]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/skbuff.h
netfilter: nft_ct: add zone id set support
[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>
1bd758eb 37#include <net/flow_dissector.h>
a60e3cc7 38#include <linux/splice.h>
72b31f72 39#include <linux/in6.h>
8b10cab6 40#include <linux/if_packet.h>
f70ea018 41#include <net/flow.h>
1da177e4 42
7a6ae71b
TH
43/* The interface for checksum offload between the stack and networking drivers
44 * is as follows...
45 *
46 * A. IP checksum related features
47 *
48 * Drivers advertise checksum offload capabilities in the features of a device.
49 * From the stack's point of view these are capabilities offered by the driver,
50 * a driver typically only advertises features that it is capable of offloading
51 * to its device.
52 *
53 * The checksum related features are:
54 *
55 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
56 * IP (one's complement) checksum for any combination
57 * of protocols or protocol layering. The checksum is
58 * computed and set in a packet per the CHECKSUM_PARTIAL
59 * interface (see below).
60 *
61 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
62 * TCP or UDP packets over IPv4. These are specifically
63 * unencapsulated packets of the form IPv4|TCP or
64 * IPv4|UDP where the Protocol field in the IPv4 header
65 * is TCP or UDP. The IPv4 header may contain IP options
66 * This feature cannot be set in features for a device
67 * with NETIF_F_HW_CSUM also set. This feature is being
68 * DEPRECATED (see below).
69 *
70 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
71 * TCP or UDP packets over IPv6. These are specifically
72 * unencapsulated packets of the form IPv6|TCP or
73 * IPv4|UDP where the Next Header field in the IPv6
74 * header is either TCP or UDP. IPv6 extension headers
75 * are not supported with this feature. This feature
76 * cannot be set in features for a device with
77 * NETIF_F_HW_CSUM also set. This feature is being
78 * DEPRECATED (see below).
79 *
80 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
81 * This flag is used only used to disable the RX checksum
82 * feature for a device. The stack will accept receive
83 * checksum indication in packets received on a device
84 * regardless of whether NETIF_F_RXCSUM is set.
85 *
86 * B. Checksumming of received packets by device. Indication of checksum
87 * verification is in set skb->ip_summed. Possible values are:
78ea85f1
DB
88 *
89 * CHECKSUM_NONE:
90 *
7a6ae71b 91 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
92 * The packet contains full (though not verified) checksum in packet but
93 * not in skb->csum. Thus, skb->csum is undefined in this case.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * The hardware you're dealing with doesn't calculate the full checksum
98 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
99 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
100 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
101 * though. A driver or device must never modify the checksum field in the
102 * packet even if checksum is verified.
77cffe23
TH
103 *
104 * CHECKSUM_UNNECESSARY is applicable to following protocols:
105 * TCP: IPv6 and IPv4.
106 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
107 * zero UDP checksum for either IPv4 or IPv6, the networking stack
108 * may perform further validation in this case.
109 * GRE: only if the checksum is present in the header.
110 * SCTP: indicates the CRC in SCTP header has been validated.
111 *
112 * skb->csum_level indicates the number of consecutive checksums found in
113 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
114 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
115 * and a device is able to verify the checksums for UDP (possibly zero),
116 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
117 * two. If the device were only able to verify the UDP checksum and not
118 * GRE, either because it doesn't support GRE checksum of because GRE
119 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
120 * not considered in this case).
78ea85f1
DB
121 *
122 * CHECKSUM_COMPLETE:
123 *
124 * This is the most generic way. The device supplied checksum of the _whole_
125 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
126 * hardware doesn't need to parse L3/L4 headers to implement this.
127 *
128 * Note: Even if device supports only some protocols, but is able to produce
129 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
130 *
131 * CHECKSUM_PARTIAL:
132 *
6edec0e6
TH
133 * A checksum is set up to be offloaded to a device as described in the
134 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 135 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
136 * on the same host, or it may be set in the input path in GRO or remote
137 * checksum offload. For the purposes of checksum verification, the checksum
138 * referred to by skb->csum_start + skb->csum_offset and any preceding
139 * checksums in the packet are considered verified. Any checksums in the
140 * packet that are after the checksum being offloaded are not considered to
141 * be verified.
78ea85f1 142 *
7a6ae71b
TH
143 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
144 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
145 *
146 * CHECKSUM_PARTIAL:
147 *
7a6ae71b 148 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 149 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
150 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
151 * csum_start and csum_offset values are valid values given the length and
152 * offset of the packet, however they should not attempt to validate that the
153 * checksum refers to a legitimate transport layer checksum-- it is the
154 * purview of the stack to validate that csum_start and csum_offset are set
155 * correctly.
156 *
157 * When the stack requests checksum offload for a packet, the driver MUST
158 * ensure that the checksum is set correctly. A driver can either offload the
159 * checksum calculation to the device, or call skb_checksum_help (in the case
160 * that the device does not support offload for a particular checksum).
161 *
162 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
163 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
164 * checksum offload capability. If a device has limited checksum capabilities
165 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
166 * described above) a helper function can be called to resolve
167 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
168 * function takes a spec argument that describes the protocol layer that is
169 * supported for checksum offload and can be called for each packet. If a
170 * packet does not match the specification for offload, skb_checksum_help
171 * is called to resolve the checksum.
78ea85f1 172 *
7a6ae71b 173 * CHECKSUM_NONE:
78ea85f1 174 *
7a6ae71b
TH
175 * The skb was already checksummed by the protocol, or a checksum is not
176 * required.
78ea85f1
DB
177 *
178 * CHECKSUM_UNNECESSARY:
179 *
7a6ae71b
TH
180 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
181 * output.
78ea85f1 182 *
7a6ae71b
TH
183 * CHECKSUM_COMPLETE:
184 * Not used in checksum output. If a driver observes a packet with this value
185 * set in skbuff, if should treat as CHECKSUM_NONE being set.
186 *
187 * D. Non-IP checksum (CRC) offloads
188 *
189 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
190 * offloading the SCTP CRC in a packet. To perform this offload the stack
191 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
192 * accordingly. Note the there is no indication in the skbuff that the
193 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
194 * both IP checksum offload and SCTP CRC offload must verify which offload
195 * is configured for a packet presumably by inspecting packet headers.
196 *
197 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
198 * offloading the FCOE CRC in a packet. To perform this offload the stack
199 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
200 * accordingly. Note the there is no indication in the skbuff that the
201 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
202 * both IP checksum offload and FCOE CRC offload must verify which offload
203 * is configured for a packet presumably by inspecting packet headers.
204 *
205 * E. Checksumming on output with GSO.
206 *
207 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
208 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
209 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
210 * part of the GSO operation is implied. If a checksum is being offloaded
211 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
212 * are set to refer to the outermost checksum being offload (two offloaded
213 * checksums are possible with UDP encapsulation).
78ea85f1
DB
214 */
215
60476372 216/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
217#define CHECKSUM_NONE 0
218#define CHECKSUM_UNNECESSARY 1
219#define CHECKSUM_COMPLETE 2
220#define CHECKSUM_PARTIAL 3
1da177e4 221
77cffe23
TH
222/* Maximum value in skb->csum_level */
223#define SKB_MAX_CSUM_LEVEL 3
224
0bec8c88 225#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 226#define SKB_WITH_OVERHEAD(X) \
deea84b0 227 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
228#define SKB_MAX_ORDER(X, ORDER) \
229 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
230#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
231#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
232
87fb4b7b
ED
233/* return minimum truesize of one skb containing X bytes of data */
234#define SKB_TRUESIZE(X) ((X) + \
235 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
236 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
237
1da177e4 238struct net_device;
716ea3a7 239struct scatterlist;
9c55e01c 240struct pipe_inode_info;
a8f820aa 241struct iov_iter;
fd11a83d 242struct napi_struct;
1da177e4 243
5f79e0f9 244#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
245struct nf_conntrack {
246 atomic_t use;
1da177e4 247};
5f79e0f9 248#endif
1da177e4 249
34666d46 250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 251struct nf_bridge_info {
bf1ac5ca 252 atomic_t use;
3eaf4025
FW
253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
7fb48c5b 257 } orig_proto:8;
72b1e5e4
FW
258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
411ffb4f 261 __u16 frag_max_size;
bf1ac5ca 262 struct net_device *physindev;
63cdbc06
FW
263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
7fb48c5b 266 union {
72b1e5e4 267 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
72b1e5e4
FW
270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
72b31f72 276 };
1da177e4
LT
277};
278#endif
279
1da177e4
LT
280struct sk_buff_head {
281 /* These two members must be first. */
282 struct sk_buff *next;
283 struct sk_buff *prev;
284
285 __u32 qlen;
286 spinlock_t lock;
287};
288
289struct sk_buff;
290
9d4dde52
IC
291/* To allow 64K frame to be packed as single skb without frag_list we
292 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
293 * buffers which do not start on a page boundary.
294 *
295 * Since GRO uses frags we allocate at least 16 regardless of page
296 * size.
a715dea3 297 */
9d4dde52 298#if (65536/PAGE_SIZE + 1) < 16
eec00954 299#define MAX_SKB_FRAGS 16UL
a715dea3 300#else
9d4dde52 301#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 302#endif
5f74f82e 303extern int sysctl_max_skb_frags;
1da177e4 304
3953c46c
MRL
305/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
306 * segment using its current segmentation instead.
307 */
308#define GSO_BY_FRAGS 0xFFFF
309
1da177e4
LT
310typedef struct skb_frag_struct skb_frag_t;
311
312struct skb_frag_struct {
a8605c60
IC
313 struct {
314 struct page *p;
315 } page;
cb4dfe56 316#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
317 __u32 page_offset;
318 __u32 size;
cb4dfe56
ED
319#else
320 __u16 page_offset;
321 __u16 size;
322#endif
1da177e4
LT
323};
324
9e903e08
ED
325static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326{
327 return frag->size;
328}
329
330static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
331{
332 frag->size = size;
333}
334
335static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
336{
337 frag->size += delta;
338}
339
340static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
341{
342 frag->size -= delta;
343}
344
ac45f602
PO
345#define HAVE_HW_TIME_STAMP
346
347/**
d3a21be8 348 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
349 * @hwtstamp: hardware time stamp transformed into duration
350 * since arbitrary point in time
ac45f602
PO
351 *
352 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 353 * skb->tstamp.
ac45f602
PO
354 *
355 * hwtstamps can only be compared against other hwtstamps from
356 * the same device.
357 *
358 * This structure is attached to packets as part of the
359 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
360 */
361struct skb_shared_hwtstamps {
362 ktime_t hwtstamp;
ac45f602
PO
363};
364
2244d07b
OH
365/* Definitions for tx_flags in struct skb_shared_info */
366enum {
367 /* generate hardware time stamp */
368 SKBTX_HW_TSTAMP = 1 << 0,
369
e7fd2885 370 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
371 SKBTX_SW_TSTAMP = 1 << 1,
372
373 /* device driver is going to provide hardware time stamp */
374 SKBTX_IN_PROGRESS = 1 << 2,
375
a6686f2f 376 /* device driver supports TX zero-copy buffers */
62b1a8ab 377 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
378
379 /* generate wifi status information (where possible) */
62b1a8ab 380 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
381
382 /* This indicates at least one fragment might be overwritten
383 * (as in vmsplice(), sendfile() ...)
384 * If we need to compute a TX checksum, we'll need to copy
385 * all frags to avoid possible bad checksum
386 */
387 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
388
389 /* generate software time stamp when entering packet scheduling */
390 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
391};
392
e1c8a607 393#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 394 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
395#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
396
a6686f2f
SM
397/*
398 * The callback notifies userspace to release buffers when skb DMA is done in
399 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
400 * The zerocopy_success argument is true if zero copy transmit occurred,
401 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
402 * The ctx field is used to track device context.
403 * The desc field is used to track userspace buffer index.
a6686f2f
SM
404 */
405struct ubuf_info {
e19d6763 406 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 407 void *ctx;
a6686f2f 408 unsigned long desc;
ac45f602
PO
409};
410
1da177e4
LT
411/* This data is invariant across clones and lives at
412 * the end of the header data, ie. at skb->end.
413 */
414struct skb_shared_info {
9f42f126
IC
415 unsigned char nr_frags;
416 __u8 tx_flags;
7967168c
HX
417 unsigned short gso_size;
418 /* Warning: this field is not always filled in (UFO)! */
419 unsigned short gso_segs;
420 unsigned short gso_type;
1da177e4 421 struct sk_buff *frag_list;
ac45f602 422 struct skb_shared_hwtstamps hwtstamps;
09c2d251 423 u32 tskey;
9f42f126 424 __be32 ip6_frag_id;
ec7d2f2c
ED
425
426 /*
427 * Warning : all fields before dataref are cleared in __alloc_skb()
428 */
429 atomic_t dataref;
430
69e3c75f
JB
431 /* Intermediate layers must ensure that destructor_arg
432 * remains valid until skb destructor */
433 void * destructor_arg;
a6686f2f 434
fed66381
ED
435 /* must be last field, see pskb_expand_head() */
436 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
437};
438
439/* We divide dataref into two halves. The higher 16 bits hold references
440 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
441 * the entire skb->data. A clone of a headerless skb holds the length of
442 * the header in skb->hdr_len.
1da177e4
LT
443 *
444 * All users must obey the rule that the skb->data reference count must be
445 * greater than or equal to the payload reference count.
446 *
447 * Holding a reference to the payload part means that the user does not
448 * care about modifications to the header part of skb->data.
