]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/skbuff.h
UBUNTU: SAUCE: LSM stacking: procfs: add smack subdir to attrs
[mirror_ubuntu-artful-kernel.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.
b4759dcd 112 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
78ea85f1
DB
123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
b4759dcd
DC
130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
134 *
135 * CHECKSUM_PARTIAL:
136 *
6edec0e6
TH
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
78ea85f1 146 *
7a6ae71b
TH
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
149 *
150 * CHECKSUM_PARTIAL:
151 *
7a6ae71b 152 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 153 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
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
dba00306
DC
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
78ea85f1
DB
217 */
218
60476372 219/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
1da177e4 224
77cffe23
TH
225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
0bec8c88 228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 229#define SKB_WITH_OVERHEAD(X) \
deea84b0 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
87fb4b7b
ED
236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
1da177e4 241struct net_device;
716ea3a7 242struct scatterlist;
9c55e01c 243struct pipe_inode_info;
a8f820aa 244struct iov_iter;
fd11a83d 245struct napi_struct;
1da177e4 246
5f79e0f9 247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
248struct nf_conntrack {
249 atomic_t use;
1da177e4 250};
5f79e0f9 251#endif
1da177e4 252
34666d46 253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 254struct nf_bridge_info {
53869ceb 255 refcount_t use;
3eaf4025
FW
256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
7fb48c5b 260 } orig_proto:8;
72b1e5e4
FW
261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
411ffb4f 264 __u16 frag_max_size;
bf1ac5ca 265 struct net_device *physindev;
63cdbc06
FW
266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
7fb48c5b 269 union {
72b1e5e4 270 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
72b1e5e4
FW
273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
72b31f72 279 };
1da177e4
LT
280};
281#endif
282
1da177e4
LT
283struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290};
291
292struct sk_buff;
293
9d4dde52
IC
294/* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
a715dea3 300 */
9d4dde52 301#if (65536/PAGE_SIZE + 1) < 16
eec00954 302#define MAX_SKB_FRAGS 16UL
a715dea3 303#else
9d4dde52 304#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 305#endif
5f74f82e 306extern int sysctl_max_skb_frags;
1da177e4 307
3953c46c
MRL
308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
1da177e4
LT
313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
a8605c60
IC
316 struct {
317 struct page *p;
318 } page;
cb4dfe56 319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
320 __u32 page_offset;
321 __u32 size;
cb4dfe56
ED
322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
1da177e4
LT
326};
327
9e903e08
ED
328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
ac45f602
PO
348#define HAVE_HW_TIME_STAMP
349
350/**
d3a21be8 351 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
352 * @hwtstamp: hardware time stamp transformed into duration
353 * since arbitrary point in time
ac45f602
PO
354 *
355 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 356 * skb->tstamp.
ac45f602
PO
357 *
358 * hwtstamps can only be compared against other hwtstamps from
359 * the same device.
360 *
361 * This structure is attached to packets as part of the
362 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
363 */
364struct skb_shared_hwtstamps {
365 ktime_t hwtstamp;
ac45f602
PO
366};
367
2244d07b
OH
368/* Definitions for tx_flags in struct skb_shared_info */
369enum {
370 /* generate hardware time stamp */
371 SKBTX_HW_TSTAMP = 1 << 0,
372
e7fd2885 373 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
374 SKBTX_SW_TSTAMP = 1 << 1,
375
376 /* device driver is going to provide hardware time stamp */
377 SKBTX_IN_PROGRESS = 1 << 2,
378
a6686f2f 379 /* device driver supports TX zero-copy buffers */
62b1a8ab 380 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
381
382 /* generate wifi status information (where possible) */
62b1a8ab 383 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
384
385 /* This indicates at least one fragment might be overwritten
386 * (as in vmsplice(), sendfile() ...)
387 * If we need to compute a TX checksum, we'll need to copy
388 * all frags to avoid possible bad checksum
389 */
390 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
391
392 /* generate software time stamp when entering packet scheduling */
393 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
394};
395
e1c8a607 396#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 397 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
398#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
399
a6686f2f
SM
400/*
401 * The callback notifies userspace to release buffers when skb DMA is done in
402 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
403 * The zerocopy_success argument is true if zero copy transmit occurred,
404 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
405 * The ctx field is used to track device context.
406 * The desc field is used to track userspace buffer index.
a6686f2f
SM
407 */
408struct ubuf_info {
e19d6763 409 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 410 void *ctx;
a6686f2f 411 unsigned long desc;
ac45f602
PO
412};
413
1da177e4
LT
414/* This data is invariant across clones and lives at
415 * the end of the header data, ie. at skb->end.
416 */
417struct skb_shared_info {
7f564528 418 unsigned short _unused;
9f42f126
IC
419 unsigned char nr_frags;
420 __u8 tx_flags;
7967168c
HX
421 unsigned short gso_size;
422 /* Warning: this field is not always filled in (UFO)! */
423 unsigned short gso_segs;
1da177e4 424 struct sk_buff *frag_list;
ac45f602 425 struct skb_shared_hwtstamps hwtstamps;
7f564528 426 unsigned int gso_type;
09c2d251 427 u32 tskey;
9f42f126 428 __be32 ip6_frag_id;
ec7d2f2c
ED
429
430 /*
431 * Warning : all fields before dataref are cleared in __alloc_skb()
432 */
433 atomic_t dataref;
434
69e3c75f
JB
435 /* Intermediate layers must ensure that destructor_arg
436 * remains valid until skb destructor */
437 void * destructor_arg;
a6686f2f 438
fed66381
ED
439 /* must be last field, see pskb_expand_head() */
440 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
441};
442
443/* We divide dataref into two halves. The higher 16 bits hold references
444 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
445 * the entire skb->data. A clone of a headerless skb holds the length of
446 * the header in skb->hdr_len.
1da177e4
LT
447 *
448 * All users must obey the rule that the skb->data reference count must be
449 * greater than or equal to the payload reference count.
450 *
451 * Holding a reference to the payload part means that the user does not
452 * care about modifications to the header part of skb->data.
453 */
454#define SKB_DATAREF_SHIFT 16
455#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
456
d179cd12
DM
457
458enum {
c8753d55
VS
459 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
460 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
461 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
462};
463
7967168c
HX
464enum {
465 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 466 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
467
468 /* This indicates the skb is from an untrusted source. */
469 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
470
471 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
472 SKB_GSO_TCP_ECN = 1 << 3,
473
cbc53e08 474 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 475
cbc53e08 476 SKB_GSO_TCPV6 = 1 << 5,
68c33163 477
cbc53e08 478 SKB_GSO_FCOE = 1 << 6,
73136267 479
cbc53e08 480 SKB_GSO_GRE = 1 << 7,
0d89d203 481
cbc53e08 482 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 483
7e13318d 484 SKB_GSO_IPXIP4 = 1 << 9,
61c1db7f 485
7e13318d 486 SKB_GSO_IPXIP6 = 1 << 10,
0f4f4ffa 487
cbc53e08 488 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 489
cbc53e08
AD
490 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
491
802ab55a
AD
492 SKB_GSO_PARTIAL = 1 << 13,
493
494 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
90017acc
MRL
495
496 SKB_GSO_SCTP = 1 << 15,
c7ef8f0c
SK
497
498 SKB_GSO_ESP = 1 << 16,
7967168c
HX
499};
500
2e07fa9c
ACM
501#if BITS_PER_LONG > 32
502#define NET_SKBUFF_DATA_USES_OFFSET 1
503#endif
504
505#ifdef NET_SKBUFF_DATA_USES_OFFSET
506typedef unsigned int sk_buff_data_t;
507#else
508typedef unsigned char *sk_buff_data_t;
509#endif
510
1da177e4
LT
511/**
512 * struct sk_buff - socket buffer
513 * @next: Next buffer in list
514 * @prev: Previous buffer in list
363ec392 515 * @tstamp: Time we arrived/left
56b17425 516 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 517 * @sk: Socket we are owned by
1da177e4 518 * @dev: Device we arrived on/are leaving by
d84e0bd7 519 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 520 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 521 * @sp: the security path, used for xfrm
1da177e4
LT
522 * @len: Length of actual data
523 * @data_len: Data length
524 * @mac_len: Length of link layer header
334a8132 525 * @hdr_len: writable header length of cloned skb
663ead3b
HX
526 * @csum: Checksum (must include start/offset pair)
527 * @csum_start: Offset from skb->head where checksumming should start
528 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 529 * @priority: Packet queueing priority
60ff7467 530 * @ignore_df: allow local fragmentation
1da177e4 531 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 532 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
533 * @nohdr: Payload reference only, must not modify header
534 * @pkt_type: Packet class
c83c2486 535 * @fclone: skbuff clone status
c83c2486 536 * @ipvs_property: skbuff is owned by ipvs
e7246e12 537 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 538 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
539 * @tc_redirected: packet was redirected by a tc action
540 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
541 * @peeked: this packet has been seen already, so stats have been
542 * done for it, don't do them again
ba9dda3a 543 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
544 * @protocol: Packet protocol from driver
545 * @destructor: Destruct function
a9e419dc 546 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 547 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 548 * @skb_iif: ifindex of device we arrived on
1da177e4 549 * @tc_index: Traffic control index
61b905da 550 * @hash: the packet hash
d84e0bd7 551 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 552 * @xmit_more: More SKBs are pending for this queue
553a5672 553 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 554 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 555 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 556 * ports.
a3b18ddb 557 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
558 * @wifi_acked_valid: wifi_acked was set
559 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 560 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
dba00306 561 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
4ff06203 562 * @dst_pending_confirm: need to confirm neighbour
06021292 563 * @napi_id: id of the NAPI struct this skb came from
984bc16c 564 * @secmark: security marking
d84e0bd7 565 * @mark: Generic packet mark
86a9bad3 566 * @vlan_proto: vlan encapsulation protocol
6aa895b0 567 * @vlan_tci: vlan tag control information
0d89d203 568 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
569 * @inner_transport_header: Inner transport layer header (encapsulation)
570 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 571 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
572 * @transport_header: Transport layer header
573 * @network_header: Network layer header
574 * @mac_header: Link layer header
575 * @tail: Tail pointer
576 * @end: End pointer
577 * @head: Head of buffer
578 * @data: Data head pointer
579 * @truesize: Buffer size
580 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
581 */
582
583struct sk_buff {
363ec392 584 union {
56b17425
ED
585 struct {
586 /* These two members must be first. */
587 struct sk_buff *next;
588 struct sk_buff *prev;
589
590 union {
591 ktime_t tstamp;
9a568de4 592 u64 skb_mstamp;
56b17425
ED
593 };
594 };
595 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 596 };
da3f5cf1 597 struct sock *sk;
1da177e4 598
c84d9490
ED
599 union {
600 struct net_device *dev;
601 /* Some protocols might use this space to store information,
602 * while device pointer would be NULL.
603 * UDP receive path is one user.
604 */
605 unsigned long dev_scratch;
606 };
1da177e4
LT
607 /*
608 * This is the control buffer. It is free to use for every
609 * layer. Please put your private variables there. If you
610 * want to keep them across layers you have to do a skb_clone()
611 * first. This is owned by whoever has the skb queued ATM.
612 */
da3f5cf1 613 char cb[48] __aligned(8);
1da177e4 614
7fee226a 615 unsigned long _skb_refdst;
b1937227 616 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
617#ifdef CONFIG_XFRM
618 struct sec_path *sp;
b1937227
ED
619#endif
620#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 621 unsigned long _nfct;
b1937227 622#endif
85224844 623#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 624 struct nf_bridge_info *nf_bridge;
da3f5cf1 625#endif
1da177e4 626 unsigned int len,
334a8132
PM
627 data_len;
628 __u16 mac_len,
629 hdr_len;
b1937227
ED
630
631 /* Following fields are _not_ copied in __copy_skb_header()
632 * Note that queue_mapping is here mostly to fill a hole.
