]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blame - include/linux/skbuff.h
UBUNTU: Ubuntu-5.3.0-29.31
[mirror_ubuntu-eoan-kernel.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
1da177e4 17#include <linux/cache.h>
56b17425 18#include <linux/rbtree.h>
51f3d02b 19#include <linux/socket.h>
c1d1b437 20#include <linux/refcount.h>
1da177e4 21
60063497 22#include <linux/atomic.h>
1da177e4
LT
23#include <asm/types.h>
24#include <linux/spinlock.h>
1da177e4 25#include <linux/net.h>
3fc7e8a6 26#include <linux/textsearch.h>
1da177e4 27#include <net/checksum.h>
a80958f4 28#include <linux/rcupdate.h>
b7aa0bf7 29#include <linux/hrtimer.h>
131ea667 30#include <linux/dma-mapping.h>
c8f44aff 31#include <linux/netdev_features.h>
363ec392 32#include <linux/sched.h>
e6017571 33#include <linux/sched/clock.h>
1bd758eb 34#include <net/flow_dissector.h>
a60e3cc7 35#include <linux/splice.h>
72b31f72 36#include <linux/in6.h>
8b10cab6 37#include <linux/if_packet.h>
f70ea018 38#include <net/flow.h>
1da177e4 39
7a6ae71b
TH
40/* The interface for checksum offload between the stack and networking drivers
41 * is as follows...
42 *
43 * A. IP checksum related features
44 *
45 * Drivers advertise checksum offload capabilities in the features of a device.
46 * From the stack's point of view these are capabilities offered by the driver,
47 * a driver typically only advertises features that it is capable of offloading
48 * to its device.
49 *
50 * The checksum related features are:
51 *
52 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
53 * IP (one's complement) checksum for any combination
54 * of protocols or protocol layering. The checksum is
55 * computed and set in a packet per the CHECKSUM_PARTIAL
56 * interface (see below).
57 *
58 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
59 * TCP or UDP packets over IPv4. These are specifically
60 * unencapsulated packets of the form IPv4|TCP or
61 * IPv4|UDP where the Protocol field in the IPv4 header
62 * is TCP or UDP. The IPv4 header may contain IP options
63 * This feature cannot be set in features for a device
64 * with NETIF_F_HW_CSUM also set. This feature is being
65 * DEPRECATED (see below).
66 *
67 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
68 * TCP or UDP packets over IPv6. These are specifically
69 * unencapsulated packets of the form IPv6|TCP or
70 * IPv4|UDP where the Next Header field in the IPv6
71 * header is either TCP or UDP. IPv6 extension headers
72 * are not supported with this feature. This feature
73 * cannot be set in features for a device with
74 * NETIF_F_HW_CSUM also set. This feature is being
75 * DEPRECATED (see below).
76 *
77 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
78 * This flag is used only used to disable the RX checksum
79 * feature for a device. The stack will accept receive
80 * checksum indication in packets received on a device
81 * regardless of whether NETIF_F_RXCSUM is set.
82 *
83 * B. Checksumming of received packets by device. Indication of checksum
84 * verification is in set skb->ip_summed. Possible values are:
78ea85f1
DB
85 *
86 * CHECKSUM_NONE:
87 *
7a6ae71b 88 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
89 * The packet contains full (though not verified) checksum in packet but
90 * not in skb->csum. Thus, skb->csum is undefined in this case.
91 *
92 * CHECKSUM_UNNECESSARY:
93 *
94 * The hardware you're dealing with doesn't calculate the full checksum
95 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
96 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
97 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
98 * though. A driver or device must never modify the checksum field in the
99 * packet even if checksum is verified.
77cffe23
TH
100 *
101 * CHECKSUM_UNNECESSARY is applicable to following protocols:
102 * TCP: IPv6 and IPv4.
103 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
104 * zero UDP checksum for either IPv4 or IPv6, the networking stack
105 * may perform further validation in this case.
106 * GRE: only if the checksum is present in the header.
107 * SCTP: indicates the CRC in SCTP header has been validated.
b4759dcd 108 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
109 *
110 * skb->csum_level indicates the number of consecutive checksums found in
111 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
112 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
113 * and a device is able to verify the checksums for UDP (possibly zero),
114 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
115 * two. If the device were only able to verify the UDP checksum and not
116 * GRE, either because it doesn't support GRE checksum of because GRE
117 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
118 * not considered in this case).
