @skb such that @i is the last fragment.
1951 *
1952 * Does not take any additional reference on the fragment.
1953 */
1954static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1955 struct page *page, int off, int size)
1956{
1957 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 skb_shinfo(skb)->nr_frags = i + 1;
1959}
1960
Joe Perches7965bd42013-09-26 14:48:15 -07001961void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1962 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001963
Jason Wangf8e617e2013-11-01 14:07:47 +08001964void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1965 unsigned int truesize);
1966
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001968#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1970
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001971#ifdef NET_SKBUFF_DATA_USES_OFFSET
1972static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1973{
1974 return skb->head + skb->tail;
1975}
1976
1977static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1978{
1979 skb->tail = skb->data - skb->head;
1980}
1981
1982static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1983{
1984 skb_reset_tail_pointer(skb);
1985 skb->tail += offset;
1986}
Simon Horman7cc46192013-05-28 20:34:29 +00001987
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001988#else /* NET_SKBUFF_DATA_USES_OFFSET */
1989static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1990{
1991 return skb->tail;
1992}
1993
1994static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1995{
1996 skb->tail = skb->data;
1997}
1998
1999static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2000{
2001 skb->tail = skb->data + offset;
2002}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002003
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002004#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2005
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006/*
2007 * Add data to an sk_buff
2008 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002009void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2010void *skb_put(struct sk_buff *skb, unsigned int len);
2011static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012{
Johannes Berg4df864c2017-06-16 14:29:21 +02002013 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 SKB_LINEAR_ASSERT(skb);
2015 skb->tail += len;
2016 skb->len += len;
2017 return tmp;
2018}
2019
yuan linyude77b962017-06-18 22:48:17 +08002020static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2021{
2022 void *tmp = __skb_put(skb, len);
2023
2024 memset(tmp, 0, len);
2025 return tmp;
2026}
2027
2028static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2029 unsigned int len)
2030{
2031 void *tmp = __skb_put(skb, len);
2032
2033 memcpy(tmp, data, len);
2034 return tmp;
2035}
2036
2037static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2038{
2039 *(u8 *)__skb_put(skb, 1) = val;
2040}
2041
Johannes Berg83ad3572017-06-14 22:17:20 +02002042static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002043{
Johannes Berg83ad3572017-06-14 22:17:20 +02002044 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002045
2046 memset(tmp, 0, len);
2047
2048 return tmp;
2049}
2050
Johannes Berg59ae1d12017-06-16 14:29:20 +02002051static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2052 unsigned int len)
2053{
2054 void *tmp = skb_put(skb, len);
2055
2056 memcpy(tmp, data, len);
2057
2058 return tmp;
2059}
2060
Johannes Berg634fef62017-06-16 14:29:24 +02002061static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2062{
2063 *(u8 *)skb_put(skb, 1) = val;
2064}
2065
Johannes Bergd58ff352017-06-16 14:29:23 +02002066void *skb_push(struct sk_buff *skb, unsigned int len);
2067static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068{
2069 skb->data -= len;
2070 skb->len += len;
2071 return skb->data;
2072}
2073
Johannes Bergaf728682017-06-16 14:29:22 +02002074void *skb_pull(struct sk_buff *skb, unsigned int len);
2075static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076{
2077 skb->len -= len;
2078 BUG_ON(skb->len < skb->data_len);
2079 return skb->data += len;
2080}
2081
Johannes Bergaf728682017-06-16 14:29:22 +02002082static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002083{
2084 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2085}
2086
Johannes Bergaf728682017-06-16 14:29:22 +02002087void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088
Johannes Bergaf728682017-06-16 14:29:22 +02002089static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090{
2091 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002092 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 return NULL;
2094 skb->len -= len;
2095 return skb->data += len;
2096}
2097
Johannes Bergaf728682017-06-16 14:29:22 +02002098static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099{
2100 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2101}
2102
2103static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2104{
2105 if (likely(len <= skb_headlen(skb)))
2106 return 1;
2107 if (unlikely(len > skb->len))
2108 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002109 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110}
2111
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002112void skb_condense(struct sk_buff *skb);
2113
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114/**
2115 * skb_headroom - bytes at buffer head
2116 * @skb: buffer to check
2117 *
2118 * Return the number of bytes of free space at the head of an &sk_buff.
2119 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002120static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121{
2122 return skb->data - skb->head;
2123}
2124
2125/**
2126 * skb_tailroom - bytes at buffer end
2127 * @skb: buffer to check
2128 *
2129 * Return the number of bytes of free space at the tail of an sk_buff
2130 */
2131static inline int skb_tailroom(const struct sk_buff *skb)
2132{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002133 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134}
2135
2136/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002137 * skb_availroom - bytes at buffer end
2138 * @skb: buffer to check
2139 *
2140 * Return the number of bytes of free space at the tail of an sk_buff
2141 * allocated by sk_stream_alloc()
2142 */
2143static inline int skb_availroom(const struct sk_buff *skb)
2144{
Eric Dumazet16fad692013-03-14 05:40:32 +00002145 if (skb_is_nonlinear(skb))
2146 return 0;
2147
2148 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002149}
2150
2151/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 * skb_reserve - adjust headroom
2153 * @skb: buffer to alter
2154 * @len: bytes to move
2155 *
2156 * Increase the headroom of an empty &sk_buff by reducing the tail
2157 * room. This is only allowed for an empty buffer.
2158 */
David S. Miller8243126c2006-01-17 02:54:21 -08002159static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160{
2161 skb->data += len;
2162 skb->tail += len;
2163}
2164
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002165/**
2166 * skb_tailroom_reserve - adjust reserved_tailroom
2167 * @skb: buffer to alter
2168 * @mtu: maximum amount of headlen permitted
2169 * @needed_tailroom: minimum amount of reserved_tailroom
2170 *
2171 * Set reserved_tailroom so that headlen can be as large as possible but
2172 * not larger than mtu and tailroom cannot be smaller than
2173 * needed_tailroom.
2174 * The required headroom should already have been reserved before using
2175 * this function.
