blob: bc486ef23f20f91ce3ed183e935399d1e4c55e18 [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 Ry5f74f822016-02-03 09:26:57 +0100306extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000316 struct {
317 struct page *p;
318 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700320 __u32 page_offset;
321 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326};
327
Eric Dumazet9e903e02011-10-18 21:00:24 +0000328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
Willem de Bruijnc613c202017-07-31 08:15:47 -0400348static inline bool skb_frag_must_loop(struct page *p)
349{
350#if defined(CONFIG_HIGHMEM)
351 if (PageHighMem(p))
352 return true;
353#endif
354 return false;
355}
356
357/**
358 * skb_frag_foreach_page - loop over pages in a fragment
359 *
360 * @f: skb frag to operate on
361 * @f_off: offset from start of f->page.p
362 * @f_len: length from f_off to loop over
363 * @p: (temp var) current page
364 * @p_off: (temp var) offset from start of current page,
365 * non-zero only on first page.
366 * @p_len: (temp var) length in current page,
367 * < PAGE_SIZE only on first and last page.
368 * @copied: (temp var) length so far, excluding current p_len.
369 *
370 * A fragment can hold a compound page, in which case per-page
371 * operations, notably kmap_atomic, must be called for each
372 * regular page.
373 */
374#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
375 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
376 p_off = (f_off) & (PAGE_SIZE - 1), \
377 p_len = skb_frag_must_loop(p) ? \
378 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
379 copied = 0; \
380 copied < f_len; \
381 copied += p_len, p++, p_off = 0, \
382 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
383
Patrick Ohlyac45f602009-02-12 05:03:37 +0000384#define HAVE_HW_TIME_STAMP
385
386/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800387 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000388 * @hwtstamp: hardware time stamp transformed into duration
389 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000390 *
391 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400392 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000393 *
394 * hwtstamps can only be compared against other hwtstamps from
395 * the same device.
396 *
397 * This structure is attached to packets as part of the
398 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
399 */
400struct skb_shared_hwtstamps {
401 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000402};
403
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000404/* Definitions for tx_flags in struct skb_shared_info */
405enum {
406 /* generate hardware time stamp */
407 SKBTX_HW_TSTAMP = 1 << 0,
408
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400409 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000410 SKBTX_SW_TSTAMP = 1 << 1,
411
412 /* device driver is going to provide hardware time stamp */
413 SKBTX_IN_PROGRESS = 1 << 2,
414
Shirley Maa6686f22011-07-06 12:22:12 +0000415 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000416 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100417
418 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000419 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000420
421 /* This indicates at least one fragment might be overwritten
422 * (as in vmsplice(), sendfile() ...)
423 * If we need to compute a TX checksum, we'll need to copy
424 * all frags to avoid possible bad checksum
425 */
426 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400427
428 /* generate software time stamp when entering packet scheduling */
429 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000430};
431
Willem de Bruijn52267792017-08-03 16:29:39 -0400432#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400433#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400434 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400435#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
436
Shirley Maa6686f22011-07-06 12:22:12 +0000437/*
438 * The callback notifies userspace to release buffers when skb DMA is done in
439 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000440 * The zerocopy_success argument is true if zero copy transmit occurred,
441 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000442 * The ctx field is used to track device context.
443 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000444 */
445struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000446 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400447 union {
448 struct {
449 unsigned long desc;
450 void *ctx;
451 };
452 struct {
453 u32 id;
454 u16 len;
455 u16 zerocopy:1;
456 u32 bytelen;
457 };
458 };
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 {
Daniel Borkmannde8f3a82017-09-25 02:25:51 +0200491 __u8 __unused;
492 __u8 meta_len;
493 __u8 nr_frags;
Ian Campbell9f42f122012-01-05 07:13:39 +0000494 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700495 unsigned short gso_size;
496 /* Warning: this field is not always filled in (UFO)! */
497 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000499 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200500 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400501 u32 tskey;
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,
Willem de Bruijn0c19f8462017-11-21 10:22:25 -0500571
572 SKB_GSO_UDP = 1 << 16,
Herbert Xu79671682006-06-22 02:40:14 -0700573};
574
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700575#if BITS_PER_LONG > 32
576#define NET_SKBUFF_DATA_USES_OFFSET 1
577#endif
578
579#ifdef NET_SKBUFF_DATA_USES_OFFSET
580typedef unsigned int sk_buff_data_t;
581#else
582typedef unsigned char *sk_buff_data_t;
583#endif
584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585/**
586 * struct sk_buff - socket buffer
587 * @next: Next buffer in list
588 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800589 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800590 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700591 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700593 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000594 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700595 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 * @len: Length of actual data
597 * @data_len: Data length
598 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700599 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700600 * @csum: Checksum (must include start/offset pair)
601 * @csum_start: Offset from skb->head where checksumming should start
602 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700603 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700604 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700606 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 * @nohdr: Payload reference only, must not modify header
608 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700609 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700610 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500611 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500612 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500613 * @tc_redirected: packet was redirected by a tc action
614 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800615 * @peeked: this packet has been seen already, so stats have been
616 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700617 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700618 * @protocol: Packet protocol from driver
619 * @destructor: Destruct function
Eric Dumazete2080072017-10-04 12:59:58 -0700620 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
Florian Westphala9e419d2017-01-23 18:21:59 +0100621 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800623 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700625 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700626 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700627 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700628 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700629 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700630 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700631 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700632 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100633 * @wifi_acked_valid: wifi_acked was set
634 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000635 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200636 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200637 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300638 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700639 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700640 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000641 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700642 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000643 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000644 * @inner_transport_header: Inner transport layer header (encapsulation)
645 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000646 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700647 * @transport_header: Transport layer header
648 * @network_header: Network layer header
649 * @mac_header: Link layer header
650 * @tail: Tail pointer
651 * @end: End pointer
652 * @head: Head of buffer
653 * @data: Data head pointer
654 * @truesize: Buffer size
655 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 */
657
658struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800659 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800660 struct {
661 /* These two members must be first. */
662 struct sk_buff *next;
663 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000664
Eric Dumazet56b17422014-11-03 08:19:53 -0800665 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700666 struct net_device *dev;
667 /* Some protocols might use this space to store information,
668 * while device pointer would be NULL.
669 * UDP receive path is one user.
670 */
671 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800672 };
673 };
674 struct rb_node rbnode; /* used in netem & tcp stack */
675 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000676 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Eric Dumazetc84d9492016-12-08 11:41:55 -0800678 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700679 ktime_t tstamp;
680 u64 skb_mstamp;
Eric Dumazetc84d9492016-12-08 11:41:55 -0800681 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 /*
683 * This is the control buffer. It is free to use for every
684 * layer. Please put your private variables there. If you
685 * want to keep them across layers you have to do a skb_clone()
686 * first. This is owned by whoever has the skb queued ATM.
687 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000688 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Eric Dumazete2080072017-10-04 12:59:58 -0700690 union {
691 struct {
692 unsigned long _skb_refdst;
693 void (*destructor)(struct sk_buff *skb);
694 };
695 struct list_head tcp_tsorted_anchor;
696 };
697
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000698#ifdef CONFIG_XFRM
699 struct sec_path *sp;
700#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700701#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100702 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700703#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200704#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700705 struct nf_bridge_info *nf_bridge;
706#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700708 data_len;
709 __u16 mac_len,
710 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700711
712 /* Following fields are _not_ copied in __copy_skb_header()
713 * Note that queue_mapping is here mostly to fill a hole.
714 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700715 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200716
717/* if you move cloned around you also must adapt those constants */
718#ifdef __BIG_ENDIAN_BITFIELD
719#define CLONED_MASK (1 << 7)
720#else
721#define CLONED_MASK 1
722#endif
723#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
724
725 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700726 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700727 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800728 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800729 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700730 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200731 xmit_more:1,
732 __unused:1; /* one bit hole */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733
Eric Dumazetb1937222014-09-28 22:18:47 -0700734 /* fields enclosed in headers_start/headers_end are copied
735 * using a single memcpy() in __copy_skb_header()
736 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700737 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700738 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700739 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200740
741/* if you move pkt_type around you also must adapt those constants */
742#ifdef __BIG_ENDIAN_BITFIELD
743#define PKT_TYPE_MAX (7 << 5)
744#else
745#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100746#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200747#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
748
749 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700750 __u8 pkt_type:3;
751 __u8 pfmemalloc:1;
752 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
Eric Dumazetb1937222014-09-28 22:18:47 -0700754 __u8 nf_trace:1;
755 __u8 ip_summed:2;
756 __u8 ooo_okay:1;
757 __u8 l4_hash:1;
758 __u8 sw_hash:1;
759 __u8 wifi_acked_valid:1;
760 __u8 wifi_acked:1;
761
762 __u8 no_fcs:1;
763 /* Indicates the inner headers are valid in the skbuff. */
764 __u8 encapsulation:1;
765 __u8 encap_hdr_csum:1;
766 __u8 csum_valid:1;
767 __u8 csum_complete_sw:1;
768 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200769 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700770
Julian Anastasov4ff06202017-02-06 23:14:12 +0200771 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700772#ifdef CONFIG_IPV6_NDISC_NODETYPE
773 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700775 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700776 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800777 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200778#ifdef CONFIG_NET_SWITCHDEV
779 __u8 offload_fwd_mark:1;
Yotam Gigiabf4bb62017-10-03 09:58:06 +0200780 __u8 offload_mr_fwd_mark:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200781#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500782#ifdef CONFIG_NET_CLS_ACT
783 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500784 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500785 __u8 tc_redirected:1;
786 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500787#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000788
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700790 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200792
Eric Dumazetb1937222014-09-28 22:18:47 -0700793 union {
794 __wsum csum;
795 struct {
796 __u16 csum_start;
797 __u16 csum_offset;
798 };
799 };
800 __u32 priority;
801 int skb_iif;
802 __u32 hash;
803 __be16 vlan_proto;
804 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800805#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
806 union {
807 unsigned int napi_id;
808 unsigned int sender_cpu;
809 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700810#endif
James Morris984bc162006-06-09 00:29:17 -0700811#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200812 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700813#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700814
Neil Horman3b885782009-10-12 13:26:31 -0700815 union {
816 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000817 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700818 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
Tom Herbert8bce6d72014-09-29 20:22:29 -0700820 union {
821 __be16 inner_protocol;
822 __u8 inner_ipproto;
823 };
824
Simon Horman1a37e412013-05-23 21:02:51 +0000825 __u16 inner_transport_header;
826 __u16 inner_network_header;
827 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700828
829 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000830 __u16 transport_header;
831 __u16 network_header;
832 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700833
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700834 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700835 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700836 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700837
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700839 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700840 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700842 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700843 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300844 refcount_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845};
846
847#ifdef __KERNEL__
848/*
849 * Handling routines are only of interest to the kernel
850 */
851#include <linux/slab.h>
852
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Mel Gormanc93bdd02012-07-31 16:44:19 -0700854#define SKB_ALLOC_FCLONE 0x01
855#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800856#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700857
858/* Returns true if the skb was allocated from PFMEMALLOC reserves */
859static inline bool skb_pfmemalloc(const struct sk_buff *skb)
860{
861 return unlikely(skb->pfmemalloc);
862}
863
Eric Dumazet7fee2262010-05-11 23:19:48 +0000864/*
865 * skb might have a dst pointer attached, refcounted or not.
866 * _skb_refdst low order bit is set if refcount was _not_ taken
867 */
868#define SKB_DST_NOREF 1UL
869#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
870
Florian Westphala9e419d2017-01-23 18:21:59 +0100871#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000872/**
873 * skb_dst - returns skb dst_entry
874 * @skb: buffer
875 *
876 * Returns skb dst_entry, regardless of reference taken or not.
877 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000878static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
879{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000880 /* If refdst was not refcounted, check we still are in a
881 * rcu_read_lock section
882 */
883 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
884 !rcu_read_lock_held() &&
885 !rcu_read_lock_bh_held());
886 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000887}
888
Eric Dumazet7fee2262010-05-11 23:19:48 +0000889/**
890 * skb_dst_set - sets skb dst
891 * @skb: buffer
892 * @dst: dst entry
893 *
894 * Sets skb dst, assuming a reference was taken on dst and should
895 * be released by skb_dst_drop()
896 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000897static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
898{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000899 skb->_skb_refdst = (unsigned long)dst;
900}
901
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200902/**
903 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
904 * @skb: buffer
905 * @dst: dst entry
906 *
907 * Sets skb dst, assuming a reference was not taken on dst.
908 * If dst entry is cached, we do not take reference and dst_release
909 * will be avoided by refdst_drop. If dst entry is not cached, we take
910 * reference, so that last dst_release can destroy the dst immediately.