78ea85f1
DB
119 *
120 * CHECKSUM_COMPLETE:
121 *
122 * This is the most generic way. The device supplied checksum of the _whole_
123 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
124 * hardware doesn't need to parse L3/L4 headers to implement this.
125 *
b4759dcd
DC
126 * Notes:
127 * - Even if device supports only some protocols, but is able to produce
128 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
129 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
130 *
131 * CHECKSUM_PARTIAL:
132 *
6edec0e6
TH
133 * A checksum is set up to be offloaded to a device as described in the
134 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 135 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
136 * on the same host, or it may be set in the input path in GRO or remote
137 * checksum offload. For the purposes of checksum verification, the checksum
138 * referred to by skb->csum_start + skb->csum_offset and any preceding
139 * checksums in the packet are considered verified. Any checksums in the
140 * packet that are after the checksum being offloaded are not considered to
141 * be verified.
78ea85f1 142 *
7a6ae71b
TH
143 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
144 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
145 *
146 * CHECKSUM_PARTIAL:
147 *
7a6ae71b 148 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 149 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
150 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
151 * csum_start and csum_offset values are valid values given the length and
152 * offset of the packet, however they should not attempt to validate that the
153 * checksum refers to a legitimate transport layer checksum-- it is the
154 * purview of the stack to validate that csum_start and csum_offset are set
155 * correctly.
156 *
157 * When the stack requests checksum offload for a packet, the driver MUST
158 * ensure that the checksum is set correctly. A driver can either offload the
159 * checksum calculation to the device, or call skb_checksum_help (in the case
160 * that the device does not support offload for a particular checksum).
161 *
162 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
163 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
164 * checksum offload capability.
165 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
166 * on network device checksumming capabilities: if a packet does not match
167 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
168 * csum_not_inet, see item D.) is called to resolve the checksum.
78ea85f1 169 *
7a6ae71b 170 * CHECKSUM_NONE:
78ea85f1 171 *
7a6ae71b
TH
172 * The skb was already checksummed by the protocol, or a checksum is not
173 * required.
78ea85f1
DB
174 *
175 * CHECKSUM_UNNECESSARY:
176 *
7a6ae71b
TH
177 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
178 * output.
78ea85f1 179 *
7a6ae71b
TH
180 * CHECKSUM_COMPLETE:
181 * Not used in checksum output. If a driver observes a packet with this value
182 * set in skbuff, if should treat as CHECKSUM_NONE being set.
183 *
184 * D. Non-IP checksum (CRC) offloads
185 *
186 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
187 * offloading the SCTP CRC in a packet. To perform this offload the stack
dba00306
DC
188 * will set set csum_start and csum_offset accordingly, set ip_summed to
189 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
190 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
191 * A driver that supports both IP checksum offload and SCTP CRC32c offload
192 * must verify which offload is configured for a packet by testing the
193 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
194 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
195 *
196 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
197 * offloading the FCOE CRC in a packet. To perform this offload the stack
198 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
199 * accordingly. Note the there is no indication in the skbuff that the
200 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
201 * both IP checksum offload and FCOE CRC offload must verify which offload
202 * is configured for a packet presumably by inspecting packet headers.
203 *
204 * E. Checksumming on output with GSO.
205 *
206 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
207 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
208 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
209 * part of the GSO operation is implied. If a checksum is being offloaded
210 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
211 * are set to refer to the outermost checksum being offload (two offloaded
212 * checksums are possible with UDP encapsulation).