633 */
fe55f6d5 634 kmemcheck_bitfield_begin(flags1);
b1937227 635 __u16 queue_mapping;
36bbef52
DB
636
637/* if you move cloned around you also must adapt those constants */
638#ifdef __BIG_ENDIAN_BITFIELD
639#define CLONED_MASK (1 << 7)
640#else
641#define CLONED_MASK 1
642#endif
643#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
644
645 __u8 __cloned_offset[0];
b1937227 646 __u8 cloned:1,
6869c4d8 647 nohdr:1,
b84f4cc9 648 fclone:2,
a59322be 649 peeked:1,
b1937227 650 head_frag:1,
36bbef52
DB
651 xmit_more:1,
652 __unused:1; /* one bit hole */
fe55f6d5 653 kmemcheck_bitfield_end(flags1);
4031ae6e 654
b1937227
ED
655 /* fields enclosed in headers_start/headers_end are copied
656 * using a single memcpy() in __copy_skb_header()
657 */
ebcf34f3 658 /* private: */
b1937227 659 __u32 headers_start[0];
ebcf34f3 660 /* public: */
4031ae6e 661
233577a2
HFS
662/* if you move pkt_type around you also must adapt those constants */
663#ifdef __BIG_ENDIAN_BITFIELD
664#define PKT_TYPE_MAX (7 << 5)
665#else
666#define PKT_TYPE_MAX 7
1da177e4 667#endif
233577a2 668#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 669
233577a2 670 __u8 __pkt_type_offset[0];
b1937227 671 __u8 pkt_type:3;
c93bdd0e 672 __u8 pfmemalloc:1;
b1937227 673 __u8 ignore_df:1;
b1937227
ED
674
675 __u8 nf_trace:1;
676 __u8 ip_summed:2;
3853b584 677 __u8 ooo_okay:1;
61b905da 678 __u8 l4_hash:1;
a3b18ddb 679 __u8 sw_hash:1;
6e3e939f
JB
680 __u8 wifi_acked_valid:1;
681 __u8 wifi_acked:1;
b1937227 682
3bdc0eba 683 __u8 no_fcs:1;
77cffe23 684 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 685 __u8 encapsulation:1;
7e2b10c1 686 __u8 encap_hdr_csum:1;
5d0c2b95 687 __u8 csum_valid:1;
7e3cead5 688 __u8 csum_complete_sw:1;
b1937227 689 __u8 csum_level:2;
dba00306 690 __u8 csum_not_inet:1;
fe55f6d5 691
4ff06203 692 __u8 dst_pending_confirm:1;
b1937227
ED
693#ifdef CONFIG_IPV6_NDISC_NODETYPE
694 __u8 ndisc_nodetype:2;
695#endif
696 __u8 ipvs_property:1;
8bce6d7d 697 __u8 inner_protocol_type:1;
e585f236 698 __u8 remcsum_offload:1;
6bc506b4
IS
699#ifdef CONFIG_NET_SWITCHDEV
700 __u8 offload_fwd_mark:1;
701#endif
e7246e12
WB
702#ifdef CONFIG_NET_CLS_ACT
703 __u8 tc_skip_classify:1;
8dc07fdb 704 __u8 tc_at_ingress:1;
bc31c905
WB
705 __u8 tc_redirected:1;
706 __u8 tc_from_ingress:1;
e7246e12 707#endif
b1937227
ED
708
709#ifdef CONFIG_NET_SCHED
710 __u16 tc_index; /* traffic control index */
b1937227 711#endif
fe55f6d5 712
b1937227
ED
713 union {
714 __wsum csum;
715 struct {
716 __u16 csum_start;
717 __u16 csum_offset;
718 };
719 };
720 __u32 priority;
721 int skb_iif;
722 __u32 hash;
723 __be16 vlan_proto;
724 __u16 vlan_tci;
2bd82484
ED
725#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
726 union {
727 unsigned int napi_id;
728 unsigned int sender_cpu;
729 };
97fc2f08 730#endif
984bc16c 731#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 732 __u32 secmark;
0c4f691f 733#endif
0c4f691f 734
3b885787
NH
735 union {
736 __u32 mark;
16fad69c 737 __u32 reserved_tailroom;
3b885787 738 };
1da177e4 739
8bce6d7d
TH
740 union {
741 __be16 inner_protocol;
742 __u8 inner_ipproto;
743 };
744
1a37e412
SH
745 __u16 inner_transport_header;
746 __u16 inner_network_header;
747 __u16 inner_mac_header;
b1937227
ED
748
749 __be16 protocol;
1a37e412
SH
750 __u16 transport_header;
751 __u16 network_header;
752 __u16 mac_header;
b1937227 753
ebcf34f3 754 /* private: */
b1937227 755 __u32 headers_end[0];
ebcf34f3 756 /* public: */
b1937227 757
1da177e4 758 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 759 sk_buff_data_t tail;
4305b541 760 sk_buff_data_t end;
1da177e4 761 unsigned char *head,
4305b541 762 *data;
27a884dc 763 unsigned int truesize;
63354797 764 refcount_t users;
1da177e4
LT
765};
766
767#ifdef __KERNEL__
768/*
769 * Handling routines are only of interest to the kernel
770 */
771#include <linux/slab.h>
772
1da177e4 773
c93bdd0e
MG
774#define SKB_ALLOC_FCLONE 0x01
775#define SKB_ALLOC_RX 0x02
fd11a83d 776#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
777
778/* Returns true if the skb was allocated from PFMEMALLOC reserves */
779static inline bool skb_pfmemalloc(const struct sk_buff *skb)
780{
781 return unlikely(skb->pfmemalloc);
782}
783
7fee226a
ED
784/*
785 * skb might have a dst pointer attached, refcounted or not.
786 * _skb_refdst low order bit is set if refcount was _not_ taken
787 */
788#define SKB_DST_NOREF 1UL
789#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
790
a9e419dc 791#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
792/**
793 * skb_dst - returns skb dst_entry
794 * @skb: buffer
795 *
796 * Returns skb dst_entry, regardless of reference taken or not.
797 */
adf30907
ED
798static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
799{
7fee226a
ED
800 /* If refdst was not refcounted, check we still are in a
801 * rcu_read_lock section
802 */
803 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
804 !rcu_read_lock_held() &&
805 !rcu_read_lock_bh_held());
806 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
807}
808
7fee226a
ED
809/**
810 * skb_dst_set - sets skb dst
811 * @skb: buffer
812 * @dst: dst entry
813 *
814 * Sets skb dst, assuming a reference was taken on dst and should
815 * be released by skb_dst_drop()
816 */
adf30907
ED
817static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
818{
7fee226a
ED
819 skb->_skb_refdst = (unsigned long)dst;
820}
821
932bc4d7
JA
822/**
823 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
824 * @skb: buffer
825 * @dst: dst entry
826 *
827 * Sets skb dst, assuming a reference was not taken on dst.
828 * If dst entry is cached, we do not take reference and dst_release
829 * will be avoided by refdst_drop. If dst entry is not cached, we take
830 * reference, so that last dst_release can destroy the dst immediately.
831 */
832static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
833{
dbfc4fb7
HFS
834 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
835 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 836}
7fee226a
ED
837
838/**
25985edc 839 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
840 * @skb: buffer
841 */
842static inline bool skb_dst_is_noref(const struct sk_buff *skb)
843{
844 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
845}
846
511c3f92
ED
847static inline struct rtable *skb_rtable(const struct sk_buff *skb)
848{
adf30907 849 return (struct rtable *)skb_dst(skb);
511c3f92
ED
850}
851
8b10cab6
JHS
852/* For mangling skb->pkt_type from user space side from applications
853 * such as nft, tc, etc, we only allow a conservative subset of
854 * possible pkt_types to be set.
855*/
856static inline bool skb_pkt_type_ok(u32 ptype)
857{
858 return ptype <= PACKET_OTHERHOST;
859}
860
90b602f8
ML
861static inline unsigned int skb_napi_id(const struct sk_buff *skb)
862{
863#ifdef CONFIG_NET_RX_BUSY_POLL
864 return skb->napi_id;
865#else
866 return 0;
867#endif
868}
869
3889a803
PA
870/* decrement the reference count and return true if we can free the skb */
871static inline bool skb_unref(struct sk_buff *skb)
872{
873 if (unlikely(!skb))
874 return false;
63354797 875 if (likely(refcount_read(&skb->users) == 1))
3889a803 876 smp_rmb();
63354797 877 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
878 return false;
879
880 return true;
881}
882
0a463c78 883void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
884void kfree_skb(struct sk_buff *skb);
885void kfree_skb_list(struct sk_buff *segs);
886void skb_tx_error(struct sk_buff *skb);
887void consume_skb(struct sk_buff *skb);
bfd6b37e 888void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 889void __kfree_skb(struct sk_buff *skb);
d7e8883c 890extern struct kmem_cache *skbuff_head_cache;
bad43ca8 891
7965bd4d
JP
892void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
893bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
894 bool *fragstolen, int *delta_truesize);
bad43ca8 895
7965bd4d
JP
896struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
897 int node);
2ea2f62c 898struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 899struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 900static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 901 gfp_t priority)
d179cd12 902{
564824b0 903 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
904}
905
2e4e4410
ED
906struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
907 unsigned long data_len,
908 int max_page_order,
909 int *errcode,
910 gfp_t gfp_mask);
911
d0bf4a9e
ED
912/* Layout of fast clones : [skb1][skb2][fclone_ref] */
913struct sk_buff_fclones {
914 struct sk_buff skb1;
915
916 struct sk_buff skb2;
917
2638595a 918 refcount_t fclone_ref;
d0bf4a9e
ED
919};
920
921/**
922 * skb_fclone_busy - check if fclone is busy
293de7de 923 * @sk: socket
d0bf4a9e
ED
924 * @skb: buffer
925 *
bda13fed 926 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
927 * Some drivers call skb_orphan() in their ndo_start_xmit(),
928 * so we also check that this didnt happen.
d0bf4a9e 929 */
39bb5e62
ED
930static inline bool skb_fclone_busy(const struct sock *sk,
931 const struct sk_buff *skb)
d0bf4a9e
ED
932{
933 const struct sk_buff_fclones *fclones;
934
935 fclones = container_of(skb, struct sk_buff_fclones, skb1);
936
937 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 938 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 939 fclones->skb2.sk == sk;
d0bf4a9e
ED
940}
941
d179cd12 942static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 943 gfp_t priority)
d179cd12 944{
c93bdd0e 945 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
946}
947
7965bd4d 948struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
949static inline struct sk_buff *alloc_skb_head(gfp_t priority)
950{
951 return __alloc_skb_head(priority, -1);
952}
953
7965bd4d
JP
954struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
955int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
956struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
957struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
958struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
959 gfp_t gfp_mask, bool fclone);
960static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
961 gfp_t gfp_mask)
962{
963 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
964}
7965bd4d
JP
965
966int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
967struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
968 unsigned int headroom);
969struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
970 int newtailroom, gfp_t priority);
48a1df65
JD
971int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
972 int offset, int len);
973int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
974 int offset, int len);
7965bd4d 975int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
976int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
977
978/**
979 * skb_pad - zero pad the tail of an skb
980 * @skb: buffer to pad
981 * @pad: space to pad
982 *
983 * Ensure that a buffer is followed by a padding area that is zero
984 * filled. Used by network drivers which may DMA or transfer data
985 * beyond the buffer end onto the wire.
986 *
987 * May return error in out of memory cases. The skb is freed on error.
988 */
989static inline int skb_pad(struct sk_buff *skb, int pad)
990{
991 return __skb_pad(skb, pad, true);
992}
ead2ceb0 993#define dev_kfree_skb(a) consume_skb(a)
1da177e4 994
7965bd4d
JP
995int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
996 int getfrag(void *from, char *to, int offset,
997 int len, int odd, struct sk_buff *skb),
998 void *from, int length);
e89e9cf5 999
be12a1fe
HFS
1000int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1001 int offset, size_t size);
1002
d94d9fee 1003struct skb_seq_state {
677e90ed
TG
1004 __u32 lower_offset;
1005 __u32 upper_offset;
1006 __u32 frag_idx;
1007 __u32 stepped_offset;
1008 struct sk_buff *root_skb;
1009 struct sk_buff *cur_skb;
1010 __u8 *frag_data;
1011};
1012
7965bd4d
JP
1013void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1014 unsigned int to, struct skb_seq_state *st);
1015unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1016 struct skb_seq_state *st);
1017void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1018
7965bd4d 1019unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1020 unsigned int to, struct ts_config *config);
3fc7e8a6 1021
09323cc4
TH
1022/*
1023 * Packet hash types specify the type of hash in skb_set_hash.
1024 *
1025 * Hash types refer to the protocol layer addresses which are used to
1026 * construct a packet's hash. The hashes are used to differentiate or identify
1027 * flows of the protocol layer for the hash type. Hash types are either
1028 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1029 *
1030 * Properties of hashes:
1031 *
1032 * 1) Two packets in different flows have different hash values
1033 * 2) Two packets in the same flow should have the same hash value
1034 *
1035 * A hash at a higher layer is considered to be more specific. A driver should
1036 * set the most specific hash possible.
1037 *
1038 * A driver cannot indicate a more specific hash than the layer at which a hash
1039 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1040 *
1041 * A driver may indicate a hash level which is less specific than the
1042 * actual layer the hash was computed on. For instance, a hash computed
1043 * at L4 may be considered an L3 hash. This should only be done if the
1044 * driver can't unambiguously determine that the HW computed the hash at
1045 * the higher layer. Note that the "should" in the second property above
1046 * permits this.
