fq: add fair queuing framework
[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
LT
303
304typedef struct skb_frag_struct skb_frag_t;
305
306struct skb_frag_struct {
a8605c60
IC
307 struct {
308 struct page *p;
309 } page;
cb4dfe56 310#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
311 __u32 page_offset;
312 __u32 size;
cb4dfe56
ED
313#else
314 __u16 page_offset;
315 __u16 size;
316#endif
1da177e4
LT
317};
318
9e903e08
ED
319static inline unsigned int skb_frag_size(const skb_frag_t *frag)
320{
321 return frag->size;
322}
323
324static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
325{
326 frag->size = size;
327}
328
329static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
330{
331 frag->size += delta;
332}
333
334static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
335{
336 frag->size -= delta;
337}
338
ac45f602
PO
339#define HAVE_HW_TIME_STAMP
340
341/**
d3a21be8 342 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
343 * @hwtstamp: hardware time stamp transformed into duration
344 * since arbitrary point in time
ac45f602
PO
345 *
346 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 347 * skb->tstamp.
ac45f602
PO
348 *
349 * hwtstamps can only be compared against other hwtstamps from
350 * the same device.
351 *
352 * This structure is attached to packets as part of the
353 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
354 */
355struct skb_shared_hwtstamps {
356 ktime_t hwtstamp;
ac45f602
PO
357};
358
2244d07b
OH
359/* Definitions for tx_flags in struct skb_shared_info */
360enum {
361 /* generate hardware time stamp */
362 SKBTX_HW_TSTAMP = 1 << 0,
363
e7fd2885 364 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
365 SKBTX_SW_TSTAMP = 1 << 1,
366
367 /* device driver is going to provide hardware time stamp */
368 SKBTX_IN_PROGRESS = 1 << 2,
369
a6686f2f 370 /* device driver supports TX zero-copy buffers */
62b1a8ab 371 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
372
373 /* generate wifi status information (where possible) */
62b1a8ab 374 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
375
376 /* This indicates at least one fragment might be overwritten
377 * (as in vmsplice(), sendfile() ...)
378 * If we need to compute a TX checksum, we'll need to copy
379 * all frags to avoid possible bad checksum
380 */
381 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
382
383 /* generate software time stamp when entering packet scheduling */
384 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
385
386 /* generate software timestamp on peer data acknowledgment */
387 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
388};
389
e1c8a607
WB
390#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
391 SKBTX_SCHED_TSTAMP | \
392 SKBTX_ACK_TSTAMP)
f24b9be5
WB
393#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
394
a6686f2f
SM
395/*
396 * The callback notifies userspace to release buffers when skb DMA is done in
397 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
398 * The zerocopy_success argument is true if zero copy transmit occurred,
399 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
400 * The ctx field is used to track device context.
401 * The desc field is used to track userspace buffer index.
a6686f2f
SM
402 */
403struct ubuf_info {
e19d6763 404 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 405 void *ctx;
a6686f2f 406 unsigned long desc;
ac45f602
PO
407};
408
1da177e4
LT
409/* This data is invariant across clones and lives at
410 * the end of the header data, ie. at skb->end.
411 */
412struct skb_shared_info {
9f42f126
IC
413 unsigned char nr_frags;
414 __u8 tx_flags;
7967168c
HX
415 unsigned short gso_size;
416 /* Warning: this field is not always filled in (UFO)! */
417 unsigned short gso_segs;
418 unsigned short gso_type;
1da177e4 419 struct sk_buff *frag_list;
ac45f602 420 struct skb_shared_hwtstamps hwtstamps;
09c2d251 421 u32 tskey;
9f42f126 422 __be32 ip6_frag_id;
ec7d2f2c
ED
423
424 /*
425 * Warning : all fields before dataref are cleared in __alloc_skb()
426 */
427 atomic_t dataref;
428
69e3c75f
JB
429 /* Intermediate layers must ensure that destructor_arg
430 * remains valid until skb destructor */
431 void * destructor_arg;
a6686f2f 432
fed66381
ED
433 /* must be last field, see pskb_expand_head() */
434 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
435};
436
437/* We divide dataref into two halves. The higher 16 bits hold references
438 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
439 * the entire skb->data. A clone of a headerless skb holds the length of
440 * the header in skb->hdr_len.
1da177e4
LT
441 *
442 * All users must obey the rule that the skb->data reference count must be
443 * greater than or equal to the payload reference count.
444 *
445 * Holding a reference to the payload part means that the user does not
446 * care about modifications to the header part of skb->data.
447 */
448#define SKB_DATAREF_SHIFT 16
449#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
450
d179cd12
DM
451
452enum {
c8753d55
VS
453 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
454 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
455 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
456};
457
7967168c
HX
458enum {
459 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 460 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
461
462 /* This indicates the skb is from an untrusted source. */
463 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
464
465 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
466 SKB_GSO_TCP_ECN = 1 << 3,
467
cbc53e08 468 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 469
cbc53e08 470 SKB_GSO_TCPV6 = 1 << 5,
68c33163 471
cbc53e08 472 SKB_GSO_FCOE = 1 << 6,
73136267 473
cbc53e08 474 SKB_GSO_GRE = 1 << 7,
0d89d203 475
cbc53e08 476 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 477
cbc53e08 478 SKB_GSO_IPIP = 1 << 9,
61c1db7f 479
cbc53e08 480 SKB_GSO_SIT = 1 << 10,
0f4f4ffa 481
cbc53e08 482 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 483
cbc53e08
AD
484 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
485
802ab55a
AD
486 SKB_GSO_PARTIAL = 1 << 13,
487
488 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
7967168c
HX
489};
490
2e07fa9c
ACM
491#if BITS_PER_LONG > 32
492#define NET_SKBUFF_DATA_USES_OFFSET 1
493#endif
494
495#ifdef NET_SKBUFF_DATA_USES_OFFSET
496typedef unsigned int sk_buff_data_t;
497#else
498typedef unsigned char *sk_buff_data_t;
499#endif
500
363ec392
ED
501/**
502 * struct skb_mstamp - multi resolution time stamps
503 * @stamp_us: timestamp in us resolution
504 * @stamp_jiffies: timestamp in jiffies
505 */
506struct skb_mstamp {
507 union {
508 u64 v64;
509 struct {
510 u32 stamp_us;
511 u32 stamp_jiffies;
512 };
513 };
514};
515
516/**
517 * skb_mstamp_get - get current timestamp
518 * @cl: place to store timestamps
519 */
520static inline void skb_mstamp_get(struct skb_mstamp *cl)
521{
522 u64 val = local_clock();
523
524 do_div(val, NSEC_PER_USEC);
525 cl->stamp_us = (u32)val;
526 cl->stamp_jiffies = (u32)jiffies;
527}
528
529/**
530 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
531 * @t1: pointer to newest sample
532 * @t0: pointer to oldest sample
533 */
534static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
535 const struct skb_mstamp *t0)
536{
537 s32 delta_us = t1->stamp_us - t0->stamp_us;
538 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
539
540 /* If delta_us is negative, this might be because interval is too big,
541 * or local_clock() drift is too big : fallback using jiffies.
542 */
543 if (delta_us <= 0 ||
544 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
545
546 delta_us = jiffies_to_usecs(delta_jiffies);
547
548 return delta_us;
549}
550
625a5e10
YC
551static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
552 const struct skb_mstamp *t0)
553{
554 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
555
556 if (!diff)
557 diff = t1->stamp_us - t0->stamp_us;
558 return diff > 0;
559}
363ec392 560
1da177e4
LT
561/**
562 * struct sk_buff - socket buffer
563 * @next: Next buffer in list
564 * @prev: Previous buffer in list
363ec392 565 * @tstamp: Time we arrived/left
56b17425 566 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 567 * @sk: Socket we are owned by
1da177e4 568 * @dev: Device we arrived on/are leaving by
d84e0bd7 569 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 570 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 571 * @sp: the security path, used for xfrm
1da177e4
LT
572 * @len: Length of actual data
573 * @data_len: Data length
574 * @mac_len: Length of link layer header
334a8132 575 * @hdr_len: writable header length of cloned skb
663ead3b
HX
576 * @csum: Checksum (must include start/offset pair)
577 * @csum_start: Offset from skb->head where checksumming should start
578 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 579 * @priority: Packet queueing priority
60ff7467 580 * @ignore_df: allow local fragmentation
1da177e4 581 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 582 * @ip_summed: Driver fed us an IP checksum
1da177e4 583 * @nohdr: Payload reference only, must not modify header
d84e0bd7 584 * @nfctinfo: Relationship of this skb to the connection
1da177e4 585 * @pkt_type: Packet class
c83c2486 586 * @fclone: skbuff clone status
c83c2486 587 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
588 * @peeked: this packet has been seen already, so stats have been
589 * done for it, don't do them again
ba9dda3a 590 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
591 * @protocol: Packet protocol from driver
592 * @destructor: Destruct function
593 * @nfct: Associated connection, if any
1da177e4 594 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 595 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
596 * @tc_index: Traffic control index
597 * @tc_verd: traffic control verdict
61b905da 598 * @hash: the packet hash
d84e0bd7 599 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 600 * @xmit_more: More SKBs are pending for this queue
553a5672 601 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 602 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 603 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 604 * ports.
a3b18ddb 605 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
606 * @wifi_acked_valid: wifi_acked was set
607 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 608 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 609 * @napi_id: id of the NAPI struct this skb came from
984bc16c 610 * @secmark: security marking
0c4f691f 611 * @offload_fwd_mark: fwding offload mark
d84e0bd7 612 * @mark: Generic packet mark
86a9bad3 613 * @vlan_proto: vlan encapsulation protocol
6aa895b0 614 * @vlan_tci: vlan tag control information
0d89d203 615 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
616 * @inner_transport_header: Inner transport layer header (encapsulation)
617 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 618 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
619 * @transport_header: Transport layer header
620 * @network_header: Network layer header
621 * @mac_header: Link layer header
622 * @tail: Tail pointer
623 * @end: End pointer
624 * @head: Head of buffer
625 * @data: Data head pointer
626 * @truesize: Buffer size
627 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
628 */
629
630struct sk_buff {
363ec392 631 union {
56b17425
ED
632 struct {
633 /* These two members must be first. */
634 struct sk_buff *next;
635 struct sk_buff *prev;
636
637 union {
638 ktime_t tstamp;
639 struct skb_mstamp skb_mstamp;
640 };
641 };
642 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 643 };
da3f5cf1 644 struct sock *sk;
1da177e4 645 struct net_device *dev;
1da177e4 646
1da177e4
LT
647 /*
648 * This is the control buffer. It is free to use for every
649 * layer. Please put your private variables there. If you
650 * want to keep them across layers you have to do a skb_clone()
651 * first. This is owned by whoever has the skb queued ATM.
652 */
da3f5cf1 653 char cb[48] __aligned(8);
1da177e4 654
7fee226a 655 unsigned long _skb_refdst;
b1937227 656 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
657#ifdef CONFIG_XFRM
658 struct sec_path *sp;
b1937227
ED
659#endif
660#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
661 struct nf_conntrack *nfct;
662#endif
85224844 663#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 664 struct nf_bridge_info *nf_bridge;
da3f5cf1 665#endif
1da177e4 666 unsigned int len,
334a8132
PM
667 data_len;
668 __u16 mac_len,
669 hdr_len;
b1937227
ED
670
671 /* Following fields are _not_ copied in __copy_skb_header()
672 * Note that queue_mapping is here mostly to fill a hole.