2176 */
2177static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2178 unsigned int needed_tailroom)
2179{
2180 SKB_LINEAR_ASSERT(skb);
2181 if (mtu < skb_tailroom(skb) - needed_tailroom)
2182 /* use at most mtu */
2183 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2184 else
2185 /* use up to all available space */
2186 skb->reserved_tailroom = needed_tailroom;
2187}
2188
Tom Herbert8bce6d72014-09-29 20:22:29 -07002189#define ENCAP_TYPE_ETHER 0
2190#define ENCAP_TYPE_IPPROTO 1
2191
2192static inline void skb_set_inner_protocol(struct sk_buff *skb,
2193 __be16 protocol)
2194{
2195 skb->inner_protocol = protocol;
2196 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2197}
2198
2199static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2200 __u8 ipproto)
2201{
2202 skb->inner_ipproto = ipproto;
2203 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2204}
2205
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002206static inline void skb_reset_inner_headers(struct sk_buff *skb)
2207{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002208 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002209 skb->inner_network_header = skb->network_header;
2210 skb->inner_transport_header = skb->transport_header;
2211}
2212
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002213static inline void skb_reset_mac_len(struct sk_buff *skb)
2214{
2215 skb->mac_len = skb->network_header - skb->mac_header;
2216}
2217
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002218static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2219 *skb)
2220{
2221 return skb->head + skb->inner_transport_header;
2222}
2223
Tom Herbert55dc5a92015-12-14 11:19:40 -08002224static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2225{
2226 return skb_inner_transport_header(skb) - skb->data;
2227}
2228
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002229static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2230{
2231 skb->inner_transport_header = skb->data - skb->head;
2232}
2233
2234static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2235 const int offset)
2236{
2237 skb_reset_inner_transport_header(skb);
2238 skb->inner_transport_header += offset;
2239}
2240
2241static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2242{
2243 return skb->head + skb->inner_network_header;
2244}
2245
2246static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2247{
2248 skb->inner_network_header = skb->data - skb->head;
2249}
2250
2251static inline void skb_set_inner_network_header(struct sk_buff *skb,
2252 const int offset)
2253{
2254 skb_reset_inner_network_header(skb);
2255 skb->inner_network_header += offset;
2256}
2257
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002258static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2259{
2260 return skb->head + skb->inner_mac_header;
2261}
2262
2263static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2264{
2265 skb->inner_mac_header = skb->data - skb->head;
2266}
2267
2268static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2269 const int offset)
2270{
2271 skb_reset_inner_mac_header(skb);
2272 skb->inner_mac_header += offset;
2273}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002274static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2275{
Cong Wang35d04612013-05-29 15:16:05 +08002276 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002277}
2278
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002279static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2280{
2281 return skb->head + skb->transport_header;
2282}
2283
2284static inline void skb_reset_transport_header(struct sk_buff *skb)
2285{
2286 skb->transport_header = skb->data - skb->head;
2287}
2288
2289static inline void skb_set_transport_header(struct sk_buff *skb,
2290 const int offset)
2291{
2292 skb_reset_transport_header(skb);
2293 skb->transport_header += offset;
2294}
2295
2296static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2297{
2298 return skb->head + skb->network_header;
2299}
2300
2301static inline void skb_reset_network_header(struct sk_buff *skb)
2302{
2303 skb->network_header = skb->data - skb->head;
2304}
2305
2306static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2307{
2308 skb_reset_network_header(skb);
2309 skb->network_header += offset;
2310}
2311
2312static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2313{
2314 return skb->head + skb->mac_header;
2315}
2316
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002317static inline int skb_mac_offset(const struct sk_buff *skb)
2318{
2319 return skb_mac_header(skb) - skb->data;
2320}
2321
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002322static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2323{
2324 return skb->network_header - skb->mac_header;
2325}
2326
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002327static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2328{
Cong Wang35d04612013-05-29 15:16:05 +08002329 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002330}
2331
2332static inline void skb_reset_mac_header(struct sk_buff *skb)
2333{
2334 skb->mac_header = skb->data - skb->head;
2335}
2336
2337static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2338{
2339 skb_reset_mac_header(skb);
2340 skb->mac_header += offset;
2341}
2342
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002343static inline void skb_pop_mac_header(struct sk_buff *skb)
2344{
2345 skb->mac_header = skb->network_header;
2346}
2347
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002348static inline void skb_probe_transport_header(struct sk_buff *skb,
2349 const int offset_hint)
2350{
2351 struct flow_keys keys;
2352
2353 if (skb_transport_header_was_set(skb))
2354 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002355 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002356 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002357 else
2358 skb_set_transport_header(skb, offset_hint);
2359}
2360
Eric Dumazet03606892012-02-23 10:55:02 +00002361static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2362{
2363 if (skb_mac_header_was_set(skb)) {
2364 const unsigned char *old_mac = skb_mac_header(skb);
2365
2366 skb_set_mac_header(skb, -skb->mac_len);
2367 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2368 }
2369}
2370
Michał Mirosław04fb4512010-12-14 15:24:08 +00002371static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2372{
2373 return skb->csum_start - skb_headroom(skb);
2374}
2375
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002376static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2377{
2378 return skb->head + skb->csum_start;
2379}
2380
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002381static inline int skb_transport_offset(const struct sk_buff *skb)
2382{
2383 return skb_transport_header(skb) - skb->data;
2384}
2385
2386static inline u32 skb_network_header_len(const struct sk_buff *skb)
2387{
2388 return skb->transport_header - skb->network_header;
2389}
2390
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002391static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2392{
2393 return skb->inner_transport_header - skb->inner_network_header;
2394}
2395
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002396static inline int skb_network_offset(const struct sk_buff *skb)
2397{
2398 return skb_network_header(skb) - skb->data;
2399}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002400
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002401static inline int skb_inner_network_offset(const struct sk_buff *skb)
2402{
2403 return skb_inner_network_header(skb) - skb->data;
2404}
2405
Changli Gaof9599ce2010-08-04 04:43:44 +00002406static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2407{
2408 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2409}
2410
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411/*
2412 * CPUs often take a performance hit when accessing unaligned memory
2413 * locations. The actual performance hit varies, it can be small if the
2414 * hardware handles it or large if we have to take an exception and fix it
2415 * in software.
2416 *
2417 * Since an ethernet header is 14 bytes network drivers often end up with
2418 * the IP header at an unaligned offset. The IP header can be aligned by
2419 * shifting the start of the packet by 2 bytes. Drivers should do this
2420 * with:
2421 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002422 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423 *
2424 * The downside to this alignment of the IP header is that the DMA is now
2425 * unaligned. On some architectures the cost of an unaligned DMA is high
2426 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002427 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2429 * to be overridden.
2430 */
2431#ifndef NET_IP_ALIGN
2432#define NET_IP_ALIGN 2
2433#endif
2434
Anton Blanchard025be812006-03-31 02:27:06 -08002435/*
2436 * The networking layer reserves some headroom in skb data (via
2437 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2438 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002439 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002440 *
2441 * Unfortunately this headroom changes the DMA alignment of the resulting
2442 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2443 * on some architectures. An architecture can override this value,
2444 * perhaps setting it to a cacheline in size (since that will maintain
2445 * cacheline alignment of the DMA). It must be a power of 2.
2446 *
David S. Millerd6301d32009-02-08 19:24:13 -08002447 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002448 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002449 *
2450 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2451 * to reduce average number of cache lines per packet.
2452 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002453 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002454 */
2455#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002456#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002457#endif
2458
Joe Perches7965bd42013-09-26 14:48:15 -07002459int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002461static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002463 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002464 WARN_ON(1);
2465 return;
2466 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002467 skb->len = len;
2468 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469}
2470
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002471static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2472{
2473 __skb_set_length(skb, len);
2474}
2475
Joe Perches7965bd42013-09-26 14:48:15 -07002476void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477
2478static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2479{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002480 if (skb->data_len)
2481 return ___pskb_trim(skb, len);
2482 __skb_trim(skb, len);
2483 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484}
2485
2486static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2487{
2488 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2489}
2490
2491/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002492 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2493 * @skb: buffer to alter
2494 * @len: new length
2495 *
2496 * This is identical to pskb_trim except that the caller knows that
2497 * the skb is not cloned so we should never get an error due to out-
2498 * of-memory.
2499 */
2500static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2501{
2502 int err = pskb_trim(skb, len);
2503 BUG_ON(err);
2504}
2505
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002506static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2507{
2508 unsigned int diff = len - skb->len;
2509
2510 if (skb_tailroom(skb) < diff) {
2511 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2512 GFP_ATOMIC);
2513 if (ret)
2514 return ret;
2515 }
2516 __skb_set_length(skb, len);
2517 return 0;
2518}
2519
Herbert Xue9fa4f72006-08-13 20:12:58 -07002520/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002521 * skb_orphan - orphan a buffer
2522 * @skb: buffer to orphan
2523 *
2524 * If a buffer currently has an owner then we call the owner's
2525 * destructor function and make the @skb unowned. The buffer continues
2526 * to exist but is no longer charged to its former owner.