1047 */
1048enum pkt_hash_types {
1049 PKT_HASH_TYPE_NONE, /* Undefined type */
1050 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1051 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1052 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1053};
1054
bcc83839 1055static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1056{
bcc83839 1057 skb->hash = 0;
a3b18ddb 1058 skb->sw_hash = 0;
bcc83839
TH
1059 skb->l4_hash = 0;
1060}
1061
1062static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1063{
1064 if (!skb->l4_hash)
1065 skb_clear_hash(skb);
1066}
1067
1068static inline void
1069__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1070{
1071 skb->l4_hash = is_l4;
1072 skb->sw_hash = is_sw;
61b905da 1073 skb->hash = hash;
09323cc4
TH
1074}
1075
bcc83839
TH
1076static inline void
1077skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1078{
1079 /* Used by drivers to set hash from HW */
1080 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1081}
1082
1083static inline void
1084__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1085{
1086 __skb_set_hash(skb, hash, true, is_l4);
1087}
1088
e5276937 1089void __skb_get_hash(struct sk_buff *skb);
b917783c 1090u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1091u32 skb_get_poff(const struct sk_buff *skb);
1092u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1093 const struct flow_keys *keys, int hlen);
1094__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1095 void *data, int hlen_proto);
1096
1097static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1098 int thoff, u8 ip_proto)
1099{
1100 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1101}
1102
1103void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1104 const struct flow_dissector_key *key,
1105 unsigned int key_count);
1106
1107bool __skb_flow_dissect(const struct sk_buff *skb,
1108 struct flow_dissector *flow_dissector,
1109 void *target_container,
cd79a238
TH
1110 void *data, __be16 proto, int nhoff, int hlen,
1111 unsigned int flags);
e5276937
TH
1112
1113static inline bool skb_flow_dissect(const struct sk_buff *skb,
1114 struct flow_dissector *flow_dissector,
cd79a238 1115 void *target_container, unsigned int flags)
e5276937
TH
1116{
1117 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1118 NULL, 0, 0, 0, flags);
e5276937
TH
1119}
1120
1121static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1122 struct flow_keys *flow,
1123 unsigned int flags)
e5276937
TH
1124{
1125 memset(flow, 0, sizeof(*flow));
1126 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1127 NULL, 0, 0, 0, flags);
e5276937
TH
1128}
1129
1130static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1131 void *data, __be16 proto,
cd79a238
TH
1132 int nhoff, int hlen,
1133 unsigned int flags)
e5276937
TH
1134{
1135 memset(flow, 0, sizeof(*flow));
1136 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1137 data, proto, nhoff, hlen, flags);
e5276937
TH
1138}
1139
3958afa1 1140static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1141{
a3b18ddb 1142 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1143 __skb_get_hash(skb);
bfb564e7 1144
61b905da 1145 return skb->hash;
bfb564e7
KK
1146}
1147
20a17bf6 1148__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1149
20a17bf6 1150static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1151{
c6cc1ca7
TH
1152 if (!skb->l4_hash && !skb->sw_hash) {
1153 struct flow_keys keys;
de4c1f8b 1154 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1155
de4c1f8b 1156 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1157 }
f70ea018
TH
1158
1159 return skb->hash;
1160}
1161
20a17bf6 1162__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1163
20a17bf6 1164static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1165{
c6cc1ca7
TH
1166 if (!skb->l4_hash && !skb->sw_hash) {
1167 struct flow_keys keys;
de4c1f8b 1168 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1169
de4c1f8b 1170 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1171 }
f70ea018
TH
1172
1173 return skb->hash;
1174}
1175
50fb7992
TH
1176__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1177
57bdf7f4
TH
1178static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1179{
61b905da 1180 return skb->hash;
57bdf7f4
TH
1181}
1182
3df7a74e
TH
1183static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1184{
61b905da 1185 to->hash = from->hash;
a3b18ddb 1186 to->sw_hash = from->sw_hash;
61b905da 1187 to->l4_hash = from->l4_hash;
3df7a74e
TH
1188};
1189
4305b541
ACM
1190#ifdef NET_SKBUFF_DATA_USES_OFFSET
1191static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1192{
1193 return skb->head + skb->end;
1194}
ec47ea82
AD
1195
1196static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1197{
1198 return skb->end;
1199}
4305b541
ACM
1200#else
1201static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1202{
1203 return skb->end;
1204}
ec47ea82
AD
1205
1206static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1207{
1208 return skb->end - skb->head;
1209}
4305b541
ACM
1210#endif
1211
1da177e4 1212/* Internal */
4305b541 1213#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1214
ac45f602
PO
1215static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1216{
1217 return &skb_shinfo(skb)->hwtstamps;
1218}
1219
1da177e4
LT
1220/**
1221 * skb_queue_empty - check if a queue is empty
1222 * @list: queue head
1223 *
1224 * Returns true if the queue is empty, false otherwise.
1225 */
1226static inline int skb_queue_empty(const struct sk_buff_head *list)
1227{
fd44b93c 1228 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1229}
1230
fc7ebb21
DM
1231/**
1232 * skb_queue_is_last - check if skb is the last entry in the queue
1233 * @list: queue head
1234 * @skb: buffer
1235 *
1236 * Returns true if @skb is the last buffer on the list.
1237 */
1238static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1239 const struct sk_buff *skb)
1240{
fd44b93c 1241 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1242}
1243
832d11c5
IJ
1244/**
1245 * skb_queue_is_first - check if skb is the first entry in the queue
1246 * @list: queue head
1247 * @skb: buffer
1248 *
1249 * Returns true if @skb is the first buffer on the list.
1250 */
1251static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1252 const struct sk_buff *skb)
1253{
fd44b93c 1254 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1255}
1256
249c8b42
DM
1257/**
1258 * skb_queue_next - return the next packet in the queue
1259 * @list: queue head
1260 * @skb: current buffer
1261 *
1262 * Return the next packet in @list after @skb. It is only valid to
1263 * call this if skb_queue_is_last() evaluates to false.
1264 */
1265static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1266 const struct sk_buff *skb)
1267{
1268 /* This BUG_ON may seem severe, but if we just return then we
1269 * are going to dereference garbage.
1270 */
1271 BUG_ON(skb_queue_is_last(list, skb));
1272 return skb->next;
1273}
1274
832d11c5
IJ
1275/**
1276 * skb_queue_prev - return the prev packet in the queue
1277 * @list: queue head
1278 * @skb: current buffer
1279 *
1280 * Return the prev packet in @list before @skb. It is only valid to
1281 * call this if skb_queue_is_first() evaluates to false.
1282 */
1283static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1284 const struct sk_buff *skb)
1285{
1286 /* This BUG_ON may seem severe, but if we just return then we
1287 * are going to dereference garbage.
1288 */
1289 BUG_ON(skb_queue_is_first(list, skb));
1290 return skb->prev;
1291}
1292
1da177e4
LT
1293/**
1294 * skb_get - reference buffer
1295 * @skb: buffer to reference
1296 *
1297 * Makes another reference to a socket buffer and returns a pointer
1298 * to the buffer.
1299 */
1300static inline struct sk_buff *skb_get(struct sk_buff *skb)
1301{
63354797 1302 refcount_inc(&skb->users);
1da177e4
LT
1303 return skb;
1304}
1305
1306/*
1307 * If users == 1, we are the only owner and are can avoid redundant
1308 * atomic change.
1309 */
1310
1da177e4
LT
1311/**
1312 * skb_cloned - is the buffer a clone
1313 * @skb: buffer to check
1314 *
1315 * Returns true if the buffer was generated with skb_clone() and is
1316 * one of multiple shared copies of the buffer. Cloned buffers are
1317 * shared data so must not be written to under normal circumstances.
1318 */
1319static inline int skb_cloned(const struct sk_buff *skb)
1320{
1321 return skb->cloned &&
1322 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1323}
1324
14bbd6a5
PS
1325static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1326{
d0164adc 1327 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1328
1329 if (skb_cloned(skb))
1330 return pskb_expand_head(skb, 0, 0, pri);
1331
1332 return 0;
1333}
1334
1da177e4
LT
1335/**
1336 * skb_header_cloned - is the header a clone
1337 * @skb: buffer to check
1338 *
1339 * Returns true if modifying the header part of the buffer requires
1340 * the data to be copied.
1341 */
1342static inline int skb_header_cloned(const struct sk_buff *skb)
1343{
1344 int dataref;
1345
1346 if (!skb->cloned)
1347 return 0;
1348
1349 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1350 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1351 return dataref != 1;
1352}
1353
9580bf2e
ED
1354static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1355{
1356 might_sleep_if(gfpflags_allow_blocking(pri));
1357
1358 if (skb_header_cloned(skb))
1359 return pskb_expand_head(skb, 0, 0, pri);
1360
1361 return 0;
1362}
1363
1da177e4
LT
1364/**
1365 * skb_header_release - release reference to header
1366 * @skb: buffer to operate on
1367 *
1368 * Drop a reference to the header part of the buffer. This is done
1369 * by acquiring a payload reference. You must not read from the header
1370 * part of skb->data after this.
f4a775d1 1371 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1372 */
1373static inline void skb_header_release(struct sk_buff *skb)
1374{
1375 BUG_ON(skb->nohdr);
1376 skb->nohdr = 1;
1377 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1378}
1379
f4a775d1
ED
1380/**
1381 * __skb_header_release - release reference to header
1382 * @skb: buffer to operate on
1383 *
1384 * Variant of skb_header_release() assuming skb is private to caller.
1385 * We can avoid one atomic operation.
1386 */
1387static inline void __skb_header_release(struct sk_buff *skb)
1388{
1389 skb->nohdr = 1;
1390 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1391}
1392
1393
1da177e4
LT
1394/**
1395 * skb_shared - is the buffer shared
1396 * @skb: buffer to check
1397 *
1398 * Returns true if more than one person has a reference to this
1399 * buffer.
1400 */
1401static inline int skb_shared(const struct sk_buff *skb)
1402{
63354797 1403 return refcount_read(&skb->users) != 1;
1da177e4
LT
1404}
1405
1406/**
1407 * skb_share_check - check if buffer is shared and if so clone it
1408 * @skb: buffer to check
1409 * @pri: priority for memory allocation
1410 *
1411 * If the buffer is shared the buffer is cloned and the old copy
1412 * drops a reference. A new clone with a single reference is returned.
1413 * If the buffer is not shared the original buffer is returned. When
1414 * being called from interrupt status or with spinlocks held pri must
1415 * be GFP_ATOMIC.
1416 *
1417 * NULL is returned on a memory allocation failure.
1418 */
47061bc4 1419static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1420{
d0164adc 1421 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1422 if (skb_shared(skb)) {
1423 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1424
1425 if (likely(nskb))
1426 consume_skb(skb);
1427 else
1428 kfree_skb(skb);
1da177e4
LT
1429 skb = nskb;
1430 }
1431 return skb;
1432}
1433
1434/*
1435 * Copy shared buffers into a new sk_buff. We effectively do COW on
1436 * packets to handle cases where we have a local reader and forward
1437 * and a couple of other messy ones. The normal one is tcpdumping
1438 * a packet thats being forwarded.
1439 */
1440
1441/**
1442 * skb_unshare - make a copy of a shared buffer
1443 * @skb: buffer to check
1444 * @pri: priority for memory allocation
1445 *
1446 * If the socket buffer is a clone then this function creates a new
1447 * copy of the data, drops a reference count on the old copy and returns
1448 * the new copy with the reference count at 1. If the buffer is not a clone
1449 * the original buffer is returned. When called with a spinlock held or
1450 * from interrupt state @pri must be %GFP_ATOMIC
1451 *
1452 * %NULL is returned on a memory allocation failure.
1453 */
e2bf521d 1454static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1455 gfp_t pri)
1da177e4 1456{
d0164adc 1457 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1458 if (skb_cloned(skb)) {
1459 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1460
1461 /* Free our shared copy */
1462 if (likely(nskb))
1463 consume_skb(skb);
1464 else
1465 kfree_skb(skb);
1da177e4
LT
1466 skb = nskb;
1467 }
1468 return skb;
1469}
1470
1471/**
1a5778aa 1472 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1473 * @list_: list to peek at
1474 *
1475 * Peek an &sk_buff. Unlike most other operations you _MUST_
1476 * be careful with this one. A peek leaves the buffer on the
1477 * list and someone else may run off with it. You must hold
1478 * the appropriate locks or have a private queue to do this.
1479 *
1480 * Returns %NULL for an empty list or a pointer to the head element.
1481 * The reference count is not incremented and the reference is therefore
1482 * volatile. Use with caution.
1483 */
05bdd2f1 1484static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1485{
18d07000
ED
1486 struct sk_buff *skb = list_->next;
1487
1488 if (skb == (struct sk_buff *)list_)
1489 skb = NULL;
1490 return skb;
1da177e4
LT
1491}
1492
da5ef6e5
PE
1493/**
1494 * skb_peek_next - peek skb following the given one from a queue
1495 * @skb: skb to start from
1496 * @list_: list to peek at
1497 *
1498 * Returns %NULL when the end of the list is met or a pointer to the
1499 * next element. The reference count is not incremented and the
1500 * reference is therefore volatile. Use with caution.
1501 */
1502static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1503 const struct sk_buff_head *list_)
1504{
1505 struct sk_buff *next = skb->next;
18d07000 1506
da5ef6e5
PE
1507 if (next == (struct sk_buff *)list_)
1508 next = NULL;
1509 return next;
1510}
1511
1da177e4 1512/**
1a5778aa 1513 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1514 * @list_: list to peek at
1515 *
1516 * Peek an &sk_buff. Unlike most other operations you _MUST_
1517 * be careful with this one. A peek leaves the buffer on the
1518 * list and someone else may run off with it. You must hold
1519 * the appropriate locks or have a private queue to do this.
1520 *
1521 * Returns %NULL for an empty list or a pointer to the tail element.
1522 * The reference count is not incremented and the reference is therefore
1523 * volatile. Use with caution.
1524 */
05bdd2f1 1525static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1526{
18d07000
ED
1527 struct sk_buff *skb = list_->prev;
1528
1529 if (skb == (struct sk_buff *)list_)
1530 skb = NULL;
1531 return skb;
1532
1da177e4
LT
1533}
1534
1535/**
1536 * skb_queue_len - get queue length
1537 * @list_: list to measure
1538 *
1539 * Return the length of an &sk_buff queue.
1540 */
1541static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1542{
1543 return list_->qlen;
1544}
1545
67fed459
DM
1546/**
1547 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1548 * @list: queue to initialize
1549 *
1550 * This initializes only the list and queue length aspects of
1551 * an sk_buff_head object. This allows to initialize the list
1552 * aspects of an sk_buff_head without reinitializing things like
1553 * the spinlock. It can also be used for on-stack sk_buff_head
1554 * objects where the spinlock is known to not be used.
1555 */
1556static inline void __skb_queue_head_init(struct sk_buff_head *list)
1557{
1558 list->prev = list->next = (struct sk_buff *)list;
1559 list->qlen = 0;
1560}
1561
76f10ad0
AV
1562/*
1563 * This function creates a split out lock class for each invocation;
1564 * this is needed for now since a whole lot of users of the skb-queue
1565 * infrastructure in drivers have different locking usage (in hardirq)
1566 * than the networking core (in softirq only). In the long run either the
1567 * network layer or drivers should need annotation to consolidate the
1568 * main types of usage into 3 classes.
1569 */
1da177e4
LT
1570static inline void skb_queue_head_init(struct sk_buff_head *list)
1571{
1572 spin_lock_init(&list->lock);
67fed459 1573 __skb_queue_head_init(list);
1da177e4
LT
1574}
1575
c2ecba71
PE
1576static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1577 struct lock_class_key *class)
1578{
1579 skb_queue_head_init(list);
1580 lockdep_set_class(&list->lock, class);
1581}
1582
1da177e4 1583/*
bf299275 1584 * Insert an sk_buff on a list.
