Merge branch 'net-smc-next'
[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>
e6017571 37#include <linux/sched/clock.h>
1bd758eb 38#include <net/flow_dissector.h>
a60e3cc7 39#include <linux/splice.h>
72b31f72 40#include <linux/in6.h>
8b10cab6 41#include <linux/if_packet.h>
f70ea018 42#include <net/flow.h>
1da177e4 43
7a6ae71b
TH
44/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
78ea85f1
DB
89 *
90 * CHECKSUM_NONE:
91 *
7a6ae71b 92 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
77cffe23
TH
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
112 *
113 * skb->csum_level indicates the number of consecutive checksums found in
114 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
115 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
116 * and a device is able to verify the checksums for UDP (possibly zero),
117 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
118 * two. If the device were only able to verify the UDP checksum and not
119 * GRE, either because it doesn't support GRE checksum of because GRE
120 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
121 * not considered in this case).
78ea85f1
DB
122 *
123 * CHECKSUM_COMPLETE:
124 *
125 * This is the most generic way. The device supplied checksum of the _whole_
126 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
127 * hardware doesn't need to parse L3/L4 headers to implement this.
128 *
129 * Note: Even if device supports only some protocols, but is able to produce
130 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
131 *
132 * CHECKSUM_PARTIAL:
133 *
6edec0e6
TH
134 * A checksum is set up to be offloaded to a device as described in the
135 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 136 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
137 * on the same host, or it may be set in the input path in GRO or remote
138 * checksum offload. For the purposes of checksum verification, the checksum
139 * referred to by skb->csum_start + skb->csum_offset and any preceding
140 * checksums in the packet are considered verified. Any checksums in the
141 * packet that are after the checksum being offloaded are not considered to
142 * be verified.
78ea85f1 143 *
7a6ae71b
TH
144 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
145 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
146 *
147 * CHECKSUM_PARTIAL:
148 *
7a6ae71b 149 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 150 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
151 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
152 * csum_start and csum_offset values are valid values given the length and
153 * offset of the packet, however they should not attempt to validate that the
154 * checksum refers to a legitimate transport layer checksum-- it is the
155 * purview of the stack to validate that csum_start and csum_offset are set
156 * correctly.
157 *
158 * When the stack requests checksum offload for a packet, the driver MUST
159 * ensure that the checksum is set correctly. A driver can either offload the
160 * checksum calculation to the device, or call skb_checksum_help (in the case
161 * that the device does not support offload for a particular checksum).
162 *
163 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
164 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
165 * checksum offload capability. If a device has limited checksum capabilities
166 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
167 * described above) a helper function can be called to resolve
168 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
169 * function takes a spec argument that describes the protocol layer that is
170 * supported for checksum offload and can be called for each packet. If a
171 * packet does not match the specification for offload, skb_checksum_help
172 * is called to resolve the checksum.
78ea85f1 173 *
7a6ae71b 174 * CHECKSUM_NONE:
78ea85f1 175 *
7a6ae71b
TH
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
78ea85f1
DB
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
7a6ae71b
TH
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
78ea85f1 183 *
7a6ae71b
TH
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
192 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
193 * accordingly. Note the there is no indication in the skbuff that the
194 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
195 * both IP checksum offload and SCTP CRC offload must verify which offload
196 * is configured for a packet presumably by inspecting packet headers.
197 *
198 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
199 * offloading the FCOE CRC in a packet. To perform this offload the stack
200 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
201 * accordingly. Note the there is no indication in the skbuff that the
202 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
203 * both IP checksum offload and FCOE CRC offload must verify which offload
204 * is configured for a packet presumably by inspecting packet headers.
205 *
206 * E. Checksumming on output with GSO.
207 *
208 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
209 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
210 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
211 * part of the GSO operation is implied. If a checksum is being offloaded
212 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
213 * are set to refer to the outermost checksum being offload (two offloaded
214 * checksums are possible with UDP encapsulation).
78ea85f1
DB
215 */
216
60476372 217/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
218#define CHECKSUM_NONE 0
219#define CHECKSUM_UNNECESSARY 1
220#define CHECKSUM_COMPLETE 2
221#define CHECKSUM_PARTIAL 3
1da177e4 222
77cffe23
TH
223/* Maximum value in skb->csum_level */
224#define SKB_MAX_CSUM_LEVEL 3
225
0bec8c88 226#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 227#define SKB_WITH_OVERHEAD(X) \
deea84b0 228 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
229#define SKB_MAX_ORDER(X, ORDER) \
230 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
231#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
232#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
233
87fb4b7b
ED
234/* return minimum truesize of one skb containing X bytes of data */
235#define SKB_TRUESIZE(X) ((X) + \
236 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
237 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
238
1da177e4 239struct net_device;
716ea3a7 240struct scatterlist;
9c55e01c 241struct pipe_inode_info;
a8f820aa 242struct iov_iter;
fd11a83d 243struct napi_struct;
1da177e4 244
5f79e0f9 245#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
246struct nf_conntrack {
247 atomic_t use;
1da177e4 248};
5f79e0f9 249#endif
1da177e4 250
34666d46 251#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 252struct nf_bridge_info {
bf1ac5ca 253 atomic_t use;
3eaf4025
FW
254 enum {
255 BRNF_PROTO_UNCHANGED,
256 BRNF_PROTO_8021Q,
257 BRNF_PROTO_PPPOE
7fb48c5b 258 } orig_proto:8;
72b1e5e4
FW
259 u8 pkt_otherhost:1;
260 u8 in_prerouting:1;
261 u8 bridged_dnat:1;
411ffb4f 262 __u16 frag_max_size;
bf1ac5ca 263 struct net_device *physindev;
63cdbc06
FW
264
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
7fb48c5b 267 union {
72b1e5e4 268 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
269 __be32 ipv4_daddr;
270 struct in6_addr ipv6_daddr;
72b1e5e4
FW
271
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
275 */
276 char neigh_header[8];
72b31f72 277 };
1da177e4
LT
278};
279#endif
280
1da177e4
LT
281struct sk_buff_head {
282 /* These two members must be first. */
283 struct sk_buff *next;
284 struct sk_buff *prev;
285
286 __u32 qlen;
287 spinlock_t lock;
288};
289
290struct sk_buff;
291
9d4dde52
IC
292/* To allow 64K frame to be packed as single skb without frag_list we
293 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
294 * buffers which do not start on a page boundary.
295 *
296 * Since GRO uses frags we allocate at least 16 regardless of page
297 * size.
a715dea3 298 */
9d4dde52 299#if (65536/PAGE_SIZE + 1) < 16
eec00954 300#define MAX_SKB_FRAGS 16UL
a715dea3 301#else
9d4dde52 302#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 303#endif
5f74f82e 304extern int sysctl_max_skb_frags;
1da177e4 305
3953c46c
MRL
306/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
307 * segment using its current segmentation instead.
308 */
309#define GSO_BY_FRAGS 0xFFFF
310
1da177e4
LT
311typedef struct skb_frag_struct skb_frag_t;
312
313struct skb_frag_struct {
a8605c60
IC
314 struct {
315 struct page *p;
316 } page;
cb4dfe56 317#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
318 __u32 page_offset;
319 __u32 size;
cb4dfe56
ED
320#else
321 __u16 page_offset;
322 __u16 size;
323#endif
1da177e4
LT
324};
325
9e903e08
ED
326static inline unsigned int skb_frag_size(const skb_frag_t *frag)
327{
328 return frag->size;
329}
330
331static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
332{
333 frag->size = size;
334}
335
336static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
337{
338 frag->size += delta;
339}
340
341static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
342{
343 frag->size -= delta;
344}
345
ac45f602
PO
346#define HAVE_HW_TIME_STAMP
347
348/**
d3a21be8 349 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
350 * @hwtstamp: hardware time stamp transformed into duration
351 * since arbitrary point in time
ac45f602
PO
352 *
353 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 354 * skb->tstamp.
ac45f602
PO
355 *
356 * hwtstamps can only be compared against other hwtstamps from
357 * the same device.
358 *
359 * This structure is attached to packets as part of the
360 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
361 */
362struct skb_shared_hwtstamps {
363 ktime_t hwtstamp;
ac45f602
PO
364};
365
2244d07b
OH
366/* Definitions for tx_flags in struct skb_shared_info */
367enum {
368 /* generate hardware time stamp */
369 SKBTX_HW_TSTAMP = 1 << 0,
370
e7fd2885 371 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
372 SKBTX_SW_TSTAMP = 1 << 1,
373
374 /* device driver is going to provide hardware time stamp */
375 SKBTX_IN_PROGRESS = 1 << 2,
376
a6686f2f 377 /* device driver supports TX zero-copy buffers */
62b1a8ab 378 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
379
380 /* generate wifi status information (where possible) */
62b1a8ab 381 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
382
383 /* This indicates at least one fragment might be overwritten
384 * (as in vmsplice(), sendfile() ...)
385 * If we need to compute a TX checksum, we'll need to copy
386 * all frags to avoid possible bad checksum
387 */
388 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
389
390 /* generate software time stamp when entering packet scheduling */
391 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
392};
393
e1c8a607 394#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 395 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
396#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
397
a6686f2f
SM
398/*
399 * The callback notifies userspace to release buffers when skb DMA is done in
400 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
401 * The zerocopy_success argument is true if zero copy transmit occurred,
402 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
403 * The ctx field is used to track device context.
404 * The desc field is used to track userspace buffer index.
a6686f2f
SM
405 */
406struct ubuf_info {
e19d6763 407 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 408 void *ctx;
a6686f2f 409 unsigned long desc;
ac45f602
PO
410};
411
1da177e4
LT
412/* This data is invariant across clones and lives at
413 * the end of the header data, ie. at skb->end.
414 */
415struct skb_shared_info {
7f564528 416 unsigned short _unused;
9f42f126
IC
417 unsigned char nr_frags;
418 __u8 tx_flags;
7967168c
HX
419 unsigned short gso_size;
420 /* Warning: this field is not always filled in (UFO)! */
421 unsigned short gso_segs;
1da177e4 422 struct sk_buff *frag_list;
ac45f602 423 struct skb_shared_hwtstamps hwtstamps;
7f564528 424 unsigned int gso_type;
09c2d251 425 u32 tskey;
9f42f126 426 __be32 ip6_frag_id;
ec7d2f2c
ED
427
428 /*
429 * Warning : all fields before dataref are cleared in __alloc_skb()
430 */
431 atomic_t dataref;
432
69e3c75f
JB
433 /* Intermediate layers must ensure that destructor_arg
434 * remains valid until skb destructor */
435 void * destructor_arg;
a6686f2f 436
fed66381
ED
437 /* must be last field, see pskb_expand_head() */
438 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
439};
440
441/* We divide dataref into two halves. The higher 16 bits hold references
442 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
443 * the entire skb->data. A clone of a headerless skb holds the length of
444 * the header in skb->hdr_len.
1da177e4
LT
445 *
446 * All users must obey the rule that the skb->data reference count must be
447 * greater than or equal to the payload reference count.
448 *
449 * Holding a reference to the payload part means that the user does not
450 * care about modifications to the header part of skb->data.
451 */
452#define SKB_DATAREF_SHIFT 16
453#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
454
d179cd12
DM
455
456enum {
c8753d55
VS
457 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
458 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
459 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
460};
461
7967168c
HX
462enum {
463 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 464 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
465
466 /* This indicates the skb is from an untrusted source. */
467 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
468
469 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
470 SKB_GSO_TCP_ECN = 1 << 3,
471
cbc53e08 472 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 473
cbc53e08 474 SKB_GSO_TCPV6 = 1 << 5,
68c33163 475
cbc53e08 476 SKB_GSO_FCOE = 1 << 6,
73136267 477
cbc53e08 478 SKB_GSO_GRE = 1 << 7,
0d89d203 479
cbc53e08 480 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 481
7e13318d 482 SKB_GSO_IPXIP4 = 1 << 9,
61c1db7f 483
7e13318d 484 SKB_GSO_IPXIP6 = 1 << 10,
0f4f4ffa 485
cbc53e08 486 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 487
cbc53e08
AD
488 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
489
802ab55a
AD
490 SKB_GSO_PARTIAL = 1 << 13,
491
492 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
90017acc
MRL
493
494 SKB_GSO_SCTP = 1 << 15,
7967168c
HX
495};
496
2e07fa9c
ACM
497#if BITS_PER_LONG > 32
498#define NET_SKBUFF_DATA_USES_OFFSET 1
499#endif
500
501#ifdef NET_SKBUFF_DATA_USES_OFFSET
502typedef unsigned int sk_buff_data_t;
503#else
504typedef unsigned char *sk_buff_data_t;
505#endif
506
363ec392
ED
507/**
508 * struct skb_mstamp - multi resolution time stamps
509 * @stamp_us: timestamp in us resolution
510 * @stamp_jiffies: timestamp in jiffies
511 */
512struct skb_mstamp {
513 union {
514 u64 v64;
515 struct {
516 u32 stamp_us;
517 u32 stamp_jiffies;
518 };
519 };
520};
521
522/**
523 * skb_mstamp_get - get current timestamp
524 * @cl: place to store timestamps
525 */
526static inline void skb_mstamp_get(struct skb_mstamp *cl)
527{
528 u64 val = local_clock();
529
530 do_div(val, NSEC_PER_USEC);
531 cl->stamp_us = (u32)val;
532 cl->stamp_jiffies = (u32)jiffies;
533}
534
535/**
536 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
537 * @t1: pointer to newest sample
538 * @t0: pointer to oldest sample
539 */
540static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
541 const struct skb_mstamp *t0)
542{
543 s32 delta_us = t1->stamp_us - t0->stamp_us;
544 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
545
546 /* If delta_us is negative, this might be because interval is too big,
547 * or local_clock() drift is too big : fallback using jiffies.
