blob: 638b0e00431074ea2c388fe61a61f73cd6b58f10 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Arun Sharma600634972011-07-26 16:09:06 -070026#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <asm/types.h>
28#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000032#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070033#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000034#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000035#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080036#include <linux/sched.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020037#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020038#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020039#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040040#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070041#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Tom Herbert7a6ae712015-12-14 11:19:47 -080043/* The interface for checksum offload between the stack and networking drivers
44 * is as follows...
45 *
46 * A. IP checksum related features
47 *
48 * Drivers advertise checksum offload capabilities in the features of a device.
49 * From the stack's point of view these are capabilities offered by the driver,
50 * a driver typically only advertises features that it is capable of offloading
51 * to its device.
52 *
53 * The checksum related features are:
54 *
55 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
56 * IP (one's complement) checksum for any combination
57 * of protocols or protocol layering. The checksum is
58 * computed and set in a packet per the CHECKSUM_PARTIAL
59 * interface (see below).
60 *
61 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
62 * TCP or UDP packets over IPv4. These are specifically
63 * unencapsulated packets of the form IPv4|TCP or
64 * IPv4|UDP where the Protocol field in the IPv4 header
65 * is TCP or UDP. The IPv4 header may contain IP options
66 * This feature cannot be set in features for a device
67 * with NETIF_F_HW_CSUM also set. This feature is being
68 * DEPRECATED (see below).
69 *
70 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
71 * TCP or UDP packets over IPv6. These are specifically
72 * unencapsulated packets of the form IPv6|TCP or
73 * IPv4|UDP where the Next Header field in the IPv6
74 * header is either TCP or UDP. IPv6 extension headers
75 * are not supported with this feature. This feature
76 * cannot be set in features for a device with
77 * NETIF_F_HW_CSUM also set. This feature is being
78 * DEPRECATED (see below).
79 *
80 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
81 * This flag is used only used to disable the RX checksum
82 * feature for a device. The stack will accept receive
83 * checksum indication in packets received on a device
84 * regardless of whether NETIF_F_RXCSUM is set.
85 *
86 * B. Checksumming of received packets by device. Indication of checksum
87 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010088 *
89 * CHECKSUM_NONE:
90 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080091 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010092 * The packet contains full (though not verified) checksum in packet but
93 * not in skb->csum. Thus, skb->csum is undefined in this case.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * The hardware you're dealing with doesn't calculate the full checksum
98 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070099 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
100 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800101 * though. A driver or device must never modify the checksum field in the
102 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700103 *
104 * CHECKSUM_UNNECESSARY is applicable to following protocols:
105 * TCP: IPv6 and IPv4.
106 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
107 * zero UDP checksum for either IPv4 or IPv6, the networking stack
108 * may perform further validation in this case.
109 * GRE: only if the checksum is present in the header.
110 * SCTP: indicates the CRC in SCTP header has been validated.
111 *
112 * skb->csum_level indicates the number of consecutive checksums found in
113 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
114 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
115 * and a device is able to verify the checksums for UDP (possibly zero),
116 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
117 * two. If the device were only able to verify the UDP checksum and not
118 * GRE, either because it doesn't support GRE checksum of because GRE
119 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
120 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100121 *
122 * CHECKSUM_COMPLETE:
123 *
124 * This is the most generic way. The device supplied checksum of the _whole_
125 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
126 * hardware doesn't need to parse L3/L4 headers to implement this.
127 *
128 * Note: Even if device supports only some protocols, but is able to produce
129 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
130 *
131 * CHECKSUM_PARTIAL:
132 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800133 * A checksum is set up to be offloaded to a device as described in the
134 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100135 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800136 * on the same host, or it may be set in the input path in GRO or remote
137 * checksum offload. For the purposes of checksum verification, the checksum
138 * referred to by skb->csum_start + skb->csum_offset and any preceding
139 * checksums in the packet are considered verified. Any checksums in the
140 * packet that are after the checksum being offloaded are not considered to
141 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100142 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800143 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
144 * in the skb->ip_summed for a packet. Values are:
145 *
146 * CHECKSUM_PARTIAL:
147 *
148 * The driver is required to checksum the packet as seen by hard_start_xmit()
149 * from skb->csum_start up to the end, and to record/write the checksum at
150 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
151 * csum_start and csum_offset values are valid values given the length and
152 * offset of the packet, however they should not attempt to validate that the
153 * checksum refers to a legitimate transport layer checksum-- it is the
154 * purview of the stack to validate that csum_start and csum_offset are set
155 * correctly.
156 *
157 * When the stack requests checksum offload for a packet, the driver MUST
158 * ensure that the checksum is set correctly. A driver can either offload the
159 * checksum calculation to the device, or call skb_checksum_help (in the case
160 * that the device does not support offload for a particular checksum).
161 *
162 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
163 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
164 * checksum offload capability. If a device has limited checksum capabilities
165 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
166 * described above) a helper function can be called to resolve
167 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
168 * function takes a spec argument that describes the protocol layer that is
169 * supported for checksum offload and can be called for each packet. If a
170 * packet does not match the specification for offload, skb_checksum_help
171 * is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100172 *
173 * CHECKSUM_NONE:
174 *
175 * The skb was already checksummed by the protocol, or a checksum is not
176 * required.
177 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100178 * CHECKSUM_UNNECESSARY:
179 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800180 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
181 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100182 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800183 * CHECKSUM_COMPLETE:
184 * Not used in checksum output. If a driver observes a packet with this value
185 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100186 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800187 * D. Non-IP checksum (CRC) offloads
188 *
189 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
190 * offloading the SCTP CRC in a packet. To perform this offload the stack
191 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
192 * accordingly. Note the there is no indication in the skbuff that the
193 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
194 * both IP checksum offload and SCTP CRC offload must verify which offload
195 * is configured for a packet presumably by inspecting packet headers.
196 *
197 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
198 * offloading the FCOE CRC in a packet. To perform this offload the stack
199 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
200 * accordingly. Note the there is no indication in the skbuff that the
201 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
202 * both IP checksum offload and FCOE CRC offload must verify which offload
203 * is configured for a packet presumably by inspecting packet headers.
204 *
205 * E. Checksumming on output with GSO.
206 *
207 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
208 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
209 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
210 * part of the GSO operation is implied. If a checksum is being offloaded
211 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
212 * are set to refer to the outermost checksum being offload (two offloaded
213 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100214 */
215
Herbert Xu60476372007-04-09 11:59:39 -0700216/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217#define CHECKSUM_NONE 0
218#define CHECKSUM_UNNECESSARY 1
219#define CHECKSUM_COMPLETE 2
220#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
Tom Herbert77cffe22014-08-27 21:26:46 -0700222/* Maximum value in skb->csum_level */
223#define SKB_MAX_CSUM_LEVEL 3
224
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200225#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700226#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700227 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700228#define SKB_MAX_ORDER(X, ORDER) \
229 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
231#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
232
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000233/* return minimum truesize of one skb containing X bytes of data */
234#define SKB_TRUESIZE(X) ((X) + \
235 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
236 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
237
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700239struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800240struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800241struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800242struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700244#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245struct nf_conntrack {
246 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700248#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000252 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200257 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200261 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000262 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200266 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200267 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200276 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277};
278#endif
279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280struct sk_buff_head {
281 /* These two members must be first. */
282 struct sk_buff *next;
283 struct sk_buff *prev;
284
285 __u32 qlen;
286 spinlock_t lock;
287};
288
289struct sk_buff;
290
Ian Campbell9d4dde52011-12-22 23:39:14 +0000291/* To allow 64K frame to be packed as single skb without frag_list we
292 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
293 * buffers which do not start on a page boundary.
294 *
295 * Since GRO uses frags we allocate at least 16 regardless of page
296 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000297 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000298#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700299#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000300#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000301#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000302#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100303extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300305/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
306 * segment using its current segmentation instead.
307 */
308#define GSO_BY_FRAGS 0xFFFF
309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310typedef struct skb_frag_struct skb_frag_t;
311
312struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000313 struct {
314 struct page *p;
315 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000316#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700317 __u32 page_offset;
318 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000319#else
320 __u16 page_offset;
321 __u16 size;
322#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323};
324
Eric Dumazet9e903e02011-10-18 21:00:24 +0000325static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326{
327 return frag->size;
328}
329
330static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
331{
332 frag->size = size;
333}
334
335static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
336{
337 frag->size += delta;
338}
339
340static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
341{
342 frag->size -= delta;
343}
344
Patrick Ohlyac45f602009-02-12 05:03:37 +0000345#define HAVE_HW_TIME_STAMP
346
347/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800348 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000349 * @hwtstamp: hardware time stamp transformed into duration
350 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000351 *
352 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400353 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000354 *
355 * hwtstamps can only be compared against other hwtstamps from
356 * the same device.
357 *
358 * This structure is attached to packets as part of the
359 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
360 */
361struct skb_shared_hwtstamps {
362 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000363};
364
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000365/* Definitions for tx_flags in struct skb_shared_info */
366enum {
367 /* generate hardware time stamp */
368 SKBTX_HW_TSTAMP = 1 << 0,
369
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400370 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000371 SKBTX_SW_TSTAMP = 1 << 1,
372
373 /* device driver is going to provide hardware time stamp */
374 SKBTX_IN_PROGRESS = 1 << 2,
375
Shirley Maa6686f22011-07-06 12:22:12 +0000376 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000377 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100378
379 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000380 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000381
382 /* This indicates at least one fragment might be overwritten
383 * (as in vmsplice(), sendfile() ...)
384 * If we need to compute a TX checksum, we'll need to copy
385 * all frags to avoid possible bad checksum
386 */
387 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400388
389 /* generate software time stamp when entering packet scheduling */
390 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000391};
392
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400393#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400394 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400395#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
396
Shirley Maa6686f22011-07-06 12:22:12 +0000397/*
398 * The callback notifies userspace to release buffers when skb DMA is done in
399 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000400 * The zerocopy_success argument is true if zero copy transmit occurred,
401 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000402 * The ctx field is used to track device context.
403 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000404 */
405struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000406 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000407 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000408 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000409};
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411/* This data is invariant across clones and lives at
412 * the end of the header data, ie. at skb->end.
413 */
414struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000415 unsigned char nr_frags;
416 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700417 unsigned short gso_size;
418 /* Warning: this field is not always filled in (UFO)! */
419 unsigned short gso_segs;
420 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000422 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400423 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000424 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700425
426 /*
427 * Warning : all fields before dataref are cleared in __alloc_skb()
428 */
429 atomic_t dataref;
430
Johann Baudy69e3c752009-05-18 22:11:22 -0700431 /* Intermediate layers must ensure that destructor_arg
432 * remains valid until skb destructor */
433 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000434
Eric Dumazetfed66382010-07-22 19:09:08 +0000435 /* must be last field, see pskb_expand_head() */
436 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437};
438
439/* We divide dataref into two halves. The higher 16 bits hold references
440 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700441 * the entire skb->data. A clone of a headerless skb holds the length of
442 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 *
444 * All users must obey the rule that the skb->data reference count must be
445 * greater than or equal to the payload reference count.
446 *
447 * Holding a reference to the payload part means that the user does not
448 * care about modifications to the header part of skb->data.
449 */
450#define SKB_DATAREF_SHIFT 16
451#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
452
David S. Millerd179cd12005-08-17 14:57:30 -0700453
454enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700455 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
456 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
457 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700458};
459
Herbert Xu79671682006-06-22 02:40:14 -0700460enum {
461 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700462 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700463
464 /* This indicates the skb is from an untrusted source. */
465 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700466
467 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700468 SKB_GSO_TCP_ECN = 1 << 3,
469
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400470 SKB_GSO_TCP_FIXEDID = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800471
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400472 SKB_GSO_TCPV6 = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000473
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400474 SKB_GSO_FCOE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000475
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400476 SKB_GSO_GRE = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000477
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400478 SKB_GSO_GRE_CSUM = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700479
Tom Herbert7e133182016-05-18 09:06:10 -0700480 SKB_GSO_IPXIP4 = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700481
Tom Herbert7e133182016-05-18 09:06:10 -0700482 SKB_GSO_IPXIP6 = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700483
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400484 SKB_GSO_UDP_TUNNEL = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700485
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400486 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
487
Alexander Duyck802ab552016-04-10 21:45:03 -0400488 SKB_GSO_PARTIAL = 1 << 13,
489
490 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300491
492 SKB_GSO_SCTP = 1 << 15,
Herbert Xu79671682006-06-22 02:40:14 -0700493};
494
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700495#if BITS_PER_LONG > 32
496#define NET_SKBUFF_DATA_USES_OFFSET 1
497#endif
498
499#ifdef NET_SKBUFF_DATA_USES_OFFSET
500typedef unsigned int sk_buff_data_t;
501#else
502typedef unsigned char *sk_buff_data_t;
503#endif
504
Eric Dumazet363ec39232014-02-26 14:02:11 -0800505/**
506 * struct skb_mstamp - multi resolution time stamps
507 * @stamp_us: timestamp in us resolution
508 * @stamp_jiffies: timestamp in jiffies
509 */
510struct skb_mstamp {
511 union {
512 u64 v64;
513 struct {
514 u32 stamp_us;
515 u32 stamp_jiffies;
516 };
517 };
518};
519
520/**
521 * skb_mstamp_get - get current timestamp
522 * @cl: place to store timestamps
523 */
524static inline void skb_mstamp_get(struct skb_mstamp *cl)
525{
526 u64 val = local_clock();
527
528 do_div(val, NSEC_PER_USEC);
529 cl->stamp_us = (u32)val;
530 cl->stamp_jiffies = (u32)jiffies;
531}
532
533/**
534 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
535 * @t1: pointer to newest sample
536 * @t0: pointer to oldest sample
537 */
538static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
539 const struct skb_mstamp *t0)
540{
541 s32 delta_us = t1->stamp_us - t0->stamp_us;
542 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
543
544 /* If delta_us is negative, this might be because interval is too big,
545 * or local_clock() drift is too big : fallback using jiffies.
