Merge tag 'pinctrl-v4.9-1' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
[linux-2.6-block.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
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 587 * @nohdr: Payload reference only, must not modify header
d84e0bd7 588 * @nfctinfo: Relationship of this skb to the connection
1da177e4 589 * @pkt_type: Packet class
c83c2486 590 * @fclone: skbuff clone status
c83c2486 591 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
592 * @peeked: this packet has been seen already, so stats have been
593 * done for it, don't do them again
ba9dda3a 594 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
595 * @protocol: Packet protocol from driver
596 * @destructor: Destruct function
597 * @nfct: Associated connection, if any
1da177e4 598 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 599 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
600 * @tc_index: Traffic control index
601 * @tc_verd: traffic control verdict
61b905da 602 * @hash: the packet hash
d84e0bd7 603 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 604 * @xmit_more: More SKBs are pending for this queue
553a5672 605 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 606 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 607 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 608 * ports.
a3b18ddb 609 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
610 * @wifi_acked_valid: wifi_acked was set
611 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 612 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 613 * @napi_id: id of the NAPI struct this skb came from
984bc16c 614 * @secmark: security marking
d84e0bd7 615 * @mark: Generic packet mark
86a9bad3 616 * @vlan_proto: vlan encapsulation protocol
6aa895b0 617 * @vlan_tci: vlan tag control information
0d89d203 618 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
619 * @inner_transport_header: Inner transport layer header (encapsulation)
620 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 621 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
622 * @transport_header: Transport layer header
623 * @network_header: Network layer header
624 * @mac_header: Link layer header
625 * @tail: Tail pointer
626 * @end: End pointer
627 * @head: Head of buffer
628 * @data: Data head pointer
629 * @truesize: Buffer size
630 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
631 */
632
633struct sk_buff {
363ec392 634 union {
56b17425
ED
635 struct {
636 /* These two members must be first. */
637 struct sk_buff *next;
638 struct sk_buff *prev;
639
640 union {
641 ktime_t tstamp;
642 struct skb_mstamp skb_mstamp;
643 };
644 };
645 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 646 };
da3f5cf1 647 struct sock *sk;
1da177e4 648 struct net_device *dev;
1da177e4 649
1da177e4
LT
650 /*
651 * This is the control buffer. It is free to use for every
652 * layer. Please put your private variables there. If you
653 * want to keep them across layers you have to do a skb_clone()
654 * first. This is owned by whoever has the skb queued ATM.
655 */
da3f5cf1 656 char cb[48] __aligned(8);
1da177e4 657
7fee226a 658 unsigned long _skb_refdst;
b1937227 659 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
660#ifdef CONFIG_XFRM
661 struct sec_path *sp;
b1937227
ED
662#endif
663#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
664 struct nf_conntrack *nfct;
665#endif
85224844 666#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 667 struct nf_bridge_info *nf_bridge;
da3f5cf1 668#endif
1da177e4 669 unsigned int len,
334a8132
PM
670 data_len;
671 __u16 mac_len,
672 hdr_len;
b1937227
ED
673
674 /* Following fields are _not_ copied in __copy_skb_header()
675 * Note that queue_mapping is here mostly to fill a hole.
676 */
fe55f6d5 677 kmemcheck_bitfield_begin(flags1);
b1937227 678 __u16 queue_mapping;
36bbef52
DB
679
680/* if you move cloned around you also must adapt those constants */
681#ifdef __BIG_ENDIAN_BITFIELD
682#define CLONED_MASK (1 << 7)
683#else
684#define CLONED_MASK 1
685#endif
686#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
687
688 __u8 __cloned_offset[0];
b1937227 689 __u8 cloned:1,
6869c4d8 690 nohdr:1,
b84f4cc9 691 fclone:2,
a59322be 692 peeked:1,
b1937227 693 head_frag:1,
36bbef52
DB
694 xmit_more:1,
695 __unused:1; /* one bit hole */
fe55f6d5 696 kmemcheck_bitfield_end(flags1);
4031ae6e 697
b1937227
ED
698 /* fields enclosed in headers_start/headers_end are copied
699 * using a single memcpy() in __copy_skb_header()
700 */
ebcf34f3 701 /* private: */
b1937227 702 __u32 headers_start[0];
ebcf34f3 703 /* public: */
4031ae6e 704
233577a2
HFS
705/* if you move pkt_type around you also must adapt those constants */
706#ifdef __BIG_ENDIAN_BITFIELD
707#define PKT_TYPE_MAX (7 << 5)
708#else
709#define PKT_TYPE_MAX 7
1da177e4 710#endif
233577a2 711#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 712
233577a2 713 __u8 __pkt_type_offset[0];
b1937227 714 __u8 pkt_type:3;
c93bdd0e 715 __u8 pfmemalloc:1;
b1937227
ED
716 __u8 ignore_df:1;
717 __u8 nfctinfo:3;
718
719 __u8 nf_trace:1;
720 __u8 ip_summed:2;
3853b584 721 __u8 ooo_okay:1;
61b905da 722 __u8 l4_hash:1;
a3b18ddb 723 __u8 sw_hash:1;
6e3e939f
JB
724 __u8 wifi_acked_valid:1;
725 __u8 wifi_acked:1;
b1937227 726
3bdc0eba 727 __u8 no_fcs:1;
77cffe23 728 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 729 __u8 encapsulation:1;
7e2b10c1 730 __u8 encap_hdr_csum:1;
5d0c2b95 731 __u8 csum_valid:1;
7e3cead5 732 __u8 csum_complete_sw:1;
b1937227
ED
733 __u8 csum_level:2;
734 __u8 csum_bad:1;
fe55f6d5 735
b1937227
ED
736#ifdef CONFIG_IPV6_NDISC_NODETYPE
737 __u8 ndisc_nodetype:2;
738#endif
739 __u8 ipvs_property:1;
8bce6d7d 740 __u8 inner_protocol_type:1;
e585f236 741 __u8 remcsum_offload:1;
6bc506b4
IS
742#ifdef CONFIG_NET_SWITCHDEV
743 __u8 offload_fwd_mark:1;
744#endif
745 /* 2, 4 or 5 bit hole */
b1937227
ED
746
747#ifdef CONFIG_NET_SCHED
748 __u16 tc_index; /* traffic control index */
749#ifdef CONFIG_NET_CLS_ACT
750 __u16 tc_verd; /* traffic control verdict */
751#endif
752#endif
fe55f6d5 753
b1937227
ED
754 union {
755 __wsum csum;
756 struct {
757 __u16 csum_start;
758 __u16 csum_offset;
759 };
760 };
761 __u32 priority;
762 int skb_iif;
763 __u32 hash;
764 __be16 vlan_proto;
765 __u16 vlan_tci;
2bd82484
ED
766#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
767 union {
768 unsigned int napi_id;
769 unsigned int sender_cpu;
770 };
97fc2f08 771#endif
984bc16c 772#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 773 __u32 secmark;
0c4f691f 774#endif
0c4f691f 775
3b885787
NH
776 union {
777 __u32 mark;
16fad69c 778 __u32 reserved_tailroom;
3b885787 779 };
1da177e4 780
8bce6d7d
TH
781 union {
782 __be16 inner_protocol;
783 __u8 inner_ipproto;
784 };
785
1a37e412
SH
786 __u16 inner_transport_header;
787 __u16 inner_network_header;
788 __u16 inner_mac_header;
b1937227
ED
789
790 __be16 protocol;
1a37e412
SH
791 __u16 transport_header;
792 __u16 network_header;
793 __u16 mac_header;
b1937227 794
ebcf34f3 795 /* private: */
b1937227 796 __u32 headers_end[0];
ebcf34f3 797 /* public: */
b1937227 798
1da177e4 799 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 800 sk_buff_data_t tail;
4305b541 801 sk_buff_data_t end;
1da177e4 802 unsigned char *head,
4305b541 803 *data;
27a884dc
ACM
804 unsigned int truesize;
805 atomic_t users;
1da177e4
LT
806};
807
808#ifdef __KERNEL__
809/*
810 * Handling routines are only of interest to the kernel
811 */
812#include <linux/slab.h>
813
1da177e4 814
c93bdd0e
MG
815#define SKB_ALLOC_FCLONE 0x01
816#define SKB_ALLOC_RX 0x02
fd11a83d 817#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
818
819/* Returns true if the skb was allocated from PFMEMALLOC reserves */
820static inline bool skb_pfmemalloc(const struct sk_buff *skb)
821{
822 return unlikely(skb->pfmemalloc);
823}
824
7fee226a
ED
825/*
826 * skb might have a dst pointer attached, refcounted or not.
827 * _skb_refdst low order bit is set if refcount was _not_ taken
828 */
829#define SKB_DST_NOREF 1UL
830#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
831
832/**
833 * skb_dst - returns skb dst_entry
834 * @skb: buffer
835 *
836 * Returns skb dst_entry, regardless of reference taken or not.
837 */
adf30907
ED
838static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
839{
7fee226a
ED
840 /* If refdst was not refcounted, check we still are in a
841 * rcu_read_lock section
842 */
843 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
844 !rcu_read_lock_held() &&
845 !rcu_read_lock_bh_held());
846 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
847}
848
7fee226a
ED
849/**
850 * skb_dst_set - sets skb dst
851 * @skb: buffer
852 * @dst: dst entry
853 *
854 * Sets skb dst, assuming a reference was taken on dst and should
855 * be released by skb_dst_drop()
856 */
adf30907
ED
857static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
858{
7fee226a
ED
859 skb->_skb_refdst = (unsigned long)dst;
860}
861
932bc4d7
JA
862/**
863 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
864 * @skb: buffer
865 * @dst: dst entry
866 *
867 * Sets skb dst, assuming a reference was not taken on dst.
868 * If dst entry is cached, we do not take reference and dst_release
869 * will be avoided by refdst_drop. If dst entry is not cached, we take
870 * reference, so that last dst_release can destroy the dst immediately.
871 */
872static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
873{
dbfc4fb7
HFS
874 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
875 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 876}
7fee226a
ED
877
878/**
25985edc 879 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
880 * @skb: buffer
881 */
882static inline bool skb_dst_is_noref(const struct sk_buff *skb)
883{
884 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
885}
886
511c3f92
ED
887static inline struct rtable *skb_rtable(const struct sk_buff *skb)
888{
adf30907 889 return (struct rtable *)skb_dst(skb);
511c3f92
ED
890}
891
8b10cab6
JHS
892/* For mangling skb->pkt_type from user space side from applications
893 * such as nft, tc, etc, we only allow a conservative subset of
894 * possible pkt_types to be set.