548 */
549 if (delta_us <= 0 ||
550 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
551
552 delta_us = jiffies_to_usecs(delta_jiffies);
553
554 return delta_us;
555}
556
625a5e10
YC
557static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
558 const struct skb_mstamp *t0)
559{
560 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
561
562 if (!diff)
563 diff = t1->stamp_us - t0->stamp_us;
564 return diff > 0;
565}
363ec392 566
1da177e4
LT
567/**
568 * struct sk_buff - socket buffer
569 * @next: Next buffer in list
570 * @prev: Previous buffer in list
363ec392 571 * @tstamp: Time we arrived/left
56b17425 572 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 573 * @sk: Socket we are owned by
1da177e4 574 * @dev: Device we arrived on/are leaving by
d84e0bd7 575 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 576 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 577 * @sp: the security path, used for xfrm
1da177e4
LT
578 * @len: Length of actual data
579 * @data_len: Data length
580 * @mac_len: Length of link layer header
334a8132 581 * @hdr_len: writable header length of cloned skb
663ead3b
HX
582 * @csum: Checksum (must include start/offset pair)
583 * @csum_start: Offset from skb->head where checksumming should start
584 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 585 * @priority: Packet queueing priority
60ff7467 586 * @ignore_df: allow local fragmentation
1da177e4 587 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 588 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
589 * @nohdr: Payload reference only, must not modify header
590 * @pkt_type: Packet class
c83c2486 591 * @fclone: skbuff clone status
c83c2486 592 * @ipvs_property: skbuff is owned by ipvs
e7246e12 593 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 594 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
595 * @tc_redirected: packet was redirected by a tc action
596 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
597 * @peeked: this packet has been seen already, so stats have been
598 * done for it, don't do them again
ba9dda3a 599 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
600 * @protocol: Packet protocol from driver
601 * @destructor: Destruct function
a9e419dc 602 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 603 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 604 * @skb_iif: ifindex of device we arrived on
1da177e4 605 * @tc_index: Traffic control index
61b905da 606 * @hash: the packet hash
d84e0bd7 607 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 608 * @xmit_more: More SKBs are pending for this queue
553a5672 609 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 610 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 611 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 612 * ports.
a3b18ddb 613 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
614 * @wifi_acked_valid: wifi_acked was set
615 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 616 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
4ff06203 617 * @dst_pending_confirm: need to confirm neighbour
06021292 618 * @napi_id: id of the NAPI struct this skb came from
984bc16c 619 * @secmark: security marking
d84e0bd7 620 * @mark: Generic packet mark
86a9bad3 621 * @vlan_proto: vlan encapsulation protocol
6aa895b0 622 * @vlan_tci: vlan tag control information
0d89d203 623 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
624 * @inner_transport_header: Inner transport layer header (encapsulation)
625 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 626 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
627 * @transport_header: Transport layer header
628 * @network_header: Network layer header
629 * @mac_header: Link layer header
630 * @tail: Tail pointer
631 * @end: End pointer
632 * @head: Head of buffer
633 * @data: Data head pointer
634 * @truesize: Buffer size
635 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
636 */
637
638struct sk_buff {
363ec392 639 union {
56b17425
ED
640 struct {
641 /* These two members must be first. */
642 struct sk_buff *next;
643 struct sk_buff *prev;
644
645 union {
646 ktime_t tstamp;
647 struct skb_mstamp skb_mstamp;
648 };
649 };
650 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 651 };
da3f5cf1 652 struct sock *sk;
1da177e4 653
c84d9490
ED
654 union {
655 struct net_device *dev;
656 /* Some protocols might use this space to store information,
657 * while device pointer would be NULL.
658 * UDP receive path is one user.
659 */
660 unsigned long dev_scratch;
661 };
1da177e4
LT
662 /*
663 * This is the control buffer. It is free to use for every
664 * layer. Please put your private variables there. If you
665 * want to keep them across layers you have to do a skb_clone()
666 * first. This is owned by whoever has the skb queued ATM.
667 */
da3f5cf1 668 char cb[48] __aligned(8);
1da177e4 669
7fee226a 670 unsigned long _skb_refdst;
b1937227 671 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
672#ifdef CONFIG_XFRM
673 struct sec_path *sp;
b1937227
ED
674#endif
675#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 676 unsigned long _nfct;
b1937227 677#endif
85224844 678#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 679 struct nf_bridge_info *nf_bridge;
da3f5cf1 680#endif
1da177e4 681 unsigned int len,
334a8132
PM
682 data_len;
683 __u16 mac_len,
684 hdr_len;
b1937227
ED
685
686 /* Following fields are _not_ copied in __copy_skb_header()
687 * Note that queue_mapping is here mostly to fill a hole.
688 */
fe55f6d5 689 kmemcheck_bitfield_begin(flags1);
b1937227 690 __u16 queue_mapping;
36bbef52
DB
691
692/* if you move cloned around you also must adapt those constants */
693#ifdef __BIG_ENDIAN_BITFIELD
694#define CLONED_MASK (1 << 7)
695#else
696#define CLONED_MASK 1
697#endif
698#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
699
700 __u8 __cloned_offset[0];
b1937227 701 __u8 cloned:1,
6869c4d8 702 nohdr:1,
b84f4cc9 703 fclone:2,
a59322be 704 peeked:1,
b1937227 705 head_frag:1,
36bbef52
DB
706 xmit_more:1,
707 __unused:1; /* one bit hole */
fe55f6d5 708 kmemcheck_bitfield_end(flags1);
4031ae6e 709
b1937227
ED
710 /* fields enclosed in headers_start/headers_end are copied
711 * using a single memcpy() in __copy_skb_header()
712 */
ebcf34f3 713 /* private: */
b1937227 714 __u32 headers_start[0];
ebcf34f3 715 /* public: */
4031ae6e 716
233577a2
HFS
717/* if you move pkt_type around you also must adapt those constants */
718#ifdef __BIG_ENDIAN_BITFIELD
719#define PKT_TYPE_MAX (7 << 5)
720#else
721#define PKT_TYPE_MAX 7
1da177e4 722#endif
233577a2 723#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 724
233577a2 725 __u8 __pkt_type_offset[0];
b1937227 726 __u8 pkt_type:3;
c93bdd0e 727 __u8 pfmemalloc:1;
b1937227 728 __u8 ignore_df:1;
b1937227
ED
729
730 __u8 nf_trace:1;
731 __u8 ip_summed:2;
3853b584 732 __u8 ooo_okay:1;
61b905da 733 __u8 l4_hash:1;
a3b18ddb 734 __u8 sw_hash:1;
6e3e939f
JB
735 __u8 wifi_acked_valid:1;
736 __u8 wifi_acked:1;
b1937227 737
3bdc0eba 738 __u8 no_fcs:1;
77cffe23 739 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 740 __u8 encapsulation:1;
7e2b10c1 741 __u8 encap_hdr_csum:1;
5d0c2b95 742 __u8 csum_valid:1;
7e3cead5 743 __u8 csum_complete_sw:1;
b1937227
ED
744 __u8 csum_level:2;
745 __u8 csum_bad:1;
fe55f6d5 746
4ff06203 747 __u8 dst_pending_confirm:1;
b1937227
ED
748#ifdef CONFIG_IPV6_NDISC_NODETYPE
749 __u8 ndisc_nodetype:2;
750#endif
751 __u8 ipvs_property:1;
8bce6d7d 752 __u8 inner_protocol_type:1;
e585f236 753 __u8 remcsum_offload:1;
6bc506b4
IS
754#ifdef CONFIG_NET_SWITCHDEV
755 __u8 offload_fwd_mark:1;
756#endif
e7246e12
WB
757#ifdef CONFIG_NET_CLS_ACT
758 __u8 tc_skip_classify:1;
8dc07fdb 759 __u8 tc_at_ingress:1;
bc31c905
WB
760 __u8 tc_redirected:1;
761 __u8 tc_from_ingress:1;
e7246e12 762#endif
b1937227
ED
763
764#ifdef CONFIG_NET_SCHED
765 __u16 tc_index; /* traffic control index */
b1937227 766#endif
fe55f6d5 767
b1937227
ED
768 union {
769 __wsum csum;
770 struct {
771 __u16 csum_start;
772 __u16 csum_offset;
773 };
774 };
775 __u32 priority;
776 int skb_iif;
777 __u32 hash;
778 __be16 vlan_proto;
779 __u16 vlan_tci;
2bd82484
ED
780#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
781 union {
782 unsigned int napi_id;
783 unsigned int sender_cpu;
784 };
97fc2f08 785#endif
984bc16c 786#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 787 __u32 secmark;
0c4f691f 788#endif
0c4f691f 789
3b885787
NH
790 union {
791 __u32 mark;
16fad69c 792 __u32 reserved_tailroom;
3b885787 793 };
1da177e4 794
8bce6d7d
TH
795 union {
796 __be16 inner_protocol;
797 __u8 inner_ipproto;
798 };
799
1a37e412
SH
800 __u16 inner_transport_header;
801 __u16 inner_network_header;
802 __u16 inner_mac_header;
b1937227
ED
803
804 __be16 protocol;
1a37e412
SH
805 __u16 transport_header;
806 __u16 network_header;
807 __u16 mac_header;
b1937227 808
ebcf34f3 809 /* private: */
b1937227 810 __u32 headers_end[0];
ebcf34f3 811 /* public: */
b1937227 812
1da177e4 813 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 814 sk_buff_data_t tail;
4305b541 815 sk_buff_data_t end;
1da177e4 816 unsigned char *head,
4305b541 817 *data;
27a884dc
ACM
818 unsigned int truesize;
819 atomic_t users;
1da177e4
LT
820};
821
822#ifdef __KERNEL__
823/*
824 * Handling routines are only of interest to the kernel
825 */
826#include <linux/slab.h>
827
1da177e4 828
c93bdd0e
MG
829#define SKB_ALLOC_FCLONE 0x01
830#define SKB_ALLOC_RX 0x02
fd11a83d 831#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
832
833/* Returns true if the skb was allocated from PFMEMALLOC reserves */
834static inline bool skb_pfmemalloc(const struct sk_buff *skb)
835{
836 return unlikely(skb->pfmemalloc);
837}
838
7fee226a
ED
839/*
840 * skb might have a dst pointer attached, refcounted or not.
841 * _skb_refdst low order bit is set if refcount was _not_ taken
842 */
843#define SKB_DST_NOREF 1UL
844#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
845
a9e419dc 846#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
847/**
848 * skb_dst - returns skb dst_entry
849 * @skb: buffer
850 *
851 * Returns skb dst_entry, regardless of reference taken or not.
852 */
adf30907
ED
853static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
854{
7fee226a
ED
855 /* If refdst was not refcounted, check we still are in a
856 * rcu_read_lock section
857 */
858 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
859 !rcu_read_lock_held() &&
860 !rcu_read_lock_bh_held());
861 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
862}
863
7fee226a
ED
864/**
865 * skb_dst_set - sets skb dst
866 * @skb: buffer
867 * @dst: dst entry
868 *
869 * Sets skb dst, assuming a reference was taken on dst and should
870 * be released by skb_dst_drop()
871 */
adf30907
ED
872static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
873{
7fee226a
ED
874 skb->_skb_refdst = (unsigned long)dst;
875}
876
932bc4d7
JA
877/**
878 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
879 * @skb: buffer
880 * @dst: dst entry
881 *
882 * Sets skb dst, assuming a reference was not taken on dst.
883 * If dst entry is cached, we do not take reference and dst_release
884 * will be avoided by refdst_drop. If dst entry is not cached, we take
885 * reference, so that last dst_release can destroy the dst immediately.
886 */
887static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
888{
dbfc4fb7
HFS
889 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
890 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 891}
7fee226a
ED
892
893/**
25985edc 894 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
895 * @skb: buffer
896 */
897static inline bool skb_dst_is_noref(const struct sk_buff *skb)
898{
899 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
900}
901
511c3f92
ED
902static inline struct rtable *skb_rtable(const struct sk_buff *skb)
903{
adf30907 904 return (struct rtable *)skb_dst(skb);
511c3f92
ED
905}
906
8b10cab6
JHS
907/* For mangling skb->pkt_type from user space side from applications
908 * such as nft, tc, etc, we only allow a conservative subset of
909 * possible pkt_types to be set.
910*/
911static inline bool skb_pkt_type_ok(u32 ptype)
912{
913 return ptype <= PACKET_OTHERHOST;
914}
915
7965bd4d
JP
916void kfree_skb(struct sk_buff *skb);
917void kfree_skb_list(struct sk_buff *segs);
918void skb_tx_error(struct sk_buff *skb);
919void consume_skb(struct sk_buff *skb);
920void __kfree_skb(struct sk_buff *skb);
d7e8883c 921extern struct kmem_cache *skbuff_head_cache;
bad43ca8 922
7965bd4d
JP
923void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
924bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
925 bool *fragstolen, int *delta_truesize);
bad43ca8 926
7965bd4d
JP
927struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
928 int node);
2ea2f62c 929struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 930struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 931static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 932 gfp_t priority)
d179cd12 933{
564824b0 934 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
935}
936
2e4e4410
ED
937struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
938 unsigned long data_len,
939 int max_page_order,
940 int *errcode,
941 gfp_t gfp_mask);
942
d0bf4a9e
ED
943/* Layout of fast clones : [skb1][skb2][fclone_ref] */
944struct sk_buff_fclones {
945 struct sk_buff skb1;
946
947 struct sk_buff skb2;
948
949 atomic_t fclone_ref;
950};
951
952/**
953 * skb_fclone_busy - check if fclone is busy
293de7de 954 * @sk: socket
d0bf4a9e
ED
955 * @skb: buffer
956 *
bda13fed 957 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
958 * Some drivers call skb_orphan() in their ndo_start_xmit(),
959 * so we also check that this didnt happen.
d0bf4a9e 960 */
39bb5e62
ED
961static inline bool skb_fclone_busy(const struct sock *sk,
962 const struct sk_buff *skb)
d0bf4a9e
ED
963{
964 const struct sk_buff_fclones *fclones;
965
966 fclones = container_of(skb, struct sk_buff_fclones, skb1);
967
968 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 969 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 970 fclones->skb2.sk == sk;
d0bf4a9e
ED
971}
972
d179cd12 973static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 974 gfp_t priority)
d179cd12 975{
c93bdd0e 976 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
977}
978
7965bd4d 979struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
980static inline struct sk_buff *alloc_skb_head(gfp_t priority)
981{
982 return __alloc_skb_head(priority, -1);
983}
984
7965bd4d
JP
985struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
986int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
987struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
988struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
989struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
990 gfp_t gfp_mask, bool fclone);
991static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
992 gfp_t gfp_mask)
993{
994 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
995}
7965bd4d
JP
996
997int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
998struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
999 unsigned int headroom);
1000struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1001 int newtailroom, gfp_t priority);
25a91d8d
FD
1002int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1003 int offset, int len);
7965bd4d
JP
1004int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
1005 int len);
1006int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1007int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 1008#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1009
7965bd4d
JP
1010int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1011 int getfrag(void *from, char *to, int offset,
1012 int len, int odd, struct sk_buff *skb),
1013 void *from, int length);
e89e9cf5 1014
be12a1fe
HFS
1015int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1016 int offset, size_t size);
1017
d94d9fee 1018struct skb_seq_state {
677e90ed
TG
1019 __u32 lower_offset;
1020 __u32 upper_offset;
1021 __u32 frag_idx;
1022 __u32 stepped_offset;
1023 struct sk_buff *root_skb;
1024 struct sk_buff *cur_skb;
1025 __u8 *frag_data;
1026};
1027
7965bd4d
JP
1028void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1029 unsigned int to, struct skb_seq_state *st);
1030unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1031 struct skb_seq_state *st);
1032void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1033
7965bd4d 1034unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1035 unsigned int to, struct ts_config *config);
3fc7e8a6 1036
09323cc4
TH
1037/*
1038 * Packet hash types specify the type of hash in skb_set_hash.