449 */
450#define SKB_DATAREF_SHIFT 16
451#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
452
d179cd12
DM
453
454enum {
c8753d55
VS
455 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
456 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
457 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
458};
459
7967168c
HX
460enum {
461 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 462 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
463
464 /* This indicates the skb is from an untrusted source. */
465 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
466
467 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
468 SKB_GSO_TCP_ECN = 1 << 3,
469
cbc53e08 470 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 471
cbc53e08 472 SKB_GSO_TCPV6 = 1 << 5,
68c33163 473
cbc53e08 474 SKB_GSO_FCOE = 1 << 6,
73136267 475
cbc53e08 476 SKB_GSO_GRE = 1 << 7,
0d89d203 477
cbc53e08 478 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 479
7e13318d 480 SKB_GSO_IPXIP4 = 1 << 9,
61c1db7f 481
7e13318d 482 SKB_GSO_IPXIP6 = 1 << 10,
0f4f4ffa 483
cbc53e08 484 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 485
cbc53e08
AD
486 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
487
802ab55a
AD
488 SKB_GSO_PARTIAL = 1 << 13,
489
490 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
90017acc
MRL
491
492 SKB_GSO_SCTP = 1 << 15,
7967168c
HX
493};
494
2e07fa9c
ACM
495#if BITS_PER_LONG > 32
496#define NET_SKBUFF_DATA_USES_OFFSET 1
497#endif
498
499#ifdef NET_SKBUFF_DATA_USES_OFFSET
500typedef unsigned int sk_buff_data_t;
501#else
502typedef unsigned char *sk_buff_data_t;
503#endif
504
363ec392
ED
505/**
506 * struct skb_mstamp - multi resolution time stamps
507 * @stamp_us: timestamp in us resolution
508 * @stamp_jiffies: timestamp in jiffies
509 */
510struct skb_mstamp {
511 union {
512 u64 v64;
513 struct {
514 u32 stamp_us;
515 u32 stamp_jiffies;
516 };
517 };
518};
519
520/**
521 * skb_mstamp_get - get current timestamp
522 * @cl: place to store timestamps
523 */
524static inline void skb_mstamp_get(struct skb_mstamp *cl)
525{
526 u64 val = local_clock();
527
528 do_div(val, NSEC_PER_USEC);
529 cl->stamp_us = (u32)val;
530 cl->stamp_jiffies = (u32)jiffies;
531}
532
533/**
534 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
535 * @t1: pointer to newest sample
536 * @t0: pointer to oldest sample
537 */
538static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
539 const struct skb_mstamp *t0)
540{
541 s32 delta_us = t1->stamp_us - t0->stamp_us;
542 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
543
544 /* If delta_us is negative, this might be because interval is too big,
545 * or local_clock() drift is too big : fallback using jiffies.
546 */
547 if (delta_us <= 0 ||
548 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
549
550 delta_us = jiffies_to_usecs(delta_jiffies);
551
552 return delta_us;
553}
554
625a5e10
YC
555static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
556 const struct skb_mstamp *t0)
557{
558 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
559
560 if (!diff)
561 diff = t1->stamp_us - t0->stamp_us;
562 return diff > 0;
563}
363ec392 564
1da177e4
LT
565/**
566 * struct sk_buff - socket buffer
567 * @next: Next buffer in list
568 * @prev: Previous buffer in list
363ec392 569 * @tstamp: Time we arrived/left
56b17425 570 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 571 * @sk: Socket we are owned by
1da177e4 572 * @dev: Device we arrived on/are leaving by
d84e0bd7 573 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 574 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 575 * @sp: the security path, used for xfrm
1da177e4
LT
576 * @len: Length of actual data
577 * @data_len: Data length
578 * @mac_len: Length of link layer header
334a8132 579 * @hdr_len: writable header length of cloned skb
663ead3b
HX
580 * @csum: Checksum (must include start/offset pair)
581 * @csum_start: Offset from skb->head where checksumming should start
582 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 583 * @priority: Packet queueing priority
60ff7467 584 * @ignore_df: allow local fragmentation
1da177e4 585 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 586 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
587 * @nohdr: Payload reference only, must not modify header
588 * @pkt_type: Packet class
c83c2486 589 * @fclone: skbuff clone status
c83c2486 590 * @ipvs_property: skbuff is owned by ipvs
e7246e12 591 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 592 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
593 * @tc_redirected: packet was redirected by a tc action
594 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
595 * @peeked: this packet has been seen already, so stats have been
596 * done for it, don't do them again
ba9dda3a 597 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
598 * @protocol: Packet protocol from driver
599 * @destructor: Destruct function
a9e419dc 600 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 601 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 602 * @skb_iif: ifindex of device we arrived on
1da177e4 603 * @tc_index: Traffic control index
61b905da 604 * @hash: the packet hash
d84e0bd7 605 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 606 * @xmit_more: More SKBs are pending for this queue
553a5672 607 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 608 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 609 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 610 * ports.
a3b18ddb 611 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
612 * @wifi_acked_valid: wifi_acked was set
613 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 614 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 615 * @napi_id: id of the NAPI struct this skb came from
984bc16c 616 * @secmark: security marking
d84e0bd7 617 * @mark: Generic packet mark
86a9bad3 618 * @vlan_proto: vlan encapsulation protocol
6aa895b0 619 * @vlan_tci: vlan tag control information
0d89d203 620 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
621 * @inner_transport_header: Inner transport layer header (encapsulation)
622 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 623 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
624 * @transport_header: Transport layer header
625 * @network_header: Network layer header
626 * @mac_header: Link layer header
627 * @tail: Tail pointer
628 * @end: End pointer
629 * @head: Head of buffer
630 * @data: Data head pointer
631 * @truesize: Buffer size
632 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
633 */
634
635struct sk_buff {
363ec392 636 union {
56b17425
ED
637 struct {
638 /* These two members must be first. */
639 struct sk_buff *next;
640 struct sk_buff *prev;
641
642 union {
643 ktime_t tstamp;
644 struct skb_mstamp skb_mstamp;
645 };
646 };
647 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 648 };
da3f5cf1 649 struct sock *sk;
1da177e4 650
c84d9490
ED
651 union {
652 struct net_device *dev;
653 /* Some protocols might use this space to store information,
654 * while device pointer would be NULL.
655 * UDP receive path is one user.
656 */
657 unsigned long dev_scratch;
658 };
1da177e4
LT
659 /*
660 * This is the control buffer. It is free to use for every
661 * layer. Please put your private variables there. If you
662 * want to keep them across layers you have to do a skb_clone()
663 * first. This is owned by whoever has the skb queued ATM.
664 */
da3f5cf1 665 char cb[48] __aligned(8);
1da177e4 666
7fee226a 667 unsigned long _skb_refdst;
b1937227 668 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
669#ifdef CONFIG_XFRM
670 struct sec_path *sp;
b1937227
ED
671#endif
672#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 673 unsigned long _nfct;
b1937227 674#endif
85224844 675#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 676 struct nf_bridge_info *nf_bridge;
da3f5cf1 677#endif
1da177e4 678 unsigned int len,
334a8132
PM
679 data_len;
680 __u16 mac_len,
681 hdr_len;
b1937227
ED
682
683 /* Following fields are _not_ copied in __copy_skb_header()
684 * Note that queue_mapping is here mostly to fill a hole.
685 */
fe55f6d5 686 kmemcheck_bitfield_begin(flags1);
b1937227 687 __u16 queue_mapping;
36bbef52
DB
688
689/* if you move cloned around you also must adapt those constants */
690#ifdef __BIG_ENDIAN_BITFIELD
691#define CLONED_MASK (1 << 7)
692#else
693#define CLONED_MASK 1
694#endif
695#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
696
697 __u8 __cloned_offset[0];
b1937227 698 __u8 cloned:1,
6869c4d8 699 nohdr:1,
b84f4cc9 700 fclone:2,
a59322be 701 peeked:1,
b1937227 702 head_frag:1,
36bbef52
DB
703 xmit_more:1,
704 __unused:1; /* one bit hole */
fe55f6d5 705 kmemcheck_bitfield_end(flags1);
4031ae6e 706
b1937227
ED
707 /* fields enclosed in headers_start/headers_end are copied
708 * using a single memcpy() in __copy_skb_header()
709 */
ebcf34f3 710 /* private: */
b1937227 711 __u32 headers_start[0];
ebcf34f3 712 /* public: */
4031ae6e 713
233577a2
HFS
714/* if you move pkt_type around you also must adapt those constants */
715#ifdef __BIG_ENDIAN_BITFIELD
716#define PKT_TYPE_MAX (7 << 5)
717#else
718#define PKT_TYPE_MAX 7
1da177e4 719#endif
233577a2 720#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 721
233577a2 722 __u8 __pkt_type_offset[0];
b1937227 723 __u8 pkt_type:3;
c93bdd0e 724 __u8 pfmemalloc:1;
b1937227 725 __u8 ignore_df:1;
b1937227
ED
726
727 __u8 nf_trace:1;
728 __u8 ip_summed:2;
3853b584 729 __u8 ooo_okay:1;
61b905da 730 __u8 l4_hash:1;
a3b18ddb 731 __u8 sw_hash:1;
6e3e939f
JB
732 __u8 wifi_acked_valid:1;
733 __u8 wifi_acked:1;
b1937227 734
3bdc0eba 735 __u8 no_fcs:1;
77cffe23 736 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 737 __u8 encapsulation:1;
7e2b10c1 738 __u8 encap_hdr_csum:1;
5d0c2b95 739 __u8 csum_valid:1;
7e3cead5 740 __u8 csum_complete_sw:1;
b1937227
ED
741 __u8 csum_level:2;
742 __u8 csum_bad:1;
fe55f6d5 743
b1937227
ED
744#ifdef CONFIG_IPV6_NDISC_NODETYPE
745 __u8 ndisc_nodetype:2;
746#endif
747 __u8 ipvs_property:1;
8bce6d7d 748 __u8 inner_protocol_type:1;
e585f236 749 __u8 remcsum_offload:1;
6bc506b4
IS
750#ifdef CONFIG_NET_SWITCHDEV
751 __u8 offload_fwd_mark:1;
752#endif
e7246e12
WB
753#ifdef CONFIG_NET_CLS_ACT
754 __u8 tc_skip_classify:1;
8dc07fdb 755 __u8 tc_at_ingress:1;
bc31c905
WB
756 __u8 tc_redirected:1;
757 __u8 tc_from_ingress:1;
e7246e12 758#endif
b1937227
ED
759
760#ifdef CONFIG_NET_SCHED
761 __u16 tc_index; /* traffic control index */
b1937227 762#endif
fe55f6d5 763
b1937227
ED
764 union {
765 __wsum csum;
766 struct {
767 __u16 csum_start;
768 __u16 csum_offset;
769 };
770 };
771 __u32 priority;
772 int skb_iif;
773 __u32 hash;
774 __be16 vlan_proto;
775 __u16 vlan_tci;
2bd82484
ED
776#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
777 union {
778 unsigned int napi_id;
779 unsigned int sender_cpu;
780 };
97fc2f08 781#endif
984bc16c 782#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 783 __u32 secmark;
0c4f691f 784#endif
0c4f691f 785
3b885787
NH
786 union {
787 __u32 mark;
16fad69c 788 __u32 reserved_tailroom;
3b885787 789 };
1da177e4 790
8bce6d7d
TH
791 union {
792 __be16 inner_protocol;
793 __u8 inner_ipproto;
794 };
795
1a37e412
SH
796 __u16 inner_transport_header;
797 __u16 inner_network_header;
798 __u16 inner_mac_header;
b1937227
ED
799
800 __be16 protocol;
1a37e412
SH
801 __u16 transport_header;
802 __u16 network_header;
803 __u16 mac_header;
b1937227 804
ebcf34f3 805 /* private: */
b1937227 806 __u32 headers_end[0];
ebcf34f3 807 /* public: */
b1937227 808
1da177e4 809 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 810 sk_buff_data_t tail;
4305b541 811 sk_buff_data_t end;
1da177e4 812 unsigned char *head,
4305b541 813 *data;
27a884dc
ACM
814 unsigned int truesize;
815 atomic_t users;
1da177e4
LT
816};
817
818#ifdef __KERNEL__
819/*
820 * Handling routines are only of interest to the kernel
821 */
822#include <linux/slab.h>
823
1da177e4 824
c93bdd0e
MG
825#define SKB_ALLOC_FCLONE 0x01
826#define SKB_ALLOC_RX 0x02
fd11a83d 827#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
828
829/* Returns true if the skb was allocated from PFMEMALLOC reserves */
830static inline bool skb_pfmemalloc(const struct sk_buff *skb)
831{
832 return unlikely(skb->pfmemalloc);
833}
834
7fee226a
ED
835/*
836 * skb might have a dst pointer attached, refcounted or not.
837 * _skb_refdst low order bit is set if refcount was _not_ taken
838 */
839#define SKB_DST_NOREF 1UL
840#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
841
a9e419dc 842#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
843/**
844 * skb_dst - returns skb dst_entry
845 * @skb: buffer
846 *
847 * Returns skb dst_entry, regardless of reference taken or not.
848 */
adf30907
ED
849static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
850{
7fee226a
ED
851 /* If refdst was not refcounted, check we still are in a
852 * rcu_read_lock section
853 */
854 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
855 !rcu_read_lock_held() &&
856 !rcu_read_lock_bh_held());
857 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
858}
859
7fee226a
ED
860/**
861 * skb_dst_set - sets skb dst
862 * @skb: buffer
863 * @dst: dst entry
864 *
865 * Sets skb dst, assuming a reference was taken on dst and should
866 * be released by skb_dst_drop()
867 */
adf30907
ED
868static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
869{
7fee226a
ED
870 skb->_skb_refdst = (unsigned long)dst;
871}
872
932bc4d7
JA
873/**
874 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
875 * @skb: buffer
876 * @dst: dst entry
877 *
878 * Sets skb dst, assuming a reference was not taken on dst.
879 * If dst entry is cached, we do not take reference and dst_release
880 * will be avoided by refdst_drop. If dst entry is not cached, we take
881 * reference, so that last dst_release can destroy the dst immediately.
882 */
883static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
884{
dbfc4fb7
HFS
885 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
886 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 887}
7fee226a
ED
888
889/**
25985edc 890 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
891 * @skb: buffer
892 */
893static inline bool skb_dst_is_noref(const struct sk_buff *skb)
894{
895 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
896}
897
511c3f92
ED
898static inline struct rtable *skb_rtable(const struct sk_buff *skb)
899{
adf30907 900 return (struct rtable *)skb_dst(skb);
511c3f92
ED
901}
902
8b10cab6
JHS
903/* For mangling skb->pkt_type from user space side from applications
904 * such as nft, tc, etc, we only allow a conservative subset of
905 * possible pkt_types to be set.
