net-next: mediatek: fix compile error inside mtk_poll_controller()
[linux-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>
f70ea018 40#include <net/flow.h>
1da177e4 41
7a6ae71b
TH
42/* The interface for checksum offload between the stack and networking drivers
43 * is as follows...
44 *
45 * A. IP checksum related features
46 *
47 * Drivers advertise checksum offload capabilities in the features of a device.
48 * From the stack's point of view these are capabilities offered by the driver,
49 * a driver typically only advertises features that it is capable of offloading
50 * to its device.
51 *
52 * The checksum related features are:
53 *
54 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
55 * IP (one's complement) checksum for any combination
56 * of protocols or protocol layering. The checksum is
57 * computed and set in a packet per the CHECKSUM_PARTIAL
58 * interface (see below).
59 *
60 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
61 * TCP or UDP packets over IPv4. These are specifically
62 * unencapsulated packets of the form IPv4|TCP or
63 * IPv4|UDP where the Protocol field in the IPv4 header
64 * is TCP or UDP. The IPv4 header may contain IP options
65 * This feature cannot be set in features for a device
66 * with NETIF_F_HW_CSUM also set. This feature is being
67 * DEPRECATED (see below).
68 *
69 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
70 * TCP or UDP packets over IPv6. These are specifically
71 * unencapsulated packets of the form IPv6|TCP or
72 * IPv4|UDP where the Next Header field in the IPv6
73 * header is either TCP or UDP. IPv6 extension headers
74 * are not supported with this feature. This feature
75 * cannot be set in features for a device with
76 * NETIF_F_HW_CSUM also set. This feature is being
77 * DEPRECATED (see below).
78 *
79 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
80 * This flag is used only used to disable the RX checksum
81 * feature for a device. The stack will accept receive
82 * checksum indication in packets received on a device
83 * regardless of whether NETIF_F_RXCSUM is set.
84 *
85 * B. Checksumming of received packets by device. Indication of checksum
86 * verification is in set skb->ip_summed. Possible values are:
78ea85f1
DB
87 *
88 * CHECKSUM_NONE:
89 *
7a6ae71b 90 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
91 * The packet contains full (though not verified) checksum in packet but
92 * not in skb->csum. Thus, skb->csum is undefined in this case.
93 *
94 * CHECKSUM_UNNECESSARY:
95 *
96 * The hardware you're dealing with doesn't calculate the full checksum
97 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
98 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
99 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
100 * though. A driver or device must never modify the checksum field in the
101 * packet even if checksum is verified.
77cffe23
TH
102 *
103 * CHECKSUM_UNNECESSARY is applicable to following protocols:
104 * TCP: IPv6 and IPv4.
105 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
106 * zero UDP checksum for either IPv4 or IPv6, the networking stack
107 * may perform further validation in this case.
108 * GRE: only if the checksum is present in the header.
109 * SCTP: indicates the CRC in SCTP header has been validated.
110 *
111 * skb->csum_level indicates the number of consecutive checksums found in
112 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
113 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
114 * and a device is able to verify the checksums for UDP (possibly zero),
115 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
116 * two. If the device were only able to verify the UDP checksum and not
117 * GRE, either because it doesn't support GRE checksum of because GRE
118 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
119 * not considered in this case).
78ea85f1
DB
120 *
121 * CHECKSUM_COMPLETE:
122 *
123 * This is the most generic way. The device supplied checksum of the _whole_
124 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
125 * hardware doesn't need to parse L3/L4 headers to implement this.
126 *
127 * Note: Even if device supports only some protocols, but is able to produce
128 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
129 *
130 * CHECKSUM_PARTIAL:
131 *
6edec0e6
TH
132 * A checksum is set up to be offloaded to a device as described in the
133 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 134 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
135 * on the same host, or it may be set in the input path in GRO or remote
136 * checksum offload. For the purposes of checksum verification, the checksum
137 * referred to by skb->csum_start + skb->csum_offset and any preceding
138 * checksums in the packet are considered verified. Any checksums in the
139 * packet that are after the checksum being offloaded are not considered to
140 * be verified.
78ea85f1 141 *
7a6ae71b
TH
142 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
143 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
144 *
145 * CHECKSUM_PARTIAL:
146 *
7a6ae71b 147 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 148 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
149 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
150 * csum_start and csum_offset values are valid values given the length and
151 * offset of the packet, however they should not attempt to validate that the
152 * checksum refers to a legitimate transport layer checksum-- it is the
153 * purview of the stack to validate that csum_start and csum_offset are set
154 * correctly.
155 *
156 * When the stack requests checksum offload for a packet, the driver MUST
157 * ensure that the checksum is set correctly. A driver can either offload the
158 * checksum calculation to the device, or call skb_checksum_help (in the case
159 * that the device does not support offload for a particular checksum).
160 *
161 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
162 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
163 * checksum offload capability. If a device has limited checksum capabilities
164 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
165 * described above) a helper function can be called to resolve
166 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
167 * function takes a spec argument that describes the protocol layer that is
168 * supported for checksum offload and can be called for each packet. If a
169 * packet does not match the specification for offload, skb_checksum_help
170 * is called to resolve the checksum.
78ea85f1 171 *
7a6ae71b 172 * CHECKSUM_NONE:
78ea85f1 173 *
7a6ae71b
TH
174 * The skb was already checksummed by the protocol, or a checksum is not
175 * required.
78ea85f1
DB
176 *
177 * CHECKSUM_UNNECESSARY:
178 *
7a6ae71b
TH
179 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
180 * output.
78ea85f1 181 *
7a6ae71b
TH
182 * CHECKSUM_COMPLETE:
183 * Not used in checksum output. If a driver observes a packet with this value
184 * set in skbuff, if should treat as CHECKSUM_NONE being set.
185 *
186 * D. Non-IP checksum (CRC) offloads
187 *
188 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
189 * offloading the SCTP CRC in a packet. To perform this offload the stack
190 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
191 * accordingly. Note the there is no indication in the skbuff that the
192 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
193 * both IP checksum offload and SCTP CRC offload must verify which offload
194 * is configured for a packet presumably by inspecting packet headers.
195 *
196 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
197 * offloading the FCOE CRC in a packet. To perform this offload the stack
198 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
199 * accordingly. Note the there is no indication in the skbuff that the
200 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
201 * both IP checksum offload and FCOE CRC offload must verify which offload
202 * is configured for a packet presumably by inspecting packet headers.
203 *
204 * E. Checksumming on output with GSO.
205 *
206 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
207 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
208 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
209 * part of the GSO operation is implied. If a checksum is being offloaded
210 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
211 * are set to refer to the outermost checksum being offload (two offloaded
212 * checksums are possible with UDP encapsulation).
78ea85f1
DB
213 */
214
60476372 215/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
216#define CHECKSUM_NONE 0
217#define CHECKSUM_UNNECESSARY 1
218#define CHECKSUM_COMPLETE 2
219#define CHECKSUM_PARTIAL 3
1da177e4 220
77cffe23
TH
221/* Maximum value in skb->csum_level */
222#define SKB_MAX_CSUM_LEVEL 3
223
0bec8c88 224#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 225#define SKB_WITH_OVERHEAD(X) \
deea84b0 226 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
227#define SKB_MAX_ORDER(X, ORDER) \
228 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
229#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
230#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
231
87fb4b7b
ED
232/* return minimum truesize of one skb containing X bytes of data */
233#define SKB_TRUESIZE(X) ((X) + \
234 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
235 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
236
1da177e4 237struct net_device;
716ea3a7 238struct scatterlist;
9c55e01c 239struct pipe_inode_info;
a8f820aa 240struct iov_iter;
fd11a83d 241struct napi_struct;
1da177e4 242
5f79e0f9 243#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
244struct nf_conntrack {
245 atomic_t use;
1da177e4 246};
5f79e0f9 247#endif
1da177e4 248
34666d46 249#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 250struct nf_bridge_info {
bf1ac5ca 251 atomic_t use;
3eaf4025
FW
252 enum {
253 BRNF_PROTO_UNCHANGED,
254 BRNF_PROTO_8021Q,
255 BRNF_PROTO_PPPOE
7fb48c5b 256 } orig_proto:8;
72b1e5e4
FW
257 u8 pkt_otherhost:1;
258 u8 in_prerouting:1;
259 u8 bridged_dnat:1;
411ffb4f 260 __u16 frag_max_size;
bf1ac5ca 261 struct net_device *physindev;
63cdbc06
FW
262
263 /* always valid & non-NULL from FORWARD on, for physdev match */
264 struct net_device *physoutdev;
7fb48c5b 265 union {
72b1e5e4 266 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
267 __be32 ipv4_daddr;
268 struct in6_addr ipv6_daddr;
72b1e5e4
FW
269
270 /* after prerouting + nat detected: store original source
271 * mac since neigh resolution overwrites it, only used while
272 * skb is out in neigh layer.
273 */
274 char neigh_header[8];
72b31f72 275 };
1da177e4
LT
276};
277#endif
278
1da177e4
LT
279struct sk_buff_head {
280 /* These two members must be first. */
281 struct sk_buff *next;
282 struct sk_buff *prev;
283
284 __u32 qlen;
285 spinlock_t lock;
286};
287
288struct sk_buff;
289
9d4dde52
IC
290/* To allow 64K frame to be packed as single skb without frag_list we
291 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
292 * buffers which do not start on a page boundary.
293 *
294 * Since GRO uses frags we allocate at least 16 regardless of page
295 * size.
a715dea3 296 */
9d4dde52 297#if (65536/PAGE_SIZE + 1) < 16
eec00954 298#define MAX_SKB_FRAGS 16UL
a715dea3 299#else
9d4dde52 300#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 301#endif
5f74f82e 302extern int sysctl_max_skb_frags;
1da177e4 303
3953c46c
MRL
304/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
305 * segment using its current segmentation instead.
306 */
307#define GSO_BY_FRAGS 0xFFFF
308
1da177e4
LT
309typedef struct skb_frag_struct skb_frag_t;
310
311struct skb_frag_struct {
a8605c60
IC
312 struct {
313 struct page *p;
314 } page;
cb4dfe56 315#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
316 __u32 page_offset;
317 __u32 size;
cb4dfe56
ED
318#else
319 __u16 page_offset;
320 __u16 size;
321#endif
1da177e4
LT
322};
323
9e903e08
ED
324static inline unsigned int skb_frag_size(const skb_frag_t *frag)
325{
326 return frag->size;
327}
328
329static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
330{
331 frag->size = size;
332}
333
334static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
335{
336 frag->size += delta;
337}
338
339static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
340{
341 frag->size -= delta;
342}
343
ac45f602
PO
344#define HAVE_HW_TIME_STAMP
345
346/**
d3a21be8 347 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
348 * @hwtstamp: hardware time stamp transformed into duration
349 * since arbitrary point in time
ac45f602
PO
350 *
351 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 352 * skb->tstamp.
ac45f602
PO
353 *
354 * hwtstamps can only be compared against other hwtstamps from
355 * the same device.
356 *
357 * This structure is attached to packets as part of the
358 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
359 */
360struct skb_shared_hwtstamps {
361 ktime_t hwtstamp;
ac45f602
PO
362};
363
2244d07b
OH
364/* Definitions for tx_flags in struct skb_shared_info */
365enum {
366 /* generate hardware time stamp */
367 SKBTX_HW_TSTAMP = 1 << 0,
368
e7fd2885 369 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
370 SKBTX_SW_TSTAMP = 1 << 1,
371
372 /* device driver is going to provide hardware time stamp */
373 SKBTX_IN_PROGRESS = 1 << 2,
374
a6686f2f 375 /* device driver supports TX zero-copy buffers */
62b1a8ab 376 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
377
378 /* generate wifi status information (where possible) */
62b1a8ab 379 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
380
381 /* This indicates at least one fragment might be overwritten
382 * (as in vmsplice(), sendfile() ...)
383 * If we need to compute a TX checksum, we'll need to copy
384 * all frags to avoid possible bad checksum
385 */
386 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
387
388 /* generate software time stamp when entering packet scheduling */
389 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
390};
391
e1c8a607 392#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 393 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
394#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
395
a6686f2f
SM
396/*
397 * The callback notifies userspace to release buffers when skb DMA is done in
398 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
399 * The zerocopy_success argument is true if zero copy transmit occurred,
400 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
401 * The ctx field is used to track device context.
402 * The desc field is used to track userspace buffer index.
a6686f2f
SM
403 */
404struct ubuf_info {
e19d6763 405 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 406 void *ctx;
a6686f2f 407 unsigned long desc;
ac45f602
PO
408};
409
1da177e4
LT
410/* This data is invariant across clones and lives at
411 * the end of the header data, ie. at skb->end.
412 */
413struct skb_shared_info {
9f42f126
IC
414 unsigned char nr_frags;
415 __u8 tx_flags;
7967168c
HX
416 unsigned short gso_size;
417 /* Warning: this field is not always filled in (UFO)! */
418 unsigned short gso_segs;
419 unsigned short gso_type;
1da177e4 420 struct sk_buff *frag_list;
ac45f602 421 struct skb_shared_hwtstamps hwtstamps;
09c2d251 422 u32 tskey;
9f42f126 423 __be32 ip6_frag_id;
ec7d2f2c
ED
424
425 /*
426 * Warning : all fields before dataref are cleared in __alloc_skb()
427 */
428 atomic_t dataref;
429
69e3c75f
JB
430 /* Intermediate layers must ensure that destructor_arg
431 * remains valid until skb destructor */
432 void * destructor_arg;
a6686f2f 433
fed66381
ED
434 /* must be last field, see pskb_expand_head() */
435 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
436};
437
438/* We divide dataref into two halves. The higher 16 bits hold references
439 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
440 * the entire skb->data. A clone of a headerless skb holds the length of
441 * the header in skb->hdr_len.