911 */
912static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
913{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100914 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
915 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200916}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000917
918/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300919 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000920 * @skb: buffer
921 */
922static inline bool skb_dst_is_noref(const struct sk_buff *skb)
923{
924 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000925}
926
Eric Dumazet511c3f92009-06-02 05:14:27 +0000927static inline struct rtable *skb_rtable(const struct sk_buff *skb)
928{
Eric Dumazetadf30902009-06-02 05:19:30 +0000929 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000930}
931
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400932/* For mangling skb->pkt_type from user space side from applications
933 * such as nft, tc, etc, we only allow a conservative subset of
934 * possible pkt_types to be set.
935*/
936static inline bool skb_pkt_type_ok(u32 ptype)
937{
938 return ptype <= PACKET_OTHERHOST;
939}
940
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200941static inline unsigned int skb_napi_id(const struct sk_buff *skb)
942{
943#ifdef CONFIG_NET_RX_BUSY_POLL
944 return skb->napi_id;
945#else
946 return 0;
947#endif
948}
949
Paolo Abeni3889a8032017-06-12 11:23:41 +0200950/* decrement the reference count and return true if we can free the skb */
951static inline bool skb_unref(struct sk_buff *skb)
952{
953 if (unlikely(!skb))
954 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +0300955 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200956 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +0300957 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200958 return false;
959
960 return true;
961}
962
Paolo Abeni0a463c72017-06-12 11:23:42 +0200963void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700964void kfree_skb(struct sk_buff *skb);
965void kfree_skb_list(struct sk_buff *segs);
966void skb_tx_error(struct sk_buff *skb);
967void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +0200968void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700969void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000970extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000971
Joe Perches7965bd42013-09-26 14:48:15 -0700972void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
973bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
974 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000975
Joe Perches7965bd42013-09-26 14:48:15 -0700976struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
977 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700978struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700979struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700980static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100981 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700982{
Eric Dumazet564824b2010-10-11 19:05:25 +0000983 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700984}
985
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700986struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
987 unsigned long data_len,
988 int max_page_order,
989 int *errcode,
990 gfp_t gfp_mask);
991
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700992/* Layout of fast clones : [skb1][skb2][fclone_ref] */
993struct sk_buff_fclones {
994 struct sk_buff skb1;
995
996 struct sk_buff skb2;
997
Reshetova, Elena26385952017-06-30 13:07:59 +0300998 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700999};
1000
1001/**
1002 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -07001003 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001004 * @skb: buffer
1005 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001006 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001007 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1008 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001009 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001010static inline bool skb_fclone_busy(const struct sock *sk,
1011 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001012{
1013 const struct sk_buff_fclones *fclones;
1014
1015 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1016
1017 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001018 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001019 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001020}
1021
David S. Millerd179cd12005-08-17 14:57:30 -07001022static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001023 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001024{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001025 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001026}
1027
Joe Perches7965bd42013-09-26 14:48:15 -07001028struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1029int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1030struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1031struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001032struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1033 gfp_t gfp_mask, bool fclone);
1034static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1035 gfp_t gfp_mask)
1036{
1037 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1038}
Eric Dumazet117632e2011-12-03 21:39:53 +00001039
Joe Perches7965bd42013-09-26 14:48:15 -07001040int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1041struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1042 unsigned int headroom);
1043struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1044 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001045int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1046 int offset, int len);
1047int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1048 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001049int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001050int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1051
1052/**
1053 * skb_pad - zero pad the tail of an skb
1054 * @skb: buffer to pad
1055 * @pad: space to pad
1056 *
1057 * Ensure that a buffer is followed by a padding area that is zero
1058 * filled. Used by network drivers which may DMA or transfer data
1059 * beyond the buffer end onto the wire.
1060 *
1061 * May return error in out of memory cases. The skb is freed on error.
1062 */
1063static inline int skb_pad(struct sk_buff *skb, int pad)
1064{
1065 return __skb_pad(skb, pad, true);
1066}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001067#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
Joe Perches7965bd42013-09-26 14:48:15 -07001069int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1070 int getfrag(void *from, char *to, int offset,
1071 int len, int odd, struct sk_buff *skb),
1072 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001073
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001074int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1075 int offset, size_t size);
1076
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001077struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001078 __u32 lower_offset;
1079 __u32 upper_offset;
1080 __u32 frag_idx;
1081 __u32 stepped_offset;
1082 struct sk_buff *root_skb;
1083 struct sk_buff *cur_skb;
1084 __u8 *frag_data;
1085};
1086
Joe Perches7965bd42013-09-26 14:48:15 -07001087void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1088 unsigned int to, struct skb_seq_state *st);
1089unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1090 struct skb_seq_state *st);
1091void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001092
Joe Perches7965bd42013-09-26 14:48:15 -07001093unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001094 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001095
Tom Herbert09323cc2013-12-15 22:16:19 -08001096/*
1097 * Packet hash types specify the type of hash in skb_set_hash.
1098 *
1099 * Hash types refer to the protocol layer addresses which are used to
1100 * construct a packet's hash. The hashes are used to differentiate or identify
1101 * flows of the protocol layer for the hash type. Hash types are either
1102 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1103 *
1104 * Properties of hashes:
1105 *
1106 * 1) Two packets in different flows have different hash values
1107 * 2) Two packets in the same flow should have the same hash value
1108 *
1109 * A hash at a higher layer is considered to be more specific. A driver should
1110 * set the most specific hash possible.
1111 *
1112 * A driver cannot indicate a more specific hash than the layer at which a hash
1113 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1114 *
1115 * A driver may indicate a hash level which is less specific than the
1116 * actual layer the hash was computed on. For instance, a hash computed
1117 * at L4 may be considered an L3 hash. This should only be done if the
1118 * driver can't unambiguously determine that the HW computed the hash at
1119 * the higher layer. Note that the "should" in the second property above
1120 * permits this.
1121 */
1122enum pkt_hash_types {
1123 PKT_HASH_TYPE_NONE, /* Undefined type */
1124 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1125 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1126 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1127};
1128
Tom Herbertbcc83832015-09-01 09:24:24 -07001129static inline void skb_clear_hash(struct sk_buff *skb)
1130{
1131 skb->hash = 0;
1132 skb->sw_hash = 0;
1133 skb->l4_hash = 0;
1134}
1135
1136static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1137{
1138 if (!skb->l4_hash)
1139 skb_clear_hash(skb);
1140}
1141
1142static inline void
1143__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1144{
1145 skb->l4_hash = is_l4;
1146 skb->sw_hash = is_sw;
1147 skb->hash = hash;
1148}
1149
Tom Herbert09323cc2013-12-15 22:16:19 -08001150static inline void
1151skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1152{
Tom Herbertbcc83832015-09-01 09:24:24 -07001153 /* Used by drivers to set hash from HW */
1154 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1155}
1156
1157static inline void
1158__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1159{
1160 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001161}
1162
Tom Herberte5276932015-09-01 09:24:23 -07001163void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001164u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001165u32 skb_get_poff(const struct sk_buff *skb);
1166u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1167 const struct flow_keys *keys, int hlen);
1168__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1169 void *data, int hlen_proto);
1170
1171static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1172 int thoff, u8 ip_proto)
1173{
1174 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1175}
1176
1177void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1178 const struct flow_dissector_key *key,
1179 unsigned int key_count);
1180
1181bool __skb_flow_dissect(const struct sk_buff *skb,
1182 struct flow_dissector *flow_dissector,
1183 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001184 void *data, __be16 proto, int nhoff, int hlen,
1185 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001186
1187static inline bool skb_flow_dissect(const struct sk_buff *skb,
1188 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001189 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001190{
1191 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001192 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001193}
1194
1195static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001196 struct flow_keys *flow,
1197 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001198{
1199 memset(flow, 0, sizeof(*flow));
1200 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001201 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001202}
1203
1204static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1205 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001206 int nhoff, int hlen,
1207 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001208{
1209 memset(flow, 0, sizeof(*flow));
1210 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001211 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001212}
1213
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001214static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001215{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001216 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001217 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001218
Tom Herbert61b905d2014-03-24 15:34:47 -07001219 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001220}
1221
David S. Miller20a17bf2015-09-01 21:19:17 -07001222static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001223{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001224 if (!skb->l4_hash && !skb->sw_hash) {
1225 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001226 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001227
Tom Herbertde4c1f82015-09-01 18:11:04 -07001228 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001229 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001230
1231 return skb->hash;
1232}
1233
Tom Herbert50fb7992015-05-01 11:30:12 -07001234__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1235
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001236static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1237{
Tom Herbert61b905d2014-03-24 15:34:47 -07001238 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001239}
1240
Tom Herbert3df7a742013-12-15 22:16:29 -08001241static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1242{
Tom Herbert61b905d2014-03-24 15:34:47 -07001243 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001244 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001245 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001246};
1247
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001248#ifdef NET_SKBUFF_DATA_USES_OFFSET
1249static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1250{
1251 return skb->head + 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;
1257}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001258#else
1259static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1260{
1261 return skb->end;
1262}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001263
1264static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1265{
1266 return skb->end - skb->head;
1267}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001268#endif
1269
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001271#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
Patrick Ohlyac45f602009-02-12 05:03:37 +00001273static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1274{
1275 return &skb_shinfo(skb)->hwtstamps;
1276}
1277
Willem de Bruijn52267792017-08-03 16:29:39 -04001278static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1279{
1280 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1281
1282 return is_zcopy ? skb_uarg(skb) : NULL;
1283}
1284
1285static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1286{
1287 if (skb && uarg && !skb_zcopy(skb)) {
1288 sock_zerocopy_get(uarg);
1289 skb_shinfo(skb)->destructor_arg = uarg;
1290 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1291 }
1292}
1293
1294/* Release a reference on a zerocopy structure */
1295static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1296{
1297 struct ubuf_info *uarg = skb_zcopy(skb);
1298
1299 if (uarg) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001300 if (uarg->callback == sock_zerocopy_callback) {
1301 uarg->zerocopy = uarg->zerocopy && zerocopy;
1302 sock_zerocopy_put(uarg);
1303 } else {
1304 uarg->callback(uarg, zerocopy);
1305 }
1306
Willem de Bruijn52267792017-08-03 16:29:39 -04001307 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1308 }
1309}
1310
1311/* Abort a zerocopy operation and revert zckey on error in send syscall */
1312static inline void skb_zcopy_abort(struct sk_buff *skb)
1313{
1314 struct ubuf_info *uarg = skb_zcopy(skb);
1315
1316 if (uarg) {
1317 sock_zerocopy_put_abort(uarg);
1318 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1319 }
1320}
1321
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322/**
1323 * skb_queue_empty - check if a queue is empty
1324 * @list: queue head
1325 *
1326 * Returns true if the queue is empty, false otherwise.
1327 */
1328static inline int skb_queue_empty(const struct sk_buff_head *list)
1329{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001330 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331}
1332
1333/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001334 * skb_queue_is_last - check if skb is the last entry in the queue
1335 * @list: queue head
1336 * @skb: buffer
1337 *
1338 * Returns true if @skb is the last buffer on the list.
1339 */
1340static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1341 const struct sk_buff *skb)
1342{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001343 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001344}
1345
1346/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001347 * skb_queue_is_first - check if skb is the first entry in the queue
1348 * @list: queue head
1349 * @skb: buffer
1350 *
1351 * Returns true if @skb is the first buffer on the list.
1352 */
1353static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1354 const struct sk_buff *skb)
1355{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001356 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001357}
1358
1359/**
David S. Miller249c8b42008-09-23 00:44:42 -07001360 * skb_queue_next - return the next packet in the queue
1361 * @list: queue head
1362 * @skb: current buffer
1363 *
1364 * Return the next packet in @list after @skb. It is only valid to
1365 * call this if skb_queue_is_last() evaluates to false.
1366 */
1367static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1368 const struct sk_buff *skb)
1369{
1370 /* This BUG_ON may seem severe, but if we just return then we
1371 * are going to dereference garbage.
1372 */
1373 BUG_ON(skb_queue_is_last(list, skb));
1374 return skb->next;
1375}
1376
1377/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001378 * skb_queue_prev - return the prev packet in the queue
1379 * @list: queue head
1380 * @skb: current buffer
1381 *
1382 * Return the prev packet in @list before @skb. It is only valid to
1383 * call this if skb_queue_is_first() evaluates to false.
1384 */
1385static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1386 const struct sk_buff *skb)
1387{
1388 /* This BUG_ON may seem severe, but if we just return then we
1389 * are going to dereference garbage.
1390 */
1391 BUG_ON(skb_queue_is_first(list, skb));
1392 return skb->prev;
1393}
1394
1395/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 * skb_get - reference buffer
1397 * @skb: buffer to reference
1398 *
1399 * Makes another reference to a socket buffer and returns a pointer
1400 * to the buffer.