78ea85f1
DB
213 */
214
60476372 215/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
216#define CHECKSUM_NONE 0
217#define CHECKSUM_UNNECESSARY 1
218#define CHECKSUM_COMPLETE 2
219#define CHECKSUM_PARTIAL 3
1da177e4 220
77cffe23
TH
221/* Maximum value in skb->csum_level */
222#define SKB_MAX_CSUM_LEVEL 3
223
0bec8c88 224#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 225#define SKB_WITH_OVERHEAD(X) \
deea84b0 226 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
227#define SKB_MAX_ORDER(X, ORDER) \
228 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
229#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
230#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
231
87fb4b7b
ED
232/* return minimum truesize of one skb containing X bytes of data */
233#define SKB_TRUESIZE(X) ((X) + \
234 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
235 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
236
1da177e4 237struct net_device;
716ea3a7 238struct scatterlist;
9c55e01c 239struct pipe_inode_info;
a8f820aa 240struct iov_iter;
fd11a83d 241struct napi_struct;
d58e468b
PP
242struct bpf_prog;
243union bpf_attr;
df5042f4 244struct skb_ext;
1da177e4 245
5f79e0f9 246#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
247struct nf_conntrack {
248 atomic_t use;
1da177e4 249};
5f79e0f9 250#endif
1da177e4 251
34666d46 252#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 253struct nf_bridge_info {
3eaf4025
FW
254 enum {
255 BRNF_PROTO_UNCHANGED,
256 BRNF_PROTO_8021Q,
257 BRNF_PROTO_PPPOE
7fb48c5b 258 } orig_proto:8;
72b1e5e4
FW
259 u8 pkt_otherhost:1;
260 u8 in_prerouting:1;
261 u8 bridged_dnat:1;
411ffb4f 262 __u16 frag_max_size;
bf1ac5ca 263 struct net_device *physindev;
63cdbc06
FW
264
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
7fb48c5b 267 union {
72b1e5e4 268 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
269 __be32 ipv4_daddr;
270 struct in6_addr ipv6_daddr;
72b1e5e4
FW
271
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
275 */
276 char neigh_header[8];
72b31f72 277 };
1da177e4
LT
278};
279#endif
280
1da177e4
LT
281struct sk_buff_head {
282 /* These two members must be first. */
283 struct sk_buff *next;
284 struct sk_buff *prev;
285
286 __u32 qlen;
287 spinlock_t lock;
288};
289
290struct sk_buff;
291
9d4dde52
IC
292/* To allow 64K frame to be packed as single skb without frag_list we
293 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
294 * buffers which do not start on a page boundary.
295 *
296 * Since GRO uses frags we allocate at least 16 regardless of page
297 * size.
a715dea3 298 */
9d4dde52 299#if (65536/PAGE_SIZE + 1) < 16
eec00954 300#define MAX_SKB_FRAGS 16UL
a715dea3 301#else
9d4dde52 302#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 303#endif
5f74f82e 304extern int sysctl_max_skb_frags;
1da177e4 305
3953c46c
MRL
306/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
307 * segment using its current segmentation instead.
308 */
309#define GSO_BY_FRAGS 0xFFFF
310
1da177e4
LT
311typedef struct skb_frag_struct skb_frag_t;
312
313struct skb_frag_struct {
a8605c60
IC
314 struct {
315 struct page *p;
316 } page;
cb4dfe56 317#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
318 __u32 page_offset;
319 __u32 size;
cb4dfe56
ED
320#else
321 __u16 page_offset;
322 __u16 size;
323#endif
1da177e4
LT
324};
325
161e6137
PT
326/**
327 * skb_frag_size - Returns the size of a skb fragment
328 * @frag: skb fragment
329 */
9e903e08
ED
330static inline unsigned int skb_frag_size(const skb_frag_t *frag)
331{
332 return frag->size;
333}
334
161e6137
PT
335/**
336 * skb_frag_size_set - Sets the size of a skb fragment
337 * @frag: skb fragment
338 * @size: size of fragment
339 */
9e903e08
ED
340static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
341{
342 frag->size = size;
343}
344
161e6137
PT
345/**
346 * skb_frag_size_add - Incrementes the size of a skb fragment by %delta
347 * @frag: skb fragment
348 * @delta: value to add
349 */
9e903e08
ED
350static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
351{
352 frag->size += delta;
353}
354
161e6137
PT
355/**
356 * skb_frag_size_sub - Decrements the size of a skb fragment by %delta
357 * @frag: skb fragment
358 * @delta: value to subtract
359 */
9e903e08
ED
360static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
361{
362 frag->size -= delta;
363}
364
161e6137
PT
365/**
366 * skb_frag_must_loop - Test if %p is a high memory page
367 * @p: fragment's page
368 */
c613c209
WB
369static inline bool skb_frag_must_loop(struct page *p)
370{
371#if defined(CONFIG_HIGHMEM)
372 if (PageHighMem(p))
373 return true;
374#endif
375 return false;
376}
377
378/**
379 * skb_frag_foreach_page - loop over pages in a fragment
380 *
381 * @f: skb frag to operate on
382 * @f_off: offset from start of f->page.p
383 * @f_len: length from f_off to loop over
384 * @p: (temp var) current page
385 * @p_off: (temp var) offset from start of current page,
386 * non-zero only on first page.