1039 *
1040 * Hash types refer to the protocol layer addresses which are used to
1041 * construct a packet's hash. The hashes are used to differentiate or identify
1042 * flows of the protocol layer for the hash type. Hash types are either
1043 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1044 *
1045 * Properties of hashes:
1046 *
1047 * 1) Two packets in different flows have different hash values
1048 * 2) Two packets in the same flow should have the same hash value
1049 *
1050 * A hash at a higher layer is considered to be more specific. A driver should
1051 * set the most specific hash possible.
1052 *
1053 * A driver cannot indicate a more specific hash than the layer at which a hash
1054 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1055 *
1056 * A driver may indicate a hash level which is less specific than the
1057 * actual layer the hash was computed on. For instance, a hash computed
1058 * at L4 may be considered an L3 hash. This should only be done if the
1059 * driver can't unambiguously determine that the HW computed the hash at
1060 * the higher layer. Note that the "should" in the second property above
1061 * permits this.
1062 */
1063enum pkt_hash_types {
1064 PKT_HASH_TYPE_NONE, /* Undefined type */
1065 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1066 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1067 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1068};
1069
bcc83839 1070static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1071{
bcc83839 1072 skb->hash = 0;
a3b18ddb 1073 skb->sw_hash = 0;
bcc83839
TH
1074 skb->l4_hash = 0;
1075}
1076
1077static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1078{
1079 if (!skb->l4_hash)
1080 skb_clear_hash(skb);
1081}
1082
1083static inline void
1084__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1085{
1086 skb->l4_hash = is_l4;
1087 skb->sw_hash = is_sw;
61b905da 1088 skb->hash = hash;
09323cc4
TH
1089}
1090
bcc83839
TH
1091static inline void
1092skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1093{
1094 /* Used by drivers to set hash from HW */
1095 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1096}
1097
1098static inline void
1099__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1100{
1101 __skb_set_hash(skb, hash, true, is_l4);
1102}
1103
e5276937 1104void __skb_get_hash(struct sk_buff *skb);
b917783c 1105u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1106u32 skb_get_poff(const struct sk_buff *skb);
1107u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1108 const struct flow_keys *keys, int hlen);
1109__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1110 void *data, int hlen_proto);
1111
1112static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1113 int thoff, u8 ip_proto)
1114{
1115 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1116}
1117
1118void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1119 const struct flow_dissector_key *key,
1120 unsigned int key_count);
1121
1122bool __skb_flow_dissect(const struct sk_buff *skb,
1123 struct flow_dissector *flow_dissector,
1124 void *target_container,
cd79a238
TH
1125 void *data, __be16 proto, int nhoff, int hlen,
1126 unsigned int flags);
e5276937
TH
1127
1128static inline bool skb_flow_dissect(const struct sk_buff *skb,
1129 struct flow_dissector *flow_dissector,
cd79a238 1130 void *target_container, unsigned int flags)
e5276937
TH
1131{
1132 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1133 NULL, 0, 0, 0, flags);
e5276937
TH
1134}
1135
1136static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1137 struct flow_keys *flow,
1138 unsigned int flags)
e5276937
TH
1139{
1140 memset(flow, 0, sizeof(*flow));
1141 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1142 NULL, 0, 0, 0, flags);
e5276937
TH
1143}
1144
1145static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1146 void *data, __be16 proto,
cd79a238
TH
1147 int nhoff, int hlen,
1148 unsigned int flags)
e5276937
TH
1149{
1150 memset(flow, 0, sizeof(*flow));
1151 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1152 data, proto, nhoff, hlen, flags);
e5276937
TH
1153}
1154
3958afa1 1155static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1156{
a3b18ddb 1157 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1158 __skb_get_hash(skb);
bfb564e7 1159
61b905da 1160 return skb->hash;
bfb564e7
KK
1161}
1162
20a17bf6 1163__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1164
20a17bf6 1165static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1166{
c6cc1ca7
TH
1167 if (!skb->l4_hash && !skb->sw_hash) {
1168 struct flow_keys keys;
de4c1f8b 1169 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1170
de4c1f8b 1171 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1172 }
f70ea018
TH
1173
1174 return skb->hash;
1175}
1176
20a17bf6 1177__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1178
20a17bf6 1179static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1180{
c6cc1ca7
TH
1181 if (!skb->l4_hash && !skb->sw_hash) {
1182 struct flow_keys keys;
de4c1f8b 1183 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1184
de4c1f8b 1185 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1186 }
f70ea018
TH
1187
1188 return skb->hash;
1189}
1190
50fb7992
TH
1191__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1192
57bdf7f4
TH
1193static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1194{
61b905da 1195 return skb->hash;
57bdf7f4
TH
1196}
1197
3df7a74e
TH
1198static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1199{
61b905da 1200 to->hash = from->hash;
a3b18ddb 1201 to->sw_hash = from->sw_hash;
61b905da 1202 to->l4_hash = from->l4_hash;
3df7a74e
TH
1203};
1204
4305b541
ACM
1205#ifdef NET_SKBUFF_DATA_USES_OFFSET
1206static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1207{
1208 return skb->head + skb->end;
1209}
ec47ea82
AD
1210
1211static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1212{
1213 return skb->end;
1214}
4305b541
ACM
1215#else
1216static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1217{
1218 return skb->end;
1219}
ec47ea82
AD
1220
1221static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1222{
1223 return skb->end - skb->head;
1224}
4305b541
ACM
1225#endif
1226
1da177e4 1227/* Internal */
4305b541 1228#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1229
ac45f602
PO
1230static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1231{
1232 return &skb_shinfo(skb)->hwtstamps;
1233}
1234
1da177e4
LT
1235/**
1236 * skb_queue_empty - check if a queue is empty
1237 * @list: queue head
1238 *
1239 * Returns true if the queue is empty, false otherwise.
1240 */
1241static inline int skb_queue_empty(const struct sk_buff_head *list)
1242{
fd44b93c 1243 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1244}
1245
fc7ebb21
DM
1246/**
1247 * skb_queue_is_last - check if skb is the last entry in the queue
1248 * @list: queue head
1249 * @skb: buffer
1250 *
1251 * Returns true if @skb is the last buffer on the list.
1252 */
1253static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1254 const struct sk_buff *skb)
1255{
fd44b93c 1256 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1257}
1258
832d11c5
IJ
1259/**
1260 * skb_queue_is_first - check if skb is the first entry in the queue
1261 * @list: queue head
1262 * @skb: buffer
1263 *
1264 * Returns true if @skb is the first buffer on the list.
1265 */
1266static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1267 const struct sk_buff *skb)
1268{
fd44b93c 1269 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1270}
1271
249c8b42
DM
1272/**
1273 * skb_queue_next - return the next packet in the queue
1274 * @list: queue head
1275 * @skb: current buffer
1276 *
1277 * Return the next packet in @list after @skb. It is only valid to
1278 * call this if skb_queue_is_last() evaluates to false.
1279 */
1280static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1281 const struct sk_buff *skb)
1282{
1283 /* This BUG_ON may seem severe, but if we just return then we
1284 * are going to dereference garbage.
1285 */
1286 BUG_ON(skb_queue_is_last(list, skb));
1287 return skb->next;
1288}
1289
832d11c5
IJ
1290/**
1291 * skb_queue_prev - return the prev packet in the queue
1292 * @list: queue head
1293 * @skb: current buffer
1294 *
1295 * Return the prev packet in @list before @skb. It is only valid to
1296 * call this if skb_queue_is_first() evaluates to false.
1297 */
1298static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1299 const struct sk_buff *skb)
1300{
1301 /* This BUG_ON may seem severe, but if we just return then we
1302 * are going to dereference garbage.
1303 */
1304 BUG_ON(skb_queue_is_first(list, skb));
1305 return skb->prev;
1306}
1307
1da177e4
LT
1308/**
1309 * skb_get - reference buffer
1310 * @skb: buffer to reference
1311 *
1312 * Makes another reference to a socket buffer and returns a pointer
1313 * to the buffer.
1314 */
1315static inline struct sk_buff *skb_get(struct sk_buff *skb)
1316{
1317 atomic_inc(&skb->users);
1318 return skb;
1319}
1320
1321/*
1322 * If users == 1, we are the only owner and are can avoid redundant
1323 * atomic change.
1324 */
1325
1da177e4
LT
1326/**
1327 * skb_cloned - is the buffer a clone
1328 * @skb: buffer to check
1329 *
1330 * Returns true if the buffer was generated with skb_clone() and is
1331 * one of multiple shared copies of the buffer. Cloned buffers are
1332 * shared data so must not be written to under normal circumstances.
1333 */
1334static inline int skb_cloned(const struct sk_buff *skb)
1335{
1336 return skb->cloned &&
1337 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1338}
1339
14bbd6a5
PS
1340static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1341{
d0164adc 1342 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1343
1344 if (skb_cloned(skb))
1345 return pskb_expand_head(skb, 0, 0, pri);
1346
1347 return 0;
1348}
1349
1da177e4
LT
1350/**
1351 * skb_header_cloned - is the header a clone
1352 * @skb: buffer to check
1353 *
1354 * Returns true if modifying the header part of the buffer requires
1355 * the data to be copied.
1356 */
1357static inline int skb_header_cloned(const struct sk_buff *skb)
1358{
1359 int dataref;
1360
1361 if (!skb->cloned)
1362 return 0;
1363
1364 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1365 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1366 return dataref != 1;
1367}
1368
9580bf2e
ED
1369static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1370{
1371 might_sleep_if(gfpflags_allow_blocking(pri));
1372
1373 if (skb_header_cloned(skb))
1374 return pskb_expand_head(skb, 0, 0, pri);
1375
1376 return 0;
1377}
1378
1da177e4
LT
1379/**
1380 * skb_header_release - release reference to header
1381 * @skb: buffer to operate on
1382 *
1383 * Drop a reference to the header part of the buffer. This is done
1384 * by acquiring a payload reference. You must not read from the header
1385 * part of skb->data after this.
f4a775d1 1386 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1387 */
1388static inline void skb_header_release(struct sk_buff *skb)
1389{
1390 BUG_ON(skb->nohdr);
1391 skb->nohdr = 1;
1392 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1393}
1394
f4a775d1
ED
1395/**
1396 * __skb_header_release - release reference to header
1397 * @skb: buffer to operate on
1398 *
1399 * Variant of skb_header_release() assuming skb is private to caller.
1400 * We can avoid one atomic operation.
1401 */
1402static inline void __skb_header_release(struct sk_buff *skb)
1403{
1404 skb->nohdr = 1;
1405 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1406}
1407
1408
1da177e4
LT
1409/**
1410 * skb_shared - is the buffer shared
1411 * @skb: buffer to check
1412 *
1413 * Returns true if more than one person has a reference to this
1414 * buffer.
1415 */
1416static inline int skb_shared(const struct sk_buff *skb)
1417{
1418 return atomic_read(&skb->users) != 1;
1419}
1420
1421/**
1422 * skb_share_check - check if buffer is shared and if so clone it
1423 * @skb: buffer to check
1424 * @pri: priority for memory allocation
1425 *
1426 * If the buffer is shared the buffer is cloned and the old copy
1427 * drops a reference. A new clone with a single reference is returned.
1428 * If the buffer is not shared the original buffer is returned. When
1429 * being called from interrupt status or with spinlocks held pri must
1430 * be GFP_ATOMIC.
1431 *
1432 * NULL is returned on a memory allocation failure.
1433 */
47061bc4 1434static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1435{
d0164adc 1436 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1437 if (skb_shared(skb)) {
1438 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1439
1440 if (likely(nskb))
1441 consume_skb(skb);
1442 else
1443 kfree_skb(skb);
1da177e4
LT
1444 skb = nskb;
1445 }
1446 return skb;
1447}
1448
1449/*
1450 * Copy shared buffers into a new sk_buff. We effectively do COW on
1451 * packets to handle cases where we have a local reader and forward
1452 * and a couple of other messy ones. The normal one is tcpdumping
1453 * a packet thats being forwarded.
1454 */
1455
1456/**
1457 * skb_unshare - make a copy of a shared buffer
1458 * @skb: buffer to check
1459 * @pri: priority for memory allocation
1460 *
1461 * If the socket buffer is a clone then this function creates a new
1462 * copy of the data, drops a reference count on the old copy and returns
1463 * the new copy with the reference count at 1. If the buffer is not a clone
1464 * the original buffer is returned. When called with a spinlock held or
1465 * from interrupt state @pri must be %GFP_ATOMIC
1466 *
1467 * %NULL is returned on a memory allocation failure.
1468 */
e2bf521d 1469static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1470 gfp_t pri)
1da177e4 1471{
d0164adc 1472 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1473 if (skb_cloned(skb)) {
1474 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1475
1476 /* Free our shared copy */
1477 if (likely(nskb))
1478 consume_skb(skb);
1479 else
1480 kfree_skb(skb);
1da177e4
LT
1481 skb = nskb;
1482 }
1483 return skb;
1484}
1485
1486/**
1a5778aa 1487 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1488 * @list_: list to peek at
1489 *
1490 * Peek an &sk_buff. Unlike most other operations you _MUST_
1491 * be careful with this one. A peek leaves the buffer on the
1492 * list and someone else may run off with it. You must hold
1493 * the appropriate locks or have a private queue to do this.