1401 */
1402static inline struct sk_buff *skb_get(struct sk_buff *skb)
1403{
Reshetova, Elena63354792017-06-30 13:07:58 +03001404 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 return skb;
1406}
1407
1408/*
1409 * If users == 1, we are the only owner and are can avoid redundant
1410 * atomic change.
1411 */
1412
1413/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 * skb_cloned - is the buffer a clone
1415 * @skb: buffer to check
1416 *
1417 * Returns true if the buffer was generated with skb_clone() and is
1418 * one of multiple shared copies of the buffer. Cloned buffers are
1419 * shared data so must not be written to under normal circumstances.
1420 */
1421static inline int skb_cloned(const struct sk_buff *skb)
1422{
1423 return skb->cloned &&
1424 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1425}
1426
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001427static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1428{
Mel Gormand0164ad2015-11-06 16:28:21 -08001429 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001430
1431 if (skb_cloned(skb))
1432 return pskb_expand_head(skb, 0, 0, pri);
1433
1434 return 0;
1435}
1436
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437/**
1438 * skb_header_cloned - is the header a clone
1439 * @skb: buffer to check
1440 *
1441 * Returns true if modifying the header part of the buffer requires
1442 * the data to be copied.
1443 */
1444static inline int skb_header_cloned(const struct sk_buff *skb)
1445{
1446 int dataref;
1447
1448 if (!skb->cloned)
1449 return 0;
1450
1451 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1452 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1453 return dataref != 1;
1454}
1455
Eric Dumazet9580bf22016-04-30 10:19:29 -07001456static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1457{
1458 might_sleep_if(gfpflags_allow_blocking(pri));
1459
1460 if (skb_header_cloned(skb))
1461 return pskb_expand_head(skb, 0, 0, pri);
1462
1463 return 0;
1464}
1465
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001467 * __skb_header_release - release reference to header
1468 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001469 */
1470static inline void __skb_header_release(struct sk_buff *skb)
1471{
1472 skb->nohdr = 1;
1473 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1474}
1475
1476
1477/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 * skb_shared - is the buffer shared
1479 * @skb: buffer to check
1480 *
1481 * Returns true if more than one person has a reference to this
1482 * buffer.
1483 */
1484static inline int skb_shared(const struct sk_buff *skb)
1485{
Reshetova, Elena63354792017-06-30 13:07:58 +03001486 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}
1488
1489/**
1490 * skb_share_check - check if buffer is shared and if so clone it
1491 * @skb: buffer to check
1492 * @pri: priority for memory allocation
1493 *
1494 * If the buffer is shared the buffer is cloned and the old copy
1495 * drops a reference. A new clone with a single reference is returned.
1496 * If the buffer is not shared the original buffer is returned. When
1497 * being called from interrupt status or with spinlocks held pri must
1498 * be GFP_ATOMIC.
1499 *
1500 * NULL is returned on a memory allocation failure.
1501 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001502static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503{
Mel Gormand0164ad2015-11-06 16:28:21 -08001504 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 if (skb_shared(skb)) {
1506 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001507
1508 if (likely(nskb))
1509 consume_skb(skb);
1510 else
1511 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 skb = nskb;
1513 }
1514 return skb;
1515}
1516
1517/*
1518 * Copy shared buffers into a new sk_buff. We effectively do COW on
1519 * packets to handle cases where we have a local reader and forward
1520 * and a couple of other messy ones. The normal one is tcpdumping
1521 * a packet thats being forwarded.
1522 */
1523
1524/**
1525 * skb_unshare - make a copy of a shared buffer
1526 * @skb: buffer to check
1527 * @pri: priority for memory allocation
1528 *
1529 * If the socket buffer is a clone then this function creates a new
1530 * copy of the data, drops a reference count on the old copy and returns
1531 * the new copy with the reference count at 1. If the buffer is not a clone
1532 * the original buffer is returned. When called with a spinlock held or
1533 * from interrupt state @pri must be %GFP_ATOMIC
1534 *
1535 * %NULL is returned on a memory allocation failure.
1536 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001537static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001538 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539{
Mel Gormand0164ad2015-11-06 16:28:21 -08001540 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 if (skb_cloned(skb)) {
1542 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001543
1544 /* Free our shared copy */
1545 if (likely(nskb))
1546 consume_skb(skb);
1547 else
1548 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 skb = nskb;
1550 }
1551 return skb;
1552}
1553
1554/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001555 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 * @list_: list to peek at
1557 *
1558 * Peek an &sk_buff. Unlike most other operations you _MUST_
1559 * be careful with this one. A peek leaves the buffer on the
1560 * list and someone else may run off with it. You must hold
1561 * the appropriate locks or have a private queue to do this.
1562 *
1563 * Returns %NULL for an empty list or a pointer to the head element.
1564 * The reference count is not incremented and the reference is therefore
1565 * volatile. Use with caution.
1566 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001567static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568{
Eric Dumazet18d07002012-04-30 16:31:46 +00001569 struct sk_buff *skb = list_->next;
1570
1571 if (skb == (struct sk_buff *)list_)
1572 skb = NULL;
1573 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574}
1575
1576/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001577 * skb_peek_next - peek skb following the given one from a queue
1578 * @skb: skb to start from
1579 * @list_: list to peek at
1580 *
1581 * Returns %NULL when the end of the list is met or a pointer to the
1582 * next element. The reference count is not incremented and the
1583 * reference is therefore volatile. Use with caution.
1584 */
1585static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1586 const struct sk_buff_head *list_)
1587{
1588 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001589
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001590 if (next == (struct sk_buff *)list_)
1591 next = NULL;
1592 return next;
1593}
1594
1595/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001596 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 * @list_: list to peek at
1598 *
1599 * Peek an &sk_buff. Unlike most other operations you _MUST_
1600 * be careful with this one. A peek leaves the buffer on the
1601 * list and someone else may run off with it. You must hold
1602 * the appropriate locks or have a private queue to do this.
1603 *
1604 * Returns %NULL for an empty list or a pointer to the tail element.
1605 * The reference count is not incremented and the reference is therefore
1606 * volatile. Use with caution.
1607 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001608static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609{
Eric Dumazet18d07002012-04-30 16:31:46 +00001610 struct sk_buff *skb = list_->prev;
1611
1612 if (skb == (struct sk_buff *)list_)
1613 skb = NULL;
1614 return skb;
1615
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616}
1617
1618/**
1619 * skb_queue_len - get queue length
1620 * @list_: list to measure
1621 *
1622 * Return the length of an &sk_buff queue.
1623 */
1624static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1625{
1626 return list_->qlen;
1627}
1628
David S. Miller67fed452008-09-21 22:36:24 -07001629/**
1630 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1631 * @list: queue to initialize
1632 *
1633 * This initializes only the list and queue length aspects of
1634 * an sk_buff_head object. This allows to initialize the list
1635 * aspects of an sk_buff_head without reinitializing things like
1636 * the spinlock. It can also be used for on-stack sk_buff_head
1637 * objects where the spinlock is known to not be used.
1638 */
1639static inline void __skb_queue_head_init(struct sk_buff_head *list)
1640{
1641 list->prev = list->next = (struct sk_buff *)list;
1642 list->qlen = 0;
1643}
1644
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001645/*
1646 * This function creates a split out lock class for each invocation;
1647 * this is needed for now since a whole lot of users of the skb-queue
1648 * infrastructure in drivers have different locking usage (in hardirq)
1649 * than the networking core (in softirq only). In the long run either the
1650 * network layer or drivers should need annotation to consolidate the
1651 * main types of usage into 3 classes.
1652 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653static inline void skb_queue_head_init(struct sk_buff_head *list)
1654{
1655 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001656 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657}
1658
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001659static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1660 struct lock_class_key *class)
1661{
1662 skb_queue_head_init(list);
1663 lockdep_set_class(&list->lock, class);
1664}
1665
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001667 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 *
1669 * The "__skb_xxxx()" functions are the non-atomic ones that
1670 * can only be called with interrupts disabled.
1671 */
Joe Perches7965bd42013-09-26 14:48:15 -07001672void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1673 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001674static inline void __skb_insert(struct sk_buff *newsk,
1675 struct sk_buff *prev, struct sk_buff *next,
1676 struct sk_buff_head *list)
1677{
1678 newsk->next = next;
1679 newsk->prev = prev;
1680 next->prev = prev->next = newsk;
1681 list->qlen++;
1682}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683
David S. Miller67fed452008-09-21 22:36:24 -07001684static inline void __skb_queue_splice(const struct sk_buff_head *list,
1685 struct sk_buff *prev,
1686 struct sk_buff *next)
1687{
1688 struct sk_buff *first = list->next;
1689 struct sk_buff *last = list->prev;
1690
1691 first->prev = prev;
1692 prev->next = first;
1693
1694 last->next = next;
1695 next->prev = last;
1696}
1697
1698/**
1699 * skb_queue_splice - join two skb lists, this is designed for stacks
1700 * @list: the new list to add
1701 * @head: the place to add it in the first list
1702 */
1703static inline void skb_queue_splice(const struct sk_buff_head *list,
1704 struct sk_buff_head *head)
1705{
1706 if (!skb_queue_empty(list)) {
1707 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001708 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001709 }
1710}
1711
1712/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001713 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001714 * @list: the new list to add
1715 * @head: the place to add it in the first list
1716 *
1717 * The list at @list is reinitialised
1718 */
1719static inline void skb_queue_splice_init(struct sk_buff_head *list,
1720 struct sk_buff_head *head)
1721{
1722 if (!skb_queue_empty(list)) {
1723 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001724 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001725 __skb_queue_head_init(list);
1726 }
1727}
1728
1729/**
1730 * skb_queue_splice_tail - join two skb lists, each list being a queue
1731 * @list: the new list to add
1732 * @head: the place to add it in the first list
1733 */
1734static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1735 struct sk_buff_head *head)
1736{
1737 if (!skb_queue_empty(list)) {
1738 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001739 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001740 }
1741}
1742
1743/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001744 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001745 * @list: the new list to add
1746 * @head: the place to add it in the first list
1747 *
1748 * Each of the lists is a queue.
1749 * The list at @list is reinitialised
1750 */
1751static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1752 struct sk_buff_head *head)
1753{
1754 if (!skb_queue_empty(list)) {
1755 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001756 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001757 __skb_queue_head_init(list);
1758 }
1759}
1760
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001762 * __skb_queue_after - queue a buffer at the list head
1763 * @list: list to use
1764 * @prev: place after this buffer
1765 * @newsk: buffer to queue
1766 *
1767 * Queue a buffer int the middle of a list. This function takes no locks
1768 * and you must therefore hold required locks before calling it.
1769 *
1770 * A buffer cannot be placed on two lists at the same time.
1771 */
1772static inline void __skb_queue_after(struct sk_buff_head *list,
1773 struct sk_buff *prev,
1774 struct sk_buff *newsk)
1775{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001776 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001777}
1778
Joe Perches7965bd42013-09-26 14:48:15 -07001779void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1780 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001781
Gerrit Renkerf5572852008-04-14 00:05:28 -07001782static inline void __skb_queue_before(struct sk_buff_head *list,
1783 struct sk_buff *next,
1784 struct sk_buff *newsk)
1785{
1786 __skb_insert(newsk, next->prev, next, list);
1787}
1788
Stephen Hemminger300ce172005-10-30 13:47:34 -08001789/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 * __skb_queue_head - queue a buffer at the list head
1791 * @list: list to use
1792 * @newsk: buffer to queue
1793 *
1794 * Queue a buffer at the start of a list. This function takes no locks
1795 * and you must therefore hold required locks before calling it.
1796 *
1797 * A buffer cannot be placed on two lists at the same time.
1798 */
Joe Perches7965bd42013-09-26 14:48:15 -07001799void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800static inline void __skb_queue_head(struct sk_buff_head *list,
1801 struct sk_buff *newsk)
1802{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001803 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804}
1805
1806/**
1807 * __skb_queue_tail - queue a buffer at the list tail
1808 * @list: list to use
1809 * @newsk: buffer to queue
1810 *
1811 * Queue a buffer at the end of a list. This function takes no locks
1812 * and you must therefore hold required locks before calling it.
1813 *
1814 * A buffer cannot be placed on two lists at the same time.
1815 */
Joe Perches7965bd42013-09-26 14:48:15 -07001816void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817static inline void __skb_queue_tail(struct sk_buff_head *list,
1818 struct sk_buff *newsk)
1819{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001820 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821}
1822
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 * remove sk_buff from list. _Must_ be called atomically, and with
1825 * the list known..