387 * @p_len: (temp var) length in current page,
388 * < PAGE_SIZE only on first and last page.
389 * @copied: (temp var) length so far, excluding current p_len.
390 *
391 * A fragment can hold a compound page, in which case per-page
392 * operations, notably kmap_atomic, must be called for each
393 * regular page.
394 */
395#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
396 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
397 p_off = (f_off) & (PAGE_SIZE - 1), \
398 p_len = skb_frag_must_loop(p) ? \
399 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
400 copied = 0; \
401 copied < f_len; \
402 copied += p_len, p++, p_off = 0, \
403 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
404
ac45f602
PO
405#define HAVE_HW_TIME_STAMP
406
407/**
d3a21be8 408 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
409 * @hwtstamp: hardware time stamp transformed into duration
410 * since arbitrary point in time
ac45f602
PO
411 *
412 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 413 * skb->tstamp.
ac45f602
PO
414 *
415 * hwtstamps can only be compared against other hwtstamps from
416 * the same device.
417 *
418 * This structure is attached to packets as part of the
419 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
420 */
421struct skb_shared_hwtstamps {
422 ktime_t hwtstamp;
ac45f602
PO
423};
424
2244d07b
OH
425/* Definitions for tx_flags in struct skb_shared_info */
426enum {
427 /* generate hardware time stamp */
428 SKBTX_HW_TSTAMP = 1 << 0,
429
e7fd2885 430 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
431 SKBTX_SW_TSTAMP = 1 << 1,
432
433 /* device driver is going to provide hardware time stamp */
434 SKBTX_IN_PROGRESS = 1 << 2,
435
a6686f2f 436 /* device driver supports TX zero-copy buffers */
62b1a8ab 437 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
438
439 /* generate wifi status information (where possible) */
62b1a8ab 440 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
441
442 /* This indicates at least one fragment might be overwritten
443 * (as in vmsplice(), sendfile() ...)
444 * If we need to compute a TX checksum, we'll need to copy
445 * all frags to avoid possible bad checksum
446 */
447 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
448
449 /* generate software time stamp when entering packet scheduling */
450 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
451};
452
52267790 453#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
e1c8a607 454#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 455 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
456#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
457
a6686f2f
SM
458/*
459 * The callback notifies userspace to release buffers when skb DMA is done in
460 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
461 * The zerocopy_success argument is true if zero copy transmit occurred,
462 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
463 * The ctx field is used to track device context.
464 * The desc field is used to track userspace buffer index.
a6686f2f
SM
465 */
466struct ubuf_info {
e19d6763 467 void (*callback)(struct ubuf_info *, bool zerocopy_success);
4ab6c99d
WB
468 union {
469 struct {
470 unsigned long desc;
471 void *ctx;
472 };
473 struct {
474 u32 id;
475 u16 len;
476 u16 zerocopy:1;
477 u32 bytelen;
478 };
479 };
c1d1b437 480 refcount_t refcnt;
a91dbff5
WB
481
482 struct mmpin {
483 struct user_struct *user;
484 unsigned int num_pg;
485 } mmp;
ac45f602
PO
486};
487
52267790
WB
488#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
489
6f89dbce
SV
490int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
491void mm_unaccount_pinned_pages(struct mmpin *mmp);
492
52267790 493struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
4ab6c99d
WB
494struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
495 struct ubuf_info *uarg);
52267790
WB
496
497static inline void sock_zerocopy_get(struct ubuf_info *uarg)
498{
c1d1b437 499 refcount_inc(&uarg->refcnt);
52267790
WB
500}
501
502void sock_zerocopy_put(struct ubuf_info *uarg);
52900d22 503void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790
WB
504
505void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
506
b5947e5d 507int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
508int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
509 struct msghdr *msg, int len,
510 struct ubuf_info *uarg);
511
1da177e4
LT
512/* This data is invariant across clones and lives at
513 * the end of the header data, ie. at skb->end.