1494 *
1495 * Returns %NULL for an empty list or a pointer to the head element.
1496 * The reference count is not incremented and the reference is therefore
1497 * volatile. Use with caution.
1498 */
05bdd2f1 1499static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1500{
18d07000
ED
1501 struct sk_buff *skb = list_->next;
1502
1503 if (skb == (struct sk_buff *)list_)
1504 skb = NULL;
1505 return skb;
1da177e4
LT
1506}
1507
da5ef6e5
PE
1508/**
1509 * skb_peek_next - peek skb following the given one from a queue
1510 * @skb: skb to start from
1511 * @list_: list to peek at
1512 *
1513 * Returns %NULL when the end of the list is met or a pointer to the
1514 * next element. The reference count is not incremented and the
1515 * reference is therefore volatile. Use with caution.
1516 */
1517static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1518 const struct sk_buff_head *list_)
1519{
1520 struct sk_buff *next = skb->next;
18d07000 1521
da5ef6e5
PE
1522 if (next == (struct sk_buff *)list_)
1523 next = NULL;
1524 return next;
1525}
1526
1da177e4 1527/**
1a5778aa 1528 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1529 * @list_: list to peek at
1530 *
1531 * Peek an &sk_buff. Unlike most other operations you _MUST_
1532 * be careful with this one. A peek leaves the buffer on the
1533 * list and someone else may run off with it. You must hold
1534 * the appropriate locks or have a private queue to do this.
1535 *
1536 * Returns %NULL for an empty list or a pointer to the tail element.
1537 * The reference count is not incremented and the reference is therefore
1538 * volatile. Use with caution.
1539 */
05bdd2f1 1540static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1541{
18d07000
ED
1542 struct sk_buff *skb = list_->prev;
1543
1544 if (skb == (struct sk_buff *)list_)
1545 skb = NULL;
1546 return skb;
1547
1da177e4
LT
1548}
1549
1550/**
1551 * skb_queue_len - get queue length
1552 * @list_: list to measure
1553 *
1554 * Return the length of an &sk_buff queue.
1555 */
1556static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1557{
1558 return list_->qlen;
1559}
1560
67fed459
DM
1561/**
1562 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1563 * @list: queue to initialize
1564 *
1565 * This initializes only the list and queue length aspects of
1566 * an sk_buff_head object. This allows to initialize the list
1567 * aspects of an sk_buff_head without reinitializing things like
1568 * the spinlock. It can also be used for on-stack sk_buff_head
1569 * objects where the spinlock is known to not be used.
1570 */
1571static inline void __skb_queue_head_init(struct sk_buff_head *list)
1572{
1573 list->prev = list->next = (struct sk_buff *)list;
1574 list->qlen = 0;
1575}
1576
76f10ad0
AV
1577/*
1578 * This function creates a split out lock class for each invocation;
1579 * this is needed for now since a whole lot of users of the skb-queue
1580 * infrastructure in drivers have different locking usage (in hardirq)
1581 * than the networking core (in softirq only). In the long run either the
1582 * network layer or drivers should need annotation to consolidate the
1583 * main types of usage into 3 classes.
1584 */
1da177e4
LT
1585static inline void skb_queue_head_init(struct sk_buff_head *list)
1586{
1587 spin_lock_init(&list->lock);
67fed459 1588 __skb_queue_head_init(list);
1da177e4
LT
1589}
1590
c2ecba71
PE
1591static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1592 struct lock_class_key *class)
1593{
1594 skb_queue_head_init(list);
1595 lockdep_set_class(&list->lock, class);
1596}
1597
1da177e4 1598/*
bf299275 1599 * Insert an sk_buff on a list.
1da177e4
LT
1600 *
1601 * The "__skb_xxxx()" functions are the non-atomic ones that
1602 * can only be called with interrupts disabled.
1603 */
7965bd4d
JP
1604void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1605 struct sk_buff_head *list);
bf299275
GR
1606static inline void __skb_insert(struct sk_buff *newsk,
1607 struct sk_buff *prev, struct sk_buff *next,
1608 struct sk_buff_head *list)
1609{
1610 newsk->next = next;
1611 newsk->prev = prev;
1612 next->prev = prev->next = newsk;
1613 list->qlen++;
1614}
1da177e4 1615
67fed459
DM
1616static inline void __skb_queue_splice(const struct sk_buff_head *list,
1617 struct sk_buff *prev,
1618 struct sk_buff *next)
1619{
1620 struct sk_buff *first = list->next;
1621 struct sk_buff *last = list->prev;
1622
1623 first->prev = prev;
1624 prev->next = first;
1625
1626 last->next = next;
1627 next->prev = last;
1628}
1629
1630/**
1631 * skb_queue_splice - join two skb lists, this is designed for stacks
1632 * @list: the new list to add
1633 * @head: the place to add it in the first list
1634 */
1635static inline void skb_queue_splice(const struct sk_buff_head *list,
1636 struct sk_buff_head *head)
1637{
1638 if (!skb_queue_empty(list)) {
1639 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1640 head->qlen += list->qlen;
67fed459
DM
1641 }
1642}
1643
1644/**
d9619496 1645 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1646 * @list: the new list to add
1647 * @head: the place to add it in the first list
1648 *
1649 * The list at @list is reinitialised
1650 */
1651static inline void skb_queue_splice_init(struct sk_buff_head *list,
1652 struct sk_buff_head *head)
1653{
1654 if (!skb_queue_empty(list)) {
1655 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1656 head->qlen += list->qlen;
67fed459
DM
1657 __skb_queue_head_init(list);
1658 }
1659}
1660
1661/**
1662 * skb_queue_splice_tail - join two skb lists, each list being a queue
1663 * @list: the new list to add
1664 * @head: the place to add it in the first list
1665 */
1666static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1667 struct sk_buff_head *head)
1668{
1669 if (!skb_queue_empty(list)) {
1670 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1671 head->qlen += list->qlen;
67fed459
DM
1672 }
1673}
1674
1675/**
d9619496 1676 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1677 * @list: the new list to add
1678 * @head: the place to add it in the first list
1679 *
1680 * Each of the lists is a queue.
1681 * The list at @list is reinitialised
1682 */
1683static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1684 struct sk_buff_head *head)
1685{
1686 if (!skb_queue_empty(list)) {
1687 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1688 head->qlen += list->qlen;
67fed459
DM
1689 __skb_queue_head_init(list);
1690 }
1691}
1692
1da177e4 1693/**
300ce174 1694 * __skb_queue_after - queue a buffer at the list head
1da177e4 1695 * @list: list to use
300ce174 1696 * @prev: place after this buffer
1da177e4
LT
1697 * @newsk: buffer to queue
1698 *
300ce174 1699 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1700 * and you must therefore hold required locks before calling it.
1701 *
1702 * A buffer cannot be placed on two lists at the same time.
1703 */
300ce174
SH
1704static inline void __skb_queue_after(struct sk_buff_head *list,
1705 struct sk_buff *prev,
1706 struct sk_buff *newsk)
1da177e4 1707{
bf299275 1708 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1709}
1710
7965bd4d
JP
1711void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1712 struct sk_buff_head *list);
7de6c033 1713
f5572855
GR
1714static inline void __skb_queue_before(struct sk_buff_head *list,
1715 struct sk_buff *next,
1716 struct sk_buff *newsk)
1717{
1718 __skb_insert(newsk, next->prev, next, list);
1719}
1720
300ce174
SH
1721/**
1722 * __skb_queue_head - queue a buffer at the list head
1723 * @list: list to use
1724 * @newsk: buffer to queue
1725 *
1726 * Queue a buffer at the start of a list. This function takes no locks
1727 * and you must therefore hold required locks before calling it.
1728 *
1729 * A buffer cannot be placed on two lists at the same time.
1730 */
7965bd4d 1731void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1732static inline void __skb_queue_head(struct sk_buff_head *list,
1733 struct sk_buff *newsk)
1734{
1735 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1736}
1737
1da177e4
LT
1738/**
1739 * __skb_queue_tail - queue a buffer at the list tail
1740 * @list: list to use
1741 * @newsk: buffer to queue
1742 *
1743 * Queue a buffer at the end of a list. This function takes no locks
1744 * and you must therefore hold required locks before calling it.
1745 *
1746 * A buffer cannot be placed on two lists at the same time.
1747 */
7965bd4d 1748void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1749static inline void __skb_queue_tail(struct sk_buff_head *list,
1750 struct sk_buff *newsk)
1751{
f5572855 1752 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1753}
1754
1da177e4
LT
1755/*
1756 * remove sk_buff from list. _Must_ be called atomically, and with
1757 * the list known..
1758 */
7965bd4d 1759void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1760static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1761{
1762 struct sk_buff *next, *prev;
1763
1764 list->qlen--;
1765 next = skb->next;
1766 prev = skb->prev;
1767 skb->next = skb->prev = NULL;
1da177e4
LT
1768 next->prev = prev;
1769 prev->next = next;
1770}
1771
f525c06d
GR
1772/**
1773 * __skb_dequeue - remove from the head of the queue
1774 * @list: list to dequeue from
1775 *
1776 * Remove the head of the list. This function does not take any locks
1777 * so must be used with appropriate locks held only. The head item is
1778 * returned or %NULL if the list is empty.
1779 */
7965bd4d 1780struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1781static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1782{
1783 struct sk_buff *skb = skb_peek(list);
1784 if (skb)
1785 __skb_unlink(skb, list);
1786 return skb;
1787}
1da177e4
LT
1788
1789/**
1790 * __skb_dequeue_tail - remove from the tail of the queue
1791 * @list: list to dequeue from
1792 *
1793 * Remove the tail of the list. This function does not take any locks
1794 * so must be used with appropriate locks held only. The tail item is
1795 * returned or %NULL if the list is empty.
1796 */
7965bd4d 1797struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1798static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1799{
1800 struct sk_buff *skb = skb_peek_tail(list);
1801 if (skb)
1802 __skb_unlink(skb, list);
1803 return skb;
1804}
1805
1806
bdcc0924 1807static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1808{
1809 return skb->data_len;
1810}
1811
1812static inline unsigned int skb_headlen(const struct sk_buff *skb)
1813{
1814 return skb->len - skb->data_len;
1815}
1816
c72d8cda 1817static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1da177e4 1818{
c72d8cda 1819 unsigned int i, len = 0;
1da177e4 1820
c72d8cda 1821 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 1822 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1823 return len + skb_headlen(skb);
1824}
1825
131ea667
IC
1826/**
1827 * __skb_fill_page_desc - initialise a paged fragment in an skb
1828 * @skb: buffer containing fragment to be initialised
1829 * @i: paged fragment index to initialise
1830 * @page: the page to use for this fragment
1831 * @off: the offset to the data with @page
1832 * @size: the length of the data
1833 *
1834 * Initialises the @i'th fragment of @skb to point to &size bytes at
1835 * offset @off within @page.
1836 *
1837 * Does not take any additional reference on the fragment.
1838 */
1839static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1840 struct page *page, int off, int size)
1da177e4
LT
1841{
1842 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1843
c48a11c7 1844 /*
2f064f34
MH
1845 * Propagate page pfmemalloc to the skb if we can. The problem is
1846 * that not all callers have unique ownership of the page but rely
1847 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1848 */
a8605c60 1849 frag->page.p = page;
1da177e4 1850 frag->page_offset = off;
9e903e08 1851 skb_frag_size_set(frag, size);
cca7af38
PE
1852
1853 page = compound_head(page);
2f064f34 1854 if (page_is_pfmemalloc(page))
cca7af38 1855 skb->pfmemalloc = true;
131ea667
IC
1856}
1857
1858/**
1859 * skb_fill_page_desc - initialise a paged fragment in an skb
1860 * @skb: buffer containing fragment to be initialised
1861 * @i: paged fragment index to initialise
1862 * @page: the page to use for this fragment
1863 * @off: the offset to the data with @page
1864 * @size: the length of the data
1865 *
1866 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1867 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1868 * addition updates @skb such that @i is the last fragment.
1869 *
1870 * Does not take any additional reference on the fragment.
1871 */
1872static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1873 struct page *page, int off, int size)
1874{
1875 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1876 skb_shinfo(skb)->nr_frags = i + 1;
1877}
1878
7965bd4d
JP
1879void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1880 int size, unsigned int truesize);
654bed16 1881
f8e617e1
JW
1882void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1883 unsigned int truesize);
1884
1da177e4 1885#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1886#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1887#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1888
27a884dc
ACM
1889#ifdef NET_SKBUFF_DATA_USES_OFFSET
1890static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1891{
1892 return skb->head + skb->tail;
1893}
1894
1895static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1896{
1897 skb->tail = skb->data - skb->head;
1898}
1899
1900static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1901{
1902 skb_reset_tail_pointer(skb);
1903 skb->tail += offset;
1904}
7cc46190 1905
27a884dc
ACM
1906#else /* NET_SKBUFF_DATA_USES_OFFSET */
1907static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1908{
1909 return skb->tail;
1910}
1911
1912static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1913{
1914 skb->tail = skb->data;
1915}
1916
1917static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1918{
1919 skb->tail = skb->data + offset;
1920}
4305b541 1921
27a884dc
ACM
1922#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1923
1da177e4
LT
1924/*
1925 * Add data to an sk_buff
1926 */
0c7ddf36 1927unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1928unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1929static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1930{
27a884dc 1931 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1932 SKB_LINEAR_ASSERT(skb);
1933 skb->tail += len;
1934 skb->len += len;
1935 return tmp;
1936}
1937
7965bd4d 1938unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1939static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1940{
1941 skb->data -= len;
1942 skb->len += len;
1943 return skb->data;
1944}
1945
7965bd4d 1946unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1947static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1948{
1949 skb->len -= len;
1950 BUG_ON(skb->len < skb->data_len);
1951 return skb->data += len;
1952}
1953
47d29646
DM
1954static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1955{
1956 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1957}
1958
7965bd4d 1959unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1960
1961static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1962{
1963 if (len > skb_headlen(skb) &&
987c402a 1964 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1965 return NULL;
1966 skb->len -= len;
1967 return skb->data += len;
1968}
1969
1970static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1971{
1972 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1973}
1974
1975static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1976{
1977 if (likely(len <= skb_headlen(skb)))
1978 return 1;
1979 if (unlikely(len > skb->len))
1980 return 0;
987c402a 1981 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1982}
1983
c8c8b127
ED
1984void skb_condense(struct sk_buff *skb);
1985
1da177e4
LT
1986/**
1987 * skb_headroom - bytes at buffer head
1988 * @skb: buffer to check
1989 *
1990 * Return the number of bytes of free space at the head of an &sk_buff.
