net: introduce skb_crc32c_csum_help
[linux-2.6-block.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
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
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
e6017571 37#include <linux/sched/clock.h>
1bd758eb 38#include <net/flow_dissector.h>
a60e3cc7 39#include <linux/splice.h>
72b31f72 40#include <linux/in6.h>
8b10cab6 41#include <linux/if_packet.h>
f70ea018 42#include <net/flow.h>
1da177e4 43
7a6ae71b
TH
44/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
78ea85f1
DB
89 *
90 * CHECKSUM_NONE:
91 *
7a6ae71b 92 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
77cffe23
TH
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
112 *
113 * skb->csum_level indicates the number of consecutive checksums found in
114 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
115 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
116 * and a device is able to verify the checksums for UDP (possibly zero),
117 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
118 * two. If the device were only able to verify the UDP checksum and not
119 * GRE, either because it doesn't support GRE checksum of because GRE
120 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
121 * not considered in this case).
78ea85f1
DB
122 *
123 * CHECKSUM_COMPLETE:
124 *
125 * This is the most generic way. The device supplied checksum of the _whole_
126 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
127 * hardware doesn't need to parse L3/L4 headers to implement this.
128 *
129 * Note: Even if device supports only some protocols, but is able to produce
130 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
131 *
132 * CHECKSUM_PARTIAL:
133 *
6edec0e6
TH
134 * A checksum is set up to be offloaded to a device as described in the
135 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 136 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
137 * on the same host, or it may be set in the input path in GRO or remote
138 * checksum offload. For the purposes of checksum verification, the checksum
139 * referred to by skb->csum_start + skb->csum_offset and any preceding
140 * checksums in the packet are considered verified. Any checksums in the
141 * packet that are after the checksum being offloaded are not considered to
142 * be verified.
78ea85f1 143 *
7a6ae71b
TH
144 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
145 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
146 *
147 * CHECKSUM_PARTIAL:
148 *
7a6ae71b 149 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 150 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
151 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
152 * csum_start and csum_offset values are valid values given the length and
153 * offset of the packet, however they should not attempt to validate that the
154 * checksum refers to a legitimate transport layer checksum-- it is the
155 * purview of the stack to validate that csum_start and csum_offset are set
156 * correctly.
157 *
158 * When the stack requests checksum offload for a packet, the driver MUST
159 * ensure that the checksum is set correctly. A driver can either offload the
160 * checksum calculation to the device, or call skb_checksum_help (in the case
161 * that the device does not support offload for a particular checksum).
162 *
163 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
164 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
165 * checksum offload capability. If a device has limited checksum capabilities
166 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
167 * described above) a helper function can be called to resolve
168 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
169 * function takes a spec argument that describes the protocol layer that is
170 * supported for checksum offload and can be called for each packet. If a
171 * packet does not match the specification for offload, skb_checksum_help
172 * is called to resolve the checksum.
78ea85f1 173 *
7a6ae71b 174 * CHECKSUM_NONE:
78ea85f1 175 *
7a6ae71b
TH
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
78ea85f1
DB
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
7a6ae71b
TH
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
78ea85f1 183 *
7a6ae71b
TH
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
192 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
193 * accordingly. Note the there is no indication in the skbuff that the
194 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
195 * both IP checksum offload and SCTP CRC offload must verify which offload
b72b5bf6
DC
196 * is configured for a packet presumably by inspecting packet headers; in
197 * case, skb_crc32c_csum_help is provided to compute CRC on SCTP packets.
7a6ae71b
TH
198 *
199 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
200 * offloading the FCOE CRC in a packet. To perform this offload the stack
201 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
202 * accordingly. Note the there is no indication in the skbuff that the
203 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
204 * both IP checksum offload and FCOE CRC offload must verify which offload
205 * is configured for a packet presumably by inspecting packet headers.
206 *
207 * E. Checksumming on output with GSO.
208 *
209 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
210 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
211 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
212 * part of the GSO operation is implied. If a checksum is being offloaded
213 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
214 * are set to refer to the outermost checksum being offload (two offloaded
215 * checksums are possible with UDP encapsulation).
78ea85f1
DB
216 */
217
60476372 218/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
219#define CHECKSUM_NONE 0
220#define CHECKSUM_UNNECESSARY 1
221#define CHECKSUM_COMPLETE 2
222#define CHECKSUM_PARTIAL 3
1da177e4 223
77cffe23
TH
224/* Maximum value in skb->csum_level */
225#define SKB_MAX_CSUM_LEVEL 3
226
0bec8c88 227#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 228#define SKB_WITH_OVERHEAD(X) \
deea84b0 229 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
230#define SKB_MAX_ORDER(X, ORDER) \
231 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
232#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
233#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
234
87fb4b7b
ED
235/* return minimum truesize of one skb containing X bytes of data */
236#define SKB_TRUESIZE(X) ((X) + \
237 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
238 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
239
1da177e4 240struct net_device;
716ea3a7 241struct scatterlist;
9c55e01c 242struct pipe_inode_info;
a8f820aa 243struct iov_iter;
fd11a83d 244struct napi_struct;
1da177e4 245
5f79e0f9 246#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
247struct nf_conntrack {
248 atomic_t use;
1da177e4 249};
5f79e0f9 250#endif
1da177e4 251
34666d46 252#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 253struct nf_bridge_info {
bf1ac5ca 254 atomic_t use;
3eaf4025
FW
255 enum {
256 BRNF_PROTO_UNCHANGED,
257 BRNF_PROTO_8021Q,
258 BRNF_PROTO_PPPOE
7fb48c5b 259 } orig_proto:8;
72b1e5e4
FW
260 u8 pkt_otherhost:1;
261 u8 in_prerouting:1;
262 u8 bridged_dnat:1;
411ffb4f 263 __u16 frag_max_size;
bf1ac5ca 264 struct net_device *physindev;
63cdbc06
FW
265
266 /* always valid & non-NULL from FORWARD on, for physdev match */
267 struct net_device *physoutdev;
7fb48c5b 268 union {
72b1e5e4 269 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
270 __be32 ipv4_daddr;
271 struct in6_addr ipv6_daddr;
72b1e5e4
FW
272
273 /* after prerouting + nat detected: store original source
274 * mac since neigh resolution overwrites it, only used while
275 * skb is out in neigh layer.
276 */
277 char neigh_header[8];
72b31f72 278 };
1da177e4
LT
279};
280#endif
281
1da177e4
LT
282struct sk_buff_head {
283 /* These two members must be first. */
284 struct sk_buff *next;
285 struct sk_buff *prev;
286
287 __u32 qlen;
288 spinlock_t lock;
289};
290
291struct sk_buff;
292
9d4dde52
IC
293/* To allow 64K frame to be packed as single skb without frag_list we
294 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
295 * buffers which do not start on a page boundary.
296 *
297 * Since GRO uses frags we allocate at least 16 regardless of page
298 * size.
a715dea3 299 */
9d4dde52 300#if (65536/PAGE_SIZE + 1) < 16
eec00954 301#define MAX_SKB_FRAGS 16UL
a715dea3 302#else
9d4dde52 303#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 304#endif
5f74f82e 305extern int sysctl_max_skb_frags;
1da177e4 306
3953c46c
MRL
307/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
308 * segment using its current segmentation instead.
309 */
310#define GSO_BY_FRAGS 0xFFFF
311
1da177e4
LT
312typedef struct skb_frag_struct skb_frag_t;
313
314struct skb_frag_struct {
a8605c60
IC
315 struct {
316 struct page *p;
317 } page;
cb4dfe56 318#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
319 __u32 page_offset;
320 __u32 size;
cb4dfe56
ED
321#else
322 __u16 page_offset;
323 __u16 size;
324#endif
1da177e4
LT
325};
326
9e903e08
ED
327static inline unsigned int skb_frag_size(const skb_frag_t *frag)
328{
329 return frag->size;
330}
331
332static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
333{
334 frag->size = size;
335}
336
337static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
338{
339 frag->size += delta;
340}
341
342static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
343{
344 frag->size -= delta;
345}
346
ac45f602
PO
347#define HAVE_HW_TIME_STAMP
348
349/**
d3a21be8 350 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
351 * @hwtstamp: hardware time stamp transformed into duration
352 * since arbitrary point in time
ac45f602
PO
353 *
354 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 355 * skb->tstamp.
ac45f602
PO
356 *
357 * hwtstamps can only be compared against other hwtstamps from
358 * the same device.
359 *
360 * This structure is attached to packets as part of the
361 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
362 */
363struct skb_shared_hwtstamps {
364 ktime_t hwtstamp;
ac45f602
PO
365};
366
2244d07b
OH
367/* Definitions for tx_flags in struct skb_shared_info */
368enum {
369 /* generate hardware time stamp */
370 SKBTX_HW_TSTAMP = 1 << 0,
371
e7fd2885 372 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
373 SKBTX_SW_TSTAMP = 1 << 1,
374
375 /* device driver is going to provide hardware time stamp */
376 SKBTX_IN_PROGRESS = 1 << 2,
377
a6686f2f 378 /* device driver supports TX zero-copy buffers */
62b1a8ab 379 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
380
381 /* generate wifi status information (where possible) */
62b1a8ab 382 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
383
384 /* This indicates at least one fragment might be overwritten
385 * (as in vmsplice(), sendfile() ...)
386 * If we need to compute a TX checksum, we'll need to copy
387 * all frags to avoid possible bad checksum
388 */
389 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
390
391 /* generate software time stamp when entering packet scheduling */
392 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
393};
394
e1c8a607 395#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 396 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
397#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
398
a6686f2f
SM
399/*
400 * The callback notifies userspace to release buffers when skb DMA is done in
401 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
402 * The zerocopy_success argument is true if zero copy transmit occurred,
403 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
404 * The ctx field is used to track device context.
405 * The desc field is used to track userspace buffer index.
a6686f2f
SM
406 */
407struct ubuf_info {
e19d6763 408 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 409 void *ctx;
a6686f2f 410 unsigned long desc;
ac45f602
PO
411};
412
1da177e4
LT
413/* This data is invariant across clones and lives at
414 * the end of the header data, ie. at skb->end.
415 */
416struct skb_shared_info {
7f564528 417 unsigned short _unused;
9f42f126
IC
418 unsigned char nr_frags;
419 __u8 tx_flags;
7967168c
HX
420 unsigned short gso_size;
421 /* Warning: this field is not always filled in (UFO)! */
422 unsigned short gso_segs;
1da177e4 423 struct sk_buff *frag_list;
ac45f602 424 struct skb_shared_hwtstamps hwtstamps;
7f564528 425 unsigned int gso_type;
09c2d251 426 u32 tskey;
9f42f126 427 __be32 ip6_frag_id;
ec7d2f2c
ED
428
429 /*
430 * Warning : all fields before dataref are cleared in __alloc_skb()
431 */
432 atomic_t dataref;
433
69e3c75f
JB
434 /* Intermediate layers must ensure that destructor_arg
435 * remains valid until skb destructor */
436 void * destructor_arg;
a6686f2f 437
fed66381
ED
438 /* must be last field, see pskb_expand_head() */
439 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
440};
441
442/* We divide dataref into two halves. The higher 16 bits hold references
443 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
444 * the entire skb->data. A clone of a headerless skb holds the length of
445 * the header in skb->hdr_len.
1da177e4
LT
446 *
447 * All users must obey the rule that the skb->data reference count must be
448 * greater than or equal to the payload reference count.
449 *
450 * Holding a reference to the payload part means that the user does not
451 * care about modifications to the header part of skb->data.
452 */
453#define SKB_DATAREF_SHIFT 16
454#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
455
d179cd12
DM
456
457enum {
c8753d55
VS
458 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
459 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
460 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
461};
462
7967168c
HX
463enum {
464 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 465 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
466
467 /* This indicates the skb is from an untrusted source. */
468 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
469
470 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
471 SKB_GSO_TCP_ECN = 1 << 3,
472
cbc53e08 473 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 474
cbc53e08 475 SKB_GSO_TCPV6 = 1 << 5,
68c33163 476
cbc53e08 477 SKB_GSO_FCOE = 1 << 6,
73136267 478
cbc53e08 479 SKB_GSO_GRE = 1 << 7,
0d89d203 480
cbc53e08 481 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 482
7e13318d 483 SKB_GSO_IPXIP4 = 1 << 9,
61c1db7f 484
7e13318d 485 SKB_GSO_IPXIP6 = 1 << 10,
0f4f4ffa 486
cbc53e08 487 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 488
cbc53e08
AD
489 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
490
802ab55a
AD
491 SKB_GSO_PARTIAL = 1 << 13,
492
493 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
90017acc
MRL
494
495 SKB_GSO_SCTP = 1 << 15,
c7ef8f0c
SK
496
497 SKB_GSO_ESP = 1 << 16,
7967168c
HX
498};
499
2e07fa9c
ACM
500#if BITS_PER_LONG > 32
501#define NET_SKBUFF_DATA_USES_OFFSET 1
502#endif
503
504#ifdef NET_SKBUFF_DATA_USES_OFFSET
505typedef unsigned int sk_buff_data_t;
506#else
507typedef unsigned char *sk_buff_data_t;
508#endif
509
1da177e4
LT
510/**
511 * struct sk_buff - socket buffer
512 * @next: Next buffer in list
513 * @prev: Previous buffer in list
363ec392 514 * @tstamp: Time we arrived/left
56b17425 515 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 516 * @sk: Socket we are owned by
1da177e4 517 * @dev: Device we arrived on/are leaving by
d84e0bd7 518 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 519 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 520 * @sp: the security path, used for xfrm
1da177e4
LT
521 * @len: Length of actual data
522 * @data_len: Data length
523 * @mac_len: Length of link layer header
334a8132 524 * @hdr_len: writable header length of cloned skb
663ead3b
HX
525 * @csum: Checksum (must include start/offset pair)
526 * @csum_start: Offset from skb->head where checksumming should start
527 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 528 * @priority: Packet queueing priority
60ff7467 529 * @ignore_df: allow local fragmentation
1da177e4 530 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 531 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
532 * @nohdr: Payload reference only, must not modify header
533 * @pkt_type: Packet class
c83c2486 534 * @fclone: skbuff clone status
c83c2486 535 * @ipvs_property: skbuff is owned by ipvs
e7246e12 536 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 537 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
538 * @tc_redirected: packet was redirected by a tc action
539 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
540 * @peeked: this packet has been seen already, so stats have been
541 * done for it, don't do them again
ba9dda3a 542 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
543 * @protocol: Packet protocol from driver
544 * @destructor: Destruct function
a9e419dc 545 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 546 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 547 * @skb_iif: ifindex of device we arrived on
1da177e4 548 * @tc_index: Traffic control index
61b905da 549 * @hash: the packet hash
d84e0bd7 550 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 551 * @xmit_more: More SKBs are pending for this queue
553a5672 552 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 553 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 554 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 555 * ports.