1da177e4
LT
1585 *
1586 * The "__skb_xxxx()" functions are the non-atomic ones that
1587 * can only be called with interrupts disabled.
1588 */
7965bd4d
JP
1589void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1590 struct sk_buff_head *list);
bf299275
GR
1591static inline void __skb_insert(struct sk_buff *newsk,
1592 struct sk_buff *prev, struct sk_buff *next,
1593 struct sk_buff_head *list)
1594{
1595 newsk->next = next;
1596 newsk->prev = prev;
1597 next->prev = prev->next = newsk;
1598 list->qlen++;
1599}
1da177e4 1600
67fed459
DM
1601static inline void __skb_queue_splice(const struct sk_buff_head *list,
1602 struct sk_buff *prev,
1603 struct sk_buff *next)
1604{
1605 struct sk_buff *first = list->next;
1606 struct sk_buff *last = list->prev;
1607
1608 first->prev = prev;
1609 prev->next = first;
1610
1611 last->next = next;
1612 next->prev = last;
1613}
1614
1615/**
1616 * skb_queue_splice - join two skb lists, this is designed for stacks
1617 * @list: the new list to add
1618 * @head: the place to add it in the first list
1619 */
1620static inline void skb_queue_splice(const struct sk_buff_head *list,
1621 struct sk_buff_head *head)
1622{
1623 if (!skb_queue_empty(list)) {
1624 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1625 head->qlen += list->qlen;
67fed459
DM
1626 }
1627}
1628
1629/**
d9619496 1630 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1631 * @list: the new list to add
1632 * @head: the place to add it in the first list
1633 *
1634 * The list at @list is reinitialised
1635 */
1636static inline void skb_queue_splice_init(struct sk_buff_head *list,
1637 struct sk_buff_head *head)
1638{
1639 if (!skb_queue_empty(list)) {
1640 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1641 head->qlen += list->qlen;
67fed459
DM
1642 __skb_queue_head_init(list);
1643 }
1644}
1645
1646/**
1647 * skb_queue_splice_tail - join two skb lists, each list being a queue
1648 * @list: the new list to add
1649 * @head: the place to add it in the first list
1650 */
1651static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1652 struct sk_buff_head *head)
1653{
1654 if (!skb_queue_empty(list)) {
1655 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1656 head->qlen += list->qlen;
67fed459
DM
1657 }
1658}
1659
1660/**
d9619496 1661 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1662 * @list: the new list to add
1663 * @head: the place to add it in the first list
1664 *
1665 * Each of the lists is a queue.
1666 * The list at @list is reinitialised
1667 */
1668static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1669 struct sk_buff_head *head)
1670{
1671 if (!skb_queue_empty(list)) {
1672 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1673 head->qlen += list->qlen;
67fed459
DM
1674 __skb_queue_head_init(list);
1675 }
1676}
1677
1da177e4 1678/**
300ce174 1679 * __skb_queue_after - queue a buffer at the list head
1da177e4 1680 * @list: list to use
300ce174 1681 * @prev: place after this buffer
1da177e4
LT
1682 * @newsk: buffer to queue
1683 *
300ce174 1684 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1685 * and you must therefore hold required locks before calling it.
1686 *
1687 * A buffer cannot be placed on two lists at the same time.
1688 */
300ce174
SH
1689static inline void __skb_queue_after(struct sk_buff_head *list,
1690 struct sk_buff *prev,
1691 struct sk_buff *newsk)
1da177e4 1692{
bf299275 1693 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1694}
1695
7965bd4d
JP
1696void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1697 struct sk_buff_head *list);
7de6c033 1698
f5572855
GR
1699static inline void __skb_queue_before(struct sk_buff_head *list,
1700 struct sk_buff *next,
1701 struct sk_buff *newsk)
1702{
1703 __skb_insert(newsk, next->prev, next, list);
1704}
1705
300ce174
SH
1706/**
1707 * __skb_queue_head - queue a buffer at the list head
1708 * @list: list to use
1709 * @newsk: buffer to queue
1710 *
1711 * Queue a buffer at the start of a list. This function takes no locks
1712 * and you must therefore hold required locks before calling it.
1713 *
1714 * A buffer cannot be placed on two lists at the same time.
1715 */
7965bd4d 1716void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1717static inline void __skb_queue_head(struct sk_buff_head *list,
1718 struct sk_buff *newsk)
1719{
1720 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1721}
1722
1da177e4
LT
1723/**
1724 * __skb_queue_tail - queue a buffer at the list tail
1725 * @list: list to use
1726 * @newsk: buffer to queue
1727 *
1728 * Queue a buffer at the end of a list. This function takes no locks
1729 * and you must therefore hold required locks before calling it.
1730 *
1731 * A buffer cannot be placed on two lists at the same time.
1732 */
7965bd4d 1733void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1734static inline void __skb_queue_tail(struct sk_buff_head *list,
1735 struct sk_buff *newsk)
1736{
f5572855 1737 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1738}
1739
1da177e4
LT
1740/*
1741 * remove sk_buff from list. _Must_ be called atomically, and with
1742 * the list known..
1743 */
7965bd4d 1744void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1745static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1746{
1747 struct sk_buff *next, *prev;
1748
1749 list->qlen--;
1750 next = skb->next;
1751 prev = skb->prev;
1752 skb->next = skb->prev = NULL;
1da177e4
LT
1753 next->prev = prev;
1754 prev->next = next;
1755}
1756
f525c06d
GR
1757/**
1758 * __skb_dequeue - remove from the head of the queue
1759 * @list: list to dequeue from
1760 *
1761 * Remove the head of the list. This function does not take any locks
1762 * so must be used with appropriate locks held only. The head item is
1763 * returned or %NULL if the list is empty.
1764 */
7965bd4d 1765struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1766static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1767{
1768 struct sk_buff *skb = skb_peek(list);
1769 if (skb)
1770 __skb_unlink(skb, list);
1771 return skb;
1772}
1da177e4
LT
1773
1774/**
1775 * __skb_dequeue_tail - remove from the tail of the queue
1776 * @list: list to dequeue from
1777 *
1778 * Remove the tail of the list. This function does not take any locks
1779 * so must be used with appropriate locks held only. The tail item is
1780 * returned or %NULL if the list is empty.
1781 */
7965bd4d 1782struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1783static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1784{
1785 struct sk_buff *skb = skb_peek_tail(list);
1786 if (skb)
1787 __skb_unlink(skb, list);
1788 return skb;
1789}
1790
1791
bdcc0924 1792static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1793{
1794 return skb->data_len;
1795}
1796
1797static inline unsigned int skb_headlen(const struct sk_buff *skb)
1798{
1799 return skb->len - skb->data_len;
1800}
1801
c72d8cda 1802static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1da177e4 1803{
c72d8cda 1804 unsigned int i, len = 0;
1da177e4 1805
c72d8cda 1806 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 1807 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1808 return len + skb_headlen(skb);
1809}
1810
131ea667
IC
1811/**
1812 * __skb_fill_page_desc - initialise a paged fragment in an skb
1813 * @skb: buffer containing fragment to be initialised
1814 * @i: paged fragment index to initialise
1815 * @page: the page to use for this fragment
1816 * @off: the offset to the data with @page
1817 * @size: the length of the data
1818 *
1819 * Initialises the @i'th fragment of @skb to point to &size bytes at
1820 * offset @off within @page.
1821 *
1822 * Does not take any additional reference on the fragment.
1823 */
1824static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1825 struct page *page, int off, int size)
1da177e4
LT
1826{
1827 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1828
c48a11c7 1829 /*
2f064f34
MH
1830 * Propagate page pfmemalloc to the skb if we can. The problem is
1831 * that not all callers have unique ownership of the page but rely
1832 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1833 */
a8605c60 1834 frag->page.p = page;
1da177e4 1835 frag->page_offset = off;
9e903e08 1836 skb_frag_size_set(frag, size);
cca7af38
PE
1837
1838 page = compound_head(page);
2f064f34 1839 if (page_is_pfmemalloc(page))
cca7af38 1840 skb->pfmemalloc = true;
131ea667
IC
1841}
1842
1843/**
1844 * skb_fill_page_desc - initialise a paged fragment in an skb
1845 * @skb: buffer containing fragment to be initialised
1846 * @i: paged fragment index to initialise
1847 * @page: the page to use for this fragment
1848 * @off: the offset to the data with @page
1849 * @size: the length of the data
1850 *
1851 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1852 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1853 * addition updates @skb such that @i is the last fragment.
1854 *
1855 * Does not take any additional reference on the fragment.
1856 */
1857static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1858 struct page *page, int off, int size)
1859{
1860 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1861 skb_shinfo(skb)->nr_frags = i + 1;
1862}
1863
7965bd4d
JP
1864void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1865 int size, unsigned int truesize);
654bed16 1866
f8e617e1
JW
1867void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1868 unsigned int truesize);
1869
1da177e4 1870#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1871#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1872#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1873
27a884dc
ACM
1874#ifdef NET_SKBUFF_DATA_USES_OFFSET
1875static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1876{
1877 return skb->head + skb->tail;
1878}
1879
1880static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1881{
1882 skb->tail = skb->data - skb->head;
1883}
1884
1885static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1886{
1887 skb_reset_tail_pointer(skb);
1888 skb->tail += offset;
1889}
7cc46190 1890
27a884dc
ACM
1891#else /* NET_SKBUFF_DATA_USES_OFFSET */
1892static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1893{
1894 return skb->tail;
1895}
1896
1897static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1898{
1899 skb->tail = skb->data;
1900}
1901
1902static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1903{
1904 skb->tail = skb->data + offset;
1905}
4305b541 1906
27a884dc
ACM
1907#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1908
1da177e4
LT
1909/*
1910 * Add data to an sk_buff
1911 */
4df864c1
JB
1912void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1913void *skb_put(struct sk_buff *skb, unsigned int len);
1914static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 1915{
4df864c1 1916 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
1917 SKB_LINEAR_ASSERT(skb);
1918 skb->tail += len;
1919 skb->len += len;
1920 return tmp;
1921}
1922
de77b966 1923static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
1924{
1925 void *tmp = __skb_put(skb, len);
1926
1927 memset(tmp, 0, len);
1928 return tmp;
1929}
1930
1931static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
1932 unsigned int len)
1933{
1934 void *tmp = __skb_put(skb, len);
1935
1936 memcpy(tmp, data, len);
1937 return tmp;
1938}
1939
1940static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
1941{
1942 *(u8 *)__skb_put(skb, 1) = val;
1943}
1944
83ad357d 1945static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 1946{
83ad357d 1947 void *tmp = skb_put(skb, len);
e45a79da
JB
1948
1949 memset(tmp, 0, len);
1950
1951 return tmp;
1952}
1953
59ae1d12
JB
1954static inline void *skb_put_data(struct sk_buff *skb, const void *data,
1955 unsigned int len)
1956{
1957 void *tmp = skb_put(skb, len);
1958
1959 memcpy(tmp, data, len);
1960
1961 return tmp;
1962}
1963
634fef61
JB
1964static inline void skb_put_u8(struct sk_buff *skb, u8 val)
1965{
1966 *(u8 *)skb_put(skb, 1) = val;
1967}
1968
d58ff351
JB
1969void *skb_push(struct sk_buff *skb, unsigned int len);
1970static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
1971{
1972 skb->data -= len;
1973 skb->len += len;
1974 return skb->data;
1975}
1976
af72868b
JB
1977void *skb_pull(struct sk_buff *skb, unsigned int len);
1978static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
1979{
1980 skb->len -= len;
1981 BUG_ON(skb->len < skb->data_len);
1982 return skb->data += len;
1983}
1984
af72868b 1985static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
1986{
1987 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1988}
1989
af72868b 1990void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 1991
af72868b 1992static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
1993{
1994 if (len > skb_headlen(skb) &&
987c402a 1995 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1996 return NULL;
1997 skb->len -= len;
1998 return skb->data += len;
1999}
2000
af72868b 2001static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2002{
2003 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2004}
2005
2006static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2007{
2008 if (likely(len <= skb_headlen(skb)))
2009 return 1;
2010 if (unlikely(len > skb->len))
2011 return 0;
987c402a 2012 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2013}
2014
c8c8b127
ED
2015void skb_condense(struct sk_buff *skb);
2016
1da177e4
LT
2017/**
2018 * skb_headroom - bytes at buffer head
2019 * @skb: buffer to check
2020 *
2021 * Return the number of bytes of free space at the head of an &sk_buff.
2022 */
c2636b4d 2023static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2024{
2025 return skb->data - skb->head;
2026}
2027
2028/**
2029 * skb_tailroom - bytes at buffer end
2030 * @skb: buffer to check
2031 *
2032 * Return the number of bytes of free space at the tail of an sk_buff
2033 */
2034static inline int skb_tailroom(const struct sk_buff *skb)
2035{
4305b541 2036 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2037}
2038
a21d4572
ED
2039/**
2040 * skb_availroom - bytes at buffer end
2041 * @skb: buffer to check
2042 *
2043 * Return the number of bytes of free space at the tail of an sk_buff
2044 * allocated by sk_stream_alloc()
2045 */
2046static inline int skb_availroom(const struct sk_buff *skb)
2047{
16fad69c
ED
2048 if (skb_is_nonlinear(skb))
2049 return 0;
2050
2051 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2052}
2053
1da177e4
LT
2054/**
2055 * skb_reserve - adjust headroom
2056 * @skb: buffer to alter
2057 * @len: bytes to move
2058 *
2059 * Increase the headroom of an empty &sk_buff by reducing the tail
2060 * room. This is only allowed for an empty buffer.