1991 */
c2636b4d 1992static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1993{
1994 return skb->data - skb->head;
1995}
1996
1997/**
1998 * skb_tailroom - bytes at buffer end
1999 * @skb: buffer to check
2000 *
2001 * Return the number of bytes of free space at the tail of an sk_buff
2002 */
2003static inline int skb_tailroom(const struct sk_buff *skb)
2004{
4305b541 2005 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2006}
2007
a21d4572
ED
2008/**
2009 * skb_availroom - bytes at buffer end
2010 * @skb: buffer to check
2011 *
2012 * Return the number of bytes of free space at the tail of an sk_buff
2013 * allocated by sk_stream_alloc()
2014 */
2015static inline int skb_availroom(const struct sk_buff *skb)
2016{
16fad69c
ED
2017 if (skb_is_nonlinear(skb))
2018 return 0;
2019
2020 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2021}
2022
1da177e4
LT
2023/**
2024 * skb_reserve - adjust headroom
2025 * @skb: buffer to alter
2026 * @len: bytes to move
2027 *
2028 * Increase the headroom of an empty &sk_buff by reducing the tail
2029 * room. This is only allowed for an empty buffer.
2030 */
8243126c 2031static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2032{
2033 skb->data += len;
2034 skb->tail += len;
2035}
2036
1837b2e2
BP
2037/**
2038 * skb_tailroom_reserve - adjust reserved_tailroom
2039 * @skb: buffer to alter
2040 * @mtu: maximum amount of headlen permitted
2041 * @needed_tailroom: minimum amount of reserved_tailroom
2042 *
2043 * Set reserved_tailroom so that headlen can be as large as possible but
2044 * not larger than mtu and tailroom cannot be smaller than
2045 * needed_tailroom.
2046 * The required headroom should already have been reserved before using
2047 * this function.
2048 */
2049static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2050 unsigned int needed_tailroom)
2051{
2052 SKB_LINEAR_ASSERT(skb);
2053 if (mtu < skb_tailroom(skb) - needed_tailroom)
2054 /* use at most mtu */
2055 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2056 else
2057 /* use up to all available space */
2058 skb->reserved_tailroom = needed_tailroom;
2059}
2060
8bce6d7d
TH
2061#define ENCAP_TYPE_ETHER 0
2062#define ENCAP_TYPE_IPPROTO 1
2063
2064static inline void skb_set_inner_protocol(struct sk_buff *skb,
2065 __be16 protocol)
2066{
2067 skb->inner_protocol = protocol;
2068 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2069}
2070
2071static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2072 __u8 ipproto)
2073{
2074 skb->inner_ipproto = ipproto;
2075 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2076}
2077
6a674e9c
JG
2078static inline void skb_reset_inner_headers(struct sk_buff *skb)
2079{
aefbd2b3 2080 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2081 skb->inner_network_header = skb->network_header;
2082 skb->inner_transport_header = skb->transport_header;
2083}
2084
0b5c9db1
JP
2085static inline void skb_reset_mac_len(struct sk_buff *skb)
2086{
2087 skb->mac_len = skb->network_header - skb->mac_header;
2088}
2089
6a674e9c
JG
2090static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2091 *skb)
2092{
2093 return skb->head + skb->inner_transport_header;
2094}
2095
55dc5a9f
TH
2096static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2097{
2098 return skb_inner_transport_header(skb) - skb->data;
2099}
2100
6a674e9c
JG
2101static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2102{
2103 skb->inner_transport_header = skb->data - skb->head;
2104}
2105
2106static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2107 const int offset)
2108{
2109 skb_reset_inner_transport_header(skb);
2110 skb->inner_transport_header += offset;
2111}
2112
2113static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2114{
2115 return skb->head + skb->inner_network_header;
2116}
2117
2118static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2119{
2120 skb->inner_network_header = skb->data - skb->head;
2121}
2122
2123static inline void skb_set_inner_network_header(struct sk_buff *skb,
2124 const int offset)
2125{
2126 skb_reset_inner_network_header(skb);
2127 skb->inner_network_header += offset;
2128}
2129
aefbd2b3
PS
2130static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2131{
2132 return skb->head + skb->inner_mac_header;
2133}
2134
2135static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2136{
2137 skb->inner_mac_header = skb->data - skb->head;
2138}
2139
2140static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2141 const int offset)
2142{
2143 skb_reset_inner_mac_header(skb);
2144 skb->inner_mac_header += offset;
2145}
fda55eca
ED
2146static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2147{
35d04610 2148 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2149}
2150
9c70220b
ACM
2151static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2152{
2e07fa9c 2153 return skb->head + skb->transport_header;
9c70220b
ACM
2154}
2155
badff6d0
ACM
2156static inline void skb_reset_transport_header(struct sk_buff *skb)
2157{
2e07fa9c 2158 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2159}
2160
967b05f6
ACM
2161static inline void skb_set_transport_header(struct sk_buff *skb,
2162 const int offset)
2163{
2e07fa9c
ACM
2164 skb_reset_transport_header(skb);
2165 skb->transport_header += offset;
ea2ae17d
ACM
2166}
2167
d56f90a7
ACM
2168static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2169{
2e07fa9c 2170 return skb->head + skb->network_header;
d56f90a7
ACM
2171}
2172
c1d2bbe1
ACM
2173static inline void skb_reset_network_header(struct sk_buff *skb)
2174{
2e07fa9c 2175 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2176}
2177
c14d2450
ACM
2178static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2179{
2e07fa9c
ACM
2180 skb_reset_network_header(skb);
2181 skb->network_header += offset;
c14d2450
ACM
2182}
2183
2e07fa9c 2184static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2185{
2e07fa9c 2186 return skb->head + skb->mac_header;
bbe735e4
ACM
2187}
2188
ea6da4fd
AV
2189static inline int skb_mac_offset(const struct sk_buff *skb)
2190{
2191 return skb_mac_header(skb) - skb->data;
2192}
2193
2e07fa9c 2194static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2195{
35d04610 2196 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2197}
2198
2199static inline void skb_reset_mac_header(struct sk_buff *skb)
2200{
2201 skb->mac_header = skb->data - skb->head;
2202}
2203
2204static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2205{
2206 skb_reset_mac_header(skb);
2207 skb->mac_header += offset;
2208}
2209
0e3da5bb
TT
2210static inline void skb_pop_mac_header(struct sk_buff *skb)
2211{
2212 skb->mac_header = skb->network_header;
2213}
2214
fbbdb8f0
YX
2215static inline void skb_probe_transport_header(struct sk_buff *skb,
2216 const int offset_hint)
2217{
2218 struct flow_keys keys;
2219
2220 if (skb_transport_header_was_set(skb))
2221 return;
cd79a238 2222 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2223 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2224 else
2225 skb_set_transport_header(skb, offset_hint);
2226}
2227
03606895
ED
2228static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2229{
2230 if (skb_mac_header_was_set(skb)) {
2231 const unsigned char *old_mac = skb_mac_header(skb);
2232
2233 skb_set_mac_header(skb, -skb->mac_len);
2234 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2235 }
2236}
2237
04fb451e
MM
2238static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2239{
2240 return skb->csum_start - skb_headroom(skb);
2241}
2242
08b64fcc
AD
2243static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2244{
2245 return skb->head + skb->csum_start;
2246}
2247
2e07fa9c
ACM
2248static inline int skb_transport_offset(const struct sk_buff *skb)
2249{
2250 return skb_transport_header(skb) - skb->data;
2251}
2252
2253static inline u32 skb_network_header_len(const struct sk_buff *skb)
2254{
2255 return skb->transport_header - skb->network_header;
2256}
2257
6a674e9c
JG
2258static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2259{
2260 return skb->inner_transport_header - skb->inner_network_header;
2261}
2262
2e07fa9c
ACM
2263static inline int skb_network_offset(const struct sk_buff *skb)
2264{
2265 return skb_network_header(skb) - skb->data;
2266}
48d49d0c 2267
6a674e9c
JG
2268static inline int skb_inner_network_offset(const struct sk_buff *skb)
2269{
2270 return skb_inner_network_header(skb) - skb->data;
2271}
2272
f9599ce1
CG
2273static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2274{
2275 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2276}
2277
1da177e4
LT
2278/*
2279 * CPUs often take a performance hit when accessing unaligned memory
2280 * locations. The actual performance hit varies, it can be small if the
2281 * hardware handles it or large if we have to take an exception and fix it
2282 * in software.
2283 *
2284 * Since an ethernet header is 14 bytes network drivers often end up with
2285 * the IP header at an unaligned offset. The IP header can be aligned by
2286 * shifting the start of the packet by 2 bytes. Drivers should do this
2287 * with:
2288 *
8660c124 2289 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2290 *
2291 * The downside to this alignment of the IP header is that the DMA is now
2292 * unaligned. On some architectures the cost of an unaligned DMA is high
2293 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2294 *
1da177e4
LT
2295 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2296 * to be overridden.
2297 */
2298#ifndef NET_IP_ALIGN
2299#define NET_IP_ALIGN 2
2300#endif
2301
025be81e
AB
2302/*
2303 * The networking layer reserves some headroom in skb data (via
2304 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2305 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2306 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2307 *
2308 * Unfortunately this headroom changes the DMA alignment of the resulting
2309 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2310 * on some architectures. An architecture can override this value,
2311 * perhaps setting it to a cacheline in size (since that will maintain
2312 * cacheline alignment of the DMA). It must be a power of 2.
2313 *
d6301d3d 2314 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2315 * headroom, you should not reduce this.
5933dd2f
ED
2316 *
2317 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2318 * to reduce average number of cache lines per packet.
2319 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2320 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2321 */
2322#ifndef NET_SKB_PAD
5933dd2f 2323#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2324#endif
2325
7965bd4d 2326int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2327
5293efe6 2328static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2329{
c4264f27 2330 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2331 WARN_ON(1);
2332 return;
2333 }
27a884dc
ACM
2334 skb->len = len;
2335 skb_set_tail_pointer(skb, len);
1da177e4
LT
2336}
2337
5293efe6
DB
2338static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2339{
2340 __skb_set_length(skb, len);
2341}
2342
7965bd4d 2343void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2344
2345static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2346{
3cc0e873
HX
2347 if (skb->data_len)
2348 return ___pskb_trim(skb, len);
2349 __skb_trim(skb, len);
2350 return 0;
1da177e4
LT
2351}
2352
2353static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2354{
2355 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2356}
2357
e9fa4f7b
HX
2358/**
2359 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2360 * @skb: buffer to alter
2361 * @len: new length
2362 *
2363 * This is identical to pskb_trim except that the caller knows that
2364 * the skb is not cloned so we should never get an error due to out-
2365 * of-memory.
2366 */
2367static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2368{
2369 int err = pskb_trim(skb, len);
2370 BUG_ON(err);
2371}
2372
5293efe6
DB
2373static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2374{
2375 unsigned int diff = len - skb->len;
2376
2377 if (skb_tailroom(skb) < diff) {
2378 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2379 GFP_ATOMIC);
2380 if (ret)
2381 return ret;
2382 }
2383 __skb_set_length(skb, len);
2384 return 0;
2385}
2386
1da177e4
LT
2387/**
2388 * skb_orphan - orphan a buffer
2389 * @skb: buffer to orphan
2390 *
2391 * If a buffer currently has an owner then we call the owner's
2392 * destructor function and make the @skb unowned. The buffer continues
2393 * to exist but is no longer charged to its former owner.
2394 */
2395static inline void skb_orphan(struct sk_buff *skb)
2396{
c34a7612 2397 if (skb->destructor) {
1da177e4 2398 skb->destructor(skb);
c34a7612
ED
2399 skb->destructor = NULL;
2400 skb->sk = NULL;
376c7311
ED
2401 } else {
2402 BUG_ON(skb->sk);
c34a7612 2403 }
1da177e4
LT
2404}
2405
a353e0ce
MT
2406/**
2407 * skb_orphan_frags - orphan the frags contained in a buffer
2408 * @skb: buffer to orphan frags from
2409 * @gfp_mask: allocation mask for replacement pages
2410 *
2411 * For each frag in the SKB which needs a destructor (i.e. has an
2412 * owner) create a copy of that frag and release the original
2413 * page by calling the destructor.
2414 */
2415static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2416{
2417 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2418 return 0;
2419 return skb_copy_ubufs(skb, gfp_mask);
2420}
2421
1da177e4
LT
2422/**
2423 * __skb_queue_purge - empty a list
2424 * @list: list to empty
2425 *
2426 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2427 * the list and one reference dropped. This function does not take the
2428 * list lock and the caller must hold the relevant locks to use it.
2429 */
7965bd4d 2430void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2431static inline void __skb_queue_purge(struct sk_buff_head *list)
2432{
2433 struct sk_buff *skb;
2434 while ((skb = __skb_dequeue(list)) != NULL)
2435 kfree_skb(skb);
2436}
2437
9f5afeae
YW
2438void skb_rbtree_purge(struct rb_root *root);
2439
7965bd4d 2440void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2441
7965bd4d
JP
2442struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2443 gfp_t gfp_mask);
8af27456
CH
2444
2445/**
2446 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2447 * @dev: network device to receive on
2448 * @length: length to allocate
2449 *
2450 * Allocate a new &sk_buff and assign it a usage count of one. The
2451 * buffer has unspecified headroom built in. Users should allocate
2452 * the headroom they think they need without accounting for the
2453 * built in space. The built in space is used for optimisations.
2454 *
2455 * %NULL is returned if there is no free memory. Although this function
2456 * allocates memory it can be called from an interrupt.