1826 */
Joe Perches7965bd42013-09-26 14:48:15 -07001827void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1829{
1830 struct sk_buff *next, *prev;
1831
1832 list->qlen--;
1833 next = skb->next;
1834 prev = skb->prev;
1835 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 next->prev = prev;
1837 prev->next = next;
1838}
1839
Gerrit Renkerf525c062008-04-14 00:04:12 -07001840/**
1841 * __skb_dequeue - remove from the head of the queue
1842 * @list: list to dequeue from
1843 *
1844 * Remove the head of the list. This function does not take any locks
1845 * so must be used with appropriate locks held only. The head item is
1846 * returned or %NULL if the list is empty.
1847 */
Joe Perches7965bd42013-09-26 14:48:15 -07001848struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001849static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1850{
1851 struct sk_buff *skb = skb_peek(list);
1852 if (skb)
1853 __skb_unlink(skb, list);
1854 return skb;
1855}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
1857/**
1858 * __skb_dequeue_tail - remove from the tail of the queue
1859 * @list: list to dequeue from
1860 *
1861 * Remove the tail of the list. This function does not take any locks
1862 * so must be used with appropriate locks held only. The tail item is
1863 * returned or %NULL if the list is empty.
1864 */
Joe Perches7965bd42013-09-26 14:48:15 -07001865struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1867{
1868 struct sk_buff *skb = skb_peek_tail(list);
1869 if (skb)
1870 __skb_unlink(skb, list);
1871 return skb;
1872}
1873
1874
David S. Millerbdcc0922012-03-07 20:53:36 -05001875static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876{
1877 return skb->data_len;
1878}
1879
1880static inline unsigned int skb_headlen(const struct sk_buff *skb)
1881{
1882 return skb->len - skb->data_len;
1883}
1884
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001885static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001887 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001889 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001890 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001891 return len;
1892}
1893
1894static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1895{
1896 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897}
1898
Ian Campbell131ea662011-08-19 06:25:00 +00001899/**
1900 * __skb_fill_page_desc - initialise a paged fragment in an skb
1901 * @skb: buffer containing fragment to be initialised
1902 * @i: paged fragment index to initialise
1903 * @page: the page to use for this fragment
1904 * @off: the offset to the data with @page
1905 * @size: the length of the data
1906 *
1907 * Initialises the @i'th fragment of @skb to point to &size bytes at
1908 * offset @off within @page.
1909 *
1910 * Does not take any additional reference on the fragment.
1911 */
1912static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1913 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914{
1915 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1916
Mel Gormanc48a11c2012-07-31 16:44:23 -07001917 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001918 * Propagate page pfmemalloc to the skb if we can. The problem is
1919 * that not all callers have unique ownership of the page but rely
1920 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001921 */
Ian Campbella8605c62011-10-19 23:01:49 +00001922 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001924 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001925
1926 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001927 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001928 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001929}
1930
1931/**
1932 * skb_fill_page_desc - initialise a paged fragment in an skb
1933 * @skb: buffer containing fragment to be initialised
1934 * @i: paged fragment index to initialise
1935 * @page: the page to use for this fragment
1936 * @off: the offset to the data with @page
1937 * @size: the length of the data
1938 *
1939 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001940 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001941 * addition updates @skb such that @i is the last fragment.
1942 *
1943 * Does not take any additional reference on the fragment.
1944 */
1945static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1946 struct page *page, int off, int size)
1947{
1948 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 skb_shinfo(skb)->nr_frags = i + 1;
1950}
1951
Joe Perches7965bd42013-09-26 14:48:15 -07001952void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1953 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001954
Jason Wangf8e617e2013-11-01 14:07:47 +08001955void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1956 unsigned int truesize);
1957
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001959#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1961
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001962#ifdef NET_SKBUFF_DATA_USES_OFFSET
1963static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1964{
1965 return skb->head + skb->tail;
1966}
1967
1968static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1969{
1970 skb->tail = skb->data - skb->head;
1971}
1972
1973static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1974{
1975 skb_reset_tail_pointer(skb);
1976 skb->tail += offset;
1977}
Simon Horman7cc46192013-05-28 20:34:29 +00001978
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001979#else /* NET_SKBUFF_DATA_USES_OFFSET */
1980static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1981{
1982 return skb->tail;
1983}
1984
1985static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1986{
1987 skb->tail = skb->data;
1988}
1989
1990static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1991{
1992 skb->tail = skb->data + offset;
1993}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001994
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001995#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1996
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997/*
1998 * Add data to an sk_buff
1999 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002000void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2001void *skb_put(struct sk_buff *skb, unsigned int len);
2002static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003{
Johannes Berg4df864c2017-06-16 14:29:21 +02002004 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 SKB_LINEAR_ASSERT(skb);
2006 skb->tail += len;
2007 skb->len += len;
2008 return tmp;
2009}
2010
yuan linyude77b962017-06-18 22:48:17 +08002011static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2012{
2013 void *tmp = __skb_put(skb, len);
2014
2015 memset(tmp, 0, len);
2016 return tmp;
2017}
2018
2019static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2020 unsigned int len)
2021{
2022 void *tmp = __skb_put(skb, len);
2023
2024 memcpy(tmp, data, len);
2025 return tmp;
2026}
2027
2028static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2029{
2030 *(u8 *)__skb_put(skb, 1) = val;
2031}
2032
Johannes Berg83ad3572017-06-14 22:17:20 +02002033static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002034{
Johannes Berg83ad3572017-06-14 22:17:20 +02002035 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002036
2037 memset(tmp, 0, len);
2038
2039 return tmp;
2040}
2041
Johannes Berg59ae1d12017-06-16 14:29:20 +02002042static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2043 unsigned int len)
2044{
2045 void *tmp = skb_put(skb, len);
2046
2047 memcpy(tmp, data, len);
2048
2049 return tmp;
2050}
2051
Johannes Berg634fef62017-06-16 14:29:24 +02002052static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2053{
2054 *(u8 *)skb_put(skb, 1) = val;
2055}
2056
Johannes Bergd58ff352017-06-16 14:29:23 +02002057void *skb_push(struct sk_buff *skb, unsigned int len);
2058static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059{
2060 skb->data -= len;
2061 skb->len += len;
2062 return skb->data;
2063}
2064
Johannes Bergaf728682017-06-16 14:29:22 +02002065void *skb_pull(struct sk_buff *skb, unsigned int len);
2066static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067{
2068 skb->len -= len;
2069 BUG_ON(skb->len < skb->data_len);
2070 return skb->data += len;
2071}
2072
Johannes Bergaf728682017-06-16 14:29:22 +02002073static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002074{
2075 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2076}
2077
Johannes Bergaf728682017-06-16 14:29:22 +02002078void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079
Johannes Bergaf728682017-06-16 14:29:22 +02002080static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081{
2082 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002083 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 return NULL;
2085 skb->len -= len;
2086 return skb->data += len;
2087}
2088
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 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2092}
2093
2094static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2095{
2096 if (likely(len <= skb_headlen(skb)))
2097 return 1;
2098 if (unlikely(len > skb->len))
2099 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002100 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101}
2102
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002103void skb_condense(struct sk_buff *skb);
2104
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105/**
2106 * skb_headroom - bytes at buffer head
2107 * @skb: buffer to check
2108 *
2109 * Return the number of bytes of free space at the head of an &sk_buff.
2110 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002111static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112{
2113 return skb->data - skb->head;
2114}
2115
2116/**
2117 * skb_tailroom - bytes at buffer end
2118 * @skb: buffer to check
2119 *
2120 * Return the number of bytes of free space at the tail of an sk_buff
2121 */
2122static inline int skb_tailroom(const struct sk_buff *skb)
2123{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002124 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125}
2126
2127/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002128 * skb_availroom - bytes at buffer end
2129 * @skb: buffer to check
2130 *
2131 * Return the number of bytes of free space at the tail of an sk_buff
2132 * allocated by sk_stream_alloc()
2133 */
2134static inline int skb_availroom(const struct sk_buff *skb)
2135{
Eric Dumazet16fad692013-03-14 05:40:32 +00002136 if (skb_is_nonlinear(skb))
2137 return 0;
2138
2139 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002140}
2141
2142/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 * skb_reserve - adjust headroom
2144 * @skb: buffer to alter
2145 * @len: bytes to move
2146 *
2147 * Increase the headroom of an empty &sk_buff by reducing the tail
2148 * room. This is only allowed for an empty buffer.
2149 */
David S. Miller8243126c2006-01-17 02:54:21 -08002150static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151{
2152 skb->data += len;
2153 skb->tail += len;
2154}
2155
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002156/**
2157 * skb_tailroom_reserve - adjust reserved_tailroom
2158 * @skb: buffer to alter
2159 * @mtu: maximum amount of headlen permitted
2160 * @needed_tailroom: minimum amount of reserved_tailroom
2161 *
2162 * Set reserved_tailroom so that headlen can be as large as possible but
2163 * not larger than mtu and tailroom cannot be smaller than
2164 * needed_tailroom.
2165 * The required headroom should already have been reserved before using
2166 * this function.
2167 */
2168static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2169 unsigned int needed_tailroom)
2170{
2171 SKB_LINEAR_ASSERT(skb);
2172 if (mtu < skb_tailroom(skb) - needed_tailroom)
2173 /* use at most mtu */
2174 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2175 else
2176 /* use up to all available space */
2177 skb->reserved_tailroom = needed_tailroom;
2178}
2179
Tom Herbert8bce6d72014-09-29 20:22:29 -07002180#define ENCAP_TYPE_ETHER 0
2181#define ENCAP_TYPE_IPPROTO 1
2182
2183static inline void skb_set_inner_protocol(struct sk_buff *skb,
2184 __be16 protocol)
2185{
2186 skb->inner_protocol = protocol;
2187 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2188}
2189
2190static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2191 __u8 ipproto)
2192{
2193 skb->inner_ipproto = ipproto;
2194 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2195}
2196
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002197static inline void skb_reset_inner_headers(struct sk_buff *skb)
2198{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002199 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002200 skb->inner_network_header = skb->network_header;
2201 skb->inner_transport_header = skb->transport_header;
2202}
2203
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002204static inline void skb_reset_mac_len(struct sk_buff *skb)
2205{
2206 skb->mac_len = skb->network_header - skb->mac_header;
2207}
2208
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002209static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2210 *skb)
2211{
2212 return skb->head + skb->inner_transport_header;
2213}
2214
Tom Herbert55dc5a92015-12-14 11:19:40 -08002215static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2216{
2217 return skb_inner_transport_header(skb) - skb->data;
2218}
2219
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002220static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2221{
2222 skb->inner_transport_header = skb->data - skb->head;
2223}
2224
2225static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2226 const int offset)
2227{
2228 skb_reset_inner_transport_header(skb);
2229 skb->inner_transport_header += offset;
2230}
2231
2232static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2233{
2234 return skb->head + skb->inner_network_header;
2235}
2236
2237static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2238{
2239 skb->inner_network_header = skb->data - skb->head;
2240}
2241
2242static inline void skb_set_inner_network_header(struct sk_buff *skb,
2243 const int offset)
2244{
2245 skb_reset_inner_network_header(skb);
2246 skb->inner_network_header += offset;
2247}
2248
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002249static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2250{
2251 return skb->head + skb->inner_mac_header;
2252}
2253
2254static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2255{
2256 skb->inner_mac_header = skb->data - skb->head;
2257}
2258
2259static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2260 const int offset)
2261{
2262 skb_reset_inner_mac_header(skb);
2263 skb->inner_mac_header += offset;
2264}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002265static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2266{
Cong Wang35d04612013-05-29 15:16:05 +08002267 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002268}
2269
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002270static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2271{
2272 return skb->head + skb->transport_header;
2273}
2274
2275static inline void skb_reset_transport_header(struct sk_buff *skb)
2276{
2277 skb->transport_header = skb->data - skb->head;
2278}
2279
2280static inline void skb_set_transport_header(struct sk_buff *skb,
2281 const int offset)
2282{
2283 skb_reset_transport_header(skb);
2284 skb->transport_header += offset;
2285}
2286
2287static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2288{
2289 return skb->head + skb->network_header;
2290}
2291
2292static inline void skb_reset_network_header(struct sk_buff *skb)
2293{
2294 skb->network_header = skb->data - skb->head;
2295}
2296
2297static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2298{
2299 skb_reset_network_header(skb);
2300 skb->network_header += offset;
2301}
2302
2303static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2304{
2305 return skb->head + skb->mac_header;
2306}
2307
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002308static inline int skb_mac_offset(const struct sk_buff *skb)
2309{
2310 return skb_mac_header(skb) - skb->data;
2311}
2312
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002313static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2314{
2315 return skb->network_header - skb->mac_header;
2316}
2317
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002318static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2319{
Cong Wang35d04612013-05-29 15:16:05 +08002320 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002321}
2322
2323static inline void skb_reset_mac_header(struct sk_buff *skb)
2324{
2325 skb->mac_header = skb->data - skb->head;
2326}
2327
2328static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2329{
2330 skb_reset_mac_header(skb);
2331 skb->mac_header += offset;
2332}
2333
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002334static inline void skb_pop_mac_header(struct sk_buff *skb)
2335{
2336 skb->mac_header = skb->network_header;
2337}
2338
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002339static inline void skb_probe_transport_header(struct sk_buff *skb,
2340 const int offset_hint)
2341{
2342 struct flow_keys keys;
2343
2344 if (skb_transport_header_was_set(skb))
2345 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002346 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002347 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002348 else
2349 skb_set_transport_header(skb, offset_hint);
2350}
2351
Eric Dumazet03606892012-02-23 10:55:02 +00002352static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2353{
2354 if (skb_mac_header_was_set(skb)) {
2355 const unsigned char *old_mac = skb_mac_header(skb);
2356
2357 skb_set_mac_header(skb, -skb->mac_len);
2358 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2359 }
2360}
2361
Michał Mirosław04fb4512010-12-14 15:24:08 +00002362static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2363{
2364 return skb->csum_start - skb_headroom(skb);
2365}
2366
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002367static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2368{
2369 return skb->head + skb->csum_start;
2370}
2371
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002372static inline int skb_transport_offset(const struct sk_buff *skb)
2373{
2374 return skb_transport_header(skb) - skb->data;
2375}
2376
2377static inline u32 skb_network_header_len(const struct sk_buff *skb)
2378{
2379 return skb->transport_header - skb->network_header;
2380}
2381
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002382static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2383{
2384 return skb->inner_transport_header - skb->inner_network_header;
2385}
2386
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002387static inline int skb_network_offset(const struct sk_buff *skb)
2388{
2389 return skb_network_header(skb) - skb->data;
2390}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002391
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002392static inline int skb_inner_network_offset(const struct sk_buff *skb)
2393{
2394 return skb_inner_network_header(skb) - skb->data;
2395}
2396
Changli Gaof9599ce2010-08-04 04:43:44 +00002397static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2398{
2399 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2400}
2401
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402/*
2403 * CPUs often take a performance hit when accessing unaligned memory
2404 * locations. The actual performance hit varies, it can be small if the
2405 * hardware handles it or large if we have to take an exception and fix it
2406 * in software.