895*/
896static inline bool skb_pkt_type_ok(u32 ptype)
897{
898 return ptype <= PACKET_OTHERHOST;
899}
900
7965bd4d
JP
901void kfree_skb(struct sk_buff *skb);
902void kfree_skb_list(struct sk_buff *segs);
903void skb_tx_error(struct sk_buff *skb);
904void consume_skb(struct sk_buff *skb);
905void __kfree_skb(struct sk_buff *skb);
d7e8883c 906extern struct kmem_cache *skbuff_head_cache;
bad43ca8 907
7965bd4d
JP
908void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
909bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
910 bool *fragstolen, int *delta_truesize);
bad43ca8 911
7965bd4d
JP
912struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
913 int node);
2ea2f62c 914struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 915struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 916static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 917 gfp_t priority)
d179cd12 918{
564824b0 919 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
920}
921
2e4e4410
ED
922struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
923 unsigned long data_len,
924 int max_page_order,
925 int *errcode,
926 gfp_t gfp_mask);
927
d0bf4a9e
ED
928/* Layout of fast clones : [skb1][skb2][fclone_ref] */
929struct sk_buff_fclones {
930 struct sk_buff skb1;
931
932 struct sk_buff skb2;
933
934 atomic_t fclone_ref;
935};
936
937/**
938 * skb_fclone_busy - check if fclone is busy
939 * @skb: buffer
940 *
bda13fed 941 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
942 * Some drivers call skb_orphan() in their ndo_start_xmit(),
943 * so we also check that this didnt happen.
d0bf4a9e 944 */
39bb5e62
ED
945static inline bool skb_fclone_busy(const struct sock *sk,
946 const struct sk_buff *skb)
d0bf4a9e
ED
947{
948 const struct sk_buff_fclones *fclones;
949
950 fclones = container_of(skb, struct sk_buff_fclones, skb1);
951
952 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 953 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 954 fclones->skb2.sk == sk;
d0bf4a9e
ED
955}
956
d179cd12 957static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 958 gfp_t priority)
d179cd12 959{
c93bdd0e 960 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
961}
962
7965bd4d 963struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
964static inline struct sk_buff *alloc_skb_head(gfp_t priority)
965{
966 return __alloc_skb_head(priority, -1);
967}
968
7965bd4d
JP
969struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
970int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
971struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
972struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
973struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
974 gfp_t gfp_mask, bool fclone);
975static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
976 gfp_t gfp_mask)
977{
978 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
979}
7965bd4d
JP
980
981int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
982struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
983 unsigned int headroom);
984struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
985 int newtailroom, gfp_t priority);
25a91d8d
FD
986int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
987 int offset, int len);
7965bd4d
JP
988int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
989 int len);
990int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
991int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 992#define dev_kfree_skb(a) consume_skb(a)
1da177e4 993
7965bd4d
JP
994int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
995 int getfrag(void *from, char *to, int offset,
996 int len, int odd, struct sk_buff *skb),
997 void *from, int length);
e89e9cf5 998
be12a1fe
HFS
999int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1000 int offset, size_t size);
1001
d94d9fee 1002struct skb_seq_state {
677e90ed
TG
1003 __u32 lower_offset;
1004 __u32 upper_offset;
1005 __u32 frag_idx;
1006 __u32 stepped_offset;
1007 struct sk_buff *root_skb;
1008 struct sk_buff *cur_skb;
1009 __u8 *frag_data;
1010};
1011
7965bd4d
JP
1012void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1013 unsigned int to, struct skb_seq_state *st);
1014unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1015 struct skb_seq_state *st);
1016void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1017
7965bd4d 1018unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1019 unsigned int to, struct ts_config *config);
3fc7e8a6 1020
09323cc4
TH
1021/*
1022 * Packet hash types specify the type of hash in skb_set_hash.
1023 *
1024 * Hash types refer to the protocol layer addresses which are used to
1025 * construct a packet's hash. The hashes are used to differentiate or identify
1026 * flows of the protocol layer for the hash type. Hash types are either
1027 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1028 *
1029 * Properties of hashes:
1030 *
1031 * 1) Two packets in different flows have different hash values
1032 * 2) Two packets in the same flow should have the same hash value
1033 *
1034 * A hash at a higher layer is considered to be more specific. A driver should
1035 * set the most specific hash possible.
1036 *
1037 * A driver cannot indicate a more specific hash than the layer at which a hash
1038 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1039 *
1040 * A driver may indicate a hash level which is less specific than the
1041 * actual layer the hash was computed on. For instance, a hash computed
1042 * at L4 may be considered an L3 hash. This should only be done if the
1043 * driver can't unambiguously determine that the HW computed the hash at
1044 * the higher layer. Note that the "should" in the second property above
1045 * permits this.
1046 */
1047enum pkt_hash_types {
1048 PKT_HASH_TYPE_NONE, /* Undefined type */
1049 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1050 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1051 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1052};
1053
bcc83839 1054static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1055{
bcc83839 1056 skb->hash = 0;
a3b18ddb 1057 skb->sw_hash = 0;
bcc83839
TH
1058 skb->l4_hash = 0;
1059}
1060
1061static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1062{
1063 if (!skb->l4_hash)
1064 skb_clear_hash(skb);
1065}
1066
1067static inline void
1068__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1069{
1070 skb->l4_hash = is_l4;
1071 skb->sw_hash = is_sw;
61b905da 1072 skb->hash = hash;
09323cc4
TH
1073}
1074
bcc83839
TH
1075static inline void
1076skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1077{
1078 /* Used by drivers to set hash from HW */
1079 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1080}
1081
1082static inline void
1083__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1084{
1085 __skb_set_hash(skb, hash, true, is_l4);
1086}
1087
e5276937 1088void __skb_get_hash(struct sk_buff *skb);
eb70db87 1089u32 __skb_get_hash_symmetric(struct sk_buff *skb);
e5276937
TH
1090u32 skb_get_poff(const struct sk_buff *skb);
1091u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1092 const struct flow_keys *keys, int hlen);
1093__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1094 void *data, int hlen_proto);
1095
1096static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1097 int thoff, u8 ip_proto)
1098{
1099 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1100}
1101
1102void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1103 const struct flow_dissector_key *key,
1104 unsigned int key_count);
1105
1106bool __skb_flow_dissect(const struct sk_buff *skb,
1107 struct flow_dissector *flow_dissector,
1108 void *target_container,
cd79a238
TH
1109 void *data, __be16 proto, int nhoff, int hlen,
1110 unsigned int flags);
e5276937
TH
1111
1112static inline bool skb_flow_dissect(const struct sk_buff *skb,
1113 struct flow_dissector *flow_dissector,
cd79a238 1114 void *target_container, unsigned int flags)
e5276937
TH
1115{
1116 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1117 NULL, 0, 0, 0, flags);
e5276937
TH
1118}
1119
1120static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1121 struct flow_keys *flow,
1122 unsigned int flags)
e5276937
TH
1123{
1124 memset(flow, 0, sizeof(*flow));
1125 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1126 NULL, 0, 0, 0, flags);
e5276937
TH
1127}
1128
1129static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1130 void *data, __be16 proto,
cd79a238
TH
1131 int nhoff, int hlen,
1132 unsigned int flags)
e5276937
TH
1133{
1134 memset(flow, 0, sizeof(*flow));
1135 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1136 data, proto, nhoff, hlen, flags);
e5276937
TH
1137}
1138
3958afa1 1139static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1140{
a3b18ddb 1141 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1142 __skb_get_hash(skb);
bfb564e7 1143
61b905da 1144 return skb->hash;
bfb564e7
KK
1145}
1146
20a17bf6 1147__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1148
20a17bf6 1149static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1150{
c6cc1ca7
TH
1151 if (!skb->l4_hash && !skb->sw_hash) {
1152 struct flow_keys keys;
de4c1f8b 1153 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1154
de4c1f8b 1155 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1156 }
f70ea018
TH
1157
1158 return skb->hash;
1159}
1160
20a17bf6 1161__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1162
20a17bf6 1163static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1164{
c6cc1ca7
TH
1165 if (!skb->l4_hash && !skb->sw_hash) {
1166 struct flow_keys keys;
de4c1f8b 1167 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1168
de4c1f8b 1169 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1170 }
f70ea018
TH
1171
1172 return skb->hash;
1173}
1174
50fb7992
TH
1175__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1176
57bdf7f4
TH
1177static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1178{
61b905da 1179 return skb->hash;
57bdf7f4
TH
1180}
1181
3df7a74e
TH
1182static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1183{
61b905da 1184 to->hash = from->hash;
a3b18ddb 1185 to->sw_hash = from->sw_hash;
61b905da 1186 to->l4_hash = from->l4_hash;
3df7a74e
TH
1187};
1188
4305b541
ACM
1189#ifdef NET_SKBUFF_DATA_USES_OFFSET
1190static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1191{
1192 return skb->head + skb->end;
1193}
ec47ea82
AD
1194
1195static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1196{
1197 return skb->end;
1198}
4305b541
ACM
1199#else
1200static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1201{
1202 return skb->end;
1203}
ec47ea82
AD
1204
1205static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1206{
1207 return skb->end - skb->head;
1208}
4305b541
ACM
1209#endif
1210
1da177e4 1211/* Internal */
4305b541 1212#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1213
ac45f602
PO
1214static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1215{
1216 return &skb_shinfo(skb)->hwtstamps;
1217}
1218
1da177e4
LT
1219/**
1220 * skb_queue_empty - check if a queue is empty
1221 * @list: queue head
1222 *
1223 * Returns true if the queue is empty, false otherwise.
1224 */
1225static inline int skb_queue_empty(const struct sk_buff_head *list)
1226{
fd44b93c 1227 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1228}
1229
fc7ebb21
DM
1230/**
1231 * skb_queue_is_last - check if skb is the last entry in the queue
1232 * @list: queue head
1233 * @skb: buffer
1234 *
1235 * Returns true if @skb is the last buffer on the list.
1236 */
1237static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1238 const struct sk_buff *skb)
1239{
fd44b93c 1240 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1241}
1242
832d11c5
IJ
1243/**
1244 * skb_queue_is_first - check if skb is the first entry in the queue
1245 * @list: queue head
1246 * @skb: buffer
1247 *
1248 * Returns true if @skb is the first buffer on the list.
1249 */
1250static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1251 const struct sk_buff *skb)
1252{
fd44b93c 1253 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1254}
1255
249c8b42
DM
1256/**
1257 * skb_queue_next - return the next packet in the queue
1258 * @list: queue head
1259 * @skb: current buffer
1260 *
1261 * Return the next packet in @list after @skb. It is only valid to
1262 * call this if skb_queue_is_last() evaluates to false.
1263 */
1264static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1265 const struct sk_buff *skb)
1266{
1267 /* This BUG_ON may seem severe, but if we just return then we
1268 * are going to dereference garbage.
1269 */
1270 BUG_ON(skb_queue_is_last(list, skb));
1271 return skb->next;
1272}
1273
832d11c5
IJ
1274/**
1275 * skb_queue_prev - return the prev packet in the queue
1276 * @list: queue head
1277 * @skb: current buffer
1278 *
1279 * Return the prev packet in @list before @skb. It is only valid to
1280 * call this if skb_queue_is_first() evaluates to false.
1281 */
1282static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1283 const struct sk_buff *skb)
1284{
1285 /* This BUG_ON may seem severe, but if we just return then we
1286 * are going to dereference garbage.
1287 */
1288 BUG_ON(skb_queue_is_first(list, skb));
1289 return skb->prev;
1290}
1291
1da177e4
LT
1292/**
1293 * skb_get - reference buffer
1294 * @skb: buffer to reference
1295 *
1296 * Makes another reference to a socket buffer and returns a pointer
1297 * to the buffer.
1298 */
1299static inline struct sk_buff *skb_get(struct sk_buff *skb)
1300{
1301 atomic_inc(&skb->users);
1302 return skb;
1303}
1304
1305/*
1306 * If users == 1, we are the only owner and are can avoid redundant
1307 * atomic change.
1308 */
1309
1da177e4
LT
1310/**
1311 * skb_cloned - is the buffer a clone
1312 * @skb: buffer to check
1313 *
1314 * Returns true if the buffer was generated with skb_clone() and is
1315 * one of multiple shared copies of the buffer. Cloned buffers are
1316 * shared data so must not be written to under normal circumstances.
1317 */
1318static inline int skb_cloned(const struct sk_buff *skb)
1319{
1320 return skb->cloned &&
1321 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1322}
1323
14bbd6a5
PS
1324static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1325{
d0164adc 1326 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1327
1328 if (skb_cloned(skb))
1329 return pskb_expand_head(skb, 0, 0, pri);
1330
1331 return 0;
1332}
1333
1da177e4
LT
1334/**
1335 * skb_header_cloned - is the header a clone
1336 * @skb: buffer to check
1337 *
1338 * Returns true if modifying the header part of the buffer requires
1339 * the data to be copied.
1340 */
1341static inline int skb_header_cloned(const struct sk_buff *skb)
1342{
1343 int dataref;
1344
1345 if (!skb->cloned)
1346 return 0;
1347
1348 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1349 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1350 return dataref != 1;
1351}
1352
9580bf2e
ED
1353static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1354{
1355 might_sleep_if(gfpflags_allow_blocking(pri));
1356
1357 if (skb_header_cloned(skb))
1358 return pskb_expand_head(skb, 0, 0, pri);
1359
1360 return 0;
1361}
1362
1da177e4
LT
1363/**
1364 * skb_header_release - release reference to header
1365 * @skb: buffer to operate on
1366 *
1367 * Drop a reference to the header part of the buffer. This is done
1368 * by acquiring a payload reference. You must not read from the header
1369 * part of skb->data after this.
f4a775d1 1370 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1371 */
1372static inline void skb_header_release(struct sk_buff *skb)
1373{
1374 BUG_ON(skb->nohdr);
1375 skb->nohdr = 1;
1376 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1377}
1378
f4a775d1
ED
1379/**
1380 * __skb_header_release - release reference to header
1381 * @skb: buffer to operate on
1382 *
1383 * Variant of skb_header_release() assuming skb is private to caller.
