]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
regulator: ab8500: Remove AB8505 USB regulator
[mirror_ubuntu-bionic-kernel.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4
LT
17#include <linux/kernel.h>
18#include <linux/compiler.h>
19#include <linux/time.h>
187f1882 20#include <linux/bug.h>
1da177e4 21#include <linux/cache.h>
56b17425 22#include <linux/rbtree.h>
51f3d02b 23#include <linux/socket.h>
c1d1b437 24#include <linux/refcount.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.
b4759dcd 112 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
78ea85f1
DB
123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
b4759dcd
DC
130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
134 *
135 * CHECKSUM_PARTIAL:
136 *
6edec0e6
TH
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
78ea85f1 146 *
7a6ae71b
TH
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
149 *
150 * CHECKSUM_PARTIAL:
151 *
7a6ae71b 152 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 153 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) 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
dba00306
DC
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
78ea85f1
DB
217 */
218
60476372 219/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
1da177e4 224
77cffe23
TH
225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
0bec8c88 228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 229#define SKB_WITH_OVERHEAD(X) \
deea84b0 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
87fb4b7b
ED
236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
1da177e4 241struct net_device;
716ea3a7 242struct scatterlist;
9c55e01c 243struct pipe_inode_info;
a8f820aa 244struct iov_iter;
fd11a83d 245struct napi_struct;
1da177e4 246
5f79e0f9 247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
248struct nf_conntrack {
249 atomic_t use;
1da177e4 250};
5f79e0f9 251#endif
1da177e4 252
34666d46 253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 254struct nf_bridge_info {
53869ceb 255 refcount_t use;
3eaf4025
FW
256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
7fb48c5b 260 } orig_proto:8;
72b1e5e4
FW
261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
411ffb4f 264 __u16 frag_max_size;
bf1ac5ca 265 struct net_device *physindev;
63cdbc06
FW
266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
7fb48c5b 269 union {
72b1e5e4 270 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
72b1e5e4
FW
273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
72b31f72 279 };
1da177e4
LT
280};
281#endif
282
1da177e4
LT
283struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290};
291
292struct sk_buff;
293
9d4dde52
IC
294/* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
a715dea3 300 */
9d4dde52 301#if (65536/PAGE_SIZE + 1) < 16
eec00954 302#define MAX_SKB_FRAGS 16UL
a715dea3 303#else
9d4dde52 304#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 305#endif
5f74f82e 306extern int sysctl_max_skb_frags;
1da177e4 307
3953c46c
MRL
308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
1da177e4
LT
313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
a8605c60
IC
316 struct {
317 struct page *p;
318 } page;
cb4dfe56 319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
320 __u32 page_offset;
321 __u32 size;
cb4dfe56
ED
322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
1da177e4
LT
326};
327
9e903e08
ED
328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
c613c209
WB
348static inline bool skb_frag_must_loop(struct page *p)
349{
350#if defined(CONFIG_HIGHMEM)
351 if (PageHighMem(p))
352 return true;
353#endif
354 return false;
355}
356
357/**
358 * skb_frag_foreach_page - loop over pages in a fragment
359 *
360 * @f: skb frag to operate on
361 * @f_off: offset from start of f->page.p
362 * @f_len: length from f_off to loop over
363 * @p: (temp var) current page
364 * @p_off: (temp var) offset from start of current page,
365 * non-zero only on first page.
366 * @p_len: (temp var) length in current page,
367 * < PAGE_SIZE only on first and last page.
368 * @copied: (temp var) length so far, excluding current p_len.
369 *
370 * A fragment can hold a compound page, in which case per-page
371 * operations, notably kmap_atomic, must be called for each
372 * regular page.
373 */
374#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
375 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
376 p_off = (f_off) & (PAGE_SIZE - 1), \
377 p_len = skb_frag_must_loop(p) ? \
378 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
379 copied = 0; \
380 copied < f_len; \
381 copied += p_len, p++, p_off = 0, \
382 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
383
ac45f602
PO
384#define HAVE_HW_TIME_STAMP
385
386/**
d3a21be8 387 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
388 * @hwtstamp: hardware time stamp transformed into duration
389 * since arbitrary point in time
ac45f602
PO
390 *
391 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 392 * skb->tstamp.
ac45f602
PO
393 *
394 * hwtstamps can only be compared against other hwtstamps from
395 * the same device.
396 *
397 * This structure is attached to packets as part of the
398 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
399 */
400struct skb_shared_hwtstamps {
401 ktime_t hwtstamp;
ac45f602
PO
402};
403
2244d07b
OH
404/* Definitions for tx_flags in struct skb_shared_info */
405enum {
406 /* generate hardware time stamp */
407 SKBTX_HW_TSTAMP = 1 << 0,
408
e7fd2885 409 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
410 SKBTX_SW_TSTAMP = 1 << 1,
411
412 /* device driver is going to provide hardware time stamp */
413 SKBTX_IN_PROGRESS = 1 << 2,
414
a6686f2f 415 /* device driver supports TX zero-copy buffers */
62b1a8ab 416 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
417
418 /* generate wifi status information (where possible) */
62b1a8ab 419 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
420
421 /* This indicates at least one fragment might be overwritten
422 * (as in vmsplice(), sendfile() ...)
423 * If we need to compute a TX checksum, we'll need to copy
424 * all frags to avoid possible bad checksum
425 */
426 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
427
428 /* generate software time stamp when entering packet scheduling */
429 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
430};
431
52267790 432#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
e1c8a607 433#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 434 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
435#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
436
a6686f2f
SM
437/*
438 * The callback notifies userspace to release buffers when skb DMA is done in
439 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
440 * The zerocopy_success argument is true if zero copy transmit occurred,
441 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
442 * The ctx field is used to track device context.
443 * The desc field is used to track userspace buffer index.
a6686f2f
SM
444 */
445struct ubuf_info {
e19d6763 446 void (*callback)(struct ubuf_info *, bool zerocopy_success);
4ab6c99d
WB
447 union {
448 struct {
449 unsigned long desc;
450 void *ctx;
451 };
452 struct {
453 u32 id;
454 u16 len;
455 u16 zerocopy:1;
456 u32 bytelen;
457 };
458 };
c1d1b437 459 refcount_t refcnt;
a91dbff5
WB
460
461 struct mmpin {
462 struct user_struct *user;
463 unsigned int num_pg;
464 } mmp;
ac45f602
PO
465};
466
52267790
WB
467#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
468
469struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
4ab6c99d
WB
470struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
471 struct ubuf_info *uarg);
52267790
WB
472
473static inline void sock_zerocopy_get(struct ubuf_info *uarg)
474{
c1d1b437 475 refcount_inc(&uarg->refcnt);
52267790
WB
476}
477
478void sock_zerocopy_put(struct ubuf_info *uarg);
479void sock_zerocopy_put_abort(struct ubuf_info *uarg);
480
481void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
482
483int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
484 struct msghdr *msg, int len,
485 struct ubuf_info *uarg);
486
1da177e4
LT
487/* This data is invariant across clones and lives at
488 * the end of the header data, ie. at skb->end.
489 */
490struct skb_shared_info {
de8f3a83
DB
491 __u8 __unused;
492 __u8 meta_len;
493 __u8 nr_frags;
9f42f126 494 __u8 tx_flags;
7967168c
HX
495 unsigned short gso_size;
496 /* Warning: this field is not always filled in (UFO)! */
497 unsigned short gso_segs;
1da177e4 498 struct sk_buff *frag_list;
ac45f602 499 struct skb_shared_hwtstamps hwtstamps;
7f564528 500 unsigned int gso_type;
09c2d251 501 u32 tskey;
ec7d2f2c
ED
502
503 /*
504 * Warning : all fields before dataref are cleared in __alloc_skb()
505 */
506 atomic_t dataref;
507
69e3c75f
JB
508 /* Intermediate layers must ensure that destructor_arg
509 * remains valid until skb destructor */
510 void * destructor_arg;
a6686f2f 511
fed66381
ED
512 /* must be last field, see pskb_expand_head() */
513 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
514};
515
516/* We divide dataref into two halves. The higher 16 bits hold references
517 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
518 * the entire skb->data. A clone of a headerless skb holds the length of
519 * the header in skb->hdr_len.
1da177e4
LT
520 *
521 * All users must obey the rule that the skb->data reference count must be
522 * greater than or equal to the payload reference count.
523 *
524 * Holding a reference to the payload part means that the user does not
525 * care about modifications to the header part of skb->data.
526 */
527#define SKB_DATAREF_SHIFT 16
528#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
529
d179cd12
DM
530
531enum {
c8753d55
VS
532 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
533 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
534 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
535};
536
7967168c
HX
537enum {
538 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
539
540 /* This indicates the skb is from an untrusted source. */
d9d30adf 541 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
542
543 /* This indicates the tcp segment has CWR set. */
d9d30adf 544 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 545
d9d30adf 546 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 547
d9d30adf 548 SKB_GSO_TCPV6 = 1 << 4,
68c33163 549
d9d30adf 550 SKB_GSO_FCOE = 1 << 5,
73136267 551
d9d30adf 552 SKB_GSO_GRE = 1 << 6,
0d89d203 553
d9d30adf 554 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 555
d9d30adf 556 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 557
d9d30adf 558 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 559
d9d30adf 560 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 561
d9d30adf 562 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 563
d9d30adf 564 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 565
d9d30adf 566 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 567
d9d30adf 568 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 569
d9d30adf 570 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
571
572 SKB_GSO_UDP = 1 << 16,
7967168c
HX
573};
574
2e07fa9c
ACM
575#if BITS_PER_LONG > 32
576#define NET_SKBUFF_DATA_USES_OFFSET 1
577#endif
578
579#ifdef NET_SKBUFF_DATA_USES_OFFSET
580typedef unsigned int sk_buff_data_t;
581#else
582typedef unsigned char *sk_buff_data_t;
583#endif
584
1da177e4
LT
585/**
586 * struct sk_buff - socket buffer
587 * @next: Next buffer in list
588 * @prev: Previous buffer in list
363ec392 589 * @tstamp: Time we arrived/left
56b17425 590 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 591 * @sk: Socket we are owned by
1da177e4 592 * @dev: Device we arrived on/are leaving by
d84e0bd7 593 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 594 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 595 * @sp: the security path, used for xfrm
1da177e4
LT
596 * @len: Length of actual data
597 * @data_len: Data length
598 * @mac_len: Length of link layer header
334a8132 599 * @hdr_len: writable header length of cloned skb
663ead3b
HX
600 * @csum: Checksum (must include start/offset pair)
601 * @csum_start: Offset from skb->head where checksumming should start
602 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 603 * @priority: Packet queueing priority
60ff7467 604 * @ignore_df: allow local fragmentation
1da177e4 605 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 606 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
607 * @nohdr: Payload reference only, must not modify header
608 * @pkt_type: Packet class
c83c2486 609 * @fclone: skbuff clone status
c83c2486 610 * @ipvs_property: skbuff is owned by ipvs
e7246e12 611 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 612 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
613 * @tc_redirected: packet was redirected by a tc action
614 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
615 * @peeked: this packet has been seen already, so stats have been
616 * done for it, don't do them again
ba9dda3a 617 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
618 * @protocol: Packet protocol from driver
619 * @destructor: Destruct function
e2080072 620 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 621 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 622 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 623 * @skb_iif: ifindex of device we arrived on
1da177e4 624 * @tc_index: Traffic control index
61b905da 625 * @hash: the packet hash
d84e0bd7 626 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 627 * @xmit_more: More SKBs are pending for this queue
a4d9a27d 628 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
553a5672 629 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 630 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 631 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 632 * ports.
a3b18ddb 633 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
634 * @wifi_acked_valid: wifi_acked was set
635 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 636 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
dba00306 637 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
4ff06203 638 * @dst_pending_confirm: need to confirm neighbour
06021292 639 * @napi_id: id of the NAPI struct this skb came from
984bc16c 640 * @secmark: security marking
d84e0bd7 641 * @mark: Generic packet mark
86a9bad3 642 * @vlan_proto: vlan encapsulation protocol
6aa895b0 643 * @vlan_tci: vlan tag control information
0d89d203 644 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
645 * @inner_transport_header: Inner transport layer header (encapsulation)
646 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 647 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
648 * @transport_header: Transport layer header
649 * @network_header: Network layer header
650 * @mac_header: Link layer header
651 * @tail: Tail pointer
652 * @end: End pointer
653 * @head: Head of buffer
654 * @data: Data head pointer
655 * @truesize: Buffer size
656 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
657 */
658
659struct sk_buff {
363ec392 660 union {
56b17425
ED
661 struct {
662 /* These two members must be first. */
663 struct sk_buff *next;
664 struct sk_buff *prev;
665
666 union {
bffa72cf
ED
667 struct net_device *dev;
668 /* Some protocols might use this space to store information,
669 * while device pointer would be NULL.
670 * UDP receive path is one user.
671 */
672 unsigned long dev_scratch;
56b17425
ED
673 };
674 };
4faadabc
PO
675 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
676 struct list_head list;
677 };
678
679 union {
680 struct sock *sk;
681 int ip_defrag_offset;
363ec392 682 };
1da177e4 683
c84d9490 684 union {
bffa72cf
ED
685 ktime_t tstamp;
686 u64 skb_mstamp;
c84d9490 687 };
1da177e4
LT
688 /*
689 * This is the control buffer. It is free to use for every
690 * layer. Please put your private variables there. If you
691 * want to keep them across layers you have to do a skb_clone()
692 * first. This is owned by whoever has the skb queued ATM.
693 */
da3f5cf1 694 char cb[48] __aligned(8);
1da177e4 695
e2080072
ED
696 union {
697 struct {
698 unsigned long _skb_refdst;
699 void (*destructor)(struct sk_buff *skb);
700 };
701 struct list_head tcp_tsorted_anchor;
702 };
703
da3f5cf1
FF
704#ifdef CONFIG_XFRM
705 struct sec_path *sp;
b1937227
ED
706#endif
707#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 708 unsigned long _nfct;
b1937227 709#endif
85224844 710#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 711 struct nf_bridge_info *nf_bridge;
da3f5cf1 712#endif
1da177e4 713 unsigned int len,
334a8132
PM
714 data_len;
715 __u16 mac_len,
716 hdr_len;
b1937227
ED
717
718 /* Following fields are _not_ copied in __copy_skb_header()
719 * Note that queue_mapping is here mostly to fill a hole.