673 */
fe55f6d5 674 kmemcheck_bitfield_begin(flags1);
b1937227
ED
675 __u16 queue_mapping;
676 __u8 cloned:1,
6869c4d8 677 nohdr:1,
b84f4cc9 678 fclone:2,
a59322be 679 peeked:1,
b1937227
ED
680 head_frag:1,
681 xmit_more:1;
682 /* one bit hole */
fe55f6d5 683 kmemcheck_bitfield_end(flags1);
4031ae6e 684
b1937227
ED
685 /* fields enclosed in headers_start/headers_end are copied
686 * using a single memcpy() in __copy_skb_header()
687 */
ebcf34f3 688 /* private: */
b1937227 689 __u32 headers_start[0];
ebcf34f3 690 /* public: */
4031ae6e 691
233577a2
HFS
692/* if you move pkt_type around you also must adapt those constants */
693#ifdef __BIG_ENDIAN_BITFIELD
694#define PKT_TYPE_MAX (7 << 5)
695#else
696#define PKT_TYPE_MAX 7
1da177e4 697#endif
233577a2 698#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 699
233577a2 700 __u8 __pkt_type_offset[0];
b1937227 701 __u8 pkt_type:3;
c93bdd0e 702 __u8 pfmemalloc:1;
b1937227
ED
703 __u8 ignore_df:1;
704 __u8 nfctinfo:3;
705
706 __u8 nf_trace:1;
707 __u8 ip_summed:2;
3853b584 708 __u8 ooo_okay:1;
61b905da 709 __u8 l4_hash:1;
a3b18ddb 710 __u8 sw_hash:1;
6e3e939f
JB
711 __u8 wifi_acked_valid:1;
712 __u8 wifi_acked:1;
b1937227 713
3bdc0eba 714 __u8 no_fcs:1;
77cffe23 715 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 716 __u8 encapsulation:1;
7e2b10c1 717 __u8 encap_hdr_csum:1;
5d0c2b95 718 __u8 csum_valid:1;
7e3cead5 719 __u8 csum_complete_sw:1;
b1937227
ED
720 __u8 csum_level:2;
721 __u8 csum_bad:1;
fe55f6d5 722
b1937227
ED
723#ifdef CONFIG_IPV6_NDISC_NODETYPE
724 __u8 ndisc_nodetype:2;
725#endif
726 __u8 ipvs_property:1;
8bce6d7d 727 __u8 inner_protocol_type:1;
e585f236
TH
728 __u8 remcsum_offload:1;
729 /* 3 or 5 bit hole */
b1937227
ED
730
731#ifdef CONFIG_NET_SCHED
732 __u16 tc_index; /* traffic control index */
733#ifdef CONFIG_NET_CLS_ACT
734 __u16 tc_verd; /* traffic control verdict */
735#endif
736#endif
fe55f6d5 737
b1937227
ED
738 union {
739 __wsum csum;
740 struct {
741 __u16 csum_start;
742 __u16 csum_offset;
743 };
744 };
745 __u32 priority;
746 int skb_iif;
747 __u32 hash;
748 __be16 vlan_proto;
749 __u16 vlan_tci;
2bd82484
ED
750#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
751 union {
752 unsigned int napi_id;
753 unsigned int sender_cpu;
754 };
97fc2f08 755#endif
0c4f691f 756 union {
984bc16c 757#ifdef CONFIG_NETWORK_SECMARK
0c4f691f
SF
758 __u32 secmark;
759#endif
760#ifdef CONFIG_NET_SWITCHDEV
761 __u32 offload_fwd_mark;
984bc16c 762#endif
0c4f691f
SF
763 };
764
3b885787
NH
765 union {
766 __u32 mark;
16fad69c 767 __u32 reserved_tailroom;
3b885787 768 };
1da177e4 769
8bce6d7d
TH
770 union {
771 __be16 inner_protocol;
772 __u8 inner_ipproto;
773 };
774
1a37e412
SH
775 __u16 inner_transport_header;
776 __u16 inner_network_header;
777 __u16 inner_mac_header;
b1937227
ED
778
779 __be16 protocol;
1a37e412
SH
780 __u16 transport_header;
781 __u16 network_header;
782 __u16 mac_header;
b1937227 783
ebcf34f3 784 /* private: */
b1937227 785 __u32 headers_end[0];
ebcf34f3 786 /* public: */
b1937227 787
1da177e4 788 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 789 sk_buff_data_t tail;
4305b541 790 sk_buff_data_t end;
1da177e4 791 unsigned char *head,
4305b541 792 *data;
27a884dc
ACM
793 unsigned int truesize;
794 atomic_t users;
1da177e4
LT
795};
796
797#ifdef __KERNEL__
798/*
799 * Handling routines are only of interest to the kernel
800 */
801#include <linux/slab.h>
802
1da177e4 803
c93bdd0e
MG
804#define SKB_ALLOC_FCLONE 0x01
805#define SKB_ALLOC_RX 0x02
fd11a83d 806#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
807
808/* Returns true if the skb was allocated from PFMEMALLOC reserves */
809static inline bool skb_pfmemalloc(const struct sk_buff *skb)
810{
811 return unlikely(skb->pfmemalloc);
812}
813
7fee226a
ED
814/*
815 * skb might have a dst pointer attached, refcounted or not.
816 * _skb_refdst low order bit is set if refcount was _not_ taken
817 */
818#define SKB_DST_NOREF 1UL
819#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
820
821/**
822 * skb_dst - returns skb dst_entry
823 * @skb: buffer
824 *
825 * Returns skb dst_entry, regardless of reference taken or not.
826 */
adf30907
ED
827static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
828{
7fee226a
ED
829 /* If refdst was not refcounted, check we still are in a
830 * rcu_read_lock section
831 */
832 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
833 !rcu_read_lock_held() &&
834 !rcu_read_lock_bh_held());
835 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
836}
837
7fee226a
ED
838/**
839 * skb_dst_set - sets skb dst
840 * @skb: buffer
841 * @dst: dst entry
842 *
843 * Sets skb dst, assuming a reference was taken on dst and should
844 * be released by skb_dst_drop()
845 */
adf30907
ED
846static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
847{
7fee226a
ED
848 skb->_skb_refdst = (unsigned long)dst;
849}
850
932bc4d7
JA
851/**
852 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
853 * @skb: buffer
854 * @dst: dst entry
855 *
856 * Sets skb dst, assuming a reference was not taken on dst.
857 * If dst entry is cached, we do not take reference and dst_release
858 * will be avoided by refdst_drop. If dst entry is not cached, we take
859 * reference, so that last dst_release can destroy the dst immediately.
860 */
861static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
862{
dbfc4fb7
HFS
863 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
864 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 865}
7fee226a
ED
866
867/**
25985edc 868 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
869 * @skb: buffer
870 */
871static inline bool skb_dst_is_noref(const struct sk_buff *skb)
872{
873 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
874}
875
511c3f92
ED
876static inline struct rtable *skb_rtable(const struct sk_buff *skb)
877{
adf30907 878 return (struct rtable *)skb_dst(skb);
511c3f92
ED
879}
880
7965bd4d
JP
881void kfree_skb(struct sk_buff *skb);
882void kfree_skb_list(struct sk_buff *segs);
883void skb_tx_error(struct sk_buff *skb);
884void consume_skb(struct sk_buff *skb);
885void __kfree_skb(struct sk_buff *skb);
d7e8883c 886extern struct kmem_cache *skbuff_head_cache;
bad43ca8 887
7965bd4d
JP
888void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
889bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
890 bool *fragstolen, int *delta_truesize);
bad43ca8 891
7965bd4d
JP
892struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
893 int node);
2ea2f62c 894struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 895struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 896static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 897 gfp_t priority)
d179cd12 898{
564824b0 899 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
900}
901
2e4e4410
ED
902struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
903 unsigned long data_len,
904 int max_page_order,
905 int *errcode,
906 gfp_t gfp_mask);
907
d0bf4a9e
ED
908/* Layout of fast clones : [skb1][skb2][fclone_ref] */
909struct sk_buff_fclones {
910 struct sk_buff skb1;
911
912 struct sk_buff skb2;
913
914 atomic_t fclone_ref;
915};
916
917/**
918 * skb_fclone_busy - check if fclone is busy
919 * @skb: buffer
920 *
bda13fed 921 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
922 * Some drivers call skb_orphan() in their ndo_start_xmit(),
923 * so we also check that this didnt happen.
d0bf4a9e 924 */
39bb5e62
ED
925static inline bool skb_fclone_busy(const struct sock *sk,
926 const struct sk_buff *skb)
d0bf4a9e
ED
927{
928 const struct sk_buff_fclones *fclones;
929
930 fclones = container_of(skb, struct sk_buff_fclones, skb1);
931
932 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 933 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 934 fclones->skb2.sk == sk;
d0bf4a9e
ED
935}
936
d179cd12 937static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 938 gfp_t priority)
d179cd12 939{
c93bdd0e 940 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
941}
942
7965bd4d 943struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
944static inline struct sk_buff *alloc_skb_head(gfp_t priority)
945{
946 return __alloc_skb_head(priority, -1);
947}
948
7965bd4d
JP
949struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
950int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
951struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
952struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
953struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
954 gfp_t gfp_mask, bool fclone);
955static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
956 gfp_t gfp_mask)
957{
958 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
959}
7965bd4d
JP
960
961int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
962struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
963 unsigned int headroom);
964struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
965 int newtailroom, gfp_t priority);
25a91d8d
FD
966int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
967 int offset, int len);
7965bd4d
JP
968int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
969 int len);
970int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
971int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 972#define dev_kfree_skb(a) consume_skb(a)
1da177e4 973
7965bd4d
JP
974int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
975 int getfrag(void *from, char *to, int offset,
976 int len, int odd, struct sk_buff *skb),
977 void *from, int length);
e89e9cf5 978
be12a1fe
HFS
979int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
980 int offset, size_t size);
981
d94d9fee 982struct skb_seq_state {
677e90ed
TG
983 __u32 lower_offset;
984 __u32 upper_offset;
985 __u32 frag_idx;
986 __u32 stepped_offset;
987 struct sk_buff *root_skb;
988 struct sk_buff *cur_skb;
989 __u8 *frag_data;
990};
991
7965bd4d
JP
992void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
993 unsigned int to, struct skb_seq_state *st);
994unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
995 struct skb_seq_state *st);
996void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 997
7965bd4d 998unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 999 unsigned int to, struct ts_config *config);
3fc7e8a6 1000
09323cc4
TH
1001/*
1002 * Packet hash types specify the type of hash in skb_set_hash.
1003 *
1004 * Hash types refer to the protocol layer addresses which are used to
1005 * construct a packet's hash. The hashes are used to differentiate or identify
1006 * flows of the protocol layer for the hash type. Hash types are either
1007 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1008 *
1009 * Properties of hashes:
1010 *
1011 * 1) Two packets in different flows have different hash values
1012 * 2) Two packets in the same flow should have the same hash value
1013 *
1014 * A hash at a higher layer is considered to be more specific. A driver should
1015 * set the most specific hash possible.
1016 *
1017 * A driver cannot indicate a more specific hash than the layer at which a hash
1018 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1019 *
1020 * A driver may indicate a hash level which is less specific than the
1021 * actual layer the hash was computed on. For instance, a hash computed
1022 * at L4 may be considered an L3 hash. This should only be done if the
1023 * driver can't unambiguously determine that the HW computed the hash at
1024 * the higher layer. Note that the "should" in the second property above
1025 * permits this.
1026 */
1027enum pkt_hash_types {
1028 PKT_HASH_TYPE_NONE, /* Undefined type */
1029 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1030 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1031 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1032};
1033
bcc83839 1034static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1035{
bcc83839 1036 skb->hash = 0;
a3b18ddb 1037 skb->sw_hash = 0;
bcc83839
TH
1038 skb->l4_hash = 0;
1039}
1040
1041static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1042{
1043 if (!skb->l4_hash)
1044 skb_clear_hash(skb);
1045}
1046
1047static inline void
1048__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1049{
1050 skb->l4_hash = is_l4;
1051 skb->sw_hash = is_sw;
61b905da 1052 skb->hash = hash;
09323cc4
TH
1053}
1054
bcc83839
TH
1055static inline void
1056skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1057{
1058 /* Used by drivers to set hash from HW */
1059 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1060}
1061
1062static inline void
1063__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1064{
1065 __skb_set_hash(skb, hash, true, is_l4);
1066}
1067
e5276937
TH
1068void __skb_get_hash(struct sk_buff *skb);
1069u32 skb_get_poff(const struct sk_buff *skb);
1070u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1071 const struct flow_keys *keys, int hlen);
1072__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1073 void *data, int hlen_proto);
1074
1075static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1076 int thoff, u8 ip_proto)
1077{
1078 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1079}
1080
1081void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1082 const struct flow_dissector_key *key,
1083 unsigned int key_count);
1084
1085bool __skb_flow_dissect(const struct sk_buff *skb,
1086 struct flow_dissector *flow_dissector,
1087 void *target_container,
cd79a238
TH
1088 void *data, __be16 proto, int nhoff, int hlen,
1089 unsigned int flags);
e5276937
TH
1090
1091static inline bool skb_flow_dissect(const struct sk_buff *skb,
1092 struct flow_dissector *flow_dissector,
cd79a238 1093 void *target_container, unsigned int flags)
e5276937
TH
1094{
1095 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1096 NULL, 0, 0, 0, flags);
e5276937
TH
1097}
1098
1099static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1100 struct flow_keys *flow,
1101 unsigned int flags)
e5276937
TH
1102{
1103 memset(flow, 0, sizeof(*flow));
1104 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1105 NULL, 0, 0, 0, flags);
e5276937
TH
1106}
1107
1108static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1109 void *data, __be16 proto,
cd79a238
TH
1110 int nhoff, int hlen,
1111 unsigned int flags)
e5276937
TH
1112{
1113 memset(flow, 0, sizeof(*flow));
1114 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1115 data, proto, nhoff, hlen, flags);
e5276937
TH
1116}
1117
3958afa1 1118static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1119{
a3b18ddb 1120 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1121 __skb_get_hash(skb);
bfb564e7 1122
61b905da 1123 return skb->hash;
bfb564e7
KK
1124}
1125
20a17bf6 1126__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1127
20a17bf6 1128static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1129{
c6cc1ca7
TH
1130 if (!skb->l4_hash && !skb->sw_hash) {
1131 struct flow_keys keys;
de4c1f8b 1132 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1133
de4c1f8b 1134 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1135 }
f70ea018
TH
1136
1137 return skb->hash;
1138}
1139
20a17bf6 1140__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1141
20a17bf6 1142static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1143{
c6cc1ca7
TH
1144 if (!skb->l4_hash && !skb->sw_hash) {
1145 struct flow_keys keys;
de4c1f8b 1146 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1147
de4c1f8b 1148 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1149 }
f70ea018
TH
1150
1151 return skb->hash;
1152}
1153
50fb7992
TH
1154__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1155
57bdf7f4
TH
1156static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1157{
61b905da 1158 return skb->hash;
57bdf7f4
TH
1159}
1160
3df7a74e
TH
1161static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1162{
61b905da 1163 to->hash = from->hash;
a3b18ddb 1164 to->sw_hash = from->sw_hash;
61b905da 1165 to->l4_hash = from->l4_hash;
3df7a74e
TH
1166};
1167
4305b541
ACM
1168#ifdef NET_SKBUFF_DATA_USES_OFFSET
1169static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1170{
1171 return skb->head + skb->end;
1172}
ec47ea82
AD
1173
1174static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1175{
1176 return skb->end;
1177}
4305b541
ACM
1178#else
1179static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1180{
1181 return skb->end;
1182}
ec47ea82
AD
1183
1184static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1185{
1186 return skb->end - skb->head;
1187}
4305b541
ACM
1188#endif
1189
1da177e4 1190/* Internal */
4305b541 1191#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1192
ac45f602
PO
1193static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1194{
1195 return &skb_shinfo(skb)->hwtstamps;
1196}
1197
1da177e4
LT
1198/**
1199 * skb_queue_empty - check if a queue is empty
1200 * @list: queue head
1201 *
1202 * Returns true if the queue is empty, false otherwise.