2061 */
8243126c 2062static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2063{
2064 skb->data += len;
2065 skb->tail += len;
2066}
2067
1837b2e2
BP
2068/**
2069 * skb_tailroom_reserve - adjust reserved_tailroom
2070 * @skb: buffer to alter
2071 * @mtu: maximum amount of headlen permitted
2072 * @needed_tailroom: minimum amount of reserved_tailroom
2073 *
2074 * Set reserved_tailroom so that headlen can be as large as possible but
2075 * not larger than mtu and tailroom cannot be smaller than
2076 * needed_tailroom.
2077 * The required headroom should already have been reserved before using
2078 * this function.
2079 */
2080static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2081 unsigned int needed_tailroom)
2082{
2083 SKB_LINEAR_ASSERT(skb);
2084 if (mtu < skb_tailroom(skb) - needed_tailroom)
2085 /* use at most mtu */
2086 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2087 else
2088 /* use up to all available space */
2089 skb->reserved_tailroom = needed_tailroom;
2090}
2091
8bce6d7d
TH
2092#define ENCAP_TYPE_ETHER 0
2093#define ENCAP_TYPE_IPPROTO 1
2094
2095static inline void skb_set_inner_protocol(struct sk_buff *skb,
2096 __be16 protocol)
2097{
2098 skb->inner_protocol = protocol;
2099 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2100}
2101
2102static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2103 __u8 ipproto)
2104{
2105 skb->inner_ipproto = ipproto;
2106 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2107}
2108
6a674e9c
JG
2109static inline void skb_reset_inner_headers(struct sk_buff *skb)
2110{
aefbd2b3 2111 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2112 skb->inner_network_header = skb->network_header;
2113 skb->inner_transport_header = skb->transport_header;
2114}
2115
0b5c9db1
JP
2116static inline void skb_reset_mac_len(struct sk_buff *skb)
2117{
2118 skb->mac_len = skb->network_header - skb->mac_header;
2119}
2120
6a674e9c
JG
2121static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2122 *skb)
2123{
2124 return skb->head + skb->inner_transport_header;
2125}
2126
55dc5a9f
TH
2127static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2128{
2129 return skb_inner_transport_header(skb) - skb->data;
2130}
2131
6a674e9c
JG
2132static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2133{
2134 skb->inner_transport_header = skb->data - skb->head;
2135}
2136
2137static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2138 const int offset)
2139{
2140 skb_reset_inner_transport_header(skb);
2141 skb->inner_transport_header += offset;
2142}
2143
2144static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2145{
2146 return skb->head + skb->inner_network_header;
2147}
2148
2149static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2150{
2151 skb->inner_network_header = skb->data - skb->head;
2152}
2153
2154static inline void skb_set_inner_network_header(struct sk_buff *skb,
2155 const int offset)
2156{
2157 skb_reset_inner_network_header(skb);
2158 skb->inner_network_header += offset;
2159}
2160
aefbd2b3
PS
2161static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2162{
2163 return skb->head + skb->inner_mac_header;
2164}
2165
2166static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2167{
2168 skb->inner_mac_header = skb->data - skb->head;
2169}
2170
2171static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2172 const int offset)
2173{
2174 skb_reset_inner_mac_header(skb);
2175 skb->inner_mac_header += offset;
2176}
fda55eca
ED
2177static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2178{
35d04610 2179 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2180}
2181
9c70220b
ACM
2182static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2183{
2e07fa9c 2184 return skb->head + skb->transport_header;
9c70220b
ACM
2185}
2186
badff6d0
ACM
2187static inline void skb_reset_transport_header(struct sk_buff *skb)
2188{
2e07fa9c 2189 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2190}
2191
967b05f6
ACM
2192static inline void skb_set_transport_header(struct sk_buff *skb,
2193 const int offset)
2194{
2e07fa9c
ACM
2195 skb_reset_transport_header(skb);
2196 skb->transport_header += offset;
ea2ae17d
ACM
2197}
2198
d56f90a7
ACM
2199static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2200{
2e07fa9c 2201 return skb->head + skb->network_header;
d56f90a7
ACM
2202}
2203
c1d2bbe1
ACM
2204static inline void skb_reset_network_header(struct sk_buff *skb)
2205{
2e07fa9c 2206 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2207}
2208
c14d2450
ACM
2209static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2210{
2e07fa9c
ACM
2211 skb_reset_network_header(skb);
2212 skb->network_header += offset;
c14d2450
ACM
2213}
2214
2e07fa9c 2215static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2216{
2e07fa9c 2217 return skb->head + skb->mac_header;
bbe735e4
ACM
2218}
2219
ea6da4fd
AV
2220static inline int skb_mac_offset(const struct sk_buff *skb)
2221{
2222 return skb_mac_header(skb) - skb->data;
2223}
2224
0daf4349
DB
2225static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2226{
2227 return skb->network_header - skb->mac_header;
2228}
2229
2e07fa9c 2230static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2231{
35d04610 2232 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2233}
2234
2235static inline void skb_reset_mac_header(struct sk_buff *skb)
2236{
2237 skb->mac_header = skb->data - skb->head;
2238}
2239
2240static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2241{
2242 skb_reset_mac_header(skb);
2243 skb->mac_header += offset;
2244}
2245
0e3da5bb
TT
2246static inline void skb_pop_mac_header(struct sk_buff *skb)
2247{
2248 skb->mac_header = skb->network_header;
2249}
2250
fbbdb8f0
YX
2251static inline void skb_probe_transport_header(struct sk_buff *skb,
2252 const int offset_hint)
2253{
2254 struct flow_keys keys;
2255
2256 if (skb_transport_header_was_set(skb))
2257 return;
cd79a238 2258 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2259 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2260 else
2261 skb_set_transport_header(skb, offset_hint);
2262}
2263
03606895
ED
2264static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2265{
2266 if (skb_mac_header_was_set(skb)) {
2267 const unsigned char *old_mac = skb_mac_header(skb);
2268
2269 skb_set_mac_header(skb, -skb->mac_len);
2270 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2271 }
2272}
2273
04fb451e
MM
2274static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2275{
2276 return skb->csum_start - skb_headroom(skb);
2277}
2278
08b64fcc
AD
2279static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2280{
2281 return skb->head + skb->csum_start;
2282}
2283
2e07fa9c
ACM
2284static inline int skb_transport_offset(const struct sk_buff *skb)
2285{
2286 return skb_transport_header(skb) - skb->data;
2287}
2288
2289static inline u32 skb_network_header_len(const struct sk_buff *skb)
2290{
2291 return skb->transport_header - skb->network_header;
2292}
2293
6a674e9c
JG
2294static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2295{
2296 return skb->inner_transport_header - skb->inner_network_header;
2297}
2298
2e07fa9c
ACM
2299static inline int skb_network_offset(const struct sk_buff *skb)
2300{
2301 return skb_network_header(skb) - skb->data;
2302}
48d49d0c 2303
6a674e9c
JG
2304static inline int skb_inner_network_offset(const struct sk_buff *skb)
2305{
2306 return skb_inner_network_header(skb) - skb->data;
2307}
2308
f9599ce1
CG
2309static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2310{
2311 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2312}
2313
1da177e4
LT
2314/*
2315 * CPUs often take a performance hit when accessing unaligned memory
2316 * locations. The actual performance hit varies, it can be small if the
2317 * hardware handles it or large if we have to take an exception and fix it
2318 * in software.
2319 *
2320 * Since an ethernet header is 14 bytes network drivers often end up with
2321 * the IP header at an unaligned offset. The IP header can be aligned by
2322 * shifting the start of the packet by 2 bytes. Drivers should do this
2323 * with:
2324 *
8660c124 2325 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2326 *
2327 * The downside to this alignment of the IP header is that the DMA is now
2328 * unaligned. On some architectures the cost of an unaligned DMA is high
2329 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2330 *
1da177e4
LT
2331 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2332 * to be overridden.
2333 */
2334#ifndef NET_IP_ALIGN
2335#define NET_IP_ALIGN 2
2336#endif
2337
025be81e
AB
2338/*
2339 * The networking layer reserves some headroom in skb data (via
2340 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2341 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2342 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2343 *
2344 * Unfortunately this headroom changes the DMA alignment of the resulting
2345 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2346 * on some architectures. An architecture can override this value,
2347 * perhaps setting it to a cacheline in size (since that will maintain
2348 * cacheline alignment of the DMA). It must be a power of 2.
2349 *
d6301d3d 2350 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2351 * headroom, you should not reduce this.
5933dd2f
ED
2352 *
2353 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2354 * to reduce average number of cache lines per packet.
2355 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2356 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2357 */
2358#ifndef NET_SKB_PAD
5933dd2f 2359#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2360#endif
2361
7965bd4d 2362int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2363
5293efe6 2364static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2365{
c4264f27 2366 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2367 WARN_ON(1);
2368 return;
2369 }
27a884dc
ACM
2370 skb->len = len;
2371 skb_set_tail_pointer(skb, len);
1da177e4
LT
2372}
2373
5293efe6
DB
2374static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2375{
2376 __skb_set_length(skb, len);
2377}
2378
7965bd4d 2379void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2380
2381static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2382{
3cc0e873
HX
2383 if (skb->data_len)
2384 return ___pskb_trim(skb, len);
2385 __skb_trim(skb, len);
2386 return 0;
1da177e4
LT
2387}
2388
2389static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2390{
2391 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2392}
2393
e9fa4f7b
HX
2394/**
2395 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2396 * @skb: buffer to alter
2397 * @len: new length
2398 *
2399 * This is identical to pskb_trim except that the caller knows that
2400 * the skb is not cloned so we should never get an error due to out-
2401 * of-memory.
2402 */
2403static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2404{
2405 int err = pskb_trim(skb, len);
2406 BUG_ON(err);
2407}
2408
5293efe6
DB
2409static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2410{
2411 unsigned int diff = len - skb->len;
2412
2413 if (skb_tailroom(skb) < diff) {
2414 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2415 GFP_ATOMIC);
2416 if (ret)
2417 return ret;
2418 }
2419 __skb_set_length(skb, len);
2420 return 0;
2421}
2422
1da177e4
LT
2423/**
2424 * skb_orphan - orphan a buffer
2425 * @skb: buffer to orphan
2426 *
2427 * If a buffer currently has an owner then we call the owner's
2428 * destructor function and make the @skb unowned. The buffer continues
2429 * to exist but is no longer charged to its former owner.
2430 */
2431static inline void skb_orphan(struct sk_buff *skb)
2432{
c34a7612 2433 if (skb->destructor) {
1da177e4 2434 skb->destructor(skb);
c34a7612
ED
2435 skb->destructor = NULL;
2436 skb->sk = NULL;
376c7311
ED
2437 } else {
2438 BUG_ON(skb->sk);
c34a7612 2439 }
1da177e4
LT
2440}
2441
a353e0ce
MT
2442/**
2443 * skb_orphan_frags - orphan the frags contained in a buffer
2444 * @skb: buffer to orphan frags from
2445 * @gfp_mask: allocation mask for replacement pages
2446 *
2447 * For each frag in the SKB which needs a destructor (i.e. has an
2448 * owner) create a copy of that frag and release the original
2449 * page by calling the destructor.
2450 */
2451static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2452{
2453 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2454 return 0;
2455 return skb_copy_ubufs(skb, gfp_mask);
2456}
2457
1da177e4
LT
2458/**
2459 * __skb_queue_purge - empty a list
2460 * @list: list to empty
2461 *
2462 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2463 * the list and one reference dropped. This function does not take the
2464 * list lock and the caller must hold the relevant locks to use it.
2465 */
7965bd4d 2466void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2467static inline void __skb_queue_purge(struct sk_buff_head *list)
2468{
2469 struct sk_buff *skb;
2470 while ((skb = __skb_dequeue(list)) != NULL)
2471 kfree_skb(skb);
2472}
2473
9f5afeae
YW
2474void skb_rbtree_purge(struct rb_root *root);
2475
7965bd4d 2476void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2477
7965bd4d
JP
2478struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2479 gfp_t gfp_mask);
8af27456
CH
2480
2481/**
2482 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2483 * @dev: network device to receive on
2484 * @length: length to allocate
2485 *
2486 * Allocate a new &sk_buff and assign it a usage count of one. The
2487 * buffer has unspecified headroom built in. Users should allocate
2488 * the headroom they think they need without accounting for the
2489 * built in space. The built in space is used for optimisations.
2490 *
2491 * %NULL is returned if there is no free memory. Although this function
2492 * allocates memory it can be called from an interrupt.
2493 */
2494static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2495 unsigned int length)
8af27456
CH
2496{
2497 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2498}
2499
6f532612
ED
2500/* legacy helper around __netdev_alloc_skb() */
2501static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2502 gfp_t gfp_mask)
2503{
2504 return __netdev_alloc_skb(NULL, length, gfp_mask);
2505}
2506
2507/* legacy helper around netdev_alloc_skb() */
2508static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2509{
2510 return netdev_alloc_skb(NULL, length);
2511}
2512
2513
4915a0de
ED
2514static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2515 unsigned int length, gfp_t gfp)
61321bbd 2516{
4915a0de 2517 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2518
2519 if (NET_IP_ALIGN && skb)
2520 skb_reserve(skb, NET_IP_ALIGN);
2521 return skb;
2522}
2523
4915a0de
ED
2524static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2525 unsigned int length)
2526{
2527 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2528}
2529
181edb2b
AD
2530static inline void skb_free_frag(void *addr)
2531{
8c2dd3e4 2532 page_frag_free(addr);
181edb2b
AD
2533}
2534
ffde7328 2535void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2536struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2537 unsigned int length, gfp_t gfp_mask);
2538static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2539 unsigned int length)
2540{
2541 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2542}
795bb1c0
JDB
2543void napi_consume_skb(struct sk_buff *skb, int budget);
2544
2545void __kfree_skb_flush(void);
15fad714 2546void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2547
71dfda58
AD
2548/**
2549 * __dev_alloc_pages - allocate page for network Rx
2550 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2551 * @order: size of the allocation
2552 *
2553 * Allocate a new page.