a3b18ddb 556 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
557 * @wifi_acked_valid: wifi_acked was set
558 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 559 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
4ff06203 560 * @dst_pending_confirm: need to confirm neighbour
06021292 561 * @napi_id: id of the NAPI struct this skb came from
984bc16c 562 * @secmark: security marking
d84e0bd7 563 * @mark: Generic packet mark
86a9bad3 564 * @vlan_proto: vlan encapsulation protocol
6aa895b0 565 * @vlan_tci: vlan tag control information
0d89d203 566 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
567 * @inner_transport_header: Inner transport layer header (encapsulation)
568 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 569 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
570 * @transport_header: Transport layer header
571 * @network_header: Network layer header
572 * @mac_header: Link layer header
573 * @tail: Tail pointer
574 * @end: End pointer
575 * @head: Head of buffer
576 * @data: Data head pointer
577 * @truesize: Buffer size
578 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
579 */
580
581struct sk_buff {
363ec392 582 union {
56b17425
ED
583 struct {
584 /* These two members must be first. */
585 struct sk_buff *next;
586 struct sk_buff *prev;
587
588 union {
589 ktime_t tstamp;
9a568de4 590 u64 skb_mstamp;
56b17425
ED
591 };
592 };
593 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 594 };
da3f5cf1 595 struct sock *sk;
1da177e4 596
c84d9490
ED
597 union {
598 struct net_device *dev;
599 /* Some protocols might use this space to store information,
600 * while device pointer would be NULL.
601 * UDP receive path is one user.
602 */
603 unsigned long dev_scratch;
604 };
1da177e4
LT
605 /*
606 * This is the control buffer. It is free to use for every
607 * layer. Please put your private variables there. If you
608 * want to keep them across layers you have to do a skb_clone()
609 * first. This is owned by whoever has the skb queued ATM.
610 */
da3f5cf1 611 char cb[48] __aligned(8);
1da177e4 612
7fee226a 613 unsigned long _skb_refdst;
b1937227 614 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
615#ifdef CONFIG_XFRM
616 struct sec_path *sp;
b1937227
ED
617#endif
618#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 619 unsigned long _nfct;
b1937227 620#endif
85224844 621#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 622 struct nf_bridge_info *nf_bridge;
da3f5cf1 623#endif
1da177e4 624 unsigned int len,
334a8132
PM
625 data_len;
626 __u16 mac_len,
627 hdr_len;
b1937227
ED
628
629 /* Following fields are _not_ copied in __copy_skb_header()
630 * Note that queue_mapping is here mostly to fill a hole.
631 */
fe55f6d5 632 kmemcheck_bitfield_begin(flags1);
b1937227 633 __u16 queue_mapping;
36bbef52
DB
634
635/* if you move cloned around you also must adapt those constants */
636#ifdef __BIG_ENDIAN_BITFIELD
637#define CLONED_MASK (1 << 7)
638#else
639#define CLONED_MASK 1
640#endif
641#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
642
643 __u8 __cloned_offset[0];
b1937227 644 __u8 cloned:1,
6869c4d8 645 nohdr:1,
b84f4cc9 646 fclone:2,
a59322be 647 peeked:1,
b1937227 648 head_frag:1,
36bbef52
DB
649 xmit_more:1,
650 __unused:1; /* one bit hole */
fe55f6d5 651 kmemcheck_bitfield_end(flags1);
4031ae6e 652
b1937227
ED
653 /* fields enclosed in headers_start/headers_end are copied
654 * using a single memcpy() in __copy_skb_header()
655 */
ebcf34f3 656 /* private: */
b1937227 657 __u32 headers_start[0];
ebcf34f3 658 /* public: */
4031ae6e 659
233577a2
HFS
660/* if you move pkt_type around you also must adapt those constants */
661#ifdef __BIG_ENDIAN_BITFIELD
662#define PKT_TYPE_MAX (7 << 5)
663#else
664#define PKT_TYPE_MAX 7
1da177e4 665#endif
233577a2 666#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 667
233577a2 668 __u8 __pkt_type_offset[0];
b1937227 669 __u8 pkt_type:3;
c93bdd0e 670 __u8 pfmemalloc:1;
b1937227 671 __u8 ignore_df:1;
b1937227
ED
672
673 __u8 nf_trace:1;
674 __u8 ip_summed:2;
3853b584 675 __u8 ooo_okay:1;
61b905da 676 __u8 l4_hash:1;
a3b18ddb 677 __u8 sw_hash:1;
6e3e939f
JB
678 __u8 wifi_acked_valid:1;
679 __u8 wifi_acked:1;
b1937227 680
3bdc0eba 681 __u8 no_fcs:1;
77cffe23 682 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 683 __u8 encapsulation:1;
7e2b10c1 684 __u8 encap_hdr_csum:1;
5d0c2b95 685 __u8 csum_valid:1;
7e3cead5 686 __u8 csum_complete_sw:1;
b1937227
ED
687 __u8 csum_level:2;
688 __u8 csum_bad:1;
fe55f6d5 689
4ff06203 690 __u8 dst_pending_confirm:1;
b1937227
ED
691#ifdef CONFIG_IPV6_NDISC_NODETYPE
692 __u8 ndisc_nodetype:2;
693#endif
694 __u8 ipvs_property:1;
8bce6d7d 695 __u8 inner_protocol_type:1;
e585f236 696 __u8 remcsum_offload:1;
6bc506b4
IS
697#ifdef CONFIG_NET_SWITCHDEV
698 __u8 offload_fwd_mark:1;
699#endif
e7246e12
WB
700#ifdef CONFIG_NET_CLS_ACT
701 __u8 tc_skip_classify:1;
8dc07fdb 702 __u8 tc_at_ingress:1;
bc31c905
WB
703 __u8 tc_redirected:1;
704 __u8 tc_from_ingress:1;
e7246e12 705#endif
b1937227
ED
706
707#ifdef CONFIG_NET_SCHED
708 __u16 tc_index; /* traffic control index */
b1937227 709#endif
fe55f6d5 710
b1937227
ED
711 union {
712 __wsum csum;
713 struct {
714 __u16 csum_start;
715 __u16 csum_offset;
716 };
717 };
718 __u32 priority;
719 int skb_iif;
720 __u32 hash;
721 __be16 vlan_proto;
722 __u16 vlan_tci;
2bd82484
ED
723#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
724 union {
725 unsigned int napi_id;
726 unsigned int sender_cpu;
727 };
97fc2f08 728#endif
984bc16c 729#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 730 __u32 secmark;
0c4f691f 731#endif
0c4f691f 732
3b885787
NH
733 union {
734 __u32 mark;
16fad69c 735 __u32 reserved_tailroom;
3b885787 736 };
1da177e4 737
8bce6d7d
TH
738 union {
739 __be16 inner_protocol;
740 __u8 inner_ipproto;
741 };
742
1a37e412
SH
743 __u16 inner_transport_header;
744 __u16 inner_network_header;
745 __u16 inner_mac_header;
b1937227
ED
746
747 __be16 protocol;
1a37e412
SH
748 __u16 transport_header;
749 __u16 network_header;
750 __u16 mac_header;
b1937227 751
ebcf34f3 752 /* private: */
b1937227 753 __u32 headers_end[0];
ebcf34f3 754 /* public: */
b1937227 755
1da177e4 756 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 757 sk_buff_data_t tail;
4305b541 758 sk_buff_data_t end;
1da177e4 759 unsigned char *head,
4305b541 760 *data;
27a884dc
ACM
761 unsigned int truesize;
762 atomic_t users;
1da177e4
LT
763};
764
765#ifdef __KERNEL__
766/*
767 * Handling routines are only of interest to the kernel
768 */
769#include <linux/slab.h>
770
1da177e4 771
c93bdd0e
MG
772#define SKB_ALLOC_FCLONE 0x01
773#define SKB_ALLOC_RX 0x02
fd11a83d 774#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
775
776/* Returns true if the skb was allocated from PFMEMALLOC reserves */
777static inline bool skb_pfmemalloc(const struct sk_buff *skb)
778{
779 return unlikely(skb->pfmemalloc);
780}
781
7fee226a
ED
782/*
783 * skb might have a dst pointer attached, refcounted or not.
784 * _skb_refdst low order bit is set if refcount was _not_ taken
785 */
786#define SKB_DST_NOREF 1UL
787#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
788
a9e419dc 789#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
790/**
791 * skb_dst - returns skb dst_entry
792 * @skb: buffer
793 *
794 * Returns skb dst_entry, regardless of reference taken or not.
795 */
adf30907
ED
796static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
797{
7fee226a
ED
798 /* If refdst was not refcounted, check we still are in a
799 * rcu_read_lock section
800 */
801 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
802 !rcu_read_lock_held() &&
803 !rcu_read_lock_bh_held());
804 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
805}
806
7fee226a
ED
807/**
808 * skb_dst_set - sets skb dst
809 * @skb: buffer
810 * @dst: dst entry
811 *
812 * Sets skb dst, assuming a reference was taken on dst and should
813 * be released by skb_dst_drop()
814 */
adf30907
ED
815static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
816{
7fee226a
ED
817 skb->_skb_refdst = (unsigned long)dst;
818}
819
932bc4d7
JA
820/**
821 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
822 * @skb: buffer
823 * @dst: dst entry
824 *
825 * Sets skb dst, assuming a reference was not taken on dst.
826 * If dst entry is cached, we do not take reference and dst_release
827 * will be avoided by refdst_drop. If dst entry is not cached, we take
828 * reference, so that last dst_release can destroy the dst immediately.
829 */
830static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
831{
dbfc4fb7
HFS
832 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
833 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 834}
7fee226a
ED
835
836/**
25985edc 837 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
838 * @skb: buffer
839 */
840static inline bool skb_dst_is_noref(const struct sk_buff *skb)
841{
842 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
843}
844
511c3f92
ED
845static inline struct rtable *skb_rtable(const struct sk_buff *skb)
846{
adf30907 847 return (struct rtable *)skb_dst(skb);
511c3f92
ED
848}
849
8b10cab6
JHS
850/* For mangling skb->pkt_type from user space side from applications
851 * such as nft, tc, etc, we only allow a conservative subset of
852 * possible pkt_types to be set.
853*/
854static inline bool skb_pkt_type_ok(u32 ptype)
855{
856 return ptype <= PACKET_OTHERHOST;
857}
858
7965bd4d
JP
859void kfree_skb(struct sk_buff *skb);
860void kfree_skb_list(struct sk_buff *segs);
861void skb_tx_error(struct sk_buff *skb);
862void consume_skb(struct sk_buff *skb);
863void __kfree_skb(struct sk_buff *skb);
d7e8883c 864extern struct kmem_cache *skbuff_head_cache;
bad43ca8 865
7965bd4d
JP
866void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
867bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
868 bool *fragstolen, int *delta_truesize);
bad43ca8 869
7965bd4d
JP
870struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
871 int node);
2ea2f62c 872struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 873struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 874static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 875 gfp_t priority)
d179cd12 876{
564824b0 877 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
878}
879
2e4e4410
ED
880struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
881 unsigned long data_len,
882 int max_page_order,
883 int *errcode,
884 gfp_t gfp_mask);
885
d0bf4a9e
ED
886/* Layout of fast clones : [skb1][skb2][fclone_ref] */
887struct sk_buff_fclones {
888 struct sk_buff skb1;
889
890 struct sk_buff skb2;
891
892 atomic_t fclone_ref;
893};
894
895/**
896 * skb_fclone_busy - check if fclone is busy
293de7de 897 * @sk: socket
d0bf4a9e
ED
898 * @skb: buffer
899 *
bda13fed 900 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
901 * Some drivers call skb_orphan() in their ndo_start_xmit(),
902 * so we also check that this didnt happen.
d0bf4a9e 903 */
39bb5e62
ED
904static inline bool skb_fclone_busy(const struct sock *sk,
905 const struct sk_buff *skb)
d0bf4a9e
ED
906{
907 const struct sk_buff_fclones *fclones;
908
909 fclones = container_of(skb, struct sk_buff_fclones, skb1);
910
911 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 912 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 913 fclones->skb2.sk == sk;
d0bf4a9e
ED
914}
915
d179cd12 916static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 917 gfp_t priority)
d179cd12 918{
c93bdd0e 919 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
920}
921
7965bd4d 922struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
923static inline struct sk_buff *alloc_skb_head(gfp_t priority)
924{
925 return __alloc_skb_head(priority, -1);
926}
927
7965bd4d
JP
928struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
929int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
930struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
931struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
932struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
933 gfp_t gfp_mask, bool fclone);
934static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
935 gfp_t gfp_mask)
936{
937 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
938}
7965bd4d
JP
939
940int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
941struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
942 unsigned int headroom);
943struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
944 int newtailroom, gfp_t priority);
25a91d8d
FD
945int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
946 int offset, int len);
7965bd4d
JP
947int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
948 int len);
949int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
950int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 951#define dev_kfree_skb(a) consume_skb(a)
1da177e4 952
7965bd4d
JP
953int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
954 int getfrag(void *from, char *to, int offset,
955 int len, int odd, struct sk_buff *skb),
956 void *from, int length);
e89e9cf5 957
be12a1fe
HFS
958int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
959 int offset, size_t size);
960
d94d9fee 961struct skb_seq_state {
677e90ed
TG
962 __u32 lower_offset;
963 __u32 upper_offset;
964 __u32 frag_idx;
965 __u32 stepped_offset;
966 struct sk_buff *root_skb;
967 struct sk_buff *cur_skb;
968 __u8 *frag_data;
969};
970
7965bd4d
JP
971void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
972 unsigned int to, struct skb_seq_state *st);
973unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
974 struct skb_seq_state *st);
975void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 976
7965bd4d 977unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 978 unsigned int to, struct ts_config *config);
3fc7e8a6 979
09323cc4
TH
980/*
981 * Packet hash types specify the type of hash in skb_set_hash.
