blob: 19e64bfb1a66974b10abc476bfcadeee804f88fe [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Eric Dumazetc1d1b432017-08-31 16:48:22 -070025#include <linux/refcount.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
Arun Sharma600634972011-07-26 16:09:06 -070027#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <asm/types.h>
29#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070031#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000033#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070034#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000035#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000036#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080037#include <linux/sched.h>
Ingo Molnare6017572017-02-01 16:36:40 +010038#include <linux/sched/clock.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020039#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020040#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020041#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040042#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070043#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
Tom Herbert7a6ae712015-12-14 11:19:47 -080045/* The interface for checksum offload between the stack and networking drivers
46 * is as follows...
47 *
48 * A. IP checksum related features
49 *
50 * Drivers advertise checksum offload capabilities in the features of a device.
51 * From the stack's point of view these are capabilities offered by the driver,
52 * a driver typically only advertises features that it is capable of offloading
53 * to its device.
54 *
55 * The checksum related features are:
56 *
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
62 *
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
67 * is TCP or UDP. The IPv4 header may contain IP options
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
71 *
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
75 * IPv4|UDP where the Next Header field in the IPv6
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
81 *
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
83 * This flag is used only used to disable the RX checksum
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
87 *
88 * B. Checksumming of received packets by device. Indication of checksum
89 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010090 *
91 * CHECKSUM_NONE:
92 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080093 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010094 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
96 *
97 * CHECKSUM_UNNECESSARY:
98 *
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700105 *
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200113 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700114 *
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
119 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
120 * two. If the device were only able to verify the UDP checksum and not
121 * GRE, either because it doesn't support GRE checksum of because GRE
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100124 *
125 * CHECKSUM_COMPLETE:
126 *
127 * This is the most generic way. The device supplied checksum of the _whole_
128 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
129 * hardware doesn't need to parse L3/L4 headers to implement this.
130 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200131 * Notes:
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100135 *
136 * CHECKSUM_PARTIAL:
137 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
146 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100147 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
150 *
151 * CHECKSUM_PARTIAL:
152 *
153 * The driver is required to checksum the packet as seen by hard_start_xmit()
154 * from skb->csum_start up to the end, and to record/write the checksum at
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
157 * offset of the packet, however they should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum-- it is the
159 * purview of the stack to validate that csum_start and csum_offset are set
160 * correctly.
161 *
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
166 *
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100174 *
175 * CHECKSUM_NONE:
176 *
177 * The skb was already checksummed by the protocol, or a checksum is not
178 * required.
179 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100180 * CHECKSUM_UNNECESSARY:
181 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800182 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
183 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100184 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800185 * CHECKSUM_COMPLETE:
186 * Not used in checksum output. If a driver observes a packet with this value
187 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100188 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800189 * D. Non-IP checksum (CRC) offloads
190 *
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
Davide Carattidba00302017-05-18 15:44:40 +0200193 * will set set csum_start and csum_offset accordingly, set ip_summed to
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800200 *
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
204 * accordingly. Note the there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
206 * both IP checksum offload and FCOE CRC offload must verify which offload
207 * is configured for a packet presumably by inspecting packet headers.
208 *
209 * E. Checksumming on output with GSO.
210 *
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
216 * are set to refer to the outermost checksum being offload (two offloaded
217 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100218 */
219
Herbert Xu60476372007-04-09 11:59:39 -0700220/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100221#define CHECKSUM_NONE 0
222#define CHECKSUM_UNNECESSARY 1
223#define CHECKSUM_COMPLETE 2
224#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
Tom Herbert77cffe22014-08-27 21:26:46 -0700226/* Maximum value in skb->csum_level */
227#define SKB_MAX_CSUM_LEVEL 3
228
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200229#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700230#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700232#define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
236
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000237/* return minimum truesize of one skb containing X bytes of data */
238#define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
241
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700243struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800244struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800245struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800246struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700248#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249struct nf_conntrack {
250 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700252#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200254#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255struct nf_bridge_info {
Reshetova, Elena53869ce2017-06-30 13:07:57 +0300256 refcount_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200257 enum {
258 BRNF_PROTO_UNCHANGED,
259 BRNF_PROTO_8021Q,
260 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200261 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200262 u8 pkt_otherhost:1;
263 u8 in_prerouting:1;
264 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200265 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000266 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200267
268 /* always valid & non-NULL from FORWARD on, for physdev match */
269 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200270 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200271 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200272 __be32 ipv4_daddr;
273 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200274
275 /* after prerouting + nat detected: store original source
276 * mac since neigh resolution overwrites it, only used while
277 * skb is out in neigh layer.
278 */
279 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200280 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281};
282#endif
283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284struct sk_buff_head {
285 /* These two members must be first. */
286 struct sk_buff *next;
287 struct sk_buff *prev;
288
289 __u32 qlen;
290 spinlock_t lock;
291};
292
293struct sk_buff;
294
Ian Campbell9d4dde52011-12-22 23:39:14 +0000295/* To allow 64K frame to be packed as single skb without frag_list we
296 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
297 * buffers which do not start on a page boundary.
298 *
299 * Since GRO uses frags we allocate at least 16 regardless of page
300 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000301 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000302#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700303#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000304#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000305#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000306#endif
Hans Westgaard Ry5f74f822016-02-03 09:26:57 +0100307extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300309/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
310 * segment using its current segmentation instead.
311 */
312#define GSO_BY_FRAGS 0xFFFF
313
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314typedef struct skb_frag_struct skb_frag_t;
315
316struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000317 struct {
318 struct page *p;
319 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000320#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700321 __u32 page_offset;
322 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000323#else
324 __u16 page_offset;
325 __u16 size;
326#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327};
328
Eric Dumazet9e903e02011-10-18 21:00:24 +0000329static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330{
331 return frag->size;
332}
333
334static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
335{
336 frag->size = size;
337}
338
339static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
340{
341 frag->size += delta;
342}
343
344static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
345{
346 frag->size -= delta;
347}
348
Willem de Bruijnc613c202017-07-31 08:15:47 -0400349static inline bool skb_frag_must_loop(struct page *p)
350{
351#if defined(CONFIG_HIGHMEM)
352 if (PageHighMem(p))
353 return true;
354#endif
355 return false;
356}
357
358/**
359 * skb_frag_foreach_page - loop over pages in a fragment
360 *
361 * @f: skb frag to operate on
362 * @f_off: offset from start of f->page.p
363 * @f_len: length from f_off to loop over
364 * @p: (temp var) current page
365 * @p_off: (temp var) offset from start of current page,
366 * non-zero only on first page.
367 * @p_len: (temp var) length in current page,
368 * < PAGE_SIZE only on first and last page.
369 * @copied: (temp var) length so far, excluding current p_len.
370 *
371 * A fragment can hold a compound page, in which case per-page
372 * operations, notably kmap_atomic, must be called for each
373 * regular page.
374 */
375#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
376 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
377 p_off = (f_off) & (PAGE_SIZE - 1), \
378 p_len = skb_frag_must_loop(p) ? \
379 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
380 copied = 0; \
381 copied < f_len; \
382 copied += p_len, p++, p_off = 0, \
383 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
384
Patrick Ohlyac45f602009-02-12 05:03:37 +0000385#define HAVE_HW_TIME_STAMP
386
387/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800388 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000389 * @hwtstamp: hardware time stamp transformed into duration
390 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000391 *
392 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400393 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000394 *
395 * hwtstamps can only be compared against other hwtstamps from
396 * the same device.
397 *
398 * This structure is attached to packets as part of the
399 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
400 */
401struct skb_shared_hwtstamps {
402 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000403};
404
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000405/* Definitions for tx_flags in struct skb_shared_info */
406enum {
407 /* generate hardware time stamp */
408 SKBTX_HW_TSTAMP = 1 << 0,
409
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400410 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000411 SKBTX_SW_TSTAMP = 1 << 1,
412
413 /* device driver is going to provide hardware time stamp */
414 SKBTX_IN_PROGRESS = 1 << 2,
415
Shirley Maa6686f22011-07-06 12:22:12 +0000416 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000417 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100418
419 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000420 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000421
422 /* This indicates at least one fragment might be overwritten
423 * (as in vmsplice(), sendfile() ...)
424 * If we need to compute a TX checksum, we'll need to copy
425 * all frags to avoid possible bad checksum
426 */
427 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400428
429 /* generate software time stamp when entering packet scheduling */
430 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000431};
432
Willem de Bruijn52267792017-08-03 16:29:39 -0400433#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400434#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400435 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400436#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
437
Shirley Maa6686f22011-07-06 12:22:12 +0000438/*
439 * The callback notifies userspace to release buffers when skb DMA is done in
440 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000441 * The zerocopy_success argument is true if zero copy transmit occurred,
442 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000443 * The ctx field is used to track device context.
444 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000445 */
446struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000447 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400448 union {
449 struct {
450 unsigned long desc;
451 void *ctx;
452 };
453 struct {
454 u32 id;
455 u16 len;
456 u16 zerocopy:1;
457 u32 bytelen;
458 };
459 };
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700460 refcount_t refcnt;
Willem de Bruijna91dbff2017-08-03 16:29:43 -0400461
462 struct mmpin {
463 struct user_struct *user;
464 unsigned int num_pg;
465 } mmp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000466};
467
Willem de Bruijn52267792017-08-03 16:29:39 -0400468#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
469
470struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400471struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
472 struct ubuf_info *uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -0400473
474static inline void sock_zerocopy_get(struct ubuf_info *uarg)
475{
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700476 refcount_inc(&uarg->refcnt);
Willem de Bruijn52267792017-08-03 16:29:39 -0400477}
478
479void sock_zerocopy_put(struct ubuf_info *uarg);
480void sock_zerocopy_put_abort(struct ubuf_info *uarg);
481
482void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
483
484int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
485 struct msghdr *msg, int len,
486 struct ubuf_info *uarg);
487
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488/* This data is invariant across clones and lives at
489 * the end of the header data, ie. at skb->end.
490 */
491struct skb_shared_info {
Daniel Borkmannde8f3a82017-09-25 02:25:51 +0200492 __u8 __unused;
493 __u8 meta_len;
494 __u8 nr_frags;
Ian Campbell9f42f122012-01-05 07:13:39 +0000495 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700496 unsigned short gso_size;
497 /* Warning: this field is not always filled in (UFO)! */
498 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000500 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200501 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400502 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000503 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700504
505 /*
506 * Warning : all fields before dataref are cleared in __alloc_skb()
507 */
508 atomic_t dataref;
509
Johann Baudy69e3c752009-05-18 22:11:22 -0700510 /* Intermediate layers must ensure that destructor_arg
511 * remains valid until skb destructor */
512 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000513
Eric Dumazetfed66382010-07-22 19:09:08 +0000514 /* must be last field, see pskb_expand_head() */
515 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516};
517
518/* We divide dataref into two halves. The higher 16 bits hold references
519 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700520 * the entire skb->data. A clone of a headerless skb holds the length of
521 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 *
523 * All users must obey the rule that the skb->data reference count must be
524 * greater than or equal to the payload reference count.
525 *
526 * Holding a reference to the payload part means that the user does not
527 * care about modifications to the header part of skb->data.
528 */
529#define SKB_DATAREF_SHIFT 16
530#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
531
David S. Millerd179cd12005-08-17 14:57:30 -0700532
533enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700534 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
535 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
536 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700537};
538
Herbert Xu79671682006-06-22 02:40:14 -0700539enum {
540 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700541
542 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700543 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700544
545 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700546 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700547
David S. Millerd9d30ad2017-07-03 07:31:57 -0700548 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800549
David S. Millerd9d30ad2017-07-03 07:31:57 -0700550 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000551
David S. Millerd9d30ad2017-07-03 07:31:57 -0700552 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000553
David S. Millerd9d30ad2017-07-03 07:31:57 -0700554 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000555
David S. Millerd9d30ad2017-07-03 07:31:57 -0700556 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700557
David S. Millerd9d30ad2017-07-03 07:31:57 -0700558 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700559
David S. Millerd9d30ad2017-07-03 07:31:57 -0700560 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700561
David S. Millerd9d30ad2017-07-03 07:31:57 -0700562 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700563
David S. Millerd9d30ad2017-07-03 07:31:57 -0700564 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400565
David S. Millerd9d30ad2017-07-03 07:31:57 -0700566 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400567
David S. Millerd9d30ad2017-07-03 07:31:57 -0700568 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300569
David S. Millerd9d30ad2017-07-03 07:31:57 -0700570 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200571
David S. Millerd9d30ad2017-07-03 07:31:57 -0700572 SKB_GSO_ESP = 1 << 15,
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
Florian Westphala9e419d2017-01-23 18:21:59 +0100620 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800622 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700624 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700625 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700626 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700627 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700628 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700629 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700630 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700631 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100632 * @wifi_acked_valid: wifi_acked was set
633 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000634 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200635 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200636 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300637 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700638 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700639 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000640 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700641 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000642 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000643 * @inner_transport_header: Inner transport layer header (encapsulation)
644 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000645 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700646 * @transport_header: Transport layer header
647 * @network_header: Network layer header
648 * @mac_header: Link layer header
649 * @tail: Tail pointer
650 * @end: End pointer
651 * @head: Head of buffer
652 * @data: Data head pointer
653 * @truesize: Buffer size
654 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 */
656
657struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800658 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800659 struct {
660 /* These two members must be first. */
661 struct sk_buff *next;
662 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000663
Eric Dumazet56b17422014-11-03 08:19:53 -0800664 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700665 struct net_device *dev;
666 /* Some protocols might use this space to store information,
667 * while device pointer would be NULL.