2554 *
2555 * %NULL is returned if there is no free memory.
2556*/
2557static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2558 unsigned int order)
2559{
2560 /* This piece of code contains several assumptions.
2561 * 1. This is for device Rx, therefor a cold page is preferred.
2562 * 2. The expectation is the user wants a compound page.
2563 * 3. If requesting a order 0 page it will not be compound
2564 * due to the check to see if order has a value in prep_new_page
2565 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2566 * code in gfp_to_alloc_flags that should be enforcing this.
2567 */
2568 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2569
2570 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2571}
2572
2573static inline struct page *dev_alloc_pages(unsigned int order)
2574{
95829b3a 2575 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2576}
2577
2578/**
2579 * __dev_alloc_page - allocate a page for network Rx
2580 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2581 *
2582 * Allocate a new page.
2583 *
2584 * %NULL is returned if there is no free memory.
2585 */
2586static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2587{
2588 return __dev_alloc_pages(gfp_mask, 0);
2589}
2590
2591static inline struct page *dev_alloc_page(void)
2592{
95829b3a 2593 return dev_alloc_pages(0);
71dfda58
AD
2594}
2595
0614002b
MG
2596/**
2597 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2598 * @page: The page that was allocated from skb_alloc_page
2599 * @skb: The skb that may need pfmemalloc set
2600 */
2601static inline void skb_propagate_pfmemalloc(struct page *page,
2602 struct sk_buff *skb)
2603{
2f064f34 2604 if (page_is_pfmemalloc(page))
0614002b
MG
2605 skb->pfmemalloc = true;
2606}
2607
131ea667 2608/**
e227867f 2609 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2610 * @frag: the paged fragment
2611 *
2612 * Returns the &struct page associated with @frag.
2613 */
2614static inline struct page *skb_frag_page(const skb_frag_t *frag)
2615{
a8605c60 2616 return frag->page.p;
131ea667
IC
2617}
2618
2619/**
2620 * __skb_frag_ref - take an addition reference on a paged fragment.
2621 * @frag: the paged fragment
2622 *
2623 * Takes an additional reference on the paged fragment @frag.
2624 */
2625static inline void __skb_frag_ref(skb_frag_t *frag)
2626{
2627 get_page(skb_frag_page(frag));
2628}
2629
2630/**
2631 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2632 * @skb: the buffer
2633 * @f: the fragment offset.
2634 *
2635 * Takes an additional reference on the @f'th paged fragment of @skb.
2636 */
2637static inline void skb_frag_ref(struct sk_buff *skb, int f)
2638{
2639 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2640}
2641
2642/**
2643 * __skb_frag_unref - release a reference on a paged fragment.
2644 * @frag: the paged fragment
2645 *
2646 * Releases a reference on the paged fragment @frag.
2647 */
2648static inline void __skb_frag_unref(skb_frag_t *frag)
2649{
2650 put_page(skb_frag_page(frag));
2651}
2652
2653/**
2654 * skb_frag_unref - release a reference on a paged fragment of an skb.
2655 * @skb: the buffer
2656 * @f: the fragment offset
2657 *
2658 * Releases a reference on the @f'th paged fragment of @skb.
2659 */
2660static inline void skb_frag_unref(struct sk_buff *skb, int f)
2661{
2662 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2663}
2664
2665/**
2666 * skb_frag_address - gets the address of the data contained in a paged fragment
2667 * @frag: the paged fragment buffer
2668 *
2669 * Returns the address of the data within @frag. The page must already
2670 * be mapped.
2671 */
2672static inline void *skb_frag_address(const skb_frag_t *frag)
2673{
2674 return page_address(skb_frag_page(frag)) + frag->page_offset;
2675}
2676
2677/**
2678 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2679 * @frag: the paged fragment buffer
2680 *
2681 * Returns the address of the data within @frag. Checks that the page
2682 * is mapped and returns %NULL otherwise.
2683 */
2684static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2685{
2686 void *ptr = page_address(skb_frag_page(frag));
2687 if (unlikely(!ptr))
2688 return NULL;
2689
2690 return ptr + frag->page_offset;
2691}
2692
2693/**
2694 * __skb_frag_set_page - sets the page contained in a paged fragment
2695 * @frag: the paged fragment
2696 * @page: the page to set
2697 *
2698 * Sets the fragment @frag to contain @page.
2699 */
2700static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2701{
a8605c60 2702 frag->page.p = page;
131ea667
IC
2703}
2704
2705/**
2706 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2707 * @skb: the buffer
2708 * @f: the fragment offset
2709 * @page: the page to set
2710 *
2711 * Sets the @f'th fragment of @skb to contain @page.
2712 */
2713static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2714 struct page *page)
2715{
2716 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2717}
2718
400dfd3a
ED
2719bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2720
131ea667
IC
2721/**
2722 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2723 * @dev: the device to map the fragment to
131ea667
IC
2724 * @frag: the paged fragment to map
2725 * @offset: the offset within the fragment (starting at the
2726 * fragment's own offset)
2727 * @size: the number of bytes to map
771b00a8 2728 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
2729 *
2730 * Maps the page associated with @frag to @device.
2731 */
2732static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2733 const skb_frag_t *frag,
2734 size_t offset, size_t size,
2735 enum dma_data_direction dir)
2736{
2737 return dma_map_page(dev, skb_frag_page(frag),
2738 frag->page_offset + offset, size, dir);
2739}
2740
117632e6
ED
2741static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2742 gfp_t gfp_mask)
2743{
2744 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2745}
2746
bad93e9d
OP
2747
2748static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2749 gfp_t gfp_mask)
2750{
2751 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2752}
2753
2754
334a8132
PM
2755/**
2756 * skb_clone_writable - is the header of a clone writable
2757 * @skb: buffer to check
2758 * @len: length up to which to write
2759 *
2760 * Returns true if modifying the header part of the cloned buffer
2761 * does not requires the data to be copied.
2762 */
05bdd2f1 2763static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2764{
2765 return !skb_header_cloned(skb) &&
2766 skb_headroom(skb) + len <= skb->hdr_len;
2767}
2768
3697649f
DB
2769static inline int skb_try_make_writable(struct sk_buff *skb,
2770 unsigned int write_len)
2771{
2772 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2773 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2774}
2775
d9cc2048
HX
2776static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2777 int cloned)
2778{
2779 int delta = 0;
2780
d9cc2048
HX
2781 if (headroom > skb_headroom(skb))
2782 delta = headroom - skb_headroom(skb);
2783
2784 if (delta || cloned)
2785 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2786 GFP_ATOMIC);
2787 return 0;
2788}
2789
1da177e4
LT
2790/**
2791 * skb_cow - copy header of skb when it is required
2792 * @skb: buffer to cow
2793 * @headroom: needed headroom
2794 *
2795 * If the skb passed lacks sufficient headroom or its data part
2796 * is shared, data is reallocated. If reallocation fails, an error
2797 * is returned and original skb is not changed.
2798 *
2799 * The result is skb with writable area skb->head...skb->tail
2800 * and at least @headroom of space at head.
2801 */
2802static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2803{
d9cc2048
HX
2804 return __skb_cow(skb, headroom, skb_cloned(skb));
2805}
1da177e4 2806
d9cc2048
HX
2807/**
2808 * skb_cow_head - skb_cow but only making the head writable
2809 * @skb: buffer to cow
2810 * @headroom: needed headroom
2811 *
2812 * This function is identical to skb_cow except that we replace the
2813 * skb_cloned check by skb_header_cloned. It should be used when
2814 * you only need to push on some header and do not need to modify
2815 * the data.
2816 */
2817static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2818{
2819 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2820}
2821
2822/**
2823 * skb_padto - pad an skbuff up to a minimal size
2824 * @skb: buffer to pad
2825 * @len: minimal length
2826 *
2827 * Pads up a buffer to ensure the trailing bytes exist and are
2828 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2829 * is untouched. Otherwise it is extended. Returns zero on
2830 * success. The skb is freed on error.
1da177e4 2831 */
5b057c6b 2832static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2833{
2834 unsigned int size = skb->len;
2835 if (likely(size >= len))
5b057c6b 2836 return 0;
987c402a 2837 return skb_pad(skb, len - size);
1da177e4
LT
2838}
2839
9c0c1124
AD
2840/**
2841 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2842 * @skb: buffer to pad
2843 * @len: minimal length
cd0a137a 2844 * @free_on_error: free buffer on error
9c0c1124
AD
2845 *
2846 * Pads up a buffer to ensure the trailing bytes exist and are
2847 * blanked. If the buffer already contains sufficient data it
2848 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 2849 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 2850 */
cd0a137a
FF
2851static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2852 bool free_on_error)
9c0c1124
AD
2853{
2854 unsigned int size = skb->len;
2855
2856 if (unlikely(size < len)) {
2857 len -= size;
cd0a137a 2858 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
2859 return -ENOMEM;
2860 __skb_put(skb, len);
2861 }
2862 return 0;
2863}
2864
cd0a137a
FF
2865/**
2866 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2867 * @skb: buffer to pad
2868 * @len: minimal length
2869 *
2870 * Pads up a buffer to ensure the trailing bytes exist and are
2871 * blanked. If the buffer already contains sufficient data it
2872 * is untouched. Otherwise it is extended. Returns zero on
2873 * success. The skb is freed on error.
2874 */
2875static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2876{
2877 return __skb_put_padto(skb, len, true);
2878}
2879
1da177e4 2880static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2881 struct iov_iter *from, int copy)
1da177e4
LT
2882{
2883 const int off = skb->len;
2884
2885 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 2886 __wsum csum = 0;
15e6cb46
AV
2887 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2888 &csum, from)) {
1da177e4
LT
2889 skb->csum = csum_block_add(skb->csum, csum, off);
2890 return 0;
2891 }
15e6cb46 2892 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
2893 return 0;
2894
2895 __skb_trim(skb, off);
2896 return -EFAULT;
2897}
2898
38ba0a65
ED
2899static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2900 const struct page *page, int off)
1da177e4
LT
2901{
2902 if (i) {
9e903e08 2903 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2904
ea2ab693 2905 return page == skb_frag_page(frag) &&
9e903e08 2906 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2907 }
38ba0a65 2908 return false;
1da177e4
LT
2909}
2910
364c6bad
HX
2911static inline int __skb_linearize(struct sk_buff *skb)
2912{
2913 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2914}
2915
1da177e4
LT
2916/**
2917 * skb_linearize - convert paged skb to linear one
2918 * @skb: buffer to linarize
1da177e4
LT
2919 *
2920 * If there is no free memory -ENOMEM is returned, otherwise zero
2921 * is returned and the old skb data released.
2922 */
364c6bad
HX
2923static inline int skb_linearize(struct sk_buff *skb)
2924{
2925 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2926}
2927
cef401de
ED
2928/**
2929 * skb_has_shared_frag - can any frag be overwritten
2930 * @skb: buffer to test
2931 *
2932 * Return true if the skb has at least one frag that might be modified
2933 * by an external entity (as in vmsplice()/sendfile())
2934 */
2935static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2936{
c9af6db4
PS
2937 return skb_is_nonlinear(skb) &&
2938 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2939}
2940
364c6bad
HX
2941/**
2942 * skb_linearize_cow - make sure skb is linear and writable
2943 * @skb: buffer to process
2944 *
2945 * If there is no free memory -ENOMEM is returned, otherwise zero
2946 * is returned and the old skb data released.
2947 */
2948static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2949{
364c6bad
HX
2950 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2951 __skb_linearize(skb) : 0;
1da177e4
LT
2952}
2953
479ffccc
DB
2954static __always_inline void
2955__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2956 unsigned int off)
2957{
2958 if (skb->ip_summed == CHECKSUM_COMPLETE)
2959 skb->csum = csum_block_sub(skb->csum,
2960 csum_partial(start, len, 0), off);
2961 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2962 skb_checksum_start_offset(skb) < 0)
2963 skb->ip_summed = CHECKSUM_NONE;
2964}
2965
1da177e4
LT
2966/**
2967 * skb_postpull_rcsum - update checksum for received skb after pull
2968 * @skb: buffer to update
2969 * @start: start of data before pull
2970 * @len: length of data pulled
2971 *
2972 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2973 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2974 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 2975 */
1da177e4 2976static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2977 const void *start, unsigned int len)
1da177e4 2978{
479ffccc 2979 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
2980}
2981
479ffccc
DB
2982static __always_inline void
2983__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2984 unsigned int off)
2985{
2986 if (skb->ip_summed == CHECKSUM_COMPLETE)
2987 skb->csum = csum_block_add(skb->csum,
2988 csum_partial(start, len, 0), off);
2989}
cbb042f9 2990
479ffccc
DB
2991/**
2992 * skb_postpush_rcsum - update checksum for received skb after push
2993 * @skb: buffer to update
2994 * @start: start of data after push
2995 * @len: length of data pushed
2996 *
2997 * After doing a push on a received packet, you need to call this to
2998 * update the CHECKSUM_COMPLETE checksum.
2999 */
f8ffad69
DB
3000static inline void skb_postpush_rcsum(struct sk_buff *skb,
3001 const void *start, unsigned int len)
3002{
479ffccc 3003 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3004}
3005
af72868b 3006void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3007
82a31b92
WC
3008/**
3009 * skb_push_rcsum - push skb and update receive checksum
3010 * @skb: buffer to update
3011 * @len: length of data pulled
3012 *
3013 * This function performs an skb_push on the packet and updates
3014 * the CHECKSUM_COMPLETE checksum. It should be used on
3015 * receive path processing instead of skb_push unless you know
3016 * that the checksum difference is zero (e.g., a valid IP header)
3017 * or you are setting ip_summed to CHECKSUM_NONE.