720 */
b1937227 721 __u16 queue_mapping;
36bbef52
DB
722
723/* if you move cloned around you also must adapt those constants */
724#ifdef __BIG_ENDIAN_BITFIELD
725#define CLONED_MASK (1 << 7)
726#else
727#define CLONED_MASK 1
728#endif
729#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
730
731 __u8 __cloned_offset[0];
b1937227 732 __u8 cloned:1,
6869c4d8 733 nohdr:1,
b84f4cc9 734 fclone:2,
a59322be 735 peeked:1,
b1937227 736 head_frag:1,
36bbef52 737 xmit_more:1,
a4d9a27d 738 pfmemalloc:1;
4031ae6e 739
b1937227
ED
740 /* fields enclosed in headers_start/headers_end are copied
741 * using a single memcpy() in __copy_skb_header()
742 */
ebcf34f3 743 /* private: */
b1937227 744 __u32 headers_start[0];
ebcf34f3 745 /* public: */
4031ae6e 746
233577a2
HFS
747/* if you move pkt_type around you also must adapt those constants */
748#ifdef __BIG_ENDIAN_BITFIELD
749#define PKT_TYPE_MAX (7 << 5)
750#else
751#define PKT_TYPE_MAX 7
1da177e4 752#endif
233577a2 753#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 754
233577a2 755 __u8 __pkt_type_offset[0];
b1937227 756 __u8 pkt_type:3;
b1937227 757 __u8 ignore_df:1;
b1937227
ED
758 __u8 nf_trace:1;
759 __u8 ip_summed:2;
3853b584 760 __u8 ooo_okay:1;
a4d9a27d 761
61b905da 762 __u8 l4_hash:1;
a3b18ddb 763 __u8 sw_hash:1;
6e3e939f
JB
764 __u8 wifi_acked_valid:1;
765 __u8 wifi_acked:1;
3bdc0eba 766 __u8 no_fcs:1;
77cffe23 767 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 768 __u8 encapsulation:1;
7e2b10c1 769 __u8 encap_hdr_csum:1;
5d0c2b95 770 __u8 csum_valid:1;
a4d9a27d 771
7e3cead5 772 __u8 csum_complete_sw:1;
b1937227 773 __u8 csum_level:2;
dba00306 774 __u8 csum_not_inet:1;
4ff06203 775 __u8 dst_pending_confirm:1;
b1937227
ED
776#ifdef CONFIG_IPV6_NDISC_NODETYPE
777 __u8 ndisc_nodetype:2;
778#endif
779 __u8 ipvs_property:1;
a4d9a27d 780
8bce6d7d 781 __u8 inner_protocol_type:1;
e585f236 782 __u8 remcsum_offload:1;
6bc506b4
IS
783#ifdef CONFIG_NET_SWITCHDEV
784 __u8 offload_fwd_mark:1;
abf4bb6b 785 __u8 offload_mr_fwd_mark:1;
6bc506b4 786#endif
e7246e12
WB
787#ifdef CONFIG_NET_CLS_ACT
788 __u8 tc_skip_classify:1;
8dc07fdb 789 __u8 tc_at_ingress:1;
bc31c905
WB
790 __u8 tc_redirected:1;
791 __u8 tc_from_ingress:1;
e7246e12 792#endif
b1937227
ED
793
794#ifdef CONFIG_NET_SCHED
795 __u16 tc_index; /* traffic control index */
b1937227 796#endif
fe55f6d5 797
b1937227
ED
798 union {
799 __wsum csum;
800 struct {
801 __u16 csum_start;
802 __u16 csum_offset;
803 };
804 };
805 __u32 priority;
806 int skb_iif;
807 __u32 hash;
808 __be16 vlan_proto;
809 __u16 vlan_tci;
2bd82484
ED
810#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
811 union {
812 unsigned int napi_id;
813 unsigned int sender_cpu;
814 };
97fc2f08 815#endif
984bc16c 816#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 817 __u32 secmark;
0c4f691f 818#endif
0c4f691f 819
3b885787
NH
820 union {
821 __u32 mark;
16fad69c 822 __u32 reserved_tailroom;
3b885787 823 };
1da177e4 824
8bce6d7d
TH
825 union {
826 __be16 inner_protocol;
827 __u8 inner_ipproto;
828 };
829
1a37e412
SH
830 __u16 inner_transport_header;
831 __u16 inner_network_header;
832 __u16 inner_mac_header;
b1937227
ED
833
834 __be16 protocol;
1a37e412
SH
835 __u16 transport_header;
836 __u16 network_header;
837 __u16 mac_header;
b1937227 838
ebcf34f3 839 /* private: */
b1937227 840 __u32 headers_end[0];
ebcf34f3 841 /* public: */
b1937227 842
1da177e4 843 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 844 sk_buff_data_t tail;
4305b541 845 sk_buff_data_t end;
1da177e4 846 unsigned char *head,
4305b541 847 *data;
27a884dc 848 unsigned int truesize;
63354797 849 refcount_t users;
1da177e4
LT
850};
851
852#ifdef __KERNEL__
853/*
854 * Handling routines are only of interest to the kernel
855 */
856#include <linux/slab.h>
857
1da177e4 858
c93bdd0e
MG
859#define SKB_ALLOC_FCLONE 0x01
860#define SKB_ALLOC_RX 0x02
fd11a83d 861#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
862
863/* Returns true if the skb was allocated from PFMEMALLOC reserves */
864static inline bool skb_pfmemalloc(const struct sk_buff *skb)
865{
866 return unlikely(skb->pfmemalloc);
867}
868
7fee226a
ED
869/*
870 * skb might have a dst pointer attached, refcounted or not.
871 * _skb_refdst low order bit is set if refcount was _not_ taken
872 */
873#define SKB_DST_NOREF 1UL
874#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
875
a9e419dc 876#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
877/**
878 * skb_dst - returns skb dst_entry
879 * @skb: buffer
880 *
881 * Returns skb dst_entry, regardless of reference taken or not.
882 */
adf30907
ED
883static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
884{
7fee226a
ED
885 /* If refdst was not refcounted, check we still are in a
886 * rcu_read_lock section
887 */
888 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
889 !rcu_read_lock_held() &&
890 !rcu_read_lock_bh_held());
891 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
892}
893
7fee226a
ED
894/**
895 * skb_dst_set - sets skb dst
896 * @skb: buffer
897 * @dst: dst entry
898 *
899 * Sets skb dst, assuming a reference was taken on dst and should
900 * be released by skb_dst_drop()
901 */
adf30907
ED
902static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
903{
7fee226a
ED
904 skb->_skb_refdst = (unsigned long)dst;
905}
906
932bc4d7
JA
907/**
908 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
909 * @skb: buffer
910 * @dst: dst entry
911 *
912 * Sets skb dst, assuming a reference was not taken on dst.
913 * If dst entry is cached, we do not take reference and dst_release
914 * will be avoided by refdst_drop. If dst entry is not cached, we take
915 * reference, so that last dst_release can destroy the dst immediately.
916 */
917static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
918{
dbfc4fb7
HFS
919 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
920 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 921}
7fee226a
ED
922
923/**
25985edc 924 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
925 * @skb: buffer
926 */
927static inline bool skb_dst_is_noref(const struct sk_buff *skb)
928{
929 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
930}
931
511c3f92
ED
932static inline struct rtable *skb_rtable(const struct sk_buff *skb)
933{
adf30907 934 return (struct rtable *)skb_dst(skb);
511c3f92
ED
935}
936
8b10cab6
JHS
937/* For mangling skb->pkt_type from user space side from applications
938 * such as nft, tc, etc, we only allow a conservative subset of
939 * possible pkt_types to be set.
940*/
941static inline bool skb_pkt_type_ok(u32 ptype)
942{
943 return ptype <= PACKET_OTHERHOST;
944}
945
90b602f8
ML
946static inline unsigned int skb_napi_id(const struct sk_buff *skb)
947{
948#ifdef CONFIG_NET_RX_BUSY_POLL
949 return skb->napi_id;
950#else
951 return 0;
952#endif
953}
954
3889a803
PA
955/* decrement the reference count and return true if we can free the skb */
956static inline bool skb_unref(struct sk_buff *skb)
957{
958 if (unlikely(!skb))
959 return false;
63354797 960 if (likely(refcount_read(&skb->users) == 1))
3889a803 961 smp_rmb();
63354797 962 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
963 return false;
964
965 return true;
966}
967
0a463c78 968void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
969void kfree_skb(struct sk_buff *skb);
970void kfree_skb_list(struct sk_buff *segs);
971void skb_tx_error(struct sk_buff *skb);
972void consume_skb(struct sk_buff *skb);
ca2c1418 973void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 974void __kfree_skb(struct sk_buff *skb);
d7e8883c 975extern struct kmem_cache *skbuff_head_cache;
bad43ca8 976
7965bd4d
JP
977void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
978bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
979 bool *fragstolen, int *delta_truesize);
bad43ca8 980
7965bd4d
JP
981struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
982 int node);
2ea2f62c 983struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 984struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 985static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 986 gfp_t priority)
d179cd12 987{
564824b0 988 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
989}
990
2e4e4410
ED
991struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
992 unsigned long data_len,
993 int max_page_order,
994 int *errcode,
995 gfp_t gfp_mask);
996
d0bf4a9e
ED
997/* Layout of fast clones : [skb1][skb2][fclone_ref] */
998struct sk_buff_fclones {
999 struct sk_buff skb1;
1000
1001 struct sk_buff skb2;
1002
2638595a 1003 refcount_t fclone_ref;
d0bf4a9e
ED
1004};
1005
1006/**
1007 * skb_fclone_busy - check if fclone is busy
293de7de 1008 * @sk: socket
d0bf4a9e
ED
1009 * @skb: buffer
1010 *
bda13fed 1011 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1012 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1013 * so we also check that this didnt happen.
d0bf4a9e 1014 */
39bb5e62
ED
1015static inline bool skb_fclone_busy(const struct sock *sk,
1016 const struct sk_buff *skb)
d0bf4a9e
ED
1017{
1018 const struct sk_buff_fclones *fclones;
1019
1020 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1021
1022 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1023 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1024 fclones->skb2.sk == sk;
d0bf4a9e
ED
1025}
1026
d179cd12 1027static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1028 gfp_t priority)
d179cd12 1029{
c93bdd0e 1030 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1031}
1032
7965bd4d
JP
1033struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1034int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1035struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1036struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1037struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1038 gfp_t gfp_mask, bool fclone);
1039static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1040 gfp_t gfp_mask)
1041{
1042 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1043}
7965bd4d
JP
1044
1045int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1046struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1047 unsigned int headroom);
1048struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1049 int newtailroom, gfp_t priority);
48a1df65
JD
1050int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1051 int offset, int len);
1052int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1053 int offset, int len);
7965bd4d 1054int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1055int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1056
1057/**
1058 * skb_pad - zero pad the tail of an skb
1059 * @skb: buffer to pad
1060 * @pad: space to pad
1061 *
1062 * Ensure that a buffer is followed by a padding area that is zero
1063 * filled. Used by network drivers which may DMA or transfer data
1064 * beyond the buffer end onto the wire.
1065 *
1066 * May return error in out of memory cases. The skb is freed on error.
1067 */
1068static inline int skb_pad(struct sk_buff *skb, int pad)
1069{
1070 return __skb_pad(skb, pad, true);
1071}
ead2ceb0 1072#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1073
7965bd4d
JP
1074int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1075 int getfrag(void *from, char *to, int offset,
1076 int len, int odd, struct sk_buff *skb),
1077 void *from, int length);
e89e9cf5 1078
be12a1fe
HFS
1079int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1080 int offset, size_t size);
1081
d94d9fee 1082struct skb_seq_state {
677e90ed
TG
1083 __u32 lower_offset;
1084 __u32 upper_offset;
1085 __u32 frag_idx;
1086 __u32 stepped_offset;
1087 struct sk_buff *root_skb;
1088 struct sk_buff *cur_skb;
1089 __u8 *frag_data;
1090};
1091
7965bd4d
JP
1092void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1093 unsigned int to, struct skb_seq_state *st);
1094unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1095 struct skb_seq_state *st);
1096void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1097
7965bd4d 1098unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1099 unsigned int to, struct ts_config *config);
3fc7e8a6 1100
09323cc4
TH
1101/*
1102 * Packet hash types specify the type of hash in skb_set_hash.
1103 *
1104 * Hash types refer to the protocol layer addresses which are used to
1105 * construct a packet's hash. The hashes are used to differentiate or identify
1106 * flows of the protocol layer for the hash type. Hash types are either
1107 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1108 *
1109 * Properties of hashes:
1110 *
1111 * 1) Two packets in different flows have different hash values
1112 * 2) Two packets in the same flow should have the same hash value
1113 *
1114 * A hash at a higher layer is considered to be more specific. A driver should
1115 * set the most specific hash possible.
1116 *
1117 * A driver cannot indicate a more specific hash than the layer at which a hash
1118 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1119 *
1120 * A driver may indicate a hash level which is less specific than the
1121 * actual layer the hash was computed on. For instance, a hash computed
1122 * at L4 may be considered an L3 hash. This should only be done if the
1123 * driver can't unambiguously determine that the HW computed the hash at
1124 * the higher layer. Note that the "should" in the second property above
1125 * permits this.