906*/
907static inline bool skb_pkt_type_ok(u32 ptype)
908{
909 return ptype <= PACKET_OTHERHOST;
910}
911
7965bd4d
JP
912void kfree_skb(struct sk_buff *skb);
913void kfree_skb_list(struct sk_buff *segs);
914void skb_tx_error(struct sk_buff *skb);
915void consume_skb(struct sk_buff *skb);
916void __kfree_skb(struct sk_buff *skb);
d7e8883c 917extern struct kmem_cache *skbuff_head_cache;
bad43ca8 918
7965bd4d
JP
919void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
920bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
921 bool *fragstolen, int *delta_truesize);
bad43ca8 922
7965bd4d
JP
923struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
924 int node);
2ea2f62c 925struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 926struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 927static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 928 gfp_t priority)
d179cd12 929{
564824b0 930 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
931}
932
2e4e4410
ED
933struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
934 unsigned long data_len,
935 int max_page_order,
936 int *errcode,
937 gfp_t gfp_mask);
938
d0bf4a9e
ED
939/* Layout of fast clones : [skb1][skb2][fclone_ref] */
940struct sk_buff_fclones {
941 struct sk_buff skb1;
942
943 struct sk_buff skb2;
944
945 atomic_t fclone_ref;
946};
947
948/**
949 * skb_fclone_busy - check if fclone is busy
293de7de 950 * @sk: socket
d0bf4a9e
ED
951 * @skb: buffer
952 *
bda13fed 953 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
954 * Some drivers call skb_orphan() in their ndo_start_xmit(),
955 * so we also check that this didnt happen.
d0bf4a9e 956 */
39bb5e62
ED
957static inline bool skb_fclone_busy(const struct sock *sk,
958 const struct sk_buff *skb)
d0bf4a9e
ED
959{
960 const struct sk_buff_fclones *fclones;
961
962 fclones = container_of(skb, struct sk_buff_fclones, skb1);
963
964 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 965 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 966 fclones->skb2.sk == sk;
d0bf4a9e
ED
967}
968
d179cd12 969static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 970 gfp_t priority)
d179cd12 971{
c93bdd0e 972 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
973}
974
7965bd4d 975struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
976static inline struct sk_buff *alloc_skb_head(gfp_t priority)
977{
978 return __alloc_skb_head(priority, -1);
979}
980
7965bd4d
JP
981struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
982int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
983struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
984struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
985struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
986 gfp_t gfp_mask, bool fclone);
987static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
988 gfp_t gfp_mask)
989{
990 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
991}
7965bd4d
JP
992
993int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
994struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
995 unsigned int headroom);
996struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
997 int newtailroom, gfp_t priority);
25a91d8d
FD
998int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
999 int offset, int len);
7965bd4d
JP
1000int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
1001 int len);
1002int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1003int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 1004#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1005
7965bd4d
JP
1006int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1007 int getfrag(void *from, char *to, int offset,
1008 int len, int odd, struct sk_buff *skb),
1009 void *from, int length);
e89e9cf5 1010
be12a1fe
HFS
1011int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1012 int offset, size_t size);
1013
d94d9fee 1014struct skb_seq_state {
677e90ed
TG
1015 __u32 lower_offset;
1016 __u32 upper_offset;
1017 __u32 frag_idx;
1018 __u32 stepped_offset;
1019 struct sk_buff *root_skb;
1020 struct sk_buff *cur_skb;
1021 __u8 *frag_data;
1022};
1023
7965bd4d
JP
1024void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1025 unsigned int to, struct skb_seq_state *st);
1026unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1027 struct skb_seq_state *st);
1028void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1029
7965bd4d 1030unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1031 unsigned int to, struct ts_config *config);
3fc7e8a6 1032
09323cc4
TH
1033/*
1034 * Packet hash types specify the type of hash in skb_set_hash.
1035 *
1036 * Hash types refer to the protocol layer addresses which are used to
1037 * construct a packet's hash. The hashes are used to differentiate or identify
1038 * flows of the protocol layer for the hash type. Hash types are either
1039 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1040 *
1041 * Properties of hashes:
1042 *
1043 * 1) Two packets in different flows have different hash values
1044 * 2) Two packets in the same flow should have the same hash value
1045 *
1046 * A hash at a higher layer is considered to be more specific. A driver should
1047 * set the most specific hash possible.
1048 *
1049 * A driver cannot indicate a more specific hash than the layer at which a hash
1050 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1051 *
1052 * A driver may indicate a hash level which is less specific than the
1053 * actual layer the hash was computed on. For instance, a hash computed
1054 * at L4 may be considered an L3 hash. This should only be done if the
1055 * driver can't unambiguously determine that the HW computed the hash at
1056 * the higher layer. Note that the "should" in the second property above
1057 * permits this.
1058 */
1059enum pkt_hash_types {
1060 PKT_HASH_TYPE_NONE, /* Undefined type */
1061 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1062 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1063 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1064};
1065
bcc83839 1066static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1067{
bcc83839 1068 skb->hash = 0;
a3b18ddb 1069 skb->sw_hash = 0;
bcc83839
TH
1070 skb->l4_hash = 0;
1071}
1072
1073static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1074{
1075 if (!skb->l4_hash)
1076 skb_clear_hash(skb);
1077}
1078
1079static inline void
1080__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1081{
1082 skb->l4_hash = is_l4;
1083 skb->sw_hash = is_sw;
61b905da 1084 skb->hash = hash;
09323cc4
TH
1085}
1086
bcc83839
TH
1087static inline void
1088skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1089{
1090 /* Used by drivers to set hash from HW */
1091 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1092}
1093
1094static inline void
1095__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1096{
1097 __skb_set_hash(skb, hash, true, is_l4);
1098}
1099
e5276937 1100void __skb_get_hash(struct sk_buff *skb);
b917783c 1101u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1102u32 skb_get_poff(const struct sk_buff *skb);
1103u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1104 const struct flow_keys *keys, int hlen);
1105__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1106 void *data, int hlen_proto);
1107
1108static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1109 int thoff, u8 ip_proto)
1110{
1111 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1112}
1113
1114void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1115 const struct flow_dissector_key *key,
1116 unsigned int key_count);
1117
1118bool __skb_flow_dissect(const struct sk_buff *skb,
1119 struct flow_dissector *flow_dissector,
1120 void *target_container,
cd79a238
TH
1121 void *data, __be16 proto, int nhoff, int hlen,
1122 unsigned int flags);
e5276937
TH
1123
1124static inline bool skb_flow_dissect(const struct sk_buff *skb,
1125 struct flow_dissector *flow_dissector,
cd79a238 1126 void *target_container, unsigned int flags)
e5276937
TH
1127{
1128 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1129 NULL, 0, 0, 0, flags);
e5276937
TH
1130}
1131
1132static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1133 struct flow_keys *flow,
1134 unsigned int flags)
e5276937
TH
1135{
1136 memset(flow, 0, sizeof(*flow));
1137 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1138 NULL, 0, 0, 0, flags);
e5276937
TH
1139}
1140
1141static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1142 void *data, __be16 proto,
cd79a238
TH
1143 int nhoff, int hlen,
1144 unsigned int flags)
e5276937
TH
1145{
1146 memset(flow, 0, sizeof(*flow));
1147 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1148 data, proto, nhoff, hlen, flags);
e5276937
TH
1149}
1150
3958afa1 1151static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1152{
a3b18ddb 1153 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1154 __skb_get_hash(skb);
bfb564e7 1155
61b905da 1156 return skb->hash;
bfb564e7
KK
1157}
1158
20a17bf6 1159__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1160
20a17bf6 1161static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1162{
c6cc1ca7
TH
1163 if (!skb->l4_hash && !skb->sw_hash) {
1164 struct flow_keys keys;
de4c1f8b 1165 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1166
de4c1f8b 1167 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1168 }
f70ea018
TH
1169
1170 return skb->hash;
1171}
1172
20a17bf6 1173__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1174
20a17bf6 1175static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1176{
c6cc1ca7
TH
1177 if (!skb->l4_hash && !skb->sw_hash) {
1178 struct flow_keys keys;
de4c1f8b 1179 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1180
de4c1f8b 1181 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1182 }
f70ea018
TH
1183
1184 return skb->hash;
1185}
1186
50fb7992
TH
1187__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1188
57bdf7f4
TH
1189static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1190{
61b905da 1191 return skb->hash;
57bdf7f4
TH
1192}
1193
3df7a74e
TH
1194static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1195{
61b905da 1196 to->hash = from->hash;
a3b18ddb 1197 to->sw_hash = from->sw_hash;
61b905da 1198 to->l4_hash = from->l4_hash;
3df7a74e
TH
1199};
1200
4305b541
ACM
1201#ifdef NET_SKBUFF_DATA_USES_OFFSET
1202static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1203{
1204 return skb->head + skb->end;
1205}
ec47ea82
AD
1206
1207static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1208{
1209 return skb->end;
1210}
4305b541
ACM
1211#else
1212static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1213{
1214 return skb->end;
1215}
ec47ea82
AD
1216
1217static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1218{
1219 return skb->end - skb->head;
1220}
4305b541
ACM
1221#endif
1222
1da177e4 1223/* Internal */
4305b541 1224#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1225
ac45f602
PO
1226static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1227{
1228 return &skb_shinfo(skb)->hwtstamps;
1229}
1230
1da177e4
LT
1231/**
1232 * skb_queue_empty - check if a queue is empty
1233 * @list: queue head
1234 *
1235 * Returns true if the queue is empty, false otherwise.
1236 */
1237static inline int skb_queue_empty(const struct sk_buff_head *list)
1238{
fd44b93c 1239 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1240}
1241
fc7ebb21
DM
1242/**
1243 * skb_queue_is_last - check if skb is the last entry in the queue
1244 * @list: queue head
1245 * @skb: buffer
1246 *
1247 * Returns true if @skb is the last buffer on the list.
1248 */
1249static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1250 const struct sk_buff *skb)
1251{
fd44b93c 1252 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1253}
1254
832d11c5
IJ
1255/**
1256 * skb_queue_is_first - check if skb is the first entry in the queue
1257 * @list: queue head
1258 * @skb: buffer
1259 *
1260 * Returns true if @skb is the first buffer on the list.
1261 */
1262static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1263 const struct sk_buff *skb)
1264{
fd44b93c 1265 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1266}
1267
249c8b42
DM
1268/**
1269 * skb_queue_next - return the next packet in the queue
1270 * @list: queue head
1271 * @skb: current buffer
1272 *
1273 * Return the next packet in @list after @skb. It is only valid to
1274 * call this if skb_queue_is_last() evaluates to false.
1275 */
1276static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1277 const struct sk_buff *skb)
1278{
1279 /* This BUG_ON may seem severe, but if we just return then we
1280 * are going to dereference garbage.
1281 */
1282 BUG_ON(skb_queue_is_last(list, skb));
1283 return skb->next;
1284}
1285
832d11c5
IJ
1286/**
1287 * skb_queue_prev - return the prev packet in the queue
1288 * @list: queue head
1289 * @skb: current buffer
1290 *
1291 * Return the prev packet in @list before @skb. It is only valid to
1292 * call this if skb_queue_is_first() evaluates to false.
1293 */
1294static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1295 const struct sk_buff *skb)
1296{
1297 /* This BUG_ON may seem severe, but if we just return then we
1298 * are going to dereference garbage.
1299 */
1300 BUG_ON(skb_queue_is_first(list, skb));
1301 return skb->prev;
1302}
1303
1da177e4
LT
1304/**
1305 * skb_get - reference buffer
1306 * @skb: buffer to reference
1307 *
1308 * Makes another reference to a socket buffer and returns a pointer
1309 * to the buffer.
1310 */
1311static inline struct sk_buff *skb_get(struct sk_buff *skb)
1312{
1313 atomic_inc(&skb->users);
1314 return skb;
1315}
1316
1317/*
1318 * If users == 1, we are the only owner and are can avoid redundant
1319 * atomic change.
1320 */
1321
1da177e4
LT
1322/**
1323 * skb_cloned - is the buffer a clone
1324 * @skb: buffer to check
1325 *
1326 * Returns true if the buffer was generated with skb_clone() and is
1327 * one of multiple shared copies of the buffer. Cloned buffers are
1328 * shared data so must not be written to under normal circumstances.
1329 */
1330static inline int skb_cloned(const struct sk_buff *skb)
1331{
1332 return skb->cloned &&
1333 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1334}
1335
14bbd6a5
PS
1336static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1337{
d0164adc 1338 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1339
1340 if (skb_cloned(skb))
1341 return pskb_expand_head(skb, 0, 0, pri);
1342
1343 return 0;
1344}
1345
1da177e4
LT
1346/**
1347 * skb_header_cloned - is the header a clone
1348 * @skb: buffer to check
1349 *
1350 * Returns true if modifying the header part of the buffer requires
1351 * the data to be copied.
1352 */
1353static inline int skb_header_cloned(const struct sk_buff *skb)
1354{
1355 int dataref;
1356
1357 if (!skb->cloned)
1358 return 0;
1359
1360 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1361 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1362 return dataref != 1;
1363}
1364
9580bf2e
ED
1365static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1366{
1367 might_sleep_if(gfpflags_allow_blocking(pri));
1368
1369 if (skb_header_cloned(skb))
1370 return pskb_expand_head(skb, 0, 0, pri);
1371
1372 return 0;
1373}
1374
1da177e4
LT
1375/**
1376 * skb_header_release - release reference to header
1377 * @skb: buffer to operate on
1378 *
1379 * Drop a reference to the header part of the buffer. This is done
1380 * by acquiring a payload reference. You must not read from the header
1381 * part of skb->data after this.
f4a775d1 1382 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1383 */
1384static inline void skb_header_release(struct sk_buff *skb)
1385{
1386 BUG_ON(skb->nohdr);
1387 skb->nohdr = 1;
1388 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1389}
1390
f4a775d1
ED
1391/**
1392 * __skb_header_release - release reference to header
1393 * @skb: buffer to operate on
1394 *
1395 * Variant of skb_header_release() assuming skb is private to caller.
1396 * We can avoid one atomic operation.
1397 */
1398static inline void __skb_header_release(struct sk_buff *skb)
1399{
1400 skb->nohdr = 1;
1401 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1402}
1403
1404
1da177e4
LT
1405/**
1406 * skb_shared - is the buffer shared
1407 * @skb: buffer to check
1408 *
1409 * Returns true if more than one person has a reference to this
1410 * buffer.
1411 */
1412static inline int skb_shared(const struct sk_buff *skb)
1413{
1414 return atomic_read(&skb->users) != 1;
1415}
1416
1417/**
1418 * skb_share_check - check if buffer is shared and if so clone it
1419 * @skb: buffer to check
1420 * @pri: priority for memory allocation
1421 *
1422 * If the buffer is shared the buffer is cloned and the old copy
1423 * drops a reference. A new clone with a single reference is returned.
1424 * If the buffer is not shared the original buffer is returned. When
1425 * being called from interrupt status or with spinlocks held pri must
1426 * be GFP_ATOMIC.
1427 *
1428 * NULL is returned on a memory allocation failure.