1384 * We can avoid one atomic operation.
1385 */
1386static inline void __skb_header_release(struct sk_buff *skb)
1387{
1388 skb->nohdr = 1;
1389 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1390}
1391
1392
1da177e4
LT
1393/**
1394 * skb_shared - is the buffer shared
1395 * @skb: buffer to check
1396 *
1397 * Returns true if more than one person has a reference to this
1398 * buffer.
1399 */
1400static inline int skb_shared(const struct sk_buff *skb)
1401{
1402 return atomic_read(&skb->users) != 1;
1403}
1404
1405/**
1406 * skb_share_check - check if buffer is shared and if so clone it
1407 * @skb: buffer to check
1408 * @pri: priority for memory allocation
1409 *
1410 * If the buffer is shared the buffer is cloned and the old copy
1411 * drops a reference. A new clone with a single reference is returned.
1412 * If the buffer is not shared the original buffer is returned. When
1413 * being called from interrupt status or with spinlocks held pri must
1414 * be GFP_ATOMIC.
1415 *
1416 * NULL is returned on a memory allocation failure.
1417 */
47061bc4 1418static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1419{
d0164adc 1420 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1421 if (skb_shared(skb)) {
1422 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1423
1424 if (likely(nskb))
1425 consume_skb(skb);
1426 else
1427 kfree_skb(skb);
1da177e4
LT
1428 skb = nskb;
1429 }
1430 return skb;
1431}
1432
1433/*
1434 * Copy shared buffers into a new sk_buff. We effectively do COW on
1435 * packets to handle cases where we have a local reader and forward
1436 * and a couple of other messy ones. The normal one is tcpdumping
1437 * a packet thats being forwarded.
1438 */
1439
1440/**
1441 * skb_unshare - make a copy of a shared buffer
1442 * @skb: buffer to check
1443 * @pri: priority for memory allocation
1444 *
1445 * If the socket buffer is a clone then this function creates a new
1446 * copy of the data, drops a reference count on the old copy and returns
1447 * the new copy with the reference count at 1. If the buffer is not a clone
1448 * the original buffer is returned. When called with a spinlock held or
1449 * from interrupt state @pri must be %GFP_ATOMIC
1450 *
1451 * %NULL is returned on a memory allocation failure.
1452 */
e2bf521d 1453static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1454 gfp_t pri)
1da177e4 1455{
d0164adc 1456 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1457 if (skb_cloned(skb)) {
1458 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1459
1460 /* Free our shared copy */
1461 if (likely(nskb))
1462 consume_skb(skb);
1463 else
1464 kfree_skb(skb);
1da177e4
LT
1465 skb = nskb;
1466 }
1467 return skb;
1468}
1469
1470/**
1a5778aa 1471 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1472 * @list_: list to peek at
1473 *
1474 * Peek an &sk_buff. Unlike most other operations you _MUST_
1475 * be careful with this one. A peek leaves the buffer on the
1476 * list and someone else may run off with it. You must hold
1477 * the appropriate locks or have a private queue to do this.
1478 *
1479 * Returns %NULL for an empty list or a pointer to the head element.
1480 * The reference count is not incremented and the reference is therefore
1481 * volatile. Use with caution.
1482 */
05bdd2f1 1483static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1484{
18d07000
ED
1485 struct sk_buff *skb = list_->next;
1486
1487 if (skb == (struct sk_buff *)list_)
1488 skb = NULL;
1489 return skb;
1da177e4
LT
1490}
1491
da5ef6e5
PE
1492/**
1493 * skb_peek_next - peek skb following the given one from a queue
1494 * @skb: skb to start from
1495 * @list_: list to peek at
1496 *
1497 * Returns %NULL when the end of the list is met or a pointer to the
1498 * next element. The reference count is not incremented and the
1499 * reference is therefore volatile. Use with caution.
1500 */
1501static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1502 const struct sk_buff_head *list_)
1503{
1504 struct sk_buff *next = skb->next;
18d07000 1505
da5ef6e5
PE
1506 if (next == (struct sk_buff *)list_)
1507 next = NULL;
1508 return next;
1509}
1510
1da177e4 1511/**
1a5778aa 1512 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1513 * @list_: list to peek at
1514 *
1515 * Peek an &sk_buff. Unlike most other operations you _MUST_
1516 * be careful with this one. A peek leaves the buffer on the
1517 * list and someone else may run off with it. You must hold
1518 * the appropriate locks or have a private queue to do this.
1519 *
1520 * Returns %NULL for an empty list or a pointer to the tail element.
1521 * The reference count is not incremented and the reference is therefore
1522 * volatile. Use with caution.
1523 */
05bdd2f1 1524static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1525{
18d07000
ED
1526 struct sk_buff *skb = list_->prev;
1527
1528 if (skb == (struct sk_buff *)list_)
1529 skb = NULL;
1530 return skb;
1531
1da177e4
LT
1532}
1533
1534/**
1535 * skb_queue_len - get queue length
1536 * @list_: list to measure
1537 *
1538 * Return the length of an &sk_buff queue.
1539 */
1540static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1541{
1542 return list_->qlen;
1543}
1544
67fed459
DM
1545/**
1546 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1547 * @list: queue to initialize
1548 *
1549 * This initializes only the list and queue length aspects of
1550 * an sk_buff_head object. This allows to initialize the list
1551 * aspects of an sk_buff_head without reinitializing things like
1552 * the spinlock. It can also be used for on-stack sk_buff_head
1553 * objects where the spinlock is known to not be used.
1554 */
1555static inline void __skb_queue_head_init(struct sk_buff_head *list)
1556{
1557 list->prev = list->next = (struct sk_buff *)list;
1558 list->qlen = 0;
1559}
1560
76f10ad0
AV
1561/*
1562 * This function creates a split out lock class for each invocation;
1563 * this is needed for now since a whole lot of users of the skb-queue
1564 * infrastructure in drivers have different locking usage (in hardirq)
1565 * than the networking core (in softirq only). In the long run either the
1566 * network layer or drivers should need annotation to consolidate the
1567 * main types of usage into 3 classes.
1568 */
1da177e4
LT
1569static inline void skb_queue_head_init(struct sk_buff_head *list)
1570{
1571 spin_lock_init(&list->lock);
67fed459 1572 __skb_queue_head_init(list);
1da177e4
LT
1573}
1574
c2ecba71
PE
1575static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1576 struct lock_class_key *class)
1577{
1578 skb_queue_head_init(list);
1579 lockdep_set_class(&list->lock, class);
1580}
1581
1da177e4 1582/*
bf299275 1583 * Insert an sk_buff on a list.
1da177e4
LT
1584 *
1585 * The "__skb_xxxx()" functions are the non-atomic ones that
1586 * can only be called with interrupts disabled.
1587 */
7965bd4d
JP
1588void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1589 struct sk_buff_head *list);
bf299275
GR
1590static inline void __skb_insert(struct sk_buff *newsk,
1591 struct sk_buff *prev, struct sk_buff *next,
1592 struct sk_buff_head *list)
1593{
1594 newsk->next = next;
1595 newsk->prev = prev;
1596 next->prev = prev->next = newsk;
1597 list->qlen++;
1598}
1da177e4 1599
67fed459
DM
1600static inline void __skb_queue_splice(const struct sk_buff_head *list,
1601 struct sk_buff *prev,
1602 struct sk_buff *next)
1603{
1604 struct sk_buff *first = list->next;
1605 struct sk_buff *last = list->prev;
1606
1607 first->prev = prev;
1608 prev->next = first;
1609
1610 last->next = next;
1611 next->prev = last;
1612}
1613
1614/**
1615 * skb_queue_splice - join two skb lists, this is designed for stacks
1616 * @list: the new list to add
1617 * @head: the place to add it in the first list
1618 */
1619static inline void skb_queue_splice(const struct sk_buff_head *list,
1620 struct sk_buff_head *head)
1621{
1622 if (!skb_queue_empty(list)) {
1623 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1624 head->qlen += list->qlen;
67fed459
DM
1625 }
1626}
1627
1628/**
d9619496 1629 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1630 * @list: the new list to add
1631 * @head: the place to add it in the first list
1632 *
1633 * The list at @list is reinitialised
1634 */
1635static inline void skb_queue_splice_init(struct sk_buff_head *list,
1636 struct sk_buff_head *head)
1637{
1638 if (!skb_queue_empty(list)) {
1639 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1640 head->qlen += list->qlen;
67fed459
DM
1641 __skb_queue_head_init(list);
1642 }
1643}
1644
1645/**
1646 * skb_queue_splice_tail - join two skb lists, each list being a queue
1647 * @list: the new list to add
1648 * @head: the place to add it in the first list
1649 */
1650static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1651 struct sk_buff_head *head)
1652{
1653 if (!skb_queue_empty(list)) {
1654 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1655 head->qlen += list->qlen;
67fed459
DM
1656 }
1657}
1658
1659/**
d9619496 1660 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1661 * @list: the new list to add
1662 * @head: the place to add it in the first list
1663 *
1664 * Each of the lists is a queue.
1665 * The list at @list is reinitialised
1666 */
1667static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1668 struct sk_buff_head *head)
1669{
1670 if (!skb_queue_empty(list)) {
1671 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1672 head->qlen += list->qlen;
67fed459
DM
1673 __skb_queue_head_init(list);
1674 }
1675}
1676
1da177e4 1677/**
300ce174 1678 * __skb_queue_after - queue a buffer at the list head
1da177e4 1679 * @list: list to use
300ce174 1680 * @prev: place after this buffer
1da177e4
LT
1681 * @newsk: buffer to queue
1682 *
300ce174 1683 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1684 * and you must therefore hold required locks before calling it.
1685 *
1686 * A buffer cannot be placed on two lists at the same time.
1687 */
300ce174
SH
1688static inline void __skb_queue_after(struct sk_buff_head *list,
1689 struct sk_buff *prev,
1690 struct sk_buff *newsk)
1da177e4 1691{
bf299275 1692 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1693}
1694
7965bd4d
JP
1695void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1696 struct sk_buff_head *list);
7de6c033 1697
f5572855
GR
1698static inline void __skb_queue_before(struct sk_buff_head *list,
1699 struct sk_buff *next,
1700 struct sk_buff *newsk)
1701{
1702 __skb_insert(newsk, next->prev, next, list);
1703}
1704
300ce174
SH
1705/**
1706 * __skb_queue_head - queue a buffer at the list head
1707 * @list: list to use
1708 * @newsk: buffer to queue
1709 *
1710 * Queue a buffer at the start of a list. This function takes no locks
1711 * and you must therefore hold required locks before calling it.
1712 *
1713 * A buffer cannot be placed on two lists at the same time.
1714 */
7965bd4d 1715void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1716static inline void __skb_queue_head(struct sk_buff_head *list,
1717 struct sk_buff *newsk)
1718{
1719 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1720}
1721
1da177e4
LT
1722/**
1723 * __skb_queue_tail - queue a buffer at the list tail
1724 * @list: list to use
1725 * @newsk: buffer to queue
1726 *
1727 * Queue a buffer at the end of a list. This function takes no locks
1728 * and you must therefore hold required locks before calling it.
1729 *
1730 * A buffer cannot be placed on two lists at the same time.
1731 */
7965bd4d 1732void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1733static inline void __skb_queue_tail(struct sk_buff_head *list,
1734 struct sk_buff *newsk)
1735{
f5572855 1736 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1737}
1738
1da177e4
LT
1739/*
1740 * remove sk_buff from list. _Must_ be called atomically, and with
1741 * the list known..
1742 */
7965bd4d 1743void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1744static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1745{
1746 struct sk_buff *next, *prev;
1747
1748 list->qlen--;
1749 next = skb->next;
1750 prev = skb->prev;
1751 skb->next = skb->prev = NULL;
1da177e4
LT
1752 next->prev = prev;
1753 prev->next = next;
1754}
1755
f525c06d
GR
1756/**
1757 * __skb_dequeue - remove from the head of the queue
1758 * @list: list to dequeue from
1759 *
1760 * Remove the head of the list. This function does not take any locks
1761 * so must be used with appropriate locks held only. The head item is
1762 * returned or %NULL if the list is empty.