546 */
547 if (delta_us <= 0 ||
548 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
549
550 delta_us = jiffies_to_usecs(delta_jiffies);
551
552 return delta_us;
553}
554
Yuchung Cheng625a5e12015-10-16 21:57:45 -0700555static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
556 const struct skb_mstamp *t0)
557{
558 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
559
560 if (!diff)
561 diff = t1->stamp_us - t0->stamp_us;
562 return diff > 0;
563}
Eric Dumazet363ec39232014-02-26 14:02:11 -0800564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565/**
566 * struct sk_buff - socket buffer
567 * @next: Next buffer in list
568 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800569 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800570 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700571 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700573 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000574 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700575 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 * @len: Length of actual data
577 * @data_len: Data length
578 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700579 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700580 * @csum: Checksum (must include start/offset pair)
581 * @csum_start: Offset from skb->head where checksumming should start
582 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700583 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700584 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700586 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700588 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700590 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700591 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800592 * @peeked: this packet has been seen already, so stats have been
593 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700594 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700595 * @protocol: Packet protocol from driver
596 * @destructor: Destruct function
597 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800599 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 * @tc_index: Traffic control index
601 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700602 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700603 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700604 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700605 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700606 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700607 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700608 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700609 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100610 * @wifi_acked_valid: wifi_acked was set
611 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000612 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Eliezer Tamir06021292013-06-10 11:39:50 +0300613 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700614 * @secmark: security marking
Scott Feldman0c4f6912015-07-18 18:24:48 -0700615 * @offload_fwd_mark: fwding offload mark
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700616 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000617 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700618 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000619 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000620 * @inner_transport_header: Inner transport layer header (encapsulation)
621 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000622 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700623 * @transport_header: Transport layer header
624 * @network_header: Network layer header
625 * @mac_header: Link layer header
626 * @tail: Tail pointer
627 * @end: End pointer
628 * @head: Head of buffer
629 * @data: Data head pointer
630 * @truesize: Buffer size
631 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 */
633
634struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800635 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800636 struct {
637 /* These two members must be first. */
638 struct sk_buff *next;
639 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000640
Eric Dumazet56b17422014-11-03 08:19:53 -0800641 union {
642 ktime_t tstamp;
643 struct skb_mstamp skb_mstamp;
644 };
645 };
646 struct rb_node rbnode; /* used in netem & tcp stack */
647 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000648 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 /*
652 * This is the control buffer. It is free to use for every
653 * layer. Please put your private variables there. If you
654 * want to keep them across layers you have to do a skb_clone()
655 * first. This is owned by whoever has the skb queued ATM.
656 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000657 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Eric Dumazet7fee2262010-05-11 23:19:48 +0000659 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700660 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000661#ifdef CONFIG_XFRM
662 struct sec_path *sp;
663#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700664#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
665 struct nf_conntrack *nfct;
666#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200667#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700668 struct nf_bridge_info *nf_bridge;
669#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700671 data_len;
672 __u16 mac_len,
673 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700674
675 /* Following fields are _not_ copied in __copy_skb_header()
676 * Note that queue_mapping is here mostly to fill a hole.
677 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200678 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700679 __u16 queue_mapping;
680 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700681 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800682 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800683 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700684 head_frag:1,
685 xmit_more:1;
686 /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200687 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Eric Dumazetb1937222014-09-28 22:18:47 -0700689 /* fields enclosed in headers_start/headers_end are copied
690 * using a single memcpy() in __copy_skb_header()
691 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700692 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700693 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700694 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200695
696/* if you move pkt_type around you also must adapt those constants */
697#ifdef __BIG_ENDIAN_BITFIELD
698#define PKT_TYPE_MAX (7 << 5)
699#else
700#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100701#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200702#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
703
704 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700705 __u8 pkt_type:3;
706 __u8 pfmemalloc:1;
707 __u8 ignore_df:1;
708 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
Eric Dumazetb1937222014-09-28 22:18:47 -0700710 __u8 nf_trace:1;
711 __u8 ip_summed:2;
712 __u8 ooo_okay:1;
713 __u8 l4_hash:1;
714 __u8 sw_hash:1;
715 __u8 wifi_acked_valid:1;
716 __u8 wifi_acked:1;
717
718 __u8 no_fcs:1;
719 /* Indicates the inner headers are valid in the skbuff. */
720 __u8 encapsulation:1;
721 __u8 encap_hdr_csum:1;
722 __u8 csum_valid:1;
723 __u8 csum_complete_sw:1;
724 __u8 csum_level:2;
725 __u8 csum_bad:1;
726
727#ifdef CONFIG_IPV6_NDISC_NODETYPE
728 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700730 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700731 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800732 __u8 remcsum_offload:1;
733 /* 3 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000734
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700736 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700738 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200741
Eric Dumazetb1937222014-09-28 22:18:47 -0700742 union {
743 __wsum csum;
744 struct {
745 __u16 csum_start;
746 __u16 csum_offset;
747 };
748 };
749 __u32 priority;
750 int skb_iif;
751 __u32 hash;
752 __be16 vlan_proto;
753 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800754#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
755 union {
756 unsigned int napi_id;
757 unsigned int sender_cpu;
758 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700759#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700760 union {
James Morris984bc162006-06-09 00:29:17 -0700761#ifdef CONFIG_NETWORK_SECMARK
Scott Feldman0c4f6912015-07-18 18:24:48 -0700762 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700763#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700764#ifdef CONFIG_NET_SWITCHDEV
765 __u32 offload_fwd_mark;
766#endif
767 };
768
Neil Horman3b885782009-10-12 13:26:31 -0700769 union {
770 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000771 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700772 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Tom Herbert8bce6d72014-09-29 20:22:29 -0700774 union {
775 __be16 inner_protocol;
776 __u8 inner_ipproto;
777 };
778
Simon Horman1a37e412013-05-23 21:02:51 +0000779 __u16 inner_transport_header;
780 __u16 inner_network_header;
781 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700782
783 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000784 __u16 transport_header;
785 __u16 network_header;
786 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700787
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700788 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700789 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700790 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700791
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700793 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700794 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700796 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700797 unsigned int truesize;
798 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799};
800
801#ifdef __KERNEL__
802/*
803 * Handling routines are only of interest to the kernel
804 */
805#include <linux/slab.h>
806
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
Mel Gormanc93bdd02012-07-31 16:44:19 -0700808#define SKB_ALLOC_FCLONE 0x01
809#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800810#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700811
812/* Returns true if the skb was allocated from PFMEMALLOC reserves */
813static inline bool skb_pfmemalloc(const struct sk_buff *skb)
814{
815 return unlikely(skb->pfmemalloc);
816}
817
Eric Dumazet7fee2262010-05-11 23:19:48 +0000818/*
819 * skb might have a dst pointer attached, refcounted or not.
820 * _skb_refdst low order bit is set if refcount was _not_ taken
821 */
822#define SKB_DST_NOREF 1UL
823#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
824
825/**
826 * skb_dst - returns skb dst_entry
827 * @skb: buffer
828 *
829 * Returns skb dst_entry, regardless of reference taken or not.
830 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000831static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
832{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000833 /* If refdst was not refcounted, check we still are in a
834 * rcu_read_lock section
835 */
836 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
837 !rcu_read_lock_held() &&
838 !rcu_read_lock_bh_held());
839 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000840}
841
Eric Dumazet7fee2262010-05-11 23:19:48 +0000842/**
843 * skb_dst_set - sets skb dst
844 * @skb: buffer
845 * @dst: dst entry
846 *
847 * Sets skb dst, assuming a reference was taken on dst and should
848 * be released by skb_dst_drop()
849 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000850static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
851{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000852 skb->_skb_refdst = (unsigned long)dst;
853}
854
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200855/**
856 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
857 * @skb: buffer
858 * @dst: dst entry
859 *
860 * Sets skb dst, assuming a reference was not taken on dst.
861 * If dst entry is cached, we do not take reference and dst_release
862 * will be avoided by refdst_drop. If dst entry is not cached, we take
863 * reference, so that last dst_release can destroy the dst immediately.
864 */
865static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
866{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100867 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
868 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200869}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000870
871/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300872 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000873 * @skb: buffer
874 */
875static inline bool skb_dst_is_noref(const struct sk_buff *skb)
876{
877 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000878}
879
Eric Dumazet511c3f92009-06-02 05:14:27 +0000880static inline struct rtable *skb_rtable(const struct sk_buff *skb)
881{
Eric Dumazetadf30902009-06-02 05:19:30 +0000882 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000883}
884
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400885/* For mangling skb->pkt_type from user space side from applications
886 * such as nft, tc, etc, we only allow a conservative subset of
887 * possible pkt_types to be set.
888*/
889static inline bool skb_pkt_type_ok(u32 ptype)
890{
891 return ptype <= PACKET_OTHERHOST;
892}
893
Joe Perches7965bd42013-09-26 14:48:15 -0700894void kfree_skb(struct sk_buff *skb);
895void kfree_skb_list(struct sk_buff *segs);
896void skb_tx_error(struct sk_buff *skb);
897void consume_skb(struct sk_buff *skb);
898void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000899extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000900
Joe Perches7965bd42013-09-26 14:48:15 -0700901void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
902bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
903 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000904
Joe Perches7965bd42013-09-26 14:48:15 -0700905struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
906 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700907struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700908struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700909static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100910 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700911{
Eric Dumazet564824b2010-10-11 19:05:25 +0000912 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700913}
914
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700915struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
916 unsigned long data_len,
917 int max_page_order,
918 int *errcode,
919 gfp_t gfp_mask);
920
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700921/* Layout of fast clones : [skb1][skb2][fclone_ref] */
922struct sk_buff_fclones {
923 struct sk_buff skb1;
924
925 struct sk_buff skb2;
926
927 atomic_t fclone_ref;
928};
929
930/**
931 * skb_fclone_busy - check if fclone is busy
932 * @skb: buffer
933 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900934 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700935 * Some drivers call skb_orphan() in their ndo_start_xmit(),
936 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700937 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700938static inline bool skb_fclone_busy(const struct sock *sk,
939 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700940{
941 const struct sk_buff_fclones *fclones;
942
943 fclones = container_of(skb, struct sk_buff_fclones, skb1);
944
945 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75eb2014-12-03 17:04:39 -0800946 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700947 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700948}
949
David S. Millerd179cd12005-08-17 14:57:30 -0700950static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100951 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700952{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700953 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700954}
955
Joe Perches7965bd42013-09-26 14:48:15 -0700956struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000957static inline struct sk_buff *alloc_skb_head(gfp_t priority)
958{
959 return __alloc_skb_head(priority, -1);
960}
961
Joe Perches7965bd42013-09-26 14:48:15 -0700962struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
963int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
964struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
965struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300966struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
967 gfp_t gfp_mask, bool fclone);
968static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
969 gfp_t gfp_mask)
970{
971 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
972}
Eric Dumazet117632e2011-12-03 21:39:53 +0000973
Joe Perches7965bd42013-09-26 14:48:15 -0700974int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
975struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
976 unsigned int headroom);
977struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
978 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800979int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
980 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700981int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
982 int len);
983int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
984int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000985#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
Joe Perches7965bd42013-09-26 14:48:15 -0700987int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
988 int getfrag(void *from, char *to, int offset,
989 int len, int odd, struct sk_buff *skb),
990 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700991
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200992int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
993 int offset, size_t size);
994
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800995struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700996 __u32 lower_offset;
997 __u32 upper_offset;
998 __u32 frag_idx;
999 __u32 stepped_offset;
1000 struct sk_buff *root_skb;
1001 struct sk_buff *cur_skb;
1002 __u8 *frag_data;
1003};
1004
Joe Perches7965bd42013-09-26 14:48:15 -07001005void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1006 unsigned int to, struct skb_seq_state *st);
1007unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1008 struct skb_seq_state *st);
1009void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001010
Joe Perches7965bd42013-09-26 14:48:15 -07001011unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001012 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001013
Tom Herbert09323cc2013-12-15 22:16:19 -08001014/*
1015 * Packet hash types specify the type of hash in skb_set_hash.