1429 */
47061bc4 1430static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1431{
d0164adc 1432 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1433 if (skb_shared(skb)) {
1434 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1435
1436 if (likely(nskb))
1437 consume_skb(skb);
1438 else
1439 kfree_skb(skb);
1da177e4
LT
1440 skb = nskb;
1441 }
1442 return skb;
1443}
1444
1445/*
1446 * Copy shared buffers into a new sk_buff. We effectively do COW on
1447 * packets to handle cases where we have a local reader and forward
1448 * and a couple of other messy ones. The normal one is tcpdumping
1449 * a packet thats being forwarded.
1450 */
1451
1452/**
1453 * skb_unshare - make a copy of a shared buffer
1454 * @skb: buffer to check
1455 * @pri: priority for memory allocation
1456 *
1457 * If the socket buffer is a clone then this function creates a new
1458 * copy of the data, drops a reference count on the old copy and returns
1459 * the new copy with the reference count at 1. If the buffer is not a clone
1460 * the original buffer is returned. When called with a spinlock held or
1461 * from interrupt state @pri must be %GFP_ATOMIC
1462 *
1463 * %NULL is returned on a memory allocation failure.
1464 */
e2bf521d 1465static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1466 gfp_t pri)
1da177e4 1467{
d0164adc 1468 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1469 if (skb_cloned(skb)) {
1470 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1471
1472 /* Free our shared copy */
1473 if (likely(nskb))
1474 consume_skb(skb);
1475 else
1476 kfree_skb(skb);
1da177e4
LT
1477 skb = nskb;
1478 }
1479 return skb;
1480}
1481
1482/**
1a5778aa 1483 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1484 * @list_: list to peek at
1485 *
1486 * Peek an &sk_buff. Unlike most other operations you _MUST_
1487 * be careful with this one. A peek leaves the buffer on the
1488 * list and someone else may run off with it. You must hold
1489 * the appropriate locks or have a private queue to do this.
1490 *
1491 * Returns %NULL for an empty list or a pointer to the head element.
1492 * The reference count is not incremented and the reference is therefore
1493 * volatile. Use with caution.
1494 */
05bdd2f1 1495static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1496{
18d07000
ED
1497 struct sk_buff *skb = list_->next;
1498
1499 if (skb == (struct sk_buff *)list_)
1500 skb = NULL;
1501 return skb;
1da177e4
LT
1502}
1503
da5ef6e5
PE
1504/**
1505 * skb_peek_next - peek skb following the given one from a queue
1506 * @skb: skb to start from
1507 * @list_: list to peek at
1508 *
1509 * Returns %NULL when the end of the list is met or a pointer to the
1510 * next element. The reference count is not incremented and the
1511 * reference is therefore volatile. Use with caution.
1512 */
1513static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1514 const struct sk_buff_head *list_)
1515{
1516 struct sk_buff *next = skb->next;
18d07000 1517
da5ef6e5
PE
1518 if (next == (struct sk_buff *)list_)
1519 next = NULL;
1520 return next;
1521}
1522
1da177e4 1523/**
1a5778aa 1524 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1525 * @list_: list to peek at
1526 *
1527 * Peek an &sk_buff. Unlike most other operations you _MUST_
1528 * be careful with this one. A peek leaves the buffer on the
1529 * list and someone else may run off with it. You must hold
1530 * the appropriate locks or have a private queue to do this.
1531 *
1532 * Returns %NULL for an empty list or a pointer to the tail element.
1533 * The reference count is not incremented and the reference is therefore
1534 * volatile. Use with caution.
1535 */
05bdd2f1 1536static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1537{
18d07000
ED
1538 struct sk_buff *skb = list_->prev;
1539
1540 if (skb == (struct sk_buff *)list_)
1541 skb = NULL;
1542 return skb;
1543
1da177e4
LT
1544}
1545
1546/**
1547 * skb_queue_len - get queue length
1548 * @list_: list to measure
1549 *
1550 * Return the length of an &sk_buff queue.
1551 */
1552static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1553{
1554 return list_->qlen;
1555}
1556
67fed459
DM
1557/**
1558 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1559 * @list: queue to initialize
1560 *
1561 * This initializes only the list and queue length aspects of
1562 * an sk_buff_head object. This allows to initialize the list
1563 * aspects of an sk_buff_head without reinitializing things like
1564 * the spinlock. It can also be used for on-stack sk_buff_head
1565 * objects where the spinlock is known to not be used.
1566 */
1567static inline void __skb_queue_head_init(struct sk_buff_head *list)
1568{
1569 list->prev = list->next = (struct sk_buff *)list;
1570 list->qlen = 0;
1571}
1572
76f10ad0
AV
1573/*
1574 * This function creates a split out lock class for each invocation;
1575 * this is needed for now since a whole lot of users of the skb-queue
1576 * infrastructure in drivers have different locking usage (in hardirq)
1577 * than the networking core (in softirq only). In the long run either the
1578 * network layer or drivers should need annotation to consolidate the
1579 * main types of usage into 3 classes.
1580 */
1da177e4
LT
1581static inline void skb_queue_head_init(struct sk_buff_head *list)
1582{
1583 spin_lock_init(&list->lock);
67fed459 1584 __skb_queue_head_init(list);
1da177e4
LT
1585}
1586
c2ecba71
PE
1587static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1588 struct lock_class_key *class)
1589{
1590 skb_queue_head_init(list);
1591 lockdep_set_class(&list->lock, class);
1592}
1593
1da177e4 1594/*
bf299275 1595 * Insert an sk_buff on a list.
1da177e4
LT
1596 *
1597 * The "__skb_xxxx()" functions are the non-atomic ones that
1598 * can only be called with interrupts disabled.
1599 */
7965bd4d
JP
1600void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1601 struct sk_buff_head *list);
bf299275
GR
1602static inline void __skb_insert(struct sk_buff *newsk,
1603 struct sk_buff *prev, struct sk_buff *next,
1604 struct sk_buff_head *list)
1605{
1606 newsk->next = next;
1607 newsk->prev = prev;
1608 next->prev = prev->next = newsk;
1609 list->qlen++;
1610}
1da177e4 1611
67fed459
DM
1612static inline void __skb_queue_splice(const struct sk_buff_head *list,
1613 struct sk_buff *prev,
1614 struct sk_buff *next)
1615{
1616 struct sk_buff *first = list->next;
1617 struct sk_buff *last = list->prev;
1618
1619 first->prev = prev;
1620 prev->next = first;
1621
1622 last->next = next;
1623 next->prev = last;
1624}
1625
1626/**
1627 * skb_queue_splice - join two skb lists, this is designed for stacks
1628 * @list: the new list to add
1629 * @head: the place to add it in the first list
1630 */
1631static inline void skb_queue_splice(const struct sk_buff_head *list,
1632 struct sk_buff_head *head)
1633{
1634 if (!skb_queue_empty(list)) {
1635 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1636 head->qlen += list->qlen;
67fed459
DM
1637 }
1638}
1639
1640/**
d9619496 1641 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1642 * @list: the new list to add
1643 * @head: the place to add it in the first list
1644 *
1645 * The list at @list is reinitialised
1646 */
1647static inline void skb_queue_splice_init(struct sk_buff_head *list,
1648 struct sk_buff_head *head)
1649{
1650 if (!skb_queue_empty(list)) {
1651 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1652 head->qlen += list->qlen;
67fed459
DM
1653 __skb_queue_head_init(list);
1654 }
1655}
1656
1657/**
1658 * skb_queue_splice_tail - join two skb lists, each list being a queue
1659 * @list: the new list to add
1660 * @head: the place to add it in the first list
1661 */
1662static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1663 struct sk_buff_head *head)
1664{
1665 if (!skb_queue_empty(list)) {
1666 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1667 head->qlen += list->qlen;
67fed459
DM
1668 }
1669}
1670
1671/**
d9619496 1672 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1673 * @list: the new list to add
1674 * @head: the place to add it in the first list
1675 *
1676 * Each of the lists is a queue.
1677 * The list at @list is reinitialised
1678 */
1679static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1680 struct sk_buff_head *head)
1681{
1682 if (!skb_queue_empty(list)) {
1683 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1684 head->qlen += list->qlen;
67fed459
DM
1685 __skb_queue_head_init(list);
1686 }
1687}
1688
1da177e4 1689/**
300ce174 1690 * __skb_queue_after - queue a buffer at the list head
1da177e4 1691 * @list: list to use
300ce174 1692 * @prev: place after this buffer
1da177e4
LT
1693 * @newsk: buffer to queue
1694 *
300ce174 1695 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1696 * and you must therefore hold required locks before calling it.
1697 *
1698 * A buffer cannot be placed on two lists at the same time.
1699 */
300ce174
SH
1700static inline void __skb_queue_after(struct sk_buff_head *list,
1701 struct sk_buff *prev,
1702 struct sk_buff *newsk)
1da177e4 1703{
bf299275 1704 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1705}
1706
7965bd4d
JP
1707void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1708 struct sk_buff_head *list);
7de6c033 1709
f5572855
GR
1710static inline void __skb_queue_before(struct sk_buff_head *list,
1711 struct sk_buff *next,
1712 struct sk_buff *newsk)
1713{
1714 __skb_insert(newsk, next->prev, next, list);
1715}
1716
300ce174
SH
1717/**
1718 * __skb_queue_head - queue a buffer at the list head
1719 * @list: list to use
1720 * @newsk: buffer to queue
1721 *
1722 * Queue a buffer at the start of a list. This function takes no locks
1723 * and you must therefore hold required locks before calling it.
1724 *
1725 * A buffer cannot be placed on two lists at the same time.
1726 */
7965bd4d 1727void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1728static inline void __skb_queue_head(struct sk_buff_head *list,
1729 struct sk_buff *newsk)
1730{
1731 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1732}
1733
1da177e4
LT
1734/**
1735 * __skb_queue_tail - queue a buffer at the list tail
1736 * @list: list to use
1737 * @newsk: buffer to queue
1738 *
1739 * Queue a buffer at the end of a list. This function takes no locks
1740 * and you must therefore hold required locks before calling it.
1741 *
1742 * A buffer cannot be placed on two lists at the same time.
1743 */
7965bd4d 1744void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1745static inline void __skb_queue_tail(struct sk_buff_head *list,
1746 struct sk_buff *newsk)
1747{
f5572855 1748 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1749}
1750
1da177e4
LT
1751/*
1752 * remove sk_buff from list. _Must_ be called atomically, and with
1753 * the list known..
1754 */
7965bd4d 1755void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1756static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1757{
1758 struct sk_buff *next, *prev;
1759
1760 list->qlen--;
1761 next = skb->next;
1762 prev = skb->prev;
1763 skb->next = skb->prev = NULL;
1da177e4
LT
1764 next->prev = prev;
1765 prev->next = next;
1766}
1767
f525c06d
GR
1768/**
1769 * __skb_dequeue - remove from the head of the queue
1770 * @list: list to dequeue from
1771 *
1772 * Remove the head of the list. This function does not take any locks
1773 * so must be used with appropriate locks held only. The head item is
1774 * returned or %NULL if the list is empty.
1775 */
7965bd4d 1776struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1777static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1778{
1779 struct sk_buff *skb = skb_peek(list);
1780 if (skb)
1781 __skb_unlink(skb, list);
1782 return skb;
1783}
1da177e4
LT
1784
1785/**
1786 * __skb_dequeue_tail - remove from the tail of the queue
1787 * @list: list to dequeue from
1788 *
1789 * Remove the tail of the list. This function does not take any locks
1790 * so must be used with appropriate locks held only. The tail item is
1791 * returned or %NULL if the list is empty.
1792 */
7965bd4d 1793struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1794static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1795{
1796 struct sk_buff *skb = skb_peek_tail(list);
1797 if (skb)
1798 __skb_unlink(skb, list);
1799 return skb;
1800}
1801
1802
bdcc0924 1803static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1804{
1805 return skb->data_len;
1806}
1807
1808static inline unsigned int skb_headlen(const struct sk_buff *skb)
1809{
1810 return skb->len - skb->data_len;
1811}
1812
c72d8cda 1813static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1da177e4 1814{
c72d8cda 1815 unsigned int i, len = 0;
1da177e4 1816
c72d8cda 1817 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 1818 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1819 return len + skb_headlen(skb);
1820}
1821
131ea667
IC
1822/**
1823 * __skb_fill_page_desc - initialise a paged fragment in an skb
1824 * @skb: buffer containing fragment to be initialised
1825 * @i: paged fragment index to initialise
1826 * @page: the page to use for this fragment
1827 * @off: the offset to the data with @page
1828 * @size: the length of the data
1829 *
1830 * Initialises the @i'th fragment of @skb to point to &size bytes at
1831 * offset @off within @page.
1832 *
1833 * Does not take any additional reference on the fragment.
1834 */
1835static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1836 struct page *page, int off, int size)
1da177e4
LT
1837{
1838 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1839
c48a11c7 1840 /*
2f064f34
MH
1841 * Propagate page pfmemalloc to the skb if we can. The problem is
1842 * that not all callers have unique ownership of the page but rely
1843 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1844 */
a8605c60 1845 frag->page.p = page;
1da177e4 1846 frag->page_offset = off;
9e903e08 1847 skb_frag_size_set(frag, size);
cca7af38
PE
1848
1849 page = compound_head(page);
2f064f34 1850 if (page_is_pfmemalloc(page))
cca7af38 1851 skb->pfmemalloc = true;
131ea667
IC
1852}
1853
1854/**
1855 * skb_fill_page_desc - initialise a paged fragment in an skb
1856 * @skb: buffer containing fragment to be initialised
1857 * @i: paged fragment index to initialise
1858 * @page: the page to use for this fragment
1859 * @off: the offset to the data with @page
1860 * @size: the length of the data
1861 *
1862 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1863 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1864 * addition updates @skb such that @i is the last fragment.
1865 *
1866 * Does not take any additional reference on the fragment.