1763 */
7965bd4d 1764struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1765static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1766{
1767 struct sk_buff *skb = skb_peek(list);
1768 if (skb)
1769 __skb_unlink(skb, list);
1770 return skb;
1771}
1da177e4
LT
1772
1773/**
1774 * __skb_dequeue_tail - remove from the tail of the queue
1775 * @list: list to dequeue from
1776 *
1777 * Remove the tail of the list. This function does not take any locks
1778 * so must be used with appropriate locks held only. The tail item is
1779 * returned or %NULL if the list is empty.
1780 */
7965bd4d 1781struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1782static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1783{
1784 struct sk_buff *skb = skb_peek_tail(list);
1785 if (skb)
1786 __skb_unlink(skb, list);
1787 return skb;
1788}
1789
1790
bdcc0924 1791static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1792{
1793 return skb->data_len;
1794}
1795
1796static inline unsigned int skb_headlen(const struct sk_buff *skb)
1797{
1798 return skb->len - skb->data_len;
1799}
1800
1801static inline int skb_pagelen(const struct sk_buff *skb)
1802{
1803 int i, len = 0;
1804
1805 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1806 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1807 return len + skb_headlen(skb);
1808}
1809
131ea667
IC
1810/**
1811 * __skb_fill_page_desc - initialise a paged fragment in an skb
1812 * @skb: buffer containing fragment to be initialised
1813 * @i: paged fragment index to initialise
1814 * @page: the page to use for this fragment
1815 * @off: the offset to the data with @page
1816 * @size: the length of the data
1817 *
1818 * Initialises the @i'th fragment of @skb to point to &size bytes at
1819 * offset @off within @page.
1820 *
1821 * Does not take any additional reference on the fragment.
1822 */
1823static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1824 struct page *page, int off, int size)
1da177e4
LT
1825{
1826 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1827
c48a11c7 1828 /*
2f064f34
MH
1829 * Propagate page pfmemalloc to the skb if we can. The problem is
1830 * that not all callers have unique ownership of the page but rely
1831 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1832 */
a8605c60 1833 frag->page.p = page;
1da177e4 1834 frag->page_offset = off;
9e903e08 1835 skb_frag_size_set(frag, size);
cca7af38
PE
1836
1837 page = compound_head(page);
2f064f34 1838 if (page_is_pfmemalloc(page))
cca7af38 1839 skb->pfmemalloc = true;
131ea667
IC
1840}
1841
1842/**
1843 * skb_fill_page_desc - initialise a paged fragment in an skb
1844 * @skb: buffer containing fragment to be initialised
1845 * @i: paged fragment index to initialise
1846 * @page: the page to use for this fragment
1847 * @off: the offset to the data with @page
1848 * @size: the length of the data
1849 *
1850 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1851 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1852 * addition updates @skb such that @i is the last fragment.
1853 *
1854 * Does not take any additional reference on the fragment.
1855 */
1856static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1857 struct page *page, int off, int size)
1858{
1859 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1860 skb_shinfo(skb)->nr_frags = i + 1;
1861}
1862
7965bd4d
JP
1863void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1864 int size, unsigned int truesize);
654bed16 1865
f8e617e1
JW
1866void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1867 unsigned int truesize);
1868
1da177e4 1869#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1870#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1871#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1872
27a884dc
ACM
1873#ifdef NET_SKBUFF_DATA_USES_OFFSET
1874static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1875{
1876 return skb->head + skb->tail;
1877}
1878
1879static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1880{
1881 skb->tail = skb->data - skb->head;
1882}
1883
1884static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1885{
1886 skb_reset_tail_pointer(skb);
1887 skb->tail += offset;
1888}
7cc46190 1889
27a884dc
ACM
1890#else /* NET_SKBUFF_DATA_USES_OFFSET */
1891static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1892{
1893 return skb->tail;
1894}
1895
1896static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1897{
1898 skb->tail = skb->data;
1899}
1900
1901static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1902{
1903 skb->tail = skb->data + offset;
1904}
4305b541 1905
27a884dc
ACM
1906#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1907
1da177e4
LT
1908/*
1909 * Add data to an sk_buff
1910 */
0c7ddf36 1911unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1912unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1913static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1914{
27a884dc 1915 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1916 SKB_LINEAR_ASSERT(skb);
1917 skb->tail += len;
1918 skb->len += len;
1919 return tmp;
1920}
1921
7965bd4d 1922unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1923static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1924{
1925 skb->data -= len;
1926 skb->len += len;
1927 return skb->data;
1928}
1929
7965bd4d 1930unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1931static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1932{
1933 skb->len -= len;
1934 BUG_ON(skb->len < skb->data_len);
1935 return skb->data += len;
1936}
1937
47d29646
DM
1938static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1939{
1940 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1941}
1942
7965bd4d 1943unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1944
1945static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1946{
1947 if (len > skb_headlen(skb) &&
987c402a 1948 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1949 return NULL;
1950 skb->len -= len;
1951 return skb->data += len;
1952}
1953
1954static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1955{
1956 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1957}
1958
1959static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1960{
1961 if (likely(len <= skb_headlen(skb)))
1962 return 1;
1963 if (unlikely(len > skb->len))
1964 return 0;
987c402a 1965 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1966}
1967
1968/**
1969 * skb_headroom - bytes at buffer head
1970 * @skb: buffer to check
1971 *
1972 * Return the number of bytes of free space at the head of an &sk_buff.
1973 */
c2636b4d 1974static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1975{
1976 return skb->data - skb->head;
1977}
1978
1979/**
1980 * skb_tailroom - bytes at buffer end
1981 * @skb: buffer to check
1982 *
1983 * Return the number of bytes of free space at the tail of an sk_buff
1984 */
1985static inline int skb_tailroom(const struct sk_buff *skb)
1986{
4305b541 1987 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1988}
1989
a21d4572
ED
1990/**
1991 * skb_availroom - bytes at buffer end
1992 * @skb: buffer to check
1993 *
1994 * Return the number of bytes of free space at the tail of an sk_buff
1995 * allocated by sk_stream_alloc()
1996 */
1997static inline int skb_availroom(const struct sk_buff *skb)
1998{
16fad69c
ED
1999 if (skb_is_nonlinear(skb))
2000 return 0;
2001
2002 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2003}
2004
1da177e4
LT
2005/**
2006 * skb_reserve - adjust headroom
2007 * @skb: buffer to alter
2008 * @len: bytes to move
2009 *
2010 * Increase the headroom of an empty &sk_buff by reducing the tail
2011 * room. This is only allowed for an empty buffer.
2012 */
8243126c 2013static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2014{
2015 skb->data += len;
2016 skb->tail += len;
2017}
2018
1837b2e2
BP
2019/**
2020 * skb_tailroom_reserve - adjust reserved_tailroom
2021 * @skb: buffer to alter
2022 * @mtu: maximum amount of headlen permitted
2023 * @needed_tailroom: minimum amount of reserved_tailroom
2024 *
2025 * Set reserved_tailroom so that headlen can be as large as possible but
2026 * not larger than mtu and tailroom cannot be smaller than
2027 * needed_tailroom.
2028 * The required headroom should already have been reserved before using
2029 * this function.
2030 */
2031static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2032 unsigned int needed_tailroom)
2033{
2034 SKB_LINEAR_ASSERT(skb);
2035 if (mtu < skb_tailroom(skb) - needed_tailroom)
2036 /* use at most mtu */
2037 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2038 else
2039 /* use up to all available space */
2040 skb->reserved_tailroom = needed_tailroom;
2041}
2042
8bce6d7d
TH
2043#define ENCAP_TYPE_ETHER 0
2044#define ENCAP_TYPE_IPPROTO 1
2045
2046static inline void skb_set_inner_protocol(struct sk_buff *skb,
2047 __be16 protocol)
2048{
2049 skb->inner_protocol = protocol;
2050 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2051}
2052
2053static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2054 __u8 ipproto)
2055{
2056 skb->inner_ipproto = ipproto;
2057 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2058}
2059
6a674e9c
JG
2060static inline void skb_reset_inner_headers(struct sk_buff *skb)
2061{
aefbd2b3 2062 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2063 skb->inner_network_header = skb->network_header;
2064 skb->inner_transport_header = skb->transport_header;
2065}
2066
0b5c9db1
JP
2067static inline void skb_reset_mac_len(struct sk_buff *skb)
2068{
2069 skb->mac_len = skb->network_header - skb->mac_header;
2070}
2071
6a674e9c
JG
2072static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2073 *skb)
2074{
2075 return skb->head + skb->inner_transport_header;
2076}
2077
55dc5a9f
TH
2078static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2079{
2080 return skb_inner_transport_header(skb) - skb->data;
2081}
2082
6a674e9c
JG
2083static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2084{
2085 skb->inner_transport_header = skb->data - skb->head;
2086}
2087
2088static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2089 const int offset)
2090{
2091 skb_reset_inner_transport_header(skb);
2092 skb->inner_transport_header += offset;
2093}
2094
2095static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2096{
2097 return skb->head + skb->inner_network_header;
2098}
2099
2100static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2101{
2102 skb->inner_network_header = skb->data - skb->head;
2103}
2104
2105static inline void skb_set_inner_network_header(struct sk_buff *skb,
2106 const int offset)
2107{
2108 skb_reset_inner_network_header(skb);
2109 skb->inner_network_header += offset;
2110}
2111
aefbd2b3
PS
2112static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2113{
2114 return skb->head + skb->inner_mac_header;
2115}
2116
2117static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2118{
2119 skb->inner_mac_header = skb->data - skb->head;
2120}
2121
2122static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2123 const int offset)
2124{
2125 skb_reset_inner_mac_header(skb);
2126 skb->inner_mac_header += offset;
2127}
fda55eca
ED
2128static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2129{
35d04610 2130 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2131}
2132
9c70220b
ACM
2133static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2134{
2e07fa9c 2135 return skb->head + skb->transport_header;
9c70220b
ACM
2136}
2137
badff6d0
ACM
2138static inline void skb_reset_transport_header(struct sk_buff *skb)
2139{
2e07fa9c 2140 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2141}
2142
967b05f6
ACM
2143static inline void skb_set_transport_header(struct sk_buff *skb,
2144 const int offset)
2145{
2e07fa9c
ACM
2146 skb_reset_transport_header(skb);
2147 skb->transport_header += offset;
ea2ae17d
ACM
2148}
2149
d56f90a7
ACM
2150static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2151{
2e07fa9c 2152 return skb->head + skb->network_header;
d56f90a7
ACM
2153}
2154
c1d2bbe1
ACM
2155static inline void skb_reset_network_header(struct sk_buff *skb)
2156{
2e07fa9c 2157 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2158}
2159
c14d2450
ACM
2160static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2161{
2e07fa9c
ACM
2162 skb_reset_network_header(skb);
2163 skb->network_header += offset;
c14d2450
ACM
2164}
2165
2e07fa9c 2166static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2167{
2e07fa9c 2168 return skb->head + skb->mac_header;
bbe735e4
ACM
2169}
2170
2e07fa9c 2171static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2172{
35d04610 2173 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2174}
2175
2176static inline void skb_reset_mac_header(struct sk_buff *skb)
2177{
2178 skb->mac_header = skb->data - skb->head;
2179}
2180
2181static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2182{
2183 skb_reset_mac_header(skb);
2184 skb->mac_header += offset;
2185}
2186
0e3da5bb
TT
2187static inline void skb_pop_mac_header(struct sk_buff *skb)
2188{
2189 skb->mac_header = skb->network_header;
2190}
2191
fbbdb8f0
YX
2192static inline void skb_probe_transport_header(struct sk_buff *skb,
2193 const int offset_hint)
2194{
2195 struct flow_keys keys;
2196
2197 if (skb_transport_header_was_set(skb))
2198 return;
cd79a238 2199 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2200 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2201 else
2202 skb_set_transport_header(skb, offset_hint);
2203}
2204
03606895
ED
2205static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2206{
2207 if (skb_mac_header_was_set(skb)) {
2208 const unsigned char *old_mac = skb_mac_header(skb);
2209
2210 skb_set_mac_header(skb, -skb->mac_len);
2211 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2212 }
2213}
2214
04fb451e
MM
2215static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2216{
2217 return skb->csum_start - skb_headroom(skb);
2218}
2219
08b64fcc
AD
2220static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2221{
2222 return skb->head + skb->csum_start;
2223}
2224
2e07fa9c
ACM
2225static inline int skb_transport_offset(const struct sk_buff *skb)
2226{
2227 return skb_transport_header(skb) - skb->data;
2228}
2229
2230static inline u32 skb_network_header_len(const struct sk_buff *skb)
2231{
2232 return skb->transport_header - skb->network_header;
2233}
2234
6a674e9c
JG
2235static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2236{
2237 return skb->inner_transport_header - skb->inner_network_header;
2238}
2239
2e07fa9c
ACM
2240static inline int skb_network_offset(const struct sk_buff *skb)
2241{
2242 return skb_network_header(skb) - skb->data;
2243}
48d49d0c 2244
6a674e9c
JG
2245static inline int skb_inner_network_offset(const struct sk_buff *skb)
2246{
2247 return skb_inner_network_header(skb) - skb->data;
2248}
2249
f9599ce1
CG
2250static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2251{
2252 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2253}
2254
1da177e4
LT
2255/*
2256 * CPUs often take a performance hit when accessing unaligned memory
2257 * locations. The actual performance hit varies, it can be small if the
2258 * hardware handles it or large if we have to take an exception and fix it
2259 * in software.