1126 */
1127enum pkt_hash_types {
1128 PKT_HASH_TYPE_NONE, /* Undefined type */
1129 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1130 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1131 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1132};
1133
bcc83839 1134static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1135{
bcc83839 1136 skb->hash = 0;
a3b18ddb 1137 skb->sw_hash = 0;
bcc83839
TH
1138 skb->l4_hash = 0;
1139}
1140
1141static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1142{
1143 if (!skb->l4_hash)
1144 skb_clear_hash(skb);
1145}
1146
1147static inline void
1148__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1149{
1150 skb->l4_hash = is_l4;
1151 skb->sw_hash = is_sw;
61b905da 1152 skb->hash = hash;
09323cc4
TH
1153}
1154
bcc83839
TH
1155static inline void
1156skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1157{
1158 /* Used by drivers to set hash from HW */
1159 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1160}
1161
1162static inline void
1163__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1164{
1165 __skb_set_hash(skb, hash, true, is_l4);
1166}
1167
e5276937 1168void __skb_get_hash(struct sk_buff *skb);
b917783c 1169u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1170u32 skb_get_poff(const struct sk_buff *skb);
1171u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1172 const struct flow_keys *keys, int hlen);
1173__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1174 void *data, int hlen_proto);
1175
1176static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1177 int thoff, u8 ip_proto)
1178{
1179 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1180}
1181
1182void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1183 const struct flow_dissector_key *key,
1184 unsigned int key_count);
1185
1186bool __skb_flow_dissect(const struct sk_buff *skb,
1187 struct flow_dissector *flow_dissector,
1188 void *target_container,
cd79a238
TH
1189 void *data, __be16 proto, int nhoff, int hlen,
1190 unsigned int flags);
e5276937
TH
1191
1192static inline bool skb_flow_dissect(const struct sk_buff *skb,
1193 struct flow_dissector *flow_dissector,
cd79a238 1194 void *target_container, unsigned int flags)
e5276937
TH
1195{
1196 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1197 NULL, 0, 0, 0, flags);
e5276937
TH
1198}
1199
1200static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1201 struct flow_keys *flow,
1202 unsigned int flags)
e5276937
TH
1203{
1204 memset(flow, 0, sizeof(*flow));
1205 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1206 NULL, 0, 0, 0, flags);
e5276937
TH
1207}
1208
1209static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1210 void *data, __be16 proto,
cd79a238
TH
1211 int nhoff, int hlen,
1212 unsigned int flags)
e5276937
TH
1213{
1214 memset(flow, 0, sizeof(*flow));
1215 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1216 data, proto, nhoff, hlen, flags);
e5276937
TH
1217}
1218
3958afa1 1219static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1220{
a3b18ddb 1221 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1222 __skb_get_hash(skb);
bfb564e7 1223
61b905da 1224 return skb->hash;
bfb564e7
KK
1225}
1226
20a17bf6 1227static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1228{
c6cc1ca7
TH
1229 if (!skb->l4_hash && !skb->sw_hash) {
1230 struct flow_keys keys;
de4c1f8b 1231 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1232
de4c1f8b 1233 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1234 }
f70ea018
TH
1235
1236 return skb->hash;
1237}
1238
d0d453f2
ED
1239__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1240 const siphash_key_t *perturb);
50fb7992 1241
57bdf7f4
TH
1242static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1243{
61b905da 1244 return skb->hash;
57bdf7f4
TH
1245}
1246
3df7a74e
TH
1247static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1248{
61b905da 1249 to->hash = from->hash;
a3b18ddb 1250 to->sw_hash = from->sw_hash;
61b905da 1251 to->l4_hash = from->l4_hash;
3df7a74e
TH
1252};
1253
4305b541
ACM
1254#ifdef NET_SKBUFF_DATA_USES_OFFSET
1255static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1256{
1257 return skb->head + skb->end;
1258}
ec47ea82
AD
1259
1260static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1261{
1262 return skb->end;
1263}
4305b541
ACM
1264#else
1265static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1266{
1267 return skb->end;
1268}
ec47ea82
AD
1269
1270static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1271{
1272 return skb->end - skb->head;
1273}
4305b541
ACM
1274#endif
1275
1da177e4 1276/* Internal */
4305b541 1277#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1278
ac45f602
PO
1279static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1280{
1281 return &skb_shinfo(skb)->hwtstamps;
1282}
1283
52267790
WB
1284static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1285{
1286 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1287
1288 return is_zcopy ? skb_uarg(skb) : NULL;
1289}
1290
1291static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1292{
1293 if (skb && uarg && !skb_zcopy(skb)) {
1294 sock_zerocopy_get(uarg);
1295 skb_shinfo(skb)->destructor_arg = uarg;
1296 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1297 }
1298}
1299
95ece2e0
WB
1300static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1301{
1302 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1303 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1304}
1305
1306static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1307{
1308 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1309}
1310
1311static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1312{
1313 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1314}
1315
52267790
WB
1316/* Release a reference on a zerocopy structure */
1317static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1318{
1319 struct ubuf_info *uarg = skb_zcopy(skb);
1320
1321 if (uarg) {
99a86317
WB
1322 if (skb_zcopy_is_nouarg(skb)) {
1323 /* no notification callback */
1324 } else if (uarg->callback == sock_zerocopy_callback) {
0a4a060b
WB
1325 uarg->zerocopy = uarg->zerocopy && zerocopy;
1326 sock_zerocopy_put(uarg);
99a86317 1327 } else {
0a4a060b
WB
1328 uarg->callback(uarg, zerocopy);
1329 }
1330
52267790
WB
1331 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1332 }
1333}
1334
1335/* Abort a zerocopy operation and revert zckey on error in send syscall */
1336static inline void skb_zcopy_abort(struct sk_buff *skb)
1337{
1338 struct ubuf_info *uarg = skb_zcopy(skb);
1339
1340 if (uarg) {
1341 sock_zerocopy_put_abort(uarg);
1342 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1343 }
1344}
1345
1da177e4
LT
1346/**
1347 * skb_queue_empty - check if a queue is empty
1348 * @list: queue head
1349 *
1350 * Returns true if the queue is empty, false otherwise.
1351 */
1352static inline int skb_queue_empty(const struct sk_buff_head *list)
1353{
fd44b93c 1354 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1355}
1356
4b27877d
ED
1357/**
1358 * skb_queue_empty_lockless - check if a queue is empty
1359 * @list: queue head
1360 *
1361 * Returns true if the queue is empty, false otherwise.
1362 * This variant can be used in lockless contexts.
1363 */
1364static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1365{
1366 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1367}
1368
1369
fc7ebb21
DM
1370/**
1371 * skb_queue_is_last - check if skb is the last entry in the queue
1372 * @list: queue head
1373 * @skb: buffer
1374 *
1375 * Returns true if @skb is the last buffer on the list.
1376 */
1377static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1378 const struct sk_buff *skb)
1379{
fd44b93c 1380 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1381}
1382
832d11c5
IJ
1383/**
1384 * skb_queue_is_first - check if skb is the first entry in the queue
1385 * @list: queue head
1386 * @skb: buffer
1387 *
1388 * Returns true if @skb is the first buffer on the list.
1389 */
1390static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1391 const struct sk_buff *skb)
1392{
fd44b93c 1393 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1394}
1395
249c8b42
DM
1396/**
1397 * skb_queue_next - return the next packet in the queue
1398 * @list: queue head
1399 * @skb: current buffer
1400 *
1401 * Return the next packet in @list after @skb. It is only valid to
1402 * call this if skb_queue_is_last() evaluates to false.
1403 */
1404static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1405 const struct sk_buff *skb)
1406{
1407 /* This BUG_ON may seem severe, but if we just return then we
1408 * are going to dereference garbage.
1409 */
1410 BUG_ON(skb_queue_is_last(list, skb));
1411 return skb->next;
1412}
1413
832d11c5
IJ
1414/**
1415 * skb_queue_prev - return the prev packet in the queue
1416 * @list: queue head
1417 * @skb: current buffer
1418 *
1419 * Return the prev packet in @list before @skb. It is only valid to
1420 * call this if skb_queue_is_first() evaluates to false.
1421 */
1422static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1423 const struct sk_buff *skb)
1424{
1425 /* This BUG_ON may seem severe, but if we just return then we
1426 * are going to dereference garbage.
1427 */
1428 BUG_ON(skb_queue_is_first(list, skb));
1429 return skb->prev;
1430}
1431
1da177e4
LT
1432/**
1433 * skb_get - reference buffer
1434 * @skb: buffer to reference
1435 *
1436 * Makes another reference to a socket buffer and returns a pointer
1437 * to the buffer.
1438 */
1439static inline struct sk_buff *skb_get(struct sk_buff *skb)
1440{
63354797 1441 refcount_inc(&skb->users);
1da177e4
LT
1442 return skb;
1443}
1444
1445/*
f8821f96 1446 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1447 */
1448
1da177e4
LT
1449/**
1450 * skb_cloned - is the buffer a clone
1451 * @skb: buffer to check
1452 *
1453 * Returns true if the buffer was generated with skb_clone() and is
1454 * one of multiple shared copies of the buffer. Cloned buffers are
1455 * shared data so must not be written to under normal circumstances.
1456 */
1457static inline int skb_cloned(const struct sk_buff *skb)
1458{
1459 return skb->cloned &&
1460 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1461}
1462
14bbd6a5
PS
1463static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1464{
d0164adc 1465 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1466
1467 if (skb_cloned(skb))
1468 return pskb_expand_head(skb, 0, 0, pri);
1469
1470 return 0;
1471}
1472
1da177e4
LT
1473/**
1474 * skb_header_cloned - is the header a clone
1475 * @skb: buffer to check
1476 *
1477 * Returns true if modifying the header part of the buffer requires
1478 * the data to be copied.
1479 */
1480static inline int skb_header_cloned(const struct sk_buff *skb)
1481{
1482 int dataref;
1483
1484 if (!skb->cloned)
1485 return 0;
1486
1487 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1488 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1489 return dataref != 1;
1490}
1491
9580bf2e
ED
1492static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1493{
1494 might_sleep_if(gfpflags_allow_blocking(pri));
1495
1496 if (skb_header_cloned(skb))
1497 return pskb_expand_head(skb, 0, 0, pri);
1498
1499 return 0;
1500}
1501
f4a775d1
ED
1502/**
1503 * __skb_header_release - release reference to header
1504 * @skb: buffer to operate on
f4a775d1
ED
1505 */
1506static inline void __skb_header_release(struct sk_buff *skb)
1507{
1508 skb->nohdr = 1;
1509 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1510}
1511
1512
1da177e4
LT
1513/**
1514 * skb_shared - is the buffer shared
1515 * @skb: buffer to check
1516 *
1517 * Returns true if more than one person has a reference to this
1518 * buffer.
1519 */
1520static inline int skb_shared(const struct sk_buff *skb)
1521{
63354797 1522 return refcount_read(&skb->users) != 1;
1da177e4
LT
1523}
1524
1525/**
1526 * skb_share_check - check if buffer is shared and if so clone it
1527 * @skb: buffer to check
1528 * @pri: priority for memory allocation
1529 *
1530 * If the buffer is shared the buffer is cloned and the old copy
1531 * drops a reference. A new clone with a single reference is returned.
1532 * If the buffer is not shared the original buffer is returned. When
1533 * being called from interrupt status or with spinlocks held pri must
1534 * be GFP_ATOMIC.
1535 *
1536 * NULL is returned on a memory allocation failure.
1537 */
47061bc4 1538static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1539{
d0164adc 1540 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1541 if (skb_shared(skb)) {
1542 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1543
1544 if (likely(nskb))
1545 consume_skb(skb);
1546 else
1547 kfree_skb(skb);
1da177e4
LT
1548 skb = nskb;
1549 }
1550 return skb;
1551}
1552
1553/*
1554 * Copy shared buffers into a new sk_buff. We effectively do COW on
1555 * packets to handle cases where we have a local reader and forward
1556 * and a couple of other messy ones. The normal one is tcpdumping
1557 * a packet thats being forwarded.
1558 */
1559
1560/**
1561 * skb_unshare - make a copy of a shared buffer
1562 * @skb: buffer to check
1563 * @pri: priority for memory allocation
1564 *
1565 * If the socket buffer is a clone then this function creates a new
1566 * copy of the data, drops a reference count on the old copy and returns
1567 * the new copy with the reference count at 1. If the buffer is not a clone
1568 * the original buffer is returned. When called with a spinlock held or
1569 * from interrupt state @pri must be %GFP_ATOMIC
1570 *
1571 * %NULL is returned on a memory allocation failure.
1572 */
e2bf521d 1573static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1574 gfp_t pri)
1da177e4 1575{
d0164adc 1576 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1577 if (skb_cloned(skb)) {
1578 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1579
1580 /* Free our shared copy */
1581 if (likely(nskb))
1582 consume_skb(skb);
1583 else
1584 kfree_skb(skb);
1da177e4
LT
1585 skb = nskb;
1586 }
1587 return skb;
1588}
1589
1590/**
1a5778aa 1591 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1592 * @list_: list to peek at
1593 *
1594 * Peek an &sk_buff. Unlike most other operations you _MUST_
1595 * be careful with this one. A peek leaves the buffer on the
1596 * list and someone else may run off with it. You must hold
1597 * the appropriate locks or have a private queue to do this.
1598 *
1599 * Returns %NULL for an empty list or a pointer to the head element.
1600 * The reference count is not incremented and the reference is therefore
1601 * volatile. Use with caution.
1602 */
05bdd2f1 1603static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1604{
18d07000
ED
1605 struct sk_buff *skb = list_->next;
1606
1607 if (skb == (struct sk_buff *)list_)
1608 skb = NULL;
1609 return skb;
1da177e4
LT
1610}
1611
da5ef6e5
PE
1612/**
1613 * skb_peek_next - peek skb following the given one from a queue
1614 * @skb: skb to start from
1615 * @list_: list to peek at
1616 *
1617 * Returns %NULL when the end of the list is met or a pointer to the
1618 * next element. The reference count is not incremented and the
1619 * reference is therefore volatile. Use with caution.
1620 */
1621static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1622 const struct sk_buff_head *list_)
1623{
1624 struct sk_buff *next = skb->next;
18d07000 1625
da5ef6e5
PE
1626 if (next == (struct sk_buff *)list_)
1627 next = NULL;
1628 return next;
1629}
1630
1da177e4 1631/**
1a5778aa 1632 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1633 * @list_: list to peek at
1634 *
1635 * Peek an &sk_buff. Unlike most other operations you _MUST_
1636 * be careful with this one. A peek leaves the buffer on the
1637 * list and someone else may run off with it. You must hold
1638 * the appropriate locks or have a private queue to do this.
1639 *
1640 * Returns %NULL for an empty list or a pointer to the tail element.
1641 * The reference count is not incremented and the reference is therefore
1642 * volatile. Use with caution.
1643 */
05bdd2f1 1644static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1645{
4333e7aa 1646 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1647
1648 if (skb == (struct sk_buff *)list_)
1649 skb = NULL;
1650 return skb;
1651
1da177e4
LT
1652}
1653
1654/**
1655 * skb_queue_len - get queue length
1656 * @list_: list to measure
1657 *
1658 * Return the length of an &sk_buff queue.
1659 */
1660static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1661{
1662 return list_->qlen;
1663}
1664
67fed459
DM
1665/**
1666 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1667 * @list: queue to initialize
1668 *
1669 * This initializes only the list and queue length aspects of
1670 * an sk_buff_head object. This allows to initialize the list
1671 * aspects of an sk_buff_head without reinitializing things like
1672 * the spinlock. It can also be used for on-stack sk_buff_head
1673 * objects where the spinlock is known to not be used.
1674 */
1675static inline void __skb_queue_head_init(struct sk_buff_head *list)
1676{
1677 list->prev = list->next = (struct sk_buff *)list;
1678 list->qlen = 0;
1679}
1680
76f10ad0
AV
1681/*
1682 * This function creates a split out lock class for each invocation;
1683 * this is needed for now since a whole lot of users of the skb-queue
1684 * infrastructure in drivers have different locking usage (in hardirq)
1685 * than the networking core (in softirq only). In the long run either the
1686 * network layer or drivers should need annotation to consolidate the
1687 * main types of usage into 3 classes.