1016 *
1017 * Hash types refer to the protocol layer addresses which are used to
1018 * construct a packet's hash. The hashes are used to differentiate or identify
1019 * flows of the protocol layer for the hash type. Hash types are either
1020 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1021 *
1022 * Properties of hashes:
1023 *
1024 * 1) Two packets in different flows have different hash values
1025 * 2) Two packets in the same flow should have the same hash value
1026 *
1027 * A hash at a higher layer is considered to be more specific. A driver should
1028 * set the most specific hash possible.
1029 *
1030 * A driver cannot indicate a more specific hash than the layer at which a hash
1031 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1032 *
1033 * A driver may indicate a hash level which is less specific than the
1034 * actual layer the hash was computed on. For instance, a hash computed
1035 * at L4 may be considered an L3 hash. This should only be done if the
1036 * driver can't unambiguously determine that the HW computed the hash at
1037 * the higher layer. Note that the "should" in the second property above
1038 * permits this.
1039 */
1040enum pkt_hash_types {
1041 PKT_HASH_TYPE_NONE, /* Undefined type */
1042 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1043 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1044 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1045};
1046
Tom Herbertbcc83832015-09-01 09:24:24 -07001047static inline void skb_clear_hash(struct sk_buff *skb)
1048{
1049 skb->hash = 0;
1050 skb->sw_hash = 0;
1051 skb->l4_hash = 0;
1052}
1053
1054static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1055{
1056 if (!skb->l4_hash)
1057 skb_clear_hash(skb);
1058}
1059
1060static inline void
1061__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1062{
1063 skb->l4_hash = is_l4;
1064 skb->sw_hash = is_sw;
1065 skb->hash = hash;
1066}
1067
Tom Herbert09323cc2013-12-15 22:16:19 -08001068static inline void
1069skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1070{
Tom Herbertbcc83832015-09-01 09:24:24 -07001071 /* Used by drivers to set hash from HW */
1072 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1073}
1074
1075static inline void
1076__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1077{
1078 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001079}
1080
Tom Herberte5276932015-09-01 09:24:23 -07001081void __skb_get_hash(struct sk_buff *skb);
1082u32 skb_get_poff(const struct sk_buff *skb);
1083u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1084 const struct flow_keys *keys, int hlen);
1085__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1086 void *data, int hlen_proto);
1087
1088static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1089 int thoff, u8 ip_proto)
1090{
1091 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1092}
1093
1094void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1095 const struct flow_dissector_key *key,
1096 unsigned int key_count);
1097
1098bool __skb_flow_dissect(const struct sk_buff *skb,
1099 struct flow_dissector *flow_dissector,
1100 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001101 void *data, __be16 proto, int nhoff, int hlen,
1102 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001103
1104static inline bool skb_flow_dissect(const struct sk_buff *skb,
1105 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001106 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001107{
1108 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001109 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001110}
1111
1112static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001113 struct flow_keys *flow,
1114 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001115{
1116 memset(flow, 0, sizeof(*flow));
1117 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001118 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001119}
1120
1121static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1122 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001123 int nhoff, int hlen,
1124 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001125{
1126 memset(flow, 0, sizeof(*flow));
1127 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001128 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001129}
1130
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001131static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001132{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001133 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001134 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001135
Tom Herbert61b905d2014-03-24 15:34:47 -07001136 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001137}
1138
David S. Miller20a17bf2015-09-01 21:19:17 -07001139__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001140
David S. Miller20a17bf2015-09-01 21:19:17 -07001141static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001142{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001143 if (!skb->l4_hash && !skb->sw_hash) {
1144 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001145 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001146
Tom Herbertde4c1f82015-09-01 18:11:04 -07001147 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001148 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001149
1150 return skb->hash;
1151}
1152
David S. Miller20a17bf2015-09-01 21:19:17 -07001153__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001154
David S. Miller20a17bf2015-09-01 21:19:17 -07001155static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001156{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001157 if (!skb->l4_hash && !skb->sw_hash) {
1158 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001159 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001160
Tom Herbertde4c1f82015-09-01 18:11:04 -07001161 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001162 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001163
1164 return skb->hash;
1165}
1166
Tom Herbert50fb7992015-05-01 11:30:12 -07001167__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1168
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001169static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1170{
Tom Herbert61b905d2014-03-24 15:34:47 -07001171 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001172}
1173
Tom Herbert3df7a742013-12-15 22:16:29 -08001174static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1175{
Tom Herbert61b905d2014-03-24 15:34:47 -07001176 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001177 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001178 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001179};
1180
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001181#ifdef NET_SKBUFF_DATA_USES_OFFSET
1182static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1183{
1184 return skb->head + skb->end;
1185}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001186
1187static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1188{
1189 return skb->end;
1190}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001191#else
1192static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1193{
1194 return skb->end;
1195}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001196
1197static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1198{
1199 return skb->end - skb->head;
1200}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001201#endif
1202
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001204#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205
Patrick Ohlyac45f602009-02-12 05:03:37 +00001206static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1207{
1208 return &skb_shinfo(skb)->hwtstamps;
1209}
1210
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211/**
1212 * skb_queue_empty - check if a queue is empty
1213 * @list: queue head
1214 *
1215 * Returns true if the queue is empty, false otherwise.
1216 */
1217static inline int skb_queue_empty(const struct sk_buff_head *list)
1218{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001219 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220}
1221
1222/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001223 * skb_queue_is_last - check if skb is the last entry in the queue
1224 * @list: queue head
1225 * @skb: buffer
1226 *
1227 * Returns true if @skb is the last buffer on the list.
1228 */
1229static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1230 const struct sk_buff *skb)
1231{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001232 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001233}
1234
1235/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001236 * skb_queue_is_first - check if skb is the first entry in the queue
1237 * @list: queue head
1238 * @skb: buffer
1239 *
1240 * Returns true if @skb is the first buffer on the list.
1241 */
1242static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1243 const struct sk_buff *skb)
1244{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001245 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001246}
1247
1248/**
David S. Miller249c8b42008-09-23 00:44:42 -07001249 * skb_queue_next - return the next packet in the queue
1250 * @list: queue head
1251 * @skb: current buffer
1252 *
1253 * Return the next packet in @list after @skb. It is only valid to
1254 * call this if skb_queue_is_last() evaluates to false.
1255 */
1256static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1257 const struct sk_buff *skb)
1258{
1259 /* This BUG_ON may seem severe, but if we just return then we
1260 * are going to dereference garbage.
1261 */
1262 BUG_ON(skb_queue_is_last(list, skb));
1263 return skb->next;
1264}
1265
1266/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001267 * skb_queue_prev - return the prev packet in the queue
1268 * @list: queue head
1269 * @skb: current buffer
1270 *
1271 * Return the prev packet in @list before @skb. It is only valid to
1272 * call this if skb_queue_is_first() evaluates to false.
1273 */
1274static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1275 const struct sk_buff *skb)
1276{
1277 /* This BUG_ON may seem severe, but if we just return then we
1278 * are going to dereference garbage.
1279 */
1280 BUG_ON(skb_queue_is_first(list, skb));
1281 return skb->prev;
1282}
1283
1284/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 * skb_get - reference buffer
1286 * @skb: buffer to reference
1287 *
1288 * Makes another reference to a socket buffer and returns a pointer
1289 * to the buffer.
1290 */
1291static inline struct sk_buff *skb_get(struct sk_buff *skb)
1292{
1293 atomic_inc(&skb->users);
1294 return skb;
1295}
1296
1297/*
1298 * If users == 1, we are the only owner and are can avoid redundant
1299 * atomic change.
1300 */
1301
1302/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 * skb_cloned - is the buffer a clone
1304 * @skb: buffer to check
1305 *
1306 * Returns true if the buffer was generated with skb_clone() and is
1307 * one of multiple shared copies of the buffer. Cloned buffers are
1308 * shared data so must not be written to under normal circumstances.
1309 */
1310static inline int skb_cloned(const struct sk_buff *skb)
1311{
1312 return skb->cloned &&
1313 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1314}
1315
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001316static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1317{
Mel Gormand0164ad2015-11-06 16:28:21 -08001318 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001319
1320 if (skb_cloned(skb))
1321 return pskb_expand_head(skb, 0, 0, pri);
1322
1323 return 0;
1324}
1325
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326/**
1327 * skb_header_cloned - is the header a clone
1328 * @skb: buffer to check
1329 *
1330 * Returns true if modifying the header part of the buffer requires
1331 * the data to be copied.
1332 */
1333static inline int skb_header_cloned(const struct sk_buff *skb)
1334{
1335 int dataref;
1336
1337 if (!skb->cloned)
1338 return 0;
1339
1340 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1341 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1342 return dataref != 1;
1343}
1344
Eric Dumazet9580bf22016-04-30 10:19:29 -07001345static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1346{
1347 might_sleep_if(gfpflags_allow_blocking(pri));
1348
1349 if (skb_header_cloned(skb))
1350 return pskb_expand_head(skb, 0, 0, pri);
1351
1352 return 0;
1353}
1354
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355/**
1356 * skb_header_release - release reference to header
1357 * @skb: buffer to operate on
1358 *
1359 * Drop a reference to the header part of the buffer. This is done
1360 * by acquiring a payload reference. You must not read from the header
1361 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001362 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 */
1364static inline void skb_header_release(struct sk_buff *skb)
1365{
1366 BUG_ON(skb->nohdr);
1367 skb->nohdr = 1;
1368 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1369}
1370
1371/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001372 * __skb_header_release - release reference to header
1373 * @skb: buffer to operate on
1374 *
1375 * Variant of skb_header_release() assuming skb is private to caller.
1376 * We can avoid one atomic operation.
1377 */
1378static inline void __skb_header_release(struct sk_buff *skb)
1379{
1380 skb->nohdr = 1;
1381 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1382}
1383
1384
1385/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 * skb_shared - is the buffer shared
1387 * @skb: buffer to check
1388 *
1389 * Returns true if more than one person has a reference to this
1390 * buffer.
1391 */
1392static inline int skb_shared(const struct sk_buff *skb)
1393{
1394 return atomic_read(&skb->users) != 1;
1395}
1396
1397/**
1398 * skb_share_check - check if buffer is shared and if so clone it
1399 * @skb: buffer to check
1400 * @pri: priority for memory allocation
1401 *
1402 * If the buffer is shared the buffer is cloned and the old copy
1403 * drops a reference. A new clone with a single reference is returned.
1404 * If the buffer is not shared the original buffer is returned. When
1405 * being called from interrupt status or with spinlocks held pri must
1406 * be GFP_ATOMIC.
1407 *
1408 * NULL is returned on a memory allocation failure.
1409 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001410static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411{
Mel Gormand0164ad2015-11-06 16:28:21 -08001412 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 if (skb_shared(skb)) {
1414 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001415
1416 if (likely(nskb))
1417 consume_skb(skb);
1418 else
1419 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 skb = nskb;
1421 }
1422 return skb;
1423}
1424
1425/*
1426 * Copy shared buffers into a new sk_buff. We effectively do COW on
1427 * packets to handle cases where we have a local reader and forward
1428 * and a couple of other messy ones. The normal one is tcpdumping
1429 * a packet thats being forwarded.
1430 */
1431
1432/**
1433 * skb_unshare - make a copy of a shared buffer
1434 * @skb: buffer to check
1435 * @pri: priority for memory allocation
1436 *
1437 * If the socket buffer is a clone then this function creates a new
1438 * copy of the data, drops a reference count on the old copy and returns
1439 * the new copy with the reference count at 1. If the buffer is not a clone
1440 * the original buffer is returned. When called with a spinlock held or
1441 * from interrupt state @pri must be %GFP_ATOMIC
1442 *
1443 * %NULL is returned on a memory allocation failure.
1444 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001445static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001446 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447{
Mel Gormand0164ad2015-11-06 16:28:21 -08001448 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 if (skb_cloned(skb)) {
1450 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001451
1452 /* Free our shared copy */
1453 if (likely(nskb))
1454 consume_skb(skb);
1455 else
1456 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 skb = nskb;
1458 }
1459 return skb;
1460}
1461
1462/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001463 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 * @list_: list to peek at
1465 *
1466 * Peek an &sk_buff. Unlike most other operations you _MUST_
1467 * be careful with this one. A peek leaves the buffer on the
1468 * list and someone else may run off with it. You must hold
1469 * the appropriate locks or have a private queue to do this.
1470 *
1471 * Returns %NULL for an empty list or a pointer to the head element.
1472 * The reference count is not incremented and the reference is therefore
1473 * volatile. Use with caution.