2260 *
2261 * Since an ethernet header is 14 bytes network drivers often end up with
2262 * the IP header at an unaligned offset. The IP header can be aligned by
2263 * shifting the start of the packet by 2 bytes. Drivers should do this
2264 * with:
2265 *
8660c124 2266 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2267 *
2268 * The downside to this alignment of the IP header is that the DMA is now
2269 * unaligned. On some architectures the cost of an unaligned DMA is high
2270 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2271 *
1da177e4
LT
2272 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2273 * to be overridden.
2274 */
2275#ifndef NET_IP_ALIGN
2276#define NET_IP_ALIGN 2
2277#endif
2278
025be81e
AB
2279/*
2280 * The networking layer reserves some headroom in skb data (via
2281 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2282 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2283 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2284 *
2285 * Unfortunately this headroom changes the DMA alignment of the resulting
2286 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2287 * on some architectures. An architecture can override this value,
2288 * perhaps setting it to a cacheline in size (since that will maintain
2289 * cacheline alignment of the DMA). It must be a power of 2.
2290 *
d6301d3d 2291 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2292 * headroom, you should not reduce this.
5933dd2f
ED
2293 *
2294 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2295 * to reduce average number of cache lines per packet.
2296 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2297 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2298 */
2299#ifndef NET_SKB_PAD
5933dd2f 2300#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2301#endif
2302
7965bd4d 2303int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2304
5293efe6 2305static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2306{
c4264f27 2307 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2308 WARN_ON(1);
2309 return;
2310 }
27a884dc
ACM
2311 skb->len = len;
2312 skb_set_tail_pointer(skb, len);
1da177e4
LT
2313}
2314
5293efe6
DB
2315static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2316{
2317 __skb_set_length(skb, len);
2318}
2319
7965bd4d 2320void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2321
2322static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2323{
3cc0e873
HX
2324 if (skb->data_len)
2325 return ___pskb_trim(skb, len);
2326 __skb_trim(skb, len);
2327 return 0;
1da177e4
LT
2328}
2329
2330static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2331{
2332 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2333}
2334
e9fa4f7b
HX
2335/**
2336 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2337 * @skb: buffer to alter
2338 * @len: new length
2339 *
2340 * This is identical to pskb_trim except that the caller knows that
2341 * the skb is not cloned so we should never get an error due to out-
2342 * of-memory.
2343 */
2344static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2345{
2346 int err = pskb_trim(skb, len);
2347 BUG_ON(err);
2348}
2349
5293efe6
DB
2350static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2351{
2352 unsigned int diff = len - skb->len;
2353
2354 if (skb_tailroom(skb) < diff) {
2355 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2356 GFP_ATOMIC);
2357 if (ret)
2358 return ret;
2359 }
2360 __skb_set_length(skb, len);
2361 return 0;
2362}
2363
1da177e4
LT
2364/**
2365 * skb_orphan - orphan a buffer
2366 * @skb: buffer to orphan
2367 *
2368 * If a buffer currently has an owner then we call the owner's
2369 * destructor function and make the @skb unowned. The buffer continues
2370 * to exist but is no longer charged to its former owner.
2371 */
2372static inline void skb_orphan(struct sk_buff *skb)
2373{
c34a7612 2374 if (skb->destructor) {
1da177e4 2375 skb->destructor(skb);
c34a7612
ED
2376 skb->destructor = NULL;
2377 skb->sk = NULL;
376c7311
ED
2378 } else {
2379 BUG_ON(skb->sk);
c34a7612 2380 }
1da177e4
LT
2381}
2382
a353e0ce
MT
2383/**
2384 * skb_orphan_frags - orphan the frags contained in a buffer
2385 * @skb: buffer to orphan frags from
2386 * @gfp_mask: allocation mask for replacement pages
2387 *
2388 * For each frag in the SKB which needs a destructor (i.e. has an
2389 * owner) create a copy of that frag and release the original
2390 * page by calling the destructor.
2391 */
2392static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2393{
2394 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2395 return 0;
2396 return skb_copy_ubufs(skb, gfp_mask);
2397}
2398
1da177e4
LT
2399/**
2400 * __skb_queue_purge - empty a list
2401 * @list: list to empty
2402 *
2403 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2404 * the list and one reference dropped. This function does not take the
2405 * list lock and the caller must hold the relevant locks to use it.
2406 */
7965bd4d 2407void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2408static inline void __skb_queue_purge(struct sk_buff_head *list)
2409{
2410 struct sk_buff *skb;
2411 while ((skb = __skb_dequeue(list)) != NULL)
2412 kfree_skb(skb);
2413}
2414
9f5afeae
YW
2415void skb_rbtree_purge(struct rb_root *root);
2416
7965bd4d 2417void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2418
7965bd4d
JP
2419struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2420 gfp_t gfp_mask);
8af27456
CH
2421
2422/**
2423 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2424 * @dev: network device to receive on
2425 * @length: length to allocate
2426 *
2427 * Allocate a new &sk_buff and assign it a usage count of one. The
2428 * buffer has unspecified headroom built in. Users should allocate
2429 * the headroom they think they need without accounting for the
2430 * built in space. The built in space is used for optimisations.
2431 *
2432 * %NULL is returned if there is no free memory. Although this function
2433 * allocates memory it can be called from an interrupt.
2434 */
2435static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2436 unsigned int length)
8af27456
CH
2437{
2438 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2439}
2440
6f532612
ED
2441/* legacy helper around __netdev_alloc_skb() */
2442static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2443 gfp_t gfp_mask)
2444{
2445 return __netdev_alloc_skb(NULL, length, gfp_mask);
2446}
2447
2448/* legacy helper around netdev_alloc_skb() */
2449static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2450{
2451 return netdev_alloc_skb(NULL, length);
2452}
2453
2454
4915a0de
ED
2455static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2456 unsigned int length, gfp_t gfp)
61321bbd 2457{
4915a0de 2458 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2459
2460 if (NET_IP_ALIGN && skb)
2461 skb_reserve(skb, NET_IP_ALIGN);
2462 return skb;
2463}
2464
4915a0de
ED
2465static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2466 unsigned int length)
2467{
2468 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2469}
2470
181edb2b
AD
2471static inline void skb_free_frag(void *addr)
2472{
2473 __free_page_frag(addr);
2474}
2475
ffde7328 2476void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2477struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2478 unsigned int length, gfp_t gfp_mask);
2479static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2480 unsigned int length)
2481{
2482 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2483}
795bb1c0
JDB
2484void napi_consume_skb(struct sk_buff *skb, int budget);
2485
2486void __kfree_skb_flush(void);
15fad714 2487void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2488
71dfda58
AD
2489/**
2490 * __dev_alloc_pages - allocate page for network Rx
2491 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2492 * @order: size of the allocation
2493 *
2494 * Allocate a new page.
2495 *
2496 * %NULL is returned if there is no free memory.
2497*/
2498static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2499 unsigned int order)
2500{
2501 /* This piece of code contains several assumptions.
2502 * 1. This is for device Rx, therefor a cold page is preferred.
2503 * 2. The expectation is the user wants a compound page.
2504 * 3. If requesting a order 0 page it will not be compound
2505 * due to the check to see if order has a value in prep_new_page
2506 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2507 * code in gfp_to_alloc_flags that should be enforcing this.
2508 */
2509 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2510
2511 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2512}
2513
2514static inline struct page *dev_alloc_pages(unsigned int order)
2515{
95829b3a 2516 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2517}
2518
2519/**
2520 * __dev_alloc_page - allocate a page for network Rx
2521 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2522 *
2523 * Allocate a new page.
2524 *
2525 * %NULL is returned if there is no free memory.
2526 */
2527static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2528{
2529 return __dev_alloc_pages(gfp_mask, 0);
2530}
2531
2532static inline struct page *dev_alloc_page(void)
2533{
95829b3a 2534 return dev_alloc_pages(0);
71dfda58
AD
2535}
2536
0614002b
MG
2537/**
2538 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2539 * @page: The page that was allocated from skb_alloc_page
2540 * @skb: The skb that may need pfmemalloc set
2541 */
2542static inline void skb_propagate_pfmemalloc(struct page *page,
2543 struct sk_buff *skb)
2544{
2f064f34 2545 if (page_is_pfmemalloc(page))
0614002b
MG
2546 skb->pfmemalloc = true;
2547}
2548
131ea667 2549/**
e227867f 2550 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2551 * @frag: the paged fragment
2552 *
2553 * Returns the &struct page associated with @frag.
2554 */
2555static inline struct page *skb_frag_page(const skb_frag_t *frag)
2556{
a8605c60 2557 return frag->page.p;
131ea667
IC
2558}
2559
2560/**
2561 * __skb_frag_ref - take an addition reference on a paged fragment.
2562 * @frag: the paged fragment
2563 *
2564 * Takes an additional reference on the paged fragment @frag.
2565 */
2566static inline void __skb_frag_ref(skb_frag_t *frag)
2567{
2568 get_page(skb_frag_page(frag));
2569}
2570
2571/**
2572 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2573 * @skb: the buffer
2574 * @f: the fragment offset.
2575 *
2576 * Takes an additional reference on the @f'th paged fragment of @skb.
2577 */
2578static inline void skb_frag_ref(struct sk_buff *skb, int f)
2579{
2580 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2581}
2582
2583/**
2584 * __skb_frag_unref - release a reference on a paged fragment.
2585 * @frag: the paged fragment
2586 *
2587 * Releases a reference on the paged fragment @frag.
2588 */
2589static inline void __skb_frag_unref(skb_frag_t *frag)
2590{
2591 put_page(skb_frag_page(frag));
2592}
2593
2594/**
2595 * skb_frag_unref - release a reference on a paged fragment of an skb.
2596 * @skb: the buffer
2597 * @f: the fragment offset
2598 *
2599 * Releases a reference on the @f'th paged fragment of @skb.
2600 */
2601static inline void skb_frag_unref(struct sk_buff *skb, int f)
2602{
2603 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2604}
2605
2606/**
2607 * skb_frag_address - gets the address of the data contained in a paged fragment
2608 * @frag: the paged fragment buffer
2609 *
2610 * Returns the address of the data within @frag. The page must already
2611 * be mapped.
2612 */
2613static inline void *skb_frag_address(const skb_frag_t *frag)
2614{
2615 return page_address(skb_frag_page(frag)) + frag->page_offset;
2616}
2617
2618/**
2619 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2620 * @frag: the paged fragment buffer
2621 *
2622 * Returns the address of the data within @frag. Checks that the page
2623 * is mapped and returns %NULL otherwise.
2624 */
2625static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2626{
2627 void *ptr = page_address(skb_frag_page(frag));
2628 if (unlikely(!ptr))
2629 return NULL;
2630
2631 return ptr + frag->page_offset;
2632}
2633
2634/**
2635 * __skb_frag_set_page - sets the page contained in a paged fragment
2636 * @frag: the paged fragment
2637 * @page: the page to set
2638 *
2639 * Sets the fragment @frag to contain @page.