1da177e4
LT
442 *
443 * All users must obey the rule that the skb->data reference count must be
444 * greater than or equal to the payload reference count.
445 *
446 * Holding a reference to the payload part means that the user does not
447 * care about modifications to the header part of skb->data.
448 */
449#define SKB_DATAREF_SHIFT 16
450#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
451
d179cd12
DM
452
453enum {
c8753d55
VS
454 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
455 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
456 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
457};
458
7967168c
HX
459enum {
460 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 461 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
462
463 /* This indicates the skb is from an untrusted source. */
464 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
465
466 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
467 SKB_GSO_TCP_ECN = 1 << 3,
468
cbc53e08 469 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 470
cbc53e08 471 SKB_GSO_TCPV6 = 1 << 5,
68c33163 472
cbc53e08 473 SKB_GSO_FCOE = 1 << 6,
73136267 474
cbc53e08 475 SKB_GSO_GRE = 1 << 7,
0d89d203 476
cbc53e08 477 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 478
7e13318d 479 SKB_GSO_IPXIP4 = 1 << 9,
61c1db7f 480
7e13318d 481 SKB_GSO_IPXIP6 = 1 << 10,
0f4f4ffa 482
cbc53e08 483 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 484
cbc53e08
AD
485 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
486
802ab55a
AD
487 SKB_GSO_PARTIAL = 1 << 13,
488
489 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
90017acc
MRL
490
491 SKB_GSO_SCTP = 1 << 15,
7967168c
HX
492};
493
2e07fa9c
ACM
494#if BITS_PER_LONG > 32
495#define NET_SKBUFF_DATA_USES_OFFSET 1
496#endif
497
498#ifdef NET_SKBUFF_DATA_USES_OFFSET
499typedef unsigned int sk_buff_data_t;
500#else
501typedef unsigned char *sk_buff_data_t;
502#endif
503
363ec392
ED
504/**
505 * struct skb_mstamp - multi resolution time stamps
506 * @stamp_us: timestamp in us resolution
507 * @stamp_jiffies: timestamp in jiffies
508 */
509struct skb_mstamp {
510 union {
511 u64 v64;
512 struct {
513 u32 stamp_us;
514 u32 stamp_jiffies;
515 };
516 };
517};
518
519/**
520 * skb_mstamp_get - get current timestamp
521 * @cl: place to store timestamps
522 */
523static inline void skb_mstamp_get(struct skb_mstamp *cl)
524{
525 u64 val = local_clock();
526
527 do_div(val, NSEC_PER_USEC);
528 cl->stamp_us = (u32)val;
529 cl->stamp_jiffies = (u32)jiffies;
530}
531
532/**
533 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
534 * @t1: pointer to newest sample
535 * @t0: pointer to oldest sample
536 */
537static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
538 const struct skb_mstamp *t0)
539{
540 s32 delta_us = t1->stamp_us - t0->stamp_us;
541 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
542
543 /* If delta_us is negative, this might be because interval is too big,
544 * or local_clock() drift is too big : fallback using jiffies.
545 */
546 if (delta_us <= 0 ||
547 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
548
549 delta_us = jiffies_to_usecs(delta_jiffies);
550
551 return delta_us;
552}
553
625a5e10
YC
554static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
555 const struct skb_mstamp *t0)
556{
557 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
558
559 if (!diff)
560 diff = t1->stamp_us - t0->stamp_us;
561 return diff > 0;
562}
363ec392 563
1da177e4
LT
564/**
565 * struct sk_buff - socket buffer
566 * @next: Next buffer in list
567 * @prev: Previous buffer in list
363ec392 568 * @tstamp: Time we arrived/left
56b17425 569 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 570 * @sk: Socket we are owned by
1da177e4 571 * @dev: Device we arrived on/are leaving by
d84e0bd7 572 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 573 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 574 * @sp: the security path, used for xfrm
1da177e4
LT
575 * @len: Length of actual data
576 * @data_len: Data length
577 * @mac_len: Length of link layer header
334a8132 578 * @hdr_len: writable header length of cloned skb
663ead3b
HX
579 * @csum: Checksum (must include start/offset pair)
580 * @csum_start: Offset from skb->head where checksumming should start
581 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 582 * @priority: Packet queueing priority
60ff7467 583 * @ignore_df: allow local fragmentation
1da177e4 584 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 585 * @ip_summed: Driver fed us an IP checksum
1da177e4 586 * @nohdr: Payload reference only, must not modify header
d84e0bd7 587 * @nfctinfo: Relationship of this skb to the connection
1da177e4 588 * @pkt_type: Packet class
c83c2486 589 * @fclone: skbuff clone status
c83c2486 590 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
591 * @peeked: this packet has been seen already, so stats have been
592 * done for it, don't do them again
ba9dda3a 593 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
594 * @protocol: Packet protocol from driver
595 * @destructor: Destruct function
596 * @nfct: Associated connection, if any
1da177e4 597 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 598 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
599 * @tc_index: Traffic control index
600 * @tc_verd: traffic control verdict
61b905da 601 * @hash: the packet hash
d84e0bd7 602 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 603 * @xmit_more: More SKBs are pending for this queue
553a5672 604 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 605 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 606 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 607 * ports.
a3b18ddb 608 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
609 * @wifi_acked_valid: wifi_acked was set
610 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 611 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 612 * @napi_id: id of the NAPI struct this skb came from
984bc16c 613 * @secmark: security marking
0c4f691f 614 * @offload_fwd_mark: fwding offload mark
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
ED
678 __u16 queue_mapping;
679 __u8 cloned:1,
6869c4d8 680 nohdr:1,
b84f4cc9 681 fclone:2,
a59322be 682 peeked:1,
b1937227
ED
683 head_frag:1,
684 xmit_more:1;
685 /* one bit hole */
fe55f6d5 686 kmemcheck_bitfield_end(flags1);
4031ae6e 687
b1937227
ED
688 /* fields enclosed in headers_start/headers_end are copied
689 * using a single memcpy() in __copy_skb_header()
690 */
ebcf34f3 691 /* private: */
b1937227 692 __u32 headers_start[0];
ebcf34f3 693 /* public: */
4031ae6e 694
233577a2
HFS
695/* if you move pkt_type around you also must adapt those constants */
696#ifdef __BIG_ENDIAN_BITFIELD
697#define PKT_TYPE_MAX (7 << 5)
698#else
699#define PKT_TYPE_MAX 7
1da177e4 700#endif
233577a2 701#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 702
233577a2 703 __u8 __pkt_type_offset[0];
b1937227 704 __u8 pkt_type:3;
c93bdd0e 705 __u8 pfmemalloc:1;
b1937227
ED
706 __u8 ignore_df:1;
707 __u8 nfctinfo:3;
708
709 __u8 nf_trace:1;
710 __u8 ip_summed:2;
3853b584 711 __u8 ooo_okay:1;
61b905da 712 __u8 l4_hash:1;
a3b18ddb 713 __u8 sw_hash:1;
6e3e939f
JB
714 __u8 wifi_acked_valid:1;
715 __u8 wifi_acked:1;
b1937227 716
3bdc0eba 717 __u8 no_fcs:1;
77cffe23 718 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 719 __u8 encapsulation:1;
7e2b10c1 720 __u8 encap_hdr_csum:1;
5d0c2b95 721 __u8 csum_valid:1;
7e3cead5 722 __u8 csum_complete_sw:1;
b1937227
ED
723 __u8 csum_level:2;
724 __u8 csum_bad:1;
fe55f6d5 725
b1937227
ED
726#ifdef CONFIG_IPV6_NDISC_NODETYPE
727 __u8 ndisc_nodetype:2;
728#endif
729 __u8 ipvs_property:1;
8bce6d7d 730 __u8 inner_protocol_type:1;
e585f236
TH
731 __u8 remcsum_offload:1;
732 /* 3 or 5 bit hole */
b1937227
ED
733
734#ifdef CONFIG_NET_SCHED
735 __u16 tc_index; /* traffic control index */
736#ifdef CONFIG_NET_CLS_ACT
737 __u16 tc_verd; /* traffic control verdict */
738#endif
739#endif
fe55f6d5 740
b1937227
ED
741 union {
742 __wsum csum;
743 struct {
744 __u16 csum_start;
745 __u16 csum_offset;
746 };
747 };
748 __u32 priority;
749 int skb_iif;
750 __u32 hash;
751 __be16 vlan_proto;
752 __u16 vlan_tci;
2bd82484
ED
753#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
754 union {
755 unsigned int napi_id;
756 unsigned int sender_cpu;
757 };
97fc2f08 758#endif
0c4f691f 759 union {
984bc16c 760#ifdef CONFIG_NETWORK_SECMARK
0c4f691f
SF
761 __u32 secmark;
762#endif
763#ifdef CONFIG_NET_SWITCHDEV
764 __u32 offload_fwd_mark;
984bc16c 765#endif
0c4f691f
SF
766 };
767
3b885787
NH
768 union {
769 __u32 mark;
16fad69c 770 __u32 reserved_tailroom;
3b885787 771 };
1da177e4 772
8bce6d7d
TH
773 union {
774 __be16 inner_protocol;
775 __u8 inner_ipproto;
776 };
777
1a37e412
SH
778 __u16 inner_transport_header;
779 __u16 inner_network_header;
780 __u16 inner_mac_header;
b1937227
ED
781
782 __be16 protocol;
1a37e412
SH
783 __u16 transport_header;
784 __u16 network_header;
785 __u16 mac_header;
b1937227 786
ebcf34f3 787 /* private: */
b1937227 788 __u32 headers_end[0];
ebcf34f3 789 /* public: */
b1937227 790
1da177e4 791 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 792 sk_buff_data_t tail;
4305b541 793 sk_buff_data_t end;
1da177e4 794 unsigned char *head,
4305b541 795 *data;
27a884dc
ACM
796 unsigned int truesize;
797 atomic_t users;
1da177e4
LT
798};
799
800#ifdef __KERNEL__
801/*
802 * Handling routines are only of interest to the kernel
803 */
804#include <linux/slab.h>
805
1da177e4 806
c93bdd0e
MG
807#define SKB_ALLOC_FCLONE 0x01
808#define SKB_ALLOC_RX 0x02
fd11a83d 809#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
810
811/* Returns true if the skb was allocated from PFMEMALLOC reserves */
812static inline bool skb_pfmemalloc(const struct sk_buff *skb)
813{
814 return unlikely(skb->pfmemalloc);
815}
816
7fee226a
ED
817/*
818 * skb might have a dst pointer attached, refcounted or not.
819 * _skb_refdst low order bit is set if refcount was _not_ taken
820 */
821#define SKB_DST_NOREF 1UL
822#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
823
824/**
825 * skb_dst - returns skb dst_entry
826 * @skb: buffer
827 *
828 * Returns skb dst_entry, regardless of reference taken or not.
829 */
adf30907
ED
830static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
831{
7fee226a
ED
832 /* If refdst was not refcounted, check we still are in a
833 * rcu_read_lock section
834 */
835 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
836 !rcu_read_lock_held() &&
837 !rcu_read_lock_bh_held());
838 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
839}
840
7fee226a
ED
841/**
842 * skb_dst_set - sets skb dst
843 * @skb: buffer
844 * @dst: dst entry
845 *
846 * Sets skb dst, assuming a reference was taken on dst and should
847 * be released by skb_dst_drop()
848 */
adf30907
ED
849static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
850{
7fee226a
ED
851 skb->_skb_refdst = (unsigned long)dst;
852}
853
932bc4d7
JA
854/**
855 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
856 * @skb: buffer
857 * @dst: dst entry
858 *
859 * Sets skb dst, assuming a reference was not taken on dst.
860 * If dst entry is cached, we do not take reference and dst_release
861 * will be avoided by refdst_drop. If dst entry is not cached, we take
862 * reference, so that last dst_release can destroy the dst immediately.
863 */
864static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
865{
dbfc4fb7
HFS
866 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
867 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 868}
7fee226a
ED
869
870/**
25985edc 871 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
872 * @skb: buffer
873 */
874static inline bool skb_dst_is_noref(const struct sk_buff *skb)
875{
876 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
877}
878
511c3f92
ED
879static inline struct rtable *skb_rtable(const struct sk_buff *skb)
880{
adf30907 881 return (struct rtable *)skb_dst(skb);
511c3f92
ED
882}
883
7965bd4d
JP
884void kfree_skb(struct sk_buff *skb);
885void kfree_skb_list(struct sk_buff *segs);
886void skb_tx_error(struct sk_buff *skb);
887void consume_skb(struct sk_buff *skb);
888void __kfree_skb(struct sk_buff *skb);
d7e8883c 889extern struct kmem_cache *skbuff_head_cache;
bad43ca8 890
7965bd4d
JP
891void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
892bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
893 bool *fragstolen, int *delta_truesize);
bad43ca8 894
7965bd4d
JP
895struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
896 int node);
2ea2f62c 897struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 898struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 899static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 900 gfp_t priority)
d179cd12 901{
564824b0 902 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
903}
904
2e4e4410
ED
905struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
906 unsigned long data_len,
907 int max_page_order,
908 int *errcode,
909 gfp_t gfp_mask);
910
d0bf4a9e
ED
911/* Layout of fast clones : [skb1][skb2][fclone_ref] */
912struct sk_buff_fclones {
913 struct sk_buff skb1;
914
915 struct sk_buff skb2;
916
917 atomic_t fclone_ref;
918};
919
920/**
921 * skb_fclone_busy - check if fclone is busy
922 * @skb: buffer
923 *
bda13fed 924 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
925 * Some drivers call skb_orphan() in their ndo_start_xmit(),
926 * so we also check that this didnt happen.
d0bf4a9e 927 */
39bb5e62
ED
928static inline bool skb_fclone_busy(const struct sock *sk,
929 const struct sk_buff *skb)
d0bf4a9e
ED
930{
931 const struct sk_buff_fclones *fclones;
932
933 fclones = container_of(skb, struct sk_buff_fclones, skb1);
934
935 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 936 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 937 fclones->skb2.sk == sk;
d0bf4a9e
ED
938}
939
d179cd12 940static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 941 gfp_t priority)
d179cd12 942{
c93bdd0e 943 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
944}
945
7965bd4d 946struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
947static inline struct sk_buff *alloc_skb_head(gfp_t priority)
948{
949 return __alloc_skb_head(priority, -1);
950}
951
7965bd4d
JP
952struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
953int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
954struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
955struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
956struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
957 gfp_t gfp_mask, bool fclone);
958static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
959 gfp_t gfp_mask)
960{
961 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
962}
7965bd4d
JP
963
964int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
965struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
966 unsigned int headroom);
967struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
968 int newtailroom, gfp_t priority);
25a91d8d
FD
969int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
970 int offset, int len);
7965bd4d
JP
971int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
972 int len);
973int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
974int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 975#define dev_kfree_skb(a) consume_skb(a)
1da177e4 976
7965bd4d
JP
977int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
978 int getfrag(void *from, char *to, int offset,
979 int len, int odd, struct sk_buff *skb),
980 void *from, int length);
e89e9cf5 981
be12a1fe
HFS
982int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
983 int offset, size_t size);
984
d94d9fee 985struct skb_seq_state {
677e90ed
TG
986 __u32 lower_offset;
987 __u32 upper_offset;
988 __u32 frag_idx;
989 __u32 stepped_offset;
990 struct sk_buff *root_skb;
991 struct sk_buff *cur_skb;
992 __u8 *frag_data;
993};
994
7965bd4d
JP
995void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
996 unsigned int to, struct skb_seq_state *st);
997unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
998 struct skb_seq_state *st);
999void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1000
7965bd4d 1001unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1002 unsigned int to, struct ts_config *config);
3fc7e8a6 1003
09323cc4
TH
1004/*
1005 * Packet hash types specify the type of hash in skb_set_hash.