1688 */
1da177e4
LT
1689static inline void skb_queue_head_init(struct sk_buff_head *list)
1690{
1691 spin_lock_init(&list->lock);
67fed459 1692 __skb_queue_head_init(list);
1da177e4
LT
1693}
1694
c2ecba71
PE
1695static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1696 struct lock_class_key *class)
1697{
1698 skb_queue_head_init(list);
1699 lockdep_set_class(&list->lock, class);
1700}
1701
1da177e4 1702/*
bf299275 1703 * Insert an sk_buff on a list.
1da177e4
LT
1704 *
1705 * The "__skb_xxxx()" functions are the non-atomic ones that
1706 * can only be called with interrupts disabled.
1707 */
7965bd4d
JP
1708void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1709 struct sk_buff_head *list);
bf299275
GR
1710static inline void __skb_insert(struct sk_buff *newsk,
1711 struct sk_buff *prev, struct sk_buff *next,
1712 struct sk_buff_head *list)
1713{
4333e7aa
ED
1714 /* See skb_queue_empty_lockless() and skb_peek_tail()
1715 * for the opposite READ_ONCE()
1716 */
4b27877d
ED
1717 WRITE_ONCE(newsk->next, next);
1718 WRITE_ONCE(newsk->prev, prev);
1719 WRITE_ONCE(next->prev, newsk);
1720 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1721 list->qlen++;
1722}
1da177e4 1723
67fed459
DM
1724static inline void __skb_queue_splice(const struct sk_buff_head *list,
1725 struct sk_buff *prev,
1726 struct sk_buff *next)
1727{
1728 struct sk_buff *first = list->next;
1729 struct sk_buff *last = list->prev;
1730
4b27877d
ED
1731 WRITE_ONCE(first->prev, prev);
1732 WRITE_ONCE(prev->next, first);
67fed459 1733
4b27877d
ED
1734 WRITE_ONCE(last->next, next);
1735 WRITE_ONCE(next->prev, last);
67fed459
DM
1736}
1737
1738/**
1739 * skb_queue_splice - join two skb lists, this is designed for stacks
1740 * @list: the new list to add
1741 * @head: the place to add it in the first list
1742 */
1743static inline void skb_queue_splice(const struct sk_buff_head *list,
1744 struct sk_buff_head *head)
1745{
1746 if (!skb_queue_empty(list)) {
1747 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1748 head->qlen += list->qlen;
67fed459
DM
1749 }
1750}
1751
1752/**
d9619496 1753 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1754 * @list: the new list to add
1755 * @head: the place to add it in the first list
1756 *
1757 * The list at @list is reinitialised
1758 */
1759static inline void skb_queue_splice_init(struct sk_buff_head *list,
1760 struct sk_buff_head *head)
1761{
1762 if (!skb_queue_empty(list)) {
1763 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1764 head->qlen += list->qlen;
67fed459
DM
1765 __skb_queue_head_init(list);
1766 }
1767}
1768
1769/**
1770 * skb_queue_splice_tail - join two skb lists, each list being a queue
1771 * @list: the new list to add
1772 * @head: the place to add it in the first list
1773 */
1774static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1775 struct sk_buff_head *head)
1776{
1777 if (!skb_queue_empty(list)) {
1778 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1779 head->qlen += list->qlen;
67fed459
DM
1780 }
1781}
1782
1783/**
d9619496 1784 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1785 * @list: the new list to add
1786 * @head: the place to add it in the first list
1787 *
1788 * Each of the lists is a queue.
1789 * The list at @list is reinitialised
1790 */
1791static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1792 struct sk_buff_head *head)
1793{
1794 if (!skb_queue_empty(list)) {
1795 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1796 head->qlen += list->qlen;
67fed459
DM
1797 __skb_queue_head_init(list);
1798 }
1799}
1800
1da177e4 1801/**
300ce174 1802 * __skb_queue_after - queue a buffer at the list head
1da177e4 1803 * @list: list to use
300ce174 1804 * @prev: place after this buffer
1da177e4
LT
1805 * @newsk: buffer to queue
1806 *
300ce174 1807 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1808 * and you must therefore hold required locks before calling it.
1809 *
1810 * A buffer cannot be placed on two lists at the same time.
1811 */
300ce174
SH
1812static inline void __skb_queue_after(struct sk_buff_head *list,
1813 struct sk_buff *prev,
1814 struct sk_buff *newsk)
1da177e4 1815{
bf299275 1816 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1817}
1818
7965bd4d
JP
1819void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1820 struct sk_buff_head *list);
7de6c033 1821
f5572855
GR
1822static inline void __skb_queue_before(struct sk_buff_head *list,
1823 struct sk_buff *next,
1824 struct sk_buff *newsk)
1825{
1826 __skb_insert(newsk, next->prev, next, list);
1827}
1828
300ce174
SH
1829/**
1830 * __skb_queue_head - queue a buffer at the list head
1831 * @list: list to use
1832 * @newsk: buffer to queue
1833 *
1834 * Queue a buffer at the start of a list. This function takes no locks
1835 * and you must therefore hold required locks before calling it.
1836 *
1837 * A buffer cannot be placed on two lists at the same time.
1838 */
7965bd4d 1839void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1840static inline void __skb_queue_head(struct sk_buff_head *list,
1841 struct sk_buff *newsk)
1842{
1843 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1844}
1845
1da177e4
LT
1846/**
1847 * __skb_queue_tail - queue a buffer at the list tail
1848 * @list: list to use
1849 * @newsk: buffer to queue
1850 *
1851 * Queue a buffer at the end of a list. This function takes no locks
1852 * and you must therefore hold required locks before calling it.
1853 *
1854 * A buffer cannot be placed on two lists at the same time.
1855 */
7965bd4d 1856void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1857static inline void __skb_queue_tail(struct sk_buff_head *list,
1858 struct sk_buff *newsk)
1859{
f5572855 1860 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1861}
1862
1da177e4
LT
1863/*
1864 * remove sk_buff from list. _Must_ be called atomically, and with
1865 * the list known..
1866 */
7965bd4d 1867void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1868static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1869{
1870 struct sk_buff *next, *prev;
1871
1872 list->qlen--;
1873 next = skb->next;
1874 prev = skb->prev;
1875 skb->next = skb->prev = NULL;
4b27877d
ED
1876 WRITE_ONCE(next->prev, prev);
1877 WRITE_ONCE(prev->next, next);
1da177e4
LT
1878}
1879
f525c06d
GR
1880/**
1881 * __skb_dequeue - remove from the head of the queue
1882 * @list: list to dequeue from
1883 *
1884 * Remove the head of the list. This function does not take any locks
1885 * so must be used with appropriate locks held only. The head item is
1886 * returned or %NULL if the list is empty.
1887 */
7965bd4d 1888struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1889static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1890{
1891 struct sk_buff *skb = skb_peek(list);
1892 if (skb)
1893 __skb_unlink(skb, list);
1894 return skb;
1895}
1da177e4
LT
1896
1897/**
1898 * __skb_dequeue_tail - remove from the tail of the queue
1899 * @list: list to dequeue from
1900 *
1901 * Remove the tail of the list. This function does not take any locks
1902 * so must be used with appropriate locks held only. The tail item is
1903 * returned or %NULL if the list is empty.
1904 */
7965bd4d 1905struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1906static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1907{
1908 struct sk_buff *skb = skb_peek_tail(list);
1909 if (skb)
1910 __skb_unlink(skb, list);
1911 return skb;
1912}
1913
1914
bdcc0924 1915static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1916{
1917 return skb->data_len;
1918}
1919
1920static inline unsigned int skb_headlen(const struct sk_buff *skb)
1921{
1922 return skb->len - skb->data_len;
1923}
1924
3ece7826 1925static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 1926{
c72d8cda 1927 unsigned int i, len = 0;
1da177e4 1928
c72d8cda 1929 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 1930 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
1931 return len;
1932}
1933
1934static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1935{
1936 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
1937}
1938
131ea667
IC
1939/**
1940 * __skb_fill_page_desc - initialise a paged fragment in an skb
1941 * @skb: buffer containing fragment to be initialised
1942 * @i: paged fragment index to initialise
1943 * @page: the page to use for this fragment
1944 * @off: the offset to the data with @page
1945 * @size: the length of the data
1946 *
1947 * Initialises the @i'th fragment of @skb to point to &size bytes at
1948 * offset @off within @page.
1949 *
1950 * Does not take any additional reference on the fragment.
1951 */
1952static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1953 struct page *page, int off, int size)
1da177e4
LT
1954{
1955 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1956
c48a11c7 1957 /*
2f064f34
MH
1958 * Propagate page pfmemalloc to the skb if we can. The problem is
1959 * that not all callers have unique ownership of the page but rely
1960 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1961 */
a8605c60 1962 frag->page.p = page;
1da177e4 1963 frag->page_offset = off;
9e903e08 1964 skb_frag_size_set(frag, size);
cca7af38
PE
1965
1966 page = compound_head(page);
2f064f34 1967 if (page_is_pfmemalloc(page))
cca7af38 1968 skb->pfmemalloc = true;
131ea667
IC
1969}
1970
1971/**
1972 * skb_fill_page_desc - initialise a paged fragment in an skb
1973 * @skb: buffer containing fragment to be initialised
1974 * @i: paged fragment index to initialise
1975 * @page: the page to use for this fragment
1976 * @off: the offset to the data with @page
1977 * @size: the length of the data
1978 *
1979 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1980 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1981 * addition updates @skb such that @i is the last fragment.
1982 *
1983 * Does not take any additional reference on the fragment.
1984 */
1985static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1986 struct page *page, int off, int size)
1987{
1988 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1989 skb_shinfo(skb)->nr_frags = i + 1;
1990}
1991
7965bd4d
JP
1992void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1993 int size, unsigned int truesize);
654bed16 1994
f8e617e1
JW
1995void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1996 unsigned int truesize);
1997
1da177e4 1998#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1999#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
2000#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2001
27a884dc
ACM
2002#ifdef NET_SKBUFF_DATA_USES_OFFSET
2003static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2004{
2005 return skb->head + skb->tail;
2006}
2007
2008static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2009{
2010 skb->tail = skb->data - skb->head;
2011}
2012
2013static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2014{
2015 skb_reset_tail_pointer(skb);
2016 skb->tail += offset;
2017}
7cc46190 2018
27a884dc
ACM
2019#else /* NET_SKBUFF_DATA_USES_OFFSET */
2020static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2021{
2022 return skb->tail;
2023}
2024
2025static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2026{
2027 skb->tail = skb->data;
2028}
2029
2030static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2031{
2032 skb->tail = skb->data + offset;
2033}
4305b541 2034
27a884dc
ACM
2035#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2036
1da177e4
LT
2037/*
2038 * Add data to an sk_buff
2039 */
4df864c1
JB
2040void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2041void *skb_put(struct sk_buff *skb, unsigned int len);
2042static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2043{
4df864c1 2044 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2045 SKB_LINEAR_ASSERT(skb);
2046 skb->tail += len;
2047 skb->len += len;
2048 return tmp;
2049}
2050
de77b966 2051static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2052{
2053 void *tmp = __skb_put(skb, len);
2054
2055 memset(tmp, 0, len);
2056 return tmp;
2057}
2058
2059static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2060 unsigned int len)
2061{
2062 void *tmp = __skb_put(skb, len);
2063
2064 memcpy(tmp, data, len);
2065 return tmp;
2066}
2067
2068static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2069{
2070 *(u8 *)__skb_put(skb, 1) = val;
2071}
2072
83ad357d 2073static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2074{
83ad357d 2075 void *tmp = skb_put(skb, len);
e45a79da
JB
2076
2077 memset(tmp, 0, len);
2078
2079 return tmp;
2080}
2081
59ae1d12
JB
2082static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2083 unsigned int len)
2084{
2085 void *tmp = skb_put(skb, len);
2086
2087 memcpy(tmp, data, len);
2088
2089 return tmp;
2090}
2091
634fef61
JB
2092static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2093{
2094 *(u8 *)skb_put(skb, 1) = val;
2095}
2096
d58ff351
JB
2097void *skb_push(struct sk_buff *skb, unsigned int len);
2098static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2099{
2100 skb->data -= len;
2101 skb->len += len;
2102 return skb->data;
2103}
2104
af72868b
JB
2105void *skb_pull(struct sk_buff *skb, unsigned int len);
2106static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2107{
2108 skb->len -= len;
2109 BUG_ON(skb->len < skb->data_len);
2110 return skb->data += len;
2111}
2112
af72868b 2113static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2114{
2115 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2116}
2117
af72868b 2118void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2119
af72868b 2120static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2121{
2122 if (len > skb_headlen(skb) &&
987c402a 2123 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2124 return NULL;
2125 skb->len -= len;
2126 return skb->data += len;
2127}
2128
af72868b 2129static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2130{
2131 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2132}
2133
2134static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2135{
2136 if (likely(len <= skb_headlen(skb)))
2137 return 1;
2138 if (unlikely(len > skb->len))
2139 return 0;
987c402a 2140 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2141}
2142
c8c8b127
ED
2143void skb_condense(struct sk_buff *skb);
2144
1da177e4
LT
2145/**
2146 * skb_headroom - bytes at buffer head
2147 * @skb: buffer to check
2148 *
2149 * Return the number of bytes of free space at the head of an &sk_buff.
2150 */
c2636b4d 2151static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2152{
2153 return skb->data - skb->head;
2154}
2155
2156/**
2157 * skb_tailroom - bytes at buffer end
2158 * @skb: buffer to check
2159 *
2160 * Return the number of bytes of free space at the tail of an sk_buff
2161 */
2162static inline int skb_tailroom(const struct sk_buff *skb)
2163{
4305b541 2164 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2165}
2166
a21d4572
ED
2167/**
2168 * skb_availroom - bytes at buffer end
2169 * @skb: buffer to check
2170 *
2171 * Return the number of bytes of free space at the tail of an sk_buff
2172 * allocated by sk_stream_alloc()
2173 */
2174static inline int skb_availroom(const struct sk_buff *skb)
2175{
16fad69c
ED
2176 if (skb_is_nonlinear(skb))
2177 return 0;
2178
2179 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2180}
2181
1da177e4
LT
2182/**
2183 * skb_reserve - adjust headroom
2184 * @skb: buffer to alter
2185 * @len: bytes to move
2186 *
2187 * Increase the headroom of an empty &sk_buff by reducing the tail
2188 * room. This is only allowed for an empty buffer.
2189 */
8243126c 2190static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2191{
2192 skb->data += len;
2193 skb->tail += len;
2194}
2195
1837b2e2
BP
2196/**
2197 * skb_tailroom_reserve - adjust reserved_tailroom
2198 * @skb: buffer to alter
2199 * @mtu: maximum amount of headlen permitted
2200 * @needed_tailroom: minimum amount of reserved_tailroom
2201 *
2202 * Set reserved_tailroom so that headlen can be as large as possible but
2203 * not larger than mtu and tailroom cannot be smaller than
2204 * needed_tailroom.