2407 *
2408 * Since an ethernet header is 14 bytes network drivers often end up with
2409 * the IP header at an unaligned offset. The IP header can be aligned by
2410 * shifting the start of the packet by 2 bytes. Drivers should do this
2411 * with:
2412 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002413 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414 *
2415 * The downside to this alignment of the IP header is that the DMA is now
2416 * unaligned. On some architectures the cost of an unaligned DMA is high
2417 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002418 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2420 * to be overridden.
2421 */
2422#ifndef NET_IP_ALIGN
2423#define NET_IP_ALIGN 2
2424#endif
2425
Anton Blanchard025be812006-03-31 02:27:06 -08002426/*
2427 * The networking layer reserves some headroom in skb data (via
2428 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2429 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002430 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002431 *
2432 * Unfortunately this headroom changes the DMA alignment of the resulting
2433 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2434 * on some architectures. An architecture can override this value,
2435 * perhaps setting it to a cacheline in size (since that will maintain
2436 * cacheline alignment of the DMA). It must be a power of 2.
2437 *
David S. Millerd6301d32009-02-08 19:24:13 -08002438 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002439 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002440 *
2441 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2442 * to reduce average number of cache lines per packet.
2443 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002444 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002445 */
2446#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002447#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002448#endif
2449
Joe Perches7965bd42013-09-26 14:48:15 -07002450int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002452static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002454 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002455 WARN_ON(1);
2456 return;
2457 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002458 skb->len = len;
2459 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460}
2461
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002462static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2463{
2464 __skb_set_length(skb, len);
2465}
2466
Joe Perches7965bd42013-09-26 14:48:15 -07002467void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468
2469static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2470{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002471 if (skb->data_len)
2472 return ___pskb_trim(skb, len);
2473 __skb_trim(skb, len);
2474 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475}
2476
2477static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2478{
2479 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2480}
2481
2482/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002483 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2484 * @skb: buffer to alter
2485 * @len: new length
2486 *
2487 * This is identical to pskb_trim except that the caller knows that
2488 * the skb is not cloned so we should never get an error due to out-
2489 * of-memory.
2490 */
2491static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2492{
2493 int err = pskb_trim(skb, len);
2494 BUG_ON(err);
2495}
2496
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002497static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2498{
2499 unsigned int diff = len - skb->len;
2500
2501 if (skb_tailroom(skb) < diff) {
2502 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2503 GFP_ATOMIC);
2504 if (ret)
2505 return ret;
2506 }
2507 __skb_set_length(skb, len);
2508 return 0;
2509}
2510
Herbert Xue9fa4f72006-08-13 20:12:58 -07002511/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002512 * skb_orphan - orphan a buffer
2513 * @skb: buffer to orphan
2514 *
2515 * If a buffer currently has an owner then we call the owner's
2516 * destructor function and make the @skb unowned. The buffer continues
2517 * to exist but is no longer charged to its former owner.
2518 */
2519static inline void skb_orphan(struct sk_buff *skb)
2520{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002521 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002523 skb->destructor = NULL;
2524 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002525 } else {
2526 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002527 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528}
2529
2530/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002531 * skb_orphan_frags - orphan the frags contained in a buffer
2532 * @skb: buffer to orphan frags from
2533 * @gfp_mask: allocation mask for replacement pages
2534 *
2535 * For each frag in the SKB which needs a destructor (i.e. has an
2536 * owner) create a copy of that frag and release the original
2537 * page by calling the destructor.
2538 */
2539static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2540{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002541 if (likely(!skb_zcopy(skb)))
2542 return 0;
2543 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2544 return 0;
2545 return skb_copy_ubufs(skb, gfp_mask);
2546}
2547
2548/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2549static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2550{
2551 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002552 return 0;
2553 return skb_copy_ubufs(skb, gfp_mask);
2554}
2555
2556/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557 * __skb_queue_purge - empty a list
2558 * @list: list to empty
2559 *
2560 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2561 * the list and one reference dropped. This function does not take the
2562 * list lock and the caller must hold the relevant locks to use it.
2563 */
Joe Perches7965bd42013-09-26 14:48:15 -07002564void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565static inline void __skb_queue_purge(struct sk_buff_head *list)
2566{
2567 struct sk_buff *skb;
2568 while ((skb = __skb_dequeue(list)) != NULL)
2569 kfree_skb(skb);
2570}
2571
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002572void skb_rbtree_purge(struct rb_root *root);
2573
Joe Perches7965bd42013-09-26 14:48:15 -07002574void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575
Joe Perches7965bd42013-09-26 14:48:15 -07002576struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2577 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002578
2579/**
2580 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2581 * @dev: network device to receive on
2582 * @length: length to allocate
2583 *
2584 * Allocate a new &sk_buff and assign it a usage count of one. The
2585 * buffer has unspecified headroom built in. Users should allocate
2586 * the headroom they think they need without accounting for the
2587 * built in space. The built in space is used for optimisations.
2588 *
2589 * %NULL is returned if there is no free memory. Although this function
2590 * allocates memory it can be called from an interrupt.
2591 */
2592static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002593 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002594{
2595 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2596}
2597
Eric Dumazet6f532612012-05-18 05:12:12 +00002598/* legacy helper around __netdev_alloc_skb() */
2599static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2600 gfp_t gfp_mask)
2601{
2602 return __netdev_alloc_skb(NULL, length, gfp_mask);
2603}
2604
2605/* legacy helper around netdev_alloc_skb() */
2606static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2607{
2608 return netdev_alloc_skb(NULL, length);
2609}
2610
2611
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002612static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2613 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002614{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002615 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002616
2617 if (NET_IP_ALIGN && skb)
2618 skb_reserve(skb, NET_IP_ALIGN);
2619 return skb;
2620}
2621
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002622static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2623 unsigned int length)
2624{
2625 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2626}
2627
Alexander Duyck181edb22015-05-06 21:12:03 -07002628static inline void skb_free_frag(void *addr)
2629{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002630 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002631}
2632
Alexander Duyckffde7322014-12-09 19:40:42 -08002633void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002634struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2635 unsigned int length, gfp_t gfp_mask);
2636static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2637 unsigned int length)
2638{
2639 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2640}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002641void napi_consume_skb(struct sk_buff *skb, int budget);
2642
2643void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002644void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002645
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002646/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002647 * __dev_alloc_pages - allocate page for network Rx
2648 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2649 * @order: size of the allocation
2650 *
2651 * Allocate a new page.
2652 *
2653 * %NULL is returned if there is no free memory.
2654*/
2655static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2656 unsigned int order)
2657{
2658 /* This piece of code contains several assumptions.
2659 * 1. This is for device Rx, therefor a cold page is preferred.
2660 * 2. The expectation is the user wants a compound page.
2661 * 3. If requesting a order 0 page it will not be compound
2662 * due to the check to see if order has a value in prep_new_page
2663 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2664 * code in gfp_to_alloc_flags that should be enforcing this.
2665 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002666 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002667
2668 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2669}
2670
2671static inline struct page *dev_alloc_pages(unsigned int order)
2672{
Neil Horman95829b32016-05-19 11:30:54 -04002673 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002674}
2675
2676/**
2677 * __dev_alloc_page - allocate a page for network Rx
2678 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2679 *
2680 * Allocate a new page.
2681 *
2682 * %NULL is returned if there is no free memory.
2683 */
2684static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2685{
2686 return __dev_alloc_pages(gfp_mask, 0);
2687}
2688
2689static inline struct page *dev_alloc_page(void)
2690{
Neil Horman95829b32016-05-19 11:30:54 -04002691 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002692}
2693
2694/**
Mel Gorman06140022012-07-31 16:44:24 -07002695 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2696 * @page: The page that was allocated from skb_alloc_page
2697 * @skb: The skb that may need pfmemalloc set
2698 */
2699static inline void skb_propagate_pfmemalloc(struct page *page,
2700 struct sk_buff *skb)
2701{
Michal Hocko2f064f32015-08-21 14:11:51 -07002702 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002703 skb->pfmemalloc = true;
2704}
2705
Eric Dumazet564824b2010-10-11 19:05:25 +00002706/**
Masanari Iidae2278672014-02-18 22:54:36 +09002707 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002708 * @frag: the paged fragment
2709 *
2710 * Returns the &struct page associated with @frag.
2711 */
2712static inline struct page *skb_frag_page(const skb_frag_t *frag)
2713{
Ian Campbella8605c62011-10-19 23:01:49 +00002714 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002715}
2716
2717/**
2718 * __skb_frag_ref - take an addition reference on a paged fragment.
2719 * @frag: the paged fragment
2720 *
2721 * Takes an additional reference on the paged fragment @frag.
2722 */
2723static inline void __skb_frag_ref(skb_frag_t *frag)
2724{
2725 get_page(skb_frag_page(frag));
2726}
2727
2728/**
2729 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2730 * @skb: the buffer
2731 * @f: the fragment offset.
2732 *
2733 * Takes an additional reference on the @f'th paged fragment of @skb.
2734 */
2735static inline void skb_frag_ref(struct sk_buff *skb, int f)
2736{
2737 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2738}
2739
2740/**
2741 * __skb_frag_unref - release a reference on a paged fragment.
2742 * @frag: the paged fragment
2743 *
2744 * Releases a reference on the paged fragment @frag.
2745 */
2746static inline void __skb_frag_unref(skb_frag_t *frag)
2747{
2748 put_page(skb_frag_page(frag));
2749}
2750
2751/**
2752 * skb_frag_unref - release a reference on a paged fragment of an skb.
2753 * @skb: the buffer
2754 * @f: the fragment offset
2755 *
2756 * Releases a reference on the @f'th paged fragment of @skb.
2757 */
2758static inline void skb_frag_unref(struct sk_buff *skb, int f)
2759{
2760 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2761}
2762
2763/**
2764 * skb_frag_address - gets the address of the data contained in a paged fragment
2765 * @frag: the paged fragment buffer
2766 *
2767 * Returns the address of the data within @frag. The page must already
2768 * be mapped.
2769 */
2770static inline void *skb_frag_address(const skb_frag_t *frag)
2771{
2772 return page_address(skb_frag_page(frag)) + frag->page_offset;
2773}
2774
2775/**
2776 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2777 * @frag: the paged fragment buffer
2778 *
2779 * Returns the address of the data within @frag. Checks that the page
2780 * is mapped and returns %NULL otherwise.
2781 */
2782static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2783{
2784 void *ptr = page_address(skb_frag_page(frag));
2785 if (unlikely(!ptr))
2786 return NULL;
2787
2788 return ptr + frag->page_offset;
2789}
2790
2791/**
2792 * __skb_frag_set_page - sets the page contained in a paged fragment
2793 * @frag: the paged fragment
2794 * @page: the page to set
2795 *
2796 * Sets the fragment @frag to contain @page.