1474 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001475static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476{
Eric Dumazet18d07002012-04-30 16:31:46 +00001477 struct sk_buff *skb = list_->next;
1478
1479 if (skb == (struct sk_buff *)list_)
1480 skb = NULL;
1481 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482}
1483
1484/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001485 * skb_peek_next - peek skb following the given one from a queue
1486 * @skb: skb to start from
1487 * @list_: list to peek at
1488 *
1489 * Returns %NULL when the end of the list is met or a pointer to the
1490 * next element. The reference count is not incremented and the
1491 * reference is therefore volatile. Use with caution.
1492 */
1493static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1494 const struct sk_buff_head *list_)
1495{
1496 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001497
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001498 if (next == (struct sk_buff *)list_)
1499 next = NULL;
1500 return next;
1501}
1502
1503/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001504 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 * @list_: list to peek at
1506 *
1507 * Peek an &sk_buff. Unlike most other operations you _MUST_
1508 * be careful with this one. A peek leaves the buffer on the
1509 * list and someone else may run off with it. You must hold
1510 * the appropriate locks or have a private queue to do this.
1511 *
1512 * Returns %NULL for an empty list or a pointer to the tail element.
1513 * The reference count is not incremented and the reference is therefore
1514 * volatile. Use with caution.
1515 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001516static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517{
Eric Dumazet18d07002012-04-30 16:31:46 +00001518 struct sk_buff *skb = list_->prev;
1519
1520 if (skb == (struct sk_buff *)list_)
1521 skb = NULL;
1522 return skb;
1523
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524}
1525
1526/**
1527 * skb_queue_len - get queue length
1528 * @list_: list to measure
1529 *
1530 * Return the length of an &sk_buff queue.
1531 */
1532static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1533{
1534 return list_->qlen;
1535}
1536
David S. Miller67fed452008-09-21 22:36:24 -07001537/**
1538 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1539 * @list: queue to initialize
1540 *
1541 * This initializes only the list and queue length aspects of
1542 * an sk_buff_head object. This allows to initialize the list
1543 * aspects of an sk_buff_head without reinitializing things like
1544 * the spinlock. It can also be used for on-stack sk_buff_head
1545 * objects where the spinlock is known to not be used.
1546 */
1547static inline void __skb_queue_head_init(struct sk_buff_head *list)
1548{
1549 list->prev = list->next = (struct sk_buff *)list;
1550 list->qlen = 0;
1551}
1552
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001553/*
1554 * This function creates a split out lock class for each invocation;
1555 * this is needed for now since a whole lot of users of the skb-queue
1556 * infrastructure in drivers have different locking usage (in hardirq)
1557 * than the networking core (in softirq only). In the long run either the
1558 * network layer or drivers should need annotation to consolidate the
1559 * main types of usage into 3 classes.
1560 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561static inline void skb_queue_head_init(struct sk_buff_head *list)
1562{
1563 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001564 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565}
1566
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001567static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1568 struct lock_class_key *class)
1569{
1570 skb_queue_head_init(list);
1571 lockdep_set_class(&list->lock, class);
1572}
1573
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001575 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 *
1577 * The "__skb_xxxx()" functions are the non-atomic ones that
1578 * can only be called with interrupts disabled.
1579 */
Joe Perches7965bd42013-09-26 14:48:15 -07001580void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1581 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001582static inline void __skb_insert(struct sk_buff *newsk,
1583 struct sk_buff *prev, struct sk_buff *next,
1584 struct sk_buff_head *list)
1585{
1586 newsk->next = next;
1587 newsk->prev = prev;
1588 next->prev = prev->next = newsk;
1589 list->qlen++;
1590}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591
David S. Miller67fed452008-09-21 22:36:24 -07001592static inline void __skb_queue_splice(const struct sk_buff_head *list,
1593 struct sk_buff *prev,
1594 struct sk_buff *next)
1595{
1596 struct sk_buff *first = list->next;
1597 struct sk_buff *last = list->prev;
1598
1599 first->prev = prev;
1600 prev->next = first;
1601
1602 last->next = next;
1603 next->prev = last;
1604}
1605
1606/**
1607 * skb_queue_splice - join two skb lists, this is designed for stacks
1608 * @list: the new list to add
1609 * @head: the place to add it in the first list
1610 */
1611static inline void skb_queue_splice(const struct sk_buff_head *list,
1612 struct sk_buff_head *head)
1613{
1614 if (!skb_queue_empty(list)) {
1615 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001616 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001617 }
1618}
1619
1620/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001621 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001622 * @list: the new list to add
1623 * @head: the place to add it in the first list
1624 *
1625 * The list at @list is reinitialised
1626 */
1627static inline void skb_queue_splice_init(struct sk_buff_head *list,
1628 struct sk_buff_head *head)
1629{
1630 if (!skb_queue_empty(list)) {
1631 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001632 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001633 __skb_queue_head_init(list);
1634 }
1635}
1636
1637/**
1638 * skb_queue_splice_tail - join two skb lists, each list being a queue
1639 * @list: the new list to add
1640 * @head: the place to add it in the first list
1641 */
1642static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1643 struct sk_buff_head *head)
1644{
1645 if (!skb_queue_empty(list)) {
1646 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001647 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001648 }
1649}
1650
1651/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001652 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001653 * @list: the new list to add
1654 * @head: the place to add it in the first list
1655 *
1656 * Each of the lists is a queue.
1657 * The list at @list is reinitialised
1658 */
1659static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1660 struct sk_buff_head *head)
1661{
1662 if (!skb_queue_empty(list)) {
1663 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001664 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001665 __skb_queue_head_init(list);
1666 }
1667}
1668
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001670 * __skb_queue_after - queue a buffer at the list head
1671 * @list: list to use
1672 * @prev: place after this buffer
1673 * @newsk: buffer to queue
1674 *
1675 * Queue a buffer int the middle of a list. This function takes no locks
1676 * and you must therefore hold required locks before calling it.
1677 *
1678 * A buffer cannot be placed on two lists at the same time.
1679 */
1680static inline void __skb_queue_after(struct sk_buff_head *list,
1681 struct sk_buff *prev,
1682 struct sk_buff *newsk)
1683{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001684 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001685}
1686
Joe Perches7965bd42013-09-26 14:48:15 -07001687void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1688 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001689
Gerrit Renkerf5572852008-04-14 00:05:28 -07001690static inline void __skb_queue_before(struct sk_buff_head *list,
1691 struct sk_buff *next,
1692 struct sk_buff *newsk)
1693{
1694 __skb_insert(newsk, next->prev, next, list);
1695}
1696
Stephen Hemminger300ce172005-10-30 13:47:34 -08001697/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 * __skb_queue_head - queue a buffer at the list head
1699 * @list: list to use
1700 * @newsk: buffer to queue
1701 *
1702 * Queue a buffer at the start of a list. This function takes no locks
1703 * and you must therefore hold required locks before calling it.
1704 *
1705 * A buffer cannot be placed on two lists at the same time.
1706 */
Joe Perches7965bd42013-09-26 14:48:15 -07001707void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708static inline void __skb_queue_head(struct sk_buff_head *list,
1709 struct sk_buff *newsk)
1710{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001711 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712}
1713
1714/**
1715 * __skb_queue_tail - queue a buffer at the list tail
1716 * @list: list to use
1717 * @newsk: buffer to queue
1718 *
1719 * Queue a buffer at the end of a list. This function takes no locks
1720 * and you must therefore hold required locks before calling it.
1721 *
1722 * A buffer cannot be placed on two lists at the same time.
1723 */
Joe Perches7965bd42013-09-26 14:48:15 -07001724void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725static inline void __skb_queue_tail(struct sk_buff_head *list,
1726 struct sk_buff *newsk)
1727{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001728 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729}
1730
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 * remove sk_buff from list. _Must_ be called atomically, and with
1733 * the list known..
1734 */
Joe Perches7965bd42013-09-26 14:48:15 -07001735void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1737{
1738 struct sk_buff *next, *prev;
1739
1740 list->qlen--;
1741 next = skb->next;
1742 prev = skb->prev;
1743 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 next->prev = prev;
1745 prev->next = next;
1746}
1747
Gerrit Renkerf525c062008-04-14 00:04:12 -07001748/**
1749 * __skb_dequeue - remove from the head of the queue
1750 * @list: list to dequeue from
1751 *
1752 * Remove the head of the list. This function does not take any locks
1753 * so must be used with appropriate locks held only. The head item is
1754 * returned or %NULL if the list is empty.
1755 */
Joe Perches7965bd42013-09-26 14:48:15 -07001756struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001757static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1758{
1759 struct sk_buff *skb = skb_peek(list);
1760 if (skb)
1761 __skb_unlink(skb, list);
1762 return skb;
1763}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764
1765/**
1766 * __skb_dequeue_tail - remove from the tail of the queue
1767 * @list: list to dequeue from
1768 *
1769 * Remove the tail of the list. This function does not take any locks
1770 * so must be used with appropriate locks held only. The tail item is
1771 * returned or %NULL if the list is empty.
1772 */
Joe Perches7965bd42013-09-26 14:48:15 -07001773struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1775{
1776 struct sk_buff *skb = skb_peek_tail(list);
1777 if (skb)
1778 __skb_unlink(skb, list);
1779 return skb;
1780}
1781
1782
David S. Millerbdcc0922012-03-07 20:53:36 -05001783static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784{
1785 return skb->data_len;
1786}
1787
1788static inline unsigned int skb_headlen(const struct sk_buff *skb)
1789{
1790 return skb->len - skb->data_len;
1791}
1792
1793static inline int skb_pagelen(const struct sk_buff *skb)
1794{
1795 int i, len = 0;
1796
1797 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001798 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 return len + skb_headlen(skb);
1800}
1801
Ian Campbell131ea662011-08-19 06:25:00 +00001802/**
1803 * __skb_fill_page_desc - initialise a paged fragment in an skb
1804 * @skb: buffer containing fragment to be initialised
1805 * @i: paged fragment index to initialise
1806 * @page: the page to use for this fragment
1807 * @off: the offset to the data with @page
1808 * @size: the length of the data
1809 *
1810 * Initialises the @i'th fragment of @skb to point to &size bytes at
1811 * offset @off within @page.
1812 *
1813 * Does not take any additional reference on the fragment.
1814 */
1815static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1816 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817{
1818 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1819
Mel Gormanc48a11c2012-07-31 16:44:23 -07001820 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001821 * Propagate page pfmemalloc to the skb if we can. The problem is
1822 * that not all callers have unique ownership of the page but rely
1823 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001824 */
Ian Campbella8605c62011-10-19 23:01:49 +00001825 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001827 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001828
1829 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001830 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001831 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001832}
1833
1834/**
1835 * skb_fill_page_desc - initialise a paged fragment in an skb
1836 * @skb: buffer containing fragment to be initialised
1837 * @i: paged fragment index to initialise
1838 * @page: the page to use for this fragment
1839 * @off: the offset to the data with @page
1840 * @size: the length of the data
1841 *
1842 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001843 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001844 * addition updates @skb such that @i is the last fragment.
1845 *
1846 * Does not take any additional reference on the fragment.
1847 */
1848static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1849 struct page *page, int off, int size)
1850{
1851 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 skb_shinfo(skb)->nr_frags = i + 1;
1853}
1854
Joe Perches7965bd42013-09-26 14:48:15 -07001855void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1856 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001857
Jason Wangf8e617e2013-11-01 14:07:47 +08001858void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1859 unsigned int truesize);
1860
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001862#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1864
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001865#ifdef NET_SKBUFF_DATA_USES_OFFSET
1866static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1867{
1868 return skb->head + skb->tail;
1869}
1870
1871static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1872{
1873 skb->tail = skb->data - skb->head;
1874}
1875
1876static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1877{
1878 skb_reset_tail_pointer(skb);
1879 skb->tail += offset;
1880}
Simon Horman7cc46192013-05-28 20:34:29 +00001881
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001882#else /* NET_SKBUFF_DATA_USES_OFFSET */
1883static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1884{
1885 return skb->tail;
1886}
1887
1888static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1889{
1890 skb->tail = skb->data;
1891}
1892
1893static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1894{
1895 skb->tail = skb->data + offset;
1896}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001897
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001898#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1899
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900/*
1901 * Add data to an sk_buff
1902 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001903unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001904unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1906{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001907 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908 SKB_LINEAR_ASSERT(skb);
1909 skb->tail += len;
1910 skb->len += len;
1911 return tmp;
1912}
1913
Joe Perches7965bd42013-09-26 14:48:15 -07001914unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1916{
1917 skb->data -= len;
1918 skb->len += len;
1919 return skb->data;
1920}
1921
Joe Perches7965bd42013-09-26 14:48:15 -07001922unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1924{
1925 skb->len -= len;
1926 BUG_ON(skb->len < skb->data_len);
1927 return skb->data += len;
1928}
1929
David S. Miller47d29642010-05-02 02:21:44 -07001930static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1931{
1932 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1933}
1934
Joe Perches7965bd42013-09-26 14:48:15 -07001935unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
1937static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1938{
1939 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001940 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 return NULL;
1942 skb->len -= len;
1943 return skb->data += len;
1944}
1945
1946static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1947{
1948 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1949}
1950
1951static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1952{
1953 if (likely(len <= skb_headlen(skb)))
1954 return 1;
1955 if (unlikely(len > skb->len))
1956 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001957 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958}
1959
1960/**
1961 * skb_headroom - bytes at buffer head
1962 * @skb: buffer to check
1963 *
1964 * Return the number of bytes of free space at the head of an &sk_buff.