2457 */
2458static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2459 unsigned int length)
8af27456
CH
2460{
2461 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2462}
2463
6f532612
ED
2464/* legacy helper around __netdev_alloc_skb() */
2465static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2466 gfp_t gfp_mask)
2467{
2468 return __netdev_alloc_skb(NULL, length, gfp_mask);
2469}
2470
2471/* legacy helper around netdev_alloc_skb() */
2472static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2473{
2474 return netdev_alloc_skb(NULL, length);
2475}
2476
2477
4915a0de
ED
2478static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2479 unsigned int length, gfp_t gfp)
61321bbd 2480{
4915a0de 2481 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2482
2483 if (NET_IP_ALIGN && skb)
2484 skb_reserve(skb, NET_IP_ALIGN);
2485 return skb;
2486}
2487
4915a0de
ED
2488static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2489 unsigned int length)
2490{
2491 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2492}
2493
181edb2b
AD
2494static inline void skb_free_frag(void *addr)
2495{
8c2dd3e4 2496 page_frag_free(addr);
181edb2b
AD
2497}
2498
ffde7328 2499void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2500struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2501 unsigned int length, gfp_t gfp_mask);
2502static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2503 unsigned int length)
2504{
2505 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2506}
795bb1c0
JDB
2507void napi_consume_skb(struct sk_buff *skb, int budget);
2508
2509void __kfree_skb_flush(void);
15fad714 2510void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2511
71dfda58
AD
2512/**
2513 * __dev_alloc_pages - allocate page for network Rx
2514 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2515 * @order: size of the allocation
2516 *
2517 * Allocate a new page.
2518 *
2519 * %NULL is returned if there is no free memory.
2520*/
2521static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2522 unsigned int order)
2523{
2524 /* This piece of code contains several assumptions.
2525 * 1. This is for device Rx, therefor a cold page is preferred.
2526 * 2. The expectation is the user wants a compound page.
2527 * 3. If requesting a order 0 page it will not be compound
2528 * due to the check to see if order has a value in prep_new_page
2529 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2530 * code in gfp_to_alloc_flags that should be enforcing this.
2531 */
2532 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2533
2534 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2535}
2536
2537static inline struct page *dev_alloc_pages(unsigned int order)
2538{
95829b3a 2539 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2540}
2541
2542/**
2543 * __dev_alloc_page - allocate a page for network Rx
2544 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2545 *
2546 * Allocate a new page.
2547 *
2548 * %NULL is returned if there is no free memory.
2549 */
2550static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2551{
2552 return __dev_alloc_pages(gfp_mask, 0);
2553}
2554
2555static inline struct page *dev_alloc_page(void)
2556{
95829b3a 2557 return dev_alloc_pages(0);
71dfda58
AD
2558}
2559
0614002b
MG
2560/**
2561 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2562 * @page: The page that was allocated from skb_alloc_page
2563 * @skb: The skb that may need pfmemalloc set
2564 */
2565static inline void skb_propagate_pfmemalloc(struct page *page,
2566 struct sk_buff *skb)
2567{
2f064f34 2568 if (page_is_pfmemalloc(page))
0614002b
MG
2569 skb->pfmemalloc = true;
2570}
2571
131ea667 2572/**
e227867f 2573 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2574 * @frag: the paged fragment
2575 *
2576 * Returns the &struct page associated with @frag.
2577 */
2578static inline struct page *skb_frag_page(const skb_frag_t *frag)
2579{
a8605c60 2580 return frag->page.p;
131ea667
IC
2581}
2582
2583/**
2584 * __skb_frag_ref - take an addition reference on a paged fragment.
2585 * @frag: the paged fragment
2586 *
2587 * Takes an additional reference on the paged fragment @frag.
2588 */
2589static inline void __skb_frag_ref(skb_frag_t *frag)
2590{
2591 get_page(skb_frag_page(frag));
2592}
2593
2594/**
2595 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2596 * @skb: the buffer
2597 * @f: the fragment offset.
2598 *
2599 * Takes an additional reference on the @f'th paged fragment of @skb.
2600 */
2601static inline void skb_frag_ref(struct sk_buff *skb, int f)
2602{
2603 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2604}
2605
2606/**
2607 * __skb_frag_unref - release a reference on a paged fragment.
2608 * @frag: the paged fragment
2609 *
2610 * Releases a reference on the paged fragment @frag.
2611 */
2612static inline void __skb_frag_unref(skb_frag_t *frag)
2613{
2614 put_page(skb_frag_page(frag));
2615}
2616
2617/**
2618 * skb_frag_unref - release a reference on a paged fragment of an skb.
2619 * @skb: the buffer
2620 * @f: the fragment offset
2621 *
2622 * Releases a reference on the @f'th paged fragment of @skb.
2623 */
2624static inline void skb_frag_unref(struct sk_buff *skb, int f)
2625{
2626 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2627}
2628
2629/**
2630 * skb_frag_address - gets the address of the data contained in a paged fragment
2631 * @frag: the paged fragment buffer
2632 *
2633 * Returns the address of the data within @frag. The page must already
2634 * be mapped.
2635 */
2636static inline void *skb_frag_address(const skb_frag_t *frag)
2637{
2638 return page_address(skb_frag_page(frag)) + frag->page_offset;
2639}
2640
2641/**
2642 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2643 * @frag: the paged fragment buffer
2644 *
2645 * Returns the address of the data within @frag. Checks that the page
2646 * is mapped and returns %NULL otherwise.
2647 */
2648static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2649{
2650 void *ptr = page_address(skb_frag_page(frag));
2651 if (unlikely(!ptr))
2652 return NULL;
2653
2654 return ptr + frag->page_offset;
2655}
2656
2657/**
2658 * __skb_frag_set_page - sets the page contained in a paged fragment
2659 * @frag: the paged fragment
2660 * @page: the page to set
2661 *
2662 * Sets the fragment @frag to contain @page.
2663 */
2664static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2665{
a8605c60 2666 frag->page.p = page;
131ea667
IC
2667}
2668
2669/**
2670 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2671 * @skb: the buffer
2672 * @f: the fragment offset
2673 * @page: the page to set
2674 *
2675 * Sets the @f'th fragment of @skb to contain @page.
2676 */
2677static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2678 struct page *page)
2679{
2680 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2681}
2682
400dfd3a
ED
2683bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2684
131ea667
IC
2685/**
2686 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2687 * @dev: the device to map the fragment to
131ea667
IC
2688 * @frag: the paged fragment to map
2689 * @offset: the offset within the fragment (starting at the
2690 * fragment's own offset)
2691 * @size: the number of bytes to map
f83347df 2692 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2693 *
2694 * Maps the page associated with @frag to @device.
2695 */
2696static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2697 const skb_frag_t *frag,
2698 size_t offset, size_t size,
2699 enum dma_data_direction dir)
2700{
2701 return dma_map_page(dev, skb_frag_page(frag),
2702 frag->page_offset + offset, size, dir);
2703}
2704
117632e6
ED
2705static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2706 gfp_t gfp_mask)
2707{
2708 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2709}
2710
bad93e9d
OP
2711
2712static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2713 gfp_t gfp_mask)
2714{
2715 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2716}
2717
2718
334a8132
PM
2719/**
2720 * skb_clone_writable - is the header of a clone writable
2721 * @skb: buffer to check
2722 * @len: length up to which to write
2723 *
2724 * Returns true if modifying the header part of the cloned buffer
2725 * does not requires the data to be copied.
2726 */
05bdd2f1 2727static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2728{
2729 return !skb_header_cloned(skb) &&
2730 skb_headroom(skb) + len <= skb->hdr_len;
2731}
2732
3697649f
DB
2733static inline int skb_try_make_writable(struct sk_buff *skb,
2734 unsigned int write_len)
2735{
2736 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2737 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2738}
2739
d9cc2048
HX
2740static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2741 int cloned)
2742{
2743 int delta = 0;
2744
d9cc2048
HX
2745 if (headroom > skb_headroom(skb))
2746 delta = headroom - skb_headroom(skb);
2747
2748 if (delta || cloned)
2749 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2750 GFP_ATOMIC);
2751 return 0;
2752}
2753
1da177e4
LT
2754/**
2755 * skb_cow - copy header of skb when it is required
2756 * @skb: buffer to cow
2757 * @headroom: needed headroom
2758 *
2759 * If the skb passed lacks sufficient headroom or its data part
2760 * is shared, data is reallocated. If reallocation fails, an error
2761 * is returned and original skb is not changed.
2762 *
2763 * The result is skb with writable area skb->head...skb->tail
2764 * and at least @headroom of space at head.
2765 */
2766static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2767{
d9cc2048
HX
2768 return __skb_cow(skb, headroom, skb_cloned(skb));
2769}
1da177e4 2770
d9cc2048
HX
2771/**
2772 * skb_cow_head - skb_cow but only making the head writable
2773 * @skb: buffer to cow
2774 * @headroom: needed headroom
2775 *
2776 * This function is identical to skb_cow except that we replace the
2777 * skb_cloned check by skb_header_cloned. It should be used when
2778 * you only need to push on some header and do not need to modify
2779 * the data.
2780 */
2781static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2782{
2783 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2784}
2785
2786/**
2787 * skb_padto - pad an skbuff up to a minimal size
2788 * @skb: buffer to pad
2789 * @len: minimal length
2790 *
2791 * Pads up a buffer to ensure the trailing bytes exist and are
2792 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2793 * is untouched. Otherwise it is extended. Returns zero on
2794 * success. The skb is freed on error.
1da177e4 2795 */
5b057c6b 2796static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2797{
2798 unsigned int size = skb->len;
2799 if (likely(size >= len))
5b057c6b 2800 return 0;
987c402a 2801 return skb_pad(skb, len - size);
1da177e4
LT
2802}
2803
9c0c1124
AD
2804/**
2805 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2806 * @skb: buffer to pad
2807 * @len: minimal length
2808 *
2809 * Pads up a buffer to ensure the trailing bytes exist and are
2810 * blanked. If the buffer already contains sufficient data it
2811 * is untouched. Otherwise it is extended. Returns zero on
2812 * success. The skb is freed on error.
2813 */
2814static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2815{
2816 unsigned int size = skb->len;
2817
2818 if (unlikely(size < len)) {
2819 len -= size;
2820 if (skb_pad(skb, len))
2821 return -ENOMEM;
2822 __skb_put(skb, len);
2823 }
2824 return 0;
2825}
2826
1da177e4 2827static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2828 struct iov_iter *from, int copy)
1da177e4
LT
2829{
2830 const int off = skb->len;
2831
2832 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 2833 __wsum csum = 0;
15e6cb46
AV
2834 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2835 &csum, from)) {
1da177e4
LT
2836 skb->csum = csum_block_add(skb->csum, csum, off);
2837 return 0;
2838 }
15e6cb46 2839 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
2840 return 0;
2841
2842 __skb_trim(skb, off);
2843 return -EFAULT;
2844}
2845
38ba0a65
ED
2846static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2847 const struct page *page, int off)
1da177e4
LT
2848{
2849 if (i) {
9e903e08 2850 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2851
ea2ab693 2852 return page == skb_frag_page(frag) &&
9e903e08 2853 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2854 }
38ba0a65 2855 return false;
1da177e4
LT
2856}
2857
364c6bad
HX
2858static inline int __skb_linearize(struct sk_buff *skb)
2859{
2860 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2861}
2862
1da177e4
LT
2863/**
2864 * skb_linearize - convert paged skb to linear one
2865 * @skb: buffer to linarize
1da177e4
LT
2866 *
2867 * If there is no free memory -ENOMEM is returned, otherwise zero
2868 * is returned and the old skb data released.
2869 */
364c6bad
HX
2870static inline int skb_linearize(struct sk_buff *skb)
2871{
2872 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2873}
2874
cef401de
ED
2875/**
2876 * skb_has_shared_frag - can any frag be overwritten
2877 * @skb: buffer to test
2878 *
2879 * Return true if the skb has at least one frag that might be modified
2880 * by an external entity (as in vmsplice()/sendfile())
2881 */
2882static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2883{
c9af6db4
PS
2884 return skb_is_nonlinear(skb) &&
2885 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2886}
2887
364c6bad
HX
2888/**
2889 * skb_linearize_cow - make sure skb is linear and writable
2890 * @skb: buffer to process
2891 *
2892 * If there is no free memory -ENOMEM is returned, otherwise zero
2893 * is returned and the old skb data released.
2894 */
2895static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2896{
364c6bad
HX
2897 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2898 __skb_linearize(skb) : 0;
1da177e4
LT
2899}
2900
479ffccc
DB
2901static __always_inline void
2902__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2903 unsigned int off)
2904{
2905 if (skb->ip_summed == CHECKSUM_COMPLETE)
2906 skb->csum = csum_block_sub(skb->csum,
2907 csum_partial(start, len, 0), off);
2908 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2909 skb_checksum_start_offset(skb) < 0)
2910 skb->ip_summed = CHECKSUM_NONE;
2911}
2912
1da177e4
LT
2913/**
2914 * skb_postpull_rcsum - update checksum for received skb after pull
2915 * @skb: buffer to update
2916 * @start: start of data before pull
2917 * @len: length of data pulled
2918 *
2919 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2920 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2921 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 2922 */
1da177e4 2923static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2924 const void *start, unsigned int len)
1da177e4 2925{
479ffccc 2926 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
2927}
2928
479ffccc
DB
2929static __always_inline void
2930__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2931 unsigned int off)
2932{
2933 if (skb->ip_summed == CHECKSUM_COMPLETE)
2934 skb->csum = csum_block_add(skb->csum,
2935 csum_partial(start, len, 0), off);
2936}
cbb042f9 2937
479ffccc
DB
2938/**
2939 * skb_postpush_rcsum - update checksum for received skb after push
2940 * @skb: buffer to update
2941 * @start: start of data after push
2942 * @len: length of data pushed
2943 *
2944 * After doing a push on a received packet, you need to call this to
2945 * update the CHECKSUM_COMPLETE checksum.