1867 */
1868static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1869 struct page *page, int off, int size)
1870{
1871 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1872 skb_shinfo(skb)->nr_frags = i + 1;
1873}
1874
7965bd4d
JP
1875void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1876 int size, unsigned int truesize);
654bed16 1877
f8e617e1
JW
1878void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1879 unsigned int truesize);
1880
1da177e4 1881#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1882#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1883#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1884
27a884dc
ACM
1885#ifdef NET_SKBUFF_DATA_USES_OFFSET
1886static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1887{
1888 return skb->head + skb->tail;
1889}
1890
1891static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1892{
1893 skb->tail = skb->data - skb->head;
1894}
1895
1896static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1897{
1898 skb_reset_tail_pointer(skb);
1899 skb->tail += offset;
1900}
7cc46190 1901
27a884dc
ACM
1902#else /* NET_SKBUFF_DATA_USES_OFFSET */
1903static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1904{
1905 return skb->tail;
1906}
1907
1908static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1909{
1910 skb->tail = skb->data;
1911}
1912
1913static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1914{
1915 skb->tail = skb->data + offset;
1916}
4305b541 1917
27a884dc
ACM
1918#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1919
1da177e4
LT
1920/*
1921 * Add data to an sk_buff
1922 */
0c7ddf36 1923unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1924unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1925static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1926{
27a884dc 1927 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1928 SKB_LINEAR_ASSERT(skb);
1929 skb->tail += len;
1930 skb->len += len;
1931 return tmp;
1932}
1933
7965bd4d 1934unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1935static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1936{
1937 skb->data -= len;
1938 skb->len += len;
1939 return skb->data;
1940}
1941
7965bd4d 1942unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1943static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1944{
1945 skb->len -= len;
1946 BUG_ON(skb->len < skb->data_len);
1947 return skb->data += len;
1948}
1949
47d29646
DM
1950static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1951{
1952 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1953}
1954
7965bd4d 1955unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1956
1957static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1958{
1959 if (len > skb_headlen(skb) &&
987c402a 1960 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1961 return NULL;
1962 skb->len -= len;
1963 return skb->data += len;
1964}
1965
1966static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1967{
1968 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1969}
1970
1971static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1972{
1973 if (likely(len <= skb_headlen(skb)))
1974 return 1;
1975 if (unlikely(len > skb->len))
1976 return 0;
987c402a 1977 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1978}
1979
c8c8b127
ED
1980void skb_condense(struct sk_buff *skb);
1981
1da177e4
LT
1982/**
1983 * skb_headroom - bytes at buffer head
1984 * @skb: buffer to check
1985 *
1986 * Return the number of bytes of free space at the head of an &sk_buff.
1987 */
c2636b4d 1988static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1989{
1990 return skb->data - skb->head;
1991}
1992
1993/**
1994 * skb_tailroom - bytes at buffer end
1995 * @skb: buffer to check
1996 *
1997 * Return the number of bytes of free space at the tail of an sk_buff
1998 */
1999static inline int skb_tailroom(const struct sk_buff *skb)
2000{
4305b541 2001 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2002}
2003
a21d4572
ED
2004/**
2005 * skb_availroom - bytes at buffer end
2006 * @skb: buffer to check
2007 *
2008 * Return the number of bytes of free space at the tail of an sk_buff
2009 * allocated by sk_stream_alloc()
2010 */
2011static inline int skb_availroom(const struct sk_buff *skb)
2012{
16fad69c
ED
2013 if (skb_is_nonlinear(skb))
2014 return 0;
2015
2016 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2017}
2018
1da177e4
LT
2019/**
2020 * skb_reserve - adjust headroom
2021 * @skb: buffer to alter
2022 * @len: bytes to move
2023 *
2024 * Increase the headroom of an empty &sk_buff by reducing the tail
2025 * room. This is only allowed for an empty buffer.
2026 */
8243126c 2027static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2028{
2029 skb->data += len;
2030 skb->tail += len;
2031}
2032
1837b2e2
BP
2033/**
2034 * skb_tailroom_reserve - adjust reserved_tailroom
2035 * @skb: buffer to alter
2036 * @mtu: maximum amount of headlen permitted
2037 * @needed_tailroom: minimum amount of reserved_tailroom
2038 *
2039 * Set reserved_tailroom so that headlen can be as large as possible but
2040 * not larger than mtu and tailroom cannot be smaller than
2041 * needed_tailroom.
2042 * The required headroom should already have been reserved before using
2043 * this function.
2044 */
2045static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2046 unsigned int needed_tailroom)
2047{
2048 SKB_LINEAR_ASSERT(skb);
2049 if (mtu < skb_tailroom(skb) - needed_tailroom)
2050 /* use at most mtu */
2051 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2052 else
2053 /* use up to all available space */
2054 skb->reserved_tailroom = needed_tailroom;
2055}
2056
8bce6d7d
TH
2057#define ENCAP_TYPE_ETHER 0
2058#define ENCAP_TYPE_IPPROTO 1
2059
2060static inline void skb_set_inner_protocol(struct sk_buff *skb,
2061 __be16 protocol)
2062{
2063 skb->inner_protocol = protocol;
2064 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2065}
2066
2067static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2068 __u8 ipproto)
2069{
2070 skb->inner_ipproto = ipproto;
2071 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2072}
2073
6a674e9c
JG
2074static inline void skb_reset_inner_headers(struct sk_buff *skb)
2075{
aefbd2b3 2076 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2077 skb->inner_network_header = skb->network_header;
2078 skb->inner_transport_header = skb->transport_header;
2079}
2080
0b5c9db1
JP
2081static inline void skb_reset_mac_len(struct sk_buff *skb)
2082{
2083 skb->mac_len = skb->network_header - skb->mac_header;
2084}
2085
6a674e9c
JG
2086static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2087 *skb)
2088{
2089 return skb->head + skb->inner_transport_header;
2090}
2091
55dc5a9f
TH
2092static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2093{
2094 return skb_inner_transport_header(skb) - skb->data;
2095}
2096
6a674e9c
JG
2097static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2098{
2099 skb->inner_transport_header = skb->data - skb->head;
2100}
2101
2102static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2103 const int offset)
2104{
2105 skb_reset_inner_transport_header(skb);
2106 skb->inner_transport_header += offset;
2107}
2108
2109static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2110{
2111 return skb->head + skb->inner_network_header;
2112}
2113
2114static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2115{
2116 skb->inner_network_header = skb->data - skb->head;
2117}
2118
2119static inline void skb_set_inner_network_header(struct sk_buff *skb,
2120 const int offset)
2121{
2122 skb_reset_inner_network_header(skb);
2123 skb->inner_network_header += offset;
2124}
2125
aefbd2b3
PS
2126static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2127{
2128 return skb->head + skb->inner_mac_header;
2129}
2130
2131static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2132{
2133 skb->inner_mac_header = skb->data - skb->head;
2134}
2135
2136static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2137 const int offset)
2138{
2139 skb_reset_inner_mac_header(skb);
2140 skb->inner_mac_header += offset;
2141}
fda55eca
ED
2142static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2143{
35d04610 2144 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2145}
2146
9c70220b
ACM
2147static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2148{
2e07fa9c 2149 return skb->head + skb->transport_header;
9c70220b
ACM
2150}
2151
badff6d0
ACM
2152static inline void skb_reset_transport_header(struct sk_buff *skb)
2153{
2e07fa9c 2154 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2155}
2156
967b05f6
ACM
2157static inline void skb_set_transport_header(struct sk_buff *skb,
2158 const int offset)
2159{
2e07fa9c
ACM
2160 skb_reset_transport_header(skb);
2161 skb->transport_header += offset;
ea2ae17d
ACM
2162}
2163
d56f90a7
ACM
2164static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2165{
2e07fa9c 2166 return skb->head + skb->network_header;
d56f90a7
ACM
2167}
2168
c1d2bbe1
ACM
2169static inline void skb_reset_network_header(struct sk_buff *skb)
2170{
2e07fa9c 2171 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2172}
2173
c14d2450
ACM
2174static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2175{
2e07fa9c
ACM
2176 skb_reset_network_header(skb);
2177 skb->network_header += offset;
c14d2450
ACM
2178}
2179
2e07fa9c 2180static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2181{
2e07fa9c 2182 return skb->head + skb->mac_header;
bbe735e4
ACM
2183}
2184
2e07fa9c 2185static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2186{
35d04610 2187 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2188}
2189
2190static inline void skb_reset_mac_header(struct sk_buff *skb)
2191{
2192 skb->mac_header = skb->data - skb->head;
2193}
2194
2195static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2196{
2197 skb_reset_mac_header(skb);
2198 skb->mac_header += offset;
2199}
2200
0e3da5bb
TT
2201static inline void skb_pop_mac_header(struct sk_buff *skb)
2202{
2203 skb->mac_header = skb->network_header;
2204}
2205
fbbdb8f0
YX
2206static inline void skb_probe_transport_header(struct sk_buff *skb,
2207 const int offset_hint)
2208{
2209 struct flow_keys keys;
2210
2211 if (skb_transport_header_was_set(skb))
2212 return;
cd79a238 2213 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2214 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2215 else
2216 skb_set_transport_header(skb, offset_hint);
2217}
2218
03606895
ED
2219static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2220{
2221 if (skb_mac_header_was_set(skb)) {
2222 const unsigned char *old_mac = skb_mac_header(skb);
2223
2224 skb_set_mac_header(skb, -skb->mac_len);
2225 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2226 }
2227}
2228
04fb451e
MM
2229static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2230{
2231 return skb->csum_start - skb_headroom(skb);
2232}
2233
08b64fcc
AD
2234static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2235{
2236 return skb->head + skb->csum_start;
2237}
2238
2e07fa9c
ACM
2239static inline int skb_transport_offset(const struct sk_buff *skb)
2240{
2241 return skb_transport_header(skb) - skb->data;
2242}
2243
2244static inline u32 skb_network_header_len(const struct sk_buff *skb)
2245{
2246 return skb->transport_header - skb->network_header;
2247}
2248
6a674e9c
JG
2249static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2250{
2251 return skb->inner_transport_header - skb->inner_network_header;
2252}
2253
2e07fa9c
ACM
2254static inline int skb_network_offset(const struct sk_buff *skb)
2255{
2256 return skb_network_header(skb) - skb->data;
2257}
48d49d0c 2258
6a674e9c
JG
2259static inline int skb_inner_network_offset(const struct sk_buff *skb)
2260{
2261 return skb_inner_network_header(skb) - skb->data;
2262}
2263
f9599ce1
CG
2264static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2265{
2266 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2267}
2268
1da177e4
LT
2269/*
2270 * CPUs often take a performance hit when accessing unaligned memory
2271 * locations. The actual performance hit varies, it can be small if the
2272 * hardware handles it or large if we have to take an exception and fix it
2273 * in software.
2274 *
2275 * Since an ethernet header is 14 bytes network drivers often end up with
2276 * the IP header at an unaligned offset. The IP header can be aligned by
2277 * shifting the start of the packet by 2 bytes. Drivers should do this
2278 * with:
2279 *
8660c124 2280 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2281 *
2282 * The downside to this alignment of the IP header is that the DMA is now
2283 * unaligned. On some architectures the cost of an unaligned DMA is high
2284 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2285 *
1da177e4
LT
2286 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2287 * to be overridden.
2288 */
2289#ifndef NET_IP_ALIGN
2290#define NET_IP_ALIGN 2
2291#endif
2292
025be81e
AB
2293/*
2294 * The networking layer reserves some headroom in skb data (via
2295 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2296 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2297 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2298 *
2299 * Unfortunately this headroom changes the DMA alignment of the resulting
2300 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2301 * on some architectures. An architecture can override this value,
2302 * perhaps setting it to a cacheline in size (since that will maintain
2303 * cacheline alignment of the DMA). It must be a power of 2.
2304 *
d6301d3d 2305 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2306 * headroom, you should not reduce this.
5933dd2f
ED
2307 *
2308 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2309 * to reduce average number of cache lines per packet.
2310 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2311 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2312 */
2313#ifndef NET_SKB_PAD
5933dd2f 2314#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2315#endif
2316
7965bd4d 2317int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2318
5293efe6 2319static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2320{
c4264f27 2321 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2322 WARN_ON(1);
2323 return;
2324 }
27a884dc
ACM
2325 skb->len = len;
2326 skb_set_tail_pointer(skb, len);
1da177e4
LT
2327}
2328
5293efe6
DB
2329static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2330{
2331 __skb_set_length(skb, len);
2332}
2333
7965bd4d 2334void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2335
2336static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2337{
3cc0e873
HX
2338 if (skb->data_len)
2339 return ___pskb_trim(skb, len);
2340 __skb_trim(skb, len);
2341 return 0;
1da177e4
LT
2342}
2343
2344static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2345{
2346 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2347}
2348
e9fa4f7b
HX
2349/**
2350 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2351 * @skb: buffer to alter
2352 * @len: new length
2353 *
2354 * This is identical to pskb_trim except that the caller knows that
2355 * the skb is not cloned so we should never get an error due to out-
2356 * of-memory.
2357 */
2358static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2359{
2360 int err = pskb_trim(skb, len);
2361 BUG_ON(err);
2362}
2363
5293efe6
DB
2364static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2365{
2366 unsigned int diff = len - skb->len;
2367
2368 if (skb_tailroom(skb) < diff) {
2369 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2370 GFP_ATOMIC);
2371 if (ret)
2372 return ret;
2373 }
2374 __skb_set_length(skb, len);
2375 return 0;
2376}
2377
1da177e4
LT
2378/**
2379 * skb_orphan - orphan a buffer
2380 * @skb: buffer to orphan
2381 *
2382 * If a buffer currently has an owner then we call the owner's
2383 * destructor function and make the @skb unowned. The buffer continues
2384 * to exist but is no longer charged to its former owner.
2385 */
2386static inline void skb_orphan(struct sk_buff *skb)
2387{
c34a7612 2388 if (skb->destructor) {
1da177e4 2389 skb->destructor(skb);
c34a7612
ED
2390 skb->destructor = NULL;
2391 skb->sk = NULL;
376c7311
ED
2392 } else {
2393 BUG_ON(skb->sk);
c34a7612 2394 }
1da177e4
LT
2395}
2396
a353e0ce
MT
2397/**
2398 * skb_orphan_frags - orphan the frags contained in a buffer
2399 * @skb: buffer to orphan frags from
2400 * @gfp_mask: allocation mask for replacement pages
2401 *
2402 * For each frag in the SKB which needs a destructor (i.e. has an
2403 * owner) create a copy of that frag and release the original
2404 * page by calling the destructor.
2405 */
2406static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2407{
2408 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2409 return 0;
2410 return skb_copy_ubufs(skb, gfp_mask);
2411}
2412
1da177e4
LT
2413/**
2414 * __skb_queue_purge - empty a list
2415 * @list: list to empty
2416 *
2417 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2418 * the list and one reference dropped. This function does not take the
2419 * list lock and the caller must hold the relevant locks to use it.