2797 */
2798static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2799{
Ian Campbella8605c62011-10-19 23:01:49 +00002800 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002801}
2802
2803/**
2804 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2805 * @skb: the buffer
2806 * @f: the fragment offset
2807 * @page: the page to set
2808 *
2809 * Sets the @f'th fragment of @skb to contain @page.
2810 */
2811static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2812 struct page *page)
2813{
2814 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2815}
2816
Eric Dumazet400dfd32013-10-17 16:27:07 -07002817bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2818
Ian Campbell131ea662011-08-19 06:25:00 +00002819/**
2820 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002821 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002822 * @frag: the paged fragment to map
2823 * @offset: the offset within the fragment (starting at the
2824 * fragment's own offset)
2825 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002826 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002827 *
2828 * Maps the page associated with @frag to @device.
2829 */
2830static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2831 const skb_frag_t *frag,
2832 size_t offset, size_t size,
2833 enum dma_data_direction dir)
2834{
2835 return dma_map_page(dev, skb_frag_page(frag),
2836 frag->page_offset + offset, size, dir);
2837}
2838
Eric Dumazet117632e2011-12-03 21:39:53 +00002839static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2840 gfp_t gfp_mask)
2841{
2842 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2843}
2844
Octavian Purdilabad93e92014-06-12 01:36:26 +03002845
2846static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2847 gfp_t gfp_mask)
2848{
2849 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2850}
2851
2852
Ian Campbell131ea662011-08-19 06:25:00 +00002853/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002854 * skb_clone_writable - is the header of a clone writable
2855 * @skb: buffer to check
2856 * @len: length up to which to write
2857 *
2858 * Returns true if modifying the header part of the cloned buffer
2859 * does not requires the data to be copied.
2860 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002861static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002862{
2863 return !skb_header_cloned(skb) &&
2864 skb_headroom(skb) + len <= skb->hdr_len;
2865}
2866
Daniel Borkmann36976492016-02-19 23:05:25 +01002867static inline int skb_try_make_writable(struct sk_buff *skb,
2868 unsigned int write_len)
2869{
2870 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2871 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2872}
2873
Herbert Xud9cc2042007-09-16 16:21:16 -07002874static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2875 int cloned)
2876{
2877 int delta = 0;
2878
Herbert Xud9cc2042007-09-16 16:21:16 -07002879 if (headroom > skb_headroom(skb))
2880 delta = headroom - skb_headroom(skb);
2881
2882 if (delta || cloned)
2883 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2884 GFP_ATOMIC);
2885 return 0;
2886}
2887
Patrick McHardy334a8132007-06-25 04:35:20 -07002888/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002889 * skb_cow - copy header of skb when it is required
2890 * @skb: buffer to cow
2891 * @headroom: needed headroom
2892 *
2893 * If the skb passed lacks sufficient headroom or its data part
2894 * is shared, data is reallocated. If reallocation fails, an error
2895 * is returned and original skb is not changed.
2896 *
2897 * The result is skb with writable area skb->head...skb->tail
2898 * and at least @headroom of space at head.
2899 */
2900static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2901{
Herbert Xud9cc2042007-09-16 16:21:16 -07002902 return __skb_cow(skb, headroom, skb_cloned(skb));
2903}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904
Herbert Xud9cc2042007-09-16 16:21:16 -07002905/**
2906 * skb_cow_head - skb_cow but only making the head writable
2907 * @skb: buffer to cow
2908 * @headroom: needed headroom
2909 *
2910 * This function is identical to skb_cow except that we replace the
2911 * skb_cloned check by skb_header_cloned. It should be used when
2912 * you only need to push on some header and do not need to modify
2913 * the data.
2914 */
2915static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2916{
2917 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002918}
2919
2920/**
2921 * skb_padto - pad an skbuff up to a minimal size
2922 * @skb: buffer to pad
2923 * @len: minimal length
2924 *
2925 * Pads up a buffer to ensure the trailing bytes exist and are
2926 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002927 * is untouched. Otherwise it is extended. Returns zero on
2928 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002929 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002930static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002931{
2932 unsigned int size = skb->len;
2933 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002934 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002935 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002936}
2937
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002938/**
2939 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2940 * @skb: buffer to pad
2941 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07002942 * @free_on_error: free buffer on error
2943 *
2944 * Pads up a buffer to ensure the trailing bytes exist and are
2945 * blanked. If the buffer already contains sufficient data it
2946 * is untouched. Otherwise it is extended. Returns zero on
2947 * success. The skb is freed on error if @free_on_error is true.
2948 */
2949static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2950 bool free_on_error)
2951{
2952 unsigned int size = skb->len;
2953
2954 if (unlikely(size < len)) {
2955 len -= size;
2956 if (__skb_pad(skb, len, free_on_error))
2957 return -ENOMEM;
2958 __skb_put(skb, len);
2959 }
2960 return 0;
2961}
2962
2963/**
2964 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2965 * @skb: buffer to pad
2966 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002967 *
2968 * Pads up a buffer to ensure the trailing bytes exist and are
2969 * blanked. If the buffer already contains sufficient data it
2970 * is untouched. Otherwise it is extended. Returns zero on
2971 * success. The skb is freed on error.
2972 */
2973static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2974{
Florian Fainellicd0a1372017-08-22 15:12:14 -07002975 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002976}
2977
Linus Torvalds1da177e2005-04-16 15:20:36 -07002978static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002979 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980{
2981 const int off = skb->len;
2982
2983 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002984 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002985 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2986 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987 skb->csum = csum_block_add(skb->csum, csum, off);
2988 return 0;
2989 }
Al Viro15e6cb42016-11-01 22:42:45 -04002990 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002991 return 0;
2992
2993 __skb_trim(skb, off);
2994 return -EFAULT;
2995}
2996
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002997static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2998 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003000 if (skb_zcopy(skb))
3001 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003002 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00003003 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003004
Ian Campbellea2ab692011-08-22 23:44:58 +00003005 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00003006 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003007 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003008 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003009}
3010
Herbert Xu364c6ba2006-06-09 16:10:40 -07003011static inline int __skb_linearize(struct sk_buff *skb)
3012{
3013 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3014}
3015
Linus Torvalds1da177e2005-04-16 15:20:36 -07003016/**
3017 * skb_linearize - convert paged skb to linear one
3018 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003019 *
3020 * If there is no free memory -ENOMEM is returned, otherwise zero
3021 * is returned and the old skb data released.
3022 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003023static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003024{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003025 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3026}
3027
3028/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003029 * skb_has_shared_frag - can any frag be overwritten
3030 * @skb: buffer to test
3031 *
3032 * Return true if the skb has at least one frag that might be modified
3033 * by an external entity (as in vmsplice()/sendfile())
3034 */
3035static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3036{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003037 return skb_is_nonlinear(skb) &&
3038 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003039}
3040
3041/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003042 * skb_linearize_cow - make sure skb is linear and writable
3043 * @skb: buffer to process
3044 *
3045 * If there is no free memory -ENOMEM is returned, otherwise zero
3046 * is returned and the old skb data released.
3047 */
3048static inline int skb_linearize_cow(struct sk_buff *skb)
3049{
3050 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3051 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003052}
3053
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003054static __always_inline void
3055__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3056 unsigned int off)
3057{
3058 if (skb->ip_summed == CHECKSUM_COMPLETE)
3059 skb->csum = csum_block_sub(skb->csum,
3060 csum_partial(start, len, 0), off);
3061 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3062 skb_checksum_start_offset(skb) < 0)
3063 skb->ip_summed = CHECKSUM_NONE;
3064}
3065
Linus Torvalds1da177e2005-04-16 15:20:36 -07003066/**
3067 * skb_postpull_rcsum - update checksum for received skb after pull
3068 * @skb: buffer to update
3069 * @start: start of data before pull
3070 * @len: length of data pulled
3071 *
3072 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003073 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3074 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003075 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003076static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003077 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003078{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003079 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003080}
3081
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003082static __always_inline void
3083__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3084 unsigned int off)
3085{
3086 if (skb->ip_summed == CHECKSUM_COMPLETE)
3087 skb->csum = csum_block_add(skb->csum,
3088 csum_partial(start, len, 0), off);
3089}
Herbert Xucbb042f2006-03-20 22:43:56 -08003090
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003091/**
3092 * skb_postpush_rcsum - update checksum for received skb after push
3093 * @skb: buffer to update
3094 * @start: start of data after push
3095 * @len: length of data pushed
3096 *
3097 * After doing a push on a received packet, you need to call this to
3098 * update the CHECKSUM_COMPLETE checksum.
3099 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003100static inline void skb_postpush_rcsum(struct sk_buff *skb,
3101 const void *start, unsigned int len)
3102{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003103 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003104}
3105
Johannes Bergaf728682017-06-16 14:29:22 +02003106void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003107
David S. Miller7ce5a272013-12-02 17:26:05 -05003108/**
WANG Cong82a31b92016-06-30 10:15:22 -07003109 * skb_push_rcsum - push skb and update receive checksum
3110 * @skb: buffer to update
3111 * @len: length of data pulled
3112 *
3113 * This function performs an skb_push on the packet and updates
3114 * the CHECKSUM_COMPLETE checksum. It should be used on
3115 * receive path processing instead of skb_push unless you know
3116 * that the checksum difference is zero (e.g., a valid IP header)
3117 * or you are setting ip_summed to CHECKSUM_NONE.
3118 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003119static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003120{
3121 skb_push(skb, len);
3122 skb_postpush_rcsum(skb, skb->data, len);
3123 return skb->data;
3124}
3125
3126/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003127 * pskb_trim_rcsum - trim received skb and update checksum
3128 * @skb: buffer to trim
3129 * @len: new length
3130 *
3131 * This is exactly the same as pskb_trim except that it ensures the
3132 * checksum of received packets are still valid after the operation.