2946 */
f8ffad69
DB
2947static inline void skb_postpush_rcsum(struct sk_buff *skb,
2948 const void *start, unsigned int len)
2949{
479ffccc 2950 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
2951}
2952
479ffccc
DB
2953unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2954
82a31b92
WC
2955/**
2956 * skb_push_rcsum - push skb and update receive checksum
2957 * @skb: buffer to update
2958 * @len: length of data pulled
2959 *
2960 * This function performs an skb_push on the packet and updates
2961 * the CHECKSUM_COMPLETE checksum. It should be used on
2962 * receive path processing instead of skb_push unless you know
2963 * that the checksum difference is zero (e.g., a valid IP header)
2964 * or you are setting ip_summed to CHECKSUM_NONE.
2965 */
2966static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2967 unsigned int len)
2968{
2969 skb_push(skb, len);
2970 skb_postpush_rcsum(skb, skb->data, len);
2971 return skb->data;
2972}
2973
7ce5a27f
DM
2974/**
2975 * pskb_trim_rcsum - trim received skb and update checksum
2976 * @skb: buffer to trim
2977 * @len: new length
2978 *
2979 * This is exactly the same as pskb_trim except that it ensures the
2980 * checksum of received packets are still valid after the operation.
2981 */
2982
2983static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2984{
2985 if (likely(len >= skb->len))
2986 return 0;
2987 if (skb->ip_summed == CHECKSUM_COMPLETE)
2988 skb->ip_summed = CHECKSUM_NONE;
2989 return __pskb_trim(skb, len);
2990}
2991
5293efe6
DB
2992static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2993{
2994 if (skb->ip_summed == CHECKSUM_COMPLETE)
2995 skb->ip_summed = CHECKSUM_NONE;
2996 __skb_trim(skb, len);
2997 return 0;
2998}
2999
3000static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3001{
3002 if (skb->ip_summed == CHECKSUM_COMPLETE)
3003 skb->ip_summed = CHECKSUM_NONE;
3004 return __skb_grow(skb, len);
3005}
3006
1da177e4
LT
3007#define skb_queue_walk(queue, skb) \
3008 for (skb = (queue)->next; \
a1e4891f 3009 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3010 skb = skb->next)
3011
46f8914e
JC
3012#define skb_queue_walk_safe(queue, skb, tmp) \
3013 for (skb = (queue)->next, tmp = skb->next; \
3014 skb != (struct sk_buff *)(queue); \
3015 skb = tmp, tmp = skb->next)
3016
1164f52a 3017#define skb_queue_walk_from(queue, skb) \
a1e4891f 3018 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3019 skb = skb->next)
3020
3021#define skb_queue_walk_from_safe(queue, skb, tmp) \
3022 for (tmp = skb->next; \
3023 skb != (struct sk_buff *)(queue); \
3024 skb = tmp, tmp = skb->next)
3025
300ce174
SH
3026#define skb_queue_reverse_walk(queue, skb) \
3027 for (skb = (queue)->prev; \
a1e4891f 3028 skb != (struct sk_buff *)(queue); \
300ce174
SH
3029 skb = skb->prev)
3030
686a2955
DM
3031#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3032 for (skb = (queue)->prev, tmp = skb->prev; \
3033 skb != (struct sk_buff *)(queue); \
3034 skb = tmp, tmp = skb->prev)
3035
3036#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3037 for (tmp = skb->prev; \
3038 skb != (struct sk_buff *)(queue); \
3039 skb = tmp, tmp = skb->prev)
1da177e4 3040
21dc3301 3041static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3042{
3043 return skb_shinfo(skb)->frag_list != NULL;
3044}
3045
3046static inline void skb_frag_list_init(struct sk_buff *skb)
3047{
3048 skb_shinfo(skb)->frag_list = NULL;
3049}
3050
ee039871
DM
3051#define skb_walk_frags(skb, iter) \
3052 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3053
ea3793ee
RW
3054
3055int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3056 const struct sk_buff *skb);
3057struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3058 void (*destructor)(struct sock *sk,
3059 struct sk_buff *skb),
ea3793ee
RW
3060 int *peeked, int *off, int *err,
3061 struct sk_buff **last);
7965bd4d 3062struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3063 void (*destructor)(struct sock *sk,
3064 struct sk_buff *skb),
7965bd4d
JP
3065 int *peeked, int *off, int *err);
3066struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3067 int *err);
3068unsigned int datagram_poll(struct file *file, struct socket *sock,
3069 struct poll_table_struct *wait);
c0371da6
AV
3070int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3071 struct iov_iter *to, int size);
51f3d02b
DM
3072static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3073 struct msghdr *msg, int size)
3074{
e5a4b0bb 3075 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3076}
e5a4b0bb
AV
3077int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3078 struct msghdr *msg);
3a654f97
AV
3079int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3080 struct iov_iter *from, int len);
3a654f97 3081int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3082void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3083void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3084static inline void skb_free_datagram_locked(struct sock *sk,
3085 struct sk_buff *skb)
3086{
3087 __skb_free_datagram_locked(sk, skb, 0);
3088}
7965bd4d 3089int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3090int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3091int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3092__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3093 int len, __wsum csum);
a60e3cc7 3094int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3095 struct pipe_inode_info *pipe, unsigned int len,
25869262 3096 unsigned int flags);
7965bd4d 3097void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3098unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3099int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3100 int len, int hlen);
7965bd4d
JP
3101void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3102int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3103void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3104unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3105bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 3106struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3107struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3108int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3109int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3110int skb_vlan_pop(struct sk_buff *skb);
3111int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3112struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3113 gfp_t gfp);
20380731 3114
6ce8e9ce
AV
3115static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3116{
21226abb 3117 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
3118}
3119
7eab8d9e
AV
3120static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3121{
e5a4b0bb 3122 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3123}
3124
2817a336
DB
3125struct skb_checksum_ops {
3126 __wsum (*update)(const void *mem, int len, __wsum wsum);
3127 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3128};
3129
3130__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3131 __wsum csum, const struct skb_checksum_ops *ops);
3132__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3133 __wsum csum);
3134
1e98a0f0
ED
3135static inline void * __must_check
3136__skb_header_pointer(const struct sk_buff *skb, int offset,
3137 int len, void *data, int hlen, void *buffer)
1da177e4 3138{
55820ee2 3139 if (hlen - offset >= len)
690e36e7 3140 return data + offset;
1da177e4 3141
690e36e7
DM
3142 if (!skb ||
3143 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3144 return NULL;
3145
3146 return buffer;
3147}
3148
1e98a0f0
ED
3149static inline void * __must_check
3150skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3151{
3152 return __skb_header_pointer(skb, offset, len, skb->data,
3153 skb_headlen(skb), buffer);
3154}
3155
4262e5cc
DB
3156/**
3157 * skb_needs_linearize - check if we need to linearize a given skb
3158 * depending on the given device features.
3159 * @skb: socket buffer to check
3160 * @features: net device features
3161 *
3162 * Returns true if either:
3163 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3164 * 2. skb is fragmented and the device does not support SG.
3165 */
3166static inline bool skb_needs_linearize(struct sk_buff *skb,
3167 netdev_features_t features)
3168{
3169 return skb_is_nonlinear(skb) &&
3170 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3171 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3172}
3173
d626f62b
ACM
3174static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3175 void *to,
3176 const unsigned int len)
3177{
3178 memcpy(to, skb->data, len);
3179}
3180
3181static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3182 const int offset, void *to,
3183 const unsigned int len)
3184{
3185 memcpy(to, skb->data + offset, len);
3186}
3187
27d7ff46
ACM
3188static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3189 const void *from,
3190 const unsigned int len)
3191{
3192 memcpy(skb->data, from, len);
3193}
3194
3195static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3196 const int offset,
3197 const void *from,
3198 const unsigned int len)
3199{
3200 memcpy(skb->data + offset, from, len);
3201}
3202
7965bd4d 3203void skb_init(void);
1da177e4 3204
ac45f602
PO
3205static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3206{
3207 return skb->tstamp;
3208}
3209
a61bbcf2
PM
3210/**
3211 * skb_get_timestamp - get timestamp from a skb
3212 * @skb: skb to get stamp from
3213 * @stamp: pointer to struct timeval to store stamp in
3214 *
3215 * Timestamps are stored in the skb as offsets to a base timestamp.
3216 * This function converts the offset back to a struct timeval and stores
3217 * it in stamp.
3218 */
ac45f602
PO
3219static inline void skb_get_timestamp(const struct sk_buff *skb,
3220 struct timeval *stamp)
a61bbcf2 3221{
b7aa0bf7 3222 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3223}
3224
ac45f602
PO
3225static inline void skb_get_timestampns(const struct sk_buff *skb,
3226 struct timespec *stamp)
3227{
3228 *stamp = ktime_to_timespec(skb->tstamp);
3229}
3230
b7aa0bf7 3231static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3232{
b7aa0bf7 3233 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3234}
3235
164891aa
SH
3236static inline ktime_t net_timedelta(ktime_t t)
3237{
3238 return ktime_sub(ktime_get_real(), t);
3239}
3240
b9ce204f
IJ
3241static inline ktime_t net_invalid_timestamp(void)
3242{
8b0e1953 3243 return 0;
b9ce204f 3244}
a61bbcf2 3245
62bccb8c
AD
3246struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3247
c1f19b51
RC
3248#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3249
7965bd4d
JP
3250void skb_clone_tx_timestamp(struct sk_buff *skb);
3251bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3252
3253#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3254
3255static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3256{
3257}
3258
3259static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3260{
3261 return false;
3262}
3263
3264#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3265
3266/**
3267 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3268 *
da92b194
RC
3269 * PHY drivers may accept clones of transmitted packets for
3270 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3271 * must call this function to return the skb back to the stack with a
3272 * timestamp.
da92b194 3273 *
c1f19b51 3274 * @skb: clone of the the original outgoing packet
7a76a021 3275 * @hwtstamps: hardware time stamps
c1f19b51
RC
3276 *
3277 */
3278void skb_complete_tx_timestamp(struct sk_buff *skb,
3279 struct skb_shared_hwtstamps *hwtstamps);
3280
e7fd2885
WB
3281void __skb_tstamp_tx(struct sk_buff *orig_skb,
3282 struct skb_shared_hwtstamps *hwtstamps,
3283 struct sock *sk, int tstype);
3284
ac45f602
PO
3285/**
3286 * skb_tstamp_tx - queue clone of skb with send time stamps
3287 * @orig_skb: the original outgoing packet
3288 * @hwtstamps: hardware time stamps, may be NULL if not available
3289 *
3290 * If the skb has a socket associated, then this function clones the
3291 * skb (thus sharing the actual data and optional structures), stores
3292 * the optional hardware time stamping information (if non NULL) or
3293 * generates a software time stamp (otherwise), then queues the clone
3294 * to the error queue of the socket. Errors are silently ignored.
3295 */
7965bd4d
JP
3296void skb_tstamp_tx(struct sk_buff *orig_skb,
3297 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3298
4507a715
RC
3299static inline void sw_tx_timestamp(struct sk_buff *skb)
3300{
2244d07b
OH
3301 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3302 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3303 skb_tstamp_tx(skb, NULL);
3304}
3305
3306/**
3307 * skb_tx_timestamp() - Driver hook for transmit timestamping
3308 *
3309 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3310 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3311 *
73409f3b
DM
3312 * Specifically, one should make absolutely sure that this function is
3313 * called before TX completion of this packet can trigger. Otherwise
3314 * the packet could potentially already be freed.
3315 *
4507a715
RC
3316 * @skb: A socket buffer.
3317 */
3318static inline void skb_tx_timestamp(struct sk_buff *skb)
3319{
c1f19b51 3320 skb_clone_tx_timestamp(skb);
4507a715
RC
3321 sw_tx_timestamp(skb);
3322}
3323
6e3e939f
JB
3324/**
3325 * skb_complete_wifi_ack - deliver skb with wifi status
3326 *
3327 * @skb: the original outgoing packet
3328 * @acked: ack status
3329 *
3330 */
3331void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3332
7965bd4d
JP
3333__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3334__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3335
60476372
HX
3336static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3337{
6edec0e6
TH
3338 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3339 skb->csum_valid ||
3340 (skb->ip_summed == CHECKSUM_PARTIAL &&
3341 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3342}
3343
fb286bb2
HX
3344/**
3345 * skb_checksum_complete - Calculate checksum of an entire packet
3346 * @skb: packet to process
3347 *
3348 * This function calculates the checksum over the entire packet plus
3349 * the value of skb->csum. The latter can be used to supply the
3350 * checksum of a pseudo header as used by TCP/UDP. It returns the
3351 * checksum.
3352 *
3353 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3354 * this function can be used to verify that checksum on received
3355 * packets. In that case the function should return zero if the
3356 * checksum is correct. In particular, this function will return zero
3357 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3358 * hardware has already verified the correctness of the checksum.
3359 */
4381ca3c 3360static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3361{
60476372
HX
3362 return skb_csum_unnecessary(skb) ?
3363 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3364}
3365
77cffe23
TH
3366static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3367{
3368 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3369 if (skb->csum_level == 0)
3370 skb->ip_summed = CHECKSUM_NONE;
3371 else
3372 skb->csum_level--;
3373 }
3374}
3375
3376static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3377{
3378 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3379 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3380 skb->csum_level++;
3381 } else if (skb->ip_summed == CHECKSUM_NONE) {
3382 skb->ip_summed = CHECKSUM_UNNECESSARY;
3383 skb->csum_level = 0;
3384 }
3385}
3386
5a212329
TH
3387static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3388{
3389 /* Mark current checksum as bad (typically called from GRO
3390 * path). In the case that ip_summed is CHECKSUM_NONE
3391 * this must be the first checksum encountered in the packet.
3392 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3393 * checksum after the last one validated. For UDP, a zero
3394 * checksum can not be marked as bad.
3395 */
3396
3397 if (skb->ip_summed == CHECKSUM_NONE ||
3398 skb->ip_summed == CHECKSUM_UNNECESSARY)
3399 skb->csum_bad = 1;
3400}
3401
76ba0aae
TH
3402/* Check if we need to perform checksum complete validation.
3403 *
3404 * Returns true if checksum complete is needed, false otherwise
3405 * (either checksum is unnecessary or zero checksum is allowed).