2420 */
7965bd4d 2421void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2422static inline void __skb_queue_purge(struct sk_buff_head *list)
2423{
2424 struct sk_buff *skb;
2425 while ((skb = __skb_dequeue(list)) != NULL)
2426 kfree_skb(skb);
2427}
2428
9f5afeae
YW
2429void skb_rbtree_purge(struct rb_root *root);
2430
7965bd4d 2431void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2432
7965bd4d
JP
2433struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2434 gfp_t gfp_mask);
8af27456
CH
2435
2436/**
2437 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2438 * @dev: network device to receive on
2439 * @length: length to allocate
2440 *
2441 * Allocate a new &sk_buff and assign it a usage count of one. The
2442 * buffer has unspecified headroom built in. Users should allocate
2443 * the headroom they think they need without accounting for the
2444 * built in space. The built in space is used for optimisations.
2445 *
2446 * %NULL is returned if there is no free memory. Although this function
2447 * allocates memory it can be called from an interrupt.
2448 */
2449static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2450 unsigned int length)
8af27456
CH
2451{
2452 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2453}
2454
6f532612
ED
2455/* legacy helper around __netdev_alloc_skb() */
2456static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2457 gfp_t gfp_mask)
2458{
2459 return __netdev_alloc_skb(NULL, length, gfp_mask);
2460}
2461
2462/* legacy helper around netdev_alloc_skb() */
2463static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2464{
2465 return netdev_alloc_skb(NULL, length);
2466}
2467
2468
4915a0de
ED
2469static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2470 unsigned int length, gfp_t gfp)
61321bbd 2471{
4915a0de 2472 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2473
2474 if (NET_IP_ALIGN && skb)
2475 skb_reserve(skb, NET_IP_ALIGN);
2476 return skb;
2477}
2478
4915a0de
ED
2479static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2480 unsigned int length)
2481{
2482 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2483}
2484
181edb2b
AD
2485static inline void skb_free_frag(void *addr)
2486{
8c2dd3e4 2487 page_frag_free(addr);
181edb2b
AD
2488}
2489
ffde7328 2490void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2491struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2492 unsigned int length, gfp_t gfp_mask);
2493static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2494 unsigned int length)
2495{
2496 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2497}
795bb1c0
JDB
2498void napi_consume_skb(struct sk_buff *skb, int budget);
2499
2500void __kfree_skb_flush(void);
15fad714 2501void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2502
71dfda58
AD
2503/**
2504 * __dev_alloc_pages - allocate page for network Rx
2505 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2506 * @order: size of the allocation
2507 *
2508 * Allocate a new page.
2509 *
2510 * %NULL is returned if there is no free memory.
2511*/
2512static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2513 unsigned int order)
2514{
2515 /* This piece of code contains several assumptions.
2516 * 1. This is for device Rx, therefor a cold page is preferred.
2517 * 2. The expectation is the user wants a compound page.
2518 * 3. If requesting a order 0 page it will not be compound
2519 * due to the check to see if order has a value in prep_new_page
2520 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2521 * code in gfp_to_alloc_flags that should be enforcing this.
2522 */
2523 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2524
2525 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2526}
2527
2528static inline struct page *dev_alloc_pages(unsigned int order)
2529{
95829b3a 2530 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2531}
2532
2533/**
2534 * __dev_alloc_page - allocate a page for network Rx
2535 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2536 *
2537 * Allocate a new page.
2538 *
2539 * %NULL is returned if there is no free memory.
2540 */
2541static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2542{
2543 return __dev_alloc_pages(gfp_mask, 0);
2544}
2545
2546static inline struct page *dev_alloc_page(void)
2547{
95829b3a 2548 return dev_alloc_pages(0);
71dfda58
AD
2549}
2550
0614002b
MG
2551/**
2552 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2553 * @page: The page that was allocated from skb_alloc_page
2554 * @skb: The skb that may need pfmemalloc set
2555 */
2556static inline void skb_propagate_pfmemalloc(struct page *page,
2557 struct sk_buff *skb)
2558{
2f064f34 2559 if (page_is_pfmemalloc(page))
0614002b
MG
2560 skb->pfmemalloc = true;
2561}
2562
131ea667 2563/**
e227867f 2564 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2565 * @frag: the paged fragment
2566 *
2567 * Returns the &struct page associated with @frag.
2568 */
2569static inline struct page *skb_frag_page(const skb_frag_t *frag)
2570{
a8605c60 2571 return frag->page.p;
131ea667
IC
2572}
2573
2574/**
2575 * __skb_frag_ref - take an addition reference on a paged fragment.
2576 * @frag: the paged fragment
2577 *
2578 * Takes an additional reference on the paged fragment @frag.
2579 */
2580static inline void __skb_frag_ref(skb_frag_t *frag)
2581{
2582 get_page(skb_frag_page(frag));
2583}
2584
2585/**
2586 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2587 * @skb: the buffer
2588 * @f: the fragment offset.
2589 *
2590 * Takes an additional reference on the @f'th paged fragment of @skb.
2591 */
2592static inline void skb_frag_ref(struct sk_buff *skb, int f)
2593{
2594 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2595}
2596
2597/**
2598 * __skb_frag_unref - release a reference on a paged fragment.
2599 * @frag: the paged fragment
2600 *
2601 * Releases a reference on the paged fragment @frag.
2602 */
2603static inline void __skb_frag_unref(skb_frag_t *frag)
2604{
2605 put_page(skb_frag_page(frag));
2606}
2607
2608/**
2609 * skb_frag_unref - release a reference on a paged fragment of an skb.
2610 * @skb: the buffer
2611 * @f: the fragment offset
2612 *
2613 * Releases a reference on the @f'th paged fragment of @skb.
2614 */
2615static inline void skb_frag_unref(struct sk_buff *skb, int f)
2616{
2617 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2618}
2619
2620/**
2621 * skb_frag_address - gets the address of the data contained in a paged fragment
2622 * @frag: the paged fragment buffer
2623 *
2624 * Returns the address of the data within @frag. The page must already
2625 * be mapped.
2626 */
2627static inline void *skb_frag_address(const skb_frag_t *frag)
2628{
2629 return page_address(skb_frag_page(frag)) + frag->page_offset;
2630}
2631
2632/**
2633 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2634 * @frag: the paged fragment buffer
2635 *
2636 * Returns the address of the data within @frag. Checks that the page
2637 * is mapped and returns %NULL otherwise.
2638 */
2639static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2640{
2641 void *ptr = page_address(skb_frag_page(frag));
2642 if (unlikely(!ptr))
2643 return NULL;
2644
2645 return ptr + frag->page_offset;
2646}
2647
2648/**
2649 * __skb_frag_set_page - sets the page contained in a paged fragment
2650 * @frag: the paged fragment
2651 * @page: the page to set
2652 *
2653 * Sets the fragment @frag to contain @page.
2654 */
2655static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2656{
a8605c60 2657 frag->page.p = page;
131ea667
IC
2658}
2659
2660/**
2661 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2662 * @skb: the buffer
2663 * @f: the fragment offset
2664 * @page: the page to set
2665 *
2666 * Sets the @f'th fragment of @skb to contain @page.
2667 */
2668static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2669 struct page *page)
2670{
2671 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2672}
2673
400dfd3a
ED
2674bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2675
131ea667
IC
2676/**
2677 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2678 * @dev: the device to map the fragment to
131ea667
IC
2679 * @frag: the paged fragment to map
2680 * @offset: the offset within the fragment (starting at the
2681 * fragment's own offset)
2682 * @size: the number of bytes to map
f83347df 2683 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2684 *
2685 * Maps the page associated with @frag to @device.
2686 */
2687static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2688 const skb_frag_t *frag,
2689 size_t offset, size_t size,
2690 enum dma_data_direction dir)
2691{
2692 return dma_map_page(dev, skb_frag_page(frag),
2693 frag->page_offset + offset, size, dir);
2694}
2695
117632e6
ED
2696static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2697 gfp_t gfp_mask)
2698{
2699 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2700}
2701
bad93e9d
OP
2702
2703static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2704 gfp_t gfp_mask)
2705{
2706 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2707}
2708
2709
334a8132
PM
2710/**
2711 * skb_clone_writable - is the header of a clone writable
2712 * @skb: buffer to check
2713 * @len: length up to which to write
2714 *
2715 * Returns true if modifying the header part of the cloned buffer
2716 * does not requires the data to be copied.
2717 */
05bdd2f1 2718static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2719{
2720 return !skb_header_cloned(skb) &&
2721 skb_headroom(skb) + len <= skb->hdr_len;
2722}
2723
3697649f
DB
2724static inline int skb_try_make_writable(struct sk_buff *skb,
2725 unsigned int write_len)
2726{
2727 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2728 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2729}
2730
d9cc2048
HX
2731static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2732 int cloned)
2733{
2734 int delta = 0;
2735
d9cc2048
HX
2736 if (headroom > skb_headroom(skb))
2737 delta = headroom - skb_headroom(skb);
2738
2739 if (delta || cloned)
2740 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2741 GFP_ATOMIC);
2742 return 0;
2743}
2744
1da177e4
LT
2745/**
2746 * skb_cow - copy header of skb when it is required
2747 * @skb: buffer to cow
2748 * @headroom: needed headroom
2749 *
2750 * If the skb passed lacks sufficient headroom or its data part
2751 * is shared, data is reallocated. If reallocation fails, an error
2752 * is returned and original skb is not changed.
2753 *
2754 * The result is skb with writable area skb->head...skb->tail
2755 * and at least @headroom of space at head.
2756 */
2757static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2758{
d9cc2048
HX
2759 return __skb_cow(skb, headroom, skb_cloned(skb));
2760}
1da177e4 2761
d9cc2048
HX
2762/**
2763 * skb_cow_head - skb_cow but only making the head writable
2764 * @skb: buffer to cow
2765 * @headroom: needed headroom
2766 *
2767 * This function is identical to skb_cow except that we replace the
2768 * skb_cloned check by skb_header_cloned. It should be used when
2769 * you only need to push on some header and do not need to modify
2770 * the data.
2771 */
2772static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2773{
2774 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2775}
2776
2777/**
2778 * skb_padto - pad an skbuff up to a minimal size
2779 * @skb: buffer to pad
2780 * @len: minimal length
2781 *
2782 * Pads up a buffer to ensure the trailing bytes exist and are
2783 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2784 * is untouched. Otherwise it is extended. Returns zero on
2785 * success. The skb is freed on error.
1da177e4 2786 */
5b057c6b 2787static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2788{
2789 unsigned int size = skb->len;
2790 if (likely(size >= len))
5b057c6b 2791 return 0;
987c402a 2792 return skb_pad(skb, len - size);
1da177e4
LT
2793}
2794
9c0c1124
AD
2795/**
2796 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2797 * @skb: buffer to pad
2798 * @len: minimal length
2799 *
2800 * Pads up a buffer to ensure the trailing bytes exist and are
2801 * blanked. If the buffer already contains sufficient data it
2802 * is untouched. Otherwise it is extended. Returns zero on
2803 * success. The skb is freed on error.
2804 */
2805static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2806{
2807 unsigned int size = skb->len;
2808
2809 if (unlikely(size < len)) {
2810 len -= size;
2811 if (skb_pad(skb, len))
2812 return -ENOMEM;
2813 __skb_put(skb, len);
2814 }
2815 return 0;
2816}
2817
1da177e4 2818static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2819 struct iov_iter *from, int copy)
1da177e4
LT
2820{
2821 const int off = skb->len;
2822
2823 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 2824 __wsum csum = 0;
15e6cb46
AV
2825 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2826 &csum, from)) {
1da177e4
LT
2827 skb->csum = csum_block_add(skb->csum, csum, off);
2828 return 0;
2829 }
15e6cb46 2830 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
2831 return 0;
2832
2833 __skb_trim(skb, off);
2834 return -EFAULT;
2835}
2836
38ba0a65
ED
2837static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2838 const struct page *page, int off)
1da177e4
LT
2839{
2840 if (i) {
9e903e08 2841 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2842
ea2ab693 2843 return page == skb_frag_page(frag) &&
9e903e08 2844 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2845 }
38ba0a65 2846 return false;
1da177e4
LT
2847}
2848
364c6bad
HX
2849static inline int __skb_linearize(struct sk_buff *skb)
2850{
2851 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2852}
2853
1da177e4
LT
2854/**
2855 * skb_linearize - convert paged skb to linear one
2856 * @skb: buffer to linarize
1da177e4
LT
2857 *
2858 * If there is no free memory -ENOMEM is returned, otherwise zero
2859 * is returned and the old skb data released.
2860 */
364c6bad
HX
2861static inline int skb_linearize(struct sk_buff *skb)
2862{
2863 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2864}
2865
cef401de
ED
2866/**
2867 * skb_has_shared_frag - can any frag be overwritten
2868 * @skb: buffer to test
2869 *
2870 * Return true if the skb has at least one frag that might be modified
2871 * by an external entity (as in vmsplice()/sendfile())
2872 */
2873static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2874{
c9af6db4
PS
2875 return skb_is_nonlinear(skb) &&
2876 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2877}
2878
364c6bad
HX
2879/**
2880 * skb_linearize_cow - make sure skb is linear and writable
2881 * @skb: buffer to process
2882 *
2883 * If there is no free memory -ENOMEM is returned, otherwise zero
2884 * is returned and the old skb data released.
2885 */
2886static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2887{
364c6bad
HX
2888 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2889 __skb_linearize(skb) : 0;
1da177e4
LT
2890}
2891
479ffccc
DB
2892static __always_inline void
2893__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2894 unsigned int off)
2895{
2896 if (skb->ip_summed == CHECKSUM_COMPLETE)
2897 skb->csum = csum_block_sub(skb->csum,
2898 csum_partial(start, len, 0), off);
2899 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2900 skb_checksum_start_offset(skb) < 0)
2901 skb->ip_summed = CHECKSUM_NONE;
2902}
2903
1da177e4
LT
2904/**
2905 * skb_postpull_rcsum - update checksum for received skb after pull
2906 * @skb: buffer to update
2907 * @start: start of data before pull
2908 * @len: length of data pulled
2909 *
2910 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2911 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2912 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 2913 */
1da177e4 2914static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2915 const void *start, unsigned int len)
1da177e4 2916{
479ffccc 2917 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
2918}
2919
479ffccc
DB
2920static __always_inline void
2921__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2922 unsigned int off)
2923{
2924 if (skb->ip_summed == CHECKSUM_COMPLETE)
2925 skb->csum = csum_block_add(skb->csum,
2926 csum_partial(start, len, 0), off);
2927}
cbb042f9 2928
479ffccc
DB
2929/**
2930 * skb_postpush_rcsum - update checksum for received skb after push
2931 * @skb: buffer to update
2932 * @start: start of data after push
2933 * @len: length of data pushed
2934 *
2935 * After doing a push on a received packet, you need to call this to
2936 * update the CHECKSUM_COMPLETE checksum.