2640 */
2641static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2642{
a8605c60 2643 frag->page.p = page;
131ea667
IC
2644}
2645
2646/**
2647 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2648 * @skb: the buffer
2649 * @f: the fragment offset
2650 * @page: the page to set
2651 *
2652 * Sets the @f'th fragment of @skb to contain @page.
2653 */
2654static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2655 struct page *page)
2656{
2657 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2658}
2659
400dfd3a
ED
2660bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2661
131ea667
IC
2662/**
2663 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2664 * @dev: the device to map the fragment to
131ea667
IC
2665 * @frag: the paged fragment to map
2666 * @offset: the offset within the fragment (starting at the
2667 * fragment's own offset)
2668 * @size: the number of bytes to map
f83347df 2669 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2670 *
2671 * Maps the page associated with @frag to @device.
2672 */
2673static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2674 const skb_frag_t *frag,
2675 size_t offset, size_t size,
2676 enum dma_data_direction dir)
2677{
2678 return dma_map_page(dev, skb_frag_page(frag),
2679 frag->page_offset + offset, size, dir);
2680}
2681
117632e6
ED
2682static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2683 gfp_t gfp_mask)
2684{
2685 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2686}
2687
bad93e9d
OP
2688
2689static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2690 gfp_t gfp_mask)
2691{
2692 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2693}
2694
2695
334a8132
PM
2696/**
2697 * skb_clone_writable - is the header of a clone writable
2698 * @skb: buffer to check
2699 * @len: length up to which to write
2700 *
2701 * Returns true if modifying the header part of the cloned buffer
2702 * does not requires the data to be copied.
2703 */
05bdd2f1 2704static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2705{
2706 return !skb_header_cloned(skb) &&
2707 skb_headroom(skb) + len <= skb->hdr_len;
2708}
2709
3697649f
DB
2710static inline int skb_try_make_writable(struct sk_buff *skb,
2711 unsigned int write_len)
2712{
2713 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2714 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2715}
2716
d9cc2048
HX
2717static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2718 int cloned)
2719{
2720 int delta = 0;
2721
d9cc2048
HX
2722 if (headroom > skb_headroom(skb))
2723 delta = headroom - skb_headroom(skb);
2724
2725 if (delta || cloned)
2726 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2727 GFP_ATOMIC);
2728 return 0;
2729}
2730
1da177e4
LT
2731/**
2732 * skb_cow - copy header of skb when it is required
2733 * @skb: buffer to cow
2734 * @headroom: needed headroom
2735 *
2736 * If the skb passed lacks sufficient headroom or its data part
2737 * is shared, data is reallocated. If reallocation fails, an error
2738 * is returned and original skb is not changed.
2739 *
2740 * The result is skb with writable area skb->head...skb->tail
2741 * and at least @headroom of space at head.
2742 */
2743static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2744{
d9cc2048
HX
2745 return __skb_cow(skb, headroom, skb_cloned(skb));
2746}
1da177e4 2747
d9cc2048
HX
2748/**
2749 * skb_cow_head - skb_cow but only making the head writable
2750 * @skb: buffer to cow
2751 * @headroom: needed headroom
2752 *
2753 * This function is identical to skb_cow except that we replace the
2754 * skb_cloned check by skb_header_cloned. It should be used when
2755 * you only need to push on some header and do not need to modify
2756 * the data.
2757 */
2758static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2759{
2760 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2761}
2762
2763/**
2764 * skb_padto - pad an skbuff up to a minimal size
2765 * @skb: buffer to pad
2766 * @len: minimal length
2767 *
2768 * Pads up a buffer to ensure the trailing bytes exist and are
2769 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2770 * is untouched. Otherwise it is extended. Returns zero on
2771 * success. The skb is freed on error.
1da177e4 2772 */
5b057c6b 2773static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2774{
2775 unsigned int size = skb->len;
2776 if (likely(size >= len))
5b057c6b 2777 return 0;
987c402a 2778 return skb_pad(skb, len - size);
1da177e4
LT
2779}
2780
9c0c1124
AD
2781/**
2782 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2783 * @skb: buffer to pad
2784 * @len: minimal length
2785 *
2786 * Pads up a buffer to ensure the trailing bytes exist and are
2787 * blanked. If the buffer already contains sufficient data it
2788 * is untouched. Otherwise it is extended. Returns zero on
2789 * success. The skb is freed on error.
2790 */
2791static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2792{
2793 unsigned int size = skb->len;
2794
2795 if (unlikely(size < len)) {
2796 len -= size;
2797 if (skb_pad(skb, len))
2798 return -ENOMEM;
2799 __skb_put(skb, len);
2800 }
2801 return 0;
2802}
2803
1da177e4 2804static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2805 struct iov_iter *from, int copy)
1da177e4
LT
2806{
2807 const int off = skb->len;
2808
2809 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2810 __wsum csum = 0;
2811 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2812 &csum, from) == copy) {
1da177e4
LT
2813 skb->csum = csum_block_add(skb->csum, csum, off);
2814 return 0;
2815 }
af2b040e 2816 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2817 return 0;
2818
2819 __skb_trim(skb, off);
2820 return -EFAULT;
2821}
2822
38ba0a65
ED
2823static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2824 const struct page *page, int off)
1da177e4
LT
2825{
2826 if (i) {
9e903e08 2827 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2828
ea2ab693 2829 return page == skb_frag_page(frag) &&
9e903e08 2830 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2831 }
38ba0a65 2832 return false;
1da177e4
LT
2833}
2834
364c6bad
HX
2835static inline int __skb_linearize(struct sk_buff *skb)
2836{
2837 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2838}
2839
1da177e4
LT
2840/**
2841 * skb_linearize - convert paged skb to linear one
2842 * @skb: buffer to linarize
1da177e4
LT
2843 *
2844 * If there is no free memory -ENOMEM is returned, otherwise zero
2845 * is returned and the old skb data released.
2846 */
364c6bad
HX
2847static inline int skb_linearize(struct sk_buff *skb)
2848{
2849 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2850}
2851
cef401de
ED
2852/**
2853 * skb_has_shared_frag - can any frag be overwritten
2854 * @skb: buffer to test
2855 *
2856 * Return true if the skb has at least one frag that might be modified
2857 * by an external entity (as in vmsplice()/sendfile())
2858 */
2859static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2860{
c9af6db4
PS
2861 return skb_is_nonlinear(skb) &&
2862 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2863}
2864
364c6bad
HX
2865/**
2866 * skb_linearize_cow - make sure skb is linear and writable
2867 * @skb: buffer to process
2868 *
2869 * If there is no free memory -ENOMEM is returned, otherwise zero
2870 * is returned and the old skb data released.
2871 */
2872static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2873{
364c6bad
HX
2874 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2875 __skb_linearize(skb) : 0;
1da177e4
LT
2876}
2877
479ffccc
DB
2878static __always_inline void
2879__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2880 unsigned int off)
2881{
2882 if (skb->ip_summed == CHECKSUM_COMPLETE)
2883 skb->csum = csum_block_sub(skb->csum,
2884 csum_partial(start, len, 0), off);
2885 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2886 skb_checksum_start_offset(skb) < 0)
2887 skb->ip_summed = CHECKSUM_NONE;
2888}
2889
1da177e4
LT
2890/**
2891 * skb_postpull_rcsum - update checksum for received skb after pull
2892 * @skb: buffer to update
2893 * @start: start of data before pull
2894 * @len: length of data pulled
2895 *
2896 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2897 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2898 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 2899 */
1da177e4 2900static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2901 const void *start, unsigned int len)
1da177e4 2902{
479ffccc 2903 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
2904}
2905
479ffccc
DB
2906static __always_inline void
2907__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2908 unsigned int off)
2909{
2910 if (skb->ip_summed == CHECKSUM_COMPLETE)
2911 skb->csum = csum_block_add(skb->csum,
2912 csum_partial(start, len, 0), off);
2913}
cbb042f9 2914
479ffccc
DB
2915/**
2916 * skb_postpush_rcsum - update checksum for received skb after push
2917 * @skb: buffer to update
2918 * @start: start of data after push
2919 * @len: length of data pushed
2920 *
2921 * After doing a push on a received packet, you need to call this to
2922 * update the CHECKSUM_COMPLETE checksum.
2923 */
f8ffad69
DB
2924static inline void skb_postpush_rcsum(struct sk_buff *skb,
2925 const void *start, unsigned int len)
2926{
479ffccc 2927 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
2928}
2929
479ffccc
DB
2930unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2931
82a31b92
WC
2932/**
2933 * skb_push_rcsum - push skb and update receive checksum
2934 * @skb: buffer to update
2935 * @len: length of data pulled
2936 *
2937 * This function performs an skb_push on the packet and updates
2938 * the CHECKSUM_COMPLETE checksum. It should be used on
2939 * receive path processing instead of skb_push unless you know
2940 * that the checksum difference is zero (e.g., a valid IP header)
2941 * or you are setting ip_summed to CHECKSUM_NONE.
2942 */
2943static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2944 unsigned int len)
2945{
2946 skb_push(skb, len);
2947 skb_postpush_rcsum(skb, skb->data, len);
2948 return skb->data;
2949}
2950
7ce5a27f
DM
2951/**
2952 * pskb_trim_rcsum - trim received skb and update checksum
2953 * @skb: buffer to trim
2954 * @len: new length
2955 *
2956 * This is exactly the same as pskb_trim except that it ensures the
2957 * checksum of received packets are still valid after the operation.