1965 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001966static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967{
1968 return skb->data - skb->head;
1969}
1970
1971/**
1972 * skb_tailroom - bytes at buffer end
1973 * @skb: buffer to check
1974 *
1975 * Return the number of bytes of free space at the tail of an sk_buff
1976 */
1977static inline int skb_tailroom(const struct sk_buff *skb)
1978{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001979 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980}
1981
1982/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001983 * skb_availroom - bytes at buffer end
1984 * @skb: buffer to check
1985 *
1986 * Return the number of bytes of free space at the tail of an sk_buff
1987 * allocated by sk_stream_alloc()
1988 */
1989static inline int skb_availroom(const struct sk_buff *skb)
1990{
Eric Dumazet16fad692013-03-14 05:40:32 +00001991 if (skb_is_nonlinear(skb))
1992 return 0;
1993
1994 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001995}
1996
1997/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 * skb_reserve - adjust headroom
1999 * @skb: buffer to alter
2000 * @len: bytes to move
2001 *
2002 * Increase the headroom of an empty &sk_buff by reducing the tail
2003 * room. This is only allowed for an empty buffer.
2004 */
David S. Miller8243126c2006-01-17 02:54:21 -08002005static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006{
2007 skb->data += len;
2008 skb->tail += len;
2009}
2010
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002011/**
2012 * skb_tailroom_reserve - adjust reserved_tailroom
2013 * @skb: buffer to alter
2014 * @mtu: maximum amount of headlen permitted
2015 * @needed_tailroom: minimum amount of reserved_tailroom
2016 *
2017 * Set reserved_tailroom so that headlen can be as large as possible but
2018 * not larger than mtu and tailroom cannot be smaller than
2019 * needed_tailroom.
2020 * The required headroom should already have been reserved before using
2021 * this function.
2022 */
2023static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2024 unsigned int needed_tailroom)
2025{
2026 SKB_LINEAR_ASSERT(skb);
2027 if (mtu < skb_tailroom(skb) - needed_tailroom)
2028 /* use at most mtu */
2029 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2030 else
2031 /* use up to all available space */
2032 skb->reserved_tailroom = needed_tailroom;
2033}
2034
Tom Herbert8bce6d72014-09-29 20:22:29 -07002035#define ENCAP_TYPE_ETHER 0
2036#define ENCAP_TYPE_IPPROTO 1
2037
2038static inline void skb_set_inner_protocol(struct sk_buff *skb,
2039 __be16 protocol)
2040{
2041 skb->inner_protocol = protocol;
2042 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2043}
2044
2045static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2046 __u8 ipproto)
2047{
2048 skb->inner_ipproto = ipproto;
2049 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2050}
2051
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002052static inline void skb_reset_inner_headers(struct sk_buff *skb)
2053{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002054 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002055 skb->inner_network_header = skb->network_header;
2056 skb->inner_transport_header = skb->transport_header;
2057}
2058
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002059static inline void skb_reset_mac_len(struct sk_buff *skb)
2060{
2061 skb->mac_len = skb->network_header - skb->mac_header;
2062}
2063
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002064static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2065 *skb)
2066{
2067 return skb->head + skb->inner_transport_header;
2068}
2069
Tom Herbert55dc5a92015-12-14 11:19:40 -08002070static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2071{
2072 return skb_inner_transport_header(skb) - skb->data;
2073}
2074
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002075static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2076{
2077 skb->inner_transport_header = skb->data - skb->head;
2078}
2079
2080static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2081 const int offset)
2082{
2083 skb_reset_inner_transport_header(skb);
2084 skb->inner_transport_header += offset;
2085}
2086
2087static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2088{
2089 return skb->head + skb->inner_network_header;
2090}
2091
2092static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2093{
2094 skb->inner_network_header = skb->data - skb->head;
2095}
2096
2097static inline void skb_set_inner_network_header(struct sk_buff *skb,
2098 const int offset)
2099{
2100 skb_reset_inner_network_header(skb);
2101 skb->inner_network_header += offset;
2102}
2103
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002104static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2105{
2106 return skb->head + skb->inner_mac_header;
2107}
2108
2109static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2110{
2111 skb->inner_mac_header = skb->data - skb->head;
2112}
2113
2114static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2115 const int offset)
2116{
2117 skb_reset_inner_mac_header(skb);
2118 skb->inner_mac_header += offset;
2119}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002120static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2121{
Cong Wang35d04612013-05-29 15:16:05 +08002122 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002123}
2124
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002125static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2126{
2127 return skb->head + skb->transport_header;
2128}
2129
2130static inline void skb_reset_transport_header(struct sk_buff *skb)
2131{
2132 skb->transport_header = skb->data - skb->head;
2133}
2134
2135static inline void skb_set_transport_header(struct sk_buff *skb,
2136 const int offset)
2137{
2138 skb_reset_transport_header(skb);
2139 skb->transport_header += offset;
2140}
2141
2142static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2143{
2144 return skb->head + skb->network_header;
2145}
2146
2147static inline void skb_reset_network_header(struct sk_buff *skb)
2148{
2149 skb->network_header = skb->data - skb->head;
2150}
2151
2152static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2153{
2154 skb_reset_network_header(skb);
2155 skb->network_header += offset;
2156}
2157
2158static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2159{
2160 return skb->head + skb->mac_header;
2161}
2162
2163static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2164{
Cong Wang35d04612013-05-29 15:16:05 +08002165 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002166}
2167
2168static inline void skb_reset_mac_header(struct sk_buff *skb)
2169{
2170 skb->mac_header = skb->data - skb->head;
2171}
2172
2173static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2174{
2175 skb_reset_mac_header(skb);
2176 skb->mac_header += offset;
2177}
2178
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002179static inline void skb_pop_mac_header(struct sk_buff *skb)
2180{
2181 skb->mac_header = skb->network_header;
2182}
2183
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002184static inline void skb_probe_transport_header(struct sk_buff *skb,
2185 const int offset_hint)
2186{
2187 struct flow_keys keys;
2188
2189 if (skb_transport_header_was_set(skb))
2190 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002191 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002192 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002193 else
2194 skb_set_transport_header(skb, offset_hint);
2195}
2196
Eric Dumazet03606892012-02-23 10:55:02 +00002197static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2198{
2199 if (skb_mac_header_was_set(skb)) {
2200 const unsigned char *old_mac = skb_mac_header(skb);
2201
2202 skb_set_mac_header(skb, -skb->mac_len);
2203 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2204 }
2205}
2206
Michał Mirosław04fb4512010-12-14 15:24:08 +00002207static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2208{
2209 return skb->csum_start - skb_headroom(skb);
2210}
2211
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002212static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2213{
2214 return skb->head + skb->csum_start;
2215}
2216
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002217static inline int skb_transport_offset(const struct sk_buff *skb)
2218{
2219 return skb_transport_header(skb) - skb->data;
2220}
2221
2222static inline u32 skb_network_header_len(const struct sk_buff *skb)
2223{
2224 return skb->transport_header - skb->network_header;
2225}
2226
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002227static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2228{
2229 return skb->inner_transport_header - skb->inner_network_header;
2230}
2231
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002232static inline int skb_network_offset(const struct sk_buff *skb)
2233{
2234 return skb_network_header(skb) - skb->data;
2235}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002236
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002237static inline int skb_inner_network_offset(const struct sk_buff *skb)
2238{
2239 return skb_inner_network_header(skb) - skb->data;
2240}
2241
Changli Gaof9599ce2010-08-04 04:43:44 +00002242static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2243{
2244 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2245}
2246
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247/*
2248 * CPUs often take a performance hit when accessing unaligned memory
2249 * locations. The actual performance hit varies, it can be small if the
2250 * hardware handles it or large if we have to take an exception and fix it
2251 * in software.
2252 *
2253 * Since an ethernet header is 14 bytes network drivers often end up with
2254 * the IP header at an unaligned offset. The IP header can be aligned by
2255 * shifting the start of the packet by 2 bytes. Drivers should do this
2256 * with:
2257 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002258 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 *
2260 * The downside to this alignment of the IP header is that the DMA is now
2261 * unaligned. On some architectures the cost of an unaligned DMA is high
2262 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002263 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002264 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2265 * to be overridden.
2266 */
2267#ifndef NET_IP_ALIGN
2268#define NET_IP_ALIGN 2
2269#endif
2270
Anton Blanchard025be812006-03-31 02:27:06 -08002271/*
2272 * The networking layer reserves some headroom in skb data (via
2273 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2274 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002275 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002276 *
2277 * Unfortunately this headroom changes the DMA alignment of the resulting
2278 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2279 * on some architectures. An architecture can override this value,
2280 * perhaps setting it to a cacheline in size (since that will maintain
2281 * cacheline alignment of the DMA). It must be a power of 2.
2282 *
David S. Millerd6301d32009-02-08 19:24:13 -08002283 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002284 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002285 *
2286 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2287 * to reduce average number of cache lines per packet.
2288 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002289 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002290 */
2291#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002292#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002293#endif
2294
Joe Perches7965bd42013-09-26 14:48:15 -07002295int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
2297static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2298{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002299 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002300 WARN_ON(1);
2301 return;
2302 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002303 skb->len = len;
2304 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305}
2306
Joe Perches7965bd42013-09-26 14:48:15 -07002307void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308
2309static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2310{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002311 if (skb->data_len)
2312 return ___pskb_trim(skb, len);
2313 __skb_trim(skb, len);
2314 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315}
2316
2317static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2318{
2319 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2320}
2321
2322/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002323 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2324 * @skb: buffer to alter
2325 * @len: new length
2326 *
2327 * This is identical to pskb_trim except that the caller knows that
2328 * the skb is not cloned so we should never get an error due to out-
2329 * of-memory.
2330 */
2331static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2332{
2333 int err = pskb_trim(skb, len);
2334 BUG_ON(err);
2335}
2336
2337/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 * skb_orphan - orphan a buffer
2339 * @skb: buffer to orphan
2340 *
2341 * If a buffer currently has an owner then we call the owner's
2342 * destructor function and make the @skb unowned. The buffer continues
2343 * to exist but is no longer charged to its former owner.
2344 */
2345static inline void skb_orphan(struct sk_buff *skb)
2346{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002347 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002349 skb->destructor = NULL;
2350 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002351 } else {
2352 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002353 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354}
2355
2356/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002357 * skb_orphan_frags - orphan the frags contained in a buffer
2358 * @skb: buffer to orphan frags from
2359 * @gfp_mask: allocation mask for replacement pages
2360 *
2361 * For each frag in the SKB which needs a destructor (i.e. has an
2362 * owner) create a copy of that frag and release the original
2363 * page by calling the destructor.
2364 */
2365static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2366{
2367 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2368 return 0;
2369 return skb_copy_ubufs(skb, gfp_mask);
2370}
2371
2372/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373 * __skb_queue_purge - empty a list
2374 * @list: list to empty
2375 *
2376 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2377 * the list and one reference dropped. This function does not take the
2378 * list lock and the caller must hold the relevant locks to use it.
2379 */
Joe Perches7965bd42013-09-26 14:48:15 -07002380void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381static inline void __skb_queue_purge(struct sk_buff_head *list)
2382{
2383 struct sk_buff *skb;
2384 while ((skb = __skb_dequeue(list)) != NULL)
2385 kfree_skb(skb);
2386}
2387
Joe Perches7965bd42013-09-26 14:48:15 -07002388void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389
Joe Perches7965bd42013-09-26 14:48:15 -07002390struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2391 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002392
2393/**
2394 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2395 * @dev: network device to receive on
2396 * @length: length to allocate
2397 *
2398 * Allocate a new &sk_buff and assign it a usage count of one. The
2399 * buffer has unspecified headroom built in. Users should allocate
2400 * the headroom they think they need without accounting for the
2401 * built in space. The built in space is used for optimisations.
2402 *
2403 * %NULL is returned if there is no free memory. Although this function
2404 * allocates memory it can be called from an interrupt.
2405 */
2406static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002407 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002408{
2409 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2410}
2411
Eric Dumazet6f532612012-05-18 05:12:12 +00002412/* legacy helper around __netdev_alloc_skb() */
2413static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2414 gfp_t gfp_mask)
2415{
2416 return __netdev_alloc_skb(NULL, length, gfp_mask);
2417}
2418
2419/* legacy helper around netdev_alloc_skb() */
2420static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2421{
2422 return netdev_alloc_skb(NULL, length);
2423}
2424
2425
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002426static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2427 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002428{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002429 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002430
2431 if (NET_IP_ALIGN && skb)
2432 skb_reserve(skb, NET_IP_ALIGN);
2433 return skb;
2434}
2435
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002436static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2437 unsigned int length)
2438{
2439 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2440}
2441
Alexander Duyck181edb22015-05-06 21:12:03 -07002442static inline void skb_free_frag(void *addr)
2443{
2444 __free_page_frag(addr);
2445}
2446
Alexander Duyckffde7322014-12-09 19:40:42 -08002447void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002448struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2449 unsigned int length, gfp_t gfp_mask);
2450static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2451 unsigned int length)
2452{
2453 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2454}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002455void napi_consume_skb(struct sk_buff *skb, int budget);
2456
2457void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002458void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002459
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002460/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002461 * __dev_alloc_pages - allocate page for network Rx
2462 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2463 * @order: size of the allocation
2464 *
2465 * Allocate a new page.