1203 */
1204static inline int skb_queue_empty(const struct sk_buff_head *list)
1205{
fd44b93c 1206 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1207}
1208
fc7ebb21
DM
1209/**
1210 * skb_queue_is_last - check if skb is the last entry in the queue
1211 * @list: queue head
1212 * @skb: buffer
1213 *
1214 * Returns true if @skb is the last buffer on the list.
1215 */
1216static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1217 const struct sk_buff *skb)
1218{
fd44b93c 1219 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1220}
1221
832d11c5
IJ
1222/**
1223 * skb_queue_is_first - check if skb is the first entry in the queue
1224 * @list: queue head
1225 * @skb: buffer
1226 *
1227 * Returns true if @skb is the first buffer on the list.
1228 */
1229static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1230 const struct sk_buff *skb)
1231{
fd44b93c 1232 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1233}
1234
249c8b42
DM
1235/**
1236 * skb_queue_next - return the next packet in the queue
1237 * @list: queue head
1238 * @skb: current buffer
1239 *
1240 * Return the next packet in @list after @skb. It is only valid to
1241 * call this if skb_queue_is_last() evaluates to false.
1242 */
1243static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1244 const struct sk_buff *skb)
1245{
1246 /* This BUG_ON may seem severe, but if we just return then we
1247 * are going to dereference garbage.
1248 */
1249 BUG_ON(skb_queue_is_last(list, skb));
1250 return skb->next;
1251}
1252
832d11c5
IJ
1253/**
1254 * skb_queue_prev - return the prev packet in the queue
1255 * @list: queue head
1256 * @skb: current buffer
1257 *
1258 * Return the prev packet in @list before @skb. It is only valid to
1259 * call this if skb_queue_is_first() evaluates to false.
1260 */
1261static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1262 const struct sk_buff *skb)
1263{
1264 /* This BUG_ON may seem severe, but if we just return then we
1265 * are going to dereference garbage.
1266 */
1267 BUG_ON(skb_queue_is_first(list, skb));
1268 return skb->prev;
1269}
1270
1da177e4
LT
1271/**
1272 * skb_get - reference buffer
1273 * @skb: buffer to reference
1274 *
1275 * Makes another reference to a socket buffer and returns a pointer
1276 * to the buffer.
1277 */
1278static inline struct sk_buff *skb_get(struct sk_buff *skb)
1279{
1280 atomic_inc(&skb->users);
1281 return skb;
1282}
1283
1284/*
1285 * If users == 1, we are the only owner and are can avoid redundant
1286 * atomic change.
1287 */
1288
1da177e4
LT
1289/**
1290 * skb_cloned - is the buffer a clone
1291 * @skb: buffer to check
1292 *
1293 * Returns true if the buffer was generated with skb_clone() and is
1294 * one of multiple shared copies of the buffer. Cloned buffers are
1295 * shared data so must not be written to under normal circumstances.
1296 */
1297static inline int skb_cloned(const struct sk_buff *skb)
1298{
1299 return skb->cloned &&
1300 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1301}
1302
14bbd6a5
PS
1303static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1304{
d0164adc 1305 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1306
1307 if (skb_cloned(skb))
1308 return pskb_expand_head(skb, 0, 0, pri);
1309
1310 return 0;
1311}
1312
1da177e4
LT
1313/**
1314 * skb_header_cloned - is the header a clone
1315 * @skb: buffer to check
1316 *
1317 * Returns true if modifying the header part of the buffer requires
1318 * the data to be copied.
1319 */
1320static inline int skb_header_cloned(const struct sk_buff *skb)
1321{
1322 int dataref;
1323
1324 if (!skb->cloned)
1325 return 0;
1326
1327 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1328 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1329 return dataref != 1;
1330}
1331
1332/**
1333 * skb_header_release - release reference to header
1334 * @skb: buffer to operate on
1335 *
1336 * Drop a reference to the header part of the buffer. This is done
1337 * by acquiring a payload reference. You must not read from the header
1338 * part of skb->data after this.
f4a775d1 1339 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1340 */
1341static inline void skb_header_release(struct sk_buff *skb)
1342{
1343 BUG_ON(skb->nohdr);
1344 skb->nohdr = 1;
1345 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1346}
1347
f4a775d1
ED
1348/**
1349 * __skb_header_release - release reference to header
1350 * @skb: buffer to operate on
1351 *
1352 * Variant of skb_header_release() assuming skb is private to caller.
1353 * We can avoid one atomic operation.
1354 */
1355static inline void __skb_header_release(struct sk_buff *skb)
1356{
1357 skb->nohdr = 1;
1358 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1359}
1360
1361
1da177e4
LT
1362/**
1363 * skb_shared - is the buffer shared
1364 * @skb: buffer to check
1365 *
1366 * Returns true if more than one person has a reference to this
1367 * buffer.
1368 */
1369static inline int skb_shared(const struct sk_buff *skb)
1370{
1371 return atomic_read(&skb->users) != 1;
1372}
1373
1374/**
1375 * skb_share_check - check if buffer is shared and if so clone it
1376 * @skb: buffer to check
1377 * @pri: priority for memory allocation
1378 *
1379 * If the buffer is shared the buffer is cloned and the old copy
1380 * drops a reference. A new clone with a single reference is returned.
1381 * If the buffer is not shared the original buffer is returned. When
1382 * being called from interrupt status or with spinlocks held pri must
1383 * be GFP_ATOMIC.
1384 *
1385 * NULL is returned on a memory allocation failure.
1386 */
47061bc4 1387static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1388{
d0164adc 1389 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1390 if (skb_shared(skb)) {
1391 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1392
1393 if (likely(nskb))
1394 consume_skb(skb);
1395 else
1396 kfree_skb(skb);
1da177e4
LT
1397 skb = nskb;
1398 }
1399 return skb;
1400}
1401
1402/*
1403 * Copy shared buffers into a new sk_buff. We effectively do COW on
1404 * packets to handle cases where we have a local reader and forward
1405 * and a couple of other messy ones. The normal one is tcpdumping
1406 * a packet thats being forwarded.
1407 */
1408
1409/**
1410 * skb_unshare - make a copy of a shared buffer
1411 * @skb: buffer to check
1412 * @pri: priority for memory allocation
1413 *
1414 * If the socket buffer is a clone then this function creates a new
1415 * copy of the data, drops a reference count on the old copy and returns
1416 * the new copy with the reference count at 1. If the buffer is not a clone
1417 * the original buffer is returned. When called with a spinlock held or
1418 * from interrupt state @pri must be %GFP_ATOMIC
1419 *
1420 * %NULL is returned on a memory allocation failure.
1421 */
e2bf521d 1422static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1423 gfp_t pri)
1da177e4 1424{
d0164adc 1425 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1426 if (skb_cloned(skb)) {
1427 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1428
1429 /* Free our shared copy */
1430 if (likely(nskb))
1431 consume_skb(skb);
1432 else
1433 kfree_skb(skb);
1da177e4
LT
1434 skb = nskb;
1435 }
1436 return skb;
1437}
1438
1439/**
1a5778aa 1440 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1441 * @list_: list to peek at
1442 *
1443 * Peek an &sk_buff. Unlike most other operations you _MUST_
1444 * be careful with this one. A peek leaves the buffer on the
1445 * list and someone else may run off with it. You must hold
1446 * the appropriate locks or have a private queue to do this.
1447 *
1448 * Returns %NULL for an empty list or a pointer to the head element.
1449 * The reference count is not incremented and the reference is therefore
1450 * volatile. Use with caution.
1451 */
05bdd2f1 1452static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1453{
18d07000
ED
1454 struct sk_buff *skb = list_->next;
1455
1456 if (skb == (struct sk_buff *)list_)
1457 skb = NULL;
1458 return skb;
1da177e4
LT
1459}
1460
da5ef6e5
PE
1461/**
1462 * skb_peek_next - peek skb following the given one from a queue
1463 * @skb: skb to start from
1464 * @list_: list to peek at
1465 *
1466 * Returns %NULL when the end of the list is met or a pointer to the
1467 * next element. The reference count is not incremented and the
1468 * reference is therefore volatile. Use with caution.
1469 */
1470static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1471 const struct sk_buff_head *list_)
1472{
1473 struct sk_buff *next = skb->next;
18d07000 1474
da5ef6e5
PE
1475 if (next == (struct sk_buff *)list_)
1476 next = NULL;
1477 return next;
1478}
1479
1da177e4 1480/**
1a5778aa 1481 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1482 * @list_: list to peek at
1483 *
1484 * Peek an &sk_buff. Unlike most other operations you _MUST_
1485 * be careful with this one. A peek leaves the buffer on the
1486 * list and someone else may run off with it. You must hold
1487 * the appropriate locks or have a private queue to do this.
1488 *
1489 * Returns %NULL for an empty list or a pointer to the tail element.
1490 * The reference count is not incremented and the reference is therefore
1491 * volatile. Use with caution.
1492 */
05bdd2f1 1493static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1494{
18d07000
ED
1495 struct sk_buff *skb = list_->prev;
1496
1497 if (skb == (struct sk_buff *)list_)
1498 skb = NULL;
1499 return skb;
1500
1da177e4
LT
1501}
1502
1503/**
1504 * skb_queue_len - get queue length
1505 * @list_: list to measure
1506 *
1507 * Return the length of an &sk_buff queue.
1508 */
1509static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1510{
1511 return list_->qlen;
1512}
1513
67fed459
DM
1514/**
1515 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1516 * @list: queue to initialize
1517 *
1518 * This initializes only the list and queue length aspects of
1519 * an sk_buff_head object. This allows to initialize the list
1520 * aspects of an sk_buff_head without reinitializing things like
1521 * the spinlock. It can also be used for on-stack sk_buff_head
1522 * objects where the spinlock is known to not be used.
1523 */
1524static inline void __skb_queue_head_init(struct sk_buff_head *list)
1525{
1526 list->prev = list->next = (struct sk_buff *)list;
1527 list->qlen = 0;
1528}
1529
76f10ad0
AV
1530/*
1531 * This function creates a split out lock class for each invocation;
1532 * this is needed for now since a whole lot of users of the skb-queue
1533 * infrastructure in drivers have different locking usage (in hardirq)
1534 * than the networking core (in softirq only). In the long run either the
1535 * network layer or drivers should need annotation to consolidate the
1536 * main types of usage into 3 classes.
1537 */
1da177e4
LT
1538static inline void skb_queue_head_init(struct sk_buff_head *list)
1539{
1540 spin_lock_init(&list->lock);
67fed459 1541 __skb_queue_head_init(list);
1da177e4
LT
1542}
1543
c2ecba71
PE
1544static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1545 struct lock_class_key *class)
1546{
1547 skb_queue_head_init(list);
1548 lockdep_set_class(&list->lock, class);
1549}
1550
1da177e4 1551/*
bf299275 1552 * Insert an sk_buff on a list.