982 *
983 * Hash types refer to the protocol layer addresses which are used to
984 * construct a packet's hash. The hashes are used to differentiate or identify
985 * flows of the protocol layer for the hash type. Hash types are either
986 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
987 *
988 * Properties of hashes:
989 *
990 * 1) Two packets in different flows have different hash values
991 * 2) Two packets in the same flow should have the same hash value
992 *
993 * A hash at a higher layer is considered to be more specific. A driver should
994 * set the most specific hash possible.
995 *
996 * A driver cannot indicate a more specific hash than the layer at which a hash
997 * was computed. For instance an L3 hash cannot be set as an L4 hash.
998 *
999 * A driver may indicate a hash level which is less specific than the
1000 * actual layer the hash was computed on. For instance, a hash computed
1001 * at L4 may be considered an L3 hash. This should only be done if the
1002 * driver can't unambiguously determine that the HW computed the hash at
1003 * the higher layer. Note that the "should" in the second property above
1004 * permits this.
1005 */
1006enum pkt_hash_types {
1007 PKT_HASH_TYPE_NONE, /* Undefined type */
1008 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1009 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1010 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1011};
1012
bcc83839 1013static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1014{
bcc83839 1015 skb->hash = 0;
a3b18ddb 1016 skb->sw_hash = 0;
bcc83839
TH
1017 skb->l4_hash = 0;
1018}
1019
1020static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1021{
1022 if (!skb->l4_hash)
1023 skb_clear_hash(skb);
1024}
1025
1026static inline void
1027__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1028{
1029 skb->l4_hash = is_l4;
1030 skb->sw_hash = is_sw;
61b905da 1031 skb->hash = hash;
09323cc4
TH
1032}
1033
bcc83839
TH
1034static inline void
1035skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1036{
1037 /* Used by drivers to set hash from HW */
1038 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1039}
1040
1041static inline void
1042__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1043{
1044 __skb_set_hash(skb, hash, true, is_l4);
1045}
1046
e5276937 1047void __skb_get_hash(struct sk_buff *skb);
b917783c 1048u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1049u32 skb_get_poff(const struct sk_buff *skb);
1050u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1051 const struct flow_keys *keys, int hlen);
1052__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1053 void *data, int hlen_proto);
1054
1055static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1056 int thoff, u8 ip_proto)
1057{
1058 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1059}
1060
1061void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1062 const struct flow_dissector_key *key,
1063 unsigned int key_count);
1064
1065bool __skb_flow_dissect(const struct sk_buff *skb,
1066 struct flow_dissector *flow_dissector,
1067 void *target_container,
cd79a238
TH
1068 void *data, __be16 proto, int nhoff, int hlen,
1069 unsigned int flags);
e5276937
TH
1070
1071static inline bool skb_flow_dissect(const struct sk_buff *skb,
1072 struct flow_dissector *flow_dissector,
cd79a238 1073 void *target_container, unsigned int flags)
e5276937
TH
1074{
1075 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1076 NULL, 0, 0, 0, flags);
e5276937
TH
1077}
1078
1079static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1080 struct flow_keys *flow,
1081 unsigned int flags)
e5276937
TH
1082{
1083 memset(flow, 0, sizeof(*flow));
1084 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1085 NULL, 0, 0, 0, flags);
e5276937
TH
1086}
1087
1088static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1089 void *data, __be16 proto,
cd79a238
TH
1090 int nhoff, int hlen,
1091 unsigned int flags)
e5276937
TH
1092{
1093 memset(flow, 0, sizeof(*flow));
1094 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1095 data, proto, nhoff, hlen, flags);
e5276937
TH
1096}
1097
3958afa1 1098static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1099{
a3b18ddb 1100 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1101 __skb_get_hash(skb);
bfb564e7 1102
61b905da 1103 return skb->hash;
bfb564e7
KK
1104}
1105
20a17bf6 1106__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1107
20a17bf6 1108static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1109{
c6cc1ca7
TH
1110 if (!skb->l4_hash && !skb->sw_hash) {
1111 struct flow_keys keys;
de4c1f8b 1112 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1113
de4c1f8b 1114 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1115 }
f70ea018
TH
1116
1117 return skb->hash;
1118}
1119
20a17bf6 1120__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1121
20a17bf6 1122static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1123{
c6cc1ca7
TH
1124 if (!skb->l4_hash && !skb->sw_hash) {
1125 struct flow_keys keys;
de4c1f8b 1126 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1127
de4c1f8b 1128 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1129 }
f70ea018
TH
1130
1131 return skb->hash;
1132}
1133
50fb7992
TH
1134__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1135
57bdf7f4
TH
1136static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1137{
61b905da 1138 return skb->hash;
57bdf7f4
TH
1139}
1140
3df7a74e
TH
1141static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1142{
61b905da 1143 to->hash = from->hash;
a3b18ddb 1144 to->sw_hash = from->sw_hash;
61b905da 1145 to->l4_hash = from->l4_hash;
3df7a74e
TH
1146};
1147
4305b541
ACM
1148#ifdef NET_SKBUFF_DATA_USES_OFFSET
1149static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1150{
1151 return skb->head + skb->end;
1152}
ec47ea82
AD
1153
1154static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1155{
1156 return skb->end;
1157}
4305b541
ACM
1158#else
1159static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1160{
1161 return skb->end;
1162}
ec47ea82
AD
1163
1164static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1165{
1166 return skb->end - skb->head;
1167}
4305b541
ACM
1168#endif
1169
1da177e4 1170/* Internal */
4305b541 1171#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1172
ac45f602
PO
1173static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1174{
1175 return &skb_shinfo(skb)->hwtstamps;
1176}
1177
1da177e4
LT
1178/**
1179 * skb_queue_empty - check if a queue is empty
1180 * @list: queue head
1181 *
1182 * Returns true if the queue is empty, false otherwise.
1183 */
1184static inline int skb_queue_empty(const struct sk_buff_head *list)
1185{
fd44b93c 1186 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1187}
1188
fc7ebb21
DM
1189/**
1190 * skb_queue_is_last - check if skb is the last entry in the queue
1191 * @list: queue head
1192 * @skb: buffer
1193 *
1194 * Returns true if @skb is the last buffer on the list.
1195 */
1196static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1197 const struct sk_buff *skb)
1198{
fd44b93c 1199 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1200}
1201
832d11c5
IJ
1202/**
1203 * skb_queue_is_first - check if skb is the first entry in the queue
1204 * @list: queue head
1205 * @skb: buffer
1206 *
1207 * Returns true if @skb is the first buffer on the list.
1208 */
1209static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1210 const struct sk_buff *skb)
1211{
fd44b93c 1212 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1213}
1214
249c8b42
DM
1215/**
1216 * skb_queue_next - return the next packet in the queue
1217 * @list: queue head
1218 * @skb: current buffer
1219 *
1220 * Return the next packet in @list after @skb. It is only valid to
1221 * call this if skb_queue_is_last() evaluates to false.
1222 */
1223static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1224 const struct sk_buff *skb)
1225{
1226 /* This BUG_ON may seem severe, but if we just return then we
1227 * are going to dereference garbage.
1228 */
1229 BUG_ON(skb_queue_is_last(list, skb));
1230 return skb->next;
1231}
1232
832d11c5
IJ
1233/**
1234 * skb_queue_prev - return the prev packet in the queue
1235 * @list: queue head
1236 * @skb: current buffer
1237 *
1238 * Return the prev packet in @list before @skb. It is only valid to
1239 * call this if skb_queue_is_first() evaluates to false.
1240 */
1241static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1242 const struct sk_buff *skb)
1243{
1244 /* This BUG_ON may seem severe, but if we just return then we
1245 * are going to dereference garbage.
1246 */
1247 BUG_ON(skb_queue_is_first(list, skb));
1248 return skb->prev;
1249}
1250
1da177e4
LT
1251/**
1252 * skb_get - reference buffer
1253 * @skb: buffer to reference
1254 *
1255 * Makes another reference to a socket buffer and returns a pointer
1256 * to the buffer.
1257 */
1258static inline struct sk_buff *skb_get(struct sk_buff *skb)
1259{
1260 atomic_inc(&skb->users);
1261 return skb;
1262}
1263
1264/*
1265 * If users == 1, we are the only owner and are can avoid redundant
1266 * atomic change.
1267 */
1268
1da177e4
LT
1269/**
1270 * skb_cloned - is the buffer a clone
1271 * @skb: buffer to check
1272 *
1273 * Returns true if the buffer was generated with skb_clone() and is
1274 * one of multiple shared copies of the buffer. Cloned buffers are
1275 * shared data so must not be written to under normal circumstances.
1276 */
1277static inline int skb_cloned(const struct sk_buff *skb)
1278{
1279 return skb->cloned &&
1280 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1281}
1282
14bbd6a5
PS
1283static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1284{
d0164adc 1285 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1286
1287 if (skb_cloned(skb))
1288 return pskb_expand_head(skb, 0, 0, pri);
1289
1290 return 0;
1291}
1292
1da177e4
LT
1293/**
1294 * skb_header_cloned - is the header a clone
1295 * @skb: buffer to check
1296 *
1297 * Returns true if modifying the header part of the buffer requires
1298 * the data to be copied.
1299 */
1300static inline int skb_header_cloned(const struct sk_buff *skb)
1301{
1302 int dataref;
1303
1304 if (!skb->cloned)
1305 return 0;
1306
1307 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1308 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1309 return dataref != 1;
1310}
1311
9580bf2e
ED
1312static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1313{
1314 might_sleep_if(gfpflags_allow_blocking(pri));
1315
1316 if (skb_header_cloned(skb))
1317 return pskb_expand_head(skb, 0, 0, pri);
1318
1319 return 0;
1320}
1321
1da177e4
LT
1322/**
1323 * skb_header_release - release reference to header
1324 * @skb: buffer to operate on
1325 *
1326 * Drop a reference to the header part of the buffer. This is done
1327 * by acquiring a payload reference. You must not read from the header
1328 * part of skb->data after this.
f4a775d1 1329 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1330 */
1331static inline void skb_header_release(struct sk_buff *skb)
1332{
1333 BUG_ON(skb->nohdr);
1334 skb->nohdr = 1;
1335 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1336}
1337
f4a775d1
ED
1338/**
1339 * __skb_header_release - release reference to header
1340 * @skb: buffer to operate on
1341 *
1342 * Variant of skb_header_release() assuming skb is private to caller.
1343 * We can avoid one atomic operation.
1344 */
1345static inline void __skb_header_release(struct sk_buff *skb)
1346{
1347 skb->nohdr = 1;
1348 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1349}
1350
1351
1da177e4
LT
1352/**
1353 * skb_shared - is the buffer shared
1354 * @skb: buffer to check
1355 *
1356 * Returns true if more than one person has a reference to this
1357 * buffer.
1358 */
1359static inline int skb_shared(const struct sk_buff *skb)
1360{
1361 return atomic_read(&skb->users) != 1;
1362}
1363
1364/**
1365 * skb_share_check - check if buffer is shared and if so clone it
1366 * @skb: buffer to check
1367 * @pri: priority for memory allocation
1368 *
1369 * If the buffer is shared the buffer is cloned and the old copy
1370 * drops a reference. A new clone with a single reference is returned.
1371 * If the buffer is not shared the original buffer is returned. When
1372 * being called from interrupt status or with spinlocks held pri must
1373 * be GFP_ATOMIC.
1374 *
1375 * NULL is returned on a memory allocation failure.
1376 */
47061bc4 1377static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1378{
d0164adc 1379 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1380 if (skb_shared(skb)) {
1381 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1382
1383 if (likely(nskb))
1384 consume_skb(skb);
1385 else
1386 kfree_skb(skb);
1da177e4
LT
1387 skb = nskb;
1388 }
1389 return skb;
1390}
1391
1392/*
1393 * Copy shared buffers into a new sk_buff. We effectively do COW on
1394 * packets to handle cases where we have a local reader and forward
1395 * and a couple of other messy ones. The normal one is tcpdumping
1396 * a packet thats being forwarded.
1397 */
1398
1399/**
1400 * skb_unshare - make a copy of a shared buffer
1401 * @skb: buffer to check
1402 * @pri: priority for memory allocation
1403 *
1404 * If the socket buffer is a clone then this function creates a new
1405 * copy of the data, drops a reference count on the old copy and returns
1406 * the new copy with the reference count at 1. If the buffer is not a clone
1407 * the original buffer is returned. When called with a spinlock held or
1408 * from interrupt state @pri must be %GFP_ATOMIC
1409 *
1410 * %NULL is returned on a memory allocation failure.
1411 */
e2bf521d 1412static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1413 gfp_t pri)
1da177e4 1414{
d0164adc 1415 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1416 if (skb_cloned(skb)) {
1417 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1418
1419 /* Free our shared copy */
1420 if (likely(nskb))
1421 consume_skb(skb);
1422 else
1423 kfree_skb(skb);
1da177e4
LT
1424 skb = nskb;
1425 }
1426 return skb;
1427}
1428
1429/**
1a5778aa 1430 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1431 * @list_: list to peek at
1432 *
1433 * Peek an &sk_buff. Unlike most other operations you _MUST_
1434 * be careful with this one. A peek leaves the buffer on the
1435 * list and someone else may run off with it. You must hold
1436 * the appropriate locks or have a private queue to do this.
1437 *
1438 * Returns %NULL for an empty list or a pointer to the head element.
1439 * The reference count is not incremented and the reference is therefore
1440 * volatile. Use with caution.