3018 */
d58ff351 3019static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3020{
3021 skb_push(skb, len);
3022 skb_postpush_rcsum(skb, skb->data, len);
3023 return skb->data;
3024}
3025
7ce5a27f
DM
3026/**
3027 * pskb_trim_rcsum - trim received skb and update checksum
3028 * @skb: buffer to trim
3029 * @len: new length
3030 *
3031 * This is exactly the same as pskb_trim except that it ensures the
3032 * checksum of received packets are still valid after the operation.
3033 */
3034
3035static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3036{
3037 if (likely(len >= skb->len))
3038 return 0;
3039 if (skb->ip_summed == CHECKSUM_COMPLETE)
3040 skb->ip_summed = CHECKSUM_NONE;
3041 return __pskb_trim(skb, len);
3042}
3043
5293efe6
DB
3044static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3045{
3046 if (skb->ip_summed == CHECKSUM_COMPLETE)
3047 skb->ip_summed = CHECKSUM_NONE;
3048 __skb_trim(skb, len);
3049 return 0;
3050}
3051
3052static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3053{
3054 if (skb->ip_summed == CHECKSUM_COMPLETE)
3055 skb->ip_summed = CHECKSUM_NONE;
3056 return __skb_grow(skb, len);
3057}
3058
1da177e4
LT
3059#define skb_queue_walk(queue, skb) \
3060 for (skb = (queue)->next; \
a1e4891f 3061 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3062 skb = skb->next)
3063
46f8914e
JC
3064#define skb_queue_walk_safe(queue, skb, tmp) \
3065 for (skb = (queue)->next, tmp = skb->next; \
3066 skb != (struct sk_buff *)(queue); \
3067 skb = tmp, tmp = skb->next)
3068
1164f52a 3069#define skb_queue_walk_from(queue, skb) \
a1e4891f 3070 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3071 skb = skb->next)
3072
3073#define skb_queue_walk_from_safe(queue, skb, tmp) \
3074 for (tmp = skb->next; \
3075 skb != (struct sk_buff *)(queue); \
3076 skb = tmp, tmp = skb->next)
3077
300ce174
SH
3078#define skb_queue_reverse_walk(queue, skb) \
3079 for (skb = (queue)->prev; \
a1e4891f 3080 skb != (struct sk_buff *)(queue); \
300ce174
SH
3081 skb = skb->prev)
3082
686a2955
DM
3083#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3084 for (skb = (queue)->prev, tmp = skb->prev; \
3085 skb != (struct sk_buff *)(queue); \
3086 skb = tmp, tmp = skb->prev)
3087
3088#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3089 for (tmp = skb->prev; \
3090 skb != (struct sk_buff *)(queue); \
3091 skb = tmp, tmp = skb->prev)
1da177e4 3092
21dc3301 3093static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3094{
3095 return skb_shinfo(skb)->frag_list != NULL;
3096}
3097
3098static inline void skb_frag_list_init(struct sk_buff *skb)
3099{
3100 skb_shinfo(skb)->frag_list = NULL;
3101}
3102
ee039871
DM
3103#define skb_walk_frags(skb, iter) \
3104 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3105
ea3793ee
RW
3106
3107int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3108 const struct sk_buff *skb);
65101aec
PA
3109struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3110 struct sk_buff_head *queue,
3111 unsigned int flags,
3112 void (*destructor)(struct sock *sk,
3113 struct sk_buff *skb),
3114 int *peeked, int *off, int *err,
3115 struct sk_buff **last);
ea3793ee 3116struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3117 void (*destructor)(struct sock *sk,
3118 struct sk_buff *skb),
ea3793ee
RW
3119 int *peeked, int *off, int *err,
3120 struct sk_buff **last);
7965bd4d 3121struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3122 void (*destructor)(struct sock *sk,
3123 struct sk_buff *skb),
7965bd4d
JP
3124 int *peeked, int *off, int *err);
3125struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3126 int *err);
3127unsigned int datagram_poll(struct file *file, struct socket *sock,
3128 struct poll_table_struct *wait);
c0371da6
AV
3129int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3130 struct iov_iter *to, int size);
51f3d02b
DM
3131static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3132 struct msghdr *msg, int size)
3133{
e5a4b0bb 3134 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3135}
e5a4b0bb
AV
3136int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3137 struct msghdr *msg);
3a654f97
AV
3138int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3139 struct iov_iter *from, int len);
3a654f97 3140int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3141void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3142void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3143static inline void skb_free_datagram_locked(struct sock *sk,
3144 struct sk_buff *skb)
3145{
3146 __skb_free_datagram_locked(sk, skb, 0);
3147}
7965bd4d 3148int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3149int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3150int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3151__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3152 int len, __wsum csum);
a60e3cc7 3153int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3154 struct pipe_inode_info *pipe, unsigned int len,
25869262 3155 unsigned int flags);
7965bd4d 3156void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3157unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3158int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3159 int len, int hlen);
7965bd4d
JP
3160void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3161int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3162void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3163unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3164bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 3165struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3166struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3167int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3168int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3169int skb_vlan_pop(struct sk_buff *skb);
3170int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3171struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3172 gfp_t gfp);
20380731 3173
6ce8e9ce
AV
3174static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3175{
3073f070 3176 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3177}
3178
7eab8d9e
AV
3179static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3180{
e5a4b0bb 3181 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3182}
3183
2817a336
DB
3184struct skb_checksum_ops {
3185 __wsum (*update)(const void *mem, int len, __wsum wsum);
3186 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3187};
3188
9617813d
DC
3189extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3190
2817a336
DB
3191__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3192 __wsum csum, const struct skb_checksum_ops *ops);
3193__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3194 __wsum csum);
3195
1e98a0f0
ED
3196static inline void * __must_check
3197__skb_header_pointer(const struct sk_buff *skb, int offset,
3198 int len, void *data, int hlen, void *buffer)
1da177e4 3199{
55820ee2 3200 if (hlen - offset >= len)
690e36e7 3201 return data + offset;
1da177e4 3202
690e36e7
DM
3203 if (!skb ||
3204 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3205 return NULL;
3206
3207 return buffer;
3208}
3209
1e98a0f0
ED
3210static inline void * __must_check
3211skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3212{
3213 return __skb_header_pointer(skb, offset, len, skb->data,
3214 skb_headlen(skb), buffer);
3215}
3216
4262e5cc
DB
3217/**
3218 * skb_needs_linearize - check if we need to linearize a given skb
3219 * depending on the given device features.
3220 * @skb: socket buffer to check
3221 * @features: net device features
3222 *
3223 * Returns true if either:
3224 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3225 * 2. skb is fragmented and the device does not support SG.
3226 */
3227static inline bool skb_needs_linearize(struct sk_buff *skb,
3228 netdev_features_t features)
3229{
3230 return skb_is_nonlinear(skb) &&
3231 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3232 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3233}
3234
d626f62b
ACM
3235static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3236 void *to,
3237 const unsigned int len)
3238{
3239 memcpy(to, skb->data, len);
3240}
3241
3242static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3243 const int offset, void *to,
3244 const unsigned int len)
3245{
3246 memcpy(to, skb->data + offset, len);
3247}
3248
27d7ff46
ACM
3249static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3250 const void *from,
3251 const unsigned int len)
3252{
3253 memcpy(skb->data, from, len);
3254}
3255
3256static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3257 const int offset,
3258 const void *from,
3259 const unsigned int len)
3260{
3261 memcpy(skb->data + offset, from, len);
3262}
3263
7965bd4d 3264void skb_init(void);
1da177e4 3265
ac45f602
PO
3266static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3267{
3268 return skb->tstamp;
3269}
3270
a61bbcf2
PM
3271/**
3272 * skb_get_timestamp - get timestamp from a skb
3273 * @skb: skb to get stamp from
3274 * @stamp: pointer to struct timeval to store stamp in
3275 *
3276 * Timestamps are stored in the skb as offsets to a base timestamp.
3277 * This function converts the offset back to a struct timeval and stores
3278 * it in stamp.
3279 */
ac45f602
PO
3280static inline void skb_get_timestamp(const struct sk_buff *skb,
3281 struct timeval *stamp)
a61bbcf2 3282{
b7aa0bf7 3283 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3284}
3285
ac45f602
PO
3286static inline void skb_get_timestampns(const struct sk_buff *skb,
3287 struct timespec *stamp)
3288{
3289 *stamp = ktime_to_timespec(skb->tstamp);
3290}
3291
b7aa0bf7 3292static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3293{
b7aa0bf7 3294 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3295}
3296
164891aa
SH
3297static inline ktime_t net_timedelta(ktime_t t)
3298{
3299 return ktime_sub(ktime_get_real(), t);
3300}
3301
b9ce204f
IJ
3302static inline ktime_t net_invalid_timestamp(void)
3303{
8b0e1953 3304 return 0;
b9ce204f 3305}
a61bbcf2 3306
62bccb8c
AD
3307struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3308
c1f19b51
RC
3309#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3310
7965bd4d
JP
3311void skb_clone_tx_timestamp(struct sk_buff *skb);
3312bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3313
3314#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3315
3316static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3317{
3318}
3319
3320static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3321{
3322 return false;
3323}
3324
3325#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3326
3327/**
3328 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3329 *
da92b194
RC
3330 * PHY drivers may accept clones of transmitted packets for
3331 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3332 * must call this function to return the skb back to the stack with a
3333 * timestamp.
da92b194 3334 *
c1f19b51 3335 * @skb: clone of the the original outgoing packet
7a76a021 3336 * @hwtstamps: hardware time stamps
c1f19b51
RC
3337 *
3338 */
3339void skb_complete_tx_timestamp(struct sk_buff *skb,
3340 struct skb_shared_hwtstamps *hwtstamps);
3341
e7fd2885
WB
3342void __skb_tstamp_tx(struct sk_buff *orig_skb,
3343 struct skb_shared_hwtstamps *hwtstamps,
3344 struct sock *sk, int tstype);
3345
ac45f602
PO
3346/**
3347 * skb_tstamp_tx - queue clone of skb with send time stamps
3348 * @orig_skb: the original outgoing packet
3349 * @hwtstamps: hardware time stamps, may be NULL if not available
3350 *
3351 * If the skb has a socket associated, then this function clones the
3352 * skb (thus sharing the actual data and optional structures), stores
3353 * the optional hardware time stamping information (if non NULL) or
3354 * generates a software time stamp (otherwise), then queues the clone
3355 * to the error queue of the socket. Errors are silently ignored.
3356 */
7965bd4d
JP
3357void skb_tstamp_tx(struct sk_buff *orig_skb,
3358 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3359
4507a715
RC
3360/**
3361 * skb_tx_timestamp() - Driver hook for transmit timestamping
3362 *
3363 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3364 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3365 *
73409f3b
DM
3366 * Specifically, one should make absolutely sure that this function is
3367 * called before TX completion of this packet can trigger. Otherwise
3368 * the packet could potentially already be freed.
3369 *
4507a715
RC
3370 * @skb: A socket buffer.
3371 */
3372static inline void skb_tx_timestamp(struct sk_buff *skb)
3373{
c1f19b51 3374 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3375 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3376 skb_tstamp_tx(skb, NULL);
4507a715
RC
3377}
3378
6e3e939f
JB
3379/**
3380 * skb_complete_wifi_ack - deliver skb with wifi status
3381 *
3382 * @skb: the original outgoing packet
3383 * @acked: ack status
3384 *
3385 */
3386void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3387
7965bd4d
JP
3388__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3389__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3390
60476372
HX
3391static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3392{
6edec0e6
TH
3393 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3394 skb->csum_valid ||
3395 (skb->ip_summed == CHECKSUM_PARTIAL &&
3396 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3397}
3398
fb286bb2
HX
3399/**
3400 * skb_checksum_complete - Calculate checksum of an entire packet
3401 * @skb: packet to process
3402 *
3403 * This function calculates the checksum over the entire packet plus
3404 * the value of skb->csum. The latter can be used to supply the
3405 * checksum of a pseudo header as used by TCP/UDP. It returns the
3406 * checksum.
3407 *
3408 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3409 * this function can be used to verify that checksum on received
3410 * packets. In that case the function should return zero if the
3411 * checksum is correct. In particular, this function will return zero
3412 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3413 * hardware has already verified the correctness of the checksum.
3414 */
4381ca3c 3415static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3416{
60476372
HX
3417 return skb_csum_unnecessary(skb) ?
3418 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3419}
3420
77cffe23
TH
3421static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3422{
3423 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3424 if (skb->csum_level == 0)
3425 skb->ip_summed = CHECKSUM_NONE;
3426 else
3427 skb->csum_level--;
3428 }
3429}
3430
3431static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3432{
3433 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3434 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3435 skb->csum_level++;
3436 } else if (skb->ip_summed == CHECKSUM_NONE) {
3437 skb->ip_summed = CHECKSUM_UNNECESSARY;
3438 skb->csum_level = 0;
3439 }
3440}
3441
76ba0aae
TH
3442/* Check if we need to perform checksum complete validation.
3443 *
3444 * Returns true if checksum complete is needed, false otherwise
3445 * (either checksum is unnecessary or zero checksum is allowed).
3446 */
3447static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3448 bool zero_okay,
3449 __sum16 check)
3450{
5d0c2b95
TH
3451 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3452 skb->csum_valid = 1;
77cffe23 3453 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3454 return false;
3455 }
3456
3457 return true;
3458}
3459
3460/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3461 * in checksum_init.
3462 */
3463#define CHECKSUM_BREAK 76
3464
4e18b9ad
TH
3465/* Unset checksum-complete
3466 *
3467 * Unset checksum complete can be done when packet is being modified
3468 * (uncompressed for instance) and checksum-complete value is
3469 * invalidated.