1006 *
1007 * Hash types refer to the protocol layer addresses which are used to
1008 * construct a packet's hash. The hashes are used to differentiate or identify
1009 * flows of the protocol layer for the hash type. Hash types are either
1010 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1011 *
1012 * Properties of hashes:
1013 *
1014 * 1) Two packets in different flows have different hash values
1015 * 2) Two packets in the same flow should have the same hash value
1016 *
1017 * A hash at a higher layer is considered to be more specific. A driver should
1018 * set the most specific hash possible.
1019 *
1020 * A driver cannot indicate a more specific hash than the layer at which a hash
1021 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1022 *
1023 * A driver may indicate a hash level which is less specific than the
1024 * actual layer the hash was computed on. For instance, a hash computed
1025 * at L4 may be considered an L3 hash. This should only be done if the
1026 * driver can't unambiguously determine that the HW computed the hash at
1027 * the higher layer. Note that the "should" in the second property above
1028 * permits this.
1029 */
1030enum pkt_hash_types {
1031 PKT_HASH_TYPE_NONE, /* Undefined type */
1032 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1033 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1034 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1035};
1036
bcc83839 1037static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1038{
bcc83839 1039 skb->hash = 0;
a3b18ddb 1040 skb->sw_hash = 0;
bcc83839
TH
1041 skb->l4_hash = 0;
1042}
1043
1044static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1045{
1046 if (!skb->l4_hash)
1047 skb_clear_hash(skb);
1048}
1049
1050static inline void
1051__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1052{
1053 skb->l4_hash = is_l4;
1054 skb->sw_hash = is_sw;
61b905da 1055 skb->hash = hash;
09323cc4
TH
1056}
1057
bcc83839
TH
1058static inline void
1059skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1060{
1061 /* Used by drivers to set hash from HW */
1062 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1063}
1064
1065static inline void
1066__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1067{
1068 __skb_set_hash(skb, hash, true, is_l4);
1069}
1070
e5276937
TH
1071void __skb_get_hash(struct sk_buff *skb);
1072u32 skb_get_poff(const struct sk_buff *skb);
1073u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1074 const struct flow_keys *keys, int hlen);
1075__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1076 void *data, int hlen_proto);
1077
1078static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1079 int thoff, u8 ip_proto)
1080{
1081 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1082}
1083
1084void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1085 const struct flow_dissector_key *key,
1086 unsigned int key_count);
1087
1088bool __skb_flow_dissect(const struct sk_buff *skb,
1089 struct flow_dissector *flow_dissector,
1090 void *target_container,
cd79a238
TH
1091 void *data, __be16 proto, int nhoff, int hlen,
1092 unsigned int flags);
e5276937
TH
1093
1094static inline bool skb_flow_dissect(const struct sk_buff *skb,
1095 struct flow_dissector *flow_dissector,
cd79a238 1096 void *target_container, unsigned int flags)
e5276937
TH
1097{
1098 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1099 NULL, 0, 0, 0, flags);
e5276937
TH
1100}
1101
1102static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1103 struct flow_keys *flow,
1104 unsigned int flags)
e5276937
TH
1105{
1106 memset(flow, 0, sizeof(*flow));
1107 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1108 NULL, 0, 0, 0, flags);
e5276937
TH
1109}
1110
1111static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1112 void *data, __be16 proto,
cd79a238
TH
1113 int nhoff, int hlen,
1114 unsigned int flags)
e5276937
TH
1115{
1116 memset(flow, 0, sizeof(*flow));
1117 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1118 data, proto, nhoff, hlen, flags);
e5276937
TH
1119}
1120
3958afa1 1121static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1122{
a3b18ddb 1123 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1124 __skb_get_hash(skb);
bfb564e7 1125
61b905da 1126 return skb->hash;
bfb564e7
KK
1127}
1128
20a17bf6 1129__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1130
20a17bf6 1131static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1132{
c6cc1ca7
TH
1133 if (!skb->l4_hash && !skb->sw_hash) {
1134 struct flow_keys keys;
de4c1f8b 1135 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1136
de4c1f8b 1137 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1138 }
f70ea018
TH
1139
1140 return skb->hash;
1141}
1142
20a17bf6 1143__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1144
20a17bf6 1145static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1146{
c6cc1ca7
TH
1147 if (!skb->l4_hash && !skb->sw_hash) {
1148 struct flow_keys keys;
de4c1f8b 1149 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1150
de4c1f8b 1151 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1152 }
f70ea018
TH
1153
1154 return skb->hash;
1155}
1156
50fb7992
TH
1157__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1158
57bdf7f4
TH
1159static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1160{
61b905da 1161 return skb->hash;
57bdf7f4
TH
1162}
1163
3df7a74e
TH
1164static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1165{
61b905da 1166 to->hash = from->hash;
a3b18ddb 1167 to->sw_hash = from->sw_hash;
61b905da 1168 to->l4_hash = from->l4_hash;
3df7a74e
TH
1169};
1170
4305b541
ACM
1171#ifdef NET_SKBUFF_DATA_USES_OFFSET
1172static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1173{
1174 return skb->head + skb->end;
1175}
ec47ea82
AD
1176
1177static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1178{
1179 return skb->end;
1180}
4305b541
ACM
1181#else
1182static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1183{
1184 return skb->end;
1185}
ec47ea82
AD
1186
1187static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1188{
1189 return skb->end - skb->head;
1190}
4305b541
ACM
1191#endif
1192
1da177e4 1193/* Internal */
4305b541 1194#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1195
ac45f602
PO
1196static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1197{
1198 return &skb_shinfo(skb)->hwtstamps;
1199}
1200
1da177e4
LT
1201/**
1202 * skb_queue_empty - check if a queue is empty
1203 * @list: queue head
1204 *
1205 * Returns true if the queue is empty, false otherwise.
1206 */
1207static inline int skb_queue_empty(const struct sk_buff_head *list)
1208{
fd44b93c 1209 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1210}
1211
fc7ebb21
DM
1212/**
1213 * skb_queue_is_last - check if skb is the last entry in the queue
1214 * @list: queue head
1215 * @skb: buffer
1216 *
1217 * Returns true if @skb is the last buffer on the list.
1218 */
1219static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1220 const struct sk_buff *skb)
1221{
fd44b93c 1222 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1223}
1224
832d11c5
IJ
1225/**
1226 * skb_queue_is_first - check if skb is the first entry in the queue
1227 * @list: queue head
1228 * @skb: buffer
1229 *
1230 * Returns true if @skb is the first buffer on the list.
1231 */
1232static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1233 const struct sk_buff *skb)
1234{
fd44b93c 1235 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1236}
1237
249c8b42
DM
1238/**
1239 * skb_queue_next - return the next packet in the queue
1240 * @list: queue head
1241 * @skb: current buffer
1242 *
1243 * Return the next packet in @list after @skb. It is only valid to
1244 * call this if skb_queue_is_last() evaluates to false.
1245 */
1246static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1247 const struct sk_buff *skb)
1248{
1249 /* This BUG_ON may seem severe, but if we just return then we
1250 * are going to dereference garbage.
1251 */
1252 BUG_ON(skb_queue_is_last(list, skb));
1253 return skb->next;
1254}
1255
832d11c5
IJ
1256/**
1257 * skb_queue_prev - return the prev packet in the queue
1258 * @list: queue head
1259 * @skb: current buffer
1260 *
1261 * Return the prev packet in @list before @skb. It is only valid to
1262 * call this if skb_queue_is_first() evaluates to false.
1263 */
1264static inline struct sk_buff *skb_queue_prev(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_first(list, skb));
1271 return skb->prev;
1272}
1273
1da177e4
LT
1274/**
1275 * skb_get - reference buffer
1276 * @skb: buffer to reference
1277 *
1278 * Makes another reference to a socket buffer and returns a pointer
1279 * to the buffer.
1280 */
1281static inline struct sk_buff *skb_get(struct sk_buff *skb)
1282{
1283 atomic_inc(&skb->users);
1284 return skb;
1285}
1286
1287/*
1288 * If users == 1, we are the only owner and are can avoid redundant
1289 * atomic change.
1290 */
1291
1da177e4
LT
1292/**
1293 * skb_cloned - is the buffer a clone
1294 * @skb: buffer to check
1295 *
1296 * Returns true if the buffer was generated with skb_clone() and is
1297 * one of multiple shared copies of the buffer. Cloned buffers are
1298 * shared data so must not be written to under normal circumstances.
1299 */
1300static inline int skb_cloned(const struct sk_buff *skb)
1301{
1302 return skb->cloned &&
1303 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1304}
1305
14bbd6a5
PS
1306static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1307{
d0164adc 1308 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1309
1310 if (skb_cloned(skb))
1311 return pskb_expand_head(skb, 0, 0, pri);
1312
1313 return 0;
1314}
1315
1da177e4
LT
1316/**
1317 * skb_header_cloned - is the header a clone
1318 * @skb: buffer to check
1319 *
1320 * Returns true if modifying the header part of the buffer requires
1321 * the data to be copied.
1322 */
1323static inline int skb_header_cloned(const struct sk_buff *skb)
1324{
1325 int dataref;
1326
1327 if (!skb->cloned)
1328 return 0;
1329
1330 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1331 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1332 return dataref != 1;
1333}
1334
9580bf2e
ED
1335static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1336{
1337 might_sleep_if(gfpflags_allow_blocking(pri));
1338
1339 if (skb_header_cloned(skb))
1340 return pskb_expand_head(skb, 0, 0, pri);
1341
1342 return 0;
1343}
1344
1da177e4
LT
1345/**
1346 * skb_header_release - release reference to header
1347 * @skb: buffer to operate on
1348 *
1349 * Drop a reference to the header part of the buffer. This is done
1350 * by acquiring a payload reference. You must not read from the header
1351 * part of skb->data after this.
f4a775d1 1352 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1353 */
1354static inline void skb_header_release(struct sk_buff *skb)
1355{
1356 BUG_ON(skb->nohdr);
1357 skb->nohdr = 1;
1358 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1359}
1360
f4a775d1
ED
1361/**
1362 * __skb_header_release - release reference to header
1363 * @skb: buffer to operate on
1364 *
1365 * Variant of skb_header_release() assuming skb is private to caller.
1366 * We can avoid one atomic operation.
1367 */
1368static inline void __skb_header_release(struct sk_buff *skb)
1369{
1370 skb->nohdr = 1;
1371 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1372}
1373
1374
1da177e4
LT
1375/**
1376 * skb_shared - is the buffer shared
1377 * @skb: buffer to check
1378 *
1379 * Returns true if more than one person has a reference to this
1380 * buffer.
1381 */
1382static inline int skb_shared(const struct sk_buff *skb)
1383{
1384 return atomic_read(&skb->users) != 1;
1385}
1386
1387/**
1388 * skb_share_check - check if buffer is shared and if so clone it
1389 * @skb: buffer to check
1390 * @pri: priority for memory allocation
1391 *
1392 * If the buffer is shared the buffer is cloned and the old copy
1393 * drops a reference. A new clone with a single reference is returned.
1394 * If the buffer is not shared the original buffer is returned. When
1395 * being called from interrupt status or with spinlocks held pri must
1396 * be GFP_ATOMIC.
1397 *
1398 * NULL is returned on a memory allocation failure.
1399 */
47061bc4 1400static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1401{
d0164adc 1402 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1403 if (skb_shared(skb)) {
1404 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1405
1406 if (likely(nskb))
1407 consume_skb(skb);
1408 else
1409 kfree_skb(skb);
1da177e4
LT
1410 skb = nskb;
1411 }
1412 return skb;
1413}
1414
1415/*
1416 * Copy shared buffers into a new sk_buff. We effectively do COW on
1417 * packets to handle cases where we have a local reader and forward
1418 * and a couple of other messy ones. The normal one is tcpdumping
1419 * a packet thats being forwarded.