1441 */
05bdd2f1 1442static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1443{
18d07000
ED
1444 struct sk_buff *skb = list_->next;
1445
1446 if (skb == (struct sk_buff *)list_)
1447 skb = NULL;
1448 return skb;
1da177e4
LT
1449}
1450
da5ef6e5
PE
1451/**
1452 * skb_peek_next - peek skb following the given one from a queue
1453 * @skb: skb to start from
1454 * @list_: list to peek at
1455 *
1456 * Returns %NULL when the end of the list is met or a pointer to the
1457 * next element. The reference count is not incremented and the
1458 * reference is therefore volatile. Use with caution.
1459 */
1460static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1461 const struct sk_buff_head *list_)
1462{
1463 struct sk_buff *next = skb->next;
18d07000 1464
da5ef6e5
PE
1465 if (next == (struct sk_buff *)list_)
1466 next = NULL;
1467 return next;
1468}
1469
1da177e4 1470/**
1a5778aa 1471 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1472 * @list_: list to peek at
1473 *
1474 * Peek an &sk_buff. Unlike most other operations you _MUST_
1475 * be careful with this one. A peek leaves the buffer on the
1476 * list and someone else may run off with it. You must hold
1477 * the appropriate locks or have a private queue to do this.
1478 *
1479 * Returns %NULL for an empty list or a pointer to the tail element.
1480 * The reference count is not incremented and the reference is therefore
1481 * volatile. Use with caution.
1482 */
05bdd2f1 1483static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1484{
18d07000
ED
1485 struct sk_buff *skb = list_->prev;
1486
1487 if (skb == (struct sk_buff *)list_)
1488 skb = NULL;
1489 return skb;
1490
1da177e4
LT
1491}
1492
1493/**
1494 * skb_queue_len - get queue length
1495 * @list_: list to measure
1496 *
1497 * Return the length of an &sk_buff queue.
1498 */
1499static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1500{
1501 return list_->qlen;
1502}
1503
67fed459
DM
1504/**
1505 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1506 * @list: queue to initialize
1507 *
1508 * This initializes only the list and queue length aspects of
1509 * an sk_buff_head object. This allows to initialize the list
1510 * aspects of an sk_buff_head without reinitializing things like
1511 * the spinlock. It can also be used for on-stack sk_buff_head
1512 * objects where the spinlock is known to not be used.
1513 */
1514static inline void __skb_queue_head_init(struct sk_buff_head *list)
1515{
1516 list->prev = list->next = (struct sk_buff *)list;
1517 list->qlen = 0;
1518}
1519
76f10ad0
AV
1520/*
1521 * This function creates a split out lock class for each invocation;
1522 * this is needed for now since a whole lot of users of the skb-queue
1523 * infrastructure in drivers have different locking usage (in hardirq)
1524 * than the networking core (in softirq only). In the long run either the
1525 * network layer or drivers should need annotation to consolidate the
1526 * main types of usage into 3 classes.
1527 */
1da177e4
LT
1528static inline void skb_queue_head_init(struct sk_buff_head *list)
1529{
1530 spin_lock_init(&list->lock);
67fed459 1531 __skb_queue_head_init(list);
1da177e4
LT
1532}
1533
c2ecba71
PE
1534static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1535 struct lock_class_key *class)
1536{
1537 skb_queue_head_init(list);
1538 lockdep_set_class(&list->lock, class);
1539}
1540
1da177e4 1541/*
bf299275 1542 * Insert an sk_buff on a list.
1da177e4
LT
1543 *
1544 * The "__skb_xxxx()" functions are the non-atomic ones that
1545 * can only be called with interrupts disabled.
1546 */
7965bd4d
JP
1547void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1548 struct sk_buff_head *list);
bf299275
GR
1549static inline void __skb_insert(struct sk_buff *newsk,
1550 struct sk_buff *prev, struct sk_buff *next,
1551 struct sk_buff_head *list)
1552{
1553 newsk->next = next;
1554 newsk->prev = prev;
1555 next->prev = prev->next = newsk;
1556 list->qlen++;
1557}
1da177e4 1558
67fed459
DM
1559static inline void __skb_queue_splice(const struct sk_buff_head *list,
1560 struct sk_buff *prev,
1561 struct sk_buff *next)
1562{
1563 struct sk_buff *first = list->next;
1564 struct sk_buff *last = list->prev;
1565
1566 first->prev = prev;
1567 prev->next = first;
1568
1569 last->next = next;
1570 next->prev = last;
1571}
1572
1573/**
1574 * skb_queue_splice - join two skb lists, this is designed for stacks
1575 * @list: the new list to add
1576 * @head: the place to add it in the first list
1577 */
1578static inline void skb_queue_splice(const struct sk_buff_head *list,
1579 struct sk_buff_head *head)
1580{
1581 if (!skb_queue_empty(list)) {
1582 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1583 head->qlen += list->qlen;
67fed459
DM
1584 }
1585}
1586
1587/**
d9619496 1588 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1589 * @list: the new list to add
1590 * @head: the place to add it in the first list
1591 *
1592 * The list at @list is reinitialised
1593 */
1594static inline void skb_queue_splice_init(struct sk_buff_head *list,
1595 struct sk_buff_head *head)
1596{
1597 if (!skb_queue_empty(list)) {
1598 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1599 head->qlen += list->qlen;
67fed459
DM
1600 __skb_queue_head_init(list);
1601 }
1602}
1603
1604/**
1605 * skb_queue_splice_tail - join two skb lists, each list being a queue
1606 * @list: the new list to add
1607 * @head: the place to add it in the first list
1608 */
1609static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1610 struct sk_buff_head *head)
1611{
1612 if (!skb_queue_empty(list)) {
1613 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1614 head->qlen += list->qlen;
67fed459
DM
1615 }
1616}
1617
1618/**
d9619496 1619 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1620 * @list: the new list to add
1621 * @head: the place to add it in the first list
1622 *
1623 * Each of the lists is a queue.
1624 * The list at @list is reinitialised
1625 */
1626static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1627 struct sk_buff_head *head)
1628{
1629 if (!skb_queue_empty(list)) {
1630 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1631 head->qlen += list->qlen;
67fed459
DM
1632 __skb_queue_head_init(list);
1633 }
1634}
1635
1da177e4 1636/**
300ce174 1637 * __skb_queue_after - queue a buffer at the list head
1da177e4 1638 * @list: list to use
300ce174 1639 * @prev: place after this buffer
1da177e4
LT
1640 * @newsk: buffer to queue
1641 *
300ce174 1642 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1643 * and you must therefore hold required locks before calling it.
1644 *
1645 * A buffer cannot be placed on two lists at the same time.
1646 */
300ce174
SH
1647static inline void __skb_queue_after(struct sk_buff_head *list,
1648 struct sk_buff *prev,
1649 struct sk_buff *newsk)
1da177e4 1650{
bf299275 1651 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1652}
1653
7965bd4d
JP
1654void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1655 struct sk_buff_head *list);
7de6c033 1656
f5572855
GR
1657static inline void __skb_queue_before(struct sk_buff_head *list,
1658 struct sk_buff *next,
1659 struct sk_buff *newsk)
1660{
1661 __skb_insert(newsk, next->prev, next, list);
1662}
1663
300ce174
SH
1664/**
1665 * __skb_queue_head - queue a buffer at the list head
1666 * @list: list to use
1667 * @newsk: buffer to queue
1668 *
1669 * Queue a buffer at the start of a list. This function takes no locks
1670 * and you must therefore hold required locks before calling it.
1671 *
1672 * A buffer cannot be placed on two lists at the same time.
1673 */
7965bd4d 1674void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1675static inline void __skb_queue_head(struct sk_buff_head *list,
1676 struct sk_buff *newsk)
1677{
1678 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1679}
1680
1da177e4
LT
1681/**
1682 * __skb_queue_tail - queue a buffer at the list tail
1683 * @list: list to use
1684 * @newsk: buffer to queue
1685 *
1686 * Queue a buffer at the end of a list. This function takes no locks
1687 * and you must therefore hold required locks before calling it.
1688 *
1689 * A buffer cannot be placed on two lists at the same time.
1690 */
7965bd4d 1691void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1692static inline void __skb_queue_tail(struct sk_buff_head *list,
1693 struct sk_buff *newsk)
1694{
f5572855 1695 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1696}
1697
1da177e4
LT
1698/*
1699 * remove sk_buff from list. _Must_ be called atomically, and with
1700 * the list known..
1701 */
7965bd4d 1702void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1703static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1704{
1705 struct sk_buff *next, *prev;
1706
1707 list->qlen--;
1708 next = skb->next;
1709 prev = skb->prev;
1710 skb->next = skb->prev = NULL;
1da177e4
LT
1711 next->prev = prev;
1712 prev->next = next;
1713}
1714
f525c06d
GR
1715/**
1716 * __skb_dequeue - remove from the head of the queue
1717 * @list: list to dequeue from
1718 *
1719 * Remove the head of the list. This function does not take any locks
1720 * so must be used with appropriate locks held only. The head item is
1721 * returned or %NULL if the list is empty.
1722 */
7965bd4d 1723struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1724static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1725{
1726 struct sk_buff *skb = skb_peek(list);
1727 if (skb)
1728 __skb_unlink(skb, list);
1729 return skb;
1730}
1da177e4
LT
1731
1732/**
1733 * __skb_dequeue_tail - remove from the tail of the queue
1734 * @list: list to dequeue from
1735 *
1736 * Remove the tail of the list. This function does not take any locks
1737 * so must be used with appropriate locks held only. The tail item is
1738 * returned or %NULL if the list is empty.
1739 */
7965bd4d 1740struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1741static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1742{
1743 struct sk_buff *skb = skb_peek_tail(list);
1744 if (skb)
1745 __skb_unlink(skb, list);
1746 return skb;
1747}
1748
1749
bdcc0924 1750static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1751{
1752 return skb->data_len;
1753}
1754
1755static inline unsigned int skb_headlen(const struct sk_buff *skb)
1756{
1757 return skb->len - skb->data_len;
1758}
1759
c72d8cda 1760static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1da177e4 1761{
c72d8cda 1762 unsigned int i, len = 0;
1da177e4 1763
c72d8cda 1764 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 1765 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1766 return len + skb_headlen(skb);
1767}
1768
131ea667
IC
1769/**
1770 * __skb_fill_page_desc - initialise a paged fragment in an skb
1771 * @skb: buffer containing fragment to be initialised
1772 * @i: paged fragment index to initialise
1773 * @page: the page to use for this fragment
1774 * @off: the offset to the data with @page
1775 * @size: the length of the data
1776 *
1777 * Initialises the @i'th fragment of @skb to point to &size bytes at
1778 * offset @off within @page.
1779 *
1780 * Does not take any additional reference on the fragment.
1781 */
1782static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1783 struct page *page, int off, int size)
1da177e4
LT
1784{
1785 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1786
c48a11c7 1787 /*
2f064f34
MH
1788 * Propagate page pfmemalloc to the skb if we can. The problem is
1789 * that not all callers have unique ownership of the page but rely
1790 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1791 */
a8605c60 1792 frag->page.p = page;
1da177e4 1793 frag->page_offset = off;
9e903e08 1794 skb_frag_size_set(frag, size);
cca7af38
PE
1795
1796 page = compound_head(page);
2f064f34 1797 if (page_is_pfmemalloc(page))
cca7af38 1798 skb->pfmemalloc = true;
131ea667
IC
1799}
1800
1801/**
1802 * skb_fill_page_desc - initialise a paged fragment in an skb
1803 * @skb: buffer containing fragment to be initialised
1804 * @i: paged fragment index to initialise
1805 * @page: the page to use for this fragment
1806 * @off: the offset to the data with @page
1807 * @size: the length of the data
1808 *
1809 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1810 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1811 * addition updates @skb such that @i is the last fragment.
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)
1817{
1818 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1819 skb_shinfo(skb)->nr_frags = i + 1;
1820}
1821
7965bd4d
JP
1822void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1823 int size, unsigned int truesize);
654bed16 1824
f8e617e1
JW
1825void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1826 unsigned int truesize);
1827
1da177e4 1828#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1829#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1830#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1831
27a884dc
ACM
1832#ifdef NET_SKBUFF_DATA_USES_OFFSET
1833static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1834{
1835 return skb->head + skb->tail;
1836}
1837
1838static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1839{
1840 skb->tail = skb->data - skb->head;
1841}
1842
1843static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1844{
1845 skb_reset_tail_pointer(skb);
1846 skb->tail += offset;
1847}
7cc46190 1848
27a884dc
ACM
1849#else /* NET_SKBUFF_DATA_USES_OFFSET */
1850static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1851{
1852 return skb->tail;
1853}
1854
1855static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1856{
1857 skb->tail = skb->data;
1858}
1859
1860static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1861{
1862 skb->tail = skb->data + offset;
1863}
4305b541 1864
27a884dc
ACM
1865#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1866
1da177e4
LT
1867/*
1868 * Add data to an sk_buff
1869 */
0c7ddf36 1870unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1871unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1872static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1873{
27a884dc 1874 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1875 SKB_LINEAR_ASSERT(skb);
1876 skb->tail += len;
1877 skb->len += len;
1878 return tmp;
1879}
1880
7965bd4d 1881unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1882static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1883{
1884 skb->data -= len;
1885 skb->len += len;
1886 return skb->data;
1887}
1888
7965bd4d 1889unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1890static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1891{
1892 skb->len -= len;
1893 BUG_ON(skb->len < skb->data_len);
1894 return skb->data += len;
1895}
1896
47d29646
DM
1897static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1898{
1899 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1900}
1901
7965bd4d 1902unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1903
1904static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1905{
1906 if (len > skb_headlen(skb) &&
987c402a 1907 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1908 return NULL;
1909 skb->len -= len;
1910 return skb->data += len;
1911}
1912
1913static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1914{
1915 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1916}
1917
1918static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1919{
1920 if (likely(len <= skb_headlen(skb)))
1921 return 1;
1922 if (unlikely(len > skb->len))
1923 return 0;
987c402a 1924 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1925}
1926
c8c8b127
ED
1927void skb_condense(struct sk_buff *skb);
1928
1da177e4
LT
1929/**
1930 * skb_headroom - bytes at buffer head
1931 * @skb: buffer to check
1932 *
1933 * Return the number of bytes of free space at the head of an &sk_buff.