668 * UDP receive path is one user.
669 */
670 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800671 };
672 };
673 struct rb_node rbnode; /* used in netem & tcp stack */
674 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000675 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
Eric Dumazetc84d9492016-12-08 11:41:55 -0800677 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700678 ktime_t tstamp;
679 u64 skb_mstamp;
Eric Dumazetc84d9492016-12-08 11:41:55 -0800680 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 /*
682 * This is the control buffer. It is free to use for every
683 * layer. Please put your private variables there. If you
684 * want to keep them across layers you have to do a skb_clone()
685 * first. This is owned by whoever has the skb queued ATM.
686 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000687 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Eric Dumazet7fee2262010-05-11 23:19:48 +0000689 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700690 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000691#ifdef CONFIG_XFRM
692 struct sec_path *sp;
693#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700694#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100695 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700696#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200697#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700698 struct nf_bridge_info *nf_bridge;
699#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700701 data_len;
702 __u16 mac_len,
703 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700704
705 /* Following fields are _not_ copied in __copy_skb_header()
706 * Note that queue_mapping is here mostly to fill a hole.
707 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200708 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700709 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200710
711/* if you move cloned around you also must adapt those constants */
712#ifdef __BIG_ENDIAN_BITFIELD
713#define CLONED_MASK (1 << 7)
714#else
715#define CLONED_MASK 1
716#endif
717#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
718
719 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700720 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700721 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800722 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800723 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700724 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200725 xmit_more:1,
726 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200727 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
Eric Dumazetb1937222014-09-28 22:18:47 -0700729 /* fields enclosed in headers_start/headers_end are copied
730 * using a single memcpy() in __copy_skb_header()
731 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700732 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700733 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700734 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200735
736/* if you move pkt_type around you also must adapt those constants */
737#ifdef __BIG_ENDIAN_BITFIELD
738#define PKT_TYPE_MAX (7 << 5)
739#else
740#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100741#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200742#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
743
744 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700745 __u8 pkt_type:3;
746 __u8 pfmemalloc:1;
747 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748
Eric Dumazetb1937222014-09-28 22:18:47 -0700749 __u8 nf_trace:1;
750 __u8 ip_summed:2;
751 __u8 ooo_okay:1;
752 __u8 l4_hash:1;
753 __u8 sw_hash:1;
754 __u8 wifi_acked_valid:1;
755 __u8 wifi_acked:1;
756
757 __u8 no_fcs:1;
758 /* Indicates the inner headers are valid in the skbuff. */
759 __u8 encapsulation:1;
760 __u8 encap_hdr_csum:1;
761 __u8 csum_valid:1;
762 __u8 csum_complete_sw:1;
763 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200764 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700765
Julian Anastasov4ff06202017-02-06 23:14:12 +0200766 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700767#ifdef CONFIG_IPV6_NDISC_NODETYPE
768 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700770 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700771 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800772 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200773#ifdef CONFIG_NET_SWITCHDEV
774 __u8 offload_fwd_mark:1;
775#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500776#ifdef CONFIG_NET_CLS_ACT
777 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500778 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500779 __u8 tc_redirected:1;
780 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500781#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700784 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200786
Eric Dumazetb1937222014-09-28 22:18:47 -0700787 union {
788 __wsum csum;
789 struct {
790 __u16 csum_start;
791 __u16 csum_offset;
792 };
793 };
794 __u32 priority;
795 int skb_iif;
796 __u32 hash;
797 __be16 vlan_proto;
798 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800799#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
800 union {
801 unsigned int napi_id;
802 unsigned int sender_cpu;
803 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700804#endif
James Morris984bc162006-06-09 00:29:17 -0700805#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200806 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700807#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700808
Neil Horman3b885782009-10-12 13:26:31 -0700809 union {
810 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000811 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700812 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Tom Herbert8bce6d72014-09-29 20:22:29 -0700814 union {
815 __be16 inner_protocol;
816 __u8 inner_ipproto;
817 };
818
Simon Horman1a37e412013-05-23 21:02:51 +0000819 __u16 inner_transport_header;
820 __u16 inner_network_header;
821 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700822
823 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000824 __u16 transport_header;
825 __u16 network_header;
826 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700827
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700828 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700829 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700830 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700831
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700833 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700834 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700836 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700837 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300838 refcount_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839};
840
841#ifdef __KERNEL__
842/*
843 * Handling routines are only of interest to the kernel
844 */
845#include <linux/slab.h>
846
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
Mel Gormanc93bdd02012-07-31 16:44:19 -0700848#define SKB_ALLOC_FCLONE 0x01
849#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800850#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700851
852/* Returns true if the skb was allocated from PFMEMALLOC reserves */
853static inline bool skb_pfmemalloc(const struct sk_buff *skb)
854{
855 return unlikely(skb->pfmemalloc);
856}
857
Eric Dumazet7fee2262010-05-11 23:19:48 +0000858/*
859 * skb might have a dst pointer attached, refcounted or not.
860 * _skb_refdst low order bit is set if refcount was _not_ taken
861 */
862#define SKB_DST_NOREF 1UL
863#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
864
Florian Westphala9e419d2017-01-23 18:21:59 +0100865#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000866/**
867 * skb_dst - returns skb dst_entry
868 * @skb: buffer
869 *
870 * Returns skb dst_entry, regardless of reference taken or not.
871 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000872static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
873{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000874 /* If refdst was not refcounted, check we still are in a
875 * rcu_read_lock section
876 */
877 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
878 !rcu_read_lock_held() &&
879 !rcu_read_lock_bh_held());
880 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000881}
882
Eric Dumazet7fee2262010-05-11 23:19:48 +0000883/**
884 * skb_dst_set - sets skb dst
885 * @skb: buffer
886 * @dst: dst entry
887 *
888 * Sets skb dst, assuming a reference was taken on dst and should
889 * be released by skb_dst_drop()
890 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000891static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
892{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000893 skb->_skb_refdst = (unsigned long)dst;
894}
895
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200896/**
897 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
898 * @skb: buffer
899 * @dst: dst entry
900 *
901 * Sets skb dst, assuming a reference was not taken on dst.
902 * If dst entry is cached, we do not take reference and dst_release
903 * will be avoided by refdst_drop. If dst entry is not cached, we take
904 * reference, so that last dst_release can destroy the dst immediately.
905 */
906static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
907{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100908 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
909 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200910}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000911
912/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300913 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000914 * @skb: buffer
915 */
916static inline bool skb_dst_is_noref(const struct sk_buff *skb)
917{
918 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000919}
920
Eric Dumazet511c3f92009-06-02 05:14:27 +0000921static inline struct rtable *skb_rtable(const struct sk_buff *skb)
922{
Eric Dumazetadf30902009-06-02 05:19:30 +0000923 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000924}
925
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400926/* For mangling skb->pkt_type from user space side from applications
927 * such as nft, tc, etc, we only allow a conservative subset of
928 * possible pkt_types to be set.
929*/
930static inline bool skb_pkt_type_ok(u32 ptype)
931{
932 return ptype <= PACKET_OTHERHOST;
933}
934
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200935static inline unsigned int skb_napi_id(const struct sk_buff *skb)
936{
937#ifdef CONFIG_NET_RX_BUSY_POLL
938 return skb->napi_id;
939#else
940 return 0;
941#endif
942}
943
Paolo Abeni3889a8032017-06-12 11:23:41 +0200944/* decrement the reference count and return true if we can free the skb */
945static inline bool skb_unref(struct sk_buff *skb)
946{
947 if (unlikely(!skb))
948 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +0300949 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200950 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +0300951 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200952 return false;
953
954 return true;
955}
956
Paolo Abeni0a463c72017-06-12 11:23:42 +0200957void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700958void kfree_skb(struct sk_buff *skb);
959void kfree_skb_list(struct sk_buff *segs);
960void skb_tx_error(struct sk_buff *skb);
961void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +0200962void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700963void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000964extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000965
Joe Perches7965bd42013-09-26 14:48:15 -0700966void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
967bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
968 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000969
Joe Perches7965bd42013-09-26 14:48:15 -0700970struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
971 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700972struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700973struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700974static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100975 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700976{
Eric Dumazet564824b2010-10-11 19:05:25 +0000977 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700978}
979
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700980struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
981 unsigned long data_len,
982 int max_page_order,
983 int *errcode,
984 gfp_t gfp_mask);
985
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700986/* Layout of fast clones : [skb1][skb2][fclone_ref] */
987struct sk_buff_fclones {
988 struct sk_buff skb1;
989
990 struct sk_buff skb2;
991
Reshetova, Elena26385952017-06-30 13:07:59 +0300992 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700993};
994
995/**
996 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700997 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700998 * @skb: buffer
999 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001000 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001001 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1002 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001003 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001004static inline bool skb_fclone_busy(const struct sock *sk,
1005 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001006{
1007 const struct sk_buff_fclones *fclones;
1008
1009 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1010
1011 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001012 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001013 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001014}
1015
David S. Millerd179cd12005-08-17 14:57:30 -07001016static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001017 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001018{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001019 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001020}
1021
Joe Perches7965bd42013-09-26 14:48:15 -07001022struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1023int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1024struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1025struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001026struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1027 gfp_t gfp_mask, bool fclone);
1028static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1029 gfp_t gfp_mask)
1030{
1031 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1032}
Eric Dumazet117632e2011-12-03 21:39:53 +00001033
Joe Perches7965bd42013-09-26 14:48:15 -07001034int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1035struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1036 unsigned int headroom);
1037struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1038 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001039int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1040 int offset, int len);
1041int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1042 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001043int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001044int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1045
1046/**
1047 * skb_pad - zero pad the tail of an skb
1048 * @skb: buffer to pad
1049 * @pad: space to pad
1050 *
1051 * Ensure that a buffer is followed by a padding area that is zero
1052 * filled. Used by network drivers which may DMA or transfer data
1053 * beyond the buffer end onto the wire.
1054 *
1055 * May return error in out of memory cases. The skb is freed on error.
1056 */
1057static inline int skb_pad(struct sk_buff *skb, int pad)
1058{
1059 return __skb_pad(skb, pad, true);
1060}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001061#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
Joe Perches7965bd42013-09-26 14:48:15 -07001063int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1064 int getfrag(void *from, char *to, int offset,
1065 int len, int odd, struct sk_buff *skb),
1066 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001067
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001068int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1069 int offset, size_t size);
1070
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001071struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001072 __u32 lower_offset;
1073 __u32 upper_offset;
1074 __u32 frag_idx;
1075 __u32 stepped_offset;
1076 struct sk_buff *root_skb;
1077 struct sk_buff *cur_skb;
1078 __u8 *frag_data;
1079};
1080
Joe Perches7965bd42013-09-26 14:48:15 -07001081void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1082 unsigned int to, struct skb_seq_state *st);
1083unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1084 struct skb_seq_state *st);
1085void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001086
Joe Perches7965bd42013-09-26 14:48:15 -07001087unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001088 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001089
Tom Herbert09323cc2013-12-15 22:16:19 -08001090/*
1091 * Packet hash types specify the type of hash in skb_set_hash.
1092 *
1093 * Hash types refer to the protocol layer addresses which are used to
1094 * construct a packet's hash. The hashes are used to differentiate or identify
1095 * flows of the protocol layer for the hash type. Hash types are either
1096 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1097 *
1098 * Properties of hashes:
1099 *
1100 * 1) Two packets in different flows have different hash values
1101 * 2) Two packets in the same flow should have the same hash value
1102 *
1103 * A hash at a higher layer is considered to be more specific. A driver should
1104 * set the most specific hash possible.
1105 *
1106 * A driver cannot indicate a more specific hash than the layer at which a hash
1107 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1108 *
1109 * A driver may indicate a hash level which is less specific than the
1110 * actual layer the hash was computed on. For instance, a hash computed
1111 * at L4 may be considered an L3 hash. This should only be done if the
1112 * driver can't unambiguously determine that the HW computed the hash at
1113 * the higher layer. Note that the "should" in the second property above
1114 * permits this.