1420 */
1421
1422/**
1423 * skb_unshare - make a copy of a shared buffer
1424 * @skb: buffer to check
1425 * @pri: priority for memory allocation
1426 *
1427 * If the socket buffer is a clone then this function creates a new
1428 * copy of the data, drops a reference count on the old copy and returns
1429 * the new copy with the reference count at 1. If the buffer is not a clone
1430 * the original buffer is returned. When called with a spinlock held or
1431 * from interrupt state @pri must be %GFP_ATOMIC
1432 *
1433 * %NULL is returned on a memory allocation failure.
1434 */
e2bf521d 1435static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1436 gfp_t pri)
1da177e4 1437{
d0164adc 1438 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1439 if (skb_cloned(skb)) {
1440 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1441
1442 /* Free our shared copy */
1443 if (likely(nskb))
1444 consume_skb(skb);
1445 else
1446 kfree_skb(skb);
1da177e4
LT
1447 skb = nskb;
1448 }
1449 return skb;
1450}
1451
1452/**
1a5778aa 1453 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1454 * @list_: list to peek at
1455 *
1456 * Peek an &sk_buff. Unlike most other operations you _MUST_
1457 * be careful with this one. A peek leaves the buffer on the
1458 * list and someone else may run off with it. You must hold
1459 * the appropriate locks or have a private queue to do this.
1460 *
1461 * Returns %NULL for an empty list or a pointer to the head element.
1462 * The reference count is not incremented and the reference is therefore
1463 * volatile. Use with caution.
1464 */
05bdd2f1 1465static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1466{
18d07000
ED
1467 struct sk_buff *skb = list_->next;
1468
1469 if (skb == (struct sk_buff *)list_)
1470 skb = NULL;
1471 return skb;
1da177e4
LT
1472}
1473
da5ef6e5
PE
1474/**
1475 * skb_peek_next - peek skb following the given one from a queue
1476 * @skb: skb to start from
1477 * @list_: list to peek at
1478 *
1479 * Returns %NULL when the end of the list is met or a pointer to the
1480 * next element. The reference count is not incremented and the
1481 * reference is therefore volatile. Use with caution.
1482 */
1483static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1484 const struct sk_buff_head *list_)
1485{
1486 struct sk_buff *next = skb->next;
18d07000 1487
da5ef6e5
PE
1488 if (next == (struct sk_buff *)list_)
1489 next = NULL;
1490 return next;
1491}
1492
1da177e4 1493/**
1a5778aa 1494 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1495 * @list_: list to peek at
1496 *
1497 * Peek an &sk_buff. Unlike most other operations you _MUST_
1498 * be careful with this one. A peek leaves the buffer on the
1499 * list and someone else may run off with it. You must hold
1500 * the appropriate locks or have a private queue to do this.
1501 *
1502 * Returns %NULL for an empty list or a pointer to the tail element.
1503 * The reference count is not incremented and the reference is therefore
1504 * volatile. Use with caution.
1505 */
05bdd2f1 1506static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1507{
18d07000
ED
1508 struct sk_buff *skb = list_->prev;
1509
1510 if (skb == (struct sk_buff *)list_)
1511 skb = NULL;
1512 return skb;
1513
1da177e4
LT
1514}
1515
1516/**
1517 * skb_queue_len - get queue length
1518 * @list_: list to measure
1519 *
1520 * Return the length of an &sk_buff queue.
1521 */
1522static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1523{
1524 return list_->qlen;
1525}
1526
67fed459
DM
1527/**
1528 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1529 * @list: queue to initialize
1530 *
1531 * This initializes only the list and queue length aspects of
1532 * an sk_buff_head object. This allows to initialize the list
1533 * aspects of an sk_buff_head without reinitializing things like
1534 * the spinlock. It can also be used for on-stack sk_buff_head
1535 * objects where the spinlock is known to not be used.
1536 */
1537static inline void __skb_queue_head_init(struct sk_buff_head *list)
1538{
1539 list->prev = list->next = (struct sk_buff *)list;
1540 list->qlen = 0;
1541}
1542
76f10ad0
AV
1543/*
1544 * This function creates a split out lock class for each invocation;
1545 * this is needed for now since a whole lot of users of the skb-queue
1546 * infrastructure in drivers have different locking usage (in hardirq)
1547 * than the networking core (in softirq only). In the long run either the
1548 * network layer or drivers should need annotation to consolidate the
1549 * main types of usage into 3 classes.
1550 */
1da177e4
LT
1551static inline void skb_queue_head_init(struct sk_buff_head *list)
1552{
1553 spin_lock_init(&list->lock);
67fed459 1554 __skb_queue_head_init(list);
1da177e4
LT
1555}
1556
c2ecba71
PE
1557static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1558 struct lock_class_key *class)
1559{
1560 skb_queue_head_init(list);
1561 lockdep_set_class(&list->lock, class);
1562}
1563
1da177e4 1564/*
bf299275 1565 * Insert an sk_buff on a list.
1da177e4
LT
1566 *
1567 * The "__skb_xxxx()" functions are the non-atomic ones that
1568 * can only be called with interrupts disabled.
1569 */
7965bd4d
JP
1570void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1571 struct sk_buff_head *list);
bf299275
GR
1572static inline void __skb_insert(struct sk_buff *newsk,
1573 struct sk_buff *prev, struct sk_buff *next,
1574 struct sk_buff_head *list)
1575{
1576 newsk->next = next;
1577 newsk->prev = prev;
1578 next->prev = prev->next = newsk;
1579 list->qlen++;
1580}
1da177e4 1581
67fed459
DM
1582static inline void __skb_queue_splice(const struct sk_buff_head *list,
1583 struct sk_buff *prev,
1584 struct sk_buff *next)
1585{
1586 struct sk_buff *first = list->next;
1587 struct sk_buff *last = list->prev;
1588
1589 first->prev = prev;
1590 prev->next = first;
1591
1592 last->next = next;
1593 next->prev = last;
1594}
1595
1596/**
1597 * skb_queue_splice - join two skb lists, this is designed for stacks
1598 * @list: the new list to add
1599 * @head: the place to add it in the first list
1600 */
1601static inline void skb_queue_splice(const struct sk_buff_head *list,
1602 struct sk_buff_head *head)
1603{
1604 if (!skb_queue_empty(list)) {
1605 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1606 head->qlen += list->qlen;
67fed459
DM
1607 }
1608}
1609
1610/**
d9619496 1611 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1612 * @list: the new list to add
1613 * @head: the place to add it in the first list
1614 *
1615 * The list at @list is reinitialised
1616 */
1617static inline void skb_queue_splice_init(struct sk_buff_head *list,
1618 struct sk_buff_head *head)
1619{
1620 if (!skb_queue_empty(list)) {
1621 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1622 head->qlen += list->qlen;
67fed459
DM
1623 __skb_queue_head_init(list);
1624 }
1625}
1626
1627/**
1628 * skb_queue_splice_tail - join two skb lists, each list being a queue
1629 * @list: the new list to add
1630 * @head: the place to add it in the first list
1631 */
1632static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1633 struct sk_buff_head *head)
1634{
1635 if (!skb_queue_empty(list)) {
1636 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1637 head->qlen += list->qlen;
67fed459
DM
1638 }
1639}
1640
1641/**
d9619496 1642 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1643 * @list: the new list to add
1644 * @head: the place to add it in the first list
1645 *
1646 * Each of the lists is a queue.
1647 * The list at @list is reinitialised
1648 */
1649static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1650 struct sk_buff_head *head)
1651{
1652 if (!skb_queue_empty(list)) {
1653 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1654 head->qlen += list->qlen;
67fed459
DM
1655 __skb_queue_head_init(list);
1656 }
1657}
1658
1da177e4 1659/**
300ce174 1660 * __skb_queue_after - queue a buffer at the list head
1da177e4 1661 * @list: list to use
300ce174 1662 * @prev: place after this buffer
1da177e4
LT
1663 * @newsk: buffer to queue
1664 *
300ce174 1665 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1666 * and you must therefore hold required locks before calling it.
1667 *
1668 * A buffer cannot be placed on two lists at the same time.
1669 */
300ce174
SH
1670static inline void __skb_queue_after(struct sk_buff_head *list,
1671 struct sk_buff *prev,
1672 struct sk_buff *newsk)
1da177e4 1673{
bf299275 1674 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1675}
1676
7965bd4d
JP
1677void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1678 struct sk_buff_head *list);
7de6c033 1679
f5572855
GR
1680static inline void __skb_queue_before(struct sk_buff_head *list,
1681 struct sk_buff *next,
1682 struct sk_buff *newsk)
1683{
1684 __skb_insert(newsk, next->prev, next, list);
1685}
1686
300ce174
SH
1687/**
1688 * __skb_queue_head - queue a buffer at the list head
1689 * @list: list to use
1690 * @newsk: buffer to queue
1691 *
1692 * Queue a buffer at the start of a list. This function takes no locks
1693 * and you must therefore hold required locks before calling it.
1694 *
1695 * A buffer cannot be placed on two lists at the same time.
1696 */
7965bd4d 1697void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1698static inline void __skb_queue_head(struct sk_buff_head *list,
1699 struct sk_buff *newsk)
1700{
1701 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1702}
1703
1da177e4
LT
1704/**
1705 * __skb_queue_tail - queue a buffer at the list tail
1706 * @list: list to use
1707 * @newsk: buffer to queue
1708 *
1709 * Queue a buffer at the end of a list. This function takes no locks
1710 * and you must therefore hold required locks before calling it.
1711 *
1712 * A buffer cannot be placed on two lists at the same time.
1713 */
7965bd4d 1714void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1715static inline void __skb_queue_tail(struct sk_buff_head *list,
1716 struct sk_buff *newsk)
1717{
f5572855 1718 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1719}
1720
1da177e4
LT
1721/*
1722 * remove sk_buff from list. _Must_ be called atomically, and with
1723 * the list known..
1724 */
7965bd4d 1725void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1726static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1727{
1728 struct sk_buff *next, *prev;
1729
1730 list->qlen--;
1731 next = skb->next;
1732 prev = skb->prev;
1733 skb->next = skb->prev = NULL;
1da177e4
LT
1734 next->prev = prev;
1735 prev->next = next;
1736}
1737
f525c06d
GR
1738/**
1739 * __skb_dequeue - remove from the head of the queue
1740 * @list: list to dequeue from
1741 *
1742 * Remove the head of the list. This function does not take any locks
1743 * so must be used with appropriate locks held only. The head item is
1744 * returned or %NULL if the list is empty.
1745 */
7965bd4d 1746struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1747static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1748{
1749 struct sk_buff *skb = skb_peek(list);
1750 if (skb)
1751 __skb_unlink(skb, list);
1752 return skb;
1753}
1da177e4
LT
1754
1755/**
1756 * __skb_dequeue_tail - remove from the tail of the queue
1757 * @list: list to dequeue from
1758 *
1759 * Remove the tail of the list. This function does not take any locks
1760 * so must be used with appropriate locks held only. The tail item is
1761 * returned or %NULL if the list is empty.
1762 */
7965bd4d 1763struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1764static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1765{
1766 struct sk_buff *skb = skb_peek_tail(list);
1767 if (skb)
1768 __skb_unlink(skb, list);
1769 return skb;
1770}
1771
1772
bdcc0924 1773static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1774{
1775 return skb->data_len;
1776}
1777
1778static inline unsigned int skb_headlen(const struct sk_buff *skb)
1779{
1780 return skb->len - skb->data_len;
1781}
1782
1783static inline int skb_pagelen(const struct sk_buff *skb)
1784{
1785 int i, len = 0;
1786
1787 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1788 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1789 return len + skb_headlen(skb);
1790}
1791
131ea667
IC
1792/**
1793 * __skb_fill_page_desc - initialise a paged fragment in an skb
1794 * @skb: buffer containing fragment to be initialised
1795 * @i: paged fragment index to initialise
1796 * @page: the page to use for this fragment
1797 * @off: the offset to the data with @page
1798 * @size: the length of the data
1799 *
1800 * Initialises the @i'th fragment of @skb to point to &size bytes at
1801 * offset @off within @page.
1802 *
1803 * Does not take any additional reference on the fragment.
1804 */
1805static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1806 struct page *page, int off, int size)
1da177e4
LT
1807{
1808 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1809
c48a11c7 1810 /*
2f064f34
MH
1811 * Propagate page pfmemalloc to the skb if we can. The problem is
1812 * that not all callers have unique ownership of the page but rely
1813 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1814 */
a8605c60 1815 frag->page.p = page;
1da177e4 1816 frag->page_offset = off;
9e903e08 1817 skb_frag_size_set(frag, size);
cca7af38
PE
1818
1819 page = compound_head(page);
2f064f34 1820 if (page_is_pfmemalloc(page))
cca7af38 1821 skb->pfmemalloc = true;
131ea667
IC
1822}
1823
1824/**
1825 * skb_fill_page_desc - initialise a paged fragment in an skb
1826 * @skb: buffer containing fragment to be initialised
1827 * @i: paged fragment index to initialise
1828 * @page: the page to use for this fragment
1829 * @off: the offset to the data with @page
1830 * @size: the length of the data
1831 *
1832 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1833 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1834 * addition updates @skb such that @i is the last fragment.
1835 *
1836 * Does not take any additional reference on the fragment.