1934 */
c2636b4d 1935static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1936{
1937 return skb->data - skb->head;
1938}
1939
1940/**
1941 * skb_tailroom - bytes at buffer end
1942 * @skb: buffer to check
1943 *
1944 * Return the number of bytes of free space at the tail of an sk_buff
1945 */
1946static inline int skb_tailroom(const struct sk_buff *skb)
1947{
4305b541 1948 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1949}
1950
a21d4572
ED
1951/**
1952 * skb_availroom - bytes at buffer end
1953 * @skb: buffer to check
1954 *
1955 * Return the number of bytes of free space at the tail of an sk_buff
1956 * allocated by sk_stream_alloc()
1957 */
1958static inline int skb_availroom(const struct sk_buff *skb)
1959{
16fad69c
ED
1960 if (skb_is_nonlinear(skb))
1961 return 0;
1962
1963 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1964}
1965
1da177e4
LT
1966/**
1967 * skb_reserve - adjust headroom
1968 * @skb: buffer to alter
1969 * @len: bytes to move
1970 *
1971 * Increase the headroom of an empty &sk_buff by reducing the tail
1972 * room. This is only allowed for an empty buffer.
1973 */
8243126c 1974static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1975{
1976 skb->data += len;
1977 skb->tail += len;
1978}
1979
1837b2e2
BP
1980/**
1981 * skb_tailroom_reserve - adjust reserved_tailroom
1982 * @skb: buffer to alter
1983 * @mtu: maximum amount of headlen permitted
1984 * @needed_tailroom: minimum amount of reserved_tailroom
1985 *
1986 * Set reserved_tailroom so that headlen can be as large as possible but
1987 * not larger than mtu and tailroom cannot be smaller than
1988 * needed_tailroom.
1989 * The required headroom should already have been reserved before using
1990 * this function.
1991 */
1992static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
1993 unsigned int needed_tailroom)
1994{
1995 SKB_LINEAR_ASSERT(skb);
1996 if (mtu < skb_tailroom(skb) - needed_tailroom)
1997 /* use at most mtu */
1998 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
1999 else
2000 /* use up to all available space */
2001 skb->reserved_tailroom = needed_tailroom;
2002}
2003
8bce6d7d
TH
2004#define ENCAP_TYPE_ETHER 0
2005#define ENCAP_TYPE_IPPROTO 1
2006
2007static inline void skb_set_inner_protocol(struct sk_buff *skb,
2008 __be16 protocol)
2009{
2010 skb->inner_protocol = protocol;
2011 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2012}
2013
2014static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2015 __u8 ipproto)
2016{
2017 skb->inner_ipproto = ipproto;
2018 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2019}
2020
6a674e9c
JG
2021static inline void skb_reset_inner_headers(struct sk_buff *skb)
2022{
aefbd2b3 2023 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2024 skb->inner_network_header = skb->network_header;
2025 skb->inner_transport_header = skb->transport_header;
2026}
2027
0b5c9db1
JP
2028static inline void skb_reset_mac_len(struct sk_buff *skb)
2029{
2030 skb->mac_len = skb->network_header - skb->mac_header;
2031}
2032
6a674e9c
JG
2033static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2034 *skb)
2035{
2036 return skb->head + skb->inner_transport_header;
2037}
2038
55dc5a9f
TH
2039static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2040{
2041 return skb_inner_transport_header(skb) - skb->data;
2042}
2043
6a674e9c
JG
2044static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2045{
2046 skb->inner_transport_header = skb->data - skb->head;
2047}
2048
2049static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2050 const int offset)
2051{
2052 skb_reset_inner_transport_header(skb);
2053 skb->inner_transport_header += offset;
2054}
2055
2056static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2057{
2058 return skb->head + skb->inner_network_header;
2059}
2060
2061static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2062{
2063 skb->inner_network_header = skb->data - skb->head;
2064}
2065
2066static inline void skb_set_inner_network_header(struct sk_buff *skb,
2067 const int offset)
2068{
2069 skb_reset_inner_network_header(skb);
2070 skb->inner_network_header += offset;
2071}
2072
aefbd2b3
PS
2073static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2074{
2075 return skb->head + skb->inner_mac_header;
2076}
2077
2078static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2079{
2080 skb->inner_mac_header = skb->data - skb->head;
2081}
2082
2083static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2084 const int offset)
2085{
2086 skb_reset_inner_mac_header(skb);
2087 skb->inner_mac_header += offset;
2088}
fda55eca
ED
2089static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2090{
35d04610 2091 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2092}
2093
9c70220b
ACM
2094static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2095{
2e07fa9c 2096 return skb->head + skb->transport_header;
9c70220b
ACM
2097}
2098
badff6d0
ACM
2099static inline void skb_reset_transport_header(struct sk_buff *skb)
2100{
2e07fa9c 2101 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2102}
2103
967b05f6
ACM
2104static inline void skb_set_transport_header(struct sk_buff *skb,
2105 const int offset)
2106{
2e07fa9c
ACM
2107 skb_reset_transport_header(skb);
2108 skb->transport_header += offset;
ea2ae17d
ACM
2109}
2110
d56f90a7
ACM
2111static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2112{
2e07fa9c 2113 return skb->head + skb->network_header;
d56f90a7
ACM
2114}
2115
c1d2bbe1
ACM
2116static inline void skb_reset_network_header(struct sk_buff *skb)
2117{
2e07fa9c 2118 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2119}
2120
c14d2450
ACM
2121static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2122{
2e07fa9c
ACM
2123 skb_reset_network_header(skb);
2124 skb->network_header += offset;
c14d2450
ACM
2125}
2126
2e07fa9c 2127static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2128{
2e07fa9c 2129 return skb->head + skb->mac_header;
bbe735e4
ACM
2130}
2131
ea6da4fd
AV
2132static inline int skb_mac_offset(const struct sk_buff *skb)
2133{
2134 return skb_mac_header(skb) - skb->data;
2135}
2136
2e07fa9c 2137static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2138{
35d04610 2139 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2140}
2141
2142static inline void skb_reset_mac_header(struct sk_buff *skb)
2143{
2144 skb->mac_header = skb->data - skb->head;
2145}
2146
2147static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2148{
2149 skb_reset_mac_header(skb);
2150 skb->mac_header += offset;
2151}
2152
0e3da5bb
TT
2153static inline void skb_pop_mac_header(struct sk_buff *skb)
2154{
2155 skb->mac_header = skb->network_header;
2156}
2157
fbbdb8f0
YX
2158static inline void skb_probe_transport_header(struct sk_buff *skb,
2159 const int offset_hint)
2160{
2161 struct flow_keys keys;
2162
2163 if (skb_transport_header_was_set(skb))
2164 return;
cd79a238 2165 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2166 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2167 else
2168 skb_set_transport_header(skb, offset_hint);
2169}
2170
03606895
ED
2171static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2172{
2173 if (skb_mac_header_was_set(skb)) {
2174 const unsigned char *old_mac = skb_mac_header(skb);
2175
2176 skb_set_mac_header(skb, -skb->mac_len);
2177 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2178 }
2179}
2180
04fb451e
MM
2181static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2182{
2183 return skb->csum_start - skb_headroom(skb);
2184}
2185
08b64fcc
AD
2186static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2187{
2188 return skb->head + skb->csum_start;
2189}
2190
2e07fa9c
ACM
2191static inline int skb_transport_offset(const struct sk_buff *skb)
2192{
2193 return skb_transport_header(skb) - skb->data;
2194}
2195
2196static inline u32 skb_network_header_len(const struct sk_buff *skb)
2197{
2198 return skb->transport_header - skb->network_header;
2199}
2200
6a674e9c
JG
2201static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2202{
2203 return skb->inner_transport_header - skb->inner_network_header;
2204}
2205
2e07fa9c
ACM
2206static inline int skb_network_offset(const struct sk_buff *skb)
2207{
2208 return skb_network_header(skb) - skb->data;
2209}
48d49d0c 2210
6a674e9c
JG
2211static inline int skb_inner_network_offset(const struct sk_buff *skb)
2212{
2213 return skb_inner_network_header(skb) - skb->data;
2214}
2215
f9599ce1
CG
2216static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2217{
2218 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2219}
2220
1da177e4
LT
2221/*
2222 * CPUs often take a performance hit when accessing unaligned memory
2223 * locations. The actual performance hit varies, it can be small if the
2224 * hardware handles it or large if we have to take an exception and fix it
2225 * in software.
2226 *
2227 * Since an ethernet header is 14 bytes network drivers often end up with
2228 * the IP header at an unaligned offset. The IP header can be aligned by
2229 * shifting the start of the packet by 2 bytes. Drivers should do this
2230 * with:
2231 *
8660c124 2232 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2233 *
2234 * The downside to this alignment of the IP header is that the DMA is now
2235 * unaligned. On some architectures the cost of an unaligned DMA is high
2236 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2237 *
1da177e4
LT
2238 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2239 * to be overridden.
2240 */
2241#ifndef NET_IP_ALIGN
2242#define NET_IP_ALIGN 2
2243#endif
2244
025be81e
AB
2245/*
2246 * The networking layer reserves some headroom in skb data (via
2247 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2248 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2249 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2250 *
2251 * Unfortunately this headroom changes the DMA alignment of the resulting
2252 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2253 * on some architectures. An architecture can override this value,
2254 * perhaps setting it to a cacheline in size (since that will maintain
2255 * cacheline alignment of the DMA). It must be a power of 2.
2256 *
d6301d3d 2257 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2258 * headroom, you should not reduce this.
5933dd2f
ED
2259 *
2260 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2261 * to reduce average number of cache lines per packet.
2262 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2263 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2264 */
2265#ifndef NET_SKB_PAD
5933dd2f 2266#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2267#endif
2268
7965bd4d 2269int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2270
5293efe6 2271static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2272{
c4264f27 2273 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2274 WARN_ON(1);
2275 return;
2276 }
27a884dc
ACM
2277 skb->len = len;
2278 skb_set_tail_pointer(skb, len);
1da177e4
LT
2279}
2280
5293efe6
DB
2281static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2282{
2283 __skb_set_length(skb, len);
2284}
2285
7965bd4d 2286void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2287
2288static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2289{
3cc0e873
HX
2290 if (skb->data_len)
2291 return ___pskb_trim(skb, len);
2292 __skb_trim(skb, len);
2293 return 0;
1da177e4
LT
2294}
2295
2296static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2297{
2298 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2299}
2300
e9fa4f7b
HX
2301/**
2302 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2303 * @skb: buffer to alter
2304 * @len: new length
2305 *
2306 * This is identical to pskb_trim except that the caller knows that
2307 * the skb is not cloned so we should never get an error due to out-
2308 * of-memory.
2309 */
2310static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2311{
2312 int err = pskb_trim(skb, len);
2313 BUG_ON(err);
2314}
2315
5293efe6
DB
2316static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2317{
2318 unsigned int diff = len - skb->len;
2319
2320 if (skb_tailroom(skb) < diff) {
2321 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2322 GFP_ATOMIC);
2323 if (ret)
2324 return ret;
2325 }
2326 __skb_set_length(skb, len);
2327 return 0;
2328}
2329
1da177e4
LT
2330/**
2331 * skb_orphan - orphan a buffer
2332 * @skb: buffer to orphan
2333 *
2334 * If a buffer currently has an owner then we call the owner's
2335 * destructor function and make the @skb unowned. The buffer continues
2336 * to exist but is no longer charged to its former owner.
2337 */
2338static inline void skb_orphan(struct sk_buff *skb)
2339{
c34a7612 2340 if (skb->destructor) {
1da177e4 2341 skb->destructor(skb);
c34a7612
ED
2342 skb->destructor = NULL;
2343 skb->sk = NULL;
376c7311
ED
2344 } else {
2345 BUG_ON(skb->sk);
c34a7612 2346 }
1da177e4
LT
2347}
2348
a353e0ce
MT
2349/**
2350 * skb_orphan_frags - orphan the frags contained in a buffer
2351 * @skb: buffer to orphan frags from
2352 * @gfp_mask: allocation mask for replacement pages
2353 *
2354 * For each frag in the SKB which needs a destructor (i.e. has an
2355 * owner) create a copy of that frag and release the original
2356 * page by calling the destructor.
2357 */
2358static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2359{
2360 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2361 return 0;
2362 return skb_copy_ubufs(skb, gfp_mask);
2363}
2364
1da177e4
LT
2365/**
2366 * __skb_queue_purge - empty a list
2367 * @list: list to empty
2368 *
2369 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2370 * the list and one reference dropped. This function does not take the
2371 * list lock and the caller must hold the relevant locks to use it.
2372 */
7965bd4d 2373void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2374static inline void __skb_queue_purge(struct sk_buff_head *list)
2375{
2376 struct sk_buff *skb;
2377 while ((skb = __skb_dequeue(list)) != NULL)
2378 kfree_skb(skb);
2379}
2380
9f5afeae
YW
2381void skb_rbtree_purge(struct rb_root *root);
2382
7965bd4d 2383void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2384
7965bd4d
JP
2385struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2386 gfp_t gfp_mask);
8af27456
CH
2387
2388/**
2389 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2390 * @dev: network device to receive on
2391 * @length: length to allocate
2392 *
2393 * Allocate a new &sk_buff and assign it a usage count of one. The
2394 * buffer has unspecified headroom built in. Users should allocate
2395 * the headroom they think they need without accounting for the
2396 * built in space. The built in space is used for optimisations.
2397 *
2398 * %NULL is returned if there is no free memory. Although this function
2399 * allocates memory it can be called from an interrupt.