1da177e4
LT
1553 *
1554 * The "__skb_xxxx()" functions are the non-atomic ones that
1555 * can only be called with interrupts disabled.
1556 */
7965bd4d
JP
1557void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1558 struct sk_buff_head *list);
bf299275
GR
1559static inline void __skb_insert(struct sk_buff *newsk,
1560 struct sk_buff *prev, struct sk_buff *next,
1561 struct sk_buff_head *list)
1562{
1563 newsk->next = next;
1564 newsk->prev = prev;
1565 next->prev = prev->next = newsk;
1566 list->qlen++;
1567}
1da177e4 1568
67fed459
DM
1569static inline void __skb_queue_splice(const struct sk_buff_head *list,
1570 struct sk_buff *prev,
1571 struct sk_buff *next)
1572{
1573 struct sk_buff *first = list->next;
1574 struct sk_buff *last = list->prev;
1575
1576 first->prev = prev;
1577 prev->next = first;
1578
1579 last->next = next;
1580 next->prev = last;
1581}
1582
1583/**
1584 * skb_queue_splice - join two skb lists, this is designed for stacks
1585 * @list: the new list to add
1586 * @head: the place to add it in the first list
1587 */
1588static inline void skb_queue_splice(const struct sk_buff_head *list,
1589 struct sk_buff_head *head)
1590{
1591 if (!skb_queue_empty(list)) {
1592 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1593 head->qlen += list->qlen;
67fed459
DM
1594 }
1595}
1596
1597/**
d9619496 1598 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1599 * @list: the new list to add
1600 * @head: the place to add it in the first list
1601 *
1602 * The list at @list is reinitialised
1603 */
1604static inline void skb_queue_splice_init(struct sk_buff_head *list,
1605 struct sk_buff_head *head)
1606{
1607 if (!skb_queue_empty(list)) {
1608 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1609 head->qlen += list->qlen;
67fed459
DM
1610 __skb_queue_head_init(list);
1611 }
1612}
1613
1614/**
1615 * skb_queue_splice_tail - join two skb lists, each list being a queue
1616 * @list: the new list to add
1617 * @head: the place to add it in the first list
1618 */
1619static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1620 struct sk_buff_head *head)
1621{
1622 if (!skb_queue_empty(list)) {
1623 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1624 head->qlen += list->qlen;
67fed459
DM
1625 }
1626}
1627
1628/**
d9619496 1629 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1630 * @list: the new list to add
1631 * @head: the place to add it in the first list
1632 *
1633 * Each of the lists is a queue.
1634 * The list at @list is reinitialised
1635 */
1636static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1637 struct sk_buff_head *head)
1638{
1639 if (!skb_queue_empty(list)) {
1640 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1641 head->qlen += list->qlen;
67fed459
DM
1642 __skb_queue_head_init(list);
1643 }
1644}
1645
1da177e4 1646/**
300ce174 1647 * __skb_queue_after - queue a buffer at the list head
1da177e4 1648 * @list: list to use
300ce174 1649 * @prev: place after this buffer
1da177e4
LT
1650 * @newsk: buffer to queue
1651 *
300ce174 1652 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1653 * and you must therefore hold required locks before calling it.
1654 *
1655 * A buffer cannot be placed on two lists at the same time.
1656 */
300ce174
SH
1657static inline void __skb_queue_after(struct sk_buff_head *list,
1658 struct sk_buff *prev,
1659 struct sk_buff *newsk)
1da177e4 1660{
bf299275 1661 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1662}
1663
7965bd4d
JP
1664void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1665 struct sk_buff_head *list);
7de6c033 1666
f5572855
GR
1667static inline void __skb_queue_before(struct sk_buff_head *list,
1668 struct sk_buff *next,
1669 struct sk_buff *newsk)
1670{
1671 __skb_insert(newsk, next->prev, next, list);
1672}
1673
300ce174
SH
1674/**
1675 * __skb_queue_head - queue a buffer at the list head
1676 * @list: list to use
1677 * @newsk: buffer to queue
1678 *
1679 * Queue a buffer at the start of a list. This function takes no locks
1680 * and you must therefore hold required locks before calling it.
1681 *
1682 * A buffer cannot be placed on two lists at the same time.
1683 */
7965bd4d 1684void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1685static inline void __skb_queue_head(struct sk_buff_head *list,
1686 struct sk_buff *newsk)
1687{
1688 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1689}
1690
1da177e4
LT
1691/**
1692 * __skb_queue_tail - queue a buffer at the list tail
1693 * @list: list to use
1694 * @newsk: buffer to queue
1695 *
1696 * Queue a buffer at the end of a list. This function takes no locks
1697 * and you must therefore hold required locks before calling it.
1698 *
1699 * A buffer cannot be placed on two lists at the same time.
1700 */
7965bd4d 1701void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1702static inline void __skb_queue_tail(struct sk_buff_head *list,
1703 struct sk_buff *newsk)
1704{
f5572855 1705 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1706}
1707
1da177e4
LT
1708/*
1709 * remove sk_buff from list. _Must_ be called atomically, and with
1710 * the list known..
1711 */
7965bd4d 1712void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1713static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1714{
1715 struct sk_buff *next, *prev;
1716
1717 list->qlen--;
1718 next = skb->next;
1719 prev = skb->prev;
1720 skb->next = skb->prev = NULL;
1da177e4
LT
1721 next->prev = prev;
1722 prev->next = next;
1723}
1724
f525c06d
GR
1725/**
1726 * __skb_dequeue - remove from the head of the queue
1727 * @list: list to dequeue from
1728 *
1729 * Remove the head of the list. This function does not take any locks
1730 * so must be used with appropriate locks held only. The head item is
1731 * returned or %NULL if the list is empty.
1732 */
7965bd4d 1733struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1734static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1735{
1736 struct sk_buff *skb = skb_peek(list);
1737 if (skb)
1738 __skb_unlink(skb, list);
1739 return skb;
1740}
1da177e4
LT
1741
1742/**
1743 * __skb_dequeue_tail - remove from the tail of the queue
1744 * @list: list to dequeue from
1745 *
1746 * Remove the tail of the list. This function does not take any locks
1747 * so must be used with appropriate locks held only. The tail item is
1748 * returned or %NULL if the list is empty.
1749 */
7965bd4d 1750struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1751static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1752{
1753 struct sk_buff *skb = skb_peek_tail(list);
1754 if (skb)
1755 __skb_unlink(skb, list);
1756 return skb;
1757}
1758
1759
bdcc0924 1760static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1761{
1762 return skb->data_len;
1763}
1764
1765static inline unsigned int skb_headlen(const struct sk_buff *skb)
1766{
1767 return skb->len - skb->data_len;
1768}
1769
1770static inline int skb_pagelen(const struct sk_buff *skb)
1771{
1772 int i, len = 0;
1773
1774 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1775 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1776 return len + skb_headlen(skb);
1777}
1778
131ea667
IC
1779/**
1780 * __skb_fill_page_desc - initialise a paged fragment in an skb
1781 * @skb: buffer containing fragment to be initialised
1782 * @i: paged fragment index to initialise
1783 * @page: the page to use for this fragment
1784 * @off: the offset to the data with @page
1785 * @size: the length of the data
1786 *
1787 * Initialises the @i'th fragment of @skb to point to &size bytes at
1788 * offset @off within @page.
1789 *
1790 * Does not take any additional reference on the fragment.
1791 */
1792static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1793 struct page *page, int off, int size)
1da177e4
LT
1794{
1795 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1796
c48a11c7 1797 /*
2f064f34
MH
1798 * Propagate page pfmemalloc to the skb if we can. The problem is
1799 * that not all callers have unique ownership of the page but rely
1800 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1801 */
a8605c60 1802 frag->page.p = page;
1da177e4 1803 frag->page_offset = off;
9e903e08 1804 skb_frag_size_set(frag, size);
cca7af38
PE
1805
1806 page = compound_head(page);
2f064f34 1807 if (page_is_pfmemalloc(page))
cca7af38 1808 skb->pfmemalloc = true;
131ea667
IC
1809}
1810
1811/**
1812 * skb_fill_page_desc - initialise a paged fragment in an skb
1813 * @skb: buffer containing fragment to be initialised
1814 * @i: paged fragment index to initialise
1815 * @page: the page to use for this fragment
1816 * @off: the offset to the data with @page
1817 * @size: the length of the data
1818 *
1819 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1820 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1821 * addition updates @skb such that @i is the last fragment.
1822 *
1823 * Does not take any additional reference on the fragment.
1824 */
1825static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1826 struct page *page, int off, int size)
1827{
1828 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1829 skb_shinfo(skb)->nr_frags = i + 1;
1830}
1831
7965bd4d
JP
1832void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1833 int size, unsigned int truesize);
654bed16 1834
f8e617e1
JW
1835void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1836 unsigned int truesize);
1837
1da177e4 1838#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1839#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1840#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1841
27a884dc
ACM
1842#ifdef NET_SKBUFF_DATA_USES_OFFSET
1843static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1844{
1845 return skb->head + skb->tail;
1846}
1847
1848static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1849{
1850 skb->tail = skb->data - skb->head;
1851}
1852
1853static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1854{
1855 skb_reset_tail_pointer(skb);
1856 skb->tail += offset;
1857}
7cc46190 1858
27a884dc
ACM
1859#else /* NET_SKBUFF_DATA_USES_OFFSET */
1860static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1861{
1862 return skb->tail;
1863}
1864
1865static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1866{
1867 skb->tail = skb->data;
1868}
1869
1870static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1871{
1872 skb->tail = skb->data + offset;
1873}
4305b541 1874
27a884dc
ACM
1875#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1876
1da177e4
LT
1877/*
1878 * Add data to an sk_buff
1879 */
0c7ddf36 1880unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1881unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1882static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1883{
27a884dc 1884 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1885 SKB_LINEAR_ASSERT(skb);
1886 skb->tail += len;
1887 skb->len += len;
1888 return tmp;
1889}
1890
7965bd4d 1891unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1892static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1893{
1894 skb->data -= len;
1895 skb->len += len;
1896 return skb->data;
1897}
1898
7965bd4d 1899unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1900static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1901{
1902 skb->len -= len;
1903 BUG_ON(skb->len < skb->data_len);
1904 return skb->data += len;
1905}
1906
47d29646
DM
1907static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1908{
1909 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1910}
1911
7965bd4d 1912unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1913
1914static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1915{
1916 if (len > skb_headlen(skb) &&
987c402a 1917 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1918 return NULL;
1919 skb->len -= len;
1920 return skb->data += len;
1921}
1922
1923static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1924{
1925 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1926}
1927
1928static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1929{
1930 if (likely(len <= skb_headlen(skb)))
1931 return 1;
1932 if (unlikely(len > skb->len))
1933 return 0;
987c402a 1934 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1935}
1936
1937/**
1938 * skb_headroom - bytes at buffer head
1939 * @skb: buffer to check
1940 *
1941 * Return the number of bytes of free space at the head of an &sk_buff.
1942 */
c2636b4d 1943static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1944{
1945 return skb->data - skb->head;
1946}
1947
1948/**
1949 * skb_tailroom - bytes at buffer end
1950 * @skb: buffer to check
1951 *
1952 * Return the number of bytes of free space at the tail of an sk_buff
1953 */
1954static inline int skb_tailroom(const struct sk_buff *skb)
1955{
4305b541 1956 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1957}
1958
a21d4572
ED
1959/**
1960 * skb_availroom - bytes at buffer end
1961 * @skb: buffer to check
1962 *
1963 * Return the number of bytes of free space at the tail of an sk_buff
1964 * allocated by sk_stream_alloc()
1965 */
1966static inline int skb_availroom(const struct sk_buff *skb)
1967{
16fad69c
ED
1968 if (skb_is_nonlinear(skb))
1969 return 0;
1970
1971 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1972}
1973
1da177e4
LT
1974/**
1975 * skb_reserve - adjust headroom
1976 * @skb: buffer to alter
1977 * @len: bytes to move
1978 *
1979 * Increase the headroom of an empty &sk_buff by reducing the tail
1980 * room. This is only allowed for an empty buffer.
1981 */
8243126c 1982static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1983{
1984 skb->data += len;
1985 skb->tail += len;
1986}
1987
1837b2e2
BP
1988/**
1989 * skb_tailroom_reserve - adjust reserved_tailroom
1990 * @skb: buffer to alter
1991 * @mtu: maximum amount of headlen permitted
1992 * @needed_tailroom: minimum amount of reserved_tailroom
1993 *
1994 * Set reserved_tailroom so that headlen can be as large as possible but
1995 * not larger than mtu and tailroom cannot be smaller than
1996 * needed_tailroom.
1997 * The required headroom should already have been reserved before using
1998 * this function.