1115 */
1116enum pkt_hash_types {
1117 PKT_HASH_TYPE_NONE, /* Undefined type */
1118 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1119 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1120 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1121};
1122
Tom Herbertbcc83832015-09-01 09:24:24 -07001123static inline void skb_clear_hash(struct sk_buff *skb)
1124{
1125 skb->hash = 0;
1126 skb->sw_hash = 0;
1127 skb->l4_hash = 0;
1128}
1129
1130static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1131{
1132 if (!skb->l4_hash)
1133 skb_clear_hash(skb);
1134}
1135
1136static inline void
1137__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1138{
1139 skb->l4_hash = is_l4;
1140 skb->sw_hash = is_sw;
1141 skb->hash = hash;
1142}
1143
Tom Herbert09323cc2013-12-15 22:16:19 -08001144static inline void
1145skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1146{
Tom Herbertbcc83832015-09-01 09:24:24 -07001147 /* Used by drivers to set hash from HW */
1148 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1149}
1150
1151static inline void
1152__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1153{
1154 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001155}
1156
Tom Herberte5276932015-09-01 09:24:23 -07001157void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001158u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001159u32 skb_get_poff(const struct sk_buff *skb);
1160u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1161 const struct flow_keys *keys, int hlen);
1162__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1163 void *data, int hlen_proto);
1164
1165static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1166 int thoff, u8 ip_proto)
1167{
1168 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1169}
1170
1171void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1172 const struct flow_dissector_key *key,
1173 unsigned int key_count);
1174
1175bool __skb_flow_dissect(const struct sk_buff *skb,
1176 struct flow_dissector *flow_dissector,
1177 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001178 void *data, __be16 proto, int nhoff, int hlen,
1179 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001180
1181static inline bool skb_flow_dissect(const struct sk_buff *skb,
1182 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001183 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001184{
1185 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001186 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001187}
1188
1189static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001190 struct flow_keys *flow,
1191 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001192{
1193 memset(flow, 0, sizeof(*flow));
1194 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001195 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001196}
1197
1198static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1199 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001200 int nhoff, int hlen,
1201 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001202{
1203 memset(flow, 0, sizeof(*flow));
1204 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001205 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001206}
1207
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001208static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001209{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001210 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001211 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001212
Tom Herbert61b905d2014-03-24 15:34:47 -07001213 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001214}
1215
David S. Miller20a17bf2015-09-01 21:19:17 -07001216static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001217{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001218 if (!skb->l4_hash && !skb->sw_hash) {
1219 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001220 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001221
Tom Herbertde4c1f82015-09-01 18:11:04 -07001222 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001223 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001224
1225 return skb->hash;
1226}
1227
Tom Herbert50fb7992015-05-01 11:30:12 -07001228__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1229
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001230static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1231{
Tom Herbert61b905d2014-03-24 15:34:47 -07001232 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001233}
1234
Tom Herbert3df7a742013-12-15 22:16:29 -08001235static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1236{
Tom Herbert61b905d2014-03-24 15:34:47 -07001237 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001238 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001239 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001240};
1241
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001242#ifdef NET_SKBUFF_DATA_USES_OFFSET
1243static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1244{
1245 return skb->head + skb->end;
1246}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001247
1248static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1249{
1250 return skb->end;
1251}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001252#else
1253static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1254{
1255 return skb->end;
1256}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001257
1258static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1259{
1260 return skb->end - skb->head;
1261}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001262#endif
1263
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001265#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Patrick Ohlyac45f602009-02-12 05:03:37 +00001267static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1268{
1269 return &skb_shinfo(skb)->hwtstamps;
1270}
1271
Willem de Bruijn52267792017-08-03 16:29:39 -04001272static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1273{
1274 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1275
1276 return is_zcopy ? skb_uarg(skb) : NULL;
1277}
1278
1279static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1280{
1281 if (skb && uarg && !skb_zcopy(skb)) {
1282 sock_zerocopy_get(uarg);
1283 skb_shinfo(skb)->destructor_arg = uarg;
1284 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1285 }
1286}
1287
1288/* Release a reference on a zerocopy structure */
1289static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1290{
1291 struct ubuf_info *uarg = skb_zcopy(skb);
1292
1293 if (uarg) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001294 if (uarg->callback == sock_zerocopy_callback) {
1295 uarg->zerocopy = uarg->zerocopy && zerocopy;
1296 sock_zerocopy_put(uarg);
1297 } else {
1298 uarg->callback(uarg, zerocopy);
1299 }
1300
Willem de Bruijn52267792017-08-03 16:29:39 -04001301 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1302 }
1303}
1304
1305/* Abort a zerocopy operation and revert zckey on error in send syscall */
1306static inline void skb_zcopy_abort(struct sk_buff *skb)
1307{
1308 struct ubuf_info *uarg = skb_zcopy(skb);
1309
1310 if (uarg) {
1311 sock_zerocopy_put_abort(uarg);
1312 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1313 }
1314}
1315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316/**
1317 * skb_queue_empty - check if a queue is empty
1318 * @list: queue head
1319 *
1320 * Returns true if the queue is empty, false otherwise.
1321 */
1322static inline int skb_queue_empty(const struct sk_buff_head *list)
1323{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001324 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325}
1326
1327/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001328 * skb_queue_is_last - check if skb is the last entry in the queue
1329 * @list: queue head
1330 * @skb: buffer
1331 *
1332 * Returns true if @skb is the last buffer on the list.
1333 */
1334static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1335 const struct sk_buff *skb)
1336{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001337 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001338}
1339
1340/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001341 * skb_queue_is_first - check if skb is the first entry in the queue
1342 * @list: queue head
1343 * @skb: buffer
1344 *
1345 * Returns true if @skb is the first buffer on the list.
1346 */
1347static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1348 const struct sk_buff *skb)
1349{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001350 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001351}
1352
1353/**
David S. Miller249c8b42008-09-23 00:44:42 -07001354 * skb_queue_next - return the next packet in the queue
1355 * @list: queue head
1356 * @skb: current buffer
1357 *
1358 * Return the next packet in @list after @skb. It is only valid to
1359 * call this if skb_queue_is_last() evaluates to false.
1360 */
1361static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1362 const struct sk_buff *skb)
1363{
1364 /* This BUG_ON may seem severe, but if we just return then we
1365 * are going to dereference garbage.
1366 */
1367 BUG_ON(skb_queue_is_last(list, skb));
1368 return skb->next;
1369}
1370
1371/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001372 * skb_queue_prev - return the prev packet in the queue
1373 * @list: queue head
1374 * @skb: current buffer
1375 *
1376 * Return the prev packet in @list before @skb. It is only valid to
1377 * call this if skb_queue_is_first() evaluates to false.
1378 */
1379static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1380 const struct sk_buff *skb)
1381{
1382 /* This BUG_ON may seem severe, but if we just return then we
1383 * are going to dereference garbage.
1384 */
1385 BUG_ON(skb_queue_is_first(list, skb));
1386 return skb->prev;
1387}
1388
1389/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 * skb_get - reference buffer
1391 * @skb: buffer to reference
1392 *
1393 * Makes another reference to a socket buffer and returns a pointer
1394 * to the buffer.
1395 */
1396static inline struct sk_buff *skb_get(struct sk_buff *skb)
1397{
Reshetova, Elena63354792017-06-30 13:07:58 +03001398 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 return skb;
1400}
1401
1402/*
1403 * If users == 1, we are the only owner and are can avoid redundant
1404 * atomic change.
1405 */
1406
1407/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 * skb_cloned - is the buffer a clone
1409 * @skb: buffer to check
1410 *
1411 * Returns true if the buffer was generated with skb_clone() and is
1412 * one of multiple shared copies of the buffer. Cloned buffers are
1413 * shared data so must not be written to under normal circumstances.
1414 */
1415static inline int skb_cloned(const struct sk_buff *skb)
1416{
1417 return skb->cloned &&
1418 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1419}
1420
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001421static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1422{
Mel Gormand0164ad2015-11-06 16:28:21 -08001423 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001424
1425 if (skb_cloned(skb))
1426 return pskb_expand_head(skb, 0, 0, pri);
1427
1428 return 0;
1429}
1430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431/**
1432 * skb_header_cloned - is the header a clone
1433 * @skb: buffer to check
1434 *
1435 * Returns true if modifying the header part of the buffer requires
1436 * the data to be copied.
1437 */
1438static inline int skb_header_cloned(const struct sk_buff *skb)
1439{
1440 int dataref;
1441
1442 if (!skb->cloned)
1443 return 0;
1444
1445 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1446 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1447 return dataref != 1;
1448}
1449
Eric Dumazet9580bf22016-04-30 10:19:29 -07001450static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1451{
1452 might_sleep_if(gfpflags_allow_blocking(pri));
1453
1454 if (skb_header_cloned(skb))
1455 return pskb_expand_head(skb, 0, 0, pri);
1456
1457 return 0;
1458}
1459
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001461 * __skb_header_release - release reference to header
1462 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001463 */
1464static inline void __skb_header_release(struct sk_buff *skb)
1465{
1466 skb->nohdr = 1;
1467 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1468}
1469
1470
1471/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 * skb_shared - is the buffer shared
1473 * @skb: buffer to check
1474 *
1475 * Returns true if more than one person has a reference to this
1476 * buffer.
1477 */
1478static inline int skb_shared(const struct sk_buff *skb)
1479{
Reshetova, Elena63354792017-06-30 13:07:58 +03001480 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481}
1482
1483/**
1484 * skb_share_check - check if buffer is shared and if so clone it
1485 * @skb: buffer to check
1486 * @pri: priority for memory allocation
1487 *
1488 * If the buffer is shared the buffer is cloned and the old copy
1489 * drops a reference. A new clone with a single reference is returned.
1490 * If the buffer is not shared the original buffer is returned. When
1491 * being called from interrupt status or with spinlocks held pri must
1492 * be GFP_ATOMIC.
1493 *
1494 * NULL is returned on a memory allocation failure.
1495 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001496static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497{
Mel Gormand0164ad2015-11-06 16:28:21 -08001498 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 if (skb_shared(skb)) {
1500 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001501
1502 if (likely(nskb))
1503 consume_skb(skb);
1504 else
1505 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 skb = nskb;
1507 }
1508 return skb;
1509}
1510
1511/*
1512 * Copy shared buffers into a new sk_buff. We effectively do COW on
1513 * packets to handle cases where we have a local reader and forward
1514 * and a couple of other messy ones. The normal one is tcpdumping
1515 * a packet thats being forwarded.
1516 */
1517
1518/**
1519 * skb_unshare - make a copy of a shared buffer
1520 * @skb: buffer to check
1521 * @pri: priority for memory allocation
1522 *
1523 * If the socket buffer is a clone then this function creates a new
1524 * copy of the data, drops a reference count on the old copy and returns
1525 * the new copy with the reference count at 1. If the buffer is not a clone
1526 * the original buffer is returned. When called with a spinlock held or
1527 * from interrupt state @pri must be %GFP_ATOMIC
1528 *
1529 * %NULL is returned on a memory allocation failure.
1530 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001531static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001532 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533{
Mel Gormand0164ad2015-11-06 16:28:21 -08001534 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 if (skb_cloned(skb)) {
1536 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001537
1538 /* Free our shared copy */
1539 if (likely(nskb))
1540 consume_skb(skb);
1541 else
1542 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 skb = nskb;
1544 }
1545 return skb;
1546}
1547
1548/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001549 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 * @list_: list to peek at
1551 *
1552 * Peek an &sk_buff. Unlike most other operations you _MUST_
1553 * be careful with this one. A peek leaves the buffer on the
1554 * list and someone else may run off with it. You must hold
1555 * the appropriate locks or have a private queue to do this.
1556 *
1557 * Returns %NULL for an empty list or a pointer to the head element.
1558 * The reference count is not incremented and the reference is therefore
1559 * volatile. Use with caution.
1560 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001561static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562{
Eric Dumazet18d07002012-04-30 16:31:46 +00001563 struct sk_buff *skb = list_->next;
1564
1565 if (skb == (struct sk_buff *)list_)
1566 skb = NULL;
1567 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568}
1569
1570/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001571 * skb_peek_next - peek skb following the given one from a queue
1572 * @skb: skb to start from
1573 * @list_: list to peek at
1574 *
1575 * Returns %NULL when the end of the list is met or a pointer to the
1576 * next element. The reference count is not incremented and the
1577 * reference is therefore volatile. Use with caution.
1578 */
1579static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1580 const struct sk_buff_head *list_)
1581{
1582 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001583
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001584 if (next == (struct sk_buff *)list_)
1585 next = NULL;
1586 return next;
1587}
1588
1589/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001590 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 * @list_: list to peek at
1592 *
1593 * Peek an &sk_buff. Unlike most other operations you _MUST_
1594 * be careful with this one. A peek leaves the buffer on the
1595 * list and someone else may run off with it. You must hold
1596 * the appropriate locks or have a private queue to do this.
1597 *
1598 * Returns %NULL for an empty list or a pointer to the tail element.
1599 * The reference count is not incremented and the reference is therefore
1600 * volatile. Use with caution.
1601 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001602static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603{
Eric Dumazet18d07002012-04-30 16:31:46 +00001604 struct sk_buff *skb = list_->prev;
1605
1606 if (skb == (struct sk_buff *)list_)
1607 skb = NULL;
1608 return skb;
1609
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610}
1611
1612/**
1613 * skb_queue_len - get queue length
1614 * @list_: list to measure
1615 *
1616 * Return the length of an &sk_buff queue.
1617 */
1618static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1619{
1620 return list_->qlen;
1621}
1622
David S. Miller67fed452008-09-21 22:36:24 -07001623/**
1624 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1625 * @list: queue to initialize
1626 *
1627 * This initializes only the list and queue length aspects of
1628 * an sk_buff_head object. This allows to initialize the list
1629 * aspects of an sk_buff_head without reinitializing things like
1630 * the spinlock. It can also be used for on-stack sk_buff_head
1631 * objects where the spinlock is known to not be used.
1632 */
1633static inline void __skb_queue_head_init(struct sk_buff_head *list)
1634{
1635 list->prev = list->next = (struct sk_buff *)list;
1636 list->qlen = 0;
1637}
1638
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001639/*
1640 * This function creates a split out lock class for each invocation;
1641 * this is needed for now since a whole lot of users of the skb-queue
1642 * infrastructure in drivers have different locking usage (in hardirq)
1643 * than the networking core (in softirq only). In the long run either the
1644 * network layer or drivers should need annotation to consolidate the
1645 * main types of usage into 3 classes.
1646 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647static inline void skb_queue_head_init(struct sk_buff_head *list)
1648{
1649 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001650 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651}
1652
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001653static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1654 struct lock_class_key *class)
1655{
1656 skb_queue_head_init(list);
1657 lockdep_set_class(&list->lock, class);
1658}
1659
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001661 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 *
1663 * The "__skb_xxxx()" functions are the non-atomic ones that
1664 * can only be called with interrupts disabled.