2400 */
2401static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2402 unsigned int length)
8af27456
CH
2403{
2404 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2405}
2406
6f532612
ED
2407/* legacy helper around __netdev_alloc_skb() */
2408static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2409 gfp_t gfp_mask)
2410{
2411 return __netdev_alloc_skb(NULL, length, gfp_mask);
2412}
2413
2414/* legacy helper around netdev_alloc_skb() */
2415static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2416{
2417 return netdev_alloc_skb(NULL, length);
2418}
2419
2420
4915a0de
ED
2421static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2422 unsigned int length, gfp_t gfp)
61321bbd 2423{
4915a0de 2424 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2425
2426 if (NET_IP_ALIGN && skb)
2427 skb_reserve(skb, NET_IP_ALIGN);
2428 return skb;
2429}
2430
4915a0de
ED
2431static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2432 unsigned int length)
2433{
2434 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2435}
2436
181edb2b
AD
2437static inline void skb_free_frag(void *addr)
2438{
8c2dd3e4 2439 page_frag_free(addr);
181edb2b
AD
2440}
2441
ffde7328 2442void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2443struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2444 unsigned int length, gfp_t gfp_mask);
2445static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2446 unsigned int length)
2447{
2448 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2449}
795bb1c0
JDB
2450void napi_consume_skb(struct sk_buff *skb, int budget);
2451
2452void __kfree_skb_flush(void);
15fad714 2453void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2454
71dfda58
AD
2455/**
2456 * __dev_alloc_pages - allocate page for network Rx
2457 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2458 * @order: size of the allocation
2459 *
2460 * Allocate a new page.
2461 *
2462 * %NULL is returned if there is no free memory.
2463*/
2464static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2465 unsigned int order)
2466{
2467 /* This piece of code contains several assumptions.
2468 * 1. This is for device Rx, therefor a cold page is preferred.
2469 * 2. The expectation is the user wants a compound page.
2470 * 3. If requesting a order 0 page it will not be compound
2471 * due to the check to see if order has a value in prep_new_page
2472 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2473 * code in gfp_to_alloc_flags that should be enforcing this.
2474 */
2475 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2476
2477 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2478}
2479
2480static inline struct page *dev_alloc_pages(unsigned int order)
2481{
95829b3a 2482 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2483}
2484
2485/**
2486 * __dev_alloc_page - allocate a page for network Rx
2487 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2488 *
2489 * Allocate a new page.
2490 *
2491 * %NULL is returned if there is no free memory.
2492 */
2493static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2494{
2495 return __dev_alloc_pages(gfp_mask, 0);
2496}
2497
2498static inline struct page *dev_alloc_page(void)
2499{
95829b3a 2500 return dev_alloc_pages(0);
71dfda58
AD
2501}
2502
0614002b
MG
2503/**
2504 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2505 * @page: The page that was allocated from skb_alloc_page
2506 * @skb: The skb that may need pfmemalloc set
2507 */
2508static inline void skb_propagate_pfmemalloc(struct page *page,
2509 struct sk_buff *skb)
2510{
2f064f34 2511 if (page_is_pfmemalloc(page))
0614002b
MG
2512 skb->pfmemalloc = true;
2513}
2514
131ea667 2515/**
e227867f 2516 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2517 * @frag: the paged fragment
2518 *
2519 * Returns the &struct page associated with @frag.
2520 */
2521static inline struct page *skb_frag_page(const skb_frag_t *frag)
2522{
a8605c60 2523 return frag->page.p;
131ea667
IC
2524}
2525
2526/**
2527 * __skb_frag_ref - take an addition reference on a paged fragment.
2528 * @frag: the paged fragment
2529 *
2530 * Takes an additional reference on the paged fragment @frag.
2531 */
2532static inline void __skb_frag_ref(skb_frag_t *frag)
2533{
2534 get_page(skb_frag_page(frag));
2535}
2536
2537/**
2538 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2539 * @skb: the buffer
2540 * @f: the fragment offset.
2541 *
2542 * Takes an additional reference on the @f'th paged fragment of @skb.
2543 */
2544static inline void skb_frag_ref(struct sk_buff *skb, int f)
2545{
2546 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2547}
2548
2549/**
2550 * __skb_frag_unref - release a reference on a paged fragment.
2551 * @frag: the paged fragment
2552 *
2553 * Releases a reference on the paged fragment @frag.
2554 */
2555static inline void __skb_frag_unref(skb_frag_t *frag)
2556{
2557 put_page(skb_frag_page(frag));
2558}
2559
2560/**
2561 * skb_frag_unref - release a reference on a paged fragment of an skb.
2562 * @skb: the buffer
2563 * @f: the fragment offset
2564 *
2565 * Releases a reference on the @f'th paged fragment of @skb.
2566 */
2567static inline void skb_frag_unref(struct sk_buff *skb, int f)
2568{
2569 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2570}
2571
2572/**
2573 * skb_frag_address - gets the address of the data contained in a paged fragment
2574 * @frag: the paged fragment buffer
2575 *
2576 * Returns the address of the data within @frag. The page must already
2577 * be mapped.
2578 */
2579static inline void *skb_frag_address(const skb_frag_t *frag)
2580{
2581 return page_address(skb_frag_page(frag)) + frag->page_offset;
2582}
2583
2584/**
2585 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2586 * @frag: the paged fragment buffer
2587 *
2588 * Returns the address of the data within @frag. Checks that the page
2589 * is mapped and returns %NULL otherwise.
2590 */
2591static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2592{
2593 void *ptr = page_address(skb_frag_page(frag));
2594 if (unlikely(!ptr))
2595 return NULL;
2596
2597 return ptr + frag->page_offset;
2598}
2599
2600/**
2601 * __skb_frag_set_page - sets the page contained in a paged fragment
2602 * @frag: the paged fragment
2603 * @page: the page to set
2604 *
2605 * Sets the fragment @frag to contain @page.
2606 */
2607static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2608{
a8605c60 2609 frag->page.p = page;
131ea667
IC
2610}
2611
2612/**
2613 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2614 * @skb: the buffer
2615 * @f: the fragment offset
2616 * @page: the page to set
2617 *
2618 * Sets the @f'th fragment of @skb to contain @page.
2619 */
2620static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2621 struct page *page)
2622{
2623 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2624}
2625
400dfd3a
ED
2626bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2627
131ea667
IC
2628/**
2629 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2630 * @dev: the device to map the fragment to
131ea667
IC
2631 * @frag: the paged fragment to map
2632 * @offset: the offset within the fragment (starting at the
2633 * fragment's own offset)
2634 * @size: the number of bytes to map
f83347df 2635 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2636 *
2637 * Maps the page associated with @frag to @device.
2638 */
2639static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2640 const skb_frag_t *frag,
2641 size_t offset, size_t size,
2642 enum dma_data_direction dir)
2643{
2644 return dma_map_page(dev, skb_frag_page(frag),
2645 frag->page_offset + offset, size, dir);
2646}
2647
117632e6
ED
2648static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2649 gfp_t gfp_mask)
2650{
2651 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2652}
2653
bad93e9d
OP
2654
2655static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2656 gfp_t gfp_mask)
2657{
2658 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2659}
2660
2661
334a8132
PM
2662/**
2663 * skb_clone_writable - is the header of a clone writable
2664 * @skb: buffer to check
2665 * @len: length up to which to write
2666 *
2667 * Returns true if modifying the header part of the cloned buffer
2668 * does not requires the data to be copied.
2669 */
05bdd2f1 2670static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2671{
2672 return !skb_header_cloned(skb) &&
2673 skb_headroom(skb) + len <= skb->hdr_len;
2674}
2675
3697649f
DB
2676static inline int skb_try_make_writable(struct sk_buff *skb,
2677 unsigned int write_len)
2678{
2679 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2680 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2681}
2682
d9cc2048
HX
2683static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2684 int cloned)
2685{
2686 int delta = 0;
2687
d9cc2048
HX
2688 if (headroom > skb_headroom(skb))
2689 delta = headroom - skb_headroom(skb);
2690
2691 if (delta || cloned)
2692 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2693 GFP_ATOMIC);
2694 return 0;
2695}
2696
1da177e4
LT
2697/**
2698 * skb_cow - copy header of skb when it is required
2699 * @skb: buffer to cow
2700 * @headroom: needed headroom
2701 *
2702 * If the skb passed lacks sufficient headroom or its data part
2703 * is shared, data is reallocated. If reallocation fails, an error
2704 * is returned and original skb is not changed.
2705 *
2706 * The result is skb with writable area skb->head...skb->tail
2707 * and at least @headroom of space at head.
2708 */
2709static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2710{
d9cc2048
HX
2711 return __skb_cow(skb, headroom, skb_cloned(skb));
2712}
1da177e4 2713
d9cc2048
HX
2714/**
2715 * skb_cow_head - skb_cow but only making the head writable
2716 * @skb: buffer to cow
2717 * @headroom: needed headroom
2718 *
2719 * This function is identical to skb_cow except that we replace the
2720 * skb_cloned check by skb_header_cloned. It should be used when
2721 * you only need to push on some header and do not need to modify
2722 * the data.
2723 */
2724static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2725{
2726 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2727}
2728
2729/**
2730 * skb_padto - pad an skbuff up to a minimal size
2731 * @skb: buffer to pad
2732 * @len: minimal length
2733 *
2734 * Pads up a buffer to ensure the trailing bytes exist and are
2735 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2736 * is untouched. Otherwise it is extended. Returns zero on
2737 * success. The skb is freed on error.
1da177e4 2738 */
5b057c6b 2739static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2740{
2741 unsigned int size = skb->len;
2742 if (likely(size >= len))
5b057c6b 2743 return 0;
987c402a 2744 return skb_pad(skb, len - size);
1da177e4
LT
2745}
2746
9c0c1124
AD
2747/**
2748 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2749 * @skb: buffer to pad
2750 * @len: minimal length
2751 *
2752 * Pads up a buffer to ensure the trailing bytes exist and are
2753 * blanked. If the buffer already contains sufficient data it
2754 * is untouched. Otherwise it is extended. Returns zero on
2755 * success. The skb is freed on error.
2756 */
2757static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2758{
2759 unsigned int size = skb->len;
2760
2761 if (unlikely(size < len)) {
2762 len -= size;
2763 if (skb_pad(skb, len))
2764 return -ENOMEM;
2765 __skb_put(skb, len);
2766 }
2767 return 0;
2768}
2769
1da177e4 2770static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2771 struct iov_iter *from, int copy)
1da177e4
LT
2772{
2773 const int off = skb->len;
2774
2775 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 2776 __wsum csum = 0;
15e6cb46
AV
2777 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2778 &csum, from)) {
1da177e4
LT
2779 skb->csum = csum_block_add(skb->csum, csum, off);
2780 return 0;
2781 }
15e6cb46 2782 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
2783 return 0;
2784
2785 __skb_trim(skb, off);
2786 return -EFAULT;
2787}
2788
38ba0a65
ED
2789static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2790 const struct page *page, int off)
1da177e4
LT
2791{
2792 if (i) {
9e903e08 2793 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2794
ea2ab693 2795 return page == skb_frag_page(frag) &&
9e903e08 2796 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2797 }
38ba0a65 2798 return false;
1da177e4
LT
2799}
2800
364c6bad
HX
2801static inline int __skb_linearize(struct sk_buff *skb)
2802{
2803 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2804}
2805
1da177e4
LT
2806/**
2807 * skb_linearize - convert paged skb to linear one
2808 * @skb: buffer to linarize
1da177e4
LT
2809 *
2810 * If there is no free memory -ENOMEM is returned, otherwise zero
2811 * is returned and the old skb data released.
2812 */
364c6bad
HX
2813static inline int skb_linearize(struct sk_buff *skb)
2814{
2815 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2816}
2817
cef401de
ED
2818/**
2819 * skb_has_shared_frag - can any frag be overwritten
2820 * @skb: buffer to test
2821 *
2822 * Return true if the skb has at least one frag that might be modified
2823 * by an external entity (as in vmsplice()/sendfile())
2824 */
2825static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2826{
c9af6db4
PS
2827 return skb_is_nonlinear(skb) &&
2828 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2829}
2830
364c6bad
HX
2831/**
2832 * skb_linearize_cow - make sure skb is linear and writable
2833 * @skb: buffer to process
2834 *
2835 * If there is no free memory -ENOMEM is returned, otherwise zero
2836 * is returned and the old skb data released.
2837 */
2838static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2839{
364c6bad
HX
2840 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2841 __skb_linearize(skb) : 0;
1da177e4
LT
2842}
2843
479ffccc
DB
2844static __always_inline void
2845__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2846 unsigned int off)
2847{
2848 if (skb->ip_summed == CHECKSUM_COMPLETE)
2849 skb->csum = csum_block_sub(skb->csum,
2850 csum_partial(start, len, 0), off);
2851 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2852 skb_checksum_start_offset(skb) < 0)
2853 skb->ip_summed = CHECKSUM_NONE;
2854}
2855
1da177e4
LT
2856/**
2857 * skb_postpull_rcsum - update checksum for received skb after pull
2858 * @skb: buffer to update
2859 * @start: start of data before pull
2860 * @len: length of data pulled
2861 *
2862 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2863 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2864 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 2865 */
1da177e4 2866static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2867 const void *start, unsigned int len)
1da177e4 2868{
479ffccc 2869 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
2870}
2871
479ffccc
DB
2872static __always_inline void
2873__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2874 unsigned int off)
2875{
2876 if (skb->ip_summed == CHECKSUM_COMPLETE)
2877 skb->csum = csum_block_add(skb->csum,
2878 csum_partial(start, len, 0), off);
2879}
cbb042f9 2880
479ffccc
DB
2881/**
2882 * skb_postpush_rcsum - update checksum for received skb after push
2883 * @skb: buffer to update
2884 * @start: start of data after push
2885 * @len: length of data pushed
2886 *
2887 * After doing a push on a received packet, you need to call this to
2888 * update the CHECKSUM_COMPLETE checksum.
2889 */
f8ffad69
DB
2890static inline void skb_postpush_rcsum(struct sk_buff *skb,
2891 const void *start, unsigned int len)
2892{
479ffccc 2893 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
2894}
2895
479ffccc
DB
2896unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2897
82a31b92
WC
2898/**
2899 * skb_push_rcsum - push skb and update receive checksum
2900 * @skb: buffer to update
2901 * @len: length of data pulled
2902 *
2903 * This function performs an skb_push on the packet and updates
2904 * the CHECKSUM_COMPLETE checksum. It should be used on
2905 * receive path processing instead of skb_push unless you know
2906 * that the checksum difference is zero (e.g., a valid IP header)
2907 * or you are setting ip_summed to CHECKSUM_NONE.