1999 */
2000static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2001 unsigned int needed_tailroom)
2002{
2003 SKB_LINEAR_ASSERT(skb);
2004 if (mtu < skb_tailroom(skb) - needed_tailroom)
2005 /* use at most mtu */
2006 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2007 else
2008 /* use up to all available space */
2009 skb->reserved_tailroom = needed_tailroom;
2010}
2011
8bce6d7d
TH
2012#define ENCAP_TYPE_ETHER 0
2013#define ENCAP_TYPE_IPPROTO 1
2014
2015static inline void skb_set_inner_protocol(struct sk_buff *skb,
2016 __be16 protocol)
2017{
2018 skb->inner_protocol = protocol;
2019 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2020}
2021
2022static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2023 __u8 ipproto)
2024{
2025 skb->inner_ipproto = ipproto;
2026 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2027}
2028
6a674e9c
JG
2029static inline void skb_reset_inner_headers(struct sk_buff *skb)
2030{
aefbd2b3 2031 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2032 skb->inner_network_header = skb->network_header;
2033 skb->inner_transport_header = skb->transport_header;
2034}
2035
0b5c9db1
JP
2036static inline void skb_reset_mac_len(struct sk_buff *skb)
2037{
2038 skb->mac_len = skb->network_header - skb->mac_header;
2039}
2040
6a674e9c
JG
2041static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2042 *skb)
2043{
2044 return skb->head + skb->inner_transport_header;
2045}
2046
55dc5a9f
TH
2047static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2048{
2049 return skb_inner_transport_header(skb) - skb->data;
2050}
2051
6a674e9c
JG
2052static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2053{
2054 skb->inner_transport_header = skb->data - skb->head;
2055}
2056
2057static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2058 const int offset)
2059{
2060 skb_reset_inner_transport_header(skb);
2061 skb->inner_transport_header += offset;
2062}
2063
2064static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2065{
2066 return skb->head + skb->inner_network_header;
2067}
2068
2069static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2070{
2071 skb->inner_network_header = skb->data - skb->head;
2072}
2073
2074static inline void skb_set_inner_network_header(struct sk_buff *skb,
2075 const int offset)
2076{
2077 skb_reset_inner_network_header(skb);
2078 skb->inner_network_header += offset;
2079}
2080
aefbd2b3
PS
2081static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2082{
2083 return skb->head + skb->inner_mac_header;
2084}
2085
2086static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2087{
2088 skb->inner_mac_header = skb->data - skb->head;
2089}
2090
2091static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2092 const int offset)
2093{
2094 skb_reset_inner_mac_header(skb);
2095 skb->inner_mac_header += offset;
2096}
fda55eca
ED
2097static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2098{
35d04610 2099 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2100}
2101
9c70220b
ACM
2102static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2103{
2e07fa9c 2104 return skb->head + skb->transport_header;
9c70220b
ACM
2105}
2106
badff6d0
ACM
2107static inline void skb_reset_transport_header(struct sk_buff *skb)
2108{
2e07fa9c 2109 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2110}
2111
967b05f6
ACM
2112static inline void skb_set_transport_header(struct sk_buff *skb,
2113 const int offset)
2114{
2e07fa9c
ACM
2115 skb_reset_transport_header(skb);
2116 skb->transport_header += offset;
ea2ae17d
ACM
2117}
2118
d56f90a7
ACM
2119static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2120{
2e07fa9c 2121 return skb->head + skb->network_header;
d56f90a7
ACM
2122}
2123
c1d2bbe1
ACM
2124static inline void skb_reset_network_header(struct sk_buff *skb)
2125{
2e07fa9c 2126 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2127}
2128
c14d2450
ACM
2129static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2130{
2e07fa9c
ACM
2131 skb_reset_network_header(skb);
2132 skb->network_header += offset;
c14d2450
ACM
2133}
2134
2e07fa9c 2135static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2136{
2e07fa9c 2137 return skb->head + skb->mac_header;
bbe735e4
ACM
2138}
2139
2e07fa9c 2140static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2141{
35d04610 2142 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2143}
2144
2145static inline void skb_reset_mac_header(struct sk_buff *skb)
2146{
2147 skb->mac_header = skb->data - skb->head;
2148}
2149
2150static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2151{
2152 skb_reset_mac_header(skb);
2153 skb->mac_header += offset;
2154}
2155
0e3da5bb
TT
2156static inline void skb_pop_mac_header(struct sk_buff *skb)
2157{
2158 skb->mac_header = skb->network_header;
2159}
2160
fbbdb8f0
YX
2161static inline void skb_probe_transport_header(struct sk_buff *skb,
2162 const int offset_hint)
2163{
2164 struct flow_keys keys;
2165
2166 if (skb_transport_header_was_set(skb))
2167 return;
cd79a238 2168 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2169 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2170 else
2171 skb_set_transport_header(skb, offset_hint);
2172}
2173
03606895
ED
2174static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2175{
2176 if (skb_mac_header_was_set(skb)) {
2177 const unsigned char *old_mac = skb_mac_header(skb);
2178
2179 skb_set_mac_header(skb, -skb->mac_len);
2180 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2181 }
2182}
2183
04fb451e
MM
2184static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2185{
2186 return skb->csum_start - skb_headroom(skb);
2187}
2188
08b64fcc
AD
2189static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2190{
2191 return skb->head + skb->csum_start;
2192}
2193
2e07fa9c
ACM
2194static inline int skb_transport_offset(const struct sk_buff *skb)
2195{
2196 return skb_transport_header(skb) - skb->data;
2197}
2198
2199static inline u32 skb_network_header_len(const struct sk_buff *skb)
2200{
2201 return skb->transport_header - skb->network_header;
2202}
2203
6a674e9c
JG
2204static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2205{
2206 return skb->inner_transport_header - skb->inner_network_header;
2207}
2208
2e07fa9c
ACM
2209static inline int skb_network_offset(const struct sk_buff *skb)
2210{
2211 return skb_network_header(skb) - skb->data;
2212}
48d49d0c 2213
6a674e9c
JG
2214static inline int skb_inner_network_offset(const struct sk_buff *skb)
2215{
2216 return skb_inner_network_header(skb) - skb->data;
2217}
2218
f9599ce1
CG
2219static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2220{
2221 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2222}
2223
1da177e4
LT
2224/*
2225 * CPUs often take a performance hit when accessing unaligned memory
2226 * locations. The actual performance hit varies, it can be small if the
2227 * hardware handles it or large if we have to take an exception and fix it
2228 * in software.
2229 *
2230 * Since an ethernet header is 14 bytes network drivers often end up with
2231 * the IP header at an unaligned offset. The IP header can be aligned by
2232 * shifting the start of the packet by 2 bytes. Drivers should do this
2233 * with:
2234 *
8660c124 2235 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2236 *
2237 * The downside to this alignment of the IP header is that the DMA is now
2238 * unaligned. On some architectures the cost of an unaligned DMA is high
2239 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2240 *
1da177e4
LT
2241 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2242 * to be overridden.
2243 */
2244#ifndef NET_IP_ALIGN
2245#define NET_IP_ALIGN 2
2246#endif
2247
025be81e
AB
2248/*
2249 * The networking layer reserves some headroom in skb data (via
2250 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2251 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2252 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2253 *
2254 * Unfortunately this headroom changes the DMA alignment of the resulting
2255 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2256 * on some architectures. An architecture can override this value,
2257 * perhaps setting it to a cacheline in size (since that will maintain
2258 * cacheline alignment of the DMA). It must be a power of 2.
2259 *
d6301d3d 2260 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2261 * headroom, you should not reduce this.
5933dd2f
ED
2262 *
2263 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2264 * to reduce average number of cache lines per packet.
2265 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2266 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2267 */
2268#ifndef NET_SKB_PAD
5933dd2f 2269#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2270#endif
2271
7965bd4d 2272int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2273
2274static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2275{
c4264f27 2276 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2277 WARN_ON(1);
2278 return;
2279 }
27a884dc
ACM
2280 skb->len = len;
2281 skb_set_tail_pointer(skb, len);
1da177e4
LT
2282}
2283
7965bd4d 2284void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2285
2286static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2287{
3cc0e873
HX
2288 if (skb->data_len)
2289 return ___pskb_trim(skb, len);
2290 __skb_trim(skb, len);
2291 return 0;
1da177e4
LT
2292}
2293
2294static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2295{
2296 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2297}
2298
e9fa4f7b
HX
2299/**
2300 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2301 * @skb: buffer to alter
2302 * @len: new length
2303 *
2304 * This is identical to pskb_trim except that the caller knows that
2305 * the skb is not cloned so we should never get an error due to out-
2306 * of-memory.
2307 */
2308static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2309{
2310 int err = pskb_trim(skb, len);
2311 BUG_ON(err);
2312}
2313
1da177e4
LT
2314/**
2315 * skb_orphan - orphan a buffer
2316 * @skb: buffer to orphan
2317 *
2318 * If a buffer currently has an owner then we call the owner's
2319 * destructor function and make the @skb unowned. The buffer continues
2320 * to exist but is no longer charged to its former owner.
2321 */
2322static inline void skb_orphan(struct sk_buff *skb)
2323{
c34a7612 2324 if (skb->destructor) {
1da177e4 2325 skb->destructor(skb);
c34a7612
ED
2326 skb->destructor = NULL;
2327 skb->sk = NULL;
376c7311
ED
2328 } else {
2329 BUG_ON(skb->sk);
c34a7612 2330 }
1da177e4
LT
2331}
2332
a353e0ce
MT
2333/**
2334 * skb_orphan_frags - orphan the frags contained in a buffer
2335 * @skb: buffer to orphan frags from
2336 * @gfp_mask: allocation mask for replacement pages
2337 *
2338 * For each frag in the SKB which needs a destructor (i.e. has an
2339 * owner) create a copy of that frag and release the original
2340 * page by calling the destructor.
2341 */
2342static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2343{
2344 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2345 return 0;
2346 return skb_copy_ubufs(skb, gfp_mask);
2347}
2348
1da177e4
LT
2349/**
2350 * __skb_queue_purge - empty a list
2351 * @list: list to empty
2352 *
2353 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2354 * the list and one reference dropped. This function does not take the
2355 * list lock and the caller must hold the relevant locks to use it.
2356 */
7965bd4d 2357void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2358static inline void __skb_queue_purge(struct sk_buff_head *list)
2359{
2360 struct sk_buff *skb;
2361 while ((skb = __skb_dequeue(list)) != NULL)
2362 kfree_skb(skb);
2363}
2364
7965bd4d 2365void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2366
7965bd4d
JP
2367struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2368 gfp_t gfp_mask);
8af27456
CH
2369
2370/**
2371 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2372 * @dev: network device to receive on
2373 * @length: length to allocate
2374 *
2375 * Allocate a new &sk_buff and assign it a usage count of one. The
2376 * buffer has unspecified headroom built in. Users should allocate
2377 * the headroom they think they need without accounting for the
2378 * built in space. The built in space is used for optimisations.
2379 *
2380 * %NULL is returned if there is no free memory. Although this function
2381 * allocates memory it can be called from an interrupt.
2382 */
2383static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2384 unsigned int length)
8af27456
CH
2385{
2386 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2387}
2388
6f532612
ED
2389/* legacy helper around __netdev_alloc_skb() */
2390static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2391 gfp_t gfp_mask)
2392{
2393 return __netdev_alloc_skb(NULL, length, gfp_mask);
2394}
2395
2396/* legacy helper around netdev_alloc_skb() */
2397static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2398{
2399 return netdev_alloc_skb(NULL, length);
2400}
2401
2402
4915a0de
ED
2403static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2404 unsigned int length, gfp_t gfp)
61321bbd 2405{
4915a0de 2406 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2407
2408 if (NET_IP_ALIGN && skb)
2409 skb_reserve(skb, NET_IP_ALIGN);
2410 return skb;
2411}
2412
4915a0de
ED
2413static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2414 unsigned int length)
2415{
2416 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2417}
2418
181edb2b
AD
2419static inline void skb_free_frag(void *addr)
2420{
2421 __free_page_frag(addr);
2422}
2423
ffde7328 2424void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2425struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2426 unsigned int length, gfp_t gfp_mask);
2427static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2428 unsigned int length)
2429{
2430 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2431}
795bb1c0
JDB
2432void napi_consume_skb(struct sk_buff *skb, int budget);
2433
2434void __kfree_skb_flush(void);
15fad714 2435void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2436
71dfda58
AD
2437/**
2438 * __dev_alloc_pages - allocate page for network Rx
2439 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2440 * @order: size of the allocation
2441 *
2442 * Allocate a new page.
2443 *
2444 * %NULL is returned if there is no free memory.
2445*/
2446static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2447 unsigned int order)
2448{
2449 /* This piece of code contains several assumptions.
2450 * 1. This is for device Rx, therefor a cold page is preferred.
2451 * 2. The expectation is the user wants a compound page.
2452 * 3. If requesting a order 0 page it will not be compound
2453 * due to the check to see if order has a value in prep_new_page
2454 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2455 * code in gfp_to_alloc_flags that should be enforcing this.