2958 */
2959
2960static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2961{
2962 if (likely(len >= skb->len))
2963 return 0;
2964 if (skb->ip_summed == CHECKSUM_COMPLETE)
2965 skb->ip_summed = CHECKSUM_NONE;
2966 return __pskb_trim(skb, len);
2967}
2968
5293efe6
DB
2969static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2970{
2971 if (skb->ip_summed == CHECKSUM_COMPLETE)
2972 skb->ip_summed = CHECKSUM_NONE;
2973 __skb_trim(skb, len);
2974 return 0;
2975}
2976
2977static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2978{
2979 if (skb->ip_summed == CHECKSUM_COMPLETE)
2980 skb->ip_summed = CHECKSUM_NONE;
2981 return __skb_grow(skb, len);
2982}
2983
1da177e4
LT
2984#define skb_queue_walk(queue, skb) \
2985 for (skb = (queue)->next; \
a1e4891f 2986 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2987 skb = skb->next)
2988
46f8914e
JC
2989#define skb_queue_walk_safe(queue, skb, tmp) \
2990 for (skb = (queue)->next, tmp = skb->next; \
2991 skb != (struct sk_buff *)(queue); \
2992 skb = tmp, tmp = skb->next)
2993
1164f52a 2994#define skb_queue_walk_from(queue, skb) \
a1e4891f 2995 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2996 skb = skb->next)
2997
2998#define skb_queue_walk_from_safe(queue, skb, tmp) \
2999 for (tmp = skb->next; \
3000 skb != (struct sk_buff *)(queue); \
3001 skb = tmp, tmp = skb->next)
3002
300ce174
SH
3003#define skb_queue_reverse_walk(queue, skb) \
3004 for (skb = (queue)->prev; \
a1e4891f 3005 skb != (struct sk_buff *)(queue); \
300ce174
SH
3006 skb = skb->prev)
3007
686a2955
DM
3008#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3009 for (skb = (queue)->prev, tmp = skb->prev; \
3010 skb != (struct sk_buff *)(queue); \
3011 skb = tmp, tmp = skb->prev)
3012
3013#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3014 for (tmp = skb->prev; \
3015 skb != (struct sk_buff *)(queue); \
3016 skb = tmp, tmp = skb->prev)
1da177e4 3017
21dc3301 3018static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3019{
3020 return skb_shinfo(skb)->frag_list != NULL;
3021}
3022
3023static inline void skb_frag_list_init(struct sk_buff *skb)
3024{
3025 skb_shinfo(skb)->frag_list = NULL;
3026}
3027
ee039871
DM
3028#define skb_walk_frags(skb, iter) \
3029 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3030
ea3793ee
RW
3031
3032int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3033 const struct sk_buff *skb);
3034struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3035 int *peeked, int *off, int *err,
3036 struct sk_buff **last);
7965bd4d
JP
3037struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3038 int *peeked, int *off, int *err);
3039struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3040 int *err);
3041unsigned int datagram_poll(struct file *file, struct socket *sock,
3042 struct poll_table_struct *wait);
c0371da6
AV
3043int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3044 struct iov_iter *to, int size);
51f3d02b
DM
3045static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3046 struct msghdr *msg, int size)
3047{
e5a4b0bb 3048 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3049}
e5a4b0bb
AV
3050int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3051 struct msghdr *msg);
3a654f97
AV
3052int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3053 struct iov_iter *from, int len);
3a654f97 3054int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3055void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3056void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3057static inline void skb_free_datagram_locked(struct sock *sk,
3058 struct sk_buff *skb)
3059{
3060 __skb_free_datagram_locked(sk, skb, 0);
3061}
7965bd4d 3062int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3063int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3064int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3065__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3066 int len, __wsum csum);
a60e3cc7
HFS
3067ssize_t skb_socket_splice(struct sock *sk,
3068 struct pipe_inode_info *pipe,
3069 struct splice_pipe_desc *spd);
3070int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3071 struct pipe_inode_info *pipe, unsigned int len,
a60e3cc7
HFS
3072 unsigned int flags,
3073 ssize_t (*splice_cb)(struct sock *,
3074 struct pipe_inode_info *,
3075 struct splice_pipe_desc *));
7965bd4d 3076void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3077unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3078int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3079 int len, int hlen);
7965bd4d
JP
3080void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3081int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3082void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3083unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3084bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 3085struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3086struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3087int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3088int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3089int skb_vlan_pop(struct sk_buff *skb);
3090int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3091struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3092 gfp_t gfp);
20380731 3093
6ce8e9ce
AV
3094static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3095{
21226abb 3096 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
3097}
3098
7eab8d9e
AV
3099static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3100{
e5a4b0bb 3101 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3102}
3103
2817a336
DB
3104struct skb_checksum_ops {
3105 __wsum (*update)(const void *mem, int len, __wsum wsum);
3106 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3107};
3108
3109__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3110 __wsum csum, const struct skb_checksum_ops *ops);
3111__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3112 __wsum csum);
3113
1e98a0f0
ED
3114static inline void * __must_check
3115__skb_header_pointer(const struct sk_buff *skb, int offset,
3116 int len, void *data, int hlen, void *buffer)
1da177e4 3117{
55820ee2 3118 if (hlen - offset >= len)
690e36e7 3119 return data + offset;
1da177e4 3120
690e36e7
DM
3121 if (!skb ||
3122 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3123 return NULL;
3124
3125 return buffer;
3126}
3127
1e98a0f0
ED
3128static inline void * __must_check
3129skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3130{
3131 return __skb_header_pointer(skb, offset, len, skb->data,
3132 skb_headlen(skb), buffer);
3133}
3134
4262e5cc
DB
3135/**
3136 * skb_needs_linearize - check if we need to linearize a given skb
3137 * depending on the given device features.
3138 * @skb: socket buffer to check
3139 * @features: net device features
3140 *
3141 * Returns true if either:
3142 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3143 * 2. skb is fragmented and the device does not support SG.
3144 */
3145static inline bool skb_needs_linearize(struct sk_buff *skb,
3146 netdev_features_t features)
3147{
3148 return skb_is_nonlinear(skb) &&
3149 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3150 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3151}
3152
d626f62b
ACM
3153static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3154 void *to,
3155 const unsigned int len)
3156{
3157 memcpy(to, skb->data, len);
3158}
3159
3160static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3161 const int offset, void *to,
3162 const unsigned int len)
3163{
3164 memcpy(to, skb->data + offset, len);
3165}
3166
27d7ff46
ACM
3167static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3168 const void *from,
3169 const unsigned int len)
3170{
3171 memcpy(skb->data, from, len);
3172}
3173
3174static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3175 const int offset,
3176 const void *from,
3177 const unsigned int len)
3178{
3179 memcpy(skb->data + offset, from, len);
3180}
3181
7965bd4d 3182void skb_init(void);
1da177e4 3183
ac45f602
PO
3184static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3185{
3186 return skb->tstamp;
3187}
3188
a61bbcf2
PM
3189/**
3190 * skb_get_timestamp - get timestamp from a skb
3191 * @skb: skb to get stamp from
3192 * @stamp: pointer to struct timeval to store stamp in
3193 *
3194 * Timestamps are stored in the skb as offsets to a base timestamp.
3195 * This function converts the offset back to a struct timeval and stores
3196 * it in stamp.
3197 */
ac45f602
PO
3198static inline void skb_get_timestamp(const struct sk_buff *skb,
3199 struct timeval *stamp)
a61bbcf2 3200{
b7aa0bf7 3201 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3202}
3203
ac45f602
PO
3204static inline void skb_get_timestampns(const struct sk_buff *skb,
3205 struct timespec *stamp)
3206{
3207 *stamp = ktime_to_timespec(skb->tstamp);
3208}
3209
b7aa0bf7 3210static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3211{
b7aa0bf7 3212 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3213}
3214
164891aa
SH
3215static inline ktime_t net_timedelta(ktime_t t)
3216{
3217 return ktime_sub(ktime_get_real(), t);
3218}
3219
b9ce204f
IJ
3220static inline ktime_t net_invalid_timestamp(void)
3221{
3222 return ktime_set(0, 0);
3223}
a61bbcf2 3224
62bccb8c
AD
3225struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3226
c1f19b51
RC
3227#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3228
7965bd4d
JP
3229void skb_clone_tx_timestamp(struct sk_buff *skb);
3230bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3231
3232#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3233
3234static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3235{
3236}
3237
3238static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3239{
3240 return false;
3241}
3242
3243#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3244
3245/**
3246 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3247 *
da92b194
RC
3248 * PHY drivers may accept clones of transmitted packets for
3249 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3250 * must call this function to return the skb back to the stack with a
3251 * timestamp.
da92b194 3252 *
c1f19b51 3253 * @skb: clone of the the original outgoing packet
7a76a021 3254 * @hwtstamps: hardware time stamps
c1f19b51
RC
3255 *
3256 */
3257void skb_complete_tx_timestamp(struct sk_buff *skb,
3258 struct skb_shared_hwtstamps *hwtstamps);
3259
e7fd2885
WB
3260void __skb_tstamp_tx(struct sk_buff *orig_skb,
3261 struct skb_shared_hwtstamps *hwtstamps,
3262 struct sock *sk, int tstype);
3263
ac45f602
PO
3264/**
3265 * skb_tstamp_tx - queue clone of skb with send time stamps
3266 * @orig_skb: the original outgoing packet
3267 * @hwtstamps: hardware time stamps, may be NULL if not available
3268 *
3269 * If the skb has a socket associated, then this function clones the
3270 * skb (thus sharing the actual data and optional structures), stores
3271 * the optional hardware time stamping information (if non NULL) or
3272 * generates a software time stamp (otherwise), then queues the clone
3273 * to the error queue of the socket. Errors are silently ignored.
3274 */
7965bd4d
JP
3275void skb_tstamp_tx(struct sk_buff *orig_skb,
3276 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3277
4507a715
RC
3278static inline void sw_tx_timestamp(struct sk_buff *skb)
3279{
2244d07b
OH
3280 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3281 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3282 skb_tstamp_tx(skb, NULL);
3283}
3284
3285/**
3286 * skb_tx_timestamp() - Driver hook for transmit timestamping
3287 *
3288 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3289 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3290 *
73409f3b
DM
3291 * Specifically, one should make absolutely sure that this function is
3292 * called before TX completion of this packet can trigger. Otherwise
3293 * the packet could potentially already be freed.
3294 *
4507a715
RC
3295 * @skb: A socket buffer.
3296 */
3297static inline void skb_tx_timestamp(struct sk_buff *skb)
3298{
c1f19b51 3299 skb_clone_tx_timestamp(skb);
4507a715
RC
3300 sw_tx_timestamp(skb);
3301}
3302
6e3e939f
JB
3303/**
3304 * skb_complete_wifi_ack - deliver skb with wifi status
3305 *
3306 * @skb: the original outgoing packet
3307 * @acked: ack status
3308 *
3309 */
3310void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3311
7965bd4d
JP
3312__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3313__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3314
60476372
HX
3315static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3316{
6edec0e6
TH
3317 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3318 skb->csum_valid ||
3319 (skb->ip_summed == CHECKSUM_PARTIAL &&
3320 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3321}
3322
fb286bb2
HX
3323/**
3324 * skb_checksum_complete - Calculate checksum of an entire packet
3325 * @skb: packet to process
3326 *
3327 * This function calculates the checksum over the entire packet plus
3328 * the value of skb->csum. The latter can be used to supply the
3329 * checksum of a pseudo header as used by TCP/UDP. It returns the
3330 * checksum.
3331 *
3332 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3333 * this function can be used to verify that checksum on received
3334 * packets. In that case the function should return zero if the
3335 * checksum is correct. In particular, this function will return zero
3336 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3337 * hardware has already verified the correctness of the checksum.
3338 */
4381ca3c 3339static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3340{
60476372
HX
3341 return skb_csum_unnecessary(skb) ?
3342 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3343}
3344
77cffe23
TH
3345static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3346{
3347 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3348 if (skb->csum_level == 0)
3349 skb->ip_summed = CHECKSUM_NONE;
3350 else
3351 skb->csum_level--;
3352 }
3353}
3354
3355static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3356{
3357 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3358 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3359 skb->csum_level++;
3360 } else if (skb->ip_summed == CHECKSUM_NONE) {
3361 skb->ip_summed = CHECKSUM_UNNECESSARY;
3362 skb->csum_level = 0;
3363 }
3364}
3365
5a212329
TH
3366static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3367{
3368 /* Mark current checksum as bad (typically called from GRO
3369 * path). In the case that ip_summed is CHECKSUM_NONE
3370 * this must be the first checksum encountered in the packet.
3371 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3372 * checksum after the last one validated. For UDP, a zero
3373 * checksum can not be marked as bad.
3374 */
3375
3376 if (skb->ip_summed == CHECKSUM_NONE ||
3377 skb->ip_summed == CHECKSUM_UNNECESSARY)
3378 skb->csum_bad = 1;
3379}
3380
76ba0aae
TH
3381/* Check if we need to perform checksum complete validation.
3382 *
3383 * Returns true if checksum complete is needed, false otherwise
3384 * (either checksum is unnecessary or zero checksum is allowed).
3385 */
3386static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3387 bool zero_okay,
3388 __sum16 check)
3389{
5d0c2b95
TH
3390 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3391 skb->csum_valid = 1;
77cffe23 3392 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3393 return false;
3394 }
3395
3396 return true;
3397}
3398
3399/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3400 * in checksum_init.
3401 */
3402#define CHECKSUM_BREAK 76
3403
4e18b9ad
TH
3404/* Unset checksum-complete
3405 *
3406 * Unset checksum complete can be done when packet is being modified
3407 * (uncompressed for instance) and checksum-complete value is
3408 * invalidated.
3409 */
3410static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3411{
3412 if (skb->ip_summed == CHECKSUM_COMPLETE)
3413 skb->ip_summed = CHECKSUM_NONE;
3414}
3415
76ba0aae
TH
3416/* Validate (init) checksum based on checksum complete.
3417 *
3418 * Return values:
3419 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3420 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3421 * checksum is stored in skb->csum for use in __skb_checksum_complete
3422 * non-zero: value of invalid checksum
3423 *
3424 */
3425static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3426 bool complete,
3427 __wsum psum)
3428{
3429 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3430 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3431 skb->csum_valid = 1;
76ba0aae
TH
3432 return 0;
3433 }
5a212329
TH
3434 } else if (skb->csum_bad) {
3435 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3436 return (__force __sum16)1;
76ba0aae
TH
3437 }
3438
3439 skb->csum = psum;
3440
5d0c2b95
TH
3441 if (complete || skb->len <= CHECKSUM_BREAK) {
3442 __sum16 csum;
3443
3444 csum = __skb_checksum_complete(skb);
3445 skb->csum_valid = !csum;
3446 return csum;
3447 }
76ba0aae
TH
3448
3449 return 0;
3450}
3451
3452static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3453{
3454 return 0;
3455}
3456
3457/* Perform checksum validate (init). Note that this is a macro since we only
3458 * want to calculate the pseudo header which is an input function if necessary.