1837 */
1838static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1839 struct page *page, int off, int size)
1840{
1841 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1842 skb_shinfo(skb)->nr_frags = i + 1;
1843}
1844
7965bd4d
JP
1845void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1846 int size, unsigned int truesize);
654bed16 1847
f8e617e1
JW
1848void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1849 unsigned int truesize);
1850
1da177e4 1851#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1852#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1853#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1854
27a884dc
ACM
1855#ifdef NET_SKBUFF_DATA_USES_OFFSET
1856static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1857{
1858 return skb->head + skb->tail;
1859}
1860
1861static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1862{
1863 skb->tail = skb->data - skb->head;
1864}
1865
1866static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1867{
1868 skb_reset_tail_pointer(skb);
1869 skb->tail += offset;
1870}
7cc46190 1871
27a884dc
ACM
1872#else /* NET_SKBUFF_DATA_USES_OFFSET */
1873static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1874{
1875 return skb->tail;
1876}
1877
1878static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1879{
1880 skb->tail = skb->data;
1881}
1882
1883static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1884{
1885 skb->tail = skb->data + offset;
1886}
4305b541 1887
27a884dc
ACM
1888#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1889
1da177e4
LT
1890/*
1891 * Add data to an sk_buff
1892 */
0c7ddf36 1893unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1894unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1895static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1896{
27a884dc 1897 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1898 SKB_LINEAR_ASSERT(skb);
1899 skb->tail += len;
1900 skb->len += len;
1901 return tmp;
1902}
1903
7965bd4d 1904unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1905static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1906{
1907 skb->data -= len;
1908 skb->len += len;
1909 return skb->data;
1910}
1911
7965bd4d 1912unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1913static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1914{
1915 skb->len -= len;
1916 BUG_ON(skb->len < skb->data_len);
1917 return skb->data += len;
1918}
1919
47d29646
DM
1920static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1921{
1922 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1923}
1924
7965bd4d 1925unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1926
1927static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1928{
1929 if (len > skb_headlen(skb) &&
987c402a 1930 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1931 return NULL;
1932 skb->len -= len;
1933 return skb->data += len;
1934}
1935
1936static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1937{
1938 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1939}
1940
1941static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1942{
1943 if (likely(len <= skb_headlen(skb)))
1944 return 1;
1945 if (unlikely(len > skb->len))
1946 return 0;
987c402a 1947 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1948}
1949
1950/**
1951 * skb_headroom - bytes at buffer head
1952 * @skb: buffer to check
1953 *
1954 * Return the number of bytes of free space at the head of an &sk_buff.
1955 */
c2636b4d 1956static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1957{
1958 return skb->data - skb->head;
1959}
1960
1961/**
1962 * skb_tailroom - bytes at buffer end
1963 * @skb: buffer to check
1964 *
1965 * Return the number of bytes of free space at the tail of an sk_buff
1966 */
1967static inline int skb_tailroom(const struct sk_buff *skb)
1968{
4305b541 1969 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1970}
1971
a21d4572
ED
1972/**
1973 * skb_availroom - bytes at buffer end
1974 * @skb: buffer to check
1975 *
1976 * Return the number of bytes of free space at the tail of an sk_buff
1977 * allocated by sk_stream_alloc()
1978 */
1979static inline int skb_availroom(const struct sk_buff *skb)
1980{
16fad69c
ED
1981 if (skb_is_nonlinear(skb))
1982 return 0;
1983
1984 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1985}
1986
1da177e4
LT
1987/**
1988 * skb_reserve - adjust headroom
1989 * @skb: buffer to alter
1990 * @len: bytes to move
1991 *
1992 * Increase the headroom of an empty &sk_buff by reducing the tail
1993 * room. This is only allowed for an empty buffer.
1994 */
8243126c 1995static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1996{
1997 skb->data += len;
1998 skb->tail += len;
1999}
2000
1837b2e2
BP
2001/**
2002 * skb_tailroom_reserve - adjust reserved_tailroom
2003 * @skb: buffer to alter
2004 * @mtu: maximum amount of headlen permitted
2005 * @needed_tailroom: minimum amount of reserved_tailroom
2006 *
2007 * Set reserved_tailroom so that headlen can be as large as possible but
2008 * not larger than mtu and tailroom cannot be smaller than
2009 * needed_tailroom.
2010 * The required headroom should already have been reserved before using
2011 * this function.
2012 */
2013static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2014 unsigned int needed_tailroom)
2015{
2016 SKB_LINEAR_ASSERT(skb);
2017 if (mtu < skb_tailroom(skb) - needed_tailroom)
2018 /* use at most mtu */
2019 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2020 else
2021 /* use up to all available space */
2022 skb->reserved_tailroom = needed_tailroom;
2023}
2024
8bce6d7d
TH
2025#define ENCAP_TYPE_ETHER 0
2026#define ENCAP_TYPE_IPPROTO 1
2027
2028static inline void skb_set_inner_protocol(struct sk_buff *skb,
2029 __be16 protocol)
2030{
2031 skb->inner_protocol = protocol;
2032 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2033}
2034
2035static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2036 __u8 ipproto)
2037{
2038 skb->inner_ipproto = ipproto;
2039 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2040}
2041
6a674e9c
JG
2042static inline void skb_reset_inner_headers(struct sk_buff *skb)
2043{
aefbd2b3 2044 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2045 skb->inner_network_header = skb->network_header;
2046 skb->inner_transport_header = skb->transport_header;
2047}
2048
0b5c9db1
JP
2049static inline void skb_reset_mac_len(struct sk_buff *skb)
2050{
2051 skb->mac_len = skb->network_header - skb->mac_header;
2052}
2053
6a674e9c
JG
2054static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2055 *skb)
2056{
2057 return skb->head + skb->inner_transport_header;
2058}
2059
55dc5a9f
TH
2060static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2061{
2062 return skb_inner_transport_header(skb) - skb->data;
2063}
2064
6a674e9c
JG
2065static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2066{
2067 skb->inner_transport_header = skb->data - skb->head;
2068}
2069
2070static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2071 const int offset)
2072{
2073 skb_reset_inner_transport_header(skb);
2074 skb->inner_transport_header += offset;
2075}
2076
2077static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2078{
2079 return skb->head + skb->inner_network_header;
2080}
2081
2082static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2083{
2084 skb->inner_network_header = skb->data - skb->head;
2085}
2086
2087static inline void skb_set_inner_network_header(struct sk_buff *skb,
2088 const int offset)
2089{
2090 skb_reset_inner_network_header(skb);
2091 skb->inner_network_header += offset;
2092}
2093
aefbd2b3
PS
2094static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2095{
2096 return skb->head + skb->inner_mac_header;
2097}
2098
2099static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2100{
2101 skb->inner_mac_header = skb->data - skb->head;
2102}
2103
2104static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2105 const int offset)
2106{
2107 skb_reset_inner_mac_header(skb);
2108 skb->inner_mac_header += offset;
2109}
fda55eca
ED
2110static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2111{
35d04610 2112 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2113}
2114
9c70220b
ACM
2115static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2116{
2e07fa9c 2117 return skb->head + skb->transport_header;
9c70220b
ACM
2118}
2119
badff6d0
ACM
2120static inline void skb_reset_transport_header(struct sk_buff *skb)
2121{
2e07fa9c 2122 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2123}
2124
967b05f6
ACM
2125static inline void skb_set_transport_header(struct sk_buff *skb,
2126 const int offset)
2127{
2e07fa9c
ACM
2128 skb_reset_transport_header(skb);
2129 skb->transport_header += offset;
ea2ae17d
ACM
2130}
2131
d56f90a7
ACM
2132static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2133{
2e07fa9c 2134 return skb->head + skb->network_header;
d56f90a7
ACM
2135}
2136
c1d2bbe1
ACM
2137static inline void skb_reset_network_header(struct sk_buff *skb)
2138{
2e07fa9c 2139 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2140}
2141
c14d2450
ACM
2142static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2143{
2e07fa9c
ACM
2144 skb_reset_network_header(skb);
2145 skb->network_header += offset;
c14d2450
ACM
2146}
2147
2e07fa9c 2148static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2149{
2e07fa9c 2150 return skb->head + skb->mac_header;
bbe735e4
ACM
2151}
2152
2e07fa9c 2153static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2154{
35d04610 2155 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2156}
2157
2158static inline void skb_reset_mac_header(struct sk_buff *skb)
2159{
2160 skb->mac_header = skb->data - skb->head;
2161}
2162
2163static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2164{
2165 skb_reset_mac_header(skb);
2166 skb->mac_header += offset;
2167}
2168
0e3da5bb
TT
2169static inline void skb_pop_mac_header(struct sk_buff *skb)
2170{
2171 skb->mac_header = skb->network_header;
2172}
2173
fbbdb8f0
YX
2174static inline void skb_probe_transport_header(struct sk_buff *skb,
2175 const int offset_hint)
2176{
2177 struct flow_keys keys;
2178
2179 if (skb_transport_header_was_set(skb))
2180 return;
cd79a238 2181 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2182 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2183 else
2184 skb_set_transport_header(skb, offset_hint);
2185}
2186
03606895
ED
2187static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2188{
2189 if (skb_mac_header_was_set(skb)) {
2190 const unsigned char *old_mac = skb_mac_header(skb);
2191
2192 skb_set_mac_header(skb, -skb->mac_len);
2193 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2194 }
2195}
2196
04fb451e
MM
2197static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2198{
2199 return skb->csum_start - skb_headroom(skb);
2200}
2201
08b64fcc
AD
2202static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2203{
2204 return skb->head + skb->csum_start;
2205}
2206
2e07fa9c
ACM
2207static inline int skb_transport_offset(const struct sk_buff *skb)
2208{
2209 return skb_transport_header(skb) - skb->data;
2210}
2211
2212static inline u32 skb_network_header_len(const struct sk_buff *skb)
2213{
2214 return skb->transport_header - skb->network_header;
2215}
2216
6a674e9c
JG
2217static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2218{
2219 return skb->inner_transport_header - skb->inner_network_header;
2220}
2221
2e07fa9c
ACM
2222static inline int skb_network_offset(const struct sk_buff *skb)
2223{
2224 return skb_network_header(skb) - skb->data;
2225}
48d49d0c 2226
6a674e9c
JG
2227static inline int skb_inner_network_offset(const struct sk_buff *skb)
2228{
2229 return skb_inner_network_header(skb) - skb->data;
2230}
2231
f9599ce1
CG
2232static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2233{
2234 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2235}
2236
1da177e4
LT
2237/*
2238 * CPUs often take a performance hit when accessing unaligned memory
2239 * locations. The actual performance hit varies, it can be small if the
2240 * hardware handles it or large if we have to take an exception and fix it
2241 * in software.
2242 *
2243 * Since an ethernet header is 14 bytes network drivers often end up with
2244 * the IP header at an unaligned offset. The IP header can be aligned by
2245 * shifting the start of the packet by 2 bytes. Drivers should do this
2246 * with:
2247 *
8660c124 2248 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2249 *
2250 * The downside to this alignment of the IP header is that the DMA is now
2251 * unaligned. On some architectures the cost of an unaligned DMA is high
2252 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2253 *
1da177e4
LT
2254 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2255 * to be overridden.
2256 */
2257#ifndef NET_IP_ALIGN
2258#define NET_IP_ALIGN 2
2259#endif
2260
025be81e
AB
2261/*
2262 * The networking layer reserves some headroom in skb data (via
2263 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2264 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2265 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2266 *
2267 * Unfortunately this headroom changes the DMA alignment of the resulting
2268 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2269 * on some architectures. An architecture can override this value,
2270 * perhaps setting it to a cacheline in size (since that will maintain
2271 * cacheline alignment of the DMA). It must be a power of 2.
2272 *
d6301d3d 2273 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2274 * headroom, you should not reduce this.
5933dd2f
ED
2275 *
2276 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2277 * to reduce average number of cache lines per packet.
2278 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2279 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2280 */
2281#ifndef NET_SKB_PAD
5933dd2f 2282#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2283#endif
2284
7965bd4d 2285int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2286
2287static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2288{
c4264f27 2289 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2290 WARN_ON(1);
2291 return;
2292 }
27a884dc
ACM
2293 skb->len = len;
2294 skb_set_tail_pointer(skb, len);
1da177e4
LT
2295}
2296
7965bd4d 2297void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2298
2299static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2300{
3cc0e873
HX
2301 if (skb->data_len)
2302 return ___pskb_trim(skb, len);
2303 __skb_trim(skb, len);
2304 return 0;
1da177e4
LT
2305}
2306
2307static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2308{
2309 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2310}
2311
e9fa4f7b
HX
2312/**
2313 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2314 * @skb: buffer to alter
2315 * @len: new length
2316 *
2317 * This is identical to pskb_trim except that the caller knows that
2318 * the skb is not cloned so we should never get an error due to out-
2319 * of-memory.
2320 */
2321static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2322{
2323 int err = pskb_trim(skb, len);
2324 BUG_ON(err);
2325}
2326
1da177e4
LT
2327/**
2328 * skb_orphan - orphan a buffer
2329 * @skb: buffer to orphan
2330 *
2331 * If a buffer currently has an owner then we call the owner's
2332 * destructor function and make the @skb unowned. The buffer continues
2333 * to exist but is no longer charged to its former owner.
2334 */
2335static inline void skb_orphan(struct sk_buff *skb)
2336{
c34a7612 2337 if (skb->destructor) {
1da177e4 2338 skb->destructor(skb);
c34a7612
ED
2339 skb->destructor = NULL;
2340 skb->sk = NULL;
376c7311
ED
2341 } else {
2342 BUG_ON(skb->sk);
c34a7612 2343 }
1da177e4
LT
2344}
2345
a353e0ce
MT
2346/**
2347 * skb_orphan_frags - orphan the frags contained in a buffer
2348 * @skb: buffer to orphan frags from
2349 * @gfp_mask: allocation mask for replacement pages
2350 *
2351 * For each frag in the SKB which needs a destructor (i.e. has an
2352 * owner) create a copy of that frag and release the original
2353 * page by calling the destructor.
2354 */
2355static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2356{
2357 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2358 return 0;
2359 return skb_copy_ubufs(skb, gfp_mask);
2360}
2361
1da177e4
LT
2362/**
2363 * __skb_queue_purge - empty a list
2364 * @list: list to empty
2365 *
2366 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2367 * the list and one reference dropped. This function does not take the
2368 * list lock and the caller must hold the relevant locks to use it.