1665 */
Joe Perches7965bd42013-09-26 14:48:15 -07001666void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1667 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001668static inline void __skb_insert(struct sk_buff *newsk,
1669 struct sk_buff *prev, struct sk_buff *next,
1670 struct sk_buff_head *list)
1671{
1672 newsk->next = next;
1673 newsk->prev = prev;
1674 next->prev = prev->next = newsk;
1675 list->qlen++;
1676}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
David S. Miller67fed452008-09-21 22:36:24 -07001678static inline void __skb_queue_splice(const struct sk_buff_head *list,
1679 struct sk_buff *prev,
1680 struct sk_buff *next)
1681{
1682 struct sk_buff *first = list->next;
1683 struct sk_buff *last = list->prev;
1684
1685 first->prev = prev;
1686 prev->next = first;
1687
1688 last->next = next;
1689 next->prev = last;
1690}
1691
1692/**
1693 * skb_queue_splice - join two skb lists, this is designed for stacks
1694 * @list: the new list to add
1695 * @head: the place to add it in the first list
1696 */
1697static inline void skb_queue_splice(const struct sk_buff_head *list,
1698 struct sk_buff_head *head)
1699{
1700 if (!skb_queue_empty(list)) {
1701 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001702 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001703 }
1704}
1705
1706/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001707 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001708 * @list: the new list to add
1709 * @head: the place to add it in the first list
1710 *
1711 * The list at @list is reinitialised
1712 */
1713static inline void skb_queue_splice_init(struct sk_buff_head *list,
1714 struct sk_buff_head *head)
1715{
1716 if (!skb_queue_empty(list)) {
1717 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001718 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001719 __skb_queue_head_init(list);
1720 }
1721}
1722
1723/**
1724 * skb_queue_splice_tail - join two skb lists, each list being a queue
1725 * @list: the new list to add
1726 * @head: the place to add it in the first list
1727 */
1728static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1729 struct sk_buff_head *head)
1730{
1731 if (!skb_queue_empty(list)) {
1732 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001733 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001734 }
1735}
1736
1737/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001738 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001739 * @list: the new list to add
1740 * @head: the place to add it in the first list
1741 *
1742 * Each of the lists is a queue.
1743 * The list at @list is reinitialised
1744 */
1745static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1746 struct sk_buff_head *head)
1747{
1748 if (!skb_queue_empty(list)) {
1749 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001750 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001751 __skb_queue_head_init(list);
1752 }
1753}
1754
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001756 * __skb_queue_after - queue a buffer at the list head
1757 * @list: list to use
1758 * @prev: place after this buffer
1759 * @newsk: buffer to queue
1760 *
1761 * Queue a buffer int the middle of a list. This function takes no locks
1762 * and you must therefore hold required locks before calling it.
1763 *
1764 * A buffer cannot be placed on two lists at the same time.
1765 */
1766static inline void __skb_queue_after(struct sk_buff_head *list,
1767 struct sk_buff *prev,
1768 struct sk_buff *newsk)
1769{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001770 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001771}
1772
Joe Perches7965bd42013-09-26 14:48:15 -07001773void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1774 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001775
Gerrit Renkerf5572852008-04-14 00:05:28 -07001776static inline void __skb_queue_before(struct sk_buff_head *list,
1777 struct sk_buff *next,
1778 struct sk_buff *newsk)
1779{
1780 __skb_insert(newsk, next->prev, next, list);
1781}
1782
Stephen Hemminger300ce172005-10-30 13:47:34 -08001783/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 * __skb_queue_head - queue a buffer at the list head
1785 * @list: list to use
1786 * @newsk: buffer to queue
1787 *
1788 * Queue a buffer at the start of a list. This function takes no locks
1789 * and you must therefore hold required locks before calling it.
1790 *
1791 * A buffer cannot be placed on two lists at the same time.
1792 */
Joe Perches7965bd42013-09-26 14:48:15 -07001793void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794static inline void __skb_queue_head(struct sk_buff_head *list,
1795 struct sk_buff *newsk)
1796{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001797 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798}
1799
1800/**
1801 * __skb_queue_tail - queue a buffer at the list tail
1802 * @list: list to use
1803 * @newsk: buffer to queue
1804 *
1805 * Queue a buffer at the end of a list. This function takes no locks
1806 * and you must therefore hold required locks before calling it.
1807 *
1808 * A buffer cannot be placed on two lists at the same time.
1809 */
Joe Perches7965bd42013-09-26 14:48:15 -07001810void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811static inline void __skb_queue_tail(struct sk_buff_head *list,
1812 struct sk_buff *newsk)
1813{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001814 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815}
1816
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 * remove sk_buff from list. _Must_ be called atomically, and with
1819 * the list known..
1820 */
Joe Perches7965bd42013-09-26 14:48:15 -07001821void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1823{
1824 struct sk_buff *next, *prev;
1825
1826 list->qlen--;
1827 next = skb->next;
1828 prev = skb->prev;
1829 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 next->prev = prev;
1831 prev->next = next;
1832}
1833
Gerrit Renkerf525c062008-04-14 00:04:12 -07001834/**
1835 * __skb_dequeue - remove from the head of the queue
1836 * @list: list to dequeue from
1837 *
1838 * Remove the head of the list. This function does not take any locks
1839 * so must be used with appropriate locks held only. The head item is
1840 * returned or %NULL if the list is empty.
1841 */
Joe Perches7965bd42013-09-26 14:48:15 -07001842struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001843static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1844{
1845 struct sk_buff *skb = skb_peek(list);
1846 if (skb)
1847 __skb_unlink(skb, list);
1848 return skb;
1849}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
1851/**
1852 * __skb_dequeue_tail - remove from the tail of the queue
1853 * @list: list to dequeue from
1854 *
1855 * Remove the tail of the list. This function does not take any locks
1856 * so must be used with appropriate locks held only. The tail item is
1857 * returned or %NULL if the list is empty.
1858 */
Joe Perches7965bd42013-09-26 14:48:15 -07001859struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1861{
1862 struct sk_buff *skb = skb_peek_tail(list);
1863 if (skb)
1864 __skb_unlink(skb, list);
1865 return skb;
1866}
1867
1868
David S. Millerbdcc0922012-03-07 20:53:36 -05001869static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870{
1871 return skb->data_len;
1872}
1873
1874static inline unsigned int skb_headlen(const struct sk_buff *skb)
1875{
1876 return skb->len - skb->data_len;
1877}
1878
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001879static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001881 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001883 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001884 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001885 return len;
1886}
1887
1888static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1889{
1890 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891}
1892
Ian Campbell131ea662011-08-19 06:25:00 +00001893/**
1894 * __skb_fill_page_desc - initialise a paged fragment in an skb
1895 * @skb: buffer containing fragment to be initialised
1896 * @i: paged fragment index to initialise
1897 * @page: the page to use for this fragment
1898 * @off: the offset to the data with @page
1899 * @size: the length of the data
1900 *
1901 * Initialises the @i'th fragment of @skb to point to &size bytes at
1902 * offset @off within @page.
1903 *
1904 * Does not take any additional reference on the fragment.
1905 */
1906static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1907 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908{
1909 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1910
Mel Gormanc48a11c2012-07-31 16:44:23 -07001911 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001912 * Propagate page pfmemalloc to the skb if we can. The problem is
1913 * that not all callers have unique ownership of the page but rely
1914 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001915 */
Ian Campbella8605c62011-10-19 23:01:49 +00001916 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001918 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001919
1920 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001921 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001922 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001923}
1924
1925/**
1926 * skb_fill_page_desc - initialise a paged fragment in an skb
1927 * @skb: buffer containing fragment to be initialised
1928 * @i: paged fragment index to initialise
1929 * @page: the page to use for this fragment
1930 * @off: the offset to the data with @page
1931 * @size: the length of the data
1932 *
1933 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001934 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001935 * addition updates @skb such that @i is the last fragment.
1936 *
1937 * Does not take any additional reference on the fragment.
1938 */
1939static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1940 struct page *page, int off, int size)
1941{
1942 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 skb_shinfo(skb)->nr_frags = i + 1;
1944}
1945
Joe Perches7965bd42013-09-26 14:48:15 -07001946void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1947 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001948
Jason Wangf8e617e2013-11-01 14:07:47 +08001949void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1950 unsigned int truesize);
1951
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001953#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1955
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001956#ifdef NET_SKBUFF_DATA_USES_OFFSET
1957static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1958{
1959 return skb->head + skb->tail;
1960}
1961
1962static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1963{
1964 skb->tail = skb->data - skb->head;
1965}
1966
1967static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1968{
1969 skb_reset_tail_pointer(skb);
1970 skb->tail += offset;
1971}
Simon Horman7cc46192013-05-28 20:34:29 +00001972
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001973#else /* NET_SKBUFF_DATA_USES_OFFSET */
1974static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1975{
1976 return skb->tail;
1977}
1978
1979static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1980{
1981 skb->tail = skb->data;
1982}
1983
1984static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1985{
1986 skb->tail = skb->data + offset;
1987}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001988
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001989#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1990
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991/*
1992 * Add data to an sk_buff
1993 */
Johannes Berg4df864c2017-06-16 14:29:21 +02001994void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1995void *skb_put(struct sk_buff *skb, unsigned int len);
1996static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997{
Johannes Berg4df864c2017-06-16 14:29:21 +02001998 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 SKB_LINEAR_ASSERT(skb);
2000 skb->tail += len;
2001 skb->len += len;
2002 return tmp;
2003}
2004
yuan linyude77b962017-06-18 22:48:17 +08002005static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2006{
2007 void *tmp = __skb_put(skb, len);
2008
2009 memset(tmp, 0, len);
2010 return tmp;
2011}
2012
2013static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2014 unsigned int len)
2015{
2016 void *tmp = __skb_put(skb, len);
2017
2018 memcpy(tmp, data, len);
2019 return tmp;
2020}
2021
2022static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2023{
2024 *(u8 *)__skb_put(skb, 1) = val;
2025}
2026
Johannes Berg83ad3572017-06-14 22:17:20 +02002027static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002028{
Johannes Berg83ad3572017-06-14 22:17:20 +02002029 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002030
2031 memset(tmp, 0, len);
2032
2033 return tmp;
2034}
2035
Johannes Berg59ae1d12017-06-16 14:29:20 +02002036static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2037 unsigned int len)
2038{
2039 void *tmp = skb_put(skb, len);
2040
2041 memcpy(tmp, data, len);
2042
2043 return tmp;
2044}
2045
Johannes Berg634fef62017-06-16 14:29:24 +02002046static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2047{
2048 *(u8 *)skb_put(skb, 1) = val;
2049}
2050
Johannes Bergd58ff352017-06-16 14:29:23 +02002051void *skb_push(struct sk_buff *skb, unsigned int len);
2052static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053{
2054 skb->data -= len;
2055 skb->len += len;
2056 return skb->data;
2057}
2058
Johannes Bergaf728682017-06-16 14:29:22 +02002059void *skb_pull(struct sk_buff *skb, unsigned int len);
2060static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061{
2062 skb->len -= len;
2063 BUG_ON(skb->len < skb->data_len);
2064 return skb->data += len;
2065}
2066
Johannes Bergaf728682017-06-16 14:29:22 +02002067static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002068{
2069 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2070}
2071
Johannes Bergaf728682017-06-16 14:29:22 +02002072void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
Johannes Bergaf728682017-06-16 14:29:22 +02002074static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075{
2076 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002077 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 return NULL;
2079 skb->len -= len;
2080 return skb->data += len;
2081}
2082
Johannes Bergaf728682017-06-16 14:29:22 +02002083static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084{
2085 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2086}
2087
2088static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2089{
2090 if (likely(len <= skb_headlen(skb)))
2091 return 1;
2092 if (unlikely(len > skb->len))
2093 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002094 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095}
2096
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002097void skb_condense(struct sk_buff *skb);
2098
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099/**
2100 * skb_headroom - bytes at buffer head
2101 * @skb: buffer to check
2102 *
2103 * Return the number of bytes of free space at the head of an &sk_buff.
2104 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002105static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106{
2107 return skb->data - skb->head;
2108}
2109
2110/**
2111 * skb_tailroom - bytes at buffer end
2112 * @skb: buffer to check
2113 *
2114 * Return the number of bytes of free space at the tail of an sk_buff
2115 */
2116static inline int skb_tailroom(const struct sk_buff *skb)
2117{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002118 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119}
2120
2121/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002122 * skb_availroom - bytes at buffer end
2123 * @skb: buffer to check
2124 *
2125 * Return the number of bytes of free space at the tail of an sk_buff
2126 * allocated by sk_stream_alloc()
2127 */
2128static inline int skb_availroom(const struct sk_buff *skb)
2129{
Eric Dumazet16fad692013-03-14 05:40:32 +00002130 if (skb_is_nonlinear(skb))
2131 return 0;
2132
2133 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002134}
2135
2136/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 * skb_reserve - adjust headroom
2138 * @skb: buffer to alter
2139 * @len: bytes to move
2140 *
2141 * Increase the headroom of an empty &sk_buff by reducing the tail
2142 * room. This is only allowed for an empty buffer.
2143 */
David S. Miller8243126c2006-01-17 02:54:21 -08002144static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145{
2146 skb->data += len;
2147 skb->tail += len;
2148}
2149
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002150/**
2151 * skb_tailroom_reserve - adjust reserved_tailroom
2152 * @skb: buffer to alter
2153 * @mtu: maximum amount of headlen permitted
2154 * @needed_tailroom: minimum amount of reserved_tailroom
2155 *
2156 * Set reserved_tailroom so that headlen can be as large as possible but
2157 * not larger than mtu and tailroom cannot be smaller than
2158 * needed_tailroom.