514 */
515struct skb_shared_info {
de8f3a83
DB
516 __u8 __unused;
517 __u8 meta_len;
518 __u8 nr_frags;
9f42f126 519 __u8 tx_flags;
7967168c
HX
520 unsigned short gso_size;
521 /* Warning: this field is not always filled in (UFO)! */
522 unsigned short gso_segs;
1da177e4 523 struct sk_buff *frag_list;
ac45f602 524 struct skb_shared_hwtstamps hwtstamps;
7f564528 525 unsigned int gso_type;
09c2d251 526 u32 tskey;
ec7d2f2c
ED
527
528 /*
529 * Warning : all fields before dataref are cleared in __alloc_skb()
530 */
531 atomic_t dataref;
532
69e3c75f
JB
533 /* Intermediate layers must ensure that destructor_arg
534 * remains valid until skb destructor */
535 void * destructor_arg;
a6686f2f 536
fed66381
ED
537 /* must be last field, see pskb_expand_head() */
538 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
539};
540
541/* We divide dataref into two halves. The higher 16 bits hold references
542 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
543 * the entire skb->data. A clone of a headerless skb holds the length of
544 * the header in skb->hdr_len.
1da177e4
LT
545 *
546 * All users must obey the rule that the skb->data reference count must be
547 * greater than or equal to the payload reference count.
548 *
549 * Holding a reference to the payload part means that the user does not
550 * care about modifications to the header part of skb->data.
551 */
552#define SKB_DATAREF_SHIFT 16
553#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
554
d179cd12
DM
555
556enum {
c8753d55
VS
557 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
558 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
559 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
560};
561
7967168c
HX
562enum {
563 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
564
565 /* This indicates the skb is from an untrusted source. */
d9d30adf 566 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
567
568 /* This indicates the tcp segment has CWR set. */
d9d30adf 569 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 570
d9d30adf 571 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 572
d9d30adf 573 SKB_GSO_TCPV6 = 1 << 4,
68c33163 574
d9d30adf 575 SKB_GSO_FCOE = 1 << 5,
73136267 576
d9d30adf 577 SKB_GSO_GRE = 1 << 6,
0d89d203 578
d9d30adf 579 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 580
d9d30adf 581 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 582
d9d30adf 583 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 584
d9d30adf 585 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 586
d9d30adf 587 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 588
d9d30adf 589 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 590
d9d30adf 591 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 592
d9d30adf 593 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 594
d9d30adf 595 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
596
597 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
598
599 SKB_GSO_UDP_L4 = 1 << 17,
7967168c
HX
600};
601
2e07fa9c
ACM
602#if BITS_PER_LONG > 32
603#define NET_SKBUFF_DATA_USES_OFFSET 1
604#endif
605
606#ifdef NET_SKBUFF_DATA_USES_OFFSET
607typedef unsigned int sk_buff_data_t;
608#else
609typedef unsigned char *sk_buff_data_t;
610#endif
611
161e6137 612/**
1da177e4
LT
613 * struct sk_buff - socket buffer
614 * @next: Next buffer in list
615 * @prev: Previous buffer in list
363ec392 616 * @tstamp: Time we arrived/left
56b17425 617 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 618 * @sk: Socket we are owned by
1da177e4 619 * @dev: Device we arrived on/are leaving by
d84e0bd7 620 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 621 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 622 * @sp: the security path, used for xfrm
1da177e4
LT
623 * @len: Length of actual data
624 * @data_len: Data length
625 * @mac_len: Length of link layer header
334a8132 626 * @hdr_len: writable header length of cloned skb
663ead3b
HX
627 * @csum: Checksum (must include start/offset pair)
628 * @csum_start: Offset from skb->head where checksumming should start
629 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 630 * @priority: Packet queueing priority
60ff7467 631 * @ignore_df: allow local fragmentation
1da177e4 632 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 633 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
634 * @nohdr: Payload reference only, must not modify header
635 * @pkt_type: Packet class
c83c2486 636 * @fclone: skbuff clone status
c83c2486 637 * @ipvs_property: skbuff is owned by ipvs
875e8939
IS
638 * @offload_fwd_mark: Packet was L2-forwarded in hardware
639 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 640 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 641 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
642 * @tc_redirected: packet was redirected by a tc action
643 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
644 * @peeked: this packet has been seen already, so stats have been
645 * done for it, don't do them again
ba9dda3a 646 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
647 * @protocol: Packet protocol from driver
648 * @destructor: Destruct function
e2080072 649 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 650 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 651 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 652 * @skb_iif: ifindex of device we arrived on
1da177e4 653 * @tc_index: Traffic control index
61b905da 654 * @hash: the packet hash
d84e0bd7 655 * @queue_mapping: Queue mapping for multiqueue devices
8b700862 656 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
df5042f4 657 * @active_extensions: active extensions (skb_ext_id types)
553a5672 658 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 659 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 660 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 661 * ports.