2527 */
2528static inline void skb_orphan(struct sk_buff *skb)
2529{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002530 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002531 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002532 skb->destructor = NULL;
2533 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002534 } else {
2535 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002536 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537}
2538
2539/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002540 * skb_orphan_frags - orphan the frags contained in a buffer
2541 * @skb: buffer to orphan frags from
2542 * @gfp_mask: allocation mask for replacement pages
2543 *
2544 * For each frag in the SKB which needs a destructor (i.e. has an
2545 * owner) create a copy of that frag and release the original
2546 * page by calling the destructor.
2547 */
2548static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2549{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002550 if (likely(!skb_zcopy(skb)))
2551 return 0;
2552 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2553 return 0;
2554 return skb_copy_ubufs(skb, gfp_mask);
2555}
2556
2557/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2558static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2559{
2560 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002561 return 0;
2562 return skb_copy_ubufs(skb, gfp_mask);
2563}
2564
2565/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002566 * __skb_queue_purge - empty a list
2567 * @list: list to empty
2568 *
2569 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2570 * the list and one reference dropped. This function does not take the
2571 * list lock and the caller must hold the relevant locks to use it.
2572 */
Joe Perches7965bd42013-09-26 14:48:15 -07002573void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002574static inline void __skb_queue_purge(struct sk_buff_head *list)
2575{
2576 struct sk_buff *skb;
2577 while ((skb = __skb_dequeue(list)) != NULL)
2578 kfree_skb(skb);
2579}
2580
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002581void skb_rbtree_purge(struct rb_root *root);
2582
Joe Perches7965bd42013-09-26 14:48:15 -07002583void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584
Joe Perches7965bd42013-09-26 14:48:15 -07002585struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2586 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002587
2588/**
2589 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2590 * @dev: network device to receive on
2591 * @length: length to allocate
2592 *
2593 * Allocate a new &sk_buff and assign it a usage count of one. The
2594 * buffer has unspecified headroom built in. Users should allocate
2595 * the headroom they think they need without accounting for the
2596 * built in space. The built in space is used for optimisations.
2597 *
2598 * %NULL is returned if there is no free memory. Although this function
2599 * allocates memory it can be called from an interrupt.
2600 */
2601static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002602 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002603{
2604 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2605}
2606
Eric Dumazet6f532612012-05-18 05:12:12 +00002607/* legacy helper around __netdev_alloc_skb() */
2608static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2609 gfp_t gfp_mask)
2610{
2611 return __netdev_alloc_skb(NULL, length, gfp_mask);
2612}
2613
2614/* legacy helper around netdev_alloc_skb() */
2615static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2616{
2617 return netdev_alloc_skb(NULL, length);
2618}
2619
2620
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002621static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2622 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002623{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002624 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002625
2626 if (NET_IP_ALIGN && skb)
2627 skb_reserve(skb, NET_IP_ALIGN);
2628 return skb;
2629}
2630
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002631static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2632 unsigned int length)
2633{
2634 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2635}
2636
Alexander Duyck181edb22015-05-06 21:12:03 -07002637static inline void skb_free_frag(void *addr)
2638{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002639 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002640}
2641
Alexander Duyckffde7322014-12-09 19:40:42 -08002642void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002643struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2644 unsigned int length, gfp_t gfp_mask);
2645static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2646 unsigned int length)
2647{
2648 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2649}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002650void napi_consume_skb(struct sk_buff *skb, int budget);
2651
2652void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002653void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002654
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002655/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002656 * __dev_alloc_pages - allocate page for network Rx
2657 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2658 * @order: size of the allocation
2659 *
2660 * Allocate a new page.
2661 *
2662 * %NULL is returned if there is no free memory.
2663*/
2664static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2665 unsigned int order)
2666{
2667 /* This piece of code contains several assumptions.
2668 * 1. This is for device Rx, therefor a cold page is preferred.
2669 * 2. The expectation is the user wants a compound page.
2670 * 3. If requesting a order 0 page it will not be compound
2671 * due to the check to see if order has a value in prep_new_page
2672 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2673 * code in gfp_to_alloc_flags that should be enforcing this.
2674 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002675 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002676
2677 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2678}
2679
2680static inline struct page *dev_alloc_pages(unsigned int order)
2681{
Neil Horman95829b32016-05-19 11:30:54 -04002682 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002683}
2684
2685/**
2686 * __dev_alloc_page - allocate a page for network Rx
2687 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2688 *
2689 * Allocate a new page.
2690 *
2691 * %NULL is returned if there is no free memory.
2692 */
2693static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2694{
2695 return __dev_alloc_pages(gfp_mask, 0);
2696}
2697
2698static inline struct page *dev_alloc_page(void)
2699{
Neil Horman95829b32016-05-19 11:30:54 -04002700 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002701}
2702
2703/**
Mel Gorman06140022012-07-31 16:44:24 -07002704 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2705 * @page: The page that was allocated from skb_alloc_page
2706 * @skb: The skb that may need pfmemalloc set
2707 */
2708static inline void skb_propagate_pfmemalloc(struct page *page,
2709 struct sk_buff *skb)
2710{
Michal Hocko2f064f32015-08-21 14:11:51 -07002711 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002712 skb->pfmemalloc = true;
2713}
2714
Eric Dumazet564824b2010-10-11 19:05:25 +00002715/**
Masanari Iidae2278672014-02-18 22:54:36 +09002716 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002717 * @frag: the paged fragment
2718 *
2719 * Returns the &struct page associated with @frag.
2720 */
2721static inline struct page *skb_frag_page(const skb_frag_t *frag)
2722{
Ian Campbella8605c62011-10-19 23:01:49 +00002723 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002724}
2725
2726/**
2727 * __skb_frag_ref - take an addition reference on a paged fragment.
2728 * @frag: the paged fragment
2729 *
2730 * Takes an additional reference on the paged fragment @frag.
2731 */
2732static inline void __skb_frag_ref(skb_frag_t *frag)
2733{
2734 get_page(skb_frag_page(frag));
2735}
2736
2737/**
2738 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2739 * @skb: the buffer
2740 * @f: the fragment offset.
2741 *
2742 * Takes an additional reference on the @f'th paged fragment of @skb.
2743 */
2744static inline void skb_frag_ref(struct sk_buff *skb, int f)
2745{
2746 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2747}
2748
2749/**
2750 * __skb_frag_unref - release a reference on a paged fragment.
2751 * @frag: the paged fragment
2752 *
2753 * Releases a reference on the paged fragment @frag.
2754 */
2755static inline void __skb_frag_unref(skb_frag_t *frag)
2756{
2757 put_page(skb_frag_page(frag));
2758}
2759
2760/**
2761 * skb_frag_unref - release a reference on a paged fragment of an skb.
2762 * @skb: the buffer
2763 * @f: the fragment offset
2764 *
2765 * Releases a reference on the @f'th paged fragment of @skb.