2159 * The required headroom should already have been reserved before using
2160 * this function.
2161 */
2162static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2163 unsigned int needed_tailroom)
2164{
2165 SKB_LINEAR_ASSERT(skb);
2166 if (mtu < skb_tailroom(skb) - needed_tailroom)
2167 /* use at most mtu */
2168 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2169 else
2170 /* use up to all available space */
2171 skb->reserved_tailroom = needed_tailroom;
2172}
2173
Tom Herbert8bce6d72014-09-29 20:22:29 -07002174#define ENCAP_TYPE_ETHER 0
2175#define ENCAP_TYPE_IPPROTO 1
2176
2177static inline void skb_set_inner_protocol(struct sk_buff *skb,
2178 __be16 protocol)
2179{
2180 skb->inner_protocol = protocol;
2181 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2182}
2183
2184static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2185 __u8 ipproto)
2186{
2187 skb->inner_ipproto = ipproto;
2188 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2189}
2190
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002191static inline void skb_reset_inner_headers(struct sk_buff *skb)
2192{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002193 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002194 skb->inner_network_header = skb->network_header;
2195 skb->inner_transport_header = skb->transport_header;
2196}
2197
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002198static inline void skb_reset_mac_len(struct sk_buff *skb)
2199{
2200 skb->mac_len = skb->network_header - skb->mac_header;
2201}
2202
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002203static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2204 *skb)
2205{
2206 return skb->head + skb->inner_transport_header;
2207}
2208
Tom Herbert55dc5a92015-12-14 11:19:40 -08002209static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2210{
2211 return skb_inner_transport_header(skb) - skb->data;
2212}
2213
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002214static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2215{
2216 skb->inner_transport_header = skb->data - skb->head;
2217}
2218
2219static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2220 const int offset)
2221{
2222 skb_reset_inner_transport_header(skb);
2223 skb->inner_transport_header += offset;
2224}
2225
2226static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2227{
2228 return skb->head + skb->inner_network_header;
2229}
2230
2231static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2232{
2233 skb->inner_network_header = skb->data - skb->head;
2234}
2235
2236static inline void skb_set_inner_network_header(struct sk_buff *skb,
2237 const int offset)
2238{
2239 skb_reset_inner_network_header(skb);
2240 skb->inner_network_header += offset;
2241}
2242
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002243static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2244{
2245 return skb->head + skb->inner_mac_header;
2246}
2247
2248static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2249{
2250 skb->inner_mac_header = skb->data - skb->head;
2251}
2252
2253static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2254 const int offset)
2255{
2256 skb_reset_inner_mac_header(skb);
2257 skb->inner_mac_header += offset;
2258}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002259static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2260{
Cong Wang35d04612013-05-29 15:16:05 +08002261 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002262}
2263
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002264static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2265{
2266 return skb->head + skb->transport_header;
2267}
2268
2269static inline void skb_reset_transport_header(struct sk_buff *skb)
2270{
2271 skb->transport_header = skb->data - skb->head;
2272}
2273
2274static inline void skb_set_transport_header(struct sk_buff *skb,
2275 const int offset)
2276{
2277 skb_reset_transport_header(skb);
2278 skb->transport_header += offset;
2279}
2280
2281static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2282{
2283 return skb->head + skb->network_header;
2284}
2285
2286static inline void skb_reset_network_header(struct sk_buff *skb)
2287{
2288 skb->network_header = skb->data - skb->head;
2289}
2290
2291static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2292{
2293 skb_reset_network_header(skb);
2294 skb->network_header += offset;
2295}
2296
2297static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2298{
2299 return skb->head + skb->mac_header;
2300}
2301
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002302static inline int skb_mac_offset(const struct sk_buff *skb)
2303{
2304 return skb_mac_header(skb) - skb->data;
2305}
2306
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002307static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2308{
2309 return skb->network_header - skb->mac_header;
2310}
2311
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002312static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2313{
Cong Wang35d04612013-05-29 15:16:05 +08002314 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002315}
2316
2317static inline void skb_reset_mac_header(struct sk_buff *skb)
2318{
2319 skb->mac_header = skb->data - skb->head;
2320}
2321
2322static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2323{
2324 skb_reset_mac_header(skb);
2325 skb->mac_header += offset;
2326}
2327
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002328static inline void skb_pop_mac_header(struct sk_buff *skb)
2329{
2330 skb->mac_header = skb->network_header;
2331}
2332
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002333static inline void skb_probe_transport_header(struct sk_buff *skb,
2334 const int offset_hint)
2335{
2336 struct flow_keys keys;
2337
2338 if (skb_transport_header_was_set(skb))
2339 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002340 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002341 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002342 else
2343 skb_set_transport_header(skb, offset_hint);
2344}
2345
Eric Dumazet03606892012-02-23 10:55:02 +00002346static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2347{
2348 if (skb_mac_header_was_set(skb)) {
2349 const unsigned char *old_mac = skb_mac_header(skb);
2350
2351 skb_set_mac_header(skb, -skb->mac_len);
2352 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2353 }
2354}
2355
Michał Mirosław04fb4512010-12-14 15:24:08 +00002356static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2357{
2358 return skb->csum_start - skb_headroom(skb);
2359}
2360
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002361static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2362{
2363 return skb->head + skb->csum_start;
2364}
2365
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002366static inline int skb_transport_offset(const struct sk_buff *skb)
2367{
2368 return skb_transport_header(skb) - skb->data;
2369}
2370
2371static inline u32 skb_network_header_len(const struct sk_buff *skb)
2372{
2373 return skb->transport_header - skb->network_header;
2374}
2375
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002376static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2377{
2378 return skb->inner_transport_header - skb->inner_network_header;
2379}
2380
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002381static inline int skb_network_offset(const struct sk_buff *skb)
2382{
2383 return skb_network_header(skb) - skb->data;
2384}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002385
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002386static inline int skb_inner_network_offset(const struct sk_buff *skb)
2387{
2388 return skb_inner_network_header(skb) - skb->data;
2389}
2390
Changli Gaof9599ce2010-08-04 04:43:44 +00002391static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2392{
2393 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2394}
2395
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396/*
2397 * CPUs often take a performance hit when accessing unaligned memory
2398 * locations. The actual performance hit varies, it can be small if the
2399 * hardware handles it or large if we have to take an exception and fix it
2400 * in software.
2401 *
2402 * Since an ethernet header is 14 bytes network drivers often end up with
2403 * the IP header at an unaligned offset. The IP header can be aligned by
2404 * shifting the start of the packet by 2 bytes. Drivers should do this
2405 * with:
2406 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002407 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408 *
2409 * The downside to this alignment of the IP header is that the DMA is now
2410 * unaligned. On some architectures the cost of an unaligned DMA is high
2411 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002412 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002413 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2414 * to be overridden.
2415 */
2416#ifndef NET_IP_ALIGN
2417#define NET_IP_ALIGN 2
2418#endif
2419
Anton Blanchard025be812006-03-31 02:27:06 -08002420/*
2421 * The networking layer reserves some headroom in skb data (via
2422 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2423 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002424 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002425 *
2426 * Unfortunately this headroom changes the DMA alignment of the resulting
2427 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2428 * on some architectures. An architecture can override this value,
2429 * perhaps setting it to a cacheline in size (since that will maintain
2430 * cacheline alignment of the DMA). It must be a power of 2.
2431 *
David S. Millerd6301d32009-02-08 19:24:13 -08002432 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002433 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002434 *
2435 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2436 * to reduce average number of cache lines per packet.
2437 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002438 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002439 */
2440#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002441#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002442#endif
2443
Joe Perches7965bd42013-09-26 14:48:15 -07002444int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002446static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002448 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002449 WARN_ON(1);
2450 return;
2451 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002452 skb->len = len;
2453 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454}
2455
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002456static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2457{
2458 __skb_set_length(skb, len);
2459}
2460
Joe Perches7965bd42013-09-26 14:48:15 -07002461void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462
2463static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2464{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002465 if (skb->data_len)
2466 return ___pskb_trim(skb, len);
2467 __skb_trim(skb, len);
2468 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469}
2470
2471static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2472{
2473 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2474}
2475
2476/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002477 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2478 * @skb: buffer to alter
2479 * @len: new length
2480 *
2481 * This is identical to pskb_trim except that the caller knows that
2482 * the skb is not cloned so we should never get an error due to out-
2483 * of-memory.
2484 */
2485static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2486{
2487 int err = pskb_trim(skb, len);
2488 BUG_ON(err);
2489}
2490
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002491static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2492{
2493 unsigned int diff = len - skb->len;
2494
2495 if (skb_tailroom(skb) < diff) {
2496 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2497 GFP_ATOMIC);
2498 if (ret)
2499 return ret;
2500 }
2501 __skb_set_length(skb, len);
2502 return 0;
2503}
2504
Herbert Xue9fa4f72006-08-13 20:12:58 -07002505/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002506 * skb_orphan - orphan a buffer
2507 * @skb: buffer to orphan
2508 *
2509 * If a buffer currently has an owner then we call the owner's
2510 * destructor function and make the @skb unowned. The buffer continues
2511 * to exist but is no longer charged to its former owner.
2512 */
2513static inline void skb_orphan(struct sk_buff *skb)
2514{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002515 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002517 skb->destructor = NULL;
2518 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002519 } else {
2520 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002521 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522}
2523
2524/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002525 * skb_orphan_frags - orphan the frags contained in a buffer
2526 * @skb: buffer to orphan frags from
2527 * @gfp_mask: allocation mask for replacement pages
2528 *
2529 * For each frag in the SKB which needs a destructor (i.e. has an
2530 * owner) create a copy of that frag and release the original
2531 * page by calling the destructor.
2532 */
2533static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2534{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002535 if (likely(!skb_zcopy(skb)))
2536 return 0;
2537 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2538 return 0;
2539 return skb_copy_ubufs(skb, gfp_mask);
2540}
2541
2542/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2543static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2544{
2545 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002546 return 0;
2547 return skb_copy_ubufs(skb, gfp_mask);
2548}
2549
2550/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551 * __skb_queue_purge - empty a list
2552 * @list: list to empty
2553 *
2554 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2555 * the list and one reference dropped. This function does not take the
2556 * list lock and the caller must hold the relevant locks to use it.
2557 */
Joe Perches7965bd42013-09-26 14:48:15 -07002558void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559static inline void __skb_queue_purge(struct sk_buff_head *list)
2560{
2561 struct sk_buff *skb;
2562 while ((skb = __skb_dequeue(list)) != NULL)
2563 kfree_skb(skb);
2564}
2565
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002566void skb_rbtree_purge(struct rb_root *root);
2567
Joe Perches7965bd42013-09-26 14:48:15 -07002568void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002569
Joe Perches7965bd42013-09-26 14:48:15 -07002570struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2571 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002572
2573/**
2574 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2575 * @dev: network device to receive on
2576 * @length: length to allocate
2577 *
2578 * Allocate a new &sk_buff and assign it a usage count of one. The
2579 * buffer has unspecified headroom built in. Users should allocate
2580 * the headroom they think they need without accounting for the
2581 * built in space. The built in space is used for optimisations.
2582 *
2583 * %NULL is returned if there is no free memory. Although this function
2584 * allocates memory it can be called from an interrupt.
2585 */
2586static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002587 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002588{
2589 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2590}
2591
Eric Dumazet6f532612012-05-18 05:12:12 +00002592/* legacy helper around __netdev_alloc_skb() */
2593static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2594 gfp_t gfp_mask)
2595{
2596 return __netdev_alloc_skb(NULL, length, gfp_mask);
2597}
2598
2599/* legacy helper around netdev_alloc_skb() */
2600static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2601{
2602 return netdev_alloc_skb(NULL, length);
2603}
2604
2605
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002606static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2607 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002608{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002609 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002610
2611 if (NET_IP_ALIGN && skb)
2612 skb_reserve(skb, NET_IP_ALIGN);
2613 return skb;
2614}
2615
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002616static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2617 unsigned int length)
2618{
2619 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2620}
2621
Alexander Duyck181edb22015-05-06 21:12:03 -07002622static inline void skb_free_frag(void *addr)
2623{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002624 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002625}
2626
Alexander Duyckffde7322014-12-09 19:40:42 -08002627void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002628struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2629 unsigned int length, gfp_t gfp_mask);
2630static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2631 unsigned int length)
2632{
2633 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2634}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002635void napi_consume_skb(struct sk_buff *skb, int budget);
2636
2637void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002638void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002639
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002640/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002641 * __dev_alloc_pages - allocate page for network Rx
2642 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2643 * @order: size of the allocation
2644 *
2645 * Allocate a new page.
2646 *
2647 * %NULL is returned if there is no free memory.
2648*/
2649static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2650 unsigned int order)
2651{
2652 /* This piece of code contains several assumptions.
2653 * 1. This is for device Rx, therefor a cold page is preferred.
2654 * 2. The expectation is the user wants a compound page.
2655 * 3. If requesting a order 0 page it will not be compound
2656 * due to the check to see if order has a value in prep_new_page
2657 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2658 * code in gfp_to_alloc_flags that should be enforcing this.
2659 */
2660 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2661
2662 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2663}
2664
2665static inline struct page *dev_alloc_pages(unsigned int order)
2666{
Neil Horman95829b32016-05-19 11:30:54 -04002667 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002668}
2669
2670/**
2671 * __dev_alloc_page - allocate a page for network Rx
2672 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2673 *
2674 * Allocate a new page.
2675 *
2676 * %NULL is returned if there is no free memory.