3133 */
3134
3135static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3136{
3137 if (likely(len >= skb->len))
3138 return 0;
3139 if (skb->ip_summed == CHECKSUM_COMPLETE)
3140 skb->ip_summed = CHECKSUM_NONE;
3141 return __pskb_trim(skb, len);
3142}
3143
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003144static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3145{
3146 if (skb->ip_summed == CHECKSUM_COMPLETE)
3147 skb->ip_summed = CHECKSUM_NONE;
3148 __skb_trim(skb, len);
3149 return 0;
3150}
3151
3152static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3153{
3154 if (skb->ip_summed == CHECKSUM_COMPLETE)
3155 skb->ip_summed = CHECKSUM_NONE;
3156 return __skb_grow(skb, len);
3157}
3158
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003159#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3160#define skb_rb_first(root) rb_to_skb(rb_first(root))
3161#define skb_rb_last(root) rb_to_skb(rb_last(root))
3162#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3163#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3164
Linus Torvalds1da177e2005-04-16 15:20:36 -07003165#define skb_queue_walk(queue, skb) \
3166 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003167 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003168 skb = skb->next)
3169
James Chapman46f89142007-04-30 00:07:31 -07003170#define skb_queue_walk_safe(queue, skb, tmp) \
3171 for (skb = (queue)->next, tmp = skb->next; \
3172 skb != (struct sk_buff *)(queue); \
3173 skb = tmp, tmp = skb->next)
3174
David S. Miller1164f522008-09-23 00:49:44 -07003175#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003176 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003177 skb = skb->next)
3178
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003179#define skb_rbtree_walk(skb, root) \
3180 for (skb = skb_rb_first(root); skb != NULL; \
3181 skb = skb_rb_next(skb))
3182
3183#define skb_rbtree_walk_from(skb) \
3184 for (; skb != NULL; \
3185 skb = skb_rb_next(skb))
3186
3187#define skb_rbtree_walk_from_safe(skb, tmp) \
3188 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3189 skb = tmp)
3190
David S. Miller1164f522008-09-23 00:49:44 -07003191#define skb_queue_walk_from_safe(queue, skb, tmp) \
3192 for (tmp = skb->next; \
3193 skb != (struct sk_buff *)(queue); \
3194 skb = tmp, tmp = skb->next)
3195
Stephen Hemminger300ce172005-10-30 13:47:34 -08003196#define skb_queue_reverse_walk(queue, skb) \
3197 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003198 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003199 skb = skb->prev)
3200
David S. Miller686a2952011-01-20 22:47:32 -08003201#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3202 for (skb = (queue)->prev, tmp = skb->prev; \
3203 skb != (struct sk_buff *)(queue); \
3204 skb = tmp, tmp = skb->prev)
3205
3206#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3207 for (tmp = skb->prev; \
3208 skb != (struct sk_buff *)(queue); \
3209 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003210
David S. Miller21dc3302010-08-23 00:13:46 -07003211static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003212{
3213 return skb_shinfo(skb)->frag_list != NULL;
3214}
3215
3216static inline void skb_frag_list_init(struct sk_buff *skb)
3217{
3218 skb_shinfo(skb)->frag_list = NULL;
3219}
3220
David S. Milleree039872009-06-09 00:17:13 -07003221#define skb_walk_frags(skb, iter) \
3222 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3223
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003224
3225int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3226 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003227struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3228 struct sk_buff_head *queue,
3229 unsigned int flags,
3230 void (*destructor)(struct sock *sk,
3231 struct sk_buff *skb),
3232 int *peeked, int *off, int *err,
3233 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003234struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003235 void (*destructor)(struct sock *sk,
3236 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003237 int *peeked, int *off, int *err,
3238 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003239struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003240 void (*destructor)(struct sock *sk,
3241 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003242 int *peeked, int *off, int *err);
3243struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3244 int *err);
3245unsigned int datagram_poll(struct file *file, struct socket *sock,
3246 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003247int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3248 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003249static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3250 struct msghdr *msg, int size)
3251{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003252 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003253}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003254int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3255 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003256int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3257 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003258int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003259void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003260void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3261static inline void skb_free_datagram_locked(struct sock *sk,
3262 struct sk_buff *skb)
3263{
3264 __skb_free_datagram_locked(sk, skb, 0);
3265}
Joe Perches7965bd42013-09-26 14:48:15 -07003266int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003267int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3268int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3269__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3270 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003271int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003272 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003273 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003274int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3275 int len);
3276int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003277void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003278unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003279int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3280 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003281void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3282int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3283void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003284unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003285bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003286struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003287struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003288int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003289int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003290int skb_vlan_pop(struct sk_buff *skb);
3291int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003292struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3293 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003294
Al Viro6ce8e9c2014-04-06 21:25:44 -04003295static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3296{
Al Viro3073f072017-02-17 23:13:25 -05003297 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003298}
3299
Al Viro7eab8d92014-04-06 21:51:23 -04003300static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3301{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003302 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003303}
3304
Daniel Borkmann2817a332013-10-30 11:50:51 +01003305struct skb_checksum_ops {
3306 __wsum (*update)(const void *mem, int len, __wsum wsum);
3307 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3308};
3309
Davide Caratti96178132017-05-18 15:44:37 +02003310extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3311
Daniel Borkmann2817a332013-10-30 11:50:51 +01003312__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3313 __wsum csum, const struct skb_checksum_ops *ops);
3314__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3315 __wsum csum);
3316
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003317static inline void * __must_check
3318__skb_header_pointer(const struct sk_buff *skb, int offset,
3319 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003321 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003322 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003323
David S. Miller690e36e2014-08-23 12:13:41 -07003324 if (!skb ||
3325 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003326 return NULL;
3327
3328 return buffer;
3329}
3330
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003331static inline void * __must_check
3332skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003333{
3334 return __skb_header_pointer(skb, offset, len, skb->data,
3335 skb_headlen(skb), buffer);
3336}
3337
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003338/**
3339 * skb_needs_linearize - check if we need to linearize a given skb
3340 * depending on the given device features.
3341 * @skb: socket buffer to check
3342 * @features: net device features
3343 *
3344 * Returns true if either:
3345 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3346 * 2. skb is fragmented and the device does not support SG.
3347 */
3348static inline bool skb_needs_linearize(struct sk_buff *skb,
3349 netdev_features_t features)
3350{
3351 return skb_is_nonlinear(skb) &&
3352 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3353 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3354}
3355
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003356static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3357 void *to,
3358 const unsigned int len)
3359{
3360 memcpy(to, skb->data, len);
3361}
3362
3363static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3364 const int offset, void *to,
3365 const unsigned int len)
3366{
3367 memcpy(to, skb->data + offset, len);
3368}
3369
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003370static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3371 const void *from,
3372 const unsigned int len)
3373{
3374 memcpy(skb->data, from, len);
3375}
3376
3377static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3378 const int offset,
3379 const void *from,
3380 const unsigned int len)
3381{
3382 memcpy(skb->data + offset, from, len);
3383}
3384
Joe Perches7965bd42013-09-26 14:48:15 -07003385void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003386
Patrick Ohlyac45f602009-02-12 05:03:37 +00003387static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3388{
3389 return skb->tstamp;
3390}
3391
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003392/**
3393 * skb_get_timestamp - get timestamp from a skb
3394 * @skb: skb to get stamp from
3395 * @stamp: pointer to struct timeval to store stamp in
3396 *
3397 * Timestamps are stored in the skb as offsets to a base timestamp.
3398 * This function converts the offset back to a struct timeval and stores
3399 * it in stamp.
3400 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003401static inline void skb_get_timestamp(const struct sk_buff *skb,
3402 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003403{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003404 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003405}
3406
Patrick Ohlyac45f602009-02-12 05:03:37 +00003407static inline void skb_get_timestampns(const struct sk_buff *skb,
3408 struct timespec *stamp)
3409{
3410 *stamp = ktime_to_timespec(skb->tstamp);
3411}
3412
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003413static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003414{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003415 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003416}
3417
Stephen Hemminger164891a2007-04-23 22:26:16 -07003418static inline ktime_t net_timedelta(ktime_t t)
3419{
3420 return ktime_sub(ktime_get_real(), t);
3421}
3422
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003423static inline ktime_t net_invalid_timestamp(void)
3424{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003425 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003426}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003427
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003428static inline u8 skb_metadata_len(const struct sk_buff *skb)
3429{
3430 return skb_shinfo(skb)->meta_len;
3431}
3432
3433static inline void *skb_metadata_end(const struct sk_buff *skb)
3434{
3435 return skb_mac_header(skb);
3436}
3437
3438static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3439 const struct sk_buff *skb_b,
3440 u8 meta_len)
3441{
3442 const void *a = skb_metadata_end(skb_a);
3443 const void *b = skb_metadata_end(skb_b);
3444 /* Using more efficient varaiant than plain call to memcmp(). */
3445#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3446 u64 diffs = 0;
3447
3448 switch (meta_len) {
3449#define __it(x, op) (x -= sizeof(u##op))
3450#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3451 case 32: diffs |= __it_diff(a, b, 64);
3452 case 24: diffs |= __it_diff(a, b, 64);
3453 case 16: diffs |= __it_diff(a, b, 64);
3454 case 8: diffs |= __it_diff(a, b, 64);
3455 break;
3456 case 28: diffs |= __it_diff(a, b, 64);
3457 case 20: diffs |= __it_diff(a, b, 64);
3458 case 12: diffs |= __it_diff(a, b, 64);
3459 case 4: diffs |= __it_diff(a, b, 32);
3460 break;
3461 }
3462 return diffs;
3463#else
3464 return memcmp(a - meta_len, b - meta_len, meta_len);
3465#endif
3466}
3467
3468static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3469 const struct sk_buff *skb_b)
3470{
3471 u8 len_a = skb_metadata_len(skb_a);
3472 u8 len_b = skb_metadata_len(skb_b);
3473
3474 if (!(len_a | len_b))
3475 return false;
3476
3477 return len_a != len_b ?
3478 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3479}
3480
3481static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3482{
3483 skb_shinfo(skb)->meta_len = meta_len;
3484}
3485
3486static inline void skb_metadata_clear(struct sk_buff *skb)
3487{
3488 skb_metadata_set(skb, 0);
3489}
3490
Alexander Duyck62bccb82014-09-04 13:31:35 -04003491struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3492
Richard Cochranc1f19b52010-07-17 08:49:36 +00003493#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3494
Joe Perches7965bd42013-09-26 14:48:15 -07003495void skb_clone_tx_timestamp(struct sk_buff *skb);
3496bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003497
3498#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3499
3500static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3501{
3502}
3503
3504static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3505{
3506 return false;
3507}
3508
3509#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3510
3511/**
3512 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3513 *
Richard Cochranda92b192011-10-21 00:49:15 +00003514 * PHY drivers may accept clones of transmitted packets for
3515 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003516 * must call this function to return the skb back to the stack with a
3517 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003518 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003519 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003520 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003521 *
3522 */
3523void skb_complete_tx_timestamp(struct sk_buff *skb,
3524 struct skb_shared_hwtstamps *hwtstamps);
3525
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003526void __skb_tstamp_tx(struct sk_buff *orig_skb,
3527 struct skb_shared_hwtstamps *hwtstamps,
3528 struct sock *sk, int tstype);
3529
Patrick Ohlyac45f602009-02-12 05:03:37 +00003530/**
3531 * skb_tstamp_tx - queue clone of skb with send time stamps
3532 * @orig_skb: the original outgoing packet
3533 * @hwtstamps: hardware time stamps, may be NULL if not available
3534 *
3535 * If the skb has a socket associated, then this function clones the
3536 * skb (thus sharing the actual data and optional structures), stores
3537 * the optional hardware time stamping information (if non NULL) or
3538 * generates a software time stamp (otherwise), then queues the clone
3539 * to the error queue of the socket. Errors are silently ignored.
3540 */
Joe Perches7965bd42013-09-26 14:48:15 -07003541void skb_tstamp_tx(struct sk_buff *orig_skb,
3542 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003543
Richard Cochran4507a712010-07-17 08:48:28 +00003544/**
3545 * skb_tx_timestamp() - Driver hook for transmit timestamping
3546 *
3547 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003548 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003549 *
David S. Miller73409f32013-12-27 13:04:33 -05003550 * Specifically, one should make absolutely sure that this function is
3551 * called before TX completion of this packet can trigger. Otherwise
3552 * the packet could potentially already be freed.
3553 *
Richard Cochran4507a712010-07-17 08:48:28 +00003554 * @skb: A socket buffer.
3555 */
3556static inline void skb_tx_timestamp(struct sk_buff *skb)
3557{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003558 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003559 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3560 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003561}
3562
Johannes Berg6e3e9392011-11-09 10:15:42 +01003563/**
3564 * skb_complete_wifi_ack - deliver skb with wifi status
3565 *
3566 * @skb: the original outgoing packet
3567 * @acked: ack status
3568 *
3569 */
3570void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3571
Joe Perches7965bd42013-09-26 14:48:15 -07003572__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3573__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003574
Herbert Xu60476372007-04-09 11:59:39 -07003575static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3576{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003577 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3578 skb->csum_valid ||
3579 (skb->ip_summed == CHECKSUM_PARTIAL &&
3580 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003581}
3582
Herbert Xufb286bb2005-11-10 13:01:24 -08003583/**
3584 * skb_checksum_complete - Calculate checksum of an entire packet
3585 * @skb: packet to process
3586 *
3587 * This function calculates the checksum over the entire packet plus
3588 * the value of skb->csum. The latter can be used to supply the
3589 * checksum of a pseudo header as used by TCP/UDP. It returns the
3590 * checksum.
3591 *
3592 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3593 * this function can be used to verify that checksum on received
3594 * packets. In that case the function should return zero if the
3595 * checksum is correct. In particular, this function will return zero
3596 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3597 * hardware has already verified the correctness of the checksum.
3598 */
Al Viro4381ca32007-07-15 21:00:11 +01003599static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003600{
Herbert Xu60476372007-04-09 11:59:39 -07003601 return skb_csum_unnecessary(skb) ?
3602 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003603}
3604
Tom Herbert77cffe22014-08-27 21:26:46 -07003605static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3606{
3607 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3608 if (skb->csum_level == 0)
3609 skb->ip_summed = CHECKSUM_NONE;
3610 else
3611 skb->csum_level--;
3612 }
3613}
3614
3615static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3616{
3617 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3618 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3619 skb->csum_level++;
3620 } else if (skb->ip_summed == CHECKSUM_NONE) {
3621 skb->ip_summed = CHECKSUM_UNNECESSARY;
3622 skb->csum_level = 0;
3623 }
3624}
3625
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003626/* Check if we need to perform checksum complete validation.
3627 *
3628 * Returns true if checksum complete is needed, false otherwise
3629 * (either checksum is unnecessary or zero checksum is allowed).
3630 */
3631static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3632 bool zero_okay,
3633 __sum16 check)
3634{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003635 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3636 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003637 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003638 return false;
3639 }
3640
3641 return true;
3642}
3643
3644/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3645 * in checksum_init.
3646 */
3647#define CHECKSUM_BREAK 76
3648
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003649/* Unset checksum-complete
3650 *
3651 * Unset checksum complete can be done when packet is being modified
3652 * (uncompressed for instance) and checksum-complete value is
3653 * invalidated.
3654 */
3655static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3656{
3657 if (skb->ip_summed == CHECKSUM_COMPLETE)
3658 skb->ip_summed = CHECKSUM_NONE;
3659}
3660
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003661/* Validate (init) checksum based on checksum complete.