2456 */
2457 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2458
2459 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2460}
2461
2462static inline struct page *dev_alloc_pages(unsigned int order)
2463{
2464 return __dev_alloc_pages(GFP_ATOMIC, order);
2465}
2466
2467/**
2468 * __dev_alloc_page - allocate a page for network Rx
2469 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2470 *
2471 * Allocate a new page.
2472 *
2473 * %NULL is returned if there is no free memory.
2474 */
2475static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2476{
2477 return __dev_alloc_pages(gfp_mask, 0);
2478}
2479
2480static inline struct page *dev_alloc_page(void)
2481{
2482 return __dev_alloc_page(GFP_ATOMIC);
2483}
2484
0614002b
MG
2485/**
2486 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2487 * @page: The page that was allocated from skb_alloc_page
2488 * @skb: The skb that may need pfmemalloc set
2489 */
2490static inline void skb_propagate_pfmemalloc(struct page *page,
2491 struct sk_buff *skb)
2492{
2f064f34 2493 if (page_is_pfmemalloc(page))
0614002b
MG
2494 skb->pfmemalloc = true;
2495}
2496
131ea667 2497/**
e227867f 2498 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2499 * @frag: the paged fragment
2500 *
2501 * Returns the &struct page associated with @frag.
2502 */
2503static inline struct page *skb_frag_page(const skb_frag_t *frag)
2504{
a8605c60 2505 return frag->page.p;
131ea667
IC
2506}
2507
2508/**
2509 * __skb_frag_ref - take an addition reference on a paged fragment.
2510 * @frag: the paged fragment
2511 *
2512 * Takes an additional reference on the paged fragment @frag.
2513 */
2514static inline void __skb_frag_ref(skb_frag_t *frag)
2515{
2516 get_page(skb_frag_page(frag));
2517}
2518
2519/**
2520 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2521 * @skb: the buffer
2522 * @f: the fragment offset.
2523 *
2524 * Takes an additional reference on the @f'th paged fragment of @skb.
2525 */
2526static inline void skb_frag_ref(struct sk_buff *skb, int f)
2527{
2528 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2529}
2530
2531/**
2532 * __skb_frag_unref - release a reference on a paged fragment.
2533 * @frag: the paged fragment
2534 *
2535 * Releases a reference on the paged fragment @frag.
2536 */
2537static inline void __skb_frag_unref(skb_frag_t *frag)
2538{
2539 put_page(skb_frag_page(frag));
2540}
2541
2542/**
2543 * skb_frag_unref - release a reference on a paged fragment of an skb.
2544 * @skb: the buffer
2545 * @f: the fragment offset
2546 *
2547 * Releases a reference on the @f'th paged fragment of @skb.
2548 */
2549static inline void skb_frag_unref(struct sk_buff *skb, int f)
2550{
2551 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2552}
2553
2554/**
2555 * skb_frag_address - gets the address of the data contained in a paged fragment
2556 * @frag: the paged fragment buffer
2557 *
2558 * Returns the address of the data within @frag. The page must already
2559 * be mapped.
2560 */
2561static inline void *skb_frag_address(const skb_frag_t *frag)
2562{
2563 return page_address(skb_frag_page(frag)) + frag->page_offset;
2564}
2565
2566/**
2567 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2568 * @frag: the paged fragment buffer
2569 *
2570 * Returns the address of the data within @frag. Checks that the page
2571 * is mapped and returns %NULL otherwise.
2572 */
2573static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2574{
2575 void *ptr = page_address(skb_frag_page(frag));
2576 if (unlikely(!ptr))
2577 return NULL;
2578
2579 return ptr + frag->page_offset;
2580}
2581
2582/**
2583 * __skb_frag_set_page - sets the page contained in a paged fragment
2584 * @frag: the paged fragment
2585 * @page: the page to set
2586 *
2587 * Sets the fragment @frag to contain @page.
2588 */
2589static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2590{
a8605c60 2591 frag->page.p = page;
131ea667
IC
2592}
2593
2594/**
2595 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2596 * @skb: the buffer
2597 * @f: the fragment offset
2598 * @page: the page to set
2599 *
2600 * Sets the @f'th fragment of @skb to contain @page.
2601 */
2602static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2603 struct page *page)
2604{
2605 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2606}
2607
400dfd3a
ED
2608bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2609
131ea667
IC
2610/**
2611 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2612 * @dev: the device to map the fragment to
131ea667
IC
2613 * @frag: the paged fragment to map
2614 * @offset: the offset within the fragment (starting at the
2615 * fragment's own offset)
2616 * @size: the number of bytes to map
f83347df 2617 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2618 *
2619 * Maps the page associated with @frag to @device.
2620 */
2621static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2622 const skb_frag_t *frag,
2623 size_t offset, size_t size,
2624 enum dma_data_direction dir)
2625{
2626 return dma_map_page(dev, skb_frag_page(frag),
2627 frag->page_offset + offset, size, dir);
2628}
2629
117632e6
ED
2630static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2631 gfp_t gfp_mask)
2632{
2633 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2634}
2635
bad93e9d
OP
2636
2637static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2638 gfp_t gfp_mask)
2639{
2640 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2641}
2642
2643
334a8132
PM
2644/**
2645 * skb_clone_writable - is the header of a clone writable
2646 * @skb: buffer to check
2647 * @len: length up to which to write
2648 *
2649 * Returns true if modifying the header part of the cloned buffer
2650 * does not requires the data to be copied.
2651 */
05bdd2f1 2652static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2653{
2654 return !skb_header_cloned(skb) &&
2655 skb_headroom(skb) + len <= skb->hdr_len;
2656}
2657
3697649f
DB
2658static inline int skb_try_make_writable(struct sk_buff *skb,
2659 unsigned int write_len)
2660{
2661 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2662 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2663}
2664
d9cc2048
HX
2665static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2666 int cloned)
2667{
2668 int delta = 0;
2669
d9cc2048
HX
2670 if (headroom > skb_headroom(skb))
2671 delta = headroom - skb_headroom(skb);
2672
2673 if (delta || cloned)
2674 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2675 GFP_ATOMIC);
2676 return 0;
2677}
2678
1da177e4
LT
2679/**
2680 * skb_cow - copy header of skb when it is required
2681 * @skb: buffer to cow
2682 * @headroom: needed headroom
2683 *
2684 * If the skb passed lacks sufficient headroom or its data part
2685 * is shared, data is reallocated. If reallocation fails, an error
2686 * is returned and original skb is not changed.
2687 *
2688 * The result is skb with writable area skb->head...skb->tail
2689 * and at least @headroom of space at head.
2690 */
2691static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2692{
d9cc2048
HX
2693 return __skb_cow(skb, headroom, skb_cloned(skb));
2694}
1da177e4 2695
d9cc2048
HX
2696/**
2697 * skb_cow_head - skb_cow but only making the head writable
2698 * @skb: buffer to cow
2699 * @headroom: needed headroom
2700 *
2701 * This function is identical to skb_cow except that we replace the
2702 * skb_cloned check by skb_header_cloned. It should be used when
2703 * you only need to push on some header and do not need to modify
2704 * the data.
2705 */
2706static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2707{
2708 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2709}
2710
2711/**
2712 * skb_padto - pad an skbuff up to a minimal size
2713 * @skb: buffer to pad
2714 * @len: minimal length
2715 *
2716 * Pads up a buffer to ensure the trailing bytes exist and are
2717 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2718 * is untouched. Otherwise it is extended. Returns zero on
2719 * success. The skb is freed on error.
1da177e4 2720 */
5b057c6b 2721static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2722{
2723 unsigned int size = skb->len;
2724 if (likely(size >= len))
5b057c6b 2725 return 0;
987c402a 2726 return skb_pad(skb, len - size);
1da177e4
LT
2727}
2728
9c0c1124
AD
2729/**
2730 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2731 * @skb: buffer to pad
2732 * @len: minimal length
2733 *
2734 * Pads up a buffer to ensure the trailing bytes exist and are
2735 * blanked. If the buffer already contains sufficient data it
2736 * is untouched. Otherwise it is extended. Returns zero on
2737 * success. The skb is freed on error.
2738 */
2739static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2740{
2741 unsigned int size = skb->len;
2742
2743 if (unlikely(size < len)) {
2744 len -= size;
2745 if (skb_pad(skb, len))
2746 return -ENOMEM;
2747 __skb_put(skb, len);
2748 }
2749 return 0;
2750}
2751
1da177e4 2752static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2753 struct iov_iter *from, int copy)
1da177e4
LT
2754{
2755 const int off = skb->len;
2756
2757 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2758 __wsum csum = 0;
2759 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2760 &csum, from) == copy) {
1da177e4
LT
2761 skb->csum = csum_block_add(skb->csum, csum, off);
2762 return 0;
2763 }
af2b040e 2764 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2765 return 0;
2766
2767 __skb_trim(skb, off);
2768 return -EFAULT;
2769}
2770
38ba0a65
ED
2771static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2772 const struct page *page, int off)
1da177e4
LT
2773{
2774 if (i) {
9e903e08 2775 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2776
ea2ab693 2777 return page == skb_frag_page(frag) &&
9e903e08 2778 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2779 }
38ba0a65 2780 return false;
1da177e4
LT
2781}
2782
364c6bad
HX
2783static inline int __skb_linearize(struct sk_buff *skb)
2784{
2785 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2786}
2787
1da177e4
LT
2788/**
2789 * skb_linearize - convert paged skb to linear one
2790 * @skb: buffer to linarize
1da177e4
LT
2791 *
2792 * If there is no free memory -ENOMEM is returned, otherwise zero
2793 * is returned and the old skb data released.
2794 */
364c6bad
HX
2795static inline int skb_linearize(struct sk_buff *skb)
2796{
2797 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2798}
2799
cef401de
ED
2800/**
2801 * skb_has_shared_frag - can any frag be overwritten
2802 * @skb: buffer to test
2803 *
2804 * Return true if the skb has at least one frag that might be modified
2805 * by an external entity (as in vmsplice()/sendfile())
2806 */
2807static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2808{
c9af6db4
PS
2809 return skb_is_nonlinear(skb) &&
2810 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2811}
2812
364c6bad
HX
2813/**
2814 * skb_linearize_cow - make sure skb is linear and writable
2815 * @skb: buffer to process
2816 *
2817 * If there is no free memory -ENOMEM is returned, otherwise zero
2818 * is returned and the old skb data released.
2819 */
2820static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2821{
364c6bad
HX
2822 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2823 __skb_linearize(skb) : 0;
1da177e4
LT
2824}
2825
2826/**
2827 * skb_postpull_rcsum - update checksum for received skb after pull
2828 * @skb: buffer to update
2829 * @start: start of data before pull
2830 * @len: length of data pulled
2831 *
2832 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2833 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2834 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2835 */
2836
2837static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2838 const void *start, unsigned int len)
1da177e4 2839{
84fa7933 2840 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4 2841 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
6ae459bd 2842 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
31b33dfb 2843 skb_checksum_start_offset(skb) < 0)
6ae459bd 2844 skb->ip_summed = CHECKSUM_NONE;
1da177e4
LT
2845}
2846
cbb042f9
HX
2847unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2848
f8ffad69
DB
2849static inline void skb_postpush_rcsum(struct sk_buff *skb,
2850 const void *start, unsigned int len)
2851{
2852 /* For performing the reverse operation to skb_postpull_rcsum(),
2853 * we can instead of ...
2854 *
2855 * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0));
2856 *
2857 * ... just use this equivalent version here to save a few
2858 * instructions. Feeding csum of 0 in csum_partial() and later
2859 * on adding skb->csum is equivalent to feed skb->csum in the
2860 * first place.
2861 */
2862 if (skb->ip_summed == CHECKSUM_COMPLETE)
2863 skb->csum = csum_partial(start, len, skb->csum);
2864}
2865
7ce5a27f
DM
2866/**
2867 * pskb_trim_rcsum - trim received skb and update checksum
2868 * @skb: buffer to trim
2869 * @len: new length
2870 *
2871 * This is exactly the same as pskb_trim except that it ensures the
2872 * checksum of received packets are still valid after the operation.