2205 * The required headroom should already have been reserved before using
2206 * this function.
2207 */
2208static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2209 unsigned int needed_tailroom)
2210{
2211 SKB_LINEAR_ASSERT(skb);
2212 if (mtu < skb_tailroom(skb) - needed_tailroom)
2213 /* use at most mtu */
2214 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2215 else
2216 /* use up to all available space */
2217 skb->reserved_tailroom = needed_tailroom;
2218}
2219
8bce6d7d
TH
2220#define ENCAP_TYPE_ETHER 0
2221#define ENCAP_TYPE_IPPROTO 1
2222
2223static inline void skb_set_inner_protocol(struct sk_buff *skb,
2224 __be16 protocol)
2225{
2226 skb->inner_protocol = protocol;
2227 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2228}
2229
2230static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2231 __u8 ipproto)
2232{
2233 skb->inner_ipproto = ipproto;
2234 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2235}
2236
6a674e9c
JG
2237static inline void skb_reset_inner_headers(struct sk_buff *skb)
2238{
aefbd2b3 2239 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2240 skb->inner_network_header = skb->network_header;
2241 skb->inner_transport_header = skb->transport_header;
2242}
2243
0b5c9db1
JP
2244static inline void skb_reset_mac_len(struct sk_buff *skb)
2245{
2246 skb->mac_len = skb->network_header - skb->mac_header;
2247}
2248
6a674e9c
JG
2249static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2250 *skb)
2251{
2252 return skb->head + skb->inner_transport_header;
2253}
2254
55dc5a9f
TH
2255static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2256{
2257 return skb_inner_transport_header(skb) - skb->data;
2258}
2259
6a674e9c
JG
2260static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2261{
2262 skb->inner_transport_header = skb->data - skb->head;
2263}
2264
2265static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2266 const int offset)
2267{
2268 skb_reset_inner_transport_header(skb);
2269 skb->inner_transport_header += offset;
2270}
2271
2272static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2273{
2274 return skb->head + skb->inner_network_header;
2275}
2276
2277static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2278{
2279 skb->inner_network_header = skb->data - skb->head;
2280}
2281
2282static inline void skb_set_inner_network_header(struct sk_buff *skb,
2283 const int offset)
2284{
2285 skb_reset_inner_network_header(skb);
2286 skb->inner_network_header += offset;
2287}
2288
aefbd2b3
PS
2289static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2290{
2291 return skb->head + skb->inner_mac_header;
2292}
2293
2294static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2295{
2296 skb->inner_mac_header = skb->data - skb->head;
2297}
2298
2299static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2300 const int offset)
2301{
2302 skb_reset_inner_mac_header(skb);
2303 skb->inner_mac_header += offset;
2304}
fda55eca
ED
2305static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2306{
35d04610 2307 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2308}
2309
9c70220b
ACM
2310static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2311{
2e07fa9c 2312 return skb->head + skb->transport_header;
9c70220b
ACM
2313}
2314
badff6d0
ACM
2315static inline void skb_reset_transport_header(struct sk_buff *skb)
2316{
2e07fa9c 2317 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2318}
2319
967b05f6
ACM
2320static inline void skb_set_transport_header(struct sk_buff *skb,
2321 const int offset)
2322{
2e07fa9c
ACM
2323 skb_reset_transport_header(skb);
2324 skb->transport_header += offset;
ea2ae17d
ACM
2325}
2326
d56f90a7
ACM
2327static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2328{
2e07fa9c 2329 return skb->head + skb->network_header;
d56f90a7
ACM
2330}
2331
c1d2bbe1
ACM
2332static inline void skb_reset_network_header(struct sk_buff *skb)
2333{
2e07fa9c 2334 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2335}
2336
c14d2450
ACM
2337static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2338{
2e07fa9c
ACM
2339 skb_reset_network_header(skb);
2340 skb->network_header += offset;
c14d2450
ACM
2341}
2342
2e07fa9c 2343static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2344{
2e07fa9c 2345 return skb->head + skb->mac_header;
bbe735e4
ACM
2346}
2347
ea6da4fd
AV
2348static inline int skb_mac_offset(const struct sk_buff *skb)
2349{
2350 return skb_mac_header(skb) - skb->data;
2351}
2352
0daf4349
DB
2353static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2354{
2355 return skb->network_header - skb->mac_header;
2356}
2357
2e07fa9c 2358static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2359{
35d04610 2360 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2361}
2362
2363static inline void skb_reset_mac_header(struct sk_buff *skb)
2364{
2365 skb->mac_header = skb->data - skb->head;
2366}
2367
2368static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2369{
2370 skb_reset_mac_header(skb);
2371 skb->mac_header += offset;
2372}
2373
0e3da5bb
TT
2374static inline void skb_pop_mac_header(struct sk_buff *skb)
2375{
2376 skb->mac_header = skb->network_header;
2377}
2378
fbbdb8f0
YX
2379static inline void skb_probe_transport_header(struct sk_buff *skb,
2380 const int offset_hint)
2381{
2382 struct flow_keys keys;
2383
2384 if (skb_transport_header_was_set(skb))
2385 return;
cd79a238 2386 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2387 skb_set_transport_header(skb, keys.control.thoff);
7ab88f32 2388 else if (offset_hint >= 0)
fbbdb8f0
YX
2389 skb_set_transport_header(skb, offset_hint);
2390}
2391
03606895
ED
2392static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2393{
2394 if (skb_mac_header_was_set(skb)) {
2395 const unsigned char *old_mac = skb_mac_header(skb);
2396
2397 skb_set_mac_header(skb, -skb->mac_len);
2398 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2399 }
2400}
2401
04fb451e
MM
2402static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2403{
2404 return skb->csum_start - skb_headroom(skb);
2405}
2406
08b64fcc
AD
2407static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2408{
2409 return skb->head + skb->csum_start;
2410}
2411
2e07fa9c
ACM
2412static inline int skb_transport_offset(const struct sk_buff *skb)
2413{
2414 return skb_transport_header(skb) - skb->data;
2415}
2416
2417static inline u32 skb_network_header_len(const struct sk_buff *skb)
2418{
2419 return skb->transport_header - skb->network_header;
2420}
2421
6a674e9c
JG
2422static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2423{
2424 return skb->inner_transport_header - skb->inner_network_header;
2425}
2426
2e07fa9c
ACM
2427static inline int skb_network_offset(const struct sk_buff *skb)
2428{
2429 return skb_network_header(skb) - skb->data;
2430}
48d49d0c 2431
6a674e9c
JG
2432static inline int skb_inner_network_offset(const struct sk_buff *skb)
2433{
2434 return skb_inner_network_header(skb) - skb->data;
2435}
2436
f9599ce1
CG
2437static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2438{
2439 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2440}
2441
1da177e4
LT
2442/*
2443 * CPUs often take a performance hit when accessing unaligned memory
2444 * locations. The actual performance hit varies, it can be small if the
2445 * hardware handles it or large if we have to take an exception and fix it
2446 * in software.
2447 *
2448 * Since an ethernet header is 14 bytes network drivers often end up with
2449 * the IP header at an unaligned offset. The IP header can be aligned by
2450 * shifting the start of the packet by 2 bytes. Drivers should do this
2451 * with:
2452 *
8660c124 2453 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2454 *
2455 * The downside to this alignment of the IP header is that the DMA is now
2456 * unaligned. On some architectures the cost of an unaligned DMA is high
2457 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2458 *
1da177e4
LT
2459 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2460 * to be overridden.
2461 */
2462#ifndef NET_IP_ALIGN
2463#define NET_IP_ALIGN 2
2464#endif
2465
025be81e
AB
2466/*
2467 * The networking layer reserves some headroom in skb data (via
2468 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2469 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2470 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2471 *
2472 * Unfortunately this headroom changes the DMA alignment of the resulting
2473 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2474 * on some architectures. An architecture can override this value,
2475 * perhaps setting it to a cacheline in size (since that will maintain
2476 * cacheline alignment of the DMA). It must be a power of 2.
2477 *
d6301d3d 2478 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2479 * headroom, you should not reduce this.
5933dd2f
ED
2480 *
2481 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2482 * to reduce average number of cache lines per packet.
2483 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2484 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2485 */
2486#ifndef NET_SKB_PAD
5933dd2f 2487#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2488#endif
2489
7965bd4d 2490int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2491
5293efe6 2492static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2493{
c4264f27 2494 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2495 WARN_ON(1);
2496 return;
2497 }
27a884dc
ACM
2498 skb->len = len;
2499 skb_set_tail_pointer(skb, len);
1da177e4
LT
2500}
2501
5293efe6
DB
2502static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2503{
2504 __skb_set_length(skb, len);
2505}
2506
7965bd4d 2507void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2508
2509static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2510{
3cc0e873
HX
2511 if (skb->data_len)
2512 return ___pskb_trim(skb, len);
2513 __skb_trim(skb, len);
2514 return 0;
1da177e4
LT
2515}
2516
2517static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2518{
2519 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2520}
2521
e9fa4f7b
HX
2522/**
2523 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2524 * @skb: buffer to alter
2525 * @len: new length
2526 *
2527 * This is identical to pskb_trim except that the caller knows that
2528 * the skb is not cloned so we should never get an error due to out-
2529 * of-memory.
2530 */
2531static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2532{
2533 int err = pskb_trim(skb, len);
2534 BUG_ON(err);
2535}
2536
5293efe6
DB
2537static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2538{
2539 unsigned int diff = len - skb->len;
2540
2541 if (skb_tailroom(skb) < diff) {
2542 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2543 GFP_ATOMIC);
2544 if (ret)
2545 return ret;
2546 }
2547 __skb_set_length(skb, len);
2548 return 0;
2549}
2550
1da177e4
LT
2551/**
2552 * skb_orphan - orphan a buffer
2553 * @skb: buffer to orphan
2554 *
2555 * If a buffer currently has an owner then we call the owner's
2556 * destructor function and make the @skb unowned. The buffer continues
2557 * to exist but is no longer charged to its former owner.
2558 */
2559static inline void skb_orphan(struct sk_buff *skb)
2560{
c34a7612 2561 if (skb->destructor) {
1da177e4 2562 skb->destructor(skb);
c34a7612
ED
2563 skb->destructor = NULL;
2564 skb->sk = NULL;
376c7311
ED
2565 } else {
2566 BUG_ON(skb->sk);
c34a7612 2567 }
1da177e4
LT
2568}
2569
a353e0ce
MT
2570/**
2571 * skb_orphan_frags - orphan the frags contained in a buffer
2572 * @skb: buffer to orphan frags from
2573 * @gfp_mask: allocation mask for replacement pages
2574 *
2575 * For each frag in the SKB which needs a destructor (i.e. has an
2576 * owner) create a copy of that frag and release the original
2577 * page by calling the destructor.
2578 */
2579static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2580{
1f8b977a
WB
2581 if (likely(!skb_zcopy(skb)))
2582 return 0;
99a86317
WB
2583 if (!skb_zcopy_is_nouarg(skb) &&
2584 skb_uarg(skb)->callback == sock_zerocopy_callback)
1f8b977a
WB
2585 return 0;
2586 return skb_copy_ubufs(skb, gfp_mask);
2587}
2588
2589/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2590static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2591{
2592 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2593 return 0;
2594 return skb_copy_ubufs(skb, gfp_mask);
2595}
2596
1da177e4
LT
2597/**
2598 * __skb_queue_purge - empty a list
2599 * @list: list to empty
2600 *
2601 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2602 * the list and one reference dropped. This function does not take the
2603 * list lock and the caller must hold the relevant locks to use it.
2604 */
7965bd4d 2605void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2606static inline void __skb_queue_purge(struct sk_buff_head *list)
2607{
2608 struct sk_buff *skb;
2609 while ((skb = __skb_dequeue(list)) != NULL)
2610 kfree_skb(skb);
2611}
2612
3c7d5142 2613unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2614
7965bd4d 2615void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2616
7965bd4d
JP
2617struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2618 gfp_t gfp_mask);
8af27456
CH
2619
2620/**
2621 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2622 * @dev: network device to receive on
2623 * @length: length to allocate
2624 *
2625 * Allocate a new &sk_buff and assign it a usage count of one. The
2626 * buffer has unspecified headroom built in. Users should allocate
2627 * the headroom they think they need without accounting for the
2628 * built in space. The built in space is used for optimisations.
2629 *
2630 * %NULL is returned if there is no free memory. Although this function
2631 * allocates memory it can be called from an interrupt.
2632 */
2633static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2634 unsigned int length)
8af27456
CH
2635{
2636 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2637}
2638
6f532612
ED
2639/* legacy helper around __netdev_alloc_skb() */
2640static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2641 gfp_t gfp_mask)
2642{
2643 return __netdev_alloc_skb(NULL, length, gfp_mask);
2644}
2645
2646/* legacy helper around netdev_alloc_skb() */
2647static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2648{
2649 return netdev_alloc_skb(NULL, length);
2650}
2651
2652
4915a0de
ED
2653static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2654 unsigned int length, gfp_t gfp)
61321bbd 2655{
4915a0de 2656 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2657
2658 if (NET_IP_ALIGN && skb)
2659 skb_reserve(skb, NET_IP_ALIGN);
2660 return skb;
2661}
2662
4915a0de
ED
2663static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2664 unsigned int length)
2665{
2666 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2667}
2668
181edb2b
AD
2669static inline void skb_free_frag(void *addr)
2670{
8c2dd3e4 2671 page_frag_free(addr);
181edb2b
AD
2672}
2673
ffde7328 2674void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2675struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2676 unsigned int length, gfp_t gfp_mask);
2677static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2678 unsigned int length)
2679{
2680 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2681}
795bb1c0
JDB
2682void napi_consume_skb(struct sk_buff *skb, int budget);
2683
2684void __kfree_skb_flush(void);
15fad714 2685void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2686
71dfda58
AD
2687/**
2688 * __dev_alloc_pages - allocate page for network Rx
2689 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2690 * @order: size of the allocation
2691 *
2692 * Allocate a new page.
2693 *
2694 * %NULL is returned if there is no free memory.
2695*/
2696static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2697 unsigned int order)
2698{
2699 /* This piece of code contains several assumptions.
2700 * 1. This is for device Rx, therefor a cold page is preferred.
2701 * 2. The expectation is the user wants a compound page.
2702 * 3. If requesting a order 0 page it will not be compound
2703 * due to the check to see if order has a value in prep_new_page
2704 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2705 * code in gfp_to_alloc_flags that should be enforcing this.