3470 */
3471static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3472{
3473 if (skb->ip_summed == CHECKSUM_COMPLETE)
3474 skb->ip_summed = CHECKSUM_NONE;
3475}
3476
76ba0aae
TH
3477/* Validate (init) checksum based on checksum complete.
3478 *
3479 * Return values:
3480 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3481 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3482 * checksum is stored in skb->csum for use in __skb_checksum_complete
3483 * non-zero: value of invalid checksum
3484 *
3485 */
3486static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3487 bool complete,
3488 __wsum psum)
3489{
3490 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3491 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3492 skb->csum_valid = 1;
76ba0aae
TH
3493 return 0;
3494 }
3495 }
3496
3497 skb->csum = psum;
3498
5d0c2b95
TH
3499 if (complete || skb->len <= CHECKSUM_BREAK) {
3500 __sum16 csum;
3501
3502 csum = __skb_checksum_complete(skb);
3503 skb->csum_valid = !csum;
3504 return csum;
3505 }
76ba0aae
TH
3506
3507 return 0;
3508}
3509
3510static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3511{
3512 return 0;
3513}
3514
3515/* Perform checksum validate (init). Note that this is a macro since we only
3516 * want to calculate the pseudo header which is an input function if necessary.
3517 * First we try to validate without any computation (checksum unnecessary) and
3518 * then calculate based on checksum complete calling the function to compute
3519 * pseudo header.
3520 *
3521 * Return values:
3522 * 0: checksum is validated or try to in skb_checksum_complete
3523 * non-zero: value of invalid checksum
3524 */
3525#define __skb_checksum_validate(skb, proto, complete, \
3526 zero_okay, check, compute_pseudo) \
3527({ \
3528 __sum16 __ret = 0; \
5d0c2b95 3529 skb->csum_valid = 0; \
76ba0aae
TH
3530 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3531 __ret = __skb_checksum_validate_complete(skb, \
3532 complete, compute_pseudo(skb, proto)); \
3533 __ret; \
3534})
3535
3536#define skb_checksum_init(skb, proto, compute_pseudo) \
3537 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3538
3539#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3540 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3541
3542#define skb_checksum_validate(skb, proto, compute_pseudo) \
3543 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3544
3545#define skb_checksum_validate_zero_check(skb, proto, check, \
3546 compute_pseudo) \
096a4cfa 3547 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3548
3549#define skb_checksum_simple_validate(skb) \
3550 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3551
d96535a1
TH
3552static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3553{
219f1d79 3554 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
3555}
3556
3557static inline void __skb_checksum_convert(struct sk_buff *skb,
3558 __sum16 check, __wsum pseudo)
3559{
3560 skb->csum = ~pseudo;
3561 skb->ip_summed = CHECKSUM_COMPLETE;
3562}
3563
3564#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3565do { \
3566 if (__skb_checksum_convert_check(skb)) \
3567 __skb_checksum_convert(skb, check, \
3568 compute_pseudo(skb, proto)); \
3569} while (0)
3570
15e2396d
TH
3571static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3572 u16 start, u16 offset)
3573{
3574 skb->ip_summed = CHECKSUM_PARTIAL;
3575 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3576 skb->csum_offset = offset - start;
3577}
3578
dcdc8994
TH
3579/* Update skbuf and packet to reflect the remote checksum offload operation.
3580 * When called, ptr indicates the starting point for skb->csum when
3581 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3582 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3583 */
3584static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3585 int start, int offset, bool nopartial)
dcdc8994
TH
3586{
3587 __wsum delta;
3588
15e2396d
TH
3589 if (!nopartial) {
3590 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3591 return;
3592 }
3593
dcdc8994
TH
3594 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3595 __skb_checksum_complete(skb);
3596 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3597 }
3598
3599 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3600
3601 /* Adjust skb->csum since we changed the packet */
3602 skb->csum = csum_add(skb->csum, delta);
3603}
3604
cb9c6836
FW
3605static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3606{
3607#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 3608 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
3609#else
3610 return NULL;
3611#endif
3612}
3613
5f79e0f9 3614#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3615void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3616static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3617{
3618 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3619 nf_conntrack_destroy(nfct);
1da177e4
LT
3620}
3621static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3622{
3623 if (nfct)
3624 atomic_inc(&nfct->use);
3625}
2fc72c7b 3626#endif
34666d46 3627#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3628static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3629{
53869ceb 3630 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
1da177e4
LT
3631 kfree(nf_bridge);
3632}
3633static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3634{
3635 if (nf_bridge)
53869ceb 3636 refcount_inc(&nf_bridge->use);
1da177e4
LT
3637}
3638#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3639static inline void nf_reset(struct sk_buff *skb)
3640{
5f79e0f9 3641#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3642 nf_conntrack_put(skb_nfct(skb));
3643 skb->_nfct = 0;
2fc72c7b 3644#endif
34666d46 3645#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3646 nf_bridge_put(skb->nf_bridge);
3647 skb->nf_bridge = NULL;
3648#endif
3649}
3650
124dff01
PM
3651static inline void nf_reset_trace(struct sk_buff *skb)
3652{
478b360a 3653#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3654 skb->nf_trace = 0;
3655#endif
a193a4ab
PM
3656}
3657
edda553c 3658/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3659static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3660 bool copy)
edda553c 3661{
5f79e0f9 3662#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3663 dst->_nfct = src->_nfct;
3664 nf_conntrack_get(skb_nfct(src));
2fc72c7b 3665#endif
34666d46 3666#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3667 dst->nf_bridge = src->nf_bridge;
3668 nf_bridge_get(src->nf_bridge);
3669#endif
478b360a 3670#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3671 if (copy)
3672 dst->nf_trace = src->nf_trace;
478b360a 3673#endif
edda553c
YK
3674}
3675
e7ac05f3
YK
3676static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3677{
e7ac05f3 3678#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 3679 nf_conntrack_put(skb_nfct(dst));
2fc72c7b 3680#endif
34666d46 3681#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3682 nf_bridge_put(dst->nf_bridge);
3683#endif
b1937227 3684 __nf_copy(dst, src, true);
e7ac05f3
YK
3685}
3686
984bc16c
JM
3687#ifdef CONFIG_NETWORK_SECMARK
3688static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3689{
3690 to->secmark = from->secmark;
3691}
3692
3693static inline void skb_init_secmark(struct sk_buff *skb)
3694{
3695 skb->secmark = 0;
3696}
3697#else
3698static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3699{ }
3700
3701static inline void skb_init_secmark(struct sk_buff *skb)
3702{ }
3703#endif
3704
574f7194
EB
3705static inline bool skb_irq_freeable(const struct sk_buff *skb)
3706{
3707 return !skb->destructor &&
3708#if IS_ENABLED(CONFIG_XFRM)
3709 !skb->sp &&
3710#endif
cb9c6836 3711 !skb_nfct(skb) &&
574f7194
EB
3712 !skb->_skb_refdst &&
3713 !skb_has_frag_list(skb);
3714}
3715
f25f4e44
PWJ
3716static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3717{
f25f4e44 3718 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3719}
3720
9247744e 3721static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3722{
4e3ab47a 3723 return skb->queue_mapping;
4e3ab47a
PE
3724}
3725
f25f4e44
PWJ
3726static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3727{
f25f4e44 3728 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3729}
3730
d5a9e24a
DM
3731static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3732{
3733 skb->queue_mapping = rx_queue + 1;
3734}
3735
9247744e 3736static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3737{
3738 return skb->queue_mapping - 1;
3739}
3740
9247744e 3741static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3742{
a02cec21 3743 return skb->queue_mapping != 0;
d5a9e24a
DM
3744}
3745
4ff06203
JA
3746static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3747{
3748 skb->dst_pending_confirm = val;
3749}
3750
3751static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3752{
3753 return skb->dst_pending_confirm != 0;
3754}
3755
def8b4fa
AD
3756static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3757{
0b3d8e08 3758#ifdef CONFIG_XFRM
def8b4fa 3759 return skb->sp;
def8b4fa 3760#else
def8b4fa 3761 return NULL;
def8b4fa 3762#endif
0b3d8e08 3763}
def8b4fa 3764
68c33163
PS
3765/* Keeps track of mac header offset relative to skb->head.
3766 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3767 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3768 * tunnel skb it points to outer mac header.
3769 * Keeps track of level of encapsulation of network headers.
3770 */
68c33163 3771struct skb_gso_cb {
802ab55a
AD
3772 union {
3773 int mac_offset;
3774 int data_offset;
3775 };
3347c960 3776 int encap_level;
76443456 3777 __wsum csum;
7e2b10c1 3778 __u16 csum_start;
68c33163 3779};
9207f9d4
KK
3780#define SKB_SGO_CB_OFFSET 32
3781#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3782
3783static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3784{
3785 return (skb_mac_header(inner_skb) - inner_skb->head) -
3786 SKB_GSO_CB(inner_skb)->mac_offset;
3787}
3788
1e2bd517
PS
3789static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3790{
3791 int new_headroom, headroom;
3792 int ret;
3793
3794 headroom = skb_headroom(skb);
3795 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3796 if (ret)
3797 return ret;
3798
3799 new_headroom = skb_headroom(skb);
3800 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3801 return 0;
3802}
3803
08b64fcc
AD
3804static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3805{
3806 /* Do not update partial checksums if remote checksum is enabled. */
3807 if (skb->remcsum_offload)
3808 return;
3809
3810 SKB_GSO_CB(skb)->csum = res;
3811 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3812}
3813
7e2b10c1
TH
3814/* Compute the checksum for a gso segment. First compute the checksum value
3815 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3816 * then add in skb->csum (checksum from csum_start to end of packet).
3817 * skb->csum and csum_start are then updated to reflect the checksum of the
3818 * resultant packet starting from the transport header-- the resultant checksum
3819 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3820 * header.
3821 */
3822static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3823{
76443456
AD
3824 unsigned char *csum_start = skb_transport_header(skb);
3825 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3826 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3827
76443456
AD
3828 SKB_GSO_CB(skb)->csum = res;
3829 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3830
76443456 3831 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3832}
3833
bdcc0924 3834static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3835{
3836 return skb_shinfo(skb)->gso_size;
3837}
3838
36a8f39e 3839/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3840static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3841{
3842 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3843}
3844
5293efe6
DB
3845static inline void skb_gso_reset(struct sk_buff *skb)
3846{
3847 skb_shinfo(skb)->gso_size = 0;
3848 skb_shinfo(skb)->gso_segs = 0;
3849 skb_shinfo(skb)->gso_type = 0;
3850}
3851
7965bd4d 3852void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3853
3854static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3855{
3856 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3857 * wanted then gso_type will be set. */
05bdd2f1
ED
3858 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3859
b78462eb
AD
3860 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3861 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3862 __skb_warn_lro_forwarding(skb);
3863 return true;
3864 }
3865 return false;
3866}
3867
35fc92a9
HX
3868static inline void skb_forward_csum(struct sk_buff *skb)
3869{
3870 /* Unfortunately we don't support this one. Any brave souls? */
3871 if (skb->ip_summed == CHECKSUM_COMPLETE)
3872 skb->ip_summed = CHECKSUM_NONE;
3873}
3874
bc8acf2c
ED
3875/**
3876 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3877 * @skb: skb to check
3878 *
3879 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3880 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3881 * use this helper, to document places where we make this assertion.
3882 */
05bdd2f1 3883static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3884{
3885#ifdef DEBUG
3886 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3887#endif
3888}
3889
f35d9d8a 3890bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3891
ed1f50c3 3892int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3893struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3894 unsigned int transport_len,
3895 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3896
3a7c1ee4
AD
3897/**
3898 * skb_head_is_locked - Determine if the skb->head is locked down
3899 * @skb: skb to check
3900 *
3901 * The head on skbs build around a head frag can be removed if they are
3902 * not cloned. This function returns true if the skb head is locked down
3903 * due to either being allocated via kmalloc, or by being a clone with
3904 * multiple references to the head.
3905 */
3906static inline bool skb_head_is_locked(const struct sk_buff *skb)
3907{
3908 return !skb->head_frag || skb_cloned(skb);
3909}
fe6cc55f
FW
3910
3911/**
3912 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3913 *
3914 * @skb: GSO skb
3915 *
3916 * skb_gso_network_seglen is used to determine the real size of the
3917 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3918 *
3919 * The MAC/L2 header is not accounted for.
3920 */
3921static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3922{
3923 unsigned int hdr_len = skb_transport_header(skb) -
3924 skb_network_header(skb);
3925 return hdr_len + skb_gso_transport_seglen(skb);
3926}
ee122c79 3927
179bc67f
EC
3928/* Local Checksum Offload.
3929 * Compute outer checksum based on the assumption that the
3930 * inner checksum will be offloaded later.
e8ae7b00
EC
3931 * See Documentation/networking/checksum-offloads.txt for
3932 * explanation of how this works.
179bc67f
EC
3933 * Fill in outer checksum adjustment (e.g. with sum of outer
3934 * pseudo-header) before calling.
3935 * Also ensure that inner checksum is in linear data area.
3936 */
3937static inline __wsum lco_csum(struct sk_buff *skb)
3938{
9e74a6da
AD
3939 unsigned char *csum_start = skb_checksum_start(skb);
3940 unsigned char *l4_hdr = skb_transport_header(skb);
3941 __wsum partial;
179bc67f
EC
3942
3943 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3944 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3945 skb->csum_offset));
3946
179bc67f 3947 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3948 * adjustment filled in by caller) and return result.
179bc67f 3949 */
9e74a6da 3950 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3951}
3952
1da177e4
LT
3953#endif /* __KERNEL__ */
3954#endif /* _LINUX_SKBUFF_H */