2466 *
2467 * %NULL is returned if there is no free memory.
2468*/
2469static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2470 unsigned int order)
2471{
2472 /* This piece of code contains several assumptions.
2473 * 1. This is for device Rx, therefor a cold page is preferred.
2474 * 2. The expectation is the user wants a compound page.
2475 * 3. If requesting a order 0 page it will not be compound
2476 * due to the check to see if order has a value in prep_new_page
2477 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2478 * code in gfp_to_alloc_flags that should be enforcing this.
2479 */
2480 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2481
2482 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2483}
2484
2485static inline struct page *dev_alloc_pages(unsigned int order)
2486{
Neil Horman95829b32016-05-19 11:30:54 -04002487 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002488}
2489
2490/**
2491 * __dev_alloc_page - allocate a page for network Rx
2492 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2493 *
2494 * Allocate a new page.
2495 *
2496 * %NULL is returned if there is no free memory.
2497 */
2498static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2499{
2500 return __dev_alloc_pages(gfp_mask, 0);
2501}
2502
2503static inline struct page *dev_alloc_page(void)
2504{
Neil Horman95829b32016-05-19 11:30:54 -04002505 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002506}
2507
2508/**
Mel Gorman06140022012-07-31 16:44:24 -07002509 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2510 * @page: The page that was allocated from skb_alloc_page
2511 * @skb: The skb that may need pfmemalloc set
2512 */
2513static inline void skb_propagate_pfmemalloc(struct page *page,
2514 struct sk_buff *skb)
2515{
Michal Hocko2f064f32015-08-21 14:11:51 -07002516 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002517 skb->pfmemalloc = true;
2518}
2519
Eric Dumazet564824b2010-10-11 19:05:25 +00002520/**
Masanari Iidae2278672014-02-18 22:54:36 +09002521 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002522 * @frag: the paged fragment
2523 *
2524 * Returns the &struct page associated with @frag.
2525 */
2526static inline struct page *skb_frag_page(const skb_frag_t *frag)
2527{
Ian Campbella8605c62011-10-19 23:01:49 +00002528 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002529}
2530
2531/**
2532 * __skb_frag_ref - take an addition reference on a paged fragment.
2533 * @frag: the paged fragment
2534 *
2535 * Takes an additional reference on the paged fragment @frag.
2536 */
2537static inline void __skb_frag_ref(skb_frag_t *frag)
2538{
2539 get_page(skb_frag_page(frag));
2540}
2541
2542/**
2543 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2544 * @skb: the buffer
2545 * @f: the fragment offset.
2546 *
2547 * Takes an additional reference on the @f'th paged fragment of @skb.
2548 */
2549static inline void skb_frag_ref(struct sk_buff *skb, int f)
2550{
2551 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2552}
2553
2554/**
2555 * __skb_frag_unref - release a reference on a paged fragment.
2556 * @frag: the paged fragment
2557 *
2558 * Releases a reference on the paged fragment @frag.
2559 */
2560static inline void __skb_frag_unref(skb_frag_t *frag)
2561{
2562 put_page(skb_frag_page(frag));
2563}
2564
2565/**
2566 * skb_frag_unref - release a reference on a paged fragment of an skb.
2567 * @skb: the buffer
2568 * @f: the fragment offset
2569 *
2570 * Releases a reference on the @f'th paged fragment of @skb.
2571 */
2572static inline void skb_frag_unref(struct sk_buff *skb, int f)
2573{
2574 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2575}
2576
2577/**
2578 * skb_frag_address - gets the address of the data contained in a paged fragment
2579 * @frag: the paged fragment buffer
2580 *
2581 * Returns the address of the data within @frag. The page must already
2582 * be mapped.
2583 */
2584static inline void *skb_frag_address(const skb_frag_t *frag)
2585{
2586 return page_address(skb_frag_page(frag)) + frag->page_offset;
2587}
2588
2589/**
2590 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2591 * @frag: the paged fragment buffer
2592 *
2593 * Returns the address of the data within @frag. Checks that the page
2594 * is mapped and returns %NULL otherwise.
2595 */
2596static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2597{
2598 void *ptr = page_address(skb_frag_page(frag));
2599 if (unlikely(!ptr))
2600 return NULL;
2601
2602 return ptr + frag->page_offset;
2603}
2604
2605/**
2606 * __skb_frag_set_page - sets the page contained in a paged fragment
2607 * @frag: the paged fragment
2608 * @page: the page to set
2609 *
2610 * Sets the fragment @frag to contain @page.
2611 */
2612static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2613{
Ian Campbella8605c62011-10-19 23:01:49 +00002614 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002615}
2616
2617/**
2618 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2619 * @skb: the buffer
2620 * @f: the fragment offset
2621 * @page: the page to set
2622 *
2623 * Sets the @f'th fragment of @skb to contain @page.
2624 */
2625static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2626 struct page *page)
2627{
2628 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2629}
2630
Eric Dumazet400dfd32013-10-17 16:27:07 -07002631bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2632
Ian Campbell131ea662011-08-19 06:25:00 +00002633/**
2634 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002635 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002636 * @frag: the paged fragment to map
2637 * @offset: the offset within the fragment (starting at the
2638 * fragment's own offset)
2639 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002640 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002641 *
2642 * Maps the page associated with @frag to @device.
2643 */
2644static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2645 const skb_frag_t *frag,
2646 size_t offset, size_t size,
2647 enum dma_data_direction dir)
2648{
2649 return dma_map_page(dev, skb_frag_page(frag),
2650 frag->page_offset + offset, size, dir);
2651}
2652
Eric Dumazet117632e2011-12-03 21:39:53 +00002653static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2654 gfp_t gfp_mask)
2655{
2656 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2657}
2658
Octavian Purdilabad93e92014-06-12 01:36:26 +03002659
2660static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2661 gfp_t gfp_mask)
2662{
2663 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2664}
2665
2666
Ian Campbell131ea662011-08-19 06:25:00 +00002667/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002668 * skb_clone_writable - is the header of a clone writable
2669 * @skb: buffer to check
2670 * @len: length up to which to write
2671 *
2672 * Returns true if modifying the header part of the cloned buffer
2673 * does not requires the data to be copied.
2674 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002675static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002676{
2677 return !skb_header_cloned(skb) &&
2678 skb_headroom(skb) + len <= skb->hdr_len;
2679}
2680
Daniel Borkmann36976492016-02-19 23:05:25 +01002681static inline int skb_try_make_writable(struct sk_buff *skb,
2682 unsigned int write_len)
2683{
2684 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2685 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2686}
2687
Herbert Xud9cc2042007-09-16 16:21:16 -07002688static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2689 int cloned)
2690{
2691 int delta = 0;
2692
Herbert Xud9cc2042007-09-16 16:21:16 -07002693 if (headroom > skb_headroom(skb))
2694 delta = headroom - skb_headroom(skb);
2695
2696 if (delta || cloned)
2697 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2698 GFP_ATOMIC);
2699 return 0;
2700}
2701
Patrick McHardy334a8132007-06-25 04:35:20 -07002702/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002703 * skb_cow - copy header of skb when it is required
2704 * @skb: buffer to cow
2705 * @headroom: needed headroom
2706 *
2707 * If the skb passed lacks sufficient headroom or its data part
2708 * is shared, data is reallocated. If reallocation fails, an error
2709 * is returned and original skb is not changed.
2710 *
2711 * The result is skb with writable area skb->head...skb->tail
2712 * and at least @headroom of space at head.
2713 */
2714static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2715{
Herbert Xud9cc2042007-09-16 16:21:16 -07002716 return __skb_cow(skb, headroom, skb_cloned(skb));
2717}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002718
Herbert Xud9cc2042007-09-16 16:21:16 -07002719/**
2720 * skb_cow_head - skb_cow but only making the head writable
2721 * @skb: buffer to cow
2722 * @headroom: needed headroom
2723 *
2724 * This function is identical to skb_cow except that we replace the
2725 * skb_cloned check by skb_header_cloned. It should be used when
2726 * you only need to push on some header and do not need to modify
2727 * the data.
2728 */
2729static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2730{
2731 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732}
2733
2734/**
2735 * skb_padto - pad an skbuff up to a minimal size
2736 * @skb: buffer to pad
2737 * @len: minimal length
2738 *
2739 * Pads up a buffer to ensure the trailing bytes exist and are
2740 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002741 * is untouched. Otherwise it is extended. Returns zero on
2742 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002743 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002744static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002745{
2746 unsigned int size = skb->len;
2747 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002748 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002749 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002750}
2751
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002752/**
2753 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2754 * @skb: buffer to pad
2755 * @len: minimal length
2756 *
2757 * Pads up a buffer to ensure the trailing bytes exist and are
2758 * blanked. If the buffer already contains sufficient data it
2759 * is untouched. Otherwise it is extended. Returns zero on
2760 * success. The skb is freed on error.
2761 */
2762static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2763{
2764 unsigned int size = skb->len;
2765
2766 if (unlikely(size < len)) {
2767 len -= size;
2768 if (skb_pad(skb, len))
2769 return -ENOMEM;
2770 __skb_put(skb, len);
2771 }
2772 return 0;
2773}
2774
Linus Torvalds1da177e2005-04-16 15:20:36 -07002775static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002776 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002777{
2778 const int off = skb->len;
2779
2780 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002781 __wsum csum = 0;
2782 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2783 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784 skb->csum = csum_block_add(skb->csum, csum, off);
2785 return 0;
2786 }
Al Viroaf2b0402014-11-27 21:44:24 -05002787 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002788 return 0;
2789
2790 __skb_trim(skb, off);
2791 return -EFAULT;
2792}
2793
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002794static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2795 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002796{
2797 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002798 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002799
Ian Campbellea2ab692011-08-22 23:44:58 +00002800 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002801 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002803 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804}
2805
Herbert Xu364c6ba2006-06-09 16:10:40 -07002806static inline int __skb_linearize(struct sk_buff *skb)
2807{
2808 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2809}
2810
Linus Torvalds1da177e2005-04-16 15:20:36 -07002811/**
2812 * skb_linearize - convert paged skb to linear one
2813 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002814 *
2815 * If there is no free memory -ENOMEM is returned, otherwise zero
2816 * is returned and the old skb data released.
2817 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002818static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002820 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2821}
2822
2823/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002824 * skb_has_shared_frag - can any frag be overwritten
2825 * @skb: buffer to test
2826 *
2827 * Return true if the skb has at least one frag that might be modified
2828 * by an external entity (as in vmsplice()/sendfile())
2829 */
2830static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2831{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002832 return skb_is_nonlinear(skb) &&
2833 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002834}
2835
2836/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002837 * skb_linearize_cow - make sure skb is linear and writable
2838 * @skb: buffer to process
2839 *
2840 * If there is no free memory -ENOMEM is returned, otherwise zero
2841 * is returned and the old skb data released.
2842 */
2843static inline int skb_linearize_cow(struct sk_buff *skb)
2844{
2845 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2846 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002847}
2848
2849/**
2850 * skb_postpull_rcsum - update checksum for received skb after pull
2851 * @skb: buffer to update
2852 * @start: start of data before pull
2853 * @len: length of data pulled
2854 *
2855 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002856 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2857 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858 */
2859
2860static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002861 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002862{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002863 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002864 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
Pravin B Shelar6ae459b2015-09-22 12:57:53 -07002865 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
Pravin B Shelar31b33df2015-09-28 17:24:25 -07002866 skb_checksum_start_offset(skb) < 0)
Pravin B Shelar6ae459b2015-09-22 12:57:53 -07002867 skb->ip_summed = CHECKSUM_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002868}
2869
Herbert Xucbb042f2006-03-20 22:43:56 -08002870unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2871
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002872static inline void skb_postpush_rcsum(struct sk_buff *skb,
2873 const void *start, unsigned int len)
2874{
2875 /* For performing the reverse operation to skb_postpull_rcsum(),
2876 * we can instead of ...
2877 *
2878 * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0));
2879 *
2880 * ... just use this equivalent version here to save a few
2881 * instructions. Feeding csum of 0 in csum_partial() and later
2882 * on adding skb->csum is equivalent to feed skb->csum in the
2883 * first place.