2369 */
7965bd4d 2370void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2371static inline void __skb_queue_purge(struct sk_buff_head *list)
2372{
2373 struct sk_buff *skb;
2374 while ((skb = __skb_dequeue(list)) != NULL)
2375 kfree_skb(skb);
2376}
2377
7965bd4d 2378void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2379
7965bd4d
JP
2380struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2381 gfp_t gfp_mask);
8af27456
CH
2382
2383/**
2384 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2385 * @dev: network device to receive on
2386 * @length: length to allocate
2387 *
2388 * Allocate a new &sk_buff and assign it a usage count of one. The
2389 * buffer has unspecified headroom built in. Users should allocate
2390 * the headroom they think they need without accounting for the
2391 * built in space. The built in space is used for optimisations.
2392 *
2393 * %NULL is returned if there is no free memory. Although this function
2394 * allocates memory it can be called from an interrupt.
2395 */
2396static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2397 unsigned int length)
8af27456
CH
2398{
2399 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2400}
2401
6f532612
ED
2402/* legacy helper around __netdev_alloc_skb() */
2403static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2404 gfp_t gfp_mask)
2405{
2406 return __netdev_alloc_skb(NULL, length, gfp_mask);
2407}
2408
2409/* legacy helper around netdev_alloc_skb() */
2410static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2411{
2412 return netdev_alloc_skb(NULL, length);
2413}
2414
2415
4915a0de
ED
2416static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2417 unsigned int length, gfp_t gfp)
61321bbd 2418{
4915a0de 2419 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2420
2421 if (NET_IP_ALIGN && skb)
2422 skb_reserve(skb, NET_IP_ALIGN);
2423 return skb;
2424}
2425
4915a0de
ED
2426static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2427 unsigned int length)
2428{
2429 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2430}
2431
181edb2b
AD
2432static inline void skb_free_frag(void *addr)
2433{
2434 __free_page_frag(addr);
2435}
2436
ffde7328 2437void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2438struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2439 unsigned int length, gfp_t gfp_mask);
2440static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2441 unsigned int length)
2442{
2443 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2444}
795bb1c0
JDB
2445void napi_consume_skb(struct sk_buff *skb, int budget);
2446
2447void __kfree_skb_flush(void);
15fad714 2448void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2449
71dfda58
AD
2450/**
2451 * __dev_alloc_pages - allocate page for network Rx
2452 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2453 * @order: size of the allocation
2454 *
2455 * Allocate a new page.
2456 *
2457 * %NULL is returned if there is no free memory.
2458*/
2459static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2460 unsigned int order)
2461{
2462 /* This piece of code contains several assumptions.
2463 * 1. This is for device Rx, therefor a cold page is preferred.
2464 * 2. The expectation is the user wants a compound page.
2465 * 3. If requesting a order 0 page it will not be compound
2466 * due to the check to see if order has a value in prep_new_page
2467 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2468 * code in gfp_to_alloc_flags that should be enforcing this.
2469 */
2470 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2471
2472 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2473}
2474
2475static inline struct page *dev_alloc_pages(unsigned int order)
2476{
95829b3a 2477 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2478}
2479
2480/**
2481 * __dev_alloc_page - allocate a page for network Rx
2482 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2483 *
2484 * Allocate a new page.
2485 *
2486 * %NULL is returned if there is no free memory.
2487 */
2488static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2489{
2490 return __dev_alloc_pages(gfp_mask, 0);
2491}
2492
2493static inline struct page *dev_alloc_page(void)
2494{
95829b3a 2495 return dev_alloc_pages(0);
71dfda58
AD
2496}
2497
0614002b
MG
2498/**
2499 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2500 * @page: The page that was allocated from skb_alloc_page
2501 * @skb: The skb that may need pfmemalloc set
2502 */
2503static inline void skb_propagate_pfmemalloc(struct page *page,
2504 struct sk_buff *skb)
2505{
2f064f34 2506 if (page_is_pfmemalloc(page))
0614002b
MG
2507 skb->pfmemalloc = true;
2508}
2509
131ea667 2510/**
e227867f 2511 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2512 * @frag: the paged fragment
2513 *
2514 * Returns the &struct page associated with @frag.
2515 */
2516static inline struct page *skb_frag_page(const skb_frag_t *frag)
2517{
a8605c60 2518 return frag->page.p;
131ea667
IC
2519}
2520
2521/**
2522 * __skb_frag_ref - take an addition reference on a paged fragment.
2523 * @frag: the paged fragment
2524 *
2525 * Takes an additional reference on the paged fragment @frag.
2526 */
2527static inline void __skb_frag_ref(skb_frag_t *frag)
2528{
2529 get_page(skb_frag_page(frag));
2530}
2531
2532/**
2533 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2534 * @skb: the buffer
2535 * @f: the fragment offset.
2536 *
2537 * Takes an additional reference on the @f'th paged fragment of @skb.
2538 */
2539static inline void skb_frag_ref(struct sk_buff *skb, int f)
2540{
2541 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2542}
2543
2544/**
2545 * __skb_frag_unref - release a reference on a paged fragment.
2546 * @frag: the paged fragment
2547 *
2548 * Releases a reference on the paged fragment @frag.
2549 */
2550static inline void __skb_frag_unref(skb_frag_t *frag)
2551{
2552 put_page(skb_frag_page(frag));
2553}
2554
2555/**
2556 * skb_frag_unref - release a reference on a paged fragment of an skb.
2557 * @skb: the buffer
2558 * @f: the fragment offset
2559 *
2560 * Releases a reference on the @f'th paged fragment of @skb.
2561 */
2562static inline void skb_frag_unref(struct sk_buff *skb, int f)
2563{
2564 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2565}
2566
2567/**
2568 * skb_frag_address - gets the address of the data contained in a paged fragment
2569 * @frag: the paged fragment buffer
2570 *
2571 * Returns the address of the data within @frag. The page must already
2572 * be mapped.
2573 */
2574static inline void *skb_frag_address(const skb_frag_t *frag)
2575{
2576 return page_address(skb_frag_page(frag)) + frag->page_offset;
2577}
2578
2579/**
2580 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2581 * @frag: the paged fragment buffer
2582 *
2583 * Returns the address of the data within @frag. Checks that the page
2584 * is mapped and returns %NULL otherwise.
2585 */
2586static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2587{
2588 void *ptr = page_address(skb_frag_page(frag));
2589 if (unlikely(!ptr))
2590 return NULL;
2591
2592 return ptr + frag->page_offset;
2593}
2594
2595/**
2596 * __skb_frag_set_page - sets the page contained in a paged fragment
2597 * @frag: the paged fragment
2598 * @page: the page to set
2599 *
2600 * Sets the fragment @frag to contain @page.
2601 */
2602static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2603{
a8605c60 2604 frag->page.p = page;
131ea667
IC
2605}
2606
2607/**
2608 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2609 * @skb: the buffer
2610 * @f: the fragment offset
2611 * @page: the page to set
2612 *
2613 * Sets the @f'th fragment of @skb to contain @page.
2614 */
2615static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2616 struct page *page)
2617{
2618 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2619}
2620
400dfd3a
ED
2621bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2622
131ea667
IC
2623/**
2624 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2625 * @dev: the device to map the fragment to
131ea667
IC
2626 * @frag: the paged fragment to map
2627 * @offset: the offset within the fragment (starting at the
2628 * fragment's own offset)
2629 * @size: the number of bytes to map
f83347df 2630 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2631 *
2632 * Maps the page associated with @frag to @device.
2633 */
2634static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2635 const skb_frag_t *frag,
2636 size_t offset, size_t size,
2637 enum dma_data_direction dir)
2638{
2639 return dma_map_page(dev, skb_frag_page(frag),
2640 frag->page_offset + offset, size, dir);
2641}
2642
117632e6
ED
2643static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2644 gfp_t gfp_mask)
2645{
2646 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2647}
2648
bad93e9d
OP
2649
2650static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2651 gfp_t gfp_mask)
2652{
2653 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2654}
2655
2656
334a8132
PM
2657/**
2658 * skb_clone_writable - is the header of a clone writable
2659 * @skb: buffer to check
2660 * @len: length up to which to write
2661 *
2662 * Returns true if modifying the header part of the cloned buffer
2663 * does not requires the data to be copied.
2664 */
05bdd2f1 2665static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2666{
2667 return !skb_header_cloned(skb) &&
2668 skb_headroom(skb) + len <= skb->hdr_len;
2669}
2670
3697649f
DB
2671static inline int skb_try_make_writable(struct sk_buff *skb,
2672 unsigned int write_len)
2673{
2674 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2675 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2676}
2677
d9cc2048
HX
2678static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2679 int cloned)
2680{
2681 int delta = 0;
2682
d9cc2048
HX
2683 if (headroom > skb_headroom(skb))
2684 delta = headroom - skb_headroom(skb);
2685
2686 if (delta || cloned)
2687 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2688 GFP_ATOMIC);
2689 return 0;
2690}
2691
1da177e4
LT
2692/**
2693 * skb_cow - copy header of skb when it is required
2694 * @skb: buffer to cow
2695 * @headroom: needed headroom
2696 *
2697 * If the skb passed lacks sufficient headroom or its data part
2698 * is shared, data is reallocated. If reallocation fails, an error
2699 * is returned and original skb is not changed.
2700 *
2701 * The result is skb with writable area skb->head...skb->tail
2702 * and at least @headroom of space at head.
2703 */
2704static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2705{
d9cc2048
HX
2706 return __skb_cow(skb, headroom, skb_cloned(skb));
2707}
1da177e4 2708
d9cc2048
HX
2709/**
2710 * skb_cow_head - skb_cow but only making the head writable
2711 * @skb: buffer to cow
2712 * @headroom: needed headroom
2713 *
2714 * This function is identical to skb_cow except that we replace the
2715 * skb_cloned check by skb_header_cloned. It should be used when
2716 * you only need to push on some header and do not need to modify
2717 * the data.
2718 */
2719static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2720{
2721 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2722}
2723
2724/**
2725 * skb_padto - pad an skbuff up to a minimal size
2726 * @skb: buffer to pad
2727 * @len: minimal length
2728 *
2729 * Pads up a buffer to ensure the trailing bytes exist and are
2730 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2731 * is untouched. Otherwise it is extended. Returns zero on
2732 * success. The skb is freed on error.
1da177e4 2733 */
5b057c6b 2734static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2735{
2736 unsigned int size = skb->len;
2737 if (likely(size >= len))
5b057c6b 2738 return 0;
987c402a 2739 return skb_pad(skb, len - size);
1da177e4
LT
2740}
2741
9c0c1124
AD
2742/**
2743 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2744 * @skb: buffer to pad
2745 * @len: minimal length
2746 *
2747 * Pads up a buffer to ensure the trailing bytes exist and are
2748 * blanked. If the buffer already contains sufficient data it
2749 * is untouched. Otherwise it is extended. Returns zero on
2750 * success. The skb is freed on error.
2751 */
2752static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2753{
2754 unsigned int size = skb->len;
2755
2756 if (unlikely(size < len)) {
2757 len -= size;
2758 if (skb_pad(skb, len))
2759 return -ENOMEM;
2760 __skb_put(skb, len);
2761 }
2762 return 0;
2763}
2764
1da177e4 2765static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2766 struct iov_iter *from, int copy)
1da177e4
LT
2767{
2768 const int off = skb->len;
2769
2770 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2771 __wsum csum = 0;
2772 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2773 &csum, from) == copy) {
1da177e4
LT
2774 skb->csum = csum_block_add(skb->csum, csum, off);
2775 return 0;
2776 }
af2b040e 2777 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2778 return 0;
2779
2780 __skb_trim(skb, off);
2781 return -EFAULT;
2782}
2783
38ba0a65
ED
2784static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2785 const struct page *page, int off)
1da177e4
LT
2786{
2787 if (i) {
9e903e08 2788 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2789
ea2ab693 2790 return page == skb_frag_page(frag) &&
9e903e08 2791 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2792 }
38ba0a65 2793 return false;
1da177e4
LT
2794}
2795
364c6bad
HX
2796static inline int __skb_linearize(struct sk_buff *skb)
2797{
2798 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2799}
2800
1da177e4
LT
2801/**
2802 * skb_linearize - convert paged skb to linear one
2803 * @skb: buffer to linarize
1da177e4
LT
2804 *
2805 * If there is no free memory -ENOMEM is returned, otherwise zero
2806 * is returned and the old skb data released.
2807 */
364c6bad
HX
2808static inline int skb_linearize(struct sk_buff *skb)
2809{
2810 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2811}
2812
cef401de
ED
2813/**
2814 * skb_has_shared_frag - can any frag be overwritten
2815 * @skb: buffer to test
2816 *
2817 * Return true if the skb has at least one frag that might be modified
2818 * by an external entity (as in vmsplice()/sendfile())
2819 */
2820static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2821{
c9af6db4
PS
2822 return skb_is_nonlinear(skb) &&
2823 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2824}
2825
364c6bad
HX
2826/**
2827 * skb_linearize_cow - make sure skb is linear and writable
2828 * @skb: buffer to process
2829 *
2830 * If there is no free memory -ENOMEM is returned, otherwise zero
2831 * is returned and the old skb data released.
2832 */
2833static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2834{
364c6bad
HX
2835 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2836 __skb_linearize(skb) : 0;
1da177e4
LT
2837}
2838
2839/**
2840 * skb_postpull_rcsum - update checksum for received skb after pull
2841 * @skb: buffer to update
2842 * @start: start of data before pull
2843 * @len: length of data pulled
2844 *
2845 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2846 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2847 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2848 */
2849
2850static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2851 const void *start, unsigned int len)
1da177e4 2852{
84fa7933 2853 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4 2854 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
6ae459bd 2855 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
31b33dfb 2856 skb_checksum_start_offset(skb) < 0)
6ae459bd 2857 skb->ip_summed = CHECKSUM_NONE;
1da177e4
LT
2858}
2859
cbb042f9
HX
2860unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2861
f8ffad69
DB
2862static inline void skb_postpush_rcsum(struct sk_buff *skb,
2863 const void *start, unsigned int len)
2864{
2865 /* For performing the reverse operation to skb_postpull_rcsum(),
2866 * we can instead of ...