2706 */
453f85d4 2707 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2708
2709 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2710}
2711
2712static inline struct page *dev_alloc_pages(unsigned int order)
2713{
95829b3a 2714 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2715}
2716
2717/**
2718 * __dev_alloc_page - allocate a page for network Rx
2719 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2720 *
2721 * Allocate a new page.
2722 *
2723 * %NULL is returned if there is no free memory.
2724 */
2725static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2726{
2727 return __dev_alloc_pages(gfp_mask, 0);
2728}
2729
2730static inline struct page *dev_alloc_page(void)
2731{
95829b3a 2732 return dev_alloc_pages(0);
71dfda58
AD
2733}
2734
0614002b
MG
2735/**
2736 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2737 * @page: The page that was allocated from skb_alloc_page
2738 * @skb: The skb that may need pfmemalloc set
2739 */
2740static inline void skb_propagate_pfmemalloc(struct page *page,
2741 struct sk_buff *skb)
2742{
2f064f34 2743 if (page_is_pfmemalloc(page))
0614002b
MG
2744 skb->pfmemalloc = true;
2745}
2746
131ea667 2747/**
e227867f 2748 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2749 * @frag: the paged fragment
2750 *
2751 * Returns the &struct page associated with @frag.
2752 */
2753static inline struct page *skb_frag_page(const skb_frag_t *frag)
2754{
a8605c60 2755 return frag->page.p;
131ea667
IC
2756}
2757
2758/**
2759 * __skb_frag_ref - take an addition reference on a paged fragment.
2760 * @frag: the paged fragment
2761 *
2762 * Takes an additional reference on the paged fragment @frag.
2763 */
2764static inline void __skb_frag_ref(skb_frag_t *frag)
2765{
2766 get_page(skb_frag_page(frag));
2767}
2768
2769/**
2770 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2771 * @skb: the buffer
2772 * @f: the fragment offset.
2773 *
2774 * Takes an additional reference on the @f'th paged fragment of @skb.
2775 */
2776static inline void skb_frag_ref(struct sk_buff *skb, int f)
2777{
2778 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2779}
2780
2781/**
2782 * __skb_frag_unref - release a reference on a paged fragment.
2783 * @frag: the paged fragment
2784 *
2785 * Releases a reference on the paged fragment @frag.
2786 */
2787static inline void __skb_frag_unref(skb_frag_t *frag)
2788{
2789 put_page(skb_frag_page(frag));
2790}
2791
2792/**
2793 * skb_frag_unref - release a reference on a paged fragment of an skb.
2794 * @skb: the buffer
2795 * @f: the fragment offset
2796 *
2797 * Releases a reference on the @f'th paged fragment of @skb.
2798 */
2799static inline void skb_frag_unref(struct sk_buff *skb, int f)
2800{
2801 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2802}
2803
2804/**
2805 * skb_frag_address - gets the address of the data contained in a paged fragment
2806 * @frag: the paged fragment buffer
2807 *
2808 * Returns the address of the data within @frag. The page must already
2809 * be mapped.
2810 */
2811static inline void *skb_frag_address(const skb_frag_t *frag)
2812{
2813 return page_address(skb_frag_page(frag)) + frag->page_offset;
2814}
2815
2816/**
2817 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2818 * @frag: the paged fragment buffer
2819 *
2820 * Returns the address of the data within @frag. Checks that the page
2821 * is mapped and returns %NULL otherwise.
2822 */
2823static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2824{
2825 void *ptr = page_address(skb_frag_page(frag));
2826 if (unlikely(!ptr))
2827 return NULL;
2828
2829 return ptr + frag->page_offset;
2830}
2831
2832/**
2833 * __skb_frag_set_page - sets the page contained in a paged fragment
2834 * @frag: the paged fragment
2835 * @page: the page to set
2836 *
2837 * Sets the fragment @frag to contain @page.
2838 */
2839static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2840{
a8605c60 2841 frag->page.p = page;
131ea667
IC
2842}
2843
2844/**
2845 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2846 * @skb: the buffer
2847 * @f: the fragment offset
2848 * @page: the page to set
2849 *
2850 * Sets the @f'th fragment of @skb to contain @page.
2851 */
2852static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2853 struct page *page)
2854{
2855 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2856}
2857
400dfd3a
ED
2858bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2859
131ea667
IC
2860/**
2861 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2862 * @dev: the device to map the fragment to
131ea667
IC
2863 * @frag: the paged fragment to map
2864 * @offset: the offset within the fragment (starting at the
2865 * fragment's own offset)
2866 * @size: the number of bytes to map
771b00a8 2867 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
2868 *
2869 * Maps the page associated with @frag to @device.
2870 */
2871static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2872 const skb_frag_t *frag,
2873 size_t offset, size_t size,
2874 enum dma_data_direction dir)
2875{
2876 return dma_map_page(dev, skb_frag_page(frag),
2877 frag->page_offset + offset, size, dir);
2878}
2879
117632e6
ED
2880static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2881 gfp_t gfp_mask)
2882{
2883 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2884}
2885
bad93e9d
OP
2886
2887static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2888 gfp_t gfp_mask)
2889{
2890 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2891}
2892
2893
334a8132
PM
2894/**
2895 * skb_clone_writable - is the header of a clone writable
2896 * @skb: buffer to check
2897 * @len: length up to which to write
2898 *
2899 * Returns true if modifying the header part of the cloned buffer
2900 * does not requires the data to be copied.
2901 */
05bdd2f1 2902static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2903{
2904 return !skb_header_cloned(skb) &&
2905 skb_headroom(skb) + len <= skb->hdr_len;
2906}
2907
3697649f
DB
2908static inline int skb_try_make_writable(struct sk_buff *skb,
2909 unsigned int write_len)
2910{
2911 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2912 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2913}
2914
d9cc2048
HX
2915static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2916 int cloned)
2917{
2918 int delta = 0;
2919
d9cc2048
HX
2920 if (headroom > skb_headroom(skb))
2921 delta = headroom - skb_headroom(skb);
2922
2923 if (delta || cloned)
2924 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2925 GFP_ATOMIC);
2926 return 0;
2927}
2928
1da177e4
LT
2929/**
2930 * skb_cow - copy header of skb when it is required
2931 * @skb: buffer to cow
2932 * @headroom: needed headroom
2933 *
2934 * If the skb passed lacks sufficient headroom or its data part
2935 * is shared, data is reallocated. If reallocation fails, an error
2936 * is returned and original skb is not changed.
2937 *
2938 * The result is skb with writable area skb->head...skb->tail
2939 * and at least @headroom of space at head.
2940 */
2941static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2942{
d9cc2048
HX
2943 return __skb_cow(skb, headroom, skb_cloned(skb));
2944}
1da177e4 2945
d9cc2048
HX
2946/**
2947 * skb_cow_head - skb_cow but only making the head writable
2948 * @skb: buffer to cow
2949 * @headroom: needed headroom
2950 *
2951 * This function is identical to skb_cow except that we replace the
2952 * skb_cloned check by skb_header_cloned. It should be used when
2953 * you only need to push on some header and do not need to modify
2954 * the data.
2955 */
2956static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2957{
2958 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2959}
2960
2961/**
2962 * skb_padto - pad an skbuff up to a minimal size
2963 * @skb: buffer to pad
2964 * @len: minimal length
2965 *
2966 * Pads up a buffer to ensure the trailing bytes exist and are
2967 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2968 * is untouched. Otherwise it is extended. Returns zero on
2969 * success. The skb is freed on error.
1da177e4 2970 */
5b057c6b 2971static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2972{
2973 unsigned int size = skb->len;
2974 if (likely(size >= len))
5b057c6b 2975 return 0;
987c402a 2976 return skb_pad(skb, len - size);
1da177e4
LT
2977}
2978
9c0c1124
AD
2979/**
2980 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2981 * @skb: buffer to pad
2982 * @len: minimal length
cd0a137a 2983 * @free_on_error: free buffer on error
9c0c1124
AD
2984 *
2985 * Pads up a buffer to ensure the trailing bytes exist and are
2986 * blanked. If the buffer already contains sufficient data it
2987 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 2988 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 2989 */
cd0a137a
FF
2990static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2991 bool free_on_error)
9c0c1124
AD
2992{
2993 unsigned int size = skb->len;
2994
2995 if (unlikely(size < len)) {
2996 len -= size;
cd0a137a 2997 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
2998 return -ENOMEM;
2999 __skb_put(skb, len);
3000 }
3001 return 0;
3002}
3003
cd0a137a
FF
3004/**
3005 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3006 * @skb: buffer to pad
3007 * @len: minimal length
3008 *
3009 * Pads up a buffer to ensure the trailing bytes exist and are
3010 * blanked. If the buffer already contains sufficient data it
3011 * is untouched. Otherwise it is extended. Returns zero on
3012 * success. The skb is freed on error.
3013 */
3014static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3015{
3016 return __skb_put_padto(skb, len, true);
3017}
3018
1da177e4 3019static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3020 struct iov_iter *from, int copy)
1da177e4
LT
3021{
3022 const int off = skb->len;
3023
3024 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3025 __wsum csum = 0;
15e6cb46
AV
3026 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3027 &csum, from)) {
1da177e4
LT
3028 skb->csum = csum_block_add(skb->csum, csum, off);
3029 return 0;
3030 }
15e6cb46 3031 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3032 return 0;
3033
3034 __skb_trim(skb, off);
3035 return -EFAULT;
3036}
3037
38ba0a65
ED
3038static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3039 const struct page *page, int off)
1da177e4 3040{
1f8b977a
WB
3041 if (skb_zcopy(skb))
3042 return false;
1da177e4 3043 if (i) {
9e903e08 3044 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3045
ea2ab693 3046 return page == skb_frag_page(frag) &&
9e903e08 3047 off == frag->page_offset + skb_frag_size(frag);
1da177e4 3048 }
38ba0a65 3049 return false;
1da177e4
LT
3050}
3051
364c6bad
HX
3052static inline int __skb_linearize(struct sk_buff *skb)
3053{
3054 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3055}
3056
1da177e4
LT
3057/**
3058 * skb_linearize - convert paged skb to linear one
3059 * @skb: buffer to linarize
1da177e4
LT
3060 *
3061 * If there is no free memory -ENOMEM is returned, otherwise zero
3062 * is returned and the old skb data released.
3063 */
364c6bad
HX
3064static inline int skb_linearize(struct sk_buff *skb)
3065{
3066 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3067}
3068
cef401de
ED
3069/**
3070 * skb_has_shared_frag - can any frag be overwritten
3071 * @skb: buffer to test
3072 *
3073 * Return true if the skb has at least one frag that might be modified
3074 * by an external entity (as in vmsplice()/sendfile())
3075 */
3076static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3077{
c9af6db4
PS
3078 return skb_is_nonlinear(skb) &&
3079 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
3080}
3081
364c6bad
HX
3082/**
3083 * skb_linearize_cow - make sure skb is linear and writable
3084 * @skb: buffer to process
3085 *
3086 * If there is no free memory -ENOMEM is returned, otherwise zero
3087 * is returned and the old skb data released.
3088 */
3089static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3090{
364c6bad
HX
3091 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3092 __skb_linearize(skb) : 0;
1da177e4
LT
3093}
3094
479ffccc
DB
3095static __always_inline void
3096__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3097 unsigned int off)
3098{
3099 if (skb->ip_summed == CHECKSUM_COMPLETE)
3100 skb->csum = csum_block_sub(skb->csum,
3101 csum_partial(start, len, 0), off);
3102 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3103 skb_checksum_start_offset(skb) < 0)
3104 skb->ip_summed = CHECKSUM_NONE;
3105}
3106
1da177e4
LT
3107/**
3108 * skb_postpull_rcsum - update checksum for received skb after pull
3109 * @skb: buffer to update
3110 * @start: start of data before pull
3111 * @len: length of data pulled
3112 *
3113 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3114 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3115 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3116 */
1da177e4 3117static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3118 const void *start, unsigned int len)
1da177e4 3119{
479ffccc 3120 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3121}
3122
479ffccc
DB
3123static __always_inline void
3124__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3125 unsigned int off)
3126{
3127 if (skb->ip_summed == CHECKSUM_COMPLETE)
3128 skb->csum = csum_block_add(skb->csum,
3129 csum_partial(start, len, 0), off);
3130}
cbb042f9 3131
479ffccc
DB
3132/**
3133 * skb_postpush_rcsum - update checksum for received skb after push
3134 * @skb: buffer to update
3135 * @start: start of data after push
3136 * @len: length of data pushed
3137 *
3138 * After doing a push on a received packet, you need to call this to
3139 * update the CHECKSUM_COMPLETE checksum.
3140 */
f8ffad69
DB
3141static inline void skb_postpush_rcsum(struct sk_buff *skb,
3142 const void *start, unsigned int len)
3143{
479ffccc 3144 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3145}
3146
af72868b 3147void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3148
82a31b92
WC
3149/**
3150 * skb_push_rcsum - push skb and update receive checksum
3151 * @skb: buffer to update
3152 * @len: length of data pulled
3153 *
3154 * This function performs an skb_push on the packet and updates
3155 * the CHECKSUM_COMPLETE checksum. It should be used on
3156 * receive path processing instead of skb_push unless you know
3157 * that the checksum difference is zero (e.g., a valid IP header)
3158 * or you are setting ip_summed to CHECKSUM_NONE.
3159 */
d58ff351 3160static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3161{
3162 skb_push(skb, len);
3163 skb_postpush_rcsum(skb, skb->data, len);
3164 return skb->data;
3165}
3166
119e56ab 3167int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3168/**
3169 * pskb_trim_rcsum - trim received skb and update checksum
3170 * @skb: buffer to trim
3171 * @len: new length
3172 *
3173 * This is exactly the same as pskb_trim except that it ensures the
3174 * checksum of received packets are still valid after the operation.
fc4dee53 3175 * It can change skb pointers.