2766 */
2767static inline void skb_frag_unref(struct sk_buff *skb, int f)
2768{
2769 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2770}
2771
2772/**
2773 * skb_frag_address - gets the address of the data contained in a paged fragment
2774 * @frag: the paged fragment buffer
2775 *
2776 * Returns the address of the data within @frag. The page must already
2777 * be mapped.
2778 */
2779static inline void *skb_frag_address(const skb_frag_t *frag)
2780{
2781 return page_address(skb_frag_page(frag)) + frag->page_offset;
2782}
2783
2784/**
2785 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2786 * @frag: the paged fragment buffer
2787 *
2788 * Returns the address of the data within @frag. Checks that the page
2789 * is mapped and returns %NULL otherwise.
2790 */
2791static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2792{
2793 void *ptr = page_address(skb_frag_page(frag));
2794 if (unlikely(!ptr))
2795 return NULL;
2796
2797 return ptr + frag->page_offset;
2798}
2799
2800/**
2801 * __skb_frag_set_page - sets the page contained in a paged fragment
2802 * @frag: the paged fragment
2803 * @page: the page to set
2804 *
2805 * Sets the fragment @frag to contain @page.
2806 */
2807static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2808{
Ian Campbella8605c62011-10-19 23:01:49 +00002809 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002810}
2811
2812/**
2813 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2814 * @skb: the buffer
2815 * @f: the fragment offset
2816 * @page: the page to set
2817 *
2818 * Sets the @f'th fragment of @skb to contain @page.
2819 */
2820static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2821 struct page *page)
2822{
2823 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2824}
2825
Eric Dumazet400dfd32013-10-17 16:27:07 -07002826bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2827
Ian Campbell131ea662011-08-19 06:25:00 +00002828/**
2829 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002830 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002831 * @frag: the paged fragment to map
2832 * @offset: the offset within the fragment (starting at the
2833 * fragment's own offset)
2834 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002835 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002836 *
2837 * Maps the page associated with @frag to @device.
2838 */
2839static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2840 const skb_frag_t *frag,
2841 size_t offset, size_t size,
2842 enum dma_data_direction dir)
2843{
2844 return dma_map_page(dev, skb_frag_page(frag),
2845 frag->page_offset + offset, size, dir);
2846}
2847
Eric Dumazet117632e2011-12-03 21:39:53 +00002848static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2849 gfp_t gfp_mask)
2850{
2851 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2852}
2853
Octavian Purdilabad93e92014-06-12 01:36:26 +03002854
2855static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2856 gfp_t gfp_mask)
2857{
2858 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2859}
2860
2861
Ian Campbell131ea662011-08-19 06:25:00 +00002862/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002863 * skb_clone_writable - is the header of a clone writable
2864 * @skb: buffer to check
2865 * @len: length up to which to write
2866 *
2867 * Returns true if modifying the header part of the cloned buffer
2868 * does not requires the data to be copied.
2869 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002870static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002871{
2872 return !skb_header_cloned(skb) &&
2873 skb_headroom(skb) + len <= skb->hdr_len;
2874}
2875
Daniel Borkmann36976492016-02-19 23:05:25 +01002876static inline int skb_try_make_writable(struct sk_buff *skb,
2877 unsigned int write_len)
2878{
2879 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2880 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2881}
2882
Herbert Xud9cc2042007-09-16 16:21:16 -07002883static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2884 int cloned)
2885{
2886 int delta = 0;
2887
Herbert Xud9cc2042007-09-16 16:21:16 -07002888 if (headroom > skb_headroom(skb))
2889 delta = headroom - skb_headroom(skb);
2890
2891 if (delta || cloned)
2892 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2893 GFP_ATOMIC);
2894 return 0;
2895}
2896
Patrick McHardy334a8132007-06-25 04:35:20 -07002897/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002898 * skb_cow - copy header of skb when it is required
2899 * @skb: buffer to cow
2900 * @headroom: needed headroom
2901 *
2902 * If the skb passed lacks sufficient headroom or its data part
2903 * is shared, data is reallocated. If reallocation fails, an error
2904 * is returned and original skb is not changed.
2905 *
2906 * The result is skb with writable area skb->head...skb->tail
2907 * and at least @headroom of space at head.
2908 */
2909static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2910{
Herbert Xud9cc2042007-09-16 16:21:16 -07002911 return __skb_cow(skb, headroom, skb_cloned(skb));
2912}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002913
Herbert Xud9cc2042007-09-16 16:21:16 -07002914/**
2915 * skb_cow_head - skb_cow but only making the head writable
2916 * @skb: buffer to cow
2917 * @headroom: needed headroom
2918 *
2919 * This function is identical to skb_cow except that we replace the
2920 * skb_cloned check by skb_header_cloned. It should be used when
2921 * you only need to push on some header and do not need to modify
2922 * the data.
2923 */
2924static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2925{
2926 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002927}
2928
2929/**
2930 * skb_padto - pad an skbuff up to a minimal size
2931 * @skb: buffer to pad
2932 * @len: minimal length
2933 *
2934 * Pads up a buffer to ensure the trailing bytes exist and are
2935 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002936 * is untouched. Otherwise it is extended. Returns zero on
2937 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002938 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002939static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002940{
2941 unsigned int size = skb->len;
2942 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002943 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002944 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002945}
2946
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002947/**
2948 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2949 * @skb: buffer to pad
2950 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07002951 * @free_on_error: free buffer on error
2952 *
2953 * Pads up a buffer to ensure the trailing bytes exist and are
2954 * blanked. If the buffer already contains sufficient data it
2955 * is untouched. Otherwise it is extended. Returns zero on
2956 * success. The skb is freed on error if @free_on_error is true.
2957 */
2958static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2959 bool free_on_error)
2960{
2961 unsigned int size = skb->len;
2962
2963 if (unlikely(size < len)) {
2964 len -= size;
2965 if (__skb_pad(skb, len, free_on_error))
2966 return -ENOMEM;
2967 __skb_put(skb, len);
2968 }
2969 return 0;
2970}
2971
2972/**
2973 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2974 * @skb: buffer to pad
2975 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002976 *
2977 * Pads up a buffer to ensure the trailing bytes exist and are
2978 * blanked. If the buffer already contains sufficient data it
2979 * is untouched. Otherwise it is extended. Returns zero on
2980 * success. The skb is freed on error.
2981 */
2982static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2983{
Florian Fainellicd0a1372017-08-22 15:12:14 -07002984 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002985}
2986
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002988 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002989{
2990 const int off = skb->len;
2991
2992 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002993 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002994 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2995 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002996 skb->csum = csum_block_add(skb->csum, csum, off);
2997 return 0;
2998 }
Al Viro15e6cb42016-11-01 22:42:45 -04002999 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003000 return 0;
3001
3002 __skb_trim(skb, off);
3003 return -EFAULT;
3004}
3005
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003006static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3007 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003008{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003009 if (skb_zcopy(skb))
3010 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003011 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00003012 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003013
Ian Campbellea2ab692011-08-22 23:44:58 +00003014 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00003015 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003016 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003017 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003018}
3019
Herbert Xu364c6ba2006-06-09 16:10:40 -07003020static inline int __skb_linearize(struct sk_buff *skb)
3021{
3022 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3023}
3024
Linus Torvalds1da177e2005-04-16 15:20:36 -07003025/**
3026 * skb_linearize - convert paged skb to linear one
3027 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003028 *
3029 * If there is no free memory -ENOMEM is returned, otherwise zero
3030 * is returned and the old skb data released.