2867 *
2868 * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0));
2869 *
2870 * ... just use this equivalent version here to save a few
2871 * instructions. Feeding csum of 0 in csum_partial() and later
2872 * on adding skb->csum is equivalent to feed skb->csum in the
2873 * first place.
2874 */
2875 if (skb->ip_summed == CHECKSUM_COMPLETE)
2876 skb->csum = csum_partial(start, len, skb->csum);
2877}
2878
7ce5a27f
DM
2879/**
2880 * pskb_trim_rcsum - trim received skb and update checksum
2881 * @skb: buffer to trim
2882 * @len: new length
2883 *
2884 * This is exactly the same as pskb_trim except that it ensures the
2885 * checksum of received packets are still valid after the operation.
2886 */
2887
2888static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2889{
2890 if (likely(len >= skb->len))
2891 return 0;
2892 if (skb->ip_summed == CHECKSUM_COMPLETE)
2893 skb->ip_summed = CHECKSUM_NONE;
2894 return __pskb_trim(skb, len);
2895}
2896
1da177e4
LT
2897#define skb_queue_walk(queue, skb) \
2898 for (skb = (queue)->next; \
a1e4891f 2899 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2900 skb = skb->next)
2901
46f8914e
JC
2902#define skb_queue_walk_safe(queue, skb, tmp) \
2903 for (skb = (queue)->next, tmp = skb->next; \
2904 skb != (struct sk_buff *)(queue); \
2905 skb = tmp, tmp = skb->next)
2906
1164f52a 2907#define skb_queue_walk_from(queue, skb) \
a1e4891f 2908 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2909 skb = skb->next)
2910
2911#define skb_queue_walk_from_safe(queue, skb, tmp) \
2912 for (tmp = skb->next; \
2913 skb != (struct sk_buff *)(queue); \
2914 skb = tmp, tmp = skb->next)
2915
300ce174
SH
2916#define skb_queue_reverse_walk(queue, skb) \
2917 for (skb = (queue)->prev; \
a1e4891f 2918 skb != (struct sk_buff *)(queue); \
300ce174
SH
2919 skb = skb->prev)
2920
686a2955
DM
2921#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2922 for (skb = (queue)->prev, tmp = skb->prev; \
2923 skb != (struct sk_buff *)(queue); \
2924 skb = tmp, tmp = skb->prev)
2925
2926#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2927 for (tmp = skb->prev; \
2928 skb != (struct sk_buff *)(queue); \
2929 skb = tmp, tmp = skb->prev)
1da177e4 2930
21dc3301 2931static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2932{
2933 return skb_shinfo(skb)->frag_list != NULL;
2934}
2935
2936static inline void skb_frag_list_init(struct sk_buff *skb)
2937{
2938 skb_shinfo(skb)->frag_list = NULL;
2939}
2940
ee039871
DM
2941#define skb_walk_frags(skb, iter) \
2942 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2943
ea3793ee
RW
2944
2945int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
2946 const struct sk_buff *skb);
2947struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
2948 int *peeked, int *off, int *err,
2949 struct sk_buff **last);
7965bd4d
JP
2950struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2951 int *peeked, int *off, int *err);
2952struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2953 int *err);
2954unsigned int datagram_poll(struct file *file, struct socket *sock,
2955 struct poll_table_struct *wait);
c0371da6
AV
2956int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2957 struct iov_iter *to, int size);
51f3d02b
DM
2958static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2959 struct msghdr *msg, int size)
2960{
e5a4b0bb 2961 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2962}
e5a4b0bb
AV
2963int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2964 struct msghdr *msg);
3a654f97
AV
2965int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2966 struct iov_iter *from, int len);
3a654f97 2967int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 2968void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 2969void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
2970static inline void skb_free_datagram_locked(struct sock *sk,
2971 struct sk_buff *skb)
2972{
2973 __skb_free_datagram_locked(sk, skb, 0);
2974}
7965bd4d 2975int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2976int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2977int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2978__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2979 int len, __wsum csum);
a60e3cc7
HFS
2980ssize_t skb_socket_splice(struct sock *sk,
2981 struct pipe_inode_info *pipe,
2982 struct splice_pipe_desc *spd);
2983int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 2984 struct pipe_inode_info *pipe, unsigned int len,
a60e3cc7
HFS
2985 unsigned int flags,
2986 ssize_t (*splice_cb)(struct sock *,
2987 struct pipe_inode_info *,
2988 struct splice_pipe_desc *));
7965bd4d 2989void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2990unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2991int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2992 int len, int hlen);
7965bd4d
JP
2993void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2994int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2995void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2996unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 2997bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 2998struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2999struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3000int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
3001int skb_vlan_pop(struct sk_buff *skb);
3002int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3003struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3004 gfp_t gfp);
20380731 3005
6ce8e9ce
AV
3006static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3007{
21226abb 3008 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
3009}
3010
7eab8d9e
AV
3011static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3012{
e5a4b0bb 3013 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3014}
3015
2817a336
DB
3016struct skb_checksum_ops {
3017 __wsum (*update)(const void *mem, int len, __wsum wsum);
3018 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3019};
3020
3021__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3022 __wsum csum, const struct skb_checksum_ops *ops);
3023__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3024 __wsum csum);
3025
1e98a0f0
ED
3026static inline void * __must_check
3027__skb_header_pointer(const struct sk_buff *skb, int offset,
3028 int len, void *data, int hlen, void *buffer)
1da177e4 3029{
55820ee2 3030 if (hlen - offset >= len)
690e36e7 3031 return data + offset;
1da177e4 3032
690e36e7
DM
3033 if (!skb ||
3034 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3035 return NULL;
3036
3037 return buffer;
3038}
3039
1e98a0f0
ED
3040static inline void * __must_check
3041skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3042{
3043 return __skb_header_pointer(skb, offset, len, skb->data,
3044 skb_headlen(skb), buffer);
3045}
3046
4262e5cc
DB
3047/**
3048 * skb_needs_linearize - check if we need to linearize a given skb
3049 * depending on the given device features.
3050 * @skb: socket buffer to check
3051 * @features: net device features
3052 *
3053 * Returns true if either:
3054 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3055 * 2. skb is fragmented and the device does not support SG.
3056 */
3057static inline bool skb_needs_linearize(struct sk_buff *skb,
3058 netdev_features_t features)
3059{
3060 return skb_is_nonlinear(skb) &&
3061 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3062 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3063}
3064
d626f62b
ACM
3065static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3066 void *to,
3067 const unsigned int len)
3068{
3069 memcpy(to, skb->data, len);
3070}
3071
3072static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3073 const int offset, void *to,
3074 const unsigned int len)
3075{
3076 memcpy(to, skb->data + offset, len);
3077}
3078
27d7ff46
ACM
3079static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3080 const void *from,
3081 const unsigned int len)
3082{
3083 memcpy(skb->data, from, len);
3084}
3085
3086static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3087 const int offset,
3088 const void *from,
3089 const unsigned int len)
3090{
3091 memcpy(skb->data + offset, from, len);
3092}
3093
7965bd4d 3094void skb_init(void);
1da177e4 3095
ac45f602
PO
3096static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3097{
3098 return skb->tstamp;
3099}
3100
a61bbcf2
PM
3101/**
3102 * skb_get_timestamp - get timestamp from a skb
3103 * @skb: skb to get stamp from
3104 * @stamp: pointer to struct timeval to store stamp in
3105 *
3106 * Timestamps are stored in the skb as offsets to a base timestamp.
3107 * This function converts the offset back to a struct timeval and stores
3108 * it in stamp.
3109 */
ac45f602
PO
3110static inline void skb_get_timestamp(const struct sk_buff *skb,
3111 struct timeval *stamp)
a61bbcf2 3112{
b7aa0bf7 3113 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3114}
3115
ac45f602
PO
3116static inline void skb_get_timestampns(const struct sk_buff *skb,
3117 struct timespec *stamp)
3118{
3119 *stamp = ktime_to_timespec(skb->tstamp);
3120}
3121
b7aa0bf7 3122static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3123{
b7aa0bf7 3124 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3125}
3126
164891aa
SH
3127static inline ktime_t net_timedelta(ktime_t t)
3128{
3129 return ktime_sub(ktime_get_real(), t);
3130}
3131
b9ce204f
IJ
3132static inline ktime_t net_invalid_timestamp(void)
3133{
3134 return ktime_set(0, 0);
3135}
a61bbcf2 3136
62bccb8c
AD
3137struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3138
c1f19b51
RC
3139#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3140
7965bd4d
JP
3141void skb_clone_tx_timestamp(struct sk_buff *skb);
3142bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3143
3144#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3145
3146static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3147{
3148}
3149
3150static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3151{
3152 return false;
3153}
3154
3155#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3156
3157/**
3158 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3159 *
da92b194
RC
3160 * PHY drivers may accept clones of transmitted packets for
3161 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3162 * must call this function to return the skb back to the stack with a
3163 * timestamp.
da92b194 3164 *
c1f19b51 3165 * @skb: clone of the the original outgoing packet
7a76a021 3166 * @hwtstamps: hardware time stamps
c1f19b51
RC
3167 *
3168 */
3169void skb_complete_tx_timestamp(struct sk_buff *skb,
3170 struct skb_shared_hwtstamps *hwtstamps);
3171
e7fd2885
WB
3172void __skb_tstamp_tx(struct sk_buff *orig_skb,
3173 struct skb_shared_hwtstamps *hwtstamps,
3174 struct sock *sk, int tstype);
3175
ac45f602
PO
3176/**
3177 * skb_tstamp_tx - queue clone of skb with send time stamps
3178 * @orig_skb: the original outgoing packet
3179 * @hwtstamps: hardware time stamps, may be NULL if not available
3180 *
3181 * If the skb has a socket associated, then this function clones the
3182 * skb (thus sharing the actual data and optional structures), stores
3183 * the optional hardware time stamping information (if non NULL) or
3184 * generates a software time stamp (otherwise), then queues the clone
3185 * to the error queue of the socket. Errors are silently ignored.
3186 */
7965bd4d
JP
3187void skb_tstamp_tx(struct sk_buff *orig_skb,
3188 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3189
4507a715
RC
3190static inline void sw_tx_timestamp(struct sk_buff *skb)
3191{
2244d07b
OH
3192 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3193 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3194 skb_tstamp_tx(skb, NULL);
3195}
3196
3197/**
3198 * skb_tx_timestamp() - Driver hook for transmit timestamping
3199 *
3200 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3201 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3202 *
73409f3b
DM
3203 * Specifically, one should make absolutely sure that this function is
3204 * called before TX completion of this packet can trigger. Otherwise
3205 * the packet could potentially already be freed.
3206 *
4507a715
RC
3207 * @skb: A socket buffer.
3208 */
3209static inline void skb_tx_timestamp(struct sk_buff *skb)
3210{
c1f19b51 3211 skb_clone_tx_timestamp(skb);
4507a715
RC
3212 sw_tx_timestamp(skb);
3213}
3214
6e3e939f
JB
3215/**
3216 * skb_complete_wifi_ack - deliver skb with wifi status
3217 *
3218 * @skb: the original outgoing packet
3219 * @acked: ack status
3220 *
3221 */
3222void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3223
7965bd4d
JP
3224__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3225__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3226
60476372
HX
3227static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3228{
6edec0e6
TH
3229 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3230 skb->csum_valid ||
3231 (skb->ip_summed == CHECKSUM_PARTIAL &&
3232 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3233}
3234
fb286bb2
HX
3235/**
3236 * skb_checksum_complete - Calculate checksum of an entire packet
3237 * @skb: packet to process
3238 *
3239 * This function calculates the checksum over the entire packet plus
3240 * the value of skb->csum. The latter can be used to supply the
3241 * checksum of a pseudo header as used by TCP/UDP. It returns the
3242 * checksum.
3243 *
3244 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3245 * this function can be used to verify that checksum on received
3246 * packets. In that case the function should return zero if the
3247 * checksum is correct. In particular, this function will return zero
3248 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3249 * hardware has already verified the correctness of the checksum.
3250 */
4381ca3c 3251static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3252{
60476372
HX
3253 return skb_csum_unnecessary(skb) ?
3254 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3255}
3256
77cffe23
TH
3257static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3258{
3259 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3260 if (skb->csum_level == 0)
3261 skb->ip_summed = CHECKSUM_NONE;
3262 else
3263 skb->csum_level--;
3264 }
3265}
3266
3267static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3268{
3269 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3270 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3271 skb->csum_level++;
3272 } else if (skb->ip_summed == CHECKSUM_NONE) {
3273 skb->ip_summed = CHECKSUM_UNNECESSARY;
3274 skb->csum_level = 0;
3275 }
3276}
3277
5a212329
TH
3278static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3279{
3280 /* Mark current checksum as bad (typically called from GRO
3281 * path). In the case that ip_summed is CHECKSUM_NONE
3282 * this must be the first checksum encountered in the packet.
3283 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3284 * checksum after the last one validated. For UDP, a zero
3285 * checksum can not be marked as bad.
3286 */
3287
3288 if (skb->ip_summed == CHECKSUM_NONE ||
3289 skb->ip_summed == CHECKSUM_UNNECESSARY)
3290 skb->csum_bad = 1;
3291}
3292
76ba0aae
TH
3293/* Check if we need to perform checksum complete validation.