2677 */
2678static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2679{
2680 return __dev_alloc_pages(gfp_mask, 0);
2681}
2682
2683static inline struct page *dev_alloc_page(void)
2684{
Neil Horman95829b32016-05-19 11:30:54 -04002685 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002686}
2687
2688/**
Mel Gorman06140022012-07-31 16:44:24 -07002689 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2690 * @page: The page that was allocated from skb_alloc_page
2691 * @skb: The skb that may need pfmemalloc set
2692 */
2693static inline void skb_propagate_pfmemalloc(struct page *page,
2694 struct sk_buff *skb)
2695{
Michal Hocko2f064f32015-08-21 14:11:51 -07002696 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002697 skb->pfmemalloc = true;
2698}
2699
Eric Dumazet564824b2010-10-11 19:05:25 +00002700/**
Masanari Iidae2278672014-02-18 22:54:36 +09002701 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002702 * @frag: the paged fragment
2703 *
2704 * Returns the &struct page associated with @frag.
2705 */
2706static inline struct page *skb_frag_page(const skb_frag_t *frag)
2707{
Ian Campbella8605c62011-10-19 23:01:49 +00002708 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002709}
2710
2711/**
2712 * __skb_frag_ref - take an addition reference on a paged fragment.
2713 * @frag: the paged fragment
2714 *
2715 * Takes an additional reference on the paged fragment @frag.
2716 */
2717static inline void __skb_frag_ref(skb_frag_t *frag)
2718{
2719 get_page(skb_frag_page(frag));
2720}
2721
2722/**
2723 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2724 * @skb: the buffer
2725 * @f: the fragment offset.
2726 *
2727 * Takes an additional reference on the @f'th paged fragment of @skb.
2728 */
2729static inline void skb_frag_ref(struct sk_buff *skb, int f)
2730{
2731 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2732}
2733
2734/**
2735 * __skb_frag_unref - release a reference on a paged fragment.
2736 * @frag: the paged fragment
2737 *
2738 * Releases a reference on the paged fragment @frag.
2739 */
2740static inline void __skb_frag_unref(skb_frag_t *frag)
2741{
2742 put_page(skb_frag_page(frag));
2743}
2744
2745/**
2746 * skb_frag_unref - release a reference on a paged fragment of an skb.
2747 * @skb: the buffer
2748 * @f: the fragment offset
2749 *
2750 * Releases a reference on the @f'th paged fragment of @skb.
2751 */
2752static inline void skb_frag_unref(struct sk_buff *skb, int f)
2753{
2754 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2755}
2756
2757/**
2758 * skb_frag_address - gets the address of the data contained in a paged fragment
2759 * @frag: the paged fragment buffer
2760 *
2761 * Returns the address of the data within @frag. The page must already
2762 * be mapped.
2763 */
2764static inline void *skb_frag_address(const skb_frag_t *frag)
2765{
2766 return page_address(skb_frag_page(frag)) + frag->page_offset;
2767}
2768
2769/**
2770 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2771 * @frag: the paged fragment buffer
2772 *
2773 * Returns the address of the data within @frag. Checks that the page
2774 * is mapped and returns %NULL otherwise.
2775 */
2776static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2777{
2778 void *ptr = page_address(skb_frag_page(frag));
2779 if (unlikely(!ptr))
2780 return NULL;
2781
2782 return ptr + frag->page_offset;
2783}
2784
2785/**
2786 * __skb_frag_set_page - sets the page contained in a paged fragment
2787 * @frag: the paged fragment
2788 * @page: the page to set
2789 *
2790 * Sets the fragment @frag to contain @page.
2791 */
2792static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2793{
Ian Campbella8605c62011-10-19 23:01:49 +00002794 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002795}
2796
2797/**
2798 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2799 * @skb: the buffer
2800 * @f: the fragment offset
2801 * @page: the page to set
2802 *
2803 * Sets the @f'th fragment of @skb to contain @page.
2804 */
2805static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2806 struct page *page)
2807{
2808 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2809}
2810
Eric Dumazet400dfd32013-10-17 16:27:07 -07002811bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2812
Ian Campbell131ea662011-08-19 06:25:00 +00002813/**
2814 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002815 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002816 * @frag: the paged fragment to map
2817 * @offset: the offset within the fragment (starting at the
2818 * fragment's own offset)
2819 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002820 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002821 *
2822 * Maps the page associated with @frag to @device.
2823 */
2824static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2825 const skb_frag_t *frag,
2826 size_t offset, size_t size,
2827 enum dma_data_direction dir)
2828{
2829 return dma_map_page(dev, skb_frag_page(frag),
2830 frag->page_offset + offset, size, dir);
2831}
2832
Eric Dumazet117632e2011-12-03 21:39:53 +00002833static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2834 gfp_t gfp_mask)
2835{
2836 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2837}
2838
Octavian Purdilabad93e92014-06-12 01:36:26 +03002839
2840static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2841 gfp_t gfp_mask)
2842{
2843 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2844}
2845
2846
Ian Campbell131ea662011-08-19 06:25:00 +00002847/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002848 * skb_clone_writable - is the header of a clone writable
2849 * @skb: buffer to check
2850 * @len: length up to which to write
2851 *
2852 * Returns true if modifying the header part of the cloned buffer
2853 * does not requires the data to be copied.
2854 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002855static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002856{
2857 return !skb_header_cloned(skb) &&
2858 skb_headroom(skb) + len <= skb->hdr_len;
2859}
2860
Daniel Borkmann36976492016-02-19 23:05:25 +01002861static inline int skb_try_make_writable(struct sk_buff *skb,
2862 unsigned int write_len)
2863{
2864 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2865 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2866}
2867
Herbert Xud9cc2042007-09-16 16:21:16 -07002868static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2869 int cloned)
2870{
2871 int delta = 0;
2872
Herbert Xud9cc2042007-09-16 16:21:16 -07002873 if (headroom > skb_headroom(skb))
2874 delta = headroom - skb_headroom(skb);
2875
2876 if (delta || cloned)
2877 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2878 GFP_ATOMIC);
2879 return 0;
2880}
2881
Patrick McHardy334a8132007-06-25 04:35:20 -07002882/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002883 * skb_cow - copy header of skb when it is required
2884 * @skb: buffer to cow
2885 * @headroom: needed headroom
2886 *
2887 * If the skb passed lacks sufficient headroom or its data part
2888 * is shared, data is reallocated. If reallocation fails, an error
2889 * is returned and original skb is not changed.
2890 *
2891 * The result is skb with writable area skb->head...skb->tail
2892 * and at least @headroom of space at head.
2893 */
2894static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2895{
Herbert Xud9cc2042007-09-16 16:21:16 -07002896 return __skb_cow(skb, headroom, skb_cloned(skb));
2897}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002898
Herbert Xud9cc2042007-09-16 16:21:16 -07002899/**
2900 * skb_cow_head - skb_cow but only making the head writable
2901 * @skb: buffer to cow
2902 * @headroom: needed headroom
2903 *
2904 * This function is identical to skb_cow except that we replace the
2905 * skb_cloned check by skb_header_cloned. It should be used when
2906 * you only need to push on some header and do not need to modify
2907 * the data.
2908 */
2909static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2910{
2911 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002912}
2913
2914/**
2915 * skb_padto - pad an skbuff up to a minimal size
2916 * @skb: buffer to pad
2917 * @len: minimal length
2918 *
2919 * Pads up a buffer to ensure the trailing bytes exist and are
2920 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002921 * is untouched. Otherwise it is extended. Returns zero on
2922 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002923 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002924static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002925{
2926 unsigned int size = skb->len;
2927 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002928 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002929 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002930}
2931
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002932/**
2933 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2934 * @skb: buffer to pad
2935 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07002936 * @free_on_error: free buffer on error
2937 *
2938 * Pads up a buffer to ensure the trailing bytes exist and are
2939 * blanked. If the buffer already contains sufficient data it
2940 * is untouched. Otherwise it is extended. Returns zero on
2941 * success. The skb is freed on error if @free_on_error is true.
2942 */
2943static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2944 bool free_on_error)
2945{
2946 unsigned int size = skb->len;
2947
2948 if (unlikely(size < len)) {
2949 len -= size;
2950 if (__skb_pad(skb, len, free_on_error))
2951 return -ENOMEM;
2952 __skb_put(skb, len);
2953 }
2954 return 0;
2955}
2956
2957/**
2958 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2959 * @skb: buffer to pad
2960 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002961 *
2962 * Pads up a buffer to ensure the trailing bytes exist and are
2963 * blanked. If the buffer already contains sufficient data it
2964 * is untouched. Otherwise it is extended. Returns zero on
2965 * success. The skb is freed on error.
2966 */
2967static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2968{
Florian Fainellicd0a1372017-08-22 15:12:14 -07002969 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002970}
2971
Linus Torvalds1da177e2005-04-16 15:20:36 -07002972static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002973 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002974{
2975 const int off = skb->len;
2976
2977 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002978 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002979 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2980 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002981 skb->csum = csum_block_add(skb->csum, csum, off);
2982 return 0;
2983 }
Al Viro15e6cb42016-11-01 22:42:45 -04002984 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002985 return 0;
2986
2987 __skb_trim(skb, off);
2988 return -EFAULT;
2989}
2990
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002991static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2992 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002993{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002994 if (skb_zcopy(skb))
2995 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002996 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002997 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002998
Ian Campbellea2ab692011-08-22 23:44:58 +00002999 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00003000 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003001 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003002 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003003}
3004
Herbert Xu364c6ba2006-06-09 16:10:40 -07003005static inline int __skb_linearize(struct sk_buff *skb)
3006{
3007 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3008}
3009
Linus Torvalds1da177e2005-04-16 15:20:36 -07003010/**
3011 * skb_linearize - convert paged skb to linear one
3012 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003013 *
3014 * If there is no free memory -ENOMEM is returned, otherwise zero
3015 * is returned and the old skb data released.
3016 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003017static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003018{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003019 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3020}
3021
3022/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003023 * skb_has_shared_frag - can any frag be overwritten
3024 * @skb: buffer to test
3025 *
3026 * Return true if the skb has at least one frag that might be modified
3027 * by an external entity (as in vmsplice()/sendfile())
3028 */
3029static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3030{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003031 return skb_is_nonlinear(skb) &&
3032 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003033}
3034
3035/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003036 * skb_linearize_cow - make sure skb is linear and writable
3037 * @skb: buffer to process
3038 *
3039 * If there is no free memory -ENOMEM is returned, otherwise zero
3040 * is returned and the old skb data released.
3041 */
3042static inline int skb_linearize_cow(struct sk_buff *skb)
3043{
3044 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3045 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003046}
3047
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003048static __always_inline void
3049__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3050 unsigned int off)
3051{
3052 if (skb->ip_summed == CHECKSUM_COMPLETE)
3053 skb->csum = csum_block_sub(skb->csum,
3054 csum_partial(start, len, 0), off);
3055 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3056 skb_checksum_start_offset(skb) < 0)
3057 skb->ip_summed = CHECKSUM_NONE;
3058}
3059
Linus Torvalds1da177e2005-04-16 15:20:36 -07003060/**
3061 * skb_postpull_rcsum - update checksum for received skb after pull
3062 * @skb: buffer to update
3063 * @start: start of data before pull
3064 * @len: length of data pulled
3065 *
3066 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003067 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3068 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003069 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003071 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003072{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003073 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003074}
3075
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003076static __always_inline void
3077__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3078 unsigned int off)
3079{
3080 if (skb->ip_summed == CHECKSUM_COMPLETE)
3081 skb->csum = csum_block_add(skb->csum,
3082 csum_partial(start, len, 0), off);
3083}
Herbert Xucbb042f2006-03-20 22:43:56 -08003084
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003085/**
3086 * skb_postpush_rcsum - update checksum for received skb after push
3087 * @skb: buffer to update
3088 * @start: start of data after push
3089 * @len: length of data pushed
3090 *
3091 * After doing a push on a received packet, you need to call this to
3092 * update the CHECKSUM_COMPLETE checksum.
3093 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003094static inline void skb_postpush_rcsum(struct sk_buff *skb,
3095 const void *start, unsigned int len)
3096{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003097 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003098}
3099
Johannes Bergaf728682017-06-16 14:29:22 +02003100void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02003101
David S. Miller7ce5a272013-12-02 17:26:05 -05003102/**
WANG Cong82a31b92016-06-30 10:15:22 -07003103 * skb_push_rcsum - push skb and update receive checksum
3104 * @skb: buffer to update
3105 * @len: length of data pulled
3106 *
3107 * This function performs an skb_push on the packet and updates
3108 * the CHECKSUM_COMPLETE checksum. It should be used on
3109 * receive path processing instead of skb_push unless you know
3110 * that the checksum difference is zero (e.g., a valid IP header)
3111 * or you are setting ip_summed to CHECKSUM_NONE.
3112 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003113static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003114{
3115 skb_push(skb, len);
3116 skb_postpush_rcsum(skb, skb->data, len);
3117 return skb->data;
3118}
3119
3120/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003121 * pskb_trim_rcsum - trim received skb and update checksum
3122 * @skb: buffer to trim
3123 * @len: new length
3124 *
3125 * This is exactly the same as pskb_trim except that it ensures the
3126 * checksum of received packets are still valid after the operation.