3406 */
3407static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3408 bool zero_okay,
3409 __sum16 check)
3410{
5d0c2b95
TH
3411 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3412 skb->csum_valid = 1;
77cffe23 3413 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3414 return false;
3415 }
3416
3417 return true;
3418}
3419
3420/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3421 * in checksum_init.
3422 */
3423#define CHECKSUM_BREAK 76
3424
4e18b9ad
TH
3425/* Unset checksum-complete
3426 *
3427 * Unset checksum complete can be done when packet is being modified
3428 * (uncompressed for instance) and checksum-complete value is
3429 * invalidated.
3430 */
3431static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3432{
3433 if (skb->ip_summed == CHECKSUM_COMPLETE)
3434 skb->ip_summed = CHECKSUM_NONE;
3435}
3436
76ba0aae
TH
3437/* Validate (init) checksum based on checksum complete.
3438 *
3439 * Return values:
3440 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3441 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3442 * checksum is stored in skb->csum for use in __skb_checksum_complete
3443 * non-zero: value of invalid checksum
3444 *
3445 */
3446static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3447 bool complete,
3448 __wsum psum)
3449{
3450 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3451 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3452 skb->csum_valid = 1;
76ba0aae
TH
3453 return 0;
3454 }
5a212329
TH
3455 } else if (skb->csum_bad) {
3456 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3457 return (__force __sum16)1;
76ba0aae
TH
3458 }
3459
3460 skb->csum = psum;
3461
5d0c2b95
TH
3462 if (complete || skb->len <= CHECKSUM_BREAK) {
3463 __sum16 csum;
3464
3465 csum = __skb_checksum_complete(skb);
3466 skb->csum_valid = !csum;
3467 return csum;
3468 }
76ba0aae
TH
3469
3470 return 0;
3471}
3472
3473static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3474{
3475 return 0;
3476}
3477
3478/* Perform checksum validate (init). Note that this is a macro since we only
3479 * want to calculate the pseudo header which is an input function if necessary.
3480 * First we try to validate without any computation (checksum unnecessary) and
3481 * then calculate based on checksum complete calling the function to compute
3482 * pseudo header.
3483 *
3484 * Return values:
3485 * 0: checksum is validated or try to in skb_checksum_complete
3486 * non-zero: value of invalid checksum
3487 */
3488#define __skb_checksum_validate(skb, proto, complete, \
3489 zero_okay, check, compute_pseudo) \
3490({ \
3491 __sum16 __ret = 0; \
5d0c2b95 3492 skb->csum_valid = 0; \
76ba0aae
TH
3493 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3494 __ret = __skb_checksum_validate_complete(skb, \
3495 complete, compute_pseudo(skb, proto)); \
3496 __ret; \
3497})
3498
3499#define skb_checksum_init(skb, proto, compute_pseudo) \
3500 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3501
3502#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3503 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3504
3505#define skb_checksum_validate(skb, proto, compute_pseudo) \
3506 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3507
3508#define skb_checksum_validate_zero_check(skb, proto, check, \
3509 compute_pseudo) \
096a4cfa 3510 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3511
3512#define skb_checksum_simple_validate(skb) \
3513 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3514
d96535a1
TH
3515static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3516{
3517 return (skb->ip_summed == CHECKSUM_NONE &&
3518 skb->csum_valid && !skb->csum_bad);
3519}
3520
3521static inline void __skb_checksum_convert(struct sk_buff *skb,
3522 __sum16 check, __wsum pseudo)
3523{
3524 skb->csum = ~pseudo;
3525 skb->ip_summed = CHECKSUM_COMPLETE;
3526}
3527
3528#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3529do { \
3530 if (__skb_checksum_convert_check(skb)) \
3531 __skb_checksum_convert(skb, check, \
3532 compute_pseudo(skb, proto)); \
3533} while (0)
3534
15e2396d
TH
3535static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3536 u16 start, u16 offset)
3537{
3538 skb->ip_summed = CHECKSUM_PARTIAL;
3539 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3540 skb->csum_offset = offset - start;
3541}
3542
dcdc8994
TH
3543/* Update skbuf and packet to reflect the remote checksum offload operation.
3544 * When called, ptr indicates the starting point for skb->csum when
3545 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3546 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3547 */
3548static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3549 int start, int offset, bool nopartial)
dcdc8994
TH
3550{
3551 __wsum delta;
3552
15e2396d
TH
3553 if (!nopartial) {
3554 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3555 return;
3556 }
3557
dcdc8994
TH
3558 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3559 __skb_checksum_complete(skb);
3560 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3561 }
3562
3563 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3564
3565 /* Adjust skb->csum since we changed the packet */
3566 skb->csum = csum_add(skb->csum, delta);
3567}
3568
cb9c6836
FW
3569static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3570{
3571#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 3572 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
3573#else
3574 return NULL;
3575#endif
3576}
3577
5f79e0f9 3578#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3579void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3580static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3581{
3582 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3583 nf_conntrack_destroy(nfct);
1da177e4
LT
3584}
3585static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3586{
3587 if (nfct)
3588 atomic_inc(&nfct->use);
3589}
2fc72c7b 3590#endif
34666d46 3591#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3592static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3593{
3594 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3595 kfree(nf_bridge);
3596}
3597static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3598{
3599 if (nf_bridge)
3600 atomic_inc(&nf_bridge->use);
3601}
3602#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3603static inline void nf_reset(struct sk_buff *skb)
3604{
5f79e0f9 3605#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3606 nf_conntrack_put(skb_nfct(skb));
3607 skb->_nfct = 0;
2fc72c7b 3608#endif
34666d46 3609#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3610 nf_bridge_put(skb->nf_bridge);
3611 skb->nf_bridge = NULL;
3612#endif
3613}
3614
124dff01
PM
3615static inline void nf_reset_trace(struct sk_buff *skb)
3616{
478b360a 3617#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3618 skb->nf_trace = 0;
3619#endif
a193a4ab
PM
3620}
3621
edda553c 3622/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3623static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3624 bool copy)
edda553c 3625{
5f79e0f9 3626#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3627 dst->_nfct = src->_nfct;
3628 nf_conntrack_get(skb_nfct(src));
2fc72c7b 3629#endif
34666d46 3630#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3631 dst->nf_bridge = src->nf_bridge;
3632 nf_bridge_get(src->nf_bridge);
3633#endif
478b360a 3634#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3635 if (copy)
3636 dst->nf_trace = src->nf_trace;
478b360a 3637#endif
edda553c
YK
3638}
3639
e7ac05f3
YK
3640static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3641{
e7ac05f3 3642#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 3643 nf_conntrack_put(skb_nfct(dst));
2fc72c7b 3644#endif
34666d46 3645#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3646 nf_bridge_put(dst->nf_bridge);
3647#endif
b1937227 3648 __nf_copy(dst, src, true);
e7ac05f3
YK
3649}
3650
984bc16c
JM
3651#ifdef CONFIG_NETWORK_SECMARK
3652static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3653{
3654 to->secmark = from->secmark;
3655}
3656
3657static inline void skb_init_secmark(struct sk_buff *skb)
3658{
3659 skb->secmark = 0;
3660}
3661#else
3662static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3663{ }
3664
3665static inline void skb_init_secmark(struct sk_buff *skb)
3666{ }
3667#endif
3668
574f7194
EB
3669static inline bool skb_irq_freeable(const struct sk_buff *skb)
3670{
3671 return !skb->destructor &&
3672#if IS_ENABLED(CONFIG_XFRM)
3673 !skb->sp &&
3674#endif
cb9c6836 3675 !skb_nfct(skb) &&
574f7194
EB
3676 !skb->_skb_refdst &&
3677 !skb_has_frag_list(skb);
3678}
3679
f25f4e44
PWJ
3680static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3681{
f25f4e44 3682 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3683}
3684
9247744e 3685static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3686{
4e3ab47a 3687 return skb->queue_mapping;
4e3ab47a
PE
3688}
3689
f25f4e44
PWJ
3690static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3691{
f25f4e44 3692 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3693}
3694
d5a9e24a
DM
3695static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3696{
3697 skb->queue_mapping = rx_queue + 1;
3698}
3699
9247744e 3700static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3701{
3702 return skb->queue_mapping - 1;
3703}
3704
9247744e 3705static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3706{
a02cec21 3707 return skb->queue_mapping != 0;
d5a9e24a
DM
3708}
3709
4ff06203
JA
3710static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3711{
3712 skb->dst_pending_confirm = val;
3713}
3714
3715static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3716{
3717 return skb->dst_pending_confirm != 0;
3718}
3719
def8b4fa
AD
3720static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3721{
0b3d8e08 3722#ifdef CONFIG_XFRM
def8b4fa 3723 return skb->sp;
def8b4fa 3724#else
def8b4fa 3725 return NULL;
def8b4fa 3726#endif
0b3d8e08 3727}
def8b4fa 3728
68c33163
PS
3729/* Keeps track of mac header offset relative to skb->head.
3730 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3731 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3732 * tunnel skb it points to outer mac header.
3733 * Keeps track of level of encapsulation of network headers.
3734 */
68c33163 3735struct skb_gso_cb {
802ab55a
AD
3736 union {
3737 int mac_offset;
3738 int data_offset;
3739 };
3347c960 3740 int encap_level;
76443456 3741 __wsum csum;
7e2b10c1 3742 __u16 csum_start;
68c33163 3743};
9207f9d4
KK
3744#define SKB_SGO_CB_OFFSET 32
3745#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3746
3747static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3748{
3749 return (skb_mac_header(inner_skb) - inner_skb->head) -
3750 SKB_GSO_CB(inner_skb)->mac_offset;
3751}
3752
1e2bd517
PS
3753static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3754{
3755 int new_headroom, headroom;
3756 int ret;
3757
3758 headroom = skb_headroom(skb);
3759 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3760 if (ret)
3761 return ret;
3762
3763 new_headroom = skb_headroom(skb);
3764 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3765 return 0;
3766}
3767
08b64fcc
AD
3768static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3769{
3770 /* Do not update partial checksums if remote checksum is enabled. */
3771 if (skb->remcsum_offload)
3772 return;
3773
3774 SKB_GSO_CB(skb)->csum = res;
3775 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3776}
3777
7e2b10c1
TH
3778/* Compute the checksum for a gso segment. First compute the checksum value
3779 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3780 * then add in skb->csum (checksum from csum_start to end of packet).
3781 * skb->csum and csum_start are then updated to reflect the checksum of the
3782 * resultant packet starting from the transport header-- the resultant checksum
3783 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3784 * header.
3785 */
3786static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3787{
76443456
AD
3788 unsigned char *csum_start = skb_transport_header(skb);
3789 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3790 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3791
76443456
AD
3792 SKB_GSO_CB(skb)->csum = res;
3793 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3794
76443456 3795 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3796}
3797
bdcc0924 3798static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3799{
3800 return skb_shinfo(skb)->gso_size;
3801}
3802
36a8f39e 3803/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3804static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3805{
3806 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3807}
3808
5293efe6
DB
3809static inline void skb_gso_reset(struct sk_buff *skb)
3810{
3811 skb_shinfo(skb)->gso_size = 0;
3812 skb_shinfo(skb)->gso_segs = 0;
3813 skb_shinfo(skb)->gso_type = 0;
3814}
3815
7965bd4d 3816void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3817
3818static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3819{
3820 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3821 * wanted then gso_type will be set. */
05bdd2f1
ED
3822 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3823
b78462eb
AD
3824 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3825 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3826 __skb_warn_lro_forwarding(skb);
3827 return true;
3828 }
3829 return false;
3830}
3831
35fc92a9
HX
3832static inline void skb_forward_csum(struct sk_buff *skb)
3833{
3834 /* Unfortunately we don't support this one. Any brave souls? */
3835 if (skb->ip_summed == CHECKSUM_COMPLETE)
3836 skb->ip_summed = CHECKSUM_NONE;
3837}
3838
bc8acf2c
ED
3839/**
3840 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3841 * @skb: skb to check
3842 *
3843 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3844 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3845 * use this helper, to document places where we make this assertion.
3846 */
05bdd2f1 3847static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3848{
3849#ifdef DEBUG
3850 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3851#endif
3852}
3853
f35d9d8a 3854bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3855
ed1f50c3 3856int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3857struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3858 unsigned int transport_len,
3859 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3860
3a7c1ee4
AD
3861/**
3862 * skb_head_is_locked - Determine if the skb->head is locked down
3863 * @skb: skb to check
3864 *
3865 * The head on skbs build around a head frag can be removed if they are
3866 * not cloned. This function returns true if the skb head is locked down
3867 * due to either being allocated via kmalloc, or by being a clone with
3868 * multiple references to the head.
3869 */
3870static inline bool skb_head_is_locked(const struct sk_buff *skb)
3871{
3872 return !skb->head_frag || skb_cloned(skb);
3873}
fe6cc55f
FW
3874
3875/**
3876 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3877 *
3878 * @skb: GSO skb
3879 *
3880 * skb_gso_network_seglen is used to determine the real size of the
3881 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3882 *
3883 * The MAC/L2 header is not accounted for.
3884 */
3885static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3886{
3887 unsigned int hdr_len = skb_transport_header(skb) -
3888 skb_network_header(skb);
3889 return hdr_len + skb_gso_transport_seglen(skb);
3890}
ee122c79 3891
179bc67f
EC
3892/* Local Checksum Offload.
3893 * Compute outer checksum based on the assumption that the
3894 * inner checksum will be offloaded later.
e8ae7b00
EC
3895 * See Documentation/networking/checksum-offloads.txt for
3896 * explanation of how this works.
179bc67f
EC
3897 * Fill in outer checksum adjustment (e.g. with sum of outer
3898 * pseudo-header) before calling.
3899 * Also ensure that inner checksum is in linear data area.
3900 */
3901static inline __wsum lco_csum(struct sk_buff *skb)
3902{
9e74a6da
AD
3903 unsigned char *csum_start = skb_checksum_start(skb);
3904 unsigned char *l4_hdr = skb_transport_header(skb);
3905 __wsum partial;
179bc67f
EC
3906
3907 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3908 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3909 skb->csum_offset));
3910
179bc67f 3911 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3912 * adjustment filled in by caller) and return result.
179bc67f 3913 */
9e74a6da 3914 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3915}
3916
1da177e4
LT
3917#endif /* __KERNEL__ */
3918#endif /* _LINUX_SKBUFF_H */