3294 *
3295 * Returns true if checksum complete is needed, false otherwise
3296 * (either checksum is unnecessary or zero checksum is allowed).
3297 */
3298static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3299 bool zero_okay,
3300 __sum16 check)
3301{
5d0c2b95
TH
3302 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3303 skb->csum_valid = 1;
77cffe23 3304 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3305 return false;
3306 }
3307
3308 return true;
3309}
3310
3311/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3312 * in checksum_init.
3313 */
3314#define CHECKSUM_BREAK 76
3315
4e18b9ad
TH
3316/* Unset checksum-complete
3317 *
3318 * Unset checksum complete can be done when packet is being modified
3319 * (uncompressed for instance) and checksum-complete value is
3320 * invalidated.
3321 */
3322static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3323{
3324 if (skb->ip_summed == CHECKSUM_COMPLETE)
3325 skb->ip_summed = CHECKSUM_NONE;
3326}
3327
76ba0aae
TH
3328/* Validate (init) checksum based on checksum complete.
3329 *
3330 * Return values:
3331 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3332 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3333 * checksum is stored in skb->csum for use in __skb_checksum_complete
3334 * non-zero: value of invalid checksum
3335 *
3336 */
3337static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3338 bool complete,
3339 __wsum psum)
3340{
3341 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3342 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3343 skb->csum_valid = 1;
76ba0aae
TH
3344 return 0;
3345 }
5a212329
TH
3346 } else if (skb->csum_bad) {
3347 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3348 return (__force __sum16)1;
76ba0aae
TH
3349 }
3350
3351 skb->csum = psum;
3352
5d0c2b95
TH
3353 if (complete || skb->len <= CHECKSUM_BREAK) {
3354 __sum16 csum;
3355
3356 csum = __skb_checksum_complete(skb);
3357 skb->csum_valid = !csum;
3358 return csum;
3359 }
76ba0aae
TH
3360
3361 return 0;
3362}
3363
3364static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3365{
3366 return 0;
3367}
3368
3369/* Perform checksum validate (init). Note that this is a macro since we only
3370 * want to calculate the pseudo header which is an input function if necessary.
3371 * First we try to validate without any computation (checksum unnecessary) and
3372 * then calculate based on checksum complete calling the function to compute
3373 * pseudo header.
3374 *
3375 * Return values:
3376 * 0: checksum is validated or try to in skb_checksum_complete
3377 * non-zero: value of invalid checksum
3378 */
3379#define __skb_checksum_validate(skb, proto, complete, \
3380 zero_okay, check, compute_pseudo) \
3381({ \
3382 __sum16 __ret = 0; \
5d0c2b95 3383 skb->csum_valid = 0; \
76ba0aae
TH
3384 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3385 __ret = __skb_checksum_validate_complete(skb, \
3386 complete, compute_pseudo(skb, proto)); \
3387 __ret; \
3388})
3389
3390#define skb_checksum_init(skb, proto, compute_pseudo) \
3391 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3392
3393#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3394 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3395
3396#define skb_checksum_validate(skb, proto, compute_pseudo) \
3397 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3398
3399#define skb_checksum_validate_zero_check(skb, proto, check, \
3400 compute_pseudo) \
096a4cfa 3401 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3402
3403#define skb_checksum_simple_validate(skb) \
3404 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3405
d96535a1
TH
3406static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3407{
3408 return (skb->ip_summed == CHECKSUM_NONE &&
3409 skb->csum_valid && !skb->csum_bad);
3410}
3411
3412static inline void __skb_checksum_convert(struct sk_buff *skb,
3413 __sum16 check, __wsum pseudo)
3414{
3415 skb->csum = ~pseudo;
3416 skb->ip_summed = CHECKSUM_COMPLETE;
3417}
3418
3419#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3420do { \
3421 if (__skb_checksum_convert_check(skb)) \
3422 __skb_checksum_convert(skb, check, \
3423 compute_pseudo(skb, proto)); \
3424} while (0)
3425
15e2396d
TH
3426static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3427 u16 start, u16 offset)
3428{
3429 skb->ip_summed = CHECKSUM_PARTIAL;
3430 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3431 skb->csum_offset = offset - start;
3432}
3433
dcdc8994
TH
3434/* Update skbuf and packet to reflect the remote checksum offload operation.
3435 * When called, ptr indicates the starting point for skb->csum when
3436 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3437 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3438 */
3439static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3440 int start, int offset, bool nopartial)
dcdc8994
TH
3441{
3442 __wsum delta;
3443
15e2396d
TH
3444 if (!nopartial) {
3445 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3446 return;
3447 }
3448
dcdc8994
TH
3449 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3450 __skb_checksum_complete(skb);
3451 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3452 }
3453
3454 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3455
3456 /* Adjust skb->csum since we changed the packet */
3457 skb->csum = csum_add(skb->csum, delta);
3458}
3459
5f79e0f9 3460#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3461void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3462static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3463{
3464 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3465 nf_conntrack_destroy(nfct);
1da177e4
LT
3466}
3467static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3468{
3469 if (nfct)
3470 atomic_inc(&nfct->use);
3471}
2fc72c7b 3472#endif
34666d46 3473#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3474static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3475{
3476 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3477 kfree(nf_bridge);
3478}
3479static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3480{
3481 if (nf_bridge)
3482 atomic_inc(&nf_bridge->use);
3483}
3484#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3485static inline void nf_reset(struct sk_buff *skb)
3486{
5f79e0f9 3487#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3488 nf_conntrack_put(skb->nfct);
3489 skb->nfct = NULL;
2fc72c7b 3490#endif
34666d46 3491#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3492 nf_bridge_put(skb->nf_bridge);
3493 skb->nf_bridge = NULL;
3494#endif
3495}
3496
124dff01
PM
3497static inline void nf_reset_trace(struct sk_buff *skb)
3498{
478b360a 3499#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3500 skb->nf_trace = 0;
3501#endif
a193a4ab
PM
3502}
3503
edda553c 3504/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3505static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3506 bool copy)
edda553c 3507{
5f79e0f9 3508#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3509 dst->nfct = src->nfct;
3510 nf_conntrack_get(src->nfct);
b1937227
ED
3511 if (copy)
3512 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3513#endif
34666d46 3514#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3515 dst->nf_bridge = src->nf_bridge;
3516 nf_bridge_get(src->nf_bridge);
3517#endif
478b360a 3518#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3519 if (copy)
3520 dst->nf_trace = src->nf_trace;
478b360a 3521#endif
edda553c
YK
3522}
3523
e7ac05f3
YK
3524static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3525{
e7ac05f3 3526#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3527 nf_conntrack_put(dst->nfct);
2fc72c7b 3528#endif
34666d46 3529#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3530 nf_bridge_put(dst->nf_bridge);
3531#endif
b1937227 3532 __nf_copy(dst, src, true);
e7ac05f3
YK
3533}
3534
984bc16c
JM
3535#ifdef CONFIG_NETWORK_SECMARK
3536static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3537{
3538 to->secmark = from->secmark;
3539}
3540
3541static inline void skb_init_secmark(struct sk_buff *skb)
3542{
3543 skb->secmark = 0;
3544}
3545#else
3546static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3547{ }
3548
3549static inline void skb_init_secmark(struct sk_buff *skb)
3550{ }
3551#endif
3552
574f7194
EB
3553static inline bool skb_irq_freeable(const struct sk_buff *skb)
3554{
3555 return !skb->destructor &&
3556#if IS_ENABLED(CONFIG_XFRM)
3557 !skb->sp &&
3558#endif
3559#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3560 !skb->nfct &&
3561#endif
3562 !skb->_skb_refdst &&
3563 !skb_has_frag_list(skb);
3564}
3565
f25f4e44
PWJ
3566static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3567{
f25f4e44 3568 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3569}
3570
9247744e 3571static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3572{
4e3ab47a 3573 return skb->queue_mapping;
4e3ab47a
PE
3574}
3575
f25f4e44
PWJ
3576static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3577{
f25f4e44 3578 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3579}
3580
d5a9e24a
DM
3581static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3582{
3583 skb->queue_mapping = rx_queue + 1;
3584}
3585
9247744e 3586static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3587{
3588 return skb->queue_mapping - 1;
3589}
3590
9247744e 3591static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3592{
a02cec21 3593 return skb->queue_mapping != 0;
d5a9e24a
DM
3594}
3595
def8b4fa
AD
3596static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3597{
0b3d8e08 3598#ifdef CONFIG_XFRM
def8b4fa 3599 return skb->sp;
def8b4fa 3600#else
def8b4fa 3601 return NULL;
def8b4fa 3602#endif
0b3d8e08 3603}
def8b4fa 3604
68c33163
PS
3605/* Keeps track of mac header offset relative to skb->head.
3606 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3607 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3608 * tunnel skb it points to outer mac header.
3609 * Keeps track of level of encapsulation of network headers.
3610 */
68c33163 3611struct skb_gso_cb {
802ab55a
AD
3612 union {
3613 int mac_offset;
3614 int data_offset;
3615 };
3347c960 3616 int encap_level;
76443456 3617 __wsum csum;
7e2b10c1 3618 __u16 csum_start;
68c33163 3619};
9207f9d4
KK
3620#define SKB_SGO_CB_OFFSET 32
3621#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3622
3623static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3624{
3625 return (skb_mac_header(inner_skb) - inner_skb->head) -
3626 SKB_GSO_CB(inner_skb)->mac_offset;
3627}
3628
1e2bd517
PS
3629static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3630{
3631 int new_headroom, headroom;
3632 int ret;
3633
3634 headroom = skb_headroom(skb);
3635 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3636 if (ret)
3637 return ret;
3638
3639 new_headroom = skb_headroom(skb);
3640 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3641 return 0;
3642}
3643
08b64fcc
AD
3644static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3645{
3646 /* Do not update partial checksums if remote checksum is enabled. */
3647 if (skb->remcsum_offload)
3648 return;
3649
3650 SKB_GSO_CB(skb)->csum = res;
3651 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3652}
3653
7e2b10c1
TH
3654/* Compute the checksum for a gso segment. First compute the checksum value
3655 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3656 * then add in skb->csum (checksum from csum_start to end of packet).
3657 * skb->csum and csum_start are then updated to reflect the checksum of the
3658 * resultant packet starting from the transport header-- the resultant checksum
3659 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3660 * header.
3661 */
3662static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3663{
76443456
AD
3664 unsigned char *csum_start = skb_transport_header(skb);
3665 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3666 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3667
76443456
AD
3668 SKB_GSO_CB(skb)->csum = res;
3669 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3670
76443456 3671 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3672}
3673
bdcc0924 3674static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3675{
3676 return skb_shinfo(skb)->gso_size;
3677}
3678
36a8f39e 3679/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3680static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3681{
3682 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3683}
3684
7965bd4d 3685void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3686
3687static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3688{
3689 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3690 * wanted then gso_type will be set. */
05bdd2f1
ED
3691 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3692
b78462eb
AD
3693 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3694 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3695 __skb_warn_lro_forwarding(skb);
3696 return true;
3697 }
3698 return false;
3699}
3700
35fc92a9
HX
3701static inline void skb_forward_csum(struct sk_buff *skb)
3702{
3703 /* Unfortunately we don't support this one. Any brave souls? */
3704 if (skb->ip_summed == CHECKSUM_COMPLETE)
3705 skb->ip_summed = CHECKSUM_NONE;
3706}
3707
bc8acf2c
ED
3708/**
3709 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3710 * @skb: skb to check
3711 *
3712 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3713 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3714 * use this helper, to document places where we make this assertion.
3715 */
05bdd2f1 3716static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3717{
3718#ifdef DEBUG
3719 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3720#endif
3721}
3722
f35d9d8a 3723bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3724
ed1f50c3 3725int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3726struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3727 unsigned int transport_len,
3728 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3729
3a7c1ee4
AD
3730/**
3731 * skb_head_is_locked - Determine if the skb->head is locked down
3732 * @skb: skb to check
3733 *
3734 * The head on skbs build around a head frag can be removed if they are
3735 * not cloned. This function returns true if the skb head is locked down
3736 * due to either being allocated via kmalloc, or by being a clone with
3737 * multiple references to the head.
3738 */
3739static inline bool skb_head_is_locked(const struct sk_buff *skb)
3740{
3741 return !skb->head_frag || skb_cloned(skb);
3742}
fe6cc55f
FW
3743
3744/**
3745 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3746 *
3747 * @skb: GSO skb
3748 *
3749 * skb_gso_network_seglen is used to determine the real size of the
3750 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3751 *
3752 * The MAC/L2 header is not accounted for.
3753 */
3754static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3755{
3756 unsigned int hdr_len = skb_transport_header(skb) -
3757 skb_network_header(skb);
3758 return hdr_len + skb_gso_transport_seglen(skb);
3759}
ee122c79 3760
179bc67f
EC
3761/* Local Checksum Offload.
3762 * Compute outer checksum based on the assumption that the
3763 * inner checksum will be offloaded later.
e8ae7b00
EC
3764 * See Documentation/networking/checksum-offloads.txt for
3765 * explanation of how this works.
179bc67f
EC
3766 * Fill in outer checksum adjustment (e.g. with sum of outer
3767 * pseudo-header) before calling.
3768 * Also ensure that inner checksum is in linear data area.
3769 */
3770static inline __wsum lco_csum(struct sk_buff *skb)
3771{
9e74a6da
AD
3772 unsigned char *csum_start = skb_checksum_start(skb);
3773 unsigned char *l4_hdr = skb_transport_header(skb);
3774 __wsum partial;
179bc67f
EC
3775
3776 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3777 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3778 skb->csum_offset));
3779
179bc67f 3780 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3781 * adjustment filled in by caller) and return result.
179bc67f 3782 */
9e74a6da 3783 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3784}
3785
1da177e4
LT
3786#endif /* __KERNEL__ */
3787#endif /* _LINUX_SKBUFF_H */