3662 *
3663 * Return values:
3664 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3665 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3666 * checksum is stored in skb->csum for use in __skb_checksum_complete
3667 * non-zero: value of invalid checksum
3668 *
3669 */
3670static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3671 bool complete,
3672 __wsum psum)
3673{
3674 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3675 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003676 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003677 return 0;
3678 }
3679 }
3680
3681 skb->csum = psum;
3682
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003683 if (complete || skb->len <= CHECKSUM_BREAK) {
3684 __sum16 csum;
3685
3686 csum = __skb_checksum_complete(skb);
3687 skb->csum_valid = !csum;
3688 return csum;
3689 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003690
3691 return 0;
3692}
3693
3694static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3695{
3696 return 0;
3697}
3698
3699/* Perform checksum validate (init). Note that this is a macro since we only
3700 * want to calculate the pseudo header which is an input function if necessary.
3701 * First we try to validate without any computation (checksum unnecessary) and
3702 * then calculate based on checksum complete calling the function to compute
3703 * pseudo header.
3704 *
3705 * Return values:
3706 * 0: checksum is validated or try to in skb_checksum_complete
3707 * non-zero: value of invalid checksum
3708 */
3709#define __skb_checksum_validate(skb, proto, complete, \
3710 zero_okay, check, compute_pseudo) \
3711({ \
3712 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003713 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003714 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3715 __ret = __skb_checksum_validate_complete(skb, \
3716 complete, compute_pseudo(skb, proto)); \
3717 __ret; \
3718})
3719
3720#define skb_checksum_init(skb, proto, compute_pseudo) \
3721 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3722
3723#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3724 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3725
3726#define skb_checksum_validate(skb, proto, compute_pseudo) \
3727 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3728
3729#define skb_checksum_validate_zero_check(skb, proto, check, \
3730 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003731 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003732
3733#define skb_checksum_simple_validate(skb) \
3734 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3735
Tom Herbertd96535a2014-08-31 15:12:42 -07003736static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3737{
Davide Caratti219f1d792017-05-18 15:44:39 +02003738 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003739}
3740
3741static inline void __skb_checksum_convert(struct sk_buff *skb,
3742 __sum16 check, __wsum pseudo)
3743{
3744 skb->csum = ~pseudo;
3745 skb->ip_summed = CHECKSUM_COMPLETE;
3746}
3747
3748#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3749do { \
3750 if (__skb_checksum_convert_check(skb)) \
3751 __skb_checksum_convert(skb, check, \
3752 compute_pseudo(skb, proto)); \
3753} while (0)
3754
Tom Herbert15e23962015-02-10 16:30:31 -08003755static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3756 u16 start, u16 offset)
3757{
3758 skb->ip_summed = CHECKSUM_PARTIAL;
3759 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3760 skb->csum_offset = offset - start;
3761}
3762
Tom Herbertdcdc8992015-02-02 16:07:34 -08003763/* Update skbuf and packet to reflect the remote checksum offload operation.
3764 * When called, ptr indicates the starting point for skb->csum when
3765 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3766 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3767 */
3768static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003769 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003770{
3771 __wsum delta;
3772
Tom Herbert15e23962015-02-10 16:30:31 -08003773 if (!nopartial) {
3774 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3775 return;
3776 }
3777
Tom Herbertdcdc8992015-02-02 16:07:34 -08003778 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3779 __skb_checksum_complete(skb);
3780 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3781 }
3782
3783 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3784
3785 /* Adjust skb->csum since we changed the packet */
3786 skb->csum = csum_add(skb->csum, delta);
3787}
3788
Florian Westphalcb9c6832017-01-23 18:21:56 +01003789static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3790{
3791#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003792 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003793#else
3794 return NULL;
3795#endif
3796}
3797
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003798#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003799void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003800static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3801{
3802 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003803 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003804}
3805static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3806{
3807 if (nfct)
3808 atomic_inc(&nfct->use);
3809}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003810#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003811#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003812static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3813{
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003814 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003815 kfree(nf_bridge);
3816}
3817static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3818{
3819 if (nf_bridge)
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003820 refcount_inc(&nf_bridge->use);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003821}
3822#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003823static inline void nf_reset(struct sk_buff *skb)
3824{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003825#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003826 nf_conntrack_put(skb_nfct(skb));
3827 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003828#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003829#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003830 nf_bridge_put(skb->nf_bridge);
3831 skb->nf_bridge = NULL;
3832#endif
3833}
3834
Patrick McHardy124dff02013-04-05 20:42:05 +02003835static inline void nf_reset_trace(struct sk_buff *skb)
3836{
Florian Westphal478b3602014-02-15 23:48:45 +01003837#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003838 skb->nf_trace = 0;
3839#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003840}
3841
Ye Yin2b5ec1a2017-10-26 16:57:05 +08003842static inline void ipvs_reset(struct sk_buff *skb)
3843{
3844#if IS_ENABLED(CONFIG_IP_VS)
3845 skb->ipvs_property = 0;
3846#endif
3847}
3848
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003849/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003850static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3851 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003852{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003853#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003854 dst->_nfct = src->_nfct;
3855 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003856#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003857#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003858 dst->nf_bridge = src->nf_bridge;
3859 nf_bridge_get(src->nf_bridge);
3860#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003861#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003862 if (copy)
3863 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003864#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003865}
3866
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003867static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3868{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003869#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003870 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003871#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003872#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003873 nf_bridge_put(dst->nf_bridge);
3874#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003875 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003876}
3877
James Morris984bc162006-06-09 00:29:17 -07003878#ifdef CONFIG_NETWORK_SECMARK
3879static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3880{
3881 to->secmark = from->secmark;
3882}
3883
3884static inline void skb_init_secmark(struct sk_buff *skb)
3885{
3886 skb->secmark = 0;
3887}
3888#else
3889static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3890{ }
3891
3892static inline void skb_init_secmark(struct sk_buff *skb)
3893{ }
3894#endif
3895
Eric W. Biederman574f7192014-04-01 12:20:24 -07003896static inline bool skb_irq_freeable(const struct sk_buff *skb)
3897{
3898 return !skb->destructor &&
3899#if IS_ENABLED(CONFIG_XFRM)
3900 !skb->sp &&
3901#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003902 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003903 !skb->_skb_refdst &&
3904 !skb_has_frag_list(skb);
3905}
3906
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003907static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3908{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003909 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003910}
3911
Stephen Hemminger92477442009-03-21 13:39:26 -07003912static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003913{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003914 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003915}
3916
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003917static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3918{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003919 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003920}
3921
David S. Millerd5a9e242009-01-27 16:22:11 -08003922static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3923{
3924 skb->queue_mapping = rx_queue + 1;
3925}
3926
Stephen Hemminger92477442009-03-21 13:39:26 -07003927static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003928{
3929 return skb->queue_mapping - 1;
3930}
3931
Stephen Hemminger92477442009-03-21 13:39:26 -07003932static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003933{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003934 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003935}
3936
Julian Anastasov4ff06202017-02-06 23:14:12 +02003937static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3938{
3939 skb->dst_pending_confirm = val;
3940}
3941
3942static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3943{
3944 return skb->dst_pending_confirm != 0;
3945}
3946
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003947static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3948{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003949#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003950 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003951#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003952 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003953#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003954}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003955
Pravin B Shelar68c33162013-02-14 14:02:41 +00003956/* Keeps track of mac header offset relative to skb->head.
3957 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3958 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003959 * tunnel skb it points to outer mac header.
3960 * Keeps track of level of encapsulation of network headers.
3961 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003962struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003963 union {
3964 int mac_offset;
3965 int data_offset;
3966 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003967 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003968 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003969 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003970};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003971#define SKB_SGO_CB_OFFSET 32
3972#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003973
3974static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3975{
3976 return (skb_mac_header(inner_skb) - inner_skb->head) -
3977 SKB_GSO_CB(inner_skb)->mac_offset;
3978}
3979
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003980static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3981{
3982 int new_headroom, headroom;
3983 int ret;
3984
3985 headroom = skb_headroom(skb);
3986 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3987 if (ret)
3988 return ret;
3989
3990 new_headroom = skb_headroom(skb);
3991 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3992 return 0;
3993}
3994
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003995static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3996{
3997 /* Do not update partial checksums if remote checksum is enabled. */
3998 if (skb->remcsum_offload)
3999 return;
4000
4001 SKB_GSO_CB(skb)->csum = res;
4002 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4003}
4004
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004005/* Compute the checksum for a gso segment. First compute the checksum value
4006 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4007 * then add in skb->csum (checksum from csum_start to end of packet).
4008 * skb->csum and csum_start are then updated to reflect the checksum of the
4009 * resultant packet starting from the transport header-- the resultant checksum
4010 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4011 * header.
4012 */
4013static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4014{
Alexander Duyck76443452016-02-05 15:27:37 -08004015 unsigned char *csum_start = skb_transport_header(skb);
4016 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4017 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004018
Alexander Duyck76443452016-02-05 15:27:37 -08004019 SKB_GSO_CB(skb)->csum = res;
4020 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004021
Alexander Duyck76443452016-02-05 15:27:37 -08004022 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004023}
4024
David S. Millerbdcc0922012-03-07 20:53:36 -05004025static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07004026{
4027 return skb_shinfo(skb)->gso_size;
4028}
4029
Eric Dumazet36a8f392013-09-29 01:21:32 -07004030/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004031static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004032{
4033 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4034}
4035
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004036static inline void skb_gso_reset(struct sk_buff *skb)
4037{
4038 skb_shinfo(skb)->gso_size = 0;
4039 skb_shinfo(skb)->gso_segs = 0;
4040 skb_shinfo(skb)->gso_type = 0;
4041}
4042
Joe Perches7965bd42013-09-26 14:48:15 -07004043void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004044
4045static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4046{
4047 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4048 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004049 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4050
Alexander Duyckb78462e2010-06-02 12:24:37 +00004051 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4052 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004053 __skb_warn_lro_forwarding(skb);
4054 return true;
4055 }
4056 return false;
4057}
4058
Herbert Xu35fc92a2007-03-26 23:22:20 -07004059static inline void skb_forward_csum(struct sk_buff *skb)
4060{
4061 /* Unfortunately we don't support this one. Any brave souls? */
4062 if (skb->ip_summed == CHECKSUM_COMPLETE)
4063 skb->ip_summed = CHECKSUM_NONE;
4064}
4065
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004066/**
4067 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4068 * @skb: skb to check
4069 *
4070 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4071 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4072 * use this helper, to document places where we make this assertion.
4073 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004074static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004075{
4076#ifdef DEBUG
4077 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4078#endif
4079}
4080
Rusty Russellf35d9d82008-02-04 23:49:54 -05004081bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004082
Paul Durranted1f50c2014-01-09 10:02:46 +00004083int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004084struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4085 unsigned int transport_len,
4086 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004087
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004088/**
4089 * skb_head_is_locked - Determine if the skb->head is locked down
4090 * @skb: skb to check
4091 *
4092 * The head on skbs build around a head frag can be removed if they are
4093 * not cloned. This function returns true if the skb head is locked down
4094 * due to either being allocated via kmalloc, or by being a clone with
4095 * multiple references to the head.
4096 */
4097static inline bool skb_head_is_locked(const struct sk_buff *skb)
4098{
4099 return !skb->head_frag || skb_cloned(skb);
4100}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004101
4102/**
4103 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4104 *
4105 * @skb: GSO skb
4106 *
4107 * skb_gso_network_seglen is used to determine the real size of the
4108 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4109 *
4110 * The MAC/L2 header is not accounted for.
4111 */
4112static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4113{
4114 unsigned int hdr_len = skb_transport_header(skb) -
4115 skb_network_header(skb);
4116 return hdr_len + skb_gso_transport_seglen(skb);
4117}
Thomas Grafee122c72015-07-21 10:43:58 +02004118
Edward Cree179bc672016-02-11 20:48:04 +00004119/* Local Checksum Offload.
4120 * Compute outer checksum based on the assumption that the
4121 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00004122 * See Documentation/networking/checksum-offloads.txt for
4123 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004124 * Fill in outer checksum adjustment (e.g. with sum of outer
4125 * pseudo-header) before calling.
4126 * Also ensure that inner checksum is in linear data area.
4127 */
4128static inline __wsum lco_csum(struct sk_buff *skb)
4129{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004130 unsigned char *csum_start = skb_checksum_start(skb);
4131 unsigned char *l4_hdr = skb_transport_header(skb);
4132 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004133
4134 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004135 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4136 skb->csum_offset));
4137
Edward Cree179bc672016-02-11 20:48:04 +00004138 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004139 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004140 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004141 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004142}
4143
Linus Torvalds1da177e2005-04-16 15:20:36 -07004144#endif /* __KERNEL__ */
4145#endif /* _LINUX_SKBUFF_H */