2873 */
2874
2875static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2876{
2877 if (likely(len >= skb->len))
2878 return 0;
2879 if (skb->ip_summed == CHECKSUM_COMPLETE)
2880 skb->ip_summed = CHECKSUM_NONE;
2881 return __pskb_trim(skb, len);
2882}
2883
1da177e4
LT
2884#define skb_queue_walk(queue, skb) \
2885 for (skb = (queue)->next; \
a1e4891f 2886 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2887 skb = skb->next)
2888
46f8914e
JC
2889#define skb_queue_walk_safe(queue, skb, tmp) \
2890 for (skb = (queue)->next, tmp = skb->next; \
2891 skb != (struct sk_buff *)(queue); \
2892 skb = tmp, tmp = skb->next)
2893
1164f52a 2894#define skb_queue_walk_from(queue, skb) \
a1e4891f 2895 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2896 skb = skb->next)
2897
2898#define skb_queue_walk_from_safe(queue, skb, tmp) \
2899 for (tmp = skb->next; \
2900 skb != (struct sk_buff *)(queue); \
2901 skb = tmp, tmp = skb->next)
2902
300ce174
SH
2903#define skb_queue_reverse_walk(queue, skb) \
2904 for (skb = (queue)->prev; \
a1e4891f 2905 skb != (struct sk_buff *)(queue); \
300ce174
SH
2906 skb = skb->prev)
2907
686a2955
DM
2908#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2909 for (skb = (queue)->prev, tmp = skb->prev; \
2910 skb != (struct sk_buff *)(queue); \
2911 skb = tmp, tmp = skb->prev)
2912
2913#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2914 for (tmp = skb->prev; \
2915 skb != (struct sk_buff *)(queue); \
2916 skb = tmp, tmp = skb->prev)
1da177e4 2917
21dc3301 2918static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2919{
2920 return skb_shinfo(skb)->frag_list != NULL;
2921}
2922
2923static inline void skb_frag_list_init(struct sk_buff *skb)
2924{
2925 skb_shinfo(skb)->frag_list = NULL;
2926}
2927
ee039871
DM
2928#define skb_walk_frags(skb, iter) \
2929 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2930
ea3793ee
RW
2931
2932int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
2933 const struct sk_buff *skb);
2934struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
2935 int *peeked, int *off, int *err,
2936 struct sk_buff **last);
7965bd4d
JP
2937struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2938 int *peeked, int *off, int *err);
2939struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2940 int *err);
2941unsigned int datagram_poll(struct file *file, struct socket *sock,
2942 struct poll_table_struct *wait);
c0371da6
AV
2943int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2944 struct iov_iter *to, int size);
51f3d02b
DM
2945static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2946 struct msghdr *msg, int size)
2947{
e5a4b0bb 2948 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2949}
e5a4b0bb
AV
2950int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2951 struct msghdr *msg);
3a654f97
AV
2952int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2953 struct iov_iter *from, int len);
3a654f97 2954int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 2955void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 2956void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
2957static inline void skb_free_datagram_locked(struct sock *sk,
2958 struct sk_buff *skb)
2959{
2960 __skb_free_datagram_locked(sk, skb, 0);
2961}
7965bd4d 2962int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2963int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2964int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2965__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2966 int len, __wsum csum);
a60e3cc7
HFS
2967ssize_t skb_socket_splice(struct sock *sk,
2968 struct pipe_inode_info *pipe,
2969 struct splice_pipe_desc *spd);
2970int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 2971 struct pipe_inode_info *pipe, unsigned int len,
a60e3cc7
HFS
2972 unsigned int flags,
2973 ssize_t (*splice_cb)(struct sock *,
2974 struct pipe_inode_info *,
2975 struct splice_pipe_desc *));
7965bd4d 2976void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2977unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2978int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2979 int len, int hlen);
7965bd4d
JP
2980void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2981int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2982void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2983unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2984struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2985struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2986int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2987int skb_vlan_pop(struct sk_buff *skb);
2988int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2989
6ce8e9ce
AV
2990static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2991{
21226abb 2992 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
2993}
2994
7eab8d9e
AV
2995static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2996{
e5a4b0bb 2997 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2998}
2999
2817a336
DB
3000struct skb_checksum_ops {
3001 __wsum (*update)(const void *mem, int len, __wsum wsum);
3002 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3003};
3004
3005__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3006 __wsum csum, const struct skb_checksum_ops *ops);
3007__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3008 __wsum csum);
3009
1e98a0f0
ED
3010static inline void * __must_check
3011__skb_header_pointer(const struct sk_buff *skb, int offset,
3012 int len, void *data, int hlen, void *buffer)
1da177e4 3013{
55820ee2 3014 if (hlen - offset >= len)
690e36e7 3015 return data + offset;
1da177e4 3016
690e36e7
DM
3017 if (!skb ||
3018 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3019 return NULL;
3020
3021 return buffer;
3022}
3023
1e98a0f0
ED
3024static inline void * __must_check
3025skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3026{
3027 return __skb_header_pointer(skb, offset, len, skb->data,
3028 skb_headlen(skb), buffer);
3029}
3030
4262e5cc
DB
3031/**
3032 * skb_needs_linearize - check if we need to linearize a given skb
3033 * depending on the given device features.
3034 * @skb: socket buffer to check
3035 * @features: net device features
3036 *
3037 * Returns true if either:
3038 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3039 * 2. skb is fragmented and the device does not support SG.
3040 */
3041static inline bool skb_needs_linearize(struct sk_buff *skb,
3042 netdev_features_t features)
3043{
3044 return skb_is_nonlinear(skb) &&
3045 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3046 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3047}
3048
d626f62b
ACM
3049static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3050 void *to,
3051 const unsigned int len)
3052{
3053 memcpy(to, skb->data, len);
3054}
3055
3056static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3057 const int offset, void *to,
3058 const unsigned int len)
3059{
3060 memcpy(to, skb->data + offset, len);
3061}
3062
27d7ff46
ACM
3063static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3064 const void *from,
3065 const unsigned int len)
3066{
3067 memcpy(skb->data, from, len);
3068}
3069
3070static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3071 const int offset,
3072 const void *from,
3073 const unsigned int len)
3074{
3075 memcpy(skb->data + offset, from, len);
3076}
3077
7965bd4d 3078void skb_init(void);
1da177e4 3079
ac45f602
PO
3080static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3081{
3082 return skb->tstamp;
3083}
3084
a61bbcf2
PM
3085/**
3086 * skb_get_timestamp - get timestamp from a skb
3087 * @skb: skb to get stamp from
3088 * @stamp: pointer to struct timeval to store stamp in
3089 *
3090 * Timestamps are stored in the skb as offsets to a base timestamp.
3091 * This function converts the offset back to a struct timeval and stores
3092 * it in stamp.
3093 */
ac45f602
PO
3094static inline void skb_get_timestamp(const struct sk_buff *skb,
3095 struct timeval *stamp)
a61bbcf2 3096{
b7aa0bf7 3097 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3098}
3099
ac45f602
PO
3100static inline void skb_get_timestampns(const struct sk_buff *skb,
3101 struct timespec *stamp)
3102{
3103 *stamp = ktime_to_timespec(skb->tstamp);
3104}
3105
b7aa0bf7 3106static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3107{
b7aa0bf7 3108 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3109}
3110
164891aa
SH
3111static inline ktime_t net_timedelta(ktime_t t)
3112{
3113 return ktime_sub(ktime_get_real(), t);
3114}
3115
b9ce204f
IJ
3116static inline ktime_t net_invalid_timestamp(void)
3117{
3118 return ktime_set(0, 0);
3119}
a61bbcf2 3120
62bccb8c
AD
3121struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3122
c1f19b51
RC
3123#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3124
7965bd4d
JP
3125void skb_clone_tx_timestamp(struct sk_buff *skb);
3126bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3127
3128#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3129
3130static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3131{
3132}
3133
3134static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3135{
3136 return false;
3137}
3138
3139#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3140
3141/**
3142 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3143 *
da92b194
RC
3144 * PHY drivers may accept clones of transmitted packets for
3145 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3146 * must call this function to return the skb back to the stack with a
3147 * timestamp.
da92b194 3148 *
c1f19b51 3149 * @skb: clone of the the original outgoing packet
7a76a021 3150 * @hwtstamps: hardware time stamps
c1f19b51
RC
3151 *
3152 */
3153void skb_complete_tx_timestamp(struct sk_buff *skb,
3154 struct skb_shared_hwtstamps *hwtstamps);
3155
e7fd2885
WB
3156void __skb_tstamp_tx(struct sk_buff *orig_skb,
3157 struct skb_shared_hwtstamps *hwtstamps,
3158 struct sock *sk, int tstype);
3159
ac45f602
PO
3160/**
3161 * skb_tstamp_tx - queue clone of skb with send time stamps
3162 * @orig_skb: the original outgoing packet
3163 * @hwtstamps: hardware time stamps, may be NULL if not available
3164 *
3165 * If the skb has a socket associated, then this function clones the
3166 * skb (thus sharing the actual data and optional structures), stores
3167 * the optional hardware time stamping information (if non NULL) or
3168 * generates a software time stamp (otherwise), then queues the clone
3169 * to the error queue of the socket. Errors are silently ignored.
3170 */
7965bd4d
JP
3171void skb_tstamp_tx(struct sk_buff *orig_skb,
3172 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3173
4507a715
RC
3174static inline void sw_tx_timestamp(struct sk_buff *skb)
3175{
2244d07b
OH
3176 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3177 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3178 skb_tstamp_tx(skb, NULL);
3179}
3180
3181/**
3182 * skb_tx_timestamp() - Driver hook for transmit timestamping
3183 *
3184 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3185 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3186 *
73409f3b
DM
3187 * Specifically, one should make absolutely sure that this function is
3188 * called before TX completion of this packet can trigger. Otherwise
3189 * the packet could potentially already be freed.
3190 *
4507a715
RC
3191 * @skb: A socket buffer.
3192 */
3193static inline void skb_tx_timestamp(struct sk_buff *skb)
3194{
c1f19b51 3195 skb_clone_tx_timestamp(skb);
4507a715
RC
3196 sw_tx_timestamp(skb);
3197}
3198
6e3e939f
JB
3199/**
3200 * skb_complete_wifi_ack - deliver skb with wifi status
3201 *
3202 * @skb: the original outgoing packet
3203 * @acked: ack status
3204 *
3205 */
3206void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3207
7965bd4d
JP
3208__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3209__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3210
60476372
HX
3211static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3212{
6edec0e6
TH
3213 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3214 skb->csum_valid ||
3215 (skb->ip_summed == CHECKSUM_PARTIAL &&
3216 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3217}
3218
fb286bb2
HX
3219/**
3220 * skb_checksum_complete - Calculate checksum of an entire packet
3221 * @skb: packet to process
3222 *
3223 * This function calculates the checksum over the entire packet plus
3224 * the value of skb->csum. The latter can be used to supply the
3225 * checksum of a pseudo header as used by TCP/UDP. It returns the
3226 * checksum.
3227 *
3228 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3229 * this function can be used to verify that checksum on received
3230 * packets. In that case the function should return zero if the
3231 * checksum is correct. In particular, this function will return zero
3232 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3233 * hardware has already verified the correctness of the checksum.
3234 */
4381ca3c 3235static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3236{
60476372
HX
3237 return skb_csum_unnecessary(skb) ?
3238 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3239}
3240
77cffe23
TH
3241static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3242{
3243 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3244 if (skb->csum_level == 0)
3245 skb->ip_summed = CHECKSUM_NONE;
3246 else
3247 skb->csum_level--;
3248 }
3249}
3250
3251static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3252{
3253 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3254 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3255 skb->csum_level++;
3256 } else if (skb->ip_summed == CHECKSUM_NONE) {
3257 skb->ip_summed = CHECKSUM_UNNECESSARY;
3258 skb->csum_level = 0;
3259 }
3260}
3261
5a212329
TH
3262static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3263{
3264 /* Mark current checksum as bad (typically called from GRO
3265 * path). In the case that ip_summed is CHECKSUM_NONE
3266 * this must be the first checksum encountered in the packet.
3267 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3268 * checksum after the last one validated. For UDP, a zero
3269 * checksum can not be marked as bad.
3270 */
3271
3272 if (skb->ip_summed == CHECKSUM_NONE ||
3273 skb->ip_summed == CHECKSUM_UNNECESSARY)
3274 skb->csum_bad = 1;
3275}
3276
76ba0aae
TH
3277/* Check if we need to perform checksum complete validation.
3278 *
3279 * Returns true if checksum complete is needed, false otherwise
3280 * (either checksum is unnecessary or zero checksum is allowed).
3281 */
3282static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3283 bool zero_okay,
3284 __sum16 check)
3285{
5d0c2b95
TH
3286 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3287 skb->csum_valid = 1;
77cffe23 3288 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3289 return false;
3290 }
3291
3292 return true;
3293}
3294
3295/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3296 * in checksum_init.
3297 */
3298#define CHECKSUM_BREAK 76
3299
4e18b9ad
TH
3300/* Unset checksum-complete
3301 *
3302 * Unset checksum complete can be done when packet is being modified
3303 * (uncompressed for instance) and checksum-complete value is
3304 * invalidated.
3305 */
3306static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3307{
3308 if (skb->ip_summed == CHECKSUM_COMPLETE)
3309 skb->ip_summed = CHECKSUM_NONE;
3310}
3311
76ba0aae
TH
3312/* Validate (init) checksum based on checksum complete.