3031 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003032static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003034 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3035}
3036
3037/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003038 * skb_has_shared_frag - can any frag be overwritten
3039 * @skb: buffer to test
3040 *
3041 * Return true if the skb has at least one frag that might be modified
3042 * by an external entity (as in vmsplice()/sendfile())
3043 */
3044static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3045{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003046 return skb_is_nonlinear(skb) &&
3047 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003048}
3049
3050/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003051 * skb_linearize_cow - make sure skb is linear and writable
3052 * @skb: buffer to process
3053 *
3054 * If there is no free memory -ENOMEM is returned, otherwise zero
3055 * is returned and the old skb data released.
3056 */
3057static inline int skb_linearize_cow(struct sk_buff *skb)
3058{
3059 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3060 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003061}
3062
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003063static __always_inline void
3064__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3065 unsigned int off)
3066{
3067 if (skb->ip_summed == CHECKSUM_COMPLETE)
3068 skb->csum = csum_block_sub(skb->csum,
3069 csum_partial(start, len, 0), off);
3070 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3071 skb_checksum_start_offset(skb) < 0)
3072 skb->ip_summed = CHECKSUM_NONE;
3073}
3074
Linus Torvalds1da177e2005-04-16 15:20:36 -07003075/**
3076 * skb_postpull_rcsum - update checksum for received skb after pull
3077 * @skb: buffer to update
3078 * @start: start of data before pull
3079 * @len: length of data pulled
3080 *
3081 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003082 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3083 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003084 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003085static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003086 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003087{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003088 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003089}
3090
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003091static __always_inline void
3092__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3093 unsigned int off)
3094{
3095 if (skb->ip_summed == CHECKSUM_COMPLETE)
3096 skb->csum = csum_block_add(skb->csum,
3097 csum_partial(start, len, 0), off);
3098}
Herbert Xucbb042f2006-03-20 22:43:56 -08003099
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003100/**
3101 * skb_postpush_rcsum - update checksum for received skb after push
3102 * @skb: buffer to update
3103 * @start: start of data after push
3104 * @len: length of data pushed
3105 *
3106 * After doing a push on a received packet, you need to call this to
3107 * update the CHECKSUM_COMPLETE checksum.
3108 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003109static inline void skb_postpush_rcsum(struct sk_buff *skb,
3110 const void *start, unsigned int len)
3111{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003112 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003113}
3114
Johannes Bergaf728682017-06-16 14:29:22 +02003115void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003116
David S. Miller7ce5a272013-12-02 17:26:05 -05003117/**
WANG Cong82a31b92016-06-30 10:15:22 -07003118 * skb_push_rcsum - push skb and update receive checksum
3119 * @skb: buffer to update
3120 * @len: length of data pulled
3121 *
3122 * This function performs an skb_push on the packet and updates
3123 * the CHECKSUM_COMPLETE checksum. It should be used on
3124 * receive path processing instead of skb_push unless you know
3125 * that the checksum difference is zero (e.g., a valid IP header)
3126 * or you are setting ip_summed to CHECKSUM_NONE.
3127 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003128static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003129{
3130 skb_push(skb, len);
3131 skb_postpush_rcsum(skb, skb->data, len);
3132 return skb->data;
3133}
3134
3135/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003136 * pskb_trim_rcsum - trim received skb and update checksum
3137 * @skb: buffer to trim
3138 * @len: new length
3139 *
3140 * This is exactly the same as pskb_trim except that it ensures the
3141 * checksum of received packets are still valid after the operation.
3142 */
3143
3144static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3145{
3146 if (likely(len >= skb->len))
3147 return 0;
3148 if (skb->ip_summed == CHECKSUM_COMPLETE)
3149 skb->ip_summed = CHECKSUM_NONE;
3150 return __pskb_trim(skb, len);
3151}
3152
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003153static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3154{
3155 if (skb->ip_summed == CHECKSUM_COMPLETE)
3156 skb->ip_summed = CHECKSUM_NONE;
3157 __skb_trim(skb, len);
3158 return 0;
3159}
3160
3161static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3162{
3163 if (skb->ip_summed == CHECKSUM_COMPLETE)
3164 skb->ip_summed = CHECKSUM_NONE;
3165 return __skb_grow(skb, len);
3166}
3167
Linus Torvalds1da177e2005-04-16 15:20:36 -07003168#define skb_queue_walk(queue, skb) \
3169 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003170 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003171 skb = skb->next)
3172
James Chapman46f89142007-04-30 00:07:31 -07003173#define skb_queue_walk_safe(queue, skb, tmp) \
3174 for (skb = (queue)->next, tmp = skb->next; \
3175 skb != (struct sk_buff *)(queue); \
3176 skb = tmp, tmp = skb->next)
3177
David S. Miller1164f522008-09-23 00:49:44 -07003178#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003179 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003180 skb = skb->next)
3181
3182#define skb_queue_walk_from_safe(queue, skb, tmp) \
3183 for (tmp = skb->next; \
3184 skb != (struct sk_buff *)(queue); \
3185 skb = tmp, tmp = skb->next)
3186
Stephen Hemminger300ce172005-10-30 13:47:34 -08003187#define skb_queue_reverse_walk(queue, skb) \
3188 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003189 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003190 skb = skb->prev)
3191
David S. Miller686a2952011-01-20 22:47:32 -08003192#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3193 for (skb = (queue)->prev, tmp = skb->prev; \
3194 skb != (struct sk_buff *)(queue); \
3195 skb = tmp, tmp = skb->prev)
3196
3197#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3198 for (tmp = skb->prev; \
3199 skb != (struct sk_buff *)(queue); \
3200 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003201
David S. Miller21dc3302010-08-23 00:13:46 -07003202static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003203{
3204 return skb_shinfo(skb)->frag_list != NULL;
3205}
3206
3207static inline void skb_frag_list_init(struct sk_buff *skb)
3208{
3209 skb_shinfo(skb)->frag_list = NULL;
3210}
3211
David S. Milleree039872009-06-09 00:17:13 -07003212#define skb_walk_frags(skb, iter) \
3213 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3214
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003215
3216int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3217 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003218struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3219 struct sk_buff_head *queue,
3220 unsigned int flags,
3221 void (*destructor)(struct sock *sk,
3222 struct sk_buff *skb),
3223 int *peeked, int *off, int *err,
3224 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003225struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003226 void (*destructor)(struct sock *sk,
3227 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003228 int *peeked, int *off, int *err,
3229 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003230struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003231 void (*destructor)(struct sock *sk,
3232 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003233 int *peeked, int *off, int *err);
3234struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3235 int *err);
3236unsigned int datagram_poll(struct file *file, struct socket *sock,
3237 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003238int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3239 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003240static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3241 struct msghdr *msg, int size)
3242{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003243 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003244}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003245int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3246 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003247int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3248 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003249int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003250void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003251void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3252static inline void skb_free_datagram_locked(struct sock *sk,
3253 struct sk_buff *skb)
3254{
3255 __skb_free_datagram_locked(sk, skb, 0);
3256}
Joe Perches7965bd42013-09-26 14:48:15 -07003257int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003258int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3259int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3260__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3261 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003262int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003263 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003264 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003265int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3266 int len);
3267int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003268void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003269unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003270int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3271 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003272void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3273int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3274void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003275unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003276bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003277struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003278struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003279int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003280int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003281int skb_vlan_pop(struct sk_buff *skb);
3282int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003283struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3284 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003285
Al Viro6ce8e9c2014-04-06 21:25:44 -04003286static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3287{
Al Viro3073f072017-02-17 23:13:25 -05003288 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003289}
3290
Al Viro7eab8d92014-04-06 21:51:23 -04003291static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3292{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003293 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003294}
3295
Daniel Borkmann2817a332013-10-30 11:50:51 +01003296struct skb_checksum_ops {
3297 __wsum (*update)(const void *mem, int len, __wsum wsum);
3298 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3299};
3300
Davide Caratti96178132017-05-18 15:44:37 +02003301extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3302
Daniel Borkmann2817a332013-10-30 11:50:51 +01003303__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3304 __wsum csum, const struct skb_checksum_ops *ops);
3305__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3306 __wsum csum);
3307
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003308static inline void * __must_check
3309__skb_header_pointer(const struct sk_buff *skb, int offset,
3310 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003311{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003312 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003313 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003314
David S. Miller690e36e2014-08-23 12:13:41 -07003315 if (!skb ||
3316 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317 return NULL;
3318
3319 return buffer;
3320}
3321
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003322static inline void * __must_check
3323skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003324{
3325 return __skb_header_pointer(skb, offset, len, skb->data,
3326 skb_headlen(skb), buffer);
3327}
3328
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003329/**
3330 * skb_needs_linearize - check if we need to linearize a given skb
3331 * depending on the given device features.