3127 */
3128
3129static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3130{
3131 if (likely(len >= skb->len))
3132 return 0;
3133 if (skb->ip_summed == CHECKSUM_COMPLETE)
3134 skb->ip_summed = CHECKSUM_NONE;
3135 return __pskb_trim(skb, len);
3136}
3137
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003138static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3139{
3140 if (skb->ip_summed == CHECKSUM_COMPLETE)
3141 skb->ip_summed = CHECKSUM_NONE;
3142 __skb_trim(skb, len);
3143 return 0;
3144}
3145
3146static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3147{
3148 if (skb->ip_summed == CHECKSUM_COMPLETE)
3149 skb->ip_summed = CHECKSUM_NONE;
3150 return __skb_grow(skb, len);
3151}
3152
Linus Torvalds1da177e2005-04-16 15:20:36 -07003153#define skb_queue_walk(queue, skb) \
3154 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003155 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003156 skb = skb->next)
3157
James Chapman46f89142007-04-30 00:07:31 -07003158#define skb_queue_walk_safe(queue, skb, tmp) \
3159 for (skb = (queue)->next, tmp = skb->next; \
3160 skb != (struct sk_buff *)(queue); \
3161 skb = tmp, tmp = skb->next)
3162
David S. Miller1164f522008-09-23 00:49:44 -07003163#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003164 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003165 skb = skb->next)
3166
3167#define skb_queue_walk_from_safe(queue, skb, tmp) \
3168 for (tmp = skb->next; \
3169 skb != (struct sk_buff *)(queue); \
3170 skb = tmp, tmp = skb->next)
3171
Stephen Hemminger300ce172005-10-30 13:47:34 -08003172#define skb_queue_reverse_walk(queue, skb) \
3173 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003174 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003175 skb = skb->prev)
3176
David S. Miller686a2952011-01-20 22:47:32 -08003177#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3178 for (skb = (queue)->prev, tmp = skb->prev; \
3179 skb != (struct sk_buff *)(queue); \
3180 skb = tmp, tmp = skb->prev)
3181
3182#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3183 for (tmp = skb->prev; \
3184 skb != (struct sk_buff *)(queue); \
3185 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003186
David S. Miller21dc3302010-08-23 00:13:46 -07003187static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003188{
3189 return skb_shinfo(skb)->frag_list != NULL;
3190}
3191
3192static inline void skb_frag_list_init(struct sk_buff *skb)
3193{
3194 skb_shinfo(skb)->frag_list = NULL;
3195}
3196
David S. Milleree039872009-06-09 00:17:13 -07003197#define skb_walk_frags(skb, iter) \
3198 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3199
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003200
3201int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3202 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003203struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3204 struct sk_buff_head *queue,
3205 unsigned int flags,
3206 void (*destructor)(struct sock *sk,
3207 struct sk_buff *skb),
3208 int *peeked, int *off, int *err,
3209 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003210struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003211 void (*destructor)(struct sock *sk,
3212 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003213 int *peeked, int *off, int *err,
3214 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003215struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003216 void (*destructor)(struct sock *sk,
3217 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003218 int *peeked, int *off, int *err);
3219struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3220 int *err);
3221unsigned int datagram_poll(struct file *file, struct socket *sock,
3222 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003223int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3224 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003225static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3226 struct msghdr *msg, int size)
3227{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003228 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003229}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003230int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3231 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003232int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3233 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003234int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003235void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003236void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3237static inline void skb_free_datagram_locked(struct sock *sk,
3238 struct sk_buff *skb)
3239{
3240 __skb_free_datagram_locked(sk, skb, 0);
3241}
Joe Perches7965bd42013-09-26 14:48:15 -07003242int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003243int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3244int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3245__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3246 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003247int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003248 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003249 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003250int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3251 int len);
3252int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003253void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003254unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003255int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3256 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003257void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3258int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3259void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003260unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003261bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003262struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003263struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003264int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003265int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003266int skb_vlan_pop(struct sk_buff *skb);
3267int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003268struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3269 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003270
Al Viro6ce8e9c2014-04-06 21:25:44 -04003271static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3272{
Al Viro3073f072017-02-17 23:13:25 -05003273 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003274}
3275
Al Viro7eab8d92014-04-06 21:51:23 -04003276static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3277{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003278 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003279}
3280
Daniel Borkmann2817a332013-10-30 11:50:51 +01003281struct skb_checksum_ops {
3282 __wsum (*update)(const void *mem, int len, __wsum wsum);
3283 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3284};
3285
Davide Caratti96178132017-05-18 15:44:37 +02003286extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3287
Daniel Borkmann2817a332013-10-30 11:50:51 +01003288__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3289 __wsum csum, const struct skb_checksum_ops *ops);
3290__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3291 __wsum csum);
3292
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003293static inline void * __must_check
3294__skb_header_pointer(const struct sk_buff *skb, int offset,
3295 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003296{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003297 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003298 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003299
David S. Miller690e36e2014-08-23 12:13:41 -07003300 if (!skb ||
3301 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003302 return NULL;
3303
3304 return buffer;
3305}
3306
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003307static inline void * __must_check
3308skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003309{
3310 return __skb_header_pointer(skb, offset, len, skb->data,
3311 skb_headlen(skb), buffer);
3312}
3313
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003314/**
3315 * skb_needs_linearize - check if we need to linearize a given skb
3316 * depending on the given device features.
3317 * @skb: socket buffer to check
3318 * @features: net device features
3319 *
3320 * Returns true if either:
3321 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3322 * 2. skb is fragmented and the device does not support SG.
3323 */
3324static inline bool skb_needs_linearize(struct sk_buff *skb,
3325 netdev_features_t features)
3326{
3327 return skb_is_nonlinear(skb) &&
3328 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3329 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3330}
3331
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003332static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3333 void *to,
3334 const unsigned int len)
3335{
3336 memcpy(to, skb->data, len);
3337}
3338
3339static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3340 const int offset, void *to,
3341 const unsigned int len)
3342{
3343 memcpy(to, skb->data + offset, len);
3344}
3345
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003346static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3347 const void *from,
3348 const unsigned int len)
3349{
3350 memcpy(skb->data, from, len);
3351}
3352
3353static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3354 const int offset,
3355 const void *from,
3356 const unsigned int len)
3357{
3358 memcpy(skb->data + offset, from, len);
3359}
3360
Joe Perches7965bd42013-09-26 14:48:15 -07003361void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003362
Patrick Ohlyac45f602009-02-12 05:03:37 +00003363static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3364{
3365 return skb->tstamp;
3366}
3367
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003368/**
3369 * skb_get_timestamp - get timestamp from a skb
3370 * @skb: skb to get stamp from
3371 * @stamp: pointer to struct timeval to store stamp in
3372 *
3373 * Timestamps are stored in the skb as offsets to a base timestamp.
3374 * This function converts the offset back to a struct timeval and stores
3375 * it in stamp.
3376 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003377static inline void skb_get_timestamp(const struct sk_buff *skb,
3378 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003379{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003380 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003381}
3382
Patrick Ohlyac45f602009-02-12 05:03:37 +00003383static inline void skb_get_timestampns(const struct sk_buff *skb,
3384 struct timespec *stamp)
3385{
3386 *stamp = ktime_to_timespec(skb->tstamp);
3387}
3388
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003389static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003390{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003391 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003392}
3393
Stephen Hemminger164891a2007-04-23 22:26:16 -07003394static inline ktime_t net_timedelta(ktime_t t)
3395{
3396 return ktime_sub(ktime_get_real(), t);
3397}
3398
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003399static inline ktime_t net_invalid_timestamp(void)
3400{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003401 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003402}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003403
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003404static inline u8 skb_metadata_len(const struct sk_buff *skb)
3405{
3406 return skb_shinfo(skb)->meta_len;
3407}
3408
3409static inline void *skb_metadata_end(const struct sk_buff *skb)
3410{
3411 return skb_mac_header(skb);
3412}
3413
3414static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3415 const struct sk_buff *skb_b,
3416 u8 meta_len)
3417{
3418 const void *a = skb_metadata_end(skb_a);
3419 const void *b = skb_metadata_end(skb_b);
3420 /* Using more efficient varaiant than plain call to memcmp(). */
3421#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3422 u64 diffs = 0;
3423
3424 switch (meta_len) {
3425#define __it(x, op) (x -= sizeof(u##op))
3426#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3427 case 32: diffs |= __it_diff(a, b, 64);
3428 case 24: diffs |= __it_diff(a, b, 64);
3429 case 16: diffs |= __it_diff(a, b, 64);
3430 case 8: diffs |= __it_diff(a, b, 64);
3431 break;
3432 case 28: diffs |= __it_diff(a, b, 64);
3433 case 20: diffs |= __it_diff(a, b, 64);
3434 case 12: diffs |= __it_diff(a, b, 64);
3435 case 4: diffs |= __it_diff(a, b, 32);
3436 break;
3437 }
3438 return diffs;
3439#else
3440 return memcmp(a - meta_len, b - meta_len, meta_len);
3441#endif
3442}
3443
3444static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3445 const struct sk_buff *skb_b)
3446{
3447 u8 len_a = skb_metadata_len(skb_a);
3448 u8 len_b = skb_metadata_len(skb_b);
3449
3450 if (!(len_a | len_b))
3451 return false;
3452
3453 return len_a != len_b ?
3454 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3455}
3456
3457static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3458{
3459 skb_shinfo(skb)->meta_len = meta_len;
3460}
3461
3462static inline void skb_metadata_clear(struct sk_buff *skb)
3463{
3464 skb_metadata_set(skb, 0);
3465}
3466
Alexander Duyck62bccb82014-09-04 13:31:35 -04003467struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3468
Richard Cochranc1f19b52010-07-17 08:49:36 +00003469#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3470
Joe Perches7965bd42013-09-26 14:48:15 -07003471void skb_clone_tx_timestamp(struct sk_buff *skb);
3472bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003473
3474#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3475
3476static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3477{
3478}
3479
3480static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3481{
3482 return false;
3483}
3484
3485#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3486
3487/**
3488 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3489 *
Richard Cochranda92b192011-10-21 00:49:15 +00003490 * PHY drivers may accept clones of transmitted packets for
3491 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003492 * must call this function to return the skb back to the stack with a
3493 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003494 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003495 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003496 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003497 *
3498 */
3499void skb_complete_tx_timestamp(struct sk_buff *skb,
3500 struct skb_shared_hwtstamps *hwtstamps);
3501
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003502void __skb_tstamp_tx(struct sk_buff *orig_skb,
3503 struct skb_shared_hwtstamps *hwtstamps,
3504 struct sock *sk, int tstype);
3505
Patrick Ohlyac45f602009-02-12 05:03:37 +00003506/**
3507 * skb_tstamp_tx - queue clone of skb with send time stamps
3508 * @orig_skb: the original outgoing packet
3509 * @hwtstamps: hardware time stamps, may be NULL if not available
3510 *
3511 * If the skb has a socket associated, then this function clones the
3512 * skb (thus sharing the actual data and optional structures), stores
3513 * the optional hardware time stamping information (if non NULL) or
3514 * generates a software time stamp (otherwise), then queues the clone
3515 * to the error queue of the socket. Errors are silently ignored.
3516 */
Joe Perches7965bd42013-09-26 14:48:15 -07003517void skb_tstamp_tx(struct sk_buff *orig_skb,
3518 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003519
Richard Cochran4507a712010-07-17 08:48:28 +00003520/**
3521 * skb_tx_timestamp() - Driver hook for transmit timestamping
3522 *
3523 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003524 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003525 *
David S. Miller73409f32013-12-27 13:04:33 -05003526 * Specifically, one should make absolutely sure that this function is
3527 * called before TX completion of this packet can trigger. Otherwise
3528 * the packet could potentially already be freed.
3529 *
Richard Cochran4507a712010-07-17 08:48:28 +00003530 * @skb: A socket buffer.
3531 */
3532static inline void skb_tx_timestamp(struct sk_buff *skb)
3533{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003534 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003535 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3536 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003537}
3538
Johannes Berg6e3e9392011-11-09 10:15:42 +01003539/**
3540 * skb_complete_wifi_ack - deliver skb with wifi status
3541 *
3542 * @skb: the original outgoing packet
3543 * @acked: ack status
3544 *
3545 */
3546void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3547
Joe Perches7965bd42013-09-26 14:48:15 -07003548__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3549__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003550
Herbert Xu60476372007-04-09 11:59:39 -07003551static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3552{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003553 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3554 skb->csum_valid ||
3555 (skb->ip_summed == CHECKSUM_PARTIAL &&
3556 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003557}
3558
Herbert Xufb286bb2005-11-10 13:01:24 -08003559/**
3560 * skb_checksum_complete - Calculate checksum of an entire packet
3561 * @skb: packet to process
3562 *
3563 * This function calculates the checksum over the entire packet plus
3564 * the value of skb->csum. The latter can be used to supply the
3565 * checksum of a pseudo header as used by TCP/UDP. It returns the
3566 * checksum.
3567 *
3568 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3569 * this function can be used to verify that checksum on received
3570 * packets. In that case the function should return zero if the
3571 * checksum is correct. In particular, this function will return zero
3572 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3573 * hardware has already verified the correctness of the checksum.
3574 */
Al Viro4381ca32007-07-15 21:00:11 +01003575static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003576{
Herbert Xu60476372007-04-09 11:59:39 -07003577 return skb_csum_unnecessary(skb) ?
3578 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003579}
3580
Tom Herbert77cffe22014-08-27 21:26:46 -07003581static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3582{
3583 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3584 if (skb->csum_level == 0)
3585 skb->ip_summed = CHECKSUM_NONE;
3586 else
3587 skb->csum_level--;
3588 }
3589}
3590
3591static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3592{
3593 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3594 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3595 skb->csum_level++;
3596 } else if (skb->ip_summed == CHECKSUM_NONE) {
3597 skb->ip_summed = CHECKSUM_UNNECESSARY;
3598 skb->csum_level = 0;
3599 }
3600}
3601
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003602/* Check if we need to perform checksum complete validation.