2884 */
2885 if (skb->ip_summed == CHECKSUM_COMPLETE)
2886 skb->csum = csum_partial(start, len, skb->csum);
2887}
2888
David S. Miller7ce5a272013-12-02 17:26:05 -05002889/**
2890 * pskb_trim_rcsum - trim received skb and update checksum
2891 * @skb: buffer to trim
2892 * @len: new length
2893 *
2894 * This is exactly the same as pskb_trim except that it ensures the
2895 * checksum of received packets are still valid after the operation.
2896 */
2897
2898static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2899{
2900 if (likely(len >= skb->len))
2901 return 0;
2902 if (skb->ip_summed == CHECKSUM_COMPLETE)
2903 skb->ip_summed = CHECKSUM_NONE;
2904 return __pskb_trim(skb, len);
2905}
2906
Linus Torvalds1da177e2005-04-16 15:20:36 -07002907#define skb_queue_walk(queue, skb) \
2908 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002909 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002910 skb = skb->next)
2911
James Chapman46f89142007-04-30 00:07:31 -07002912#define skb_queue_walk_safe(queue, skb, tmp) \
2913 for (skb = (queue)->next, tmp = skb->next; \
2914 skb != (struct sk_buff *)(queue); \
2915 skb = tmp, tmp = skb->next)
2916
David S. Miller1164f522008-09-23 00:49:44 -07002917#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002918 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002919 skb = skb->next)
2920
2921#define skb_queue_walk_from_safe(queue, skb, tmp) \
2922 for (tmp = skb->next; \
2923 skb != (struct sk_buff *)(queue); \
2924 skb = tmp, tmp = skb->next)
2925
Stephen Hemminger300ce172005-10-30 13:47:34 -08002926#define skb_queue_reverse_walk(queue, skb) \
2927 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002928 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002929 skb = skb->prev)
2930
David S. Miller686a2952011-01-20 22:47:32 -08002931#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2932 for (skb = (queue)->prev, tmp = skb->prev; \
2933 skb != (struct sk_buff *)(queue); \
2934 skb = tmp, tmp = skb->prev)
2935
2936#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2937 for (tmp = skb->prev; \
2938 skb != (struct sk_buff *)(queue); \
2939 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002940
David S. Miller21dc3302010-08-23 00:13:46 -07002941static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002942{
2943 return skb_shinfo(skb)->frag_list != NULL;
2944}
2945
2946static inline void skb_frag_list_init(struct sk_buff *skb)
2947{
2948 skb_shinfo(skb)->frag_list = NULL;
2949}
2950
David S. Milleree039872009-06-09 00:17:13 -07002951#define skb_walk_frags(skb, iter) \
2952 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2953
Rainer Weikusatea3793e2015-12-06 21:11:34 +00002954
2955int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
2956 const struct sk_buff *skb);
2957struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
2958 int *peeked, int *off, int *err,
2959 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07002960struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2961 int *peeked, int *off, int *err);
2962struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2963 int *err);
2964unsigned int datagram_poll(struct file *file, struct socket *sock,
2965 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05002966int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2967 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05002968static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2969 struct msghdr *msg, int size)
2970{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002971 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002972}
Al Viroe5a4b0b2014-11-24 18:17:55 -05002973int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2974 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04002975int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2976 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04002977int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07002978void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04002979void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
2980static inline void skb_free_datagram_locked(struct sock *sk,
2981 struct sk_buff *skb)
2982{
2983 __skb_free_datagram_locked(sk, skb, 0);
2984}
Joe Perches7965bd42013-09-26 14:48:15 -07002985int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002986int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2987int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2988__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2989 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002990ssize_t skb_socket_splice(struct sock *sk,
2991 struct pipe_inode_info *pipe,
2992 struct splice_pipe_desc *spd);
2993int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07002994 struct pipe_inode_info *pipe, unsigned int len,
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002995 unsigned int flags,
2996 ssize_t (*splice_cb)(struct sock *,
2997 struct pipe_inode_info *,
2998 struct splice_pipe_desc *));
Joe Perches7965bd42013-09-26 14:48:15 -07002999void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003000unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003001int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3002 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003003void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3004int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3005void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003006unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003007bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003008struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003009struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003010int skb_ensure_writable(struct sk_buff *skb, int write_len);
Jiri Pirko93515d52014-11-19 14:05:02 +01003011int skb_vlan_pop(struct sk_buff *skb);
3012int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003013struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3014 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003015
Al Viro6ce8e9c2014-04-06 21:25:44 -04003016static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3017{
Al Viro21226ab2014-11-28 15:48:29 -05003018 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003019}
3020
Al Viro7eab8d92014-04-06 21:51:23 -04003021static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3022{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003023 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003024}
3025
Daniel Borkmann2817a332013-10-30 11:50:51 +01003026struct skb_checksum_ops {
3027 __wsum (*update)(const void *mem, int len, __wsum wsum);
3028 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3029};
3030
3031__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3032 __wsum csum, const struct skb_checksum_ops *ops);
3033__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3034 __wsum csum);
3035
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003036static inline void * __must_check
3037__skb_header_pointer(const struct sk_buff *skb, int offset,
3038 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003039{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003040 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003041 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003042
David S. Miller690e36e2014-08-23 12:13:41 -07003043 if (!skb ||
3044 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003045 return NULL;
3046
3047 return buffer;
3048}
3049
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003050static inline void * __must_check
3051skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003052{
3053 return __skb_header_pointer(skb, offset, len, skb->data,
3054 skb_headlen(skb), buffer);
3055}
3056
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003057/**
3058 * skb_needs_linearize - check if we need to linearize a given skb
3059 * depending on the given device features.
3060 * @skb: socket buffer to check
3061 * @features: net device features
3062 *
3063 * Returns true if either:
3064 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3065 * 2. skb is fragmented and the device does not support SG.
3066 */
3067static inline bool skb_needs_linearize(struct sk_buff *skb,
3068 netdev_features_t features)
3069{
3070 return skb_is_nonlinear(skb) &&
3071 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3072 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3073}
3074
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003075static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3076 void *to,
3077 const unsigned int len)
3078{
3079 memcpy(to, skb->data, len);
3080}
3081
3082static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3083 const int offset, void *to,
3084 const unsigned int len)
3085{
3086 memcpy(to, skb->data + offset, len);
3087}
3088
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003089static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3090 const void *from,
3091 const unsigned int len)
3092{
3093 memcpy(skb->data, from, len);
3094}
3095
3096static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3097 const int offset,
3098 const void *from,
3099 const unsigned int len)
3100{
3101 memcpy(skb->data + offset, from, len);
3102}
3103
Joe Perches7965bd42013-09-26 14:48:15 -07003104void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003105
Patrick Ohlyac45f602009-02-12 05:03:37 +00003106static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3107{
3108 return skb->tstamp;
3109}
3110
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003111/**
3112 * skb_get_timestamp - get timestamp from a skb
3113 * @skb: skb to get stamp from
3114 * @stamp: pointer to struct timeval to store stamp in
3115 *
3116 * Timestamps are stored in the skb as offsets to a base timestamp.
3117 * This function converts the offset back to a struct timeval and stores
3118 * it in stamp.
3119 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003120static inline void skb_get_timestamp(const struct sk_buff *skb,
3121 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003122{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003123 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003124}
3125
Patrick Ohlyac45f602009-02-12 05:03:37 +00003126static inline void skb_get_timestampns(const struct sk_buff *skb,
3127 struct timespec *stamp)
3128{
3129 *stamp = ktime_to_timespec(skb->tstamp);
3130}
3131
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003132static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003133{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003134 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003135}
3136
Stephen Hemminger164891a2007-04-23 22:26:16 -07003137static inline ktime_t net_timedelta(ktime_t t)
3138{
3139 return ktime_sub(ktime_get_real(), t);
3140}
3141
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003142static inline ktime_t net_invalid_timestamp(void)
3143{
3144 return ktime_set(0, 0);
3145}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003146
Alexander Duyck62bccb82014-09-04 13:31:35 -04003147struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3148
Richard Cochranc1f19b52010-07-17 08:49:36 +00003149#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3150
Joe Perches7965bd42013-09-26 14:48:15 -07003151void skb_clone_tx_timestamp(struct sk_buff *skb);
3152bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003153
3154#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3155
3156static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3157{
3158}
3159
3160static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3161{
3162 return false;
3163}
3164
3165#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3166
3167/**
3168 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3169 *
Richard Cochranda92b192011-10-21 00:49:15 +00003170 * PHY drivers may accept clones of transmitted packets for
3171 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003172 * must call this function to return the skb back to the stack with a
3173 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003174 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003175 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003176 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003177 *
3178 */
3179void skb_complete_tx_timestamp(struct sk_buff *skb,
3180 struct skb_shared_hwtstamps *hwtstamps);
3181
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003182void __skb_tstamp_tx(struct sk_buff *orig_skb,
3183 struct skb_shared_hwtstamps *hwtstamps,
3184 struct sock *sk, int tstype);
3185
Patrick Ohlyac45f602009-02-12 05:03:37 +00003186/**
3187 * skb_tstamp_tx - queue clone of skb with send time stamps
3188 * @orig_skb: the original outgoing packet
3189 * @hwtstamps: hardware time stamps, may be NULL if not available
3190 *
3191 * If the skb has a socket associated, then this function clones the
3192 * skb (thus sharing the actual data and optional structures), stores
3193 * the optional hardware time stamping information (if non NULL) or
3194 * generates a software time stamp (otherwise), then queues the clone
3195 * to the error queue of the socket. Errors are silently ignored.
3196 */
Joe Perches7965bd42013-09-26 14:48:15 -07003197void skb_tstamp_tx(struct sk_buff *orig_skb,
3198 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003199
Richard Cochran4507a712010-07-17 08:48:28 +00003200static inline void sw_tx_timestamp(struct sk_buff *skb)
3201{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00003202 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3203 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00003204 skb_tstamp_tx(skb, NULL);
3205}
3206
3207/**
3208 * skb_tx_timestamp() - Driver hook for transmit timestamping
3209 *
3210 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003211 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003212 *
David S. Miller73409f32013-12-27 13:04:33 -05003213 * Specifically, one should make absolutely sure that this function is
3214 * called before TX completion of this packet can trigger. Otherwise
3215 * the packet could potentially already be freed.
3216 *
Richard Cochran4507a712010-07-17 08:48:28 +00003217 * @skb: A socket buffer.
3218 */
3219static inline void skb_tx_timestamp(struct sk_buff *skb)
3220{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003221 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00003222 sw_tx_timestamp(skb);
3223}
3224
Johannes Berg6e3e9392011-11-09 10:15:42 +01003225/**
3226 * skb_complete_wifi_ack - deliver skb with wifi status
3227 *
3228 * @skb: the original outgoing packet
3229 * @acked: ack status
3230 *
3231 */
3232void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3233
Joe Perches7965bd42013-09-26 14:48:15 -07003234__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3235__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003236
Herbert Xu60476372007-04-09 11:59:39 -07003237static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3238{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003239 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3240 skb->csum_valid ||
3241 (skb->ip_summed == CHECKSUM_PARTIAL &&
3242 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003243}
3244
Herbert Xufb286bb2005-11-10 13:01:24 -08003245/**
3246 * skb_checksum_complete - Calculate checksum of an entire packet
3247 * @skb: packet to process
3248 *
3249 * This function calculates the checksum over the entire packet plus
3250 * the value of skb->csum. The latter can be used to supply the
3251 * checksum of a pseudo header as used by TCP/UDP. It returns the
3252 * checksum.
3253 *
3254 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3255 * this function can be used to verify that checksum on received
3256 * packets. In that case the function should return zero if the
3257 * checksum is correct. In particular, this function will return zero
3258 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3259 * hardware has already verified the correctness of the checksum.
3260 */
Al Viro4381ca32007-07-15 21:00:11 +01003261static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003262{
Herbert Xu60476372007-04-09 11:59:39 -07003263 return skb_csum_unnecessary(skb) ?
3264 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003265}
3266
Tom Herbert77cffe22014-08-27 21:26:46 -07003267static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3268{
3269 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3270 if (skb->csum_level == 0)
3271 skb->ip_summed = CHECKSUM_NONE;
3272 else
3273 skb->csum_level--;
3274 }
3275}
3276
3277static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3278{
3279 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3280 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3281 skb->csum_level++;
3282 } else if (skb->ip_summed == CHECKSUM_NONE) {
3283 skb->ip_summed = CHECKSUM_UNNECESSARY;
3284 skb->csum_level = 0;
3285 }
3286}
3287
Tom Herbert5a212322014-08-31 15:12:41 -07003288static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3289{
3290 /* Mark current checksum as bad (typically called from GRO
3291 * path). In the case that ip_summed is CHECKSUM_NONE
3292 * this must be the first checksum encountered in the packet.
3293 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3294 * checksum after the last one validated. For UDP, a zero
3295 * checksum can not be marked as bad.
3296 */
3297
3298 if (skb->ip_summed == CHECKSUM_NONE ||
3299 skb->ip_summed == CHECKSUM_UNNECESSARY)
3300 skb->csum_bad = 1;
3301}
3302
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003303/* Check if we need to perform checksum complete validation.