3332 * @skb: socket buffer to check
3333 * @features: net device features
3334 *
3335 * Returns true if either:
3336 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3337 * 2. skb is fragmented and the device does not support SG.
3338 */
3339static inline bool skb_needs_linearize(struct sk_buff *skb,
3340 netdev_features_t features)
3341{
3342 return skb_is_nonlinear(skb) &&
3343 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3344 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3345}
3346
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003347static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3348 void *to,
3349 const unsigned int len)
3350{
3351 memcpy(to, skb->data, len);
3352}
3353
3354static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3355 const int offset, void *to,
3356 const unsigned int len)
3357{
3358 memcpy(to, skb->data + offset, len);
3359}
3360
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003361static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3362 const void *from,
3363 const unsigned int len)
3364{
3365 memcpy(skb->data, from, len);
3366}
3367
3368static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3369 const int offset,
3370 const void *from,
3371 const unsigned int len)
3372{
3373 memcpy(skb->data + offset, from, len);
3374}
3375
Joe Perches7965bd42013-09-26 14:48:15 -07003376void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003377
Patrick Ohlyac45f602009-02-12 05:03:37 +00003378static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3379{
3380 return skb->tstamp;
3381}
3382
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003383/**
3384 * skb_get_timestamp - get timestamp from a skb
3385 * @skb: skb to get stamp from
3386 * @stamp: pointer to struct timeval to store stamp in
3387 *
3388 * Timestamps are stored in the skb as offsets to a base timestamp.
3389 * This function converts the offset back to a struct timeval and stores
3390 * it in stamp.
3391 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003392static inline void skb_get_timestamp(const struct sk_buff *skb,
3393 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003394{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003395 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003396}
3397
Patrick Ohlyac45f602009-02-12 05:03:37 +00003398static inline void skb_get_timestampns(const struct sk_buff *skb,
3399 struct timespec *stamp)
3400{
3401 *stamp = ktime_to_timespec(skb->tstamp);
3402}
3403
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003404static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003405{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003406 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003407}
3408
Stephen Hemminger164891a2007-04-23 22:26:16 -07003409static inline ktime_t net_timedelta(ktime_t t)
3410{
3411 return ktime_sub(ktime_get_real(), t);
3412}
3413
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003414static inline ktime_t net_invalid_timestamp(void)
3415{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003416 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003417}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003418
Alexander Duyck62bccb82014-09-04 13:31:35 -04003419struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3420
Richard Cochranc1f19b52010-07-17 08:49:36 +00003421#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3422
Joe Perches7965bd42013-09-26 14:48:15 -07003423void skb_clone_tx_timestamp(struct sk_buff *skb);
3424bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003425
3426#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3427
3428static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3429{
3430}
3431
3432static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3433{
3434 return false;
3435}
3436
3437#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3438
3439/**
3440 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3441 *
Richard Cochranda92b192011-10-21 00:49:15 +00003442 * PHY drivers may accept clones of transmitted packets for
3443 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003444 * must call this function to return the skb back to the stack with a
3445 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003446 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003447 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003448 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003449 *
3450 */
3451void skb_complete_tx_timestamp(struct sk_buff *skb,
3452 struct skb_shared_hwtstamps *hwtstamps);
3453
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003454void __skb_tstamp_tx(struct sk_buff *orig_skb,
3455 struct skb_shared_hwtstamps *hwtstamps,
3456 struct sock *sk, int tstype);
3457
Patrick Ohlyac45f602009-02-12 05:03:37 +00003458/**
3459 * skb_tstamp_tx - queue clone of skb with send time stamps
3460 * @orig_skb: the original outgoing packet
3461 * @hwtstamps: hardware time stamps, may be NULL if not available
3462 *
3463 * If the skb has a socket associated, then this function clones the
3464 * skb (thus sharing the actual data and optional structures), stores
3465 * the optional hardware time stamping information (if non NULL) or
3466 * generates a software time stamp (otherwise), then queues the clone
3467 * to the error queue of the socket. Errors are silently ignored.
3468 */
Joe Perches7965bd42013-09-26 14:48:15 -07003469void skb_tstamp_tx(struct sk_buff *orig_skb,
3470 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003471
Richard Cochran4507a712010-07-17 08:48:28 +00003472/**
3473 * skb_tx_timestamp() - Driver hook for transmit timestamping
3474 *
3475 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003476 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003477 *
David S. Miller73409f32013-12-27 13:04:33 -05003478 * Specifically, one should make absolutely sure that this function is
3479 * called before TX completion of this packet can trigger. Otherwise
3480 * the packet could potentially already be freed.
3481 *
Richard Cochran4507a712010-07-17 08:48:28 +00003482 * @skb: A socket buffer.
3483 */
3484static inline void skb_tx_timestamp(struct sk_buff *skb)
3485{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003486 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003487 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3488 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003489}
3490
Johannes Berg6e3e9392011-11-09 10:15:42 +01003491/**
3492 * skb_complete_wifi_ack - deliver skb with wifi status
3493 *
3494 * @skb: the original outgoing packet
3495 * @acked: ack status
3496 *
3497 */
3498void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3499
Joe Perches7965bd42013-09-26 14:48:15 -07003500__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3501__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003502
Herbert Xu60476372007-04-09 11:59:39 -07003503static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3504{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003505 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3506 skb->csum_valid ||
3507 (skb->ip_summed == CHECKSUM_PARTIAL &&
3508 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003509}
3510
Herbert Xufb286bb2005-11-10 13:01:24 -08003511/**
3512 * skb_checksum_complete - Calculate checksum of an entire packet
3513 * @skb: packet to process
3514 *
3515 * This function calculates the checksum over the entire packet plus
3516 * the value of skb->csum. The latter can be used to supply the
3517 * checksum of a pseudo header as used by TCP/UDP. It returns the
3518 * checksum.
3519 *
3520 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3521 * this function can be used to verify that checksum on received
3522 * packets. In that case the function should return zero if the
3523 * checksum is correct. In particular, this function will return zero
3524 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3525 * hardware has already verified the correctness of the checksum.