7ce5a27f
DM
3176 */
3177
3178static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3179{
3180 if (likely(len >= skb->len))
3181 return 0;
119e56ab 3182 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3183}
3184
5293efe6
DB
3185static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3186{
3187 if (skb->ip_summed == CHECKSUM_COMPLETE)
3188 skb->ip_summed = CHECKSUM_NONE;
3189 __skb_trim(skb, len);
3190 return 0;
3191}
3192
3193static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3194{
3195 if (skb->ip_summed == CHECKSUM_COMPLETE)
3196 skb->ip_summed = CHECKSUM_NONE;
3197 return __skb_grow(skb, len);
3198}
3199
18a4c0ea
ED
3200#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3201#define skb_rb_first(root) rb_to_skb(rb_first(root))
3202#define skb_rb_last(root) rb_to_skb(rb_last(root))
3203#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3204#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3205
1da177e4
LT
3206#define skb_queue_walk(queue, skb) \
3207 for (skb = (queue)->next; \
a1e4891f 3208 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3209 skb = skb->next)
3210
46f8914e
JC
3211#define skb_queue_walk_safe(queue, skb, tmp) \
3212 for (skb = (queue)->next, tmp = skb->next; \
3213 skb != (struct sk_buff *)(queue); \
3214 skb = tmp, tmp = skb->next)
3215
1164f52a 3216#define skb_queue_walk_from(queue, skb) \
a1e4891f 3217 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3218 skb = skb->next)
3219
18a4c0ea
ED
3220#define skb_rbtree_walk(skb, root) \
3221 for (skb = skb_rb_first(root); skb != NULL; \
3222 skb = skb_rb_next(skb))
3223
3224#define skb_rbtree_walk_from(skb) \
3225 for (; skb != NULL; \
3226 skb = skb_rb_next(skb))
3227
3228#define skb_rbtree_walk_from_safe(skb, tmp) \
3229 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3230 skb = tmp)
3231
1164f52a
DM
3232#define skb_queue_walk_from_safe(queue, skb, tmp) \
3233 for (tmp = skb->next; \
3234 skb != (struct sk_buff *)(queue); \
3235 skb = tmp, tmp = skb->next)
3236
300ce174
SH
3237#define skb_queue_reverse_walk(queue, skb) \
3238 for (skb = (queue)->prev; \
a1e4891f 3239 skb != (struct sk_buff *)(queue); \
300ce174
SH
3240 skb = skb->prev)
3241
686a2955
DM
3242#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3243 for (skb = (queue)->prev, tmp = skb->prev; \
3244 skb != (struct sk_buff *)(queue); \
3245 skb = tmp, tmp = skb->prev)
3246
3247#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3248 for (tmp = skb->prev; \
3249 skb != (struct sk_buff *)(queue); \
3250 skb = tmp, tmp = skb->prev)
1da177e4 3251
21dc3301 3252static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3253{
3254 return skb_shinfo(skb)->frag_list != NULL;
3255}
3256
3257static inline void skb_frag_list_init(struct sk_buff *skb)
3258{
3259 skb_shinfo(skb)->frag_list = NULL;
3260}
3261
ee039871
DM
3262#define skb_walk_frags(skb, iter) \
3263 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3264
ea3793ee
RW
3265
3266int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3267 const struct sk_buff *skb);
65101aec
PA
3268struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3269 struct sk_buff_head *queue,
3270 unsigned int flags,
3271 void (*destructor)(struct sock *sk,
3272 struct sk_buff *skb),
3273 int *peeked, int *off, int *err,
3274 struct sk_buff **last);
ea3793ee 3275struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3276 void (*destructor)(struct sock *sk,
3277 struct sk_buff *skb),
ea3793ee
RW
3278 int *peeked, int *off, int *err,
3279 struct sk_buff **last);
7965bd4d 3280struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3281 void (*destructor)(struct sock *sk,
3282 struct sk_buff *skb),
7965bd4d
JP
3283 int *peeked, int *off, int *err);
3284struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3285 int *err);
3286unsigned int datagram_poll(struct file *file, struct socket *sock,
3287 struct poll_table_struct *wait);
c0371da6
AV
3288int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3289 struct iov_iter *to, int size);
51f3d02b
DM
3290static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3291 struct msghdr *msg, int size)
3292{
e5a4b0bb 3293 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3294}
e5a4b0bb
AV
3295int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3296 struct msghdr *msg);
3a654f97
AV
3297int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3298 struct iov_iter *from, int len);
3a654f97 3299int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3300void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3301void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3302static inline void skb_free_datagram_locked(struct sock *sk,
3303 struct sk_buff *skb)
3304{
3305 __skb_free_datagram_locked(sk, skb, 0);
3306}
7965bd4d 3307int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3308int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3309int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3310__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3311 int len, __wsum csum);
a60e3cc7 3312int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3313 struct pipe_inode_info *pipe, unsigned int len,
25869262 3314 unsigned int flags);
20bf50de
TH
3315int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3316 int len);
3317int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
7965bd4d 3318void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3319unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3320int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3321 int len, int hlen);
7965bd4d
JP
3322void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3323int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3324void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3325unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3326bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
9f4e2543 3327bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3328struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3329struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3330int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3331int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3332int skb_vlan_pop(struct sk_buff *skb);
3333int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3334struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3335 gfp_t gfp);
20380731 3336
6ce8e9ce
AV
3337static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3338{
3073f070 3339 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3340}
3341
7eab8d9e
AV
3342static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3343{
e5a4b0bb 3344 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3345}
3346
2817a336
DB
3347struct skb_checksum_ops {
3348 __wsum (*update)(const void *mem, int len, __wsum wsum);
3349 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3350};
3351
9617813d
DC
3352extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3353
2817a336
DB
3354__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3355 __wsum csum, const struct skb_checksum_ops *ops);
3356__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3357 __wsum csum);
3358
1e98a0f0
ED
3359static inline void * __must_check
3360__skb_header_pointer(const struct sk_buff *skb, int offset,
3361 int len, void *data, int hlen, void *buffer)
1da177e4 3362{
55820ee2 3363 if (hlen - offset >= len)
690e36e7 3364 return data + offset;
1da177e4 3365
690e36e7
DM
3366 if (!skb ||
3367 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3368 return NULL;
3369
3370 return buffer;
3371}
3372
1e98a0f0
ED
3373static inline void * __must_check
3374skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3375{
3376 return __skb_header_pointer(skb, offset, len, skb->data,
3377 skb_headlen(skb), buffer);
3378}
3379
4262e5cc
DB
3380/**
3381 * skb_needs_linearize - check if we need to linearize a given skb
3382 * depending on the given device features.
3383 * @skb: socket buffer to check
3384 * @features: net device features
3385 *
3386 * Returns true if either:
3387 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3388 * 2. skb is fragmented and the device does not support SG.
3389 */
3390static inline bool skb_needs_linearize(struct sk_buff *skb,
3391 netdev_features_t features)
3392{
3393 return skb_is_nonlinear(skb) &&
3394 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3395 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3396}
3397
d626f62b
ACM
3398static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3399 void *to,
3400 const unsigned int len)
3401{
3402 memcpy(to, skb->data, len);
3403}
3404
3405static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3406 const int offset, void *to,
3407 const unsigned int len)
3408{
3409 memcpy(to, skb->data + offset, len);
3410}
3411
27d7ff46
ACM
3412static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3413 const void *from,
3414 const unsigned int len)
3415{
3416 memcpy(skb->data, from, len);
3417}
3418
3419static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3420 const int offset,
3421 const void *from,
3422 const unsigned int len)
3423{
3424 memcpy(skb->data + offset, from, len);
3425}
3426
7965bd4d 3427void skb_init(void);
1da177e4 3428
ac45f602
PO
3429static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3430{
3431 return skb->tstamp;
3432}
3433
a61bbcf2
PM
3434/**
3435 * skb_get_timestamp - get timestamp from a skb
3436 * @skb: skb to get stamp from
3437 * @stamp: pointer to struct timeval to store stamp in
3438 *
3439 * Timestamps are stored in the skb as offsets to a base timestamp.
3440 * This function converts the offset back to a struct timeval and stores
3441 * it in stamp.
3442 */
ac45f602
PO
3443static inline void skb_get_timestamp(const struct sk_buff *skb,
3444 struct timeval *stamp)
a61bbcf2 3445{
b7aa0bf7 3446 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3447}
3448
ac45f602
PO
3449static inline void skb_get_timestampns(const struct sk_buff *skb,
3450 struct timespec *stamp)
3451{
3452 *stamp = ktime_to_timespec(skb->tstamp);
3453}
3454
b7aa0bf7 3455static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3456{
b7aa0bf7 3457 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3458}
3459
164891aa
SH
3460static inline ktime_t net_timedelta(ktime_t t)
3461{
3462 return ktime_sub(ktime_get_real(), t);
3463}
3464
b9ce204f
IJ
3465static inline ktime_t net_invalid_timestamp(void)
3466{
8b0e1953 3467 return 0;
b9ce204f 3468}
a61bbcf2 3469
de8f3a83
DB
3470static inline u8 skb_metadata_len(const struct sk_buff *skb)
3471{
3472 return skb_shinfo(skb)->meta_len;
3473}
3474
3475static inline void *skb_metadata_end(const struct sk_buff *skb)
3476{
3477 return skb_mac_header(skb);
3478}
3479
3480static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3481 const struct sk_buff *skb_b,
3482 u8 meta_len)
3483{
3484 const void *a = skb_metadata_end(skb_a);
3485 const void *b = skb_metadata_end(skb_b);
3486 /* Using more efficient varaiant than plain call to memcmp(). */
3487#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3488 u64 diffs = 0;
3489
3490 switch (meta_len) {
3491#define __it(x, op) (x -= sizeof(u##op))
3492#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3493 case 32: diffs |= __it_diff(a, b, 64);
3494 case 24: diffs |= __it_diff(a, b, 64);
3495 case 16: diffs |= __it_diff(a, b, 64);
3496 case 8: diffs |= __it_diff(a, b, 64);
3497 break;
3498 case 28: diffs |= __it_diff(a, b, 64);
3499 case 20: diffs |= __it_diff(a, b, 64);
3500 case 12: diffs |= __it_diff(a, b, 64);
3501 case 4: diffs |= __it_diff(a, b, 32);
3502 break;
3503 }
3504 return diffs;
3505#else
3506 return memcmp(a - meta_len, b - meta_len, meta_len);
3507#endif
3508}
3509
3510static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3511 const struct sk_buff *skb_b)
3512{
3513 u8 len_a = skb_metadata_len(skb_a);
3514 u8 len_b = skb_metadata_len(skb_b);
3515
3516 if (!(len_a | len_b))
3517 return false;
3518
3519 return len_a != len_b ?
3520 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3521}
3522
3523static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3524{
3525 skb_shinfo(skb)->meta_len = meta_len;
3526}
3527
3528static inline void skb_metadata_clear(struct sk_buff *skb)
3529{
3530 skb_metadata_set(skb, 0);
3531}
3532
62bccb8c
AD
3533struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3534
c1f19b51
RC
3535#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3536
7965bd4d
JP
3537void skb_clone_tx_timestamp(struct sk_buff *skb);
3538bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3539
3540#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3541
3542static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3543{
3544}
3545
3546static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3547{
3548 return false;
3549}
3550
3551#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3552
3553/**
3554 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3555 *
da92b194
RC
3556 * PHY drivers may accept clones of transmitted packets for
3557 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3558 * must call this function to return the skb back to the stack with a
3559 * timestamp.
da92b194 3560 *
c1f19b51 3561 * @skb: clone of the the original outgoing packet
7a76a021 3562 * @hwtstamps: hardware time stamps
c1f19b51
RC
3563 *
3564 */
3565void skb_complete_tx_timestamp(struct sk_buff *skb,
3566 struct skb_shared_hwtstamps *hwtstamps);
3567
e7fd2885
WB
3568void __skb_tstamp_tx(struct sk_buff *orig_skb,
3569 struct skb_shared_hwtstamps *hwtstamps,
3570 struct sock *sk, int tstype);
3571
ac45f602
PO
3572/**
3573 * skb_tstamp_tx - queue clone of skb with send time stamps
3574 * @orig_skb: the original outgoing packet
3575 * @hwtstamps: hardware time stamps, may be NULL if not available
3576 *
3577 * If the skb has a socket associated, then this function clones the
3578 * skb (thus sharing the actual data and optional structures), stores
3579 * the optional hardware time stamping information (if non NULL) or
3580 * generates a software time stamp (otherwise), then queues the clone
3581 * to the error queue of the socket. Errors are silently ignored.
3582 */
7965bd4d
JP
3583void skb_tstamp_tx(struct sk_buff *orig_skb,
3584 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3585
4507a715
RC
3586/**
3587 * skb_tx_timestamp() - Driver hook for transmit timestamping
3588 *
3589 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3590 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3591 *
73409f3b
DM
3592 * Specifically, one should make absolutely sure that this function is
3593 * called before TX completion of this packet can trigger. Otherwise
3594 * the packet could potentially already be freed.
3595 *
4507a715
RC
3596 * @skb: A socket buffer.
3597 */
3598static inline void skb_tx_timestamp(struct sk_buff *skb)
3599{
c1f19b51 3600 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3601 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3602 skb_tstamp_tx(skb, NULL);
4507a715
RC
3603}
3604
6e3e939f
JB
3605/**
3606 * skb_complete_wifi_ack - deliver skb with wifi status
3607 *
3608 * @skb: the original outgoing packet
3609 * @acked: ack status
3610 *
3611 */
3612void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3613
7965bd4d
JP
3614__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3615__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3616
60476372
HX
3617static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3618{
6edec0e6
TH
3619 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3620 skb->csum_valid ||
3621 (skb->ip_summed == CHECKSUM_PARTIAL &&
3622 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3623}
3624
fb286bb2
HX
3625/**
3626 * skb_checksum_complete - Calculate checksum of an entire packet
3627 * @skb: packet to process
3628 *
3629 * This function calculates the checksum over the entire packet plus
3630 * the value of skb->csum. The latter can be used to supply the
3631 * checksum of a pseudo header as used by TCP/UDP. It returns the
3632 * checksum.
3633 *
3634 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3635 * this function can be used to verify that checksum on received
3636 * packets. In that case the function should return zero if the
3637 * checksum is correct. In particular, this function will return zero
3638 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3639 * hardware has already verified the correctness of the checksum.
3640 */
4381ca3c 3641static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3642{
60476372
HX
3643 return skb_csum_unnecessary(skb) ?
3644 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3645}
3646
77cffe23
TH
3647static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3648{
3649 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3650 if (skb->csum_level == 0)
3651 skb->ip_summed = CHECKSUM_NONE;
3652 else
3653 skb->csum_level--;
3654 }
3655}
3656
3657static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3658{
3659 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3660 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3661 skb->csum_level++;
3662 } else if (skb->ip_summed == CHECKSUM_NONE) {
3663 skb->ip_summed = CHECKSUM_UNNECESSARY;
3664 skb->csum_level = 0;
3665 }
3666}
3667
76ba0aae
TH
3668/* Check if we need to perform checksum complete validation.
3669 *
3670 * Returns true if checksum complete is needed, false otherwise
3671 * (either checksum is unnecessary or zero checksum is allowed).
3672 */
3673static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3674 bool zero_okay,
3675 __sum16 check)
3676{
5d0c2b95
TH
3677 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3678 skb->csum_valid = 1;
77cffe23 3679 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3680 return false;
3681 }
3682
3683 return true;
3684}
3685
3686/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3687 * in checksum_init.