3313 *
3314 * Return values:
3315 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3316 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3317 * checksum is stored in skb->csum for use in __skb_checksum_complete
3318 * non-zero: value of invalid checksum
3319 *
3320 */
3321static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3322 bool complete,
3323 __wsum psum)
3324{
3325 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3326 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3327 skb->csum_valid = 1;
76ba0aae
TH
3328 return 0;
3329 }
5a212329
TH
3330 } else if (skb->csum_bad) {
3331 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3332 return (__force __sum16)1;
76ba0aae
TH
3333 }
3334
3335 skb->csum = psum;
3336
5d0c2b95
TH
3337 if (complete || skb->len <= CHECKSUM_BREAK) {
3338 __sum16 csum;
3339
3340 csum = __skb_checksum_complete(skb);
3341 skb->csum_valid = !csum;
3342 return csum;
3343 }
76ba0aae
TH
3344
3345 return 0;
3346}
3347
3348static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3349{
3350 return 0;
3351}
3352
3353/* Perform checksum validate (init). Note that this is a macro since we only
3354 * want to calculate the pseudo header which is an input function if necessary.
3355 * First we try to validate without any computation (checksum unnecessary) and
3356 * then calculate based on checksum complete calling the function to compute
3357 * pseudo header.
3358 *
3359 * Return values:
3360 * 0: checksum is validated or try to in skb_checksum_complete
3361 * non-zero: value of invalid checksum
3362 */
3363#define __skb_checksum_validate(skb, proto, complete, \
3364 zero_okay, check, compute_pseudo) \
3365({ \
3366 __sum16 __ret = 0; \
5d0c2b95 3367 skb->csum_valid = 0; \
76ba0aae
TH
3368 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3369 __ret = __skb_checksum_validate_complete(skb, \
3370 complete, compute_pseudo(skb, proto)); \
3371 __ret; \
3372})
3373
3374#define skb_checksum_init(skb, proto, compute_pseudo) \
3375 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3376
3377#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3378 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3379
3380#define skb_checksum_validate(skb, proto, compute_pseudo) \
3381 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3382
3383#define skb_checksum_validate_zero_check(skb, proto, check, \
3384 compute_pseudo) \
096a4cfa 3385 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3386
3387#define skb_checksum_simple_validate(skb) \
3388 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3389
d96535a1
TH
3390static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3391{
3392 return (skb->ip_summed == CHECKSUM_NONE &&
3393 skb->csum_valid && !skb->csum_bad);
3394}
3395
3396static inline void __skb_checksum_convert(struct sk_buff *skb,
3397 __sum16 check, __wsum pseudo)
3398{
3399 skb->csum = ~pseudo;
3400 skb->ip_summed = CHECKSUM_COMPLETE;
3401}
3402
3403#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3404do { \
3405 if (__skb_checksum_convert_check(skb)) \
3406 __skb_checksum_convert(skb, check, \
3407 compute_pseudo(skb, proto)); \
3408} while (0)
3409
15e2396d
TH
3410static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3411 u16 start, u16 offset)
3412{
3413 skb->ip_summed = CHECKSUM_PARTIAL;
3414 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3415 skb->csum_offset = offset - start;
3416}
3417
dcdc8994
TH
3418/* Update skbuf and packet to reflect the remote checksum offload operation.
3419 * When called, ptr indicates the starting point for skb->csum when
3420 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3421 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3422 */
3423static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3424 int start, int offset, bool nopartial)
dcdc8994
TH
3425{
3426 __wsum delta;
3427
15e2396d
TH
3428 if (!nopartial) {
3429 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3430 return;
3431 }
3432
dcdc8994
TH
3433 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3434 __skb_checksum_complete(skb);
3435 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3436 }
3437
3438 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3439
3440 /* Adjust skb->csum since we changed the packet */
3441 skb->csum = csum_add(skb->csum, delta);
3442}
3443
5f79e0f9 3444#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3445void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3446static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3447{
3448 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3449 nf_conntrack_destroy(nfct);
1da177e4
LT
3450}
3451static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3452{
3453 if (nfct)
3454 atomic_inc(&nfct->use);
3455}
2fc72c7b 3456#endif
34666d46 3457#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3458static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3459{
3460 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3461 kfree(nf_bridge);
3462}
3463static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3464{
3465 if (nf_bridge)
3466 atomic_inc(&nf_bridge->use);
3467}
3468#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3469static inline void nf_reset(struct sk_buff *skb)
3470{
5f79e0f9 3471#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3472 nf_conntrack_put(skb->nfct);
3473 skb->nfct = NULL;
2fc72c7b 3474#endif
34666d46 3475#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3476 nf_bridge_put(skb->nf_bridge);
3477 skb->nf_bridge = NULL;
3478#endif
3479}
3480
124dff01
PM
3481static inline void nf_reset_trace(struct sk_buff *skb)
3482{
478b360a 3483#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3484 skb->nf_trace = 0;
3485#endif
a193a4ab
PM
3486}
3487
edda553c 3488/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3489static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3490 bool copy)
edda553c 3491{
5f79e0f9 3492#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3493 dst->nfct = src->nfct;
3494 nf_conntrack_get(src->nfct);
b1937227
ED
3495 if (copy)
3496 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3497#endif
34666d46 3498#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3499 dst->nf_bridge = src->nf_bridge;
3500 nf_bridge_get(src->nf_bridge);
3501#endif
478b360a 3502#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3503 if (copy)
3504 dst->nf_trace = src->nf_trace;
478b360a 3505#endif
edda553c
YK
3506}
3507
e7ac05f3
YK
3508static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3509{
e7ac05f3 3510#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3511 nf_conntrack_put(dst->nfct);
2fc72c7b 3512#endif
34666d46 3513#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3514 nf_bridge_put(dst->nf_bridge);
3515#endif
b1937227 3516 __nf_copy(dst, src, true);
e7ac05f3
YK
3517}
3518
984bc16c
JM
3519#ifdef CONFIG_NETWORK_SECMARK
3520static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3521{
3522 to->secmark = from->secmark;
3523}
3524
3525static inline void skb_init_secmark(struct sk_buff *skb)
3526{
3527 skb->secmark = 0;
3528}
3529#else
3530static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3531{ }
3532
3533static inline void skb_init_secmark(struct sk_buff *skb)
3534{ }
3535#endif
3536
574f7194
EB
3537static inline bool skb_irq_freeable(const struct sk_buff *skb)
3538{
3539 return !skb->destructor &&
3540#if IS_ENABLED(CONFIG_XFRM)
3541 !skb->sp &&
3542#endif
3543#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3544 !skb->nfct &&
3545#endif
3546 !skb->_skb_refdst &&
3547 !skb_has_frag_list(skb);
3548}
3549
f25f4e44
PWJ
3550static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3551{
f25f4e44 3552 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3553}
3554
9247744e 3555static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3556{
4e3ab47a 3557 return skb->queue_mapping;
4e3ab47a
PE
3558}
3559
f25f4e44
PWJ
3560static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3561{
f25f4e44 3562 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3563}
3564
d5a9e24a
DM
3565static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3566{
3567 skb->queue_mapping = rx_queue + 1;
3568}
3569
9247744e 3570static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3571{
3572 return skb->queue_mapping - 1;
3573}
3574
9247744e 3575static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3576{
a02cec21 3577 return skb->queue_mapping != 0;
d5a9e24a
DM
3578}
3579
def8b4fa
AD
3580static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3581{
0b3d8e08 3582#ifdef CONFIG_XFRM
def8b4fa 3583 return skb->sp;
def8b4fa 3584#else
def8b4fa 3585 return NULL;
def8b4fa 3586#endif
0b3d8e08 3587}
def8b4fa 3588
68c33163
PS
3589/* Keeps track of mac header offset relative to skb->head.
3590 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3591 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3592 * tunnel skb it points to outer mac header.
3593 * Keeps track of level of encapsulation of network headers.
3594 */
68c33163 3595struct skb_gso_cb {
802ab55a
AD
3596 union {
3597 int mac_offset;
3598 int data_offset;
3599 };
3347c960 3600 int encap_level;
76443456 3601 __wsum csum;
7e2b10c1 3602 __u16 csum_start;
68c33163 3603};
9207f9d4
KK
3604#define SKB_SGO_CB_OFFSET 32
3605#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3606
3607static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3608{
3609 return (skb_mac_header(inner_skb) - inner_skb->head) -
3610 SKB_GSO_CB(inner_skb)->mac_offset;
3611}
3612
1e2bd517
PS
3613static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3614{
3615 int new_headroom, headroom;
3616 int ret;
3617
3618 headroom = skb_headroom(skb);
3619 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3620 if (ret)
3621 return ret;
3622
3623 new_headroom = skb_headroom(skb);
3624 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3625 return 0;
3626}
3627
08b64fcc
AD
3628static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3629{
3630 /* Do not update partial checksums if remote checksum is enabled. */
3631 if (skb->remcsum_offload)
3632 return;
3633
3634 SKB_GSO_CB(skb)->csum = res;
3635 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3636}
3637
7e2b10c1
TH
3638/* Compute the checksum for a gso segment. First compute the checksum value
3639 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3640 * then add in skb->csum (checksum from csum_start to end of packet).
3641 * skb->csum and csum_start are then updated to reflect the checksum of the
3642 * resultant packet starting from the transport header-- the resultant checksum
3643 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3644 * header.
3645 */
3646static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3647{
76443456
AD
3648 unsigned char *csum_start = skb_transport_header(skb);
3649 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3650 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3651
76443456
AD
3652 SKB_GSO_CB(skb)->csum = res;
3653 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3654
76443456 3655 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3656}
3657
bdcc0924 3658static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3659{
3660 return skb_shinfo(skb)->gso_size;
3661}
3662
36a8f39e 3663/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3664static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3665{
3666 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3667}
3668
7965bd4d 3669void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3670
3671static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3672{
3673 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3674 * wanted then gso_type will be set. */
05bdd2f1
ED
3675 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3676
b78462eb
AD
3677 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3678 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3679 __skb_warn_lro_forwarding(skb);
3680 return true;
3681 }
3682 return false;
3683}
3684
35fc92a9
HX
3685static inline void skb_forward_csum(struct sk_buff *skb)
3686{
3687 /* Unfortunately we don't support this one. Any brave souls? */
3688 if (skb->ip_summed == CHECKSUM_COMPLETE)
3689 skb->ip_summed = CHECKSUM_NONE;
3690}
3691
bc8acf2c
ED
3692/**
3693 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3694 * @skb: skb to check
3695 *
3696 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3697 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3698 * use this helper, to document places where we make this assertion.
3699 */
05bdd2f1 3700static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3701{
3702#ifdef DEBUG
3703 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3704#endif
3705}
3706
f35d9d8a 3707bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3708
ed1f50c3 3709int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3710struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3711 unsigned int transport_len,
3712 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3713
3a7c1ee4
AD
3714/**
3715 * skb_head_is_locked - Determine if the skb->head is locked down
3716 * @skb: skb to check
3717 *
3718 * The head on skbs build around a head frag can be removed if they are
3719 * not cloned. This function returns true if the skb head is locked down
3720 * due to either being allocated via kmalloc, or by being a clone with
3721 * multiple references to the head.
3722 */
3723static inline bool skb_head_is_locked(const struct sk_buff *skb)
3724{
3725 return !skb->head_frag || skb_cloned(skb);
3726}
fe6cc55f
FW
3727
3728/**
3729 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3730 *
3731 * @skb: GSO skb
3732 *
3733 * skb_gso_network_seglen is used to determine the real size of the
3734 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3735 *
3736 * The MAC/L2 header is not accounted for.
3737 */
3738static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3739{
3740 unsigned int hdr_len = skb_transport_header(skb) -
3741 skb_network_header(skb);
3742 return hdr_len + skb_gso_transport_seglen(skb);
3743}
ee122c79 3744
179bc67f
EC
3745/* Local Checksum Offload.
3746 * Compute outer checksum based on the assumption that the
3747 * inner checksum will be offloaded later.
e8ae7b00
EC
3748 * See Documentation/networking/checksum-offloads.txt for
3749 * explanation of how this works.
179bc67f
EC
3750 * Fill in outer checksum adjustment (e.g. with sum of outer
3751 * pseudo-header) before calling.
3752 * Also ensure that inner checksum is in linear data area.
3753 */
3754static inline __wsum lco_csum(struct sk_buff *skb)
3755{
9e74a6da
AD
3756 unsigned char *csum_start = skb_checksum_start(skb);
3757 unsigned char *l4_hdr = skb_transport_header(skb);
3758 __wsum partial;
179bc67f
EC
3759
3760 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3761 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3762 skb->csum_offset));
3763
179bc67f 3764 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3765 * adjustment filled in by caller) and return result.
179bc67f 3766 */
9e74a6da 3767 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3768}
3769
1da177e4
LT
3770#endif /* __KERNEL__ */
3771#endif /* _LINUX_SKBUFF_H */