2908 */
2909static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2910 unsigned int len)
2911{
2912 skb_push(skb, len);
2913 skb_postpush_rcsum(skb, skb->data, len);
2914 return skb->data;
2915}
2916
7ce5a27f
DM
2917/**
2918 * pskb_trim_rcsum - trim received skb and update checksum
2919 * @skb: buffer to trim
2920 * @len: new length
2921 *
2922 * This is exactly the same as pskb_trim except that it ensures the
2923 * checksum of received packets are still valid after the operation.
2924 */
2925
2926static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2927{
2928 if (likely(len >= skb->len))
2929 return 0;
2930 if (skb->ip_summed == CHECKSUM_COMPLETE)
2931 skb->ip_summed = CHECKSUM_NONE;
2932 return __pskb_trim(skb, len);
2933}
2934
5293efe6
DB
2935static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2936{
2937 if (skb->ip_summed == CHECKSUM_COMPLETE)
2938 skb->ip_summed = CHECKSUM_NONE;
2939 __skb_trim(skb, len);
2940 return 0;
2941}
2942
2943static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2944{
2945 if (skb->ip_summed == CHECKSUM_COMPLETE)
2946 skb->ip_summed = CHECKSUM_NONE;
2947 return __skb_grow(skb, len);
2948}
2949
1da177e4
LT
2950#define skb_queue_walk(queue, skb) \
2951 for (skb = (queue)->next; \
a1e4891f 2952 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2953 skb = skb->next)
2954
46f8914e
JC
2955#define skb_queue_walk_safe(queue, skb, tmp) \
2956 for (skb = (queue)->next, tmp = skb->next; \
2957 skb != (struct sk_buff *)(queue); \
2958 skb = tmp, tmp = skb->next)
2959
1164f52a 2960#define skb_queue_walk_from(queue, skb) \
a1e4891f 2961 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2962 skb = skb->next)
2963
2964#define skb_queue_walk_from_safe(queue, skb, tmp) \
2965 for (tmp = skb->next; \
2966 skb != (struct sk_buff *)(queue); \
2967 skb = tmp, tmp = skb->next)
2968
300ce174
SH
2969#define skb_queue_reverse_walk(queue, skb) \
2970 for (skb = (queue)->prev; \
a1e4891f 2971 skb != (struct sk_buff *)(queue); \
300ce174
SH
2972 skb = skb->prev)
2973
686a2955
DM
2974#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2975 for (skb = (queue)->prev, tmp = skb->prev; \
2976 skb != (struct sk_buff *)(queue); \
2977 skb = tmp, tmp = skb->prev)
2978
2979#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2980 for (tmp = skb->prev; \
2981 skb != (struct sk_buff *)(queue); \
2982 skb = tmp, tmp = skb->prev)
1da177e4 2983
21dc3301 2984static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2985{
2986 return skb_shinfo(skb)->frag_list != NULL;
2987}
2988
2989static inline void skb_frag_list_init(struct sk_buff *skb)
2990{
2991 skb_shinfo(skb)->frag_list = NULL;
2992}
2993
ee039871
DM
2994#define skb_walk_frags(skb, iter) \
2995 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2996
ea3793ee
RW
2997
2998int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
2999 const struct sk_buff *skb);
65101aec
PA
3000struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3001 struct sk_buff_head *queue,
3002 unsigned int flags,
3003 void (*destructor)(struct sock *sk,
3004 struct sk_buff *skb),
3005 int *peeked, int *off, int *err,
3006 struct sk_buff **last);
ea3793ee 3007struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3008 void (*destructor)(struct sock *sk,
3009 struct sk_buff *skb),
ea3793ee
RW
3010 int *peeked, int *off, int *err,
3011 struct sk_buff **last);
7965bd4d 3012struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3013 void (*destructor)(struct sock *sk,
3014 struct sk_buff *skb),
7965bd4d
JP
3015 int *peeked, int *off, int *err);
3016struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3017 int *err);
3018unsigned int datagram_poll(struct file *file, struct socket *sock,
3019 struct poll_table_struct *wait);
c0371da6
AV
3020int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3021 struct iov_iter *to, int size);
51f3d02b
DM
3022static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3023 struct msghdr *msg, int size)
3024{
e5a4b0bb 3025 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3026}
e5a4b0bb
AV
3027int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3028 struct msghdr *msg);
3a654f97
AV
3029int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3030 struct iov_iter *from, int len);
3a654f97 3031int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3032void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3033void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3034static inline void skb_free_datagram_locked(struct sock *sk,
3035 struct sk_buff *skb)
3036{
3037 __skb_free_datagram_locked(sk, skb, 0);
3038}
7965bd4d 3039int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3040int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3041int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3042__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3043 int len, __wsum csum);
a60e3cc7 3044int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3045 struct pipe_inode_info *pipe, unsigned int len,
25869262 3046 unsigned int flags);
7965bd4d 3047void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3048unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3049int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3050 int len, int hlen);
7965bd4d
JP
3051void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3052int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3053void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3054unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3055bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 3056struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3057struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3058int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3059int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3060int skb_vlan_pop(struct sk_buff *skb);
3061int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3062struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3063 gfp_t gfp);
20380731 3064
6ce8e9ce
AV
3065static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3066{
3073f070 3067 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3068}
3069
7eab8d9e
AV
3070static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3071{
e5a4b0bb 3072 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3073}
3074
2817a336
DB
3075struct skb_checksum_ops {
3076 __wsum (*update)(const void *mem, int len, __wsum wsum);
3077 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3078};
3079
9617813d
DC
3080extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3081
2817a336
DB
3082__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3083 __wsum csum, const struct skb_checksum_ops *ops);
3084__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3085 __wsum csum);
3086
1e98a0f0
ED
3087static inline void * __must_check
3088__skb_header_pointer(const struct sk_buff *skb, int offset,
3089 int len, void *data, int hlen, void *buffer)
1da177e4 3090{
55820ee2 3091 if (hlen - offset >= len)
690e36e7 3092 return data + offset;
1da177e4 3093
690e36e7
DM
3094 if (!skb ||
3095 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3096 return NULL;
3097
3098 return buffer;
3099}
3100
1e98a0f0
ED
3101static inline void * __must_check
3102skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3103{
3104 return __skb_header_pointer(skb, offset, len, skb->data,
3105 skb_headlen(skb), buffer);
3106}
3107
4262e5cc
DB
3108/**
3109 * skb_needs_linearize - check if we need to linearize a given skb
3110 * depending on the given device features.
3111 * @skb: socket buffer to check
3112 * @features: net device features
3113 *
3114 * Returns true if either:
3115 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3116 * 2. skb is fragmented and the device does not support SG.
3117 */
3118static inline bool skb_needs_linearize(struct sk_buff *skb,
3119 netdev_features_t features)
3120{
3121 return skb_is_nonlinear(skb) &&
3122 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3123 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3124}
3125
d626f62b
ACM
3126static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3127 void *to,
3128 const unsigned int len)
3129{
3130 memcpy(to, skb->data, len);
3131}
3132
3133static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3134 const int offset, void *to,
3135 const unsigned int len)
3136{
3137 memcpy(to, skb->data + offset, len);
3138}
3139
27d7ff46
ACM
3140static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3141 const void *from,
3142 const unsigned int len)
3143{
3144 memcpy(skb->data, from, len);
3145}
3146
3147static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3148 const int offset,
3149 const void *from,
3150 const unsigned int len)
3151{
3152 memcpy(skb->data + offset, from, len);
3153}
3154
7965bd4d 3155void skb_init(void);
1da177e4 3156
ac45f602
PO
3157static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3158{
3159 return skb->tstamp;
3160}
3161
a61bbcf2
PM
3162/**
3163 * skb_get_timestamp - get timestamp from a skb
3164 * @skb: skb to get stamp from
3165 * @stamp: pointer to struct timeval to store stamp in
3166 *
3167 * Timestamps are stored in the skb as offsets to a base timestamp.
3168 * This function converts the offset back to a struct timeval and stores
3169 * it in stamp.
3170 */
ac45f602
PO
3171static inline void skb_get_timestamp(const struct sk_buff *skb,
3172 struct timeval *stamp)
a61bbcf2 3173{
b7aa0bf7 3174 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3175}
3176
ac45f602
PO
3177static inline void skb_get_timestampns(const struct sk_buff *skb,
3178 struct timespec *stamp)
3179{
3180 *stamp = ktime_to_timespec(skb->tstamp);
3181}
3182
b7aa0bf7 3183static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3184{
b7aa0bf7 3185 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3186}
3187
164891aa
SH
3188static inline ktime_t net_timedelta(ktime_t t)
3189{
3190 return ktime_sub(ktime_get_real(), t);
3191}
3192
b9ce204f
IJ
3193static inline ktime_t net_invalid_timestamp(void)
3194{
8b0e1953 3195 return 0;
b9ce204f 3196}
a61bbcf2 3197
62bccb8c
AD
3198struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3199
c1f19b51
RC
3200#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3201
7965bd4d
JP
3202void skb_clone_tx_timestamp(struct sk_buff *skb);
3203bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3204
3205#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3206
3207static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3208{
3209}
3210
3211static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3212{
3213 return false;
3214}
3215
3216#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3217
3218/**
3219 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3220 *
da92b194
RC
3221 * PHY drivers may accept clones of transmitted packets for
3222 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3223 * must call this function to return the skb back to the stack with a
3224 * timestamp.
da92b194 3225 *
c1f19b51 3226 * @skb: clone of the the original outgoing packet
7a76a021 3227 * @hwtstamps: hardware time stamps
c1f19b51
RC
3228 *
3229 */
3230void skb_complete_tx_timestamp(struct sk_buff *skb,
3231 struct skb_shared_hwtstamps *hwtstamps);
3232
e7fd2885
WB
3233void __skb_tstamp_tx(struct sk_buff *orig_skb,
3234 struct skb_shared_hwtstamps *hwtstamps,
3235 struct sock *sk, int tstype);
3236
ac45f602
PO
3237/**
3238 * skb_tstamp_tx - queue clone of skb with send time stamps
3239 * @orig_skb: the original outgoing packet
3240 * @hwtstamps: hardware time stamps, may be NULL if not available
3241 *
3242 * If the skb has a socket associated, then this function clones the
3243 * skb (thus sharing the actual data and optional structures), stores
3244 * the optional hardware time stamping information (if non NULL) or
3245 * generates a software time stamp (otherwise), then queues the clone
3246 * to the error queue of the socket. Errors are silently ignored.
3247 */
7965bd4d
JP
3248void skb_tstamp_tx(struct sk_buff *orig_skb,
3249 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3250
4507a715
RC
3251static inline void sw_tx_timestamp(struct sk_buff *skb)
3252{
2244d07b
OH
3253 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3254 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3255 skb_tstamp_tx(skb, NULL);
3256}
3257
3258/**
3259 * skb_tx_timestamp() - Driver hook for transmit timestamping
3260 *
3261 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3262 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3263 *
73409f3b
DM
3264 * Specifically, one should make absolutely sure that this function is
3265 * called before TX completion of this packet can trigger. Otherwise
3266 * the packet could potentially already be freed.
3267 *
4507a715
RC
3268 * @skb: A socket buffer.
3269 */
3270static inline void skb_tx_timestamp(struct sk_buff *skb)
3271{
c1f19b51 3272 skb_clone_tx_timestamp(skb);
4507a715
RC
3273 sw_tx_timestamp(skb);
3274}
3275
6e3e939f
JB
3276/**
3277 * skb_complete_wifi_ack - deliver skb with wifi status
3278 *
3279 * @skb: the original outgoing packet
3280 * @acked: ack status
3281 *
3282 */
3283void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3284
7965bd4d
JP
3285__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3286__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3287
60476372
HX
3288static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3289{
6edec0e6
TH
3290 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3291 skb->csum_valid ||
3292 (skb->ip_summed == CHECKSUM_PARTIAL &&
3293 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3294}
3295
fb286bb2
HX
3296/**
3297 * skb_checksum_complete - Calculate checksum of an entire packet
3298 * @skb: packet to process
3299 *
3300 * This function calculates the checksum over the entire packet plus
3301 * the value of skb->csum. The latter can be used to supply the
3302 * checksum of a pseudo header as used by TCP/UDP. It returns the
3303 * checksum.
3304 *
3305 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3306 * this function can be used to verify that checksum on received
3307 * packets. In that case the function should return zero if the
3308 * checksum is correct. In particular, this function will return zero
3309 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3310 * hardware has already verified the correctness of the checksum.
3311 */
4381ca3c 3312static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3313{
60476372
HX
3314 return skb_csum_unnecessary(skb) ?
3315 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3316}
3317
77cffe23
TH
3318static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3319{
3320 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3321 if (skb->csum_level == 0)
3322 skb->ip_summed = CHECKSUM_NONE;
3323 else
3324 skb->csum_level--;
3325 }
3326}
3327
3328static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3329{
3330 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3331 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3332 skb->csum_level++;
3333 } else if (skb->ip_summed == CHECKSUM_NONE) {
3334 skb->ip_summed = CHECKSUM_UNNECESSARY;
3335 skb->csum_level = 0;
3336 }
3337}
3338
5a212329
TH
3339static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3340{
3341 /* Mark current checksum as bad (typically called from GRO
3342 * path). In the case that ip_summed is CHECKSUM_NONE
3343 * this must be the first checksum encountered in the packet.
3344 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3345 * checksum after the last one validated. For UDP, a zero
3346 * checksum can not be marked as bad.
3347 */
3348
3349 if (skb->ip_summed == CHECKSUM_NONE ||
3350 skb->ip_summed == CHECKSUM_UNNECESSARY)
3351 skb->csum_bad = 1;
3352}
3353
76ba0aae
TH
3354/* Check if we need to perform checksum complete validation.
3355 *
3356 * Returns true if checksum complete is needed, false otherwise
3357 * (either checksum is unnecessary or zero checksum is allowed).
3358 */
3359static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3360 bool zero_okay,
3361 __sum16 check)
3362{
5d0c2b95
TH
3363 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3364 skb->csum_valid = 1;
77cffe23 3365 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3366 return false;
3367 }
3368
3369 return true;
3370}
3371
3372/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3373 * in checksum_init.