a3b18ddb 662 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
663 * @wifi_acked_valid: wifi_acked was set
664 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 665 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
dba00306 666 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
4ff06203 667 * @dst_pending_confirm: need to confirm neighbour
a48d189e 668 * @decrypted: Decrypted SKB
161e6137 669 * @napi_id: id of the NAPI struct this skb came from
984bc16c 670 * @secmark: security marking
d84e0bd7 671 * @mark: Generic packet mark
86a9bad3 672 * @vlan_proto: vlan encapsulation protocol
6aa895b0 673 * @vlan_tci: vlan tag control information
0d89d203 674 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
675 * @inner_transport_header: Inner transport layer header (encapsulation)
676 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 677 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
678 * @transport_header: Transport layer header
679 * @network_header: Network layer header
680 * @mac_header: Link layer header
681 * @tail: Tail pointer
682 * @end: End pointer
683 * @head: Head of buffer
684 * @data: Data head pointer
685 * @truesize: Buffer size
686 * @users: User count - see {datagram,tcp}.c
df5042f4 687 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
688 */
689
690struct sk_buff {
363ec392 691 union {
56b17425
ED
692 struct {
693 /* These two members must be first. */
694 struct sk_buff *next;
695 struct sk_buff *prev;
696
697 union {
bffa72cf
ED
698 struct net_device *dev;
699 /* Some protocols might use this space to store information,
700 * while device pointer would be NULL.
701 * UDP receive path is one user.
702 */
703 unsigned long dev_scratch;
56b17425
ED
704 };
705 };
fa0f5273 706 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 707 struct list_head list;
363ec392 708 };
fa0f5273
PO
709
710 union {
711 struct sock *sk;
712 int ip_defrag_offset;
713 };
1da177e4 714
c84d9490 715 union {
bffa72cf 716 ktime_t tstamp;
d3edd06e 717 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 718 };
1da177e4
LT
719 /*
720 * This is the control buffer. It is free to use for every
721 * layer. Please put your private variables there. If you
722 * want to keep them across layers you have to do a skb_clone()
723 * first. This is owned by whoever has the skb queued ATM.
724 */
da3f5cf1 725 char cb[48] __aligned(8);
1da177e4 726
e2080072
ED
727 union {
728 struct {
729 unsigned long _skb_refdst;
730 void (*destructor)(struct sk_buff *skb);
731 };
732 struct list_head tcp_tsorted_anchor;
733 };
734
b1937227 735#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 736 unsigned long _nfct;
da3f5cf1 737#endif
1da177e4 738 unsigned int len,
334a8132
PM
739 data_len;
740 __u16 mac_len,
741 hdr_len;
b1937227
ED
742
743 /* Following fields are _not_ copied in __copy_skb_header()
744 * Note that queue_mapping is here mostly to fill a hole.
745 */
b1937227 746 __u16 queue_mapping;
36bbef52
DB
747
748/* if you move cloned around you also must adapt those constants */
749#ifdef __BIG_ENDIAN_BITFIELD
750#define CLONED_MASK (1 << 7)
751#else
752#define CLONED_MASK 1
753#endif
754#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
755
756 __u8 __cloned_offset[0];
b1937227 757 __u8 cloned:1,
6869c4d8 758 nohdr:1,
b84f4cc9 759 fclone:2,
a59322be 760 peeked:1,
b1937227 761 head_frag:1,
8b700862 762 pfmemalloc:1;
df5042f4
FW
763#ifdef CONFIG_SKB_EXTENSIONS
764 __u8 active_extensions;
765#endif
b1937227
ED
766 /* fields enclosed in headers_start/headers_end are copied
767 * using a single memcpy() in __copy_skb_header()
768 */
ebcf34f3 769 /* private: */
b1937227 770 __u32 headers_start[0];
ebcf34f3 771 /* public: */
4031ae6e 772
233577a2
HFS
773/* if you move pkt_type around you also must adapt those constants */
774#ifdef __BIG_ENDIAN_BITFIELD