2937 */
f8ffad69
DB
2938static inline void skb_postpush_rcsum(struct sk_buff *skb,
2939 const void *start, unsigned int len)
2940{
479ffccc 2941 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
2942}
2943
479ffccc
DB
2944unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2945
82a31b92
WC
2946/**
2947 * skb_push_rcsum - push skb and update receive checksum
2948 * @skb: buffer to update
2949 * @len: length of data pulled
2950 *
2951 * This function performs an skb_push on the packet and updates
2952 * the CHECKSUM_COMPLETE checksum. It should be used on
2953 * receive path processing instead of skb_push unless you know
2954 * that the checksum difference is zero (e.g., a valid IP header)
2955 * or you are setting ip_summed to CHECKSUM_NONE.
2956 */
2957static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2958 unsigned int len)
2959{
2960 skb_push(skb, len);
2961 skb_postpush_rcsum(skb, skb->data, len);
2962 return skb->data;
2963}
2964
7ce5a27f
DM
2965/**
2966 * pskb_trim_rcsum - trim received skb and update checksum
2967 * @skb: buffer to trim
2968 * @len: new length
2969 *
2970 * This is exactly the same as pskb_trim except that it ensures the
2971 * checksum of received packets are still valid after the operation.
2972 */
2973
2974static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2975{
2976 if (likely(len >= skb->len))
2977 return 0;
2978 if (skb->ip_summed == CHECKSUM_COMPLETE)
2979 skb->ip_summed = CHECKSUM_NONE;
2980 return __pskb_trim(skb, len);
2981}
2982
5293efe6
DB
2983static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2984{
2985 if (skb->ip_summed == CHECKSUM_COMPLETE)
2986 skb->ip_summed = CHECKSUM_NONE;
2987 __skb_trim(skb, len);
2988 return 0;
2989}
2990
2991static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2992{
2993 if (skb->ip_summed == CHECKSUM_COMPLETE)
2994 skb->ip_summed = CHECKSUM_NONE;
2995 return __skb_grow(skb, len);
2996}
2997
1da177e4
LT
2998#define skb_queue_walk(queue, skb) \
2999 for (skb = (queue)->next; \
a1e4891f 3000 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3001 skb = skb->next)
3002
46f8914e
JC
3003#define skb_queue_walk_safe(queue, skb, tmp) \
3004 for (skb = (queue)->next, tmp = skb->next; \
3005 skb != (struct sk_buff *)(queue); \
3006 skb = tmp, tmp = skb->next)
3007
1164f52a 3008#define skb_queue_walk_from(queue, skb) \
a1e4891f 3009 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3010 skb = skb->next)
3011
3012#define skb_queue_walk_from_safe(queue, skb, tmp) \
3013 for (tmp = skb->next; \
3014 skb != (struct sk_buff *)(queue); \
3015 skb = tmp, tmp = skb->next)
3016
300ce174
SH
3017#define skb_queue_reverse_walk(queue, skb) \
3018 for (skb = (queue)->prev; \
a1e4891f 3019 skb != (struct sk_buff *)(queue); \
300ce174
SH
3020 skb = skb->prev)
3021
686a2955
DM
3022#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3023 for (skb = (queue)->prev, tmp = skb->prev; \
3024 skb != (struct sk_buff *)(queue); \
3025 skb = tmp, tmp = skb->prev)
3026
3027#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3028 for (tmp = skb->prev; \
3029 skb != (struct sk_buff *)(queue); \
3030 skb = tmp, tmp = skb->prev)
1da177e4 3031
21dc3301 3032static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3033{
3034 return skb_shinfo(skb)->frag_list != NULL;
3035}
3036
3037static inline void skb_frag_list_init(struct sk_buff *skb)
3038{
3039 skb_shinfo(skb)->frag_list = NULL;
3040}
3041
ee039871
DM
3042#define skb_walk_frags(skb, iter) \
3043 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3044
ea3793ee
RW
3045
3046int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3047 const struct sk_buff *skb);
3048struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3049 void (*destructor)(struct sock *sk,
3050 struct sk_buff *skb),
ea3793ee
RW
3051 int *peeked, int *off, int *err,
3052 struct sk_buff **last);
7965bd4d 3053struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3054 void (*destructor)(struct sock *sk,
3055 struct sk_buff *skb),
7965bd4d
JP
3056 int *peeked, int *off, int *err);
3057struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3058 int *err);
3059unsigned int datagram_poll(struct file *file, struct socket *sock,
3060 struct poll_table_struct *wait);
c0371da6
AV
3061int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3062 struct iov_iter *to, int size);
51f3d02b
DM
3063static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3064 struct msghdr *msg, int size)
3065{
e5a4b0bb 3066 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3067}
e5a4b0bb
AV
3068int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3069 struct msghdr *msg);
3a654f97
AV
3070int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3071 struct iov_iter *from, int len);
3a654f97 3072int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3073void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3074void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3075static inline void skb_free_datagram_locked(struct sock *sk,
3076 struct sk_buff *skb)
3077{
3078 __skb_free_datagram_locked(sk, skb, 0);
3079}
7965bd4d 3080int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3081int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3082int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3083__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3084 int len, __wsum csum);
a60e3cc7 3085int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3086 struct pipe_inode_info *pipe, unsigned int len,
25869262 3087 unsigned int flags);
7965bd4d 3088void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3089unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3090int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3091 int len, int hlen);
7965bd4d
JP
3092void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3093int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3094void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3095unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3096bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 3097struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3098struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3099int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3100int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3101int skb_vlan_pop(struct sk_buff *skb);
3102int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3103struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3104 gfp_t gfp);
20380731 3105
6ce8e9ce
AV
3106static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3107{
21226abb 3108 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
3109}
3110
7eab8d9e
AV
3111static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3112{
e5a4b0bb 3113 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3114}
3115
2817a336
DB
3116struct skb_checksum_ops {
3117 __wsum (*update)(const void *mem, int len, __wsum wsum);
3118 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3119};
3120
3121__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3122 __wsum csum, const struct skb_checksum_ops *ops);
3123__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3124 __wsum csum);
3125
1e98a0f0
ED
3126static inline void * __must_check
3127__skb_header_pointer(const struct sk_buff *skb, int offset,
3128 int len, void *data, int hlen, void *buffer)
1da177e4 3129{
55820ee2 3130 if (hlen - offset >= len)
690e36e7 3131 return data + offset;
1da177e4 3132
690e36e7
DM
3133 if (!skb ||
3134 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3135 return NULL;
3136
3137 return buffer;
3138}
3139
1e98a0f0
ED
3140static inline void * __must_check
3141skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3142{
3143 return __skb_header_pointer(skb, offset, len, skb->data,
3144 skb_headlen(skb), buffer);
3145}
3146
4262e5cc
DB
3147/**
3148 * skb_needs_linearize - check if we need to linearize a given skb
3149 * depending on the given device features.
3150 * @skb: socket buffer to check
3151 * @features: net device features
3152 *
3153 * Returns true if either:
3154 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3155 * 2. skb is fragmented and the device does not support SG.
3156 */
3157static inline bool skb_needs_linearize(struct sk_buff *skb,
3158 netdev_features_t features)
3159{
3160 return skb_is_nonlinear(skb) &&
3161 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3162 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3163}
3164
d626f62b
ACM
3165static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3166 void *to,
3167 const unsigned int len)
3168{
3169 memcpy(to, skb->data, len);
3170}
3171
3172static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3173 const int offset, void *to,
3174 const unsigned int len)
3175{
3176 memcpy(to, skb->data + offset, len);
3177}
3178
27d7ff46
ACM
3179static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3180 const void *from,
3181 const unsigned int len)
3182{
3183 memcpy(skb->data, from, len);
3184}
3185
3186static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3187 const int offset,
3188 const void *from,
3189 const unsigned int len)
3190{
3191 memcpy(skb->data + offset, from, len);
3192}
3193
7965bd4d 3194void skb_init(void);
1da177e4 3195
ac45f602
PO
3196static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3197{
3198 return skb->tstamp;
3199}
3200
a61bbcf2
PM
3201/**
3202 * skb_get_timestamp - get timestamp from a skb
3203 * @skb: skb to get stamp from
3204 * @stamp: pointer to struct timeval to store stamp in
3205 *
3206 * Timestamps are stored in the skb as offsets to a base timestamp.
3207 * This function converts the offset back to a struct timeval and stores
3208 * it in stamp.
3209 */
ac45f602
PO
3210static inline void skb_get_timestamp(const struct sk_buff *skb,
3211 struct timeval *stamp)
a61bbcf2 3212{
b7aa0bf7 3213 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3214}
3215
ac45f602
PO
3216static inline void skb_get_timestampns(const struct sk_buff *skb,
3217 struct timespec *stamp)
3218{
3219 *stamp = ktime_to_timespec(skb->tstamp);
3220}
3221
b7aa0bf7 3222static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3223{
b7aa0bf7 3224 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3225}
3226
164891aa
SH
3227static inline ktime_t net_timedelta(ktime_t t)
3228{
3229 return ktime_sub(ktime_get_real(), t);
3230}
3231
b9ce204f
IJ
3232static inline ktime_t net_invalid_timestamp(void)
3233{
8b0e1953 3234 return 0;
b9ce204f 3235}
a61bbcf2 3236
62bccb8c
AD
3237struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3238
c1f19b51
RC
3239#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3240
7965bd4d
JP
3241void skb_clone_tx_timestamp(struct sk_buff *skb);
3242bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3243
3244#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3245
3246static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3247{
3248}
3249
3250static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3251{
3252 return false;
3253}
3254
3255#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3256
3257/**
3258 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3259 *
da92b194
RC
3260 * PHY drivers may accept clones of transmitted packets for
3261 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3262 * must call this function to return the skb back to the stack with a
3263 * timestamp.
da92b194 3264 *
c1f19b51 3265 * @skb: clone of the the original outgoing packet
7a76a021 3266 * @hwtstamps: hardware time stamps
c1f19b51
RC
3267 *
3268 */
3269void skb_complete_tx_timestamp(struct sk_buff *skb,
3270 struct skb_shared_hwtstamps *hwtstamps);
3271
e7fd2885
WB
3272void __skb_tstamp_tx(struct sk_buff *orig_skb,
3273 struct skb_shared_hwtstamps *hwtstamps,
3274 struct sock *sk, int tstype);
3275
ac45f602
PO
3276/**
3277 * skb_tstamp_tx - queue clone of skb with send time stamps
3278 * @orig_skb: the original outgoing packet
3279 * @hwtstamps: hardware time stamps, may be NULL if not available
3280 *
3281 * If the skb has a socket associated, then this function clones the
3282 * skb (thus sharing the actual data and optional structures), stores
3283 * the optional hardware time stamping information (if non NULL) or
3284 * generates a software time stamp (otherwise), then queues the clone
3285 * to the error queue of the socket. Errors are silently ignored.
3286 */
7965bd4d
JP
3287void skb_tstamp_tx(struct sk_buff *orig_skb,
3288 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3289
4507a715
RC
3290static inline void sw_tx_timestamp(struct sk_buff *skb)
3291{
2244d07b
OH
3292 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3293 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3294 skb_tstamp_tx(skb, NULL);
3295}
3296
3297/**
3298 * skb_tx_timestamp() - Driver hook for transmit timestamping
3299 *
3300 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3301 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3302 *
73409f3b
DM
3303 * Specifically, one should make absolutely sure that this function is
3304 * called before TX completion of this packet can trigger. Otherwise
3305 * the packet could potentially already be freed.
3306 *
4507a715
RC
3307 * @skb: A socket buffer.
3308 */
3309static inline void skb_tx_timestamp(struct sk_buff *skb)
3310{
c1f19b51 3311 skb_clone_tx_timestamp(skb);
4507a715
RC
3312 sw_tx_timestamp(skb);
3313}
3314
6e3e939f
JB
3315/**
3316 * skb_complete_wifi_ack - deliver skb with wifi status
3317 *
3318 * @skb: the original outgoing packet
3319 * @acked: ack status
3320 *
3321 */
3322void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3323
7965bd4d
JP
3324__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3325__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3326
60476372
HX
3327static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3328{
6edec0e6
TH
3329 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3330 skb->csum_valid ||
3331 (skb->ip_summed == CHECKSUM_PARTIAL &&
3332 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3333}
3334
fb286bb2
HX
3335/**
3336 * skb_checksum_complete - Calculate checksum of an entire packet
3337 * @skb: packet to process
3338 *
3339 * This function calculates the checksum over the entire packet plus
3340 * the value of skb->csum. The latter can be used to supply the
3341 * checksum of a pseudo header as used by TCP/UDP. It returns the
3342 * checksum.
3343 *
3344 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3345 * this function can be used to verify that checksum on received
3346 * packets. In that case the function should return zero if the
3347 * checksum is correct. In particular, this function will return zero
3348 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3349 * hardware has already verified the correctness of the checksum.
3350 */
4381ca3c 3351static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3352{
60476372
HX
3353 return skb_csum_unnecessary(skb) ?
3354 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3355}
3356
77cffe23
TH
3357static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3358{
3359 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3360 if (skb->csum_level == 0)
3361 skb->ip_summed = CHECKSUM_NONE;
3362 else
3363 skb->csum_level--;
3364 }
3365}
3366
3367static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3368{
3369 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3370 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3371 skb->csum_level++;
3372 } else if (skb->ip_summed == CHECKSUM_NONE) {
3373 skb->ip_summed = CHECKSUM_UNNECESSARY;
3374 skb->csum_level = 0;
3375 }
3376}
3377
5a212329
TH
3378static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3379{
3380 /* Mark current checksum as bad (typically called from GRO
3381 * path). In the case that ip_summed is CHECKSUM_NONE
3382 * this must be the first checksum encountered in the packet.
3383 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3384 * checksum after the last one validated. For UDP, a zero
3385 * checksum can not be marked as bad.
3386 */
3387
3388 if (skb->ip_summed == CHECKSUM_NONE ||
3389 skb->ip_summed == CHECKSUM_UNNECESSARY)
3390 skb->csum_bad = 1;
3391}
3392
76ba0aae
TH
3393/* Check if we need to perform checksum complete validation.