3374 */
3375#define CHECKSUM_BREAK 76
3376
4e18b9ad
TH
3377/* Unset checksum-complete
3378 *
3379 * Unset checksum complete can be done when packet is being modified
3380 * (uncompressed for instance) and checksum-complete value is
3381 * invalidated.
3382 */
3383static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3384{
3385 if (skb->ip_summed == CHECKSUM_COMPLETE)
3386 skb->ip_summed = CHECKSUM_NONE;
3387}
3388
76ba0aae
TH
3389/* Validate (init) checksum based on checksum complete.
3390 *
3391 * Return values:
3392 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3393 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3394 * checksum is stored in skb->csum for use in __skb_checksum_complete
3395 * non-zero: value of invalid checksum
3396 *
3397 */
3398static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3399 bool complete,
3400 __wsum psum)
3401{
3402 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3403 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3404 skb->csum_valid = 1;
76ba0aae
TH
3405 return 0;
3406 }
5a212329
TH
3407 } else if (skb->csum_bad) {
3408 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3409 return (__force __sum16)1;
76ba0aae
TH
3410 }
3411
3412 skb->csum = psum;
3413
5d0c2b95
TH
3414 if (complete || skb->len <= CHECKSUM_BREAK) {
3415 __sum16 csum;
3416
3417 csum = __skb_checksum_complete(skb);
3418 skb->csum_valid = !csum;
3419 return csum;
3420 }
76ba0aae
TH
3421
3422 return 0;
3423}
3424
3425static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3426{
3427 return 0;
3428}
3429
3430/* Perform checksum validate (init). Note that this is a macro since we only
3431 * want to calculate the pseudo header which is an input function if necessary.
3432 * First we try to validate without any computation (checksum unnecessary) and
3433 * then calculate based on checksum complete calling the function to compute
3434 * pseudo header.
3435 *
3436 * Return values:
3437 * 0: checksum is validated or try to in skb_checksum_complete
3438 * non-zero: value of invalid checksum
3439 */
3440#define __skb_checksum_validate(skb, proto, complete, \
3441 zero_okay, check, compute_pseudo) \
3442({ \
3443 __sum16 __ret = 0; \
5d0c2b95 3444 skb->csum_valid = 0; \
76ba0aae
TH
3445 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3446 __ret = __skb_checksum_validate_complete(skb, \
3447 complete, compute_pseudo(skb, proto)); \
3448 __ret; \
3449})
3450
3451#define skb_checksum_init(skb, proto, compute_pseudo) \
3452 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3453
3454#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3455 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3456
3457#define skb_checksum_validate(skb, proto, compute_pseudo) \
3458 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3459
3460#define skb_checksum_validate_zero_check(skb, proto, check, \
3461 compute_pseudo) \
096a4cfa 3462 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3463
3464#define skb_checksum_simple_validate(skb) \
3465 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3466
d96535a1
TH
3467static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3468{
3469 return (skb->ip_summed == CHECKSUM_NONE &&
3470 skb->csum_valid && !skb->csum_bad);
3471}
3472
3473static inline void __skb_checksum_convert(struct sk_buff *skb,
3474 __sum16 check, __wsum pseudo)
3475{
3476 skb->csum = ~pseudo;
3477 skb->ip_summed = CHECKSUM_COMPLETE;
3478}
3479
3480#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3481do { \
3482 if (__skb_checksum_convert_check(skb)) \
3483 __skb_checksum_convert(skb, check, \
3484 compute_pseudo(skb, proto)); \
3485} while (0)
3486
15e2396d
TH
3487static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3488 u16 start, u16 offset)
3489{
3490 skb->ip_summed = CHECKSUM_PARTIAL;
3491 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3492 skb->csum_offset = offset - start;
3493}
3494
dcdc8994
TH
3495/* Update skbuf and packet to reflect the remote checksum offload operation.
3496 * When called, ptr indicates the starting point for skb->csum when
3497 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3498 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3499 */
3500static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3501 int start, int offset, bool nopartial)
dcdc8994
TH
3502{
3503 __wsum delta;
3504
15e2396d
TH
3505 if (!nopartial) {
3506 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3507 return;
3508 }
3509
dcdc8994
TH
3510 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3511 __skb_checksum_complete(skb);
3512 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3513 }
3514
3515 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3516
3517 /* Adjust skb->csum since we changed the packet */
3518 skb->csum = csum_add(skb->csum, delta);
3519}
3520
cb9c6836
FW
3521static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3522{
3523#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 3524 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
3525#else
3526 return NULL;
3527#endif
3528}
3529
5f79e0f9 3530#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3531void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3532static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3533{
3534 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3535 nf_conntrack_destroy(nfct);
1da177e4
LT
3536}
3537static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3538{
3539 if (nfct)
3540 atomic_inc(&nfct->use);
3541}
2fc72c7b 3542#endif
34666d46 3543#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3544static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3545{
3546 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3547 kfree(nf_bridge);
3548}
3549static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3550{
3551 if (nf_bridge)
3552 atomic_inc(&nf_bridge->use);
3553}
3554#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3555static inline void nf_reset(struct sk_buff *skb)
3556{
5f79e0f9 3557#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3558 nf_conntrack_put(skb_nfct(skb));
3559 skb->_nfct = 0;
2fc72c7b 3560#endif
34666d46 3561#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3562 nf_bridge_put(skb->nf_bridge);
3563 skb->nf_bridge = NULL;
3564#endif
3565}
3566
124dff01
PM
3567static inline void nf_reset_trace(struct sk_buff *skb)
3568{
478b360a 3569#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3570 skb->nf_trace = 0;
3571#endif
a193a4ab
PM
3572}
3573
edda553c 3574/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3575static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3576 bool copy)
edda553c 3577{
5f79e0f9 3578#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3579 dst->_nfct = src->_nfct;
3580 nf_conntrack_get(skb_nfct(src));
2fc72c7b 3581#endif
34666d46 3582#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3583 dst->nf_bridge = src->nf_bridge;
3584 nf_bridge_get(src->nf_bridge);
3585#endif
478b360a 3586#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3587 if (copy)
3588 dst->nf_trace = src->nf_trace;
478b360a 3589#endif
edda553c
YK
3590}
3591
e7ac05f3
YK
3592static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3593{
e7ac05f3 3594#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 3595 nf_conntrack_put(skb_nfct(dst));
2fc72c7b 3596#endif
34666d46 3597#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3598 nf_bridge_put(dst->nf_bridge);
3599#endif
b1937227 3600 __nf_copy(dst, src, true);
e7ac05f3
YK
3601}
3602
984bc16c
JM
3603#ifdef CONFIG_NETWORK_SECMARK
3604static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3605{
3606 to->secmark = from->secmark;
3607}
3608
3609static inline void skb_init_secmark(struct sk_buff *skb)
3610{
3611 skb->secmark = 0;
3612}
3613#else
3614static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3615{ }
3616
3617static inline void skb_init_secmark(struct sk_buff *skb)
3618{ }
3619#endif
3620
574f7194
EB
3621static inline bool skb_irq_freeable(const struct sk_buff *skb)
3622{
3623 return !skb->destructor &&
3624#if IS_ENABLED(CONFIG_XFRM)
3625 !skb->sp &&
3626#endif
cb9c6836 3627 !skb_nfct(skb) &&
574f7194
EB
3628 !skb->_skb_refdst &&
3629 !skb_has_frag_list(skb);
3630}
3631
f25f4e44
PWJ
3632static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3633{
f25f4e44 3634 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3635}
3636
9247744e 3637static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3638{
4e3ab47a 3639 return skb->queue_mapping;
4e3ab47a
PE
3640}
3641
f25f4e44
PWJ
3642static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3643{
f25f4e44 3644 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3645}
3646
d5a9e24a
DM
3647static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3648{
3649 skb->queue_mapping = rx_queue + 1;
3650}
3651
9247744e 3652static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3653{
3654 return skb->queue_mapping - 1;
3655}
3656
9247744e 3657static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3658{
a02cec21 3659 return skb->queue_mapping != 0;
d5a9e24a
DM
3660}
3661
4ff06203
JA
3662static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3663{
3664 skb->dst_pending_confirm = val;
3665}
3666
3667static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3668{
3669 return skb->dst_pending_confirm != 0;
3670}
3671
def8b4fa
AD
3672static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3673{
0b3d8e08 3674#ifdef CONFIG_XFRM
def8b4fa 3675 return skb->sp;
def8b4fa 3676#else
def8b4fa 3677 return NULL;
def8b4fa 3678#endif
0b3d8e08 3679}
def8b4fa 3680
68c33163
PS
3681/* Keeps track of mac header offset relative to skb->head.
3682 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3683 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3684 * tunnel skb it points to outer mac header.
3685 * Keeps track of level of encapsulation of network headers.
3686 */
68c33163 3687struct skb_gso_cb {
802ab55a
AD
3688 union {
3689 int mac_offset;
3690 int data_offset;
3691 };
3347c960 3692 int encap_level;
76443456 3693 __wsum csum;
7e2b10c1 3694 __u16 csum_start;
68c33163 3695};
9207f9d4
KK
3696#define SKB_SGO_CB_OFFSET 32
3697#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3698
3699static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3700{
3701 return (skb_mac_header(inner_skb) - inner_skb->head) -
3702 SKB_GSO_CB(inner_skb)->mac_offset;
3703}
3704
1e2bd517
PS
3705static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3706{
3707 int new_headroom, headroom;
3708 int ret;
3709
3710 headroom = skb_headroom(skb);
3711 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3712 if (ret)
3713 return ret;
3714
3715 new_headroom = skb_headroom(skb);
3716 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3717 return 0;
3718}
3719
08b64fcc
AD
3720static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3721{
3722 /* Do not update partial checksums if remote checksum is enabled. */
3723 if (skb->remcsum_offload)
3724 return;
3725
3726 SKB_GSO_CB(skb)->csum = res;
3727 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3728}
3729
7e2b10c1
TH
3730/* Compute the checksum for a gso segment. First compute the checksum value
3731 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3732 * then add in skb->csum (checksum from csum_start to end of packet).
3733 * skb->csum and csum_start are then updated to reflect the checksum of the
3734 * resultant packet starting from the transport header-- the resultant checksum
3735 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3736 * header.
3737 */
3738static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3739{
76443456
AD
3740 unsigned char *csum_start = skb_transport_header(skb);
3741 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3742 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3743
76443456
AD
3744 SKB_GSO_CB(skb)->csum = res;
3745 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3746
76443456 3747 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3748}
3749
bdcc0924 3750static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3751{
3752 return skb_shinfo(skb)->gso_size;
3753}
3754
36a8f39e 3755/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3756static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3757{
3758 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3759}
3760
5293efe6
DB
3761static inline void skb_gso_reset(struct sk_buff *skb)
3762{
3763 skb_shinfo(skb)->gso_size = 0;
3764 skb_shinfo(skb)->gso_segs = 0;
3765 skb_shinfo(skb)->gso_type = 0;
3766}
3767
7965bd4d 3768void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3769
3770static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3771{
3772 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3773 * wanted then gso_type will be set. */
05bdd2f1
ED
3774 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3775
b78462eb
AD
3776 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3777 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3778 __skb_warn_lro_forwarding(skb);
3779 return true;
3780 }
3781 return false;
3782}
3783
35fc92a9
HX
3784static inline void skb_forward_csum(struct sk_buff *skb)
3785{
3786 /* Unfortunately we don't support this one. Any brave souls? */
3787 if (skb->ip_summed == CHECKSUM_COMPLETE)
3788 skb->ip_summed = CHECKSUM_NONE;
3789}
3790
bc8acf2c
ED
3791/**
3792 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3793 * @skb: skb to check
3794 *
3795 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3796 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3797 * use this helper, to document places where we make this assertion.
3798 */
05bdd2f1 3799static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3800{
3801#ifdef DEBUG
3802 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3803#endif
3804}
3805
f35d9d8a 3806bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3807
ed1f50c3 3808int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3809struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3810 unsigned int transport_len,
3811 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3812
3a7c1ee4
AD
3813/**
3814 * skb_head_is_locked - Determine if the skb->head is locked down
3815 * @skb: skb to check
3816 *
3817 * The head on skbs build around a head frag can be removed if they are
3818 * not cloned. This function returns true if the skb head is locked down
3819 * due to either being allocated via kmalloc, or by being a clone with
3820 * multiple references to the head.
3821 */
3822static inline bool skb_head_is_locked(const struct sk_buff *skb)
3823{
3824 return !skb->head_frag || skb_cloned(skb);
3825}
fe6cc55f
FW
3826
3827/**
3828 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3829 *
3830 * @skb: GSO skb
3831 *
3832 * skb_gso_network_seglen is used to determine the real size of the
3833 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3834 *
3835 * The MAC/L2 header is not accounted for.
3836 */
3837static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3838{
3839 unsigned int hdr_len = skb_transport_header(skb) -
3840 skb_network_header(skb);
3841 return hdr_len + skb_gso_transport_seglen(skb);
3842}
ee122c79 3843
179bc67f
EC
3844/* Local Checksum Offload.
3845 * Compute outer checksum based on the assumption that the
3846 * inner checksum will be offloaded later.
e8ae7b00
EC
3847 * See Documentation/networking/checksum-offloads.txt for
3848 * explanation of how this works.
179bc67f
EC
3849 * Fill in outer checksum adjustment (e.g. with sum of outer
3850 * pseudo-header) before calling.
3851 * Also ensure that inner checksum is in linear data area.
3852 */
3853static inline __wsum lco_csum(struct sk_buff *skb)
3854{
9e74a6da
AD
3855 unsigned char *csum_start = skb_checksum_start(skb);
3856 unsigned char *l4_hdr = skb_transport_header(skb);
3857 __wsum partial;
179bc67f
EC
3858
3859 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3860 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3861 skb->csum_offset));
3862
179bc67f 3863 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3864 * adjustment filled in by caller) and return result.
179bc67f 3865 */
9e74a6da 3866 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3867}
3868
1da177e4
LT
3869#endif /* __KERNEL__ */
3870#endif /* _LINUX_SKBUFF_H */