775#define PKT_TYPE_MAX (7 << 5)
776#else
777#define PKT_TYPE_MAX 7
1da177e4 778#endif
233577a2 779#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 780
233577a2 781 __u8 __pkt_type_offset[0];
b1937227 782 __u8 pkt_type:3;
b1937227 783 __u8 ignore_df:1;
b1937227
ED
784 __u8 nf_trace:1;
785 __u8 ip_summed:2;
3853b584 786 __u8 ooo_okay:1;
8b700862 787
61b905da 788 __u8 l4_hash:1;
a3b18ddb 789 __u8 sw_hash:1;
6e3e939f
JB
790 __u8 wifi_acked_valid:1;
791 __u8 wifi_acked:1;
3bdc0eba 792 __u8 no_fcs:1;
77cffe23 793 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 794 __u8 encapsulation:1;
7e2b10c1 795 __u8 encap_hdr_csum:1;
5d0c2b95 796 __u8 csum_valid:1;
8b700862 797
0c4b2d37
MM
798#ifdef __BIG_ENDIAN_BITFIELD
799#define PKT_VLAN_PRESENT_BIT 7
800#else
801#define PKT_VLAN_PRESENT_BIT 0
802#endif
803#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
804 __u8 __pkt_vlan_present_offset[0];
805 __u8 vlan_present:1;
7e3cead5 806 __u8 csum_complete_sw:1;
b1937227 807 __u8 csum_level:2;
dba00306 808 __u8 csum_not_inet:1;
4ff06203 809 __u8 dst_pending_confirm:1;
b1937227
ED
810#ifdef CONFIG_IPV6_NDISC_NODETYPE
811 __u8 ndisc_nodetype:2;
812#endif
8b700862 813
0c4b2d37 814 __u8 ipvs_property:1;
8bce6d7d 815 __u8 inner_protocol_type:1;
e585f236 816 __u8 remcsum_offload:1;
6bc506b4
IS
817#ifdef CONFIG_NET_SWITCHDEV
818 __u8 offload_fwd_mark:1;
875e8939 819 __u8 offload_l3_fwd_mark:1;
6bc506b4 820#endif
e7246e12
WB
821#ifdef CONFIG_NET_CLS_ACT
822 __u8 tc_skip_classify:1;
8dc07fdb 823 __u8 tc_at_ingress:1;
bc31c905
WB
824 __u8 tc_redirected:1;
825 __u8 tc_from_ingress:1;
e7246e12 826#endif
a48d189e
SB
827#ifdef CONFIG_TLS_DEVICE
828 __u8 decrypted:1;
829#endif
b1937227
ED
830
831#ifdef CONFIG_NET_SCHED
832 __u16 tc_index; /* traffic control index */
b1937227 833#endif
fe55f6d5 834
b1937227
ED
835 union {
836 __wsum csum;
837 struct {
838 __u16 csum_start;
839 __u16 csum_offset;
840 };
841 };
842 __u32 priority;
843 int skb_iif;
844 __u32 hash;
845 __be16 vlan_proto;
846 __u16 vlan_tci;
2bd82484
ED
847#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
848 union {
849 unsigned int napi_id;
850 unsigned int sender_cpu;
851 };
97fc2f08 852#endif
984bc16c 853#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 854 __u32 secmark;
0c4f691f 855#endif
0c4f691f 856
3b885787
NH
857 union {
858 __u32 mark;
16fad69c 859 __u32 reserved_tailroom;
3b885787 860 };
1da177e4 861
8bce6d7d
TH
862 union {
863 __be16 inner_protocol;
864 __u8 inner_ipproto;
865 };
866
1a37e412
SH
867 __u16 inner_transport_header;
868 __u16 inner_network_header;
869 __u16 inner_mac_header;
b1937227
ED
870
871 __be16 protocol;
1a37e412
SH
872 __u16 transport_header;
873 __u16 network_header;
874 __u16 mac_header;
b1937227 875
ebcf34f3 876 /* private: */
b1937227 877 __u32 headers_end[0];
ebcf34f3 878 /* public: */
b1937227 879
1da177e4 880 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 881 sk_buff_data_t tail;
4305b541 882 sk_buff_data_t end;
1da177e4 883 unsigned char *head,
4305b541 884 *data;
27a884dc 885 unsigned int truesize;
63354797 886 refcount_t users;
df5042f4
FW
887
888#ifdef CONFIG_SKB_EXTENSIONS
889 /* only useable after checking ->active_extensions != 0 */
890 struct skb_ext *extensions;
891#endif
1da177e4
LT
892};
893
894#ifdef __KERNEL__
895/*
896 * Handling routines are only of interest to the kernel
897 */
1da177e4 898
c93bdd0e
MG
899#define SKB_ALLOC_FCLONE 0x01
900#define SKB_ALLOC_RX 0x02
fd11a83d 901#define SKB_ALLOC_NAPI 0x04
c93bdd0e 902
161e6137
PT
903/**
904 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
905 * @skb: buffer
906 */
c93bdd0e
MG
907static inline bool skb_pfmemalloc(const struct sk_buff *skb)
908{
909 return unlikely(skb->pfmemalloc);
910}
911
7fee226a
ED
912/*
913 * skb might have a dst pointer attached, refcounted or not.