3688 */
3689#define CHECKSUM_BREAK 76
3690
4e18b9ad
TH
3691/* Unset checksum-complete
3692 *
3693 * Unset checksum complete can be done when packet is being modified
3694 * (uncompressed for instance) and checksum-complete value is
3695 * invalidated.
3696 */
3697static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3698{
3699 if (skb->ip_summed == CHECKSUM_COMPLETE)
3700 skb->ip_summed = CHECKSUM_NONE;
3701}
3702
76ba0aae
TH
3703/* Validate (init) checksum based on checksum complete.
3704 *
3705 * Return values:
3706 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3707 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3708 * checksum is stored in skb->csum for use in __skb_checksum_complete
3709 * non-zero: value of invalid checksum
3710 *
3711 */
3712static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3713 bool complete,
3714 __wsum psum)
3715{
3716 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3717 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3718 skb->csum_valid = 1;
76ba0aae
TH
3719 return 0;
3720 }
3721 }
3722
3723 skb->csum = psum;
3724
5d0c2b95
TH
3725 if (complete || skb->len <= CHECKSUM_BREAK) {
3726 __sum16 csum;
3727
3728 csum = __skb_checksum_complete(skb);
3729 skb->csum_valid = !csum;
3730 return csum;
3731 }
76ba0aae
TH
3732
3733 return 0;
3734}
3735
3736static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3737{
3738 return 0;
3739}
3740
3741/* Perform checksum validate (init). Note that this is a macro since we only
3742 * want to calculate the pseudo header which is an input function if necessary.
3743 * First we try to validate without any computation (checksum unnecessary) and
3744 * then calculate based on checksum complete calling the function to compute
3745 * pseudo header.
3746 *
3747 * Return values:
3748 * 0: checksum is validated or try to in skb_checksum_complete
3749 * non-zero: value of invalid checksum
3750 */
3751#define __skb_checksum_validate(skb, proto, complete, \
3752 zero_okay, check, compute_pseudo) \
3753({ \
3754 __sum16 __ret = 0; \
5d0c2b95 3755 skb->csum_valid = 0; \
76ba0aae
TH
3756 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3757 __ret = __skb_checksum_validate_complete(skb, \
3758 complete, compute_pseudo(skb, proto)); \
3759 __ret; \
3760})
3761
3762#define skb_checksum_init(skb, proto, compute_pseudo) \
3763 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3764
3765#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3766 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3767
3768#define skb_checksum_validate(skb, proto, compute_pseudo) \
3769 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3770
3771#define skb_checksum_validate_zero_check(skb, proto, check, \
3772 compute_pseudo) \
096a4cfa 3773 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3774
3775#define skb_checksum_simple_validate(skb) \
3776 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3777
d96535a1
TH
3778static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3779{
219f1d79 3780 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
3781}
3782
3783static inline void __skb_checksum_convert(struct sk_buff *skb,
3784 __sum16 check, __wsum pseudo)
3785{
3786 skb->csum = ~pseudo;
3787 skb->ip_summed = CHECKSUM_COMPLETE;
3788}
3789
3790#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3791do { \
3792 if (__skb_checksum_convert_check(skb)) \
3793 __skb_checksum_convert(skb, check, \
3794 compute_pseudo(skb, proto)); \
3795} while (0)
3796
15e2396d
TH
3797static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3798 u16 start, u16 offset)
3799{
3800 skb->ip_summed = CHECKSUM_PARTIAL;
3801 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3802 skb->csum_offset = offset - start;
3803}
3804
dcdc8994
TH
3805/* Update skbuf and packet to reflect the remote checksum offload operation.
3806 * When called, ptr indicates the starting point for skb->csum when
3807 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3808 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3809 */
3810static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3811 int start, int offset, bool nopartial)
dcdc8994
TH
3812{
3813 __wsum delta;
3814
15e2396d
TH
3815 if (!nopartial) {
3816 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3817 return;
3818 }
3819
dcdc8994
TH
3820 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3821 __skb_checksum_complete(skb);
3822 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3823 }
3824
3825 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3826
3827 /* Adjust skb->csum since we changed the packet */
3828 skb->csum = csum_add(skb->csum, delta);
3829}
3830
cb9c6836
FW
3831static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3832{
3833#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 3834 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
3835#else
3836 return NULL;
3837#endif
3838}
3839
5f79e0f9 3840#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3841void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3842static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3843{
3844 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3845 nf_conntrack_destroy(nfct);
1da177e4
LT
3846}
3847static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3848{
3849 if (nfct)
3850 atomic_inc(&nfct->use);
3851}
2fc72c7b 3852#endif
34666d46 3853#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3854static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3855{
53869ceb 3856 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
1da177e4
LT
3857 kfree(nf_bridge);
3858}
3859static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3860{
3861 if (nf_bridge)
53869ceb 3862 refcount_inc(&nf_bridge->use);
1da177e4
LT
3863}
3864#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3865static inline void nf_reset(struct sk_buff *skb)
3866{
5f79e0f9 3867#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3868 nf_conntrack_put(skb_nfct(skb));
3869 skb->_nfct = 0;
2fc72c7b 3870#endif
34666d46 3871#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3872 nf_bridge_put(skb->nf_bridge);
3873 skb->nf_bridge = NULL;
3874#endif
3875}
3876
124dff01
PM
3877static inline void nf_reset_trace(struct sk_buff *skb)
3878{
478b360a 3879#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3880 skb->nf_trace = 0;
3881#endif
a193a4ab
PM
3882}
3883
2b5ec1a5
YY
3884static inline void ipvs_reset(struct sk_buff *skb)
3885{
3886#if IS_ENABLED(CONFIG_IP_VS)
3887 skb->ipvs_property = 0;
3888#endif
3889}
3890
edda553c 3891/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3892static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3893 bool copy)
edda553c 3894{
5f79e0f9 3895#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3896 dst->_nfct = src->_nfct;
3897 nf_conntrack_get(skb_nfct(src));
2fc72c7b 3898#endif
34666d46 3899#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3900 dst->nf_bridge = src->nf_bridge;
3901 nf_bridge_get(src->nf_bridge);
3902#endif
478b360a 3903#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3904 if (copy)
3905 dst->nf_trace = src->nf_trace;
478b360a 3906#endif
edda553c
YK
3907}
3908
e7ac05f3
YK
3909static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3910{
e7ac05f3 3911#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 3912 nf_conntrack_put(skb_nfct(dst));
2fc72c7b 3913#endif
34666d46 3914#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3915 nf_bridge_put(dst->nf_bridge);
3916#endif
b1937227 3917 __nf_copy(dst, src, true);
e7ac05f3
YK
3918}
3919
984bc16c
JM
3920#ifdef CONFIG_NETWORK_SECMARK
3921static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3922{
3923 to->secmark = from->secmark;
3924}
3925
3926static inline void skb_init_secmark(struct sk_buff *skb)
3927{
3928 skb->secmark = 0;
3929}
3930#else
3931static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3932{ }
3933
3934static inline void skb_init_secmark(struct sk_buff *skb)
3935{ }
3936#endif
3937
574f7194
EB
3938static inline bool skb_irq_freeable(const struct sk_buff *skb)
3939{
3940 return !skb->destructor &&
3941#if IS_ENABLED(CONFIG_XFRM)
3942 !skb->sp &&
3943#endif
cb9c6836 3944 !skb_nfct(skb) &&
574f7194
EB
3945 !skb->_skb_refdst &&
3946 !skb_has_frag_list(skb);
3947}
3948
f25f4e44
PWJ
3949static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3950{
f25f4e44 3951 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3952}
3953
9247744e 3954static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3955{
4e3ab47a 3956 return skb->queue_mapping;
4e3ab47a
PE
3957}
3958
f25f4e44
PWJ
3959static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3960{
f25f4e44 3961 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3962}
3963
d5a9e24a
DM
3964static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3965{
3966 skb->queue_mapping = rx_queue + 1;
3967}
3968
9247744e 3969static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3970{
3971 return skb->queue_mapping - 1;
3972}
3973
9247744e 3974static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3975{
a02cec21 3976 return skb->queue_mapping != 0;
d5a9e24a
DM
3977}
3978
4ff06203
JA
3979static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3980{
3981 skb->dst_pending_confirm = val;
3982}
3983
3984static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3985{
3986 return skb->dst_pending_confirm != 0;
3987}
3988
def8b4fa
AD
3989static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3990{
0b3d8e08 3991#ifdef CONFIG_XFRM
def8b4fa 3992 return skb->sp;
def8b4fa 3993#else
def8b4fa 3994 return NULL;
def8b4fa 3995#endif
0b3d8e08 3996}
def8b4fa 3997
68c33163
PS
3998/* Keeps track of mac header offset relative to skb->head.
3999 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4000 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4001 * tunnel skb it points to outer mac header.
4002 * Keeps track of level of encapsulation of network headers.
4003 */
68c33163 4004struct skb_gso_cb {
802ab55a
AD
4005 union {
4006 int mac_offset;
4007 int data_offset;
4008 };
3347c960 4009 int encap_level;
76443456 4010 __wsum csum;
7e2b10c1 4011 __u16 csum_start;
68c33163 4012};
9207f9d4
KK
4013#define SKB_SGO_CB_OFFSET 32
4014#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
4015
4016static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4017{
4018 return (skb_mac_header(inner_skb) - inner_skb->head) -
4019 SKB_GSO_CB(inner_skb)->mac_offset;
4020}
4021
1e2bd517
PS
4022static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4023{
4024 int new_headroom, headroom;
4025 int ret;
4026
4027 headroom = skb_headroom(skb);
4028 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4029 if (ret)
4030 return ret;
4031
4032 new_headroom = skb_headroom(skb);
4033 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4034 return 0;
4035}
4036
08b64fcc
AD
4037static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4038{
4039 /* Do not update partial checksums if remote checksum is enabled. */
4040 if (skb->remcsum_offload)
4041 return;
4042
4043 SKB_GSO_CB(skb)->csum = res;
4044 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4045}
4046
7e2b10c1
TH
4047/* Compute the checksum for a gso segment. First compute the checksum value
4048 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4049 * then add in skb->csum (checksum from csum_start to end of packet).
4050 * skb->csum and csum_start are then updated to reflect the checksum of the
4051 * resultant packet starting from the transport header-- the resultant checksum
4052 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4053 * header.
4054 */
4055static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4056{
76443456
AD
4057 unsigned char *csum_start = skb_transport_header(skb);
4058 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4059 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4060
76443456
AD
4061 SKB_GSO_CB(skb)->csum = res;
4062 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4063
76443456 4064 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4065}
4066
bdcc0924 4067static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4068{
4069 return skb_shinfo(skb)->gso_size;
4070}
4071
36a8f39e 4072/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4073static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4074{
4075 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4076}
4077
5293efe6
DB
4078static inline void skb_gso_reset(struct sk_buff *skb)
4079{
4080 skb_shinfo(skb)->gso_size = 0;
4081 skb_shinfo(skb)->gso_segs = 0;
4082 skb_shinfo(skb)->gso_type = 0;
4083}
4084
7965bd4d 4085void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4086
4087static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4088{
4089 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4090 * wanted then gso_type will be set. */
05bdd2f1
ED
4091 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4092
b78462eb
AD
4093 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4094 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4095 __skb_warn_lro_forwarding(skb);
4096 return true;
4097 }
4098 return false;
4099}
4100
35fc92a9
HX
4101static inline void skb_forward_csum(struct sk_buff *skb)
4102{
4103 /* Unfortunately we don't support this one. Any brave souls? */
4104 if (skb->ip_summed == CHECKSUM_COMPLETE)
4105 skb->ip_summed = CHECKSUM_NONE;
4106}
4107
bc8acf2c
ED
4108/**
4109 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4110 * @skb: skb to check
4111 *
4112 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4113 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4114 * use this helper, to document places where we make this assertion.
4115 */
05bdd2f1 4116static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4117{
4118#ifdef DEBUG
4119 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4120#endif
4121}
4122
f35d9d8a 4123bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4124
ed1f50c3 4125int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4126struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4127 unsigned int transport_len,
4128 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4129
3a7c1ee4
AD
4130/**
4131 * skb_head_is_locked - Determine if the skb->head is locked down
4132 * @skb: skb to check
4133 *
4134 * The head on skbs build around a head frag can be removed if they are
4135 * not cloned. This function returns true if the skb head is locked down
4136 * due to either being allocated via kmalloc, or by being a clone with
4137 * multiple references to the head.
4138 */
4139static inline bool skb_head_is_locked(const struct sk_buff *skb)
4140{
4141 return !skb->head_frag || skb_cloned(skb);
4142}
fe6cc55f
FW
4143
4144/**
4145 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4146 *
4147 * @skb: GSO skb
4148 *
4149 * skb_gso_network_seglen is used to determine the real size of the
4150 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4151 *
4152 * The MAC/L2 header is not accounted for.
4153 */
4154static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4155{
4156 unsigned int hdr_len = skb_transport_header(skb) -
4157 skb_network_header(skb);
4158 return hdr_len + skb_gso_transport_seglen(skb);
4159}
ee122c79 4160
9f4e2543
DA
4161/**
4162 * skb_gso_mac_seglen - Return length of individual segments of a gso packet
4163 *
4164 * @skb: GSO skb
4165 *
4166 * skb_gso_mac_seglen is used to determine the real size of the
4167 * individual segments, including MAC/L2, Layer3 (IP, IPv6) and L4
4168 * headers (TCP/UDP).
4169 */
4170static inline unsigned int skb_gso_mac_seglen(const struct sk_buff *skb)
4171{
4172 unsigned int hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
4173 return hdr_len + skb_gso_transport_seglen(skb);
4174}
4175
179bc67f
EC
4176/* Local Checksum Offload.
4177 * Compute outer checksum based on the assumption that the
4178 * inner checksum will be offloaded later.
e8ae7b00
EC
4179 * See Documentation/networking/checksum-offloads.txt for
4180 * explanation of how this works.
179bc67f
EC
4181 * Fill in outer checksum adjustment (e.g. with sum of outer
4182 * pseudo-header) before calling.
4183 * Also ensure that inner checksum is in linear data area.
4184 */
4185static inline __wsum lco_csum(struct sk_buff *skb)
4186{
9e74a6da
AD
4187 unsigned char *csum_start = skb_checksum_start(skb);
4188 unsigned char *l4_hdr = skb_transport_header(skb);
4189 __wsum partial;
179bc67f
EC
4190
4191 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4192 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4193 skb->csum_offset));
4194
179bc67f 4195 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4196 * adjustment filled in by caller) and return result.
179bc67f 4197 */
9e74a6da 4198 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4199}
4200
1da177e4
LT
4201#endif /* __KERNEL__ */
4202#endif /* _LINUX_SKBUFF_H */