3394 *
3395 * Returns true if checksum complete is needed, false otherwise
3396 * (either checksum is unnecessary or zero checksum is allowed).
3397 */
3398static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3399 bool zero_okay,
3400 __sum16 check)
3401{
5d0c2b95
TH
3402 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3403 skb->csum_valid = 1;
77cffe23 3404 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3405 return false;
3406 }
3407
3408 return true;
3409}
3410
3411/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3412 * in checksum_init.
3413 */
3414#define CHECKSUM_BREAK 76
3415
4e18b9ad
TH
3416/* Unset checksum-complete
3417 *
3418 * Unset checksum complete can be done when packet is being modified
3419 * (uncompressed for instance) and checksum-complete value is
3420 * invalidated.
3421 */
3422static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3423{
3424 if (skb->ip_summed == CHECKSUM_COMPLETE)
3425 skb->ip_summed = CHECKSUM_NONE;
3426}
3427
76ba0aae
TH
3428/* Validate (init) checksum based on checksum complete.
3429 *
3430 * Return values:
3431 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3432 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3433 * checksum is stored in skb->csum for use in __skb_checksum_complete
3434 * non-zero: value of invalid checksum
3435 *
3436 */
3437static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3438 bool complete,
3439 __wsum psum)
3440{
3441 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3442 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3443 skb->csum_valid = 1;
76ba0aae
TH
3444 return 0;
3445 }
5a212329
TH
3446 } else if (skb->csum_bad) {
3447 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3448 return (__force __sum16)1;
76ba0aae
TH
3449 }
3450
3451 skb->csum = psum;
3452
5d0c2b95
TH
3453 if (complete || skb->len <= CHECKSUM_BREAK) {
3454 __sum16 csum;
3455
3456 csum = __skb_checksum_complete(skb);
3457 skb->csum_valid = !csum;
3458 return csum;
3459 }
76ba0aae
TH
3460
3461 return 0;
3462}
3463
3464static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3465{
3466 return 0;
3467}
3468
3469/* Perform checksum validate (init). Note that this is a macro since we only
3470 * want to calculate the pseudo header which is an input function if necessary.
3471 * First we try to validate without any computation (checksum unnecessary) and
3472 * then calculate based on checksum complete calling the function to compute
3473 * pseudo header.
3474 *
3475 * Return values:
3476 * 0: checksum is validated or try to in skb_checksum_complete
3477 * non-zero: value of invalid checksum
3478 */
3479#define __skb_checksum_validate(skb, proto, complete, \
3480 zero_okay, check, compute_pseudo) \
3481({ \
3482 __sum16 __ret = 0; \
5d0c2b95 3483 skb->csum_valid = 0; \
76ba0aae
TH
3484 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3485 __ret = __skb_checksum_validate_complete(skb, \
3486 complete, compute_pseudo(skb, proto)); \
3487 __ret; \
3488})
3489
3490#define skb_checksum_init(skb, proto, compute_pseudo) \
3491 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3492
3493#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3494 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3495
3496#define skb_checksum_validate(skb, proto, compute_pseudo) \
3497 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3498
3499#define skb_checksum_validate_zero_check(skb, proto, check, \
3500 compute_pseudo) \
096a4cfa 3501 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3502
3503#define skb_checksum_simple_validate(skb) \
3504 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3505
d96535a1
TH
3506static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3507{
3508 return (skb->ip_summed == CHECKSUM_NONE &&
3509 skb->csum_valid && !skb->csum_bad);
3510}
3511
3512static inline void __skb_checksum_convert(struct sk_buff *skb,
3513 __sum16 check, __wsum pseudo)
3514{
3515 skb->csum = ~pseudo;
3516 skb->ip_summed = CHECKSUM_COMPLETE;
3517}
3518
3519#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3520do { \
3521 if (__skb_checksum_convert_check(skb)) \
3522 __skb_checksum_convert(skb, check, \
3523 compute_pseudo(skb, proto)); \
3524} while (0)
3525
15e2396d
TH
3526static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3527 u16 start, u16 offset)
3528{
3529 skb->ip_summed = CHECKSUM_PARTIAL;
3530 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3531 skb->csum_offset = offset - start;
3532}
3533
dcdc8994
TH
3534/* Update skbuf and packet to reflect the remote checksum offload operation.
3535 * When called, ptr indicates the starting point for skb->csum when
3536 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3537 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3538 */
3539static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3540 int start, int offset, bool nopartial)
dcdc8994
TH
3541{
3542 __wsum delta;
3543
15e2396d
TH
3544 if (!nopartial) {
3545 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3546 return;
3547 }
3548
dcdc8994
TH
3549 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3550 __skb_checksum_complete(skb);
3551 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3552 }
3553
3554 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3555
3556 /* Adjust skb->csum since we changed the packet */
3557 skb->csum = csum_add(skb->csum, delta);
3558}
3559
cb9c6836
FW
3560static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3561{
3562#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 3563 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
3564#else
3565 return NULL;
3566#endif
3567}
3568
5f79e0f9 3569#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3570void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3571static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3572{
3573 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3574 nf_conntrack_destroy(nfct);
1da177e4
LT
3575}
3576static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3577{
3578 if (nfct)
3579 atomic_inc(&nfct->use);
3580}
2fc72c7b 3581#endif
34666d46 3582#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3583static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3584{
3585 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3586 kfree(nf_bridge);
3587}
3588static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3589{
3590 if (nf_bridge)
3591 atomic_inc(&nf_bridge->use);
3592}
3593#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3594static inline void nf_reset(struct sk_buff *skb)
3595{
5f79e0f9 3596#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3597 nf_conntrack_put(skb_nfct(skb));
3598 skb->_nfct = 0;
2fc72c7b 3599#endif
34666d46 3600#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3601 nf_bridge_put(skb->nf_bridge);
3602 skb->nf_bridge = NULL;
3603#endif
3604}
3605
124dff01
PM
3606static inline void nf_reset_trace(struct sk_buff *skb)
3607{
478b360a 3608#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3609 skb->nf_trace = 0;
3610#endif
a193a4ab
PM
3611}
3612
edda553c 3613/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3614static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3615 bool copy)
edda553c 3616{
5f79e0f9 3617#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3618 dst->_nfct = src->_nfct;
3619 nf_conntrack_get(skb_nfct(src));
2fc72c7b 3620#endif
34666d46 3621#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3622 dst->nf_bridge = src->nf_bridge;
3623 nf_bridge_get(src->nf_bridge);
3624#endif
478b360a 3625#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3626 if (copy)
3627 dst->nf_trace = src->nf_trace;
478b360a 3628#endif
edda553c
YK
3629}
3630
e7ac05f3
YK
3631static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3632{
e7ac05f3 3633#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 3634 nf_conntrack_put(skb_nfct(dst));
2fc72c7b 3635#endif
34666d46 3636#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3637 nf_bridge_put(dst->nf_bridge);
3638#endif
b1937227 3639 __nf_copy(dst, src, true);
e7ac05f3
YK
3640}
3641
984bc16c
JM
3642#ifdef CONFIG_NETWORK_SECMARK
3643static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3644{
3645 to->secmark = from->secmark;
3646}
3647
3648static inline void skb_init_secmark(struct sk_buff *skb)
3649{
3650 skb->secmark = 0;
3651}
3652#else
3653static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3654{ }
3655
3656static inline void skb_init_secmark(struct sk_buff *skb)
3657{ }
3658#endif
3659
574f7194
EB
3660static inline bool skb_irq_freeable(const struct sk_buff *skb)
3661{
3662 return !skb->destructor &&
3663#if IS_ENABLED(CONFIG_XFRM)
3664 !skb->sp &&
3665#endif
cb9c6836 3666 !skb_nfct(skb) &&
574f7194
EB
3667 !skb->_skb_refdst &&
3668 !skb_has_frag_list(skb);
3669}
3670
f25f4e44
PWJ
3671static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3672{
f25f4e44 3673 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3674}
3675
9247744e 3676static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3677{
4e3ab47a 3678 return skb->queue_mapping;
4e3ab47a
PE
3679}
3680
f25f4e44
PWJ
3681static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3682{
f25f4e44 3683 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3684}
3685
d5a9e24a
DM
3686static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3687{
3688 skb->queue_mapping = rx_queue + 1;
3689}
3690
9247744e 3691static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3692{
3693 return skb->queue_mapping - 1;
3694}
3695
9247744e 3696static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3697{
a02cec21 3698 return skb->queue_mapping != 0;
d5a9e24a
DM
3699}
3700
def8b4fa
AD
3701static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3702{
0b3d8e08 3703#ifdef CONFIG_XFRM
def8b4fa 3704 return skb->sp;
def8b4fa 3705#else
def8b4fa 3706 return NULL;
def8b4fa 3707#endif
0b3d8e08 3708}
def8b4fa 3709
68c33163
PS
3710/* Keeps track of mac header offset relative to skb->head.
3711 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3712 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3713 * tunnel skb it points to outer mac header.
3714 * Keeps track of level of encapsulation of network headers.
3715 */
68c33163 3716struct skb_gso_cb {
802ab55a
AD
3717 union {
3718 int mac_offset;
3719 int data_offset;
3720 };
3347c960 3721 int encap_level;
76443456 3722 __wsum csum;
7e2b10c1 3723 __u16 csum_start;
68c33163 3724};
9207f9d4
KK
3725#define SKB_SGO_CB_OFFSET 32
3726#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3727
3728static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3729{
3730 return (skb_mac_header(inner_skb) - inner_skb->head) -
3731 SKB_GSO_CB(inner_skb)->mac_offset;
3732}
3733
1e2bd517
PS
3734static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3735{
3736 int new_headroom, headroom;
3737 int ret;
3738
3739 headroom = skb_headroom(skb);
3740 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3741 if (ret)
3742 return ret;
3743
3744 new_headroom = skb_headroom(skb);
3745 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3746 return 0;
3747}
3748
08b64fcc
AD
3749static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3750{
3751 /* Do not update partial checksums if remote checksum is enabled. */
3752 if (skb->remcsum_offload)
3753 return;
3754
3755 SKB_GSO_CB(skb)->csum = res;
3756 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3757}
3758
7e2b10c1
TH
3759/* Compute the checksum for a gso segment. First compute the checksum value
3760 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3761 * then add in skb->csum (checksum from csum_start to end of packet).
3762 * skb->csum and csum_start are then updated to reflect the checksum of the
3763 * resultant packet starting from the transport header-- the resultant checksum
3764 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3765 * header.
3766 */
3767static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3768{
76443456
AD
3769 unsigned char *csum_start = skb_transport_header(skb);
3770 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3771 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3772
76443456
AD
3773 SKB_GSO_CB(skb)->csum = res;
3774 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3775
76443456 3776 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3777}
3778
bdcc0924 3779static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3780{
3781 return skb_shinfo(skb)->gso_size;
3782}
3783
36a8f39e 3784/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3785static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3786{
3787 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3788}
3789
5293efe6
DB
3790static inline void skb_gso_reset(struct sk_buff *skb)
3791{
3792 skb_shinfo(skb)->gso_size = 0;
3793 skb_shinfo(skb)->gso_segs = 0;
3794 skb_shinfo(skb)->gso_type = 0;
3795}
3796
7965bd4d 3797void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3798
3799static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3800{
3801 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3802 * wanted then gso_type will be set. */
05bdd2f1
ED
3803 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3804
b78462eb
AD
3805 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3806 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3807 __skb_warn_lro_forwarding(skb);
3808 return true;
3809 }
3810 return false;
3811}
3812
35fc92a9
HX
3813static inline void skb_forward_csum(struct sk_buff *skb)
3814{
3815 /* Unfortunately we don't support this one. Any brave souls? */
3816 if (skb->ip_summed == CHECKSUM_COMPLETE)
3817 skb->ip_summed = CHECKSUM_NONE;
3818}
3819
bc8acf2c
ED
3820/**
3821 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3822 * @skb: skb to check
3823 *
3824 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3825 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3826 * use this helper, to document places where we make this assertion.
3827 */
05bdd2f1 3828static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3829{
3830#ifdef DEBUG
3831 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3832#endif
3833}
3834
f35d9d8a 3835bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3836
ed1f50c3 3837int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3838struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3839 unsigned int transport_len,
3840 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3841
3a7c1ee4
AD
3842/**
3843 * skb_head_is_locked - Determine if the skb->head is locked down
3844 * @skb: skb to check
3845 *
3846 * The head on skbs build around a head frag can be removed if they are
3847 * not cloned. This function returns true if the skb head is locked down
3848 * due to either being allocated via kmalloc, or by being a clone with
3849 * multiple references to the head.
3850 */
3851static inline bool skb_head_is_locked(const struct sk_buff *skb)
3852{
3853 return !skb->head_frag || skb_cloned(skb);
3854}
fe6cc55f
FW
3855
3856/**
3857 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3858 *
3859 * @skb: GSO skb
3860 *
3861 * skb_gso_network_seglen is used to determine the real size of the
3862 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3863 *
3864 * The MAC/L2 header is not accounted for.
3865 */
3866static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3867{
3868 unsigned int hdr_len = skb_transport_header(skb) -
3869 skb_network_header(skb);
3870 return hdr_len + skb_gso_transport_seglen(skb);
3871}
ee122c79 3872
179bc67f
EC
3873/* Local Checksum Offload.
3874 * Compute outer checksum based on the assumption that the
3875 * inner checksum will be offloaded later.
e8ae7b00
EC
3876 * See Documentation/networking/checksum-offloads.txt for
3877 * explanation of how this works.
179bc67f
EC
3878 * Fill in outer checksum adjustment (e.g. with sum of outer
3879 * pseudo-header) before calling.
3880 * Also ensure that inner checksum is in linear data area.
3881 */
3882static inline __wsum lco_csum(struct sk_buff *skb)
3883{
9e74a6da
AD
3884 unsigned char *csum_start = skb_checksum_start(skb);
3885 unsigned char *l4_hdr = skb_transport_header(skb);
3886 __wsum partial;
179bc67f
EC
3887
3888 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3889 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3890 skb->csum_offset));
3891
179bc67f 3892 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3893 * adjustment filled in by caller) and return result.
179bc67f 3894 */
9e74a6da 3895 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3896}
3897
1da177e4
LT
3898#endif /* __KERNEL__ */
3899#endif /* _LINUX_SKBUFF_H */