3459 * First we try to validate without any computation (checksum unnecessary) and
3460 * then calculate based on checksum complete calling the function to compute
3461 * pseudo header.
3462 *
3463 * Return values:
3464 * 0: checksum is validated or try to in skb_checksum_complete
3465 * non-zero: value of invalid checksum
3466 */
3467#define __skb_checksum_validate(skb, proto, complete, \
3468 zero_okay, check, compute_pseudo) \
3469({ \
3470 __sum16 __ret = 0; \
5d0c2b95 3471 skb->csum_valid = 0; \
76ba0aae
TH
3472 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3473 __ret = __skb_checksum_validate_complete(skb, \
3474 complete, compute_pseudo(skb, proto)); \
3475 __ret; \
3476})
3477
3478#define skb_checksum_init(skb, proto, compute_pseudo) \
3479 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3480
3481#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3482 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3483
3484#define skb_checksum_validate(skb, proto, compute_pseudo) \
3485 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3486
3487#define skb_checksum_validate_zero_check(skb, proto, check, \
3488 compute_pseudo) \
096a4cfa 3489 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3490
3491#define skb_checksum_simple_validate(skb) \
3492 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3493
d96535a1
TH
3494static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3495{
3496 return (skb->ip_summed == CHECKSUM_NONE &&
3497 skb->csum_valid && !skb->csum_bad);
3498}
3499
3500static inline void __skb_checksum_convert(struct sk_buff *skb,
3501 __sum16 check, __wsum pseudo)
3502{
3503 skb->csum = ~pseudo;
3504 skb->ip_summed = CHECKSUM_COMPLETE;
3505}
3506
3507#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3508do { \
3509 if (__skb_checksum_convert_check(skb)) \
3510 __skb_checksum_convert(skb, check, \
3511 compute_pseudo(skb, proto)); \
3512} while (0)
3513
15e2396d
TH
3514static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3515 u16 start, u16 offset)
3516{
3517 skb->ip_summed = CHECKSUM_PARTIAL;
3518 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3519 skb->csum_offset = offset - start;
3520}
3521
dcdc8994
TH
3522/* Update skbuf and packet to reflect the remote checksum offload operation.
3523 * When called, ptr indicates the starting point for skb->csum when
3524 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3525 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3526 */
3527static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3528 int start, int offset, bool nopartial)
dcdc8994
TH
3529{
3530 __wsum delta;
3531
15e2396d
TH
3532 if (!nopartial) {
3533 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3534 return;
3535 }
3536
dcdc8994
TH
3537 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3538 __skb_checksum_complete(skb);
3539 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3540 }
3541
3542 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3543
3544 /* Adjust skb->csum since we changed the packet */
3545 skb->csum = csum_add(skb->csum, delta);
3546}
3547
5f79e0f9 3548#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3549void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3550static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3551{
3552 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3553 nf_conntrack_destroy(nfct);
1da177e4
LT
3554}
3555static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3556{
3557 if (nfct)
3558 atomic_inc(&nfct->use);
3559}
2fc72c7b 3560#endif
34666d46 3561#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3562static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3563{
3564 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3565 kfree(nf_bridge);
3566}
3567static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3568{
3569 if (nf_bridge)
3570 atomic_inc(&nf_bridge->use);
3571}
3572#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3573static inline void nf_reset(struct sk_buff *skb)
3574{
5f79e0f9 3575#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3576 nf_conntrack_put(skb->nfct);
3577 skb->nfct = NULL;
2fc72c7b 3578#endif
34666d46 3579#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3580 nf_bridge_put(skb->nf_bridge);
3581 skb->nf_bridge = NULL;
3582#endif
3583}
3584
124dff01
PM
3585static inline void nf_reset_trace(struct sk_buff *skb)
3586{
478b360a 3587#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3588 skb->nf_trace = 0;
3589#endif
a193a4ab
PM
3590}
3591
edda553c 3592/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3593static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3594 bool copy)
edda553c 3595{
5f79e0f9 3596#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3597 dst->nfct = src->nfct;
3598 nf_conntrack_get(src->nfct);
b1937227
ED
3599 if (copy)
3600 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3601#endif
34666d46 3602#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3603 dst->nf_bridge = src->nf_bridge;
3604 nf_bridge_get(src->nf_bridge);
3605#endif
478b360a 3606#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3607 if (copy)
3608 dst->nf_trace = src->nf_trace;
478b360a 3609#endif
edda553c
YK
3610}
3611
e7ac05f3
YK
3612static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3613{
e7ac05f3 3614#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3615 nf_conntrack_put(dst->nfct);
2fc72c7b 3616#endif
34666d46 3617#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3618 nf_bridge_put(dst->nf_bridge);
3619#endif
b1937227 3620 __nf_copy(dst, src, true);
e7ac05f3
YK
3621}
3622
984bc16c
JM
3623#ifdef CONFIG_NETWORK_SECMARK
3624static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3625{
3626 to->secmark = from->secmark;
3627}
3628
3629static inline void skb_init_secmark(struct sk_buff *skb)
3630{
3631 skb->secmark = 0;
3632}
3633#else
3634static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3635{ }
3636
3637static inline void skb_init_secmark(struct sk_buff *skb)
3638{ }
3639#endif
3640
574f7194
EB
3641static inline bool skb_irq_freeable(const struct sk_buff *skb)
3642{
3643 return !skb->destructor &&
3644#if IS_ENABLED(CONFIG_XFRM)
3645 !skb->sp &&
3646#endif
3647#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3648 !skb->nfct &&
3649#endif
3650 !skb->_skb_refdst &&
3651 !skb_has_frag_list(skb);
3652}
3653
f25f4e44
PWJ
3654static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3655{
f25f4e44 3656 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3657}
3658
9247744e 3659static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3660{
4e3ab47a 3661 return skb->queue_mapping;
4e3ab47a
PE
3662}
3663
f25f4e44
PWJ
3664static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3665{
f25f4e44 3666 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3667}
3668
d5a9e24a
DM
3669static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3670{
3671 skb->queue_mapping = rx_queue + 1;
3672}
3673
9247744e 3674static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3675{
3676 return skb->queue_mapping - 1;
3677}
3678
9247744e 3679static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3680{
a02cec21 3681 return skb->queue_mapping != 0;
d5a9e24a
DM
3682}
3683
def8b4fa
AD
3684static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3685{
0b3d8e08 3686#ifdef CONFIG_XFRM
def8b4fa 3687 return skb->sp;
def8b4fa 3688#else
def8b4fa 3689 return NULL;
def8b4fa 3690#endif
0b3d8e08 3691}
def8b4fa 3692
68c33163
PS
3693/* Keeps track of mac header offset relative to skb->head.
3694 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3695 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3696 * tunnel skb it points to outer mac header.
3697 * Keeps track of level of encapsulation of network headers.
3698 */
68c33163 3699struct skb_gso_cb {
802ab55a
AD
3700 union {
3701 int mac_offset;
3702 int data_offset;
3703 };
3347c960 3704 int encap_level;
76443456 3705 __wsum csum;
7e2b10c1 3706 __u16 csum_start;
68c33163 3707};
9207f9d4
KK
3708#define SKB_SGO_CB_OFFSET 32
3709#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3710
3711static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3712{
3713 return (skb_mac_header(inner_skb) - inner_skb->head) -
3714 SKB_GSO_CB(inner_skb)->mac_offset;
3715}
3716
1e2bd517
PS
3717static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3718{
3719 int new_headroom, headroom;
3720 int ret;
3721
3722 headroom = skb_headroom(skb);
3723 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3724 if (ret)
3725 return ret;
3726
3727 new_headroom = skb_headroom(skb);
3728 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3729 return 0;
3730}
3731
08b64fcc
AD
3732static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3733{
3734 /* Do not update partial checksums if remote checksum is enabled. */
3735 if (skb->remcsum_offload)
3736 return;
3737
3738 SKB_GSO_CB(skb)->csum = res;
3739 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3740}
3741
7e2b10c1
TH
3742/* Compute the checksum for a gso segment. First compute the checksum value
3743 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3744 * then add in skb->csum (checksum from csum_start to end of packet).
3745 * skb->csum and csum_start are then updated to reflect the checksum of the
3746 * resultant packet starting from the transport header-- the resultant checksum
3747 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3748 * header.
3749 */
3750static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3751{
76443456
AD
3752 unsigned char *csum_start = skb_transport_header(skb);
3753 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3754 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3755
76443456
AD
3756 SKB_GSO_CB(skb)->csum = res;
3757 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3758
76443456 3759 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3760}
3761
bdcc0924 3762static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3763{
3764 return skb_shinfo(skb)->gso_size;
3765}
3766
36a8f39e 3767/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3768static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3769{
3770 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3771}
3772
5293efe6
DB
3773static inline void skb_gso_reset(struct sk_buff *skb)
3774{
3775 skb_shinfo(skb)->gso_size = 0;
3776 skb_shinfo(skb)->gso_segs = 0;
3777 skb_shinfo(skb)->gso_type = 0;
3778}
3779
7965bd4d 3780void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3781
3782static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3783{
3784 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3785 * wanted then gso_type will be set. */
05bdd2f1
ED
3786 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3787
b78462eb
AD
3788 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3789 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3790 __skb_warn_lro_forwarding(skb);
3791 return true;
3792 }
3793 return false;
3794}
3795
35fc92a9
HX
3796static inline void skb_forward_csum(struct sk_buff *skb)
3797{
3798 /* Unfortunately we don't support this one. Any brave souls? */
3799 if (skb->ip_summed == CHECKSUM_COMPLETE)
3800 skb->ip_summed = CHECKSUM_NONE;
3801}
3802
bc8acf2c
ED
3803/**
3804 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3805 * @skb: skb to check
3806 *
3807 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3808 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3809 * use this helper, to document places where we make this assertion.
3810 */
05bdd2f1 3811static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3812{
3813#ifdef DEBUG
3814 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3815#endif
3816}
3817
f35d9d8a 3818bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3819
ed1f50c3 3820int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3821struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3822 unsigned int transport_len,
3823 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3824
3a7c1ee4
AD
3825/**
3826 * skb_head_is_locked - Determine if the skb->head is locked down
3827 * @skb: skb to check
3828 *
3829 * The head on skbs build around a head frag can be removed if they are
3830 * not cloned. This function returns true if the skb head is locked down
3831 * due to either being allocated via kmalloc, or by being a clone with
3832 * multiple references to the head.
3833 */
3834static inline bool skb_head_is_locked(const struct sk_buff *skb)
3835{
3836 return !skb->head_frag || skb_cloned(skb);
3837}
fe6cc55f
FW
3838
3839/**
3840 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3841 *
3842 * @skb: GSO skb
3843 *
3844 * skb_gso_network_seglen is used to determine the real size of the
3845 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3846 *
3847 * The MAC/L2 header is not accounted for.
3848 */
3849static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3850{
3851 unsigned int hdr_len = skb_transport_header(skb) -
3852 skb_network_header(skb);
3853 return hdr_len + skb_gso_transport_seglen(skb);
3854}
ee122c79 3855
179bc67f
EC
3856/* Local Checksum Offload.
3857 * Compute outer checksum based on the assumption that the
3858 * inner checksum will be offloaded later.
e8ae7b00
EC
3859 * See Documentation/networking/checksum-offloads.txt for
3860 * explanation of how this works.
179bc67f
EC
3861 * Fill in outer checksum adjustment (e.g. with sum of outer
3862 * pseudo-header) before calling.
3863 * Also ensure that inner checksum is in linear data area.
3864 */
3865static inline __wsum lco_csum(struct sk_buff *skb)
3866{
9e74a6da
AD
3867 unsigned char *csum_start = skb_checksum_start(skb);
3868 unsigned char *l4_hdr = skb_transport_header(skb);
3869 __wsum partial;
179bc67f
EC
3870
3871 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3872 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3873 skb->csum_offset));
3874
179bc67f 3875 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3876 * adjustment filled in by caller) and return result.
179bc67f 3877 */
9e74a6da 3878 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3879}
3880
1da177e4
LT
3881#endif /* __KERNEL__ */
3882#endif /* _LINUX_SKBUFF_H */