3526 */
Al Viro4381ca32007-07-15 21:00:11 +01003527static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003528{
Herbert Xu60476372007-04-09 11:59:39 -07003529 return skb_csum_unnecessary(skb) ?
3530 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003531}
3532
Tom Herbert77cffe22014-08-27 21:26:46 -07003533static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3534{
3535 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3536 if (skb->csum_level == 0)
3537 skb->ip_summed = CHECKSUM_NONE;
3538 else
3539 skb->csum_level--;
3540 }
3541}
3542
3543static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3544{
3545 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3546 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3547 skb->csum_level++;
3548 } else if (skb->ip_summed == CHECKSUM_NONE) {
3549 skb->ip_summed = CHECKSUM_UNNECESSARY;
3550 skb->csum_level = 0;
3551 }
3552}
3553
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003554/* Check if we need to perform checksum complete validation.
3555 *
3556 * Returns true if checksum complete is needed, false otherwise
3557 * (either checksum is unnecessary or zero checksum is allowed).
3558 */
3559static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3560 bool zero_okay,
3561 __sum16 check)
3562{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003563 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3564 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003565 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003566 return false;
3567 }
3568
3569 return true;
3570}
3571
3572/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3573 * in checksum_init.
3574 */
3575#define CHECKSUM_BREAK 76
3576
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003577/* Unset checksum-complete
3578 *
3579 * Unset checksum complete can be done when packet is being modified
3580 * (uncompressed for instance) and checksum-complete value is
3581 * invalidated.
3582 */
3583static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3584{
3585 if (skb->ip_summed == CHECKSUM_COMPLETE)
3586 skb->ip_summed = CHECKSUM_NONE;
3587}
3588
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003589/* Validate (init) checksum based on checksum complete.
3590 *
3591 * Return values:
3592 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3593 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3594 * checksum is stored in skb->csum for use in __skb_checksum_complete
3595 * non-zero: value of invalid checksum
3596 *
3597 */
3598static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3599 bool complete,
3600 __wsum psum)
3601{
3602 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3603 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003604 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003605 return 0;
3606 }
3607 }
3608
3609 skb->csum = psum;
3610
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003611 if (complete || skb->len <= CHECKSUM_BREAK) {
3612 __sum16 csum;
3613
3614 csum = __skb_checksum_complete(skb);
3615 skb->csum_valid = !csum;
3616 return csum;
3617 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003618
3619 return 0;
3620}
3621
3622static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3623{
3624 return 0;
3625}
3626
3627/* Perform checksum validate (init). Note that this is a macro since we only
3628 * want to calculate the pseudo header which is an input function if necessary.
3629 * First we try to validate without any computation (checksum unnecessary) and
3630 * then calculate based on checksum complete calling the function to compute
3631 * pseudo header.
3632 *
3633 * Return values:
3634 * 0: checksum is validated or try to in skb_checksum_complete
3635 * non-zero: value of invalid checksum
3636 */
3637#define __skb_checksum_validate(skb, proto, complete, \
3638 zero_okay, check, compute_pseudo) \
3639({ \
3640 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003641 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003642 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3643 __ret = __skb_checksum_validate_complete(skb, \
3644 complete, compute_pseudo(skb, proto)); \
3645 __ret; \
3646})
3647
3648#define skb_checksum_init(skb, proto, compute_pseudo) \
3649 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3650
3651#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3652 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3653
3654#define skb_checksum_validate(skb, proto, compute_pseudo) \
3655 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3656
3657#define skb_checksum_validate_zero_check(skb, proto, check, \
3658 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003659 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003660
3661#define skb_checksum_simple_validate(skb) \
3662 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3663
Tom Herbertd96535a2014-08-31 15:12:42 -07003664static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3665{
Davide Caratti219f1d792017-05-18 15:44:39 +02003666 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003667}
3668
3669static inline void __skb_checksum_convert(struct sk_buff *skb,
3670 __sum16 check, __wsum pseudo)
3671{
3672 skb->csum = ~pseudo;
3673 skb->ip_summed = CHECKSUM_COMPLETE;
3674}
3675
3676#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3677do { \
3678 if (__skb_checksum_convert_check(skb)) \
3679 __skb_checksum_convert(skb, check, \
3680 compute_pseudo(skb, proto)); \
3681} while (0)
3682
Tom Herbert15e23962015-02-10 16:30:31 -08003683static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3684 u16 start, u16 offset)
3685{
3686 skb->ip_summed = CHECKSUM_PARTIAL;
3687 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3688 skb->csum_offset = offset - start;
3689}
3690
Tom Herbertdcdc8992015-02-02 16:07:34 -08003691/* Update skbuf and packet to reflect the remote checksum offload operation.
3692 * When called, ptr indicates the starting point for skb->csum when
3693 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3694 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3695 */
3696static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003697 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003698{
3699 __wsum delta;
3700
Tom Herbert15e23962015-02-10 16:30:31 -08003701 if (!nopartial) {
3702 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3703 return;
3704 }
3705
Tom Herbertdcdc8992015-02-02 16:07:34 -08003706 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3707 __skb_checksum_complete(skb);
3708 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3709 }
3710
3711 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3712
3713 /* Adjust skb->csum since we changed the packet */
3714 skb->csum = csum_add(skb->csum, delta);
3715}
3716
Florian Westphalcb9c6832017-01-23 18:21:56 +01003717static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3718{
3719#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003720 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003721#else
3722 return NULL;
3723#endif
3724}
3725
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003726#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003727void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003728static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3729{
3730 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003731 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003732}
3733static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3734{
3735 if (nfct)
3736 atomic_inc(&nfct->use);
3737}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003738#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003739#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003740static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3741{
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003742 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003743 kfree(nf_bridge);
3744}
3745static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3746{
3747 if (nf_bridge)
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003748 refcount_inc(&nf_bridge->use);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003749}
3750#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003751static inline void nf_reset(struct sk_buff *skb)
3752{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003753#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003754 nf_conntrack_put(skb_nfct(skb));
3755 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003756#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003757#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003758 nf_bridge_put(skb->nf_bridge);
3759 skb->nf_bridge = NULL;
3760#endif
3761}
3762
Patrick McHardy124dff02013-04-05 20:42:05 +02003763static inline void nf_reset_trace(struct sk_buff *skb)
3764{
Florian Westphal478b3602014-02-15 23:48:45 +01003765#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003766 skb->nf_trace = 0;
3767#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003768}
3769
Ye Yin2b5ec1a2017-10-26 16:57:05 +08003770static inline void ipvs_reset(struct sk_buff *skb)
3771{
3772#if IS_ENABLED(CONFIG_IP_VS)
3773 skb->ipvs_property = 0;
3774#endif
3775}
3776
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003777/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003778static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3779 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003780{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003781#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003782 dst->_nfct = src->_nfct;
3783 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003784#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003785#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003786 dst->nf_bridge = src->nf_bridge;
3787 nf_bridge_get(src->nf_bridge);
3788#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003789#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003790 if (copy)
3791 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003792#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003793}
3794
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003795static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3796{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003797#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003798 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003799#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003800#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003801 nf_bridge_put(dst->nf_bridge);
3802#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003803 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003804}
3805
James Morris984bc162006-06-09 00:29:17 -07003806#ifdef CONFIG_NETWORK_SECMARK
3807static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3808{
3809 to->secmark = from->secmark;
3810}
3811
3812static inline void skb_init_secmark(struct sk_buff *skb)
3813{
3814 skb->secmark = 0;
3815}
3816#else
3817static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3818{ }
3819
3820static inline void skb_init_secmark(struct sk_buff *skb)
3821{ }
3822#endif
3823
Eric W. Biederman574f7192014-04-01 12:20:24 -07003824static inline bool skb_irq_freeable(const struct sk_buff *skb)
3825{
3826 return !skb->destructor &&
3827#if IS_ENABLED(CONFIG_XFRM)
3828 !skb->sp &&
3829#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003830 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003831 !skb->_skb_refdst &&
3832 !skb_has_frag_list(skb);
3833}
3834
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003835static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3836{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003837 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003838}
3839
Stephen Hemminger92477442009-03-21 13:39:26 -07003840static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003841{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003842 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003843}
3844
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003845static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3846{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003847 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003848}
3849
David S. Millerd5a9e242009-01-27 16:22:11 -08003850static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3851{
3852 skb->queue_mapping = rx_queue + 1;
3853}
3854
Stephen Hemminger92477442009-03-21 13:39:26 -07003855static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003856{
3857 return skb->queue_mapping - 1;
3858}
3859
Stephen Hemminger92477442009-03-21 13:39:26 -07003860static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003861{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003862 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003863}
3864
Julian Anastasov4ff06202017-02-06 23:14:12 +02003865static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3866{
3867 skb->dst_pending_confirm = val;
3868}
3869
3870static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3871{
3872 return skb->dst_pending_confirm != 0;
3873}
3874
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003875static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3876{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003877#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003878 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003879#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003880 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003881#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003882}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003883
Pravin B Shelar68c33162013-02-14 14:02:41 +00003884/* Keeps track of mac header offset relative to skb->head.