3603 *
3604 * Returns true if checksum complete is needed, false otherwise
3605 * (either checksum is unnecessary or zero checksum is allowed).
3606 */
3607static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3608 bool zero_okay,
3609 __sum16 check)
3610{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003611 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3612 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003613 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003614 return false;
3615 }
3616
3617 return true;
3618}
3619
3620/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3621 * in checksum_init.
3622 */
3623#define CHECKSUM_BREAK 76
3624
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003625/* Unset checksum-complete
3626 *
3627 * Unset checksum complete can be done when packet is being modified
3628 * (uncompressed for instance) and checksum-complete value is
3629 * invalidated.
3630 */
3631static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3632{
3633 if (skb->ip_summed == CHECKSUM_COMPLETE)
3634 skb->ip_summed = CHECKSUM_NONE;
3635}
3636
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003637/* Validate (init) checksum based on checksum complete.
3638 *
3639 * Return values:
3640 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3641 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3642 * checksum is stored in skb->csum for use in __skb_checksum_complete
3643 * non-zero: value of invalid checksum
3644 *
3645 */
3646static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3647 bool complete,
3648 __wsum psum)
3649{
3650 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3651 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003652 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003653 return 0;
3654 }
3655 }
3656
3657 skb->csum = psum;
3658
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003659 if (complete || skb->len <= CHECKSUM_BREAK) {
3660 __sum16 csum;
3661
3662 csum = __skb_checksum_complete(skb);
3663 skb->csum_valid = !csum;
3664 return csum;
3665 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003666
3667 return 0;
3668}
3669
3670static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3671{
3672 return 0;
3673}
3674
3675/* Perform checksum validate (init). Note that this is a macro since we only
3676 * want to calculate the pseudo header which is an input function if necessary.
3677 * First we try to validate without any computation (checksum unnecessary) and
3678 * then calculate based on checksum complete calling the function to compute
3679 * pseudo header.
3680 *
3681 * Return values:
3682 * 0: checksum is validated or try to in skb_checksum_complete
3683 * non-zero: value of invalid checksum
3684 */
3685#define __skb_checksum_validate(skb, proto, complete, \
3686 zero_okay, check, compute_pseudo) \
3687({ \
3688 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003689 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003690 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3691 __ret = __skb_checksum_validate_complete(skb, \
3692 complete, compute_pseudo(skb, proto)); \
3693 __ret; \
3694})
3695
3696#define skb_checksum_init(skb, proto, compute_pseudo) \
3697 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3698
3699#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3700 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3701
3702#define skb_checksum_validate(skb, proto, compute_pseudo) \
3703 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3704
3705#define skb_checksum_validate_zero_check(skb, proto, check, \
3706 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003707 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003708
3709#define skb_checksum_simple_validate(skb) \
3710 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3711
Tom Herbertd96535a2014-08-31 15:12:42 -07003712static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3713{
Davide Caratti219f1d792017-05-18 15:44:39 +02003714 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003715}
3716
3717static inline void __skb_checksum_convert(struct sk_buff *skb,
3718 __sum16 check, __wsum pseudo)
3719{
3720 skb->csum = ~pseudo;
3721 skb->ip_summed = CHECKSUM_COMPLETE;
3722}
3723
3724#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3725do { \
3726 if (__skb_checksum_convert_check(skb)) \
3727 __skb_checksum_convert(skb, check, \
3728 compute_pseudo(skb, proto)); \
3729} while (0)
3730
Tom Herbert15e23962015-02-10 16:30:31 -08003731static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3732 u16 start, u16 offset)
3733{
3734 skb->ip_summed = CHECKSUM_PARTIAL;
3735 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3736 skb->csum_offset = offset - start;
3737}
3738
Tom Herbertdcdc8992015-02-02 16:07:34 -08003739/* Update skbuf and packet to reflect the remote checksum offload operation.
3740 * When called, ptr indicates the starting point for skb->csum when
3741 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3742 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3743 */
3744static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003745 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003746{
3747 __wsum delta;
3748
Tom Herbert15e23962015-02-10 16:30:31 -08003749 if (!nopartial) {
3750 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3751 return;
3752 }
3753
Tom Herbertdcdc8992015-02-02 16:07:34 -08003754 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3755 __skb_checksum_complete(skb);
3756 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3757 }
3758
3759 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3760
3761 /* Adjust skb->csum since we changed the packet */
3762 skb->csum = csum_add(skb->csum, delta);
3763}
3764
Florian Westphalcb9c6832017-01-23 18:21:56 +01003765static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3766{
3767#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003768 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003769#else
3770 return NULL;
3771#endif
3772}
3773
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003774#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003775void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003776static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3777{
3778 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003779 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003780}
3781static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3782{
3783 if (nfct)
3784 atomic_inc(&nfct->use);
3785}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003786#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003787#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003788static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3789{
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003790 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003791 kfree(nf_bridge);
3792}
3793static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3794{
3795 if (nf_bridge)
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003796 refcount_inc(&nf_bridge->use);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003797}
3798#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003799static inline void nf_reset(struct sk_buff *skb)
3800{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003801#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003802 nf_conntrack_put(skb_nfct(skb));
3803 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003804#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003805#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003806 nf_bridge_put(skb->nf_bridge);
3807 skb->nf_bridge = NULL;
3808#endif
3809}
3810
Patrick McHardy124dff02013-04-05 20:42:05 +02003811static inline void nf_reset_trace(struct sk_buff *skb)
3812{
Florian Westphal478b3602014-02-15 23:48:45 +01003813#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003814 skb->nf_trace = 0;
3815#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003816}
3817
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003818/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003819static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3820 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003821{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003822#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003823 dst->_nfct = src->_nfct;
3824 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003825#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003826#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003827 dst->nf_bridge = src->nf_bridge;
3828 nf_bridge_get(src->nf_bridge);
3829#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003830#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003831 if (copy)
3832 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003833#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003834}
3835
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003836static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3837{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003838#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003839 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003840#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003841#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003842 nf_bridge_put(dst->nf_bridge);
3843#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003844 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003845}
3846
James Morris984bc162006-06-09 00:29:17 -07003847#ifdef CONFIG_NETWORK_SECMARK
3848static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3849{
3850 to->secmark = from->secmark;
3851}
3852
3853static inline void skb_init_secmark(struct sk_buff *skb)
3854{
3855 skb->secmark = 0;
3856}
3857#else
3858static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3859{ }
3860
3861static inline void skb_init_secmark(struct sk_buff *skb)
3862{ }
3863#endif
3864
Eric W. Biederman574f7192014-04-01 12:20:24 -07003865static inline bool skb_irq_freeable(const struct sk_buff *skb)
3866{
3867 return !skb->destructor &&
3868#if IS_ENABLED(CONFIG_XFRM)
3869 !skb->sp &&
3870#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003871 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003872 !skb->_skb_refdst &&
3873 !skb_has_frag_list(skb);
3874}
3875
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003876static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3877{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003878 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003879}
3880
Stephen Hemminger92477442009-03-21 13:39:26 -07003881static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003882{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003883 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003884}
3885
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003886static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3887{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003888 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003889}
3890
David S. Millerd5a9e242009-01-27 16:22:11 -08003891static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3892{
3893 skb->queue_mapping = rx_queue + 1;
3894}
3895
Stephen Hemminger92477442009-03-21 13:39:26 -07003896static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003897{
3898 return skb->queue_mapping - 1;
3899}
3900
Stephen Hemminger92477442009-03-21 13:39:26 -07003901static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003902{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003903 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003904}
3905
Julian Anastasov4ff06202017-02-06 23:14:12 +02003906static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3907{
3908 skb->dst_pending_confirm = val;
3909}
3910
3911static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3912{
3913 return skb->dst_pending_confirm != 0;
3914}
3915
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003916static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3917{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003918#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003919 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003920#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003921 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003922#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003923}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003924
Pravin B Shelar68c33162013-02-14 14:02:41 +00003925/* Keeps track of mac header offset relative to skb->head.
3926 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3927 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003928 * tunnel skb it points to outer mac header.
3929 * Keeps track of level of encapsulation of network headers.
3930 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003931struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003932 union {
3933 int mac_offset;
3934 int data_offset;
3935 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003936 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003937 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003938 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003939};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003940#define SKB_SGO_CB_OFFSET 32
3941#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003942
3943static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3944{
3945 return (skb_mac_header(inner_skb) - inner_skb->head) -
3946 SKB_GSO_CB(inner_skb)->mac_offset;
3947}
3948
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003949static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3950{
3951 int new_headroom, headroom;
3952 int ret;
3953
3954 headroom = skb_headroom(skb);
3955 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3956 if (ret)
3957 return ret;
3958
3959 new_headroom = skb_headroom(skb);
3960 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3961 return 0;
3962}
3963
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003964static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3965{
3966 /* Do not update partial checksums if remote checksum is enabled. */
3967 if (skb->remcsum_offload)
3968 return;
3969
3970 SKB_GSO_CB(skb)->csum = res;
3971 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3972}
3973
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003974/* Compute the checksum for a gso segment. First compute the checksum value
3975 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3976 * then add in skb->csum (checksum from csum_start to end of packet).
3977 * skb->csum and csum_start are then updated to reflect the checksum of the
3978 * resultant packet starting from the transport header-- the resultant checksum
3979 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3980 * header.
3981 */
3982static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3983{
Alexander Duyck76443452016-02-05 15:27:37 -08003984 unsigned char *csum_start = skb_transport_header(skb);
3985 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3986 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003987
Alexander Duyck76443452016-02-05 15:27:37 -08003988 SKB_GSO_CB(skb)->csum = res;
3989 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003990
Alexander Duyck76443452016-02-05 15:27:37 -08003991 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003992}
3993
David S. Millerbdcc0922012-03-07 20:53:36 -05003994static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003995{
3996 return skb_shinfo(skb)->gso_size;
3997}
3998
Eric Dumazet36a8f392013-09-29 01:21:32 -07003999/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004000static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004001{
4002 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4003}
4004
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004005static inline void skb_gso_reset(struct sk_buff *skb)
4006{
4007 skb_shinfo(skb)->gso_size = 0;
4008 skb_shinfo(skb)->gso_segs = 0;
4009 skb_shinfo(skb)->gso_type = 0;
4010}
4011
Joe Perches7965bd42013-09-26 14:48:15 -07004012void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004013
4014static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4015{
4016 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4017 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004018 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4019
Alexander Duyckb78462e2010-06-02 12:24:37 +00004020 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4021 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004022 __skb_warn_lro_forwarding(skb);
4023 return true;
4024 }
4025 return false;
4026}
4027
Herbert Xu35fc92a2007-03-26 23:22:20 -07004028static inline void skb_forward_csum(struct sk_buff *skb)
4029{
4030 /* Unfortunately we don't support this one. Any brave souls? */
4031 if (skb->ip_summed == CHECKSUM_COMPLETE)
4032 skb->ip_summed = CHECKSUM_NONE;
4033}
4034
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004035/**
4036 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4037 * @skb: skb to check
4038 *
4039 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4040 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4041 * use this helper, to document places where we make this assertion.
4042 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004043static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004044{
4045#ifdef DEBUG
4046 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4047#endif
4048}
4049
Rusty Russellf35d9d82008-02-04 23:49:54 -05004050bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004051
Paul Durranted1f50c2014-01-09 10:02:46 +00004052int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004053struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4054 unsigned int transport_len,
4055 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004056
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004057/**
4058 * skb_head_is_locked - Determine if the skb->head is locked down
4059 * @skb: skb to check
4060 *
4061 * The head on skbs build around a head frag can be removed if they are
4062 * not cloned. This function returns true if the skb head is locked down
4063 * due to either being allocated via kmalloc, or by being a clone with
4064 * multiple references to the head.
4065 */
4066static inline bool skb_head_is_locked(const struct sk_buff *skb)
4067{
4068 return !skb->head_frag || skb_cloned(skb);
4069}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004070
4071/**
4072 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4073 *
4074 * @skb: GSO skb
4075 *
4076 * skb_gso_network_seglen is used to determine the real size of the
4077 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4078 *
4079 * The MAC/L2 header is not accounted for.
4080 */
4081static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4082{
4083 unsigned int hdr_len = skb_transport_header(skb) -
4084 skb_network_header(skb);
4085 return hdr_len + skb_gso_transport_seglen(skb);
4086}
Thomas Grafee122c72015-07-21 10:43:58 +02004087
Edward Cree179bc672016-02-11 20:48:04 +00004088/* Local Checksum Offload.
4089 * Compute outer checksum based on the assumption that the
4090 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00004091 * See Documentation/networking/checksum-offloads.txt for
4092 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004093 * Fill in outer checksum adjustment (e.g. with sum of outer
4094 * pseudo-header) before calling.
4095 * Also ensure that inner checksum is in linear data area.
4096 */
4097static inline __wsum lco_csum(struct sk_buff *skb)
4098{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004099 unsigned char *csum_start = skb_checksum_start(skb);
4100 unsigned char *l4_hdr = skb_transport_header(skb);
4101 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004102
4103 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004104 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4105 skb->csum_offset));
4106
Edward Cree179bc672016-02-11 20:48:04 +00004107 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004108 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004109 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004110 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004111}
4112
Linus Torvalds1da177e2005-04-16 15:20:36 -07004113#endif /* __KERNEL__ */
4114#endif /* _LINUX_SKBUFF_H */