3304 *
3305 * Returns true if checksum complete is needed, false otherwise
3306 * (either checksum is unnecessary or zero checksum is allowed).
3307 */
3308static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3309 bool zero_okay,
3310 __sum16 check)
3311{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003312 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3313 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003314 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003315 return false;
3316 }
3317
3318 return true;
3319}
3320
3321/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3322 * in checksum_init.
3323 */
3324#define CHECKSUM_BREAK 76
3325
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003326/* Unset checksum-complete
3327 *
3328 * Unset checksum complete can be done when packet is being modified
3329 * (uncompressed for instance) and checksum-complete value is
3330 * invalidated.
3331 */
3332static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3333{
3334 if (skb->ip_summed == CHECKSUM_COMPLETE)
3335 skb->ip_summed = CHECKSUM_NONE;
3336}
3337
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003338/* Validate (init) checksum based on checksum complete.
3339 *
3340 * Return values:
3341 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3342 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3343 * checksum is stored in skb->csum for use in __skb_checksum_complete
3344 * non-zero: value of invalid checksum
3345 *
3346 */
3347static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3348 bool complete,
3349 __wsum psum)
3350{
3351 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3352 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003353 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003354 return 0;
3355 }
Tom Herbert5a212322014-08-31 15:12:41 -07003356 } else if (skb->csum_bad) {
3357 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003358 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003359 }
3360
3361 skb->csum = psum;
3362
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003363 if (complete || skb->len <= CHECKSUM_BREAK) {
3364 __sum16 csum;
3365
3366 csum = __skb_checksum_complete(skb);
3367 skb->csum_valid = !csum;
3368 return csum;
3369 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003370
3371 return 0;
3372}
3373
3374static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3375{
3376 return 0;
3377}
3378
3379/* Perform checksum validate (init). Note that this is a macro since we only
3380 * want to calculate the pseudo header which is an input function if necessary.
3381 * First we try to validate without any computation (checksum unnecessary) and
3382 * then calculate based on checksum complete calling the function to compute
3383 * pseudo header.
3384 *
3385 * Return values:
3386 * 0: checksum is validated or try to in skb_checksum_complete
3387 * non-zero: value of invalid checksum
3388 */
3389#define __skb_checksum_validate(skb, proto, complete, \
3390 zero_okay, check, compute_pseudo) \
3391({ \
3392 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003393 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003394 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3395 __ret = __skb_checksum_validate_complete(skb, \
3396 complete, compute_pseudo(skb, proto)); \
3397 __ret; \
3398})
3399
3400#define skb_checksum_init(skb, proto, compute_pseudo) \
3401 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3402
3403#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3404 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3405
3406#define skb_checksum_validate(skb, proto, compute_pseudo) \
3407 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3408
3409#define skb_checksum_validate_zero_check(skb, proto, check, \
3410 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003411 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003412
3413#define skb_checksum_simple_validate(skb) \
3414 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3415
Tom Herbertd96535a2014-08-31 15:12:42 -07003416static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3417{
3418 return (skb->ip_summed == CHECKSUM_NONE &&
3419 skb->csum_valid && !skb->csum_bad);
3420}
3421
3422static inline void __skb_checksum_convert(struct sk_buff *skb,
3423 __sum16 check, __wsum pseudo)
3424{
3425 skb->csum = ~pseudo;
3426 skb->ip_summed = CHECKSUM_COMPLETE;
3427}
3428
3429#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3430do { \
3431 if (__skb_checksum_convert_check(skb)) \
3432 __skb_checksum_convert(skb, check, \
3433 compute_pseudo(skb, proto)); \
3434} while (0)
3435
Tom Herbert15e23962015-02-10 16:30:31 -08003436static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3437 u16 start, u16 offset)
3438{
3439 skb->ip_summed = CHECKSUM_PARTIAL;
3440 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3441 skb->csum_offset = offset - start;
3442}
3443
Tom Herbertdcdc8992015-02-02 16:07:34 -08003444/* Update skbuf and packet to reflect the remote checksum offload operation.
3445 * When called, ptr indicates the starting point for skb->csum when
3446 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3447 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3448 */
3449static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003450 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003451{
3452 __wsum delta;
3453
Tom Herbert15e23962015-02-10 16:30:31 -08003454 if (!nopartial) {
3455 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3456 return;
3457 }
3458
Tom Herbertdcdc8992015-02-02 16:07:34 -08003459 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3460 __skb_checksum_complete(skb);
3461 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3462 }
3463
3464 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3465
3466 /* Adjust skb->csum since we changed the packet */
3467 skb->csum = csum_add(skb->csum, delta);
3468}
3469
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003470#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003471void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003472static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3473{
3474 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003475 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003476}
3477static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3478{
3479 if (nfct)
3480 atomic_inc(&nfct->use);
3481}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003482#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003483#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003484static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3485{
3486 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3487 kfree(nf_bridge);
3488}
3489static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3490{
3491 if (nf_bridge)
3492 atomic_inc(&nf_bridge->use);
3493}
3494#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003495static inline void nf_reset(struct sk_buff *skb)
3496{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003497#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003498 nf_conntrack_put(skb->nfct);
3499 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003500#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003501#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003502 nf_bridge_put(skb->nf_bridge);
3503 skb->nf_bridge = NULL;
3504#endif
3505}
3506
Patrick McHardy124dff02013-04-05 20:42:05 +02003507static inline void nf_reset_trace(struct sk_buff *skb)
3508{
Florian Westphal478b3602014-02-15 23:48:45 +01003509#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003510 skb->nf_trace = 0;
3511#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003512}
3513
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003514/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003515static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3516 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003517{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003518#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003519 dst->nfct = src->nfct;
3520 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003521 if (copy)
3522 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003523#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003524#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003525 dst->nf_bridge = src->nf_bridge;
3526 nf_bridge_get(src->nf_bridge);
3527#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003528#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003529 if (copy)
3530 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003531#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003532}
3533
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003534static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3535{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003536#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003537 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003538#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003539#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003540 nf_bridge_put(dst->nf_bridge);
3541#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003542 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003543}
3544
James Morris984bc162006-06-09 00:29:17 -07003545#ifdef CONFIG_NETWORK_SECMARK
3546static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3547{
3548 to->secmark = from->secmark;
3549}
3550
3551static inline void skb_init_secmark(struct sk_buff *skb)
3552{
3553 skb->secmark = 0;
3554}
3555#else
3556static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3557{ }
3558
3559static inline void skb_init_secmark(struct sk_buff *skb)
3560{ }
3561#endif
3562
Eric W. Biederman574f7192014-04-01 12:20:24 -07003563static inline bool skb_irq_freeable(const struct sk_buff *skb)
3564{
3565 return !skb->destructor &&
3566#if IS_ENABLED(CONFIG_XFRM)
3567 !skb->sp &&
3568#endif
3569#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3570 !skb->nfct &&
3571#endif
3572 !skb->_skb_refdst &&
3573 !skb_has_frag_list(skb);
3574}
3575
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003576static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3577{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003578 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003579}
3580
Stephen Hemminger92477442009-03-21 13:39:26 -07003581static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003582{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003583 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003584}
3585
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003586static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3587{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003588 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003589}
3590
David S. Millerd5a9e242009-01-27 16:22:11 -08003591static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3592{
3593 skb->queue_mapping = rx_queue + 1;
3594}
3595
Stephen Hemminger92477442009-03-21 13:39:26 -07003596static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003597{
3598 return skb->queue_mapping - 1;
3599}
3600
Stephen Hemminger92477442009-03-21 13:39:26 -07003601static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003602{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003603 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003604}
3605
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003606static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3607{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003608#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003609 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003610#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003611 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003612#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003613}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003614
Pravin B Shelar68c33162013-02-14 14:02:41 +00003615/* Keeps track of mac header offset relative to skb->head.
3616 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3617 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003618 * tunnel skb it points to outer mac header.
3619 * Keeps track of level of encapsulation of network headers.
3620 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003621struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003622 union {
3623 int mac_offset;
3624 int data_offset;
3625 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003626 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003627 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003628 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003629};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003630#define SKB_SGO_CB_OFFSET 32
3631#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003632
3633static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3634{
3635 return (skb_mac_header(inner_skb) - inner_skb->head) -
3636 SKB_GSO_CB(inner_skb)->mac_offset;
3637}
3638
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003639static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3640{
3641 int new_headroom, headroom;
3642 int ret;
3643
3644 headroom = skb_headroom(skb);
3645 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3646 if (ret)
3647 return ret;
3648
3649 new_headroom = skb_headroom(skb);
3650 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3651 return 0;
3652}
3653
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003654static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3655{
3656 /* Do not update partial checksums if remote checksum is enabled. */
3657 if (skb->remcsum_offload)
3658 return;
3659
3660 SKB_GSO_CB(skb)->csum = res;
3661 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3662}
3663
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003664/* Compute the checksum for a gso segment. First compute the checksum value
3665 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3666 * then add in skb->csum (checksum from csum_start to end of packet).
3667 * skb->csum and csum_start are then updated to reflect the checksum of the
3668 * resultant packet starting from the transport header-- the resultant checksum
3669 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3670 * header.
3671 */
3672static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3673{
Alexander Duyck76443452016-02-05 15:27:37 -08003674 unsigned char *csum_start = skb_transport_header(skb);
3675 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3676 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003677
Alexander Duyck76443452016-02-05 15:27:37 -08003678 SKB_GSO_CB(skb)->csum = res;
3679 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003680
Alexander Duyck76443452016-02-05 15:27:37 -08003681 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003682}
3683
David S. Millerbdcc0922012-03-07 20:53:36 -05003684static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003685{
3686 return skb_shinfo(skb)->gso_size;
3687}
3688
Eric Dumazet36a8f392013-09-29 01:21:32 -07003689/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003690static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003691{
3692 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3693}
3694
Joe Perches7965bd42013-09-26 14:48:15 -07003695void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003696
3697static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3698{
3699 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3700 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003701 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3702
Alexander Duyckb78462e2010-06-02 12:24:37 +00003703 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3704 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003705 __skb_warn_lro_forwarding(skb);
3706 return true;
3707 }
3708 return false;
3709}
3710
Herbert Xu35fc92a2007-03-26 23:22:20 -07003711static inline void skb_forward_csum(struct sk_buff *skb)
3712{
3713 /* Unfortunately we don't support this one. Any brave souls? */
3714 if (skb->ip_summed == CHECKSUM_COMPLETE)
3715 skb->ip_summed = CHECKSUM_NONE;
3716}
3717
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003718/**
3719 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3720 * @skb: skb to check
3721 *
3722 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3723 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3724 * use this helper, to document places where we make this assertion.
3725 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003726static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003727{
3728#ifdef DEBUG
3729 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3730#endif
3731}
3732
Rusty Russellf35d9d82008-02-04 23:49:54 -05003733bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003734
Paul Durranted1f50c2014-01-09 10:02:46 +00003735int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003736struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3737 unsigned int transport_len,
3738 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003739
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003740/**
3741 * skb_head_is_locked - Determine if the skb->head is locked down
3742 * @skb: skb to check
3743 *
3744 * The head on skbs build around a head frag can be removed if they are
3745 * not cloned. This function returns true if the skb head is locked down
3746 * due to either being allocated via kmalloc, or by being a clone with
3747 * multiple references to the head.
3748 */
3749static inline bool skb_head_is_locked(const struct sk_buff *skb)
3750{
3751 return !skb->head_frag || skb_cloned(skb);
3752}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003753
3754/**
3755 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3756 *
3757 * @skb: GSO skb
3758 *
3759 * skb_gso_network_seglen is used to determine the real size of the
3760 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3761 *
3762 * The MAC/L2 header is not accounted for.
3763 */
3764static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3765{
3766 unsigned int hdr_len = skb_transport_header(skb) -
3767 skb_network_header(skb);
3768 return hdr_len + skb_gso_transport_seglen(skb);
3769}
Thomas Grafee122c72015-07-21 10:43:58 +02003770
Edward Cree179bc672016-02-11 20:48:04 +00003771/* Local Checksum Offload.
3772 * Compute outer checksum based on the assumption that the
3773 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003774 * See Documentation/networking/checksum-offloads.txt for
3775 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003776 * Fill in outer checksum adjustment (e.g. with sum of outer
3777 * pseudo-header) before calling.
3778 * Also ensure that inner checksum is in linear data area.
3779 */
3780static inline __wsum lco_csum(struct sk_buff *skb)
3781{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003782 unsigned char *csum_start = skb_checksum_start(skb);
3783 unsigned char *l4_hdr = skb_transport_header(skb);
3784 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003785
3786 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003787 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3788 skb->csum_offset));
3789
Edward Cree179bc672016-02-11 20:48:04 +00003790 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003791 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003792 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003793 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003794}
3795
Linus Torvalds1da177e2005-04-16 15:20:36 -07003796#endif /* __KERNEL__ */
3797#endif /* _LINUX_SKBUFF_H */