3885 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3886 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003887 * tunnel skb it points to outer mac header.
3888 * Keeps track of level of encapsulation of network headers.
3889 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003890struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003891 union {
3892 int mac_offset;
3893 int data_offset;
3894 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003895 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003896 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003897 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003898};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003899#define SKB_SGO_CB_OFFSET 32
3900#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003901
3902static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3903{
3904 return (skb_mac_header(inner_skb) - inner_skb->head) -
3905 SKB_GSO_CB(inner_skb)->mac_offset;
3906}
3907
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003908static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3909{
3910 int new_headroom, headroom;
3911 int ret;
3912
3913 headroom = skb_headroom(skb);
3914 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3915 if (ret)
3916 return ret;
3917
3918 new_headroom = skb_headroom(skb);
3919 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3920 return 0;
3921}
3922
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003923static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3924{
3925 /* Do not update partial checksums if remote checksum is enabled. */
3926 if (skb->remcsum_offload)
3927 return;
3928
3929 SKB_GSO_CB(skb)->csum = res;
3930 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3931}
3932
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003933/* Compute the checksum for a gso segment. First compute the checksum value
3934 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3935 * then add in skb->csum (checksum from csum_start to end of packet).
3936 * skb->csum and csum_start are then updated to reflect the checksum of the
3937 * resultant packet starting from the transport header-- the resultant checksum
3938 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3939 * header.
3940 */
3941static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3942{
Alexander Duyck76443452016-02-05 15:27:37 -08003943 unsigned char *csum_start = skb_transport_header(skb);
3944 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3945 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003946
Alexander Duyck76443452016-02-05 15:27:37 -08003947 SKB_GSO_CB(skb)->csum = res;
3948 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003949
Alexander Duyck76443452016-02-05 15:27:37 -08003950 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003951}
3952
David S. Millerbdcc0922012-03-07 20:53:36 -05003953static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003954{
3955 return skb_shinfo(skb)->gso_size;
3956}
3957
Eric Dumazet36a8f392013-09-29 01:21:32 -07003958/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003959static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003960{
3961 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3962}
3963
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003964static inline void skb_gso_reset(struct sk_buff *skb)
3965{
3966 skb_shinfo(skb)->gso_size = 0;
3967 skb_shinfo(skb)->gso_segs = 0;
3968 skb_shinfo(skb)->gso_type = 0;
3969}
3970
Joe Perches7965bd42013-09-26 14:48:15 -07003971void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003972
3973static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3974{
3975 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3976 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003977 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3978
Alexander Duyckb78462e2010-06-02 12:24:37 +00003979 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3980 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003981 __skb_warn_lro_forwarding(skb);
3982 return true;
3983 }
3984 return false;
3985}
3986
Herbert Xu35fc92a2007-03-26 23:22:20 -07003987static inline void skb_forward_csum(struct sk_buff *skb)
3988{
3989 /* Unfortunately we don't support this one. Any brave souls? */
3990 if (skb->ip_summed == CHECKSUM_COMPLETE)
3991 skb->ip_summed = CHECKSUM_NONE;
3992}
3993
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003994/**
3995 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3996 * @skb: skb to check
3997 *
3998 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3999 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4000 * use this helper, to document places where we make this assertion.
4001 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004002static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004003{
4004#ifdef DEBUG
4005 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4006#endif
4007}
4008
Rusty Russellf35d9d82008-02-04 23:49:54 -05004009bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004010
Paul Durranted1f50c2014-01-09 10:02:46 +00004011int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004012struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4013 unsigned int transport_len,
4014 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004015
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004016/**
4017 * skb_head_is_locked - Determine if the skb->head is locked down
4018 * @skb: skb to check
4019 *
4020 * The head on skbs build around a head frag can be removed if they are
4021 * not cloned. This function returns true if the skb head is locked down
4022 * due to either being allocated via kmalloc, or by being a clone with
4023 * multiple references to the head.
4024 */
4025static inline bool skb_head_is_locked(const struct sk_buff *skb)
4026{
4027 return !skb->head_frag || skb_cloned(skb);
4028}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004029
4030/**
4031 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4032 *
4033 * @skb: GSO skb
4034 *
4035 * skb_gso_network_seglen is used to determine the real size of the
4036 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4037 *
4038 * The MAC/L2 header is not accounted for.
4039 */
4040static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4041{
4042 unsigned int hdr_len = skb_transport_header(skb) -
4043 skb_network_header(skb);
4044 return hdr_len + skb_gso_transport_seglen(skb);
4045}
Thomas Grafee122c72015-07-21 10:43:58 +02004046
Edward Cree179bc672016-02-11 20:48:04 +00004047/* Local Checksum Offload.
4048 * Compute outer checksum based on the assumption that the
4049 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00004050 * See Documentation/networking/checksum-offloads.txt for
4051 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004052 * Fill in outer checksum adjustment (e.g. with sum of outer
4053 * pseudo-header) before calling.
4054 * Also ensure that inner checksum is in linear data area.
4055 */
4056static inline __wsum lco_csum(struct sk_buff *skb)
4057{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004058 unsigned char *csum_start = skb_checksum_start(skb);
4059 unsigned char *l4_hdr = skb_transport_header(skb);
4060 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004061
4062 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004063 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4064 skb->csum_offset));
4065
Edward Cree179bc672016-02-11 20:48:04 +00004066 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004067 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004068 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004069 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004070}
4071
Linus Torvalds1da177e2005-04-16 15:20:36 -07004072#endif /* __KERNEL__ */
4073#endif /* _LINUX_SKBUFF_H */