914 * _skb_refdst low order bit is set if refcount was _not_ taken
915 */
916#define SKB_DST_NOREF 1UL
917#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
918
a9e419dc 919#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
920/**
921 * skb_dst - returns skb dst_entry
922 * @skb: buffer
923 *
924 * Returns skb dst_entry, regardless of reference taken or not.
925 */
adf30907
ED
926static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
927{
161e6137 928 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
929 * rcu_read_lock section
930 */
931 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
932 !rcu_read_lock_held() &&
933 !rcu_read_lock_bh_held());
934 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
935}
936
7fee226a
ED
937/**
938 * skb_dst_set - sets skb dst
939 * @skb: buffer
940 * @dst: dst entry
941 *
942 * Sets skb dst, assuming a reference was taken on dst and should
943 * be released by skb_dst_drop()
944 */
adf30907
ED
945static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
946{
7fee226a
ED
947 skb->_skb_refdst = (unsigned long)dst;
948}
949
932bc4d7
JA
950/**
951 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
952 * @skb: buffer
953 * @dst: dst entry
954 *
955 * Sets skb dst, assuming a reference was not taken on dst.
956 * If dst entry is cached, we do not take reference and dst_release
957 * will be avoided by refdst_drop. If dst entry is not cached, we take
958 * reference, so that last dst_release can destroy the dst immediately.
959 */
960static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
961{
dbfc4fb7
HFS
962 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
963 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 964}
7fee226a
ED
965
966/**
25985edc 967 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
968 * @skb: buffer
969 */
970static inline bool skb_dst_is_noref(const struct sk_buff *skb)
971{
972 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
973}
974
161e6137
PT
975/**
976 * skb_rtable - Returns the skb &rtable
977 * @skb: buffer
978 */
511c3f92
ED
979static inline struct rtable *skb_rtable(const struct sk_buff *skb)
980{
adf30907 981 return (struct rtable *)skb_dst(skb);
511c3f92
ED
982}
983
8b10cab6
JHS
984/* For mangling skb->pkt_type from user space side from applications
985 * such as nft, tc, etc, we only allow a conservative subset of
986 * possible pkt_types to be set.
987*/
988static inline bool skb_pkt_type_ok(u32 ptype)
989{
990 return ptype <= PACKET_OTHERHOST;
991}
992
161e6137
PT
993/**
994 * skb_napi_id - Returns the skb's NAPI id
995 * @skb: buffer
996 */
90b602f8
ML
997static inline unsigned int skb_napi_id(const struct sk_buff *skb)
998{
999#ifdef CONFIG_NET_RX_BUSY_POLL
1000 return skb->napi_id;
1001#else
1002 return 0;
1003#endif
1004}
1005
161e6137
PT
1006/**
1007 * skb_unref - decrement the skb's reference count
1008 * @skb: buffer
1009 *
1010 * Returns true if we can free the skb.
1011 */
3889a803
PA
1012static inline bool skb_unref(struct sk_buff *skb)
1013{
1014 if (unlikely(!skb))
1015 return false;
63354797 1016 if (likely(refcount_read(&skb->users) == 1))
3889a803 1017 smp_rmb();
63354797 1018 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1019 return false;
1020
1021 return true;
1022}
1023
0a463c78 1024void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1025void kfree_skb(struct sk_buff *skb);
1026void kfree_skb_list(struct sk_buff *segs);
6413139d 1027void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d
JP
1028void skb_tx_error(struct sk_buff *skb);
1029void consume_skb(struct sk_buff *skb);
ca2c1418 1030void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1031void __kfree_skb(struct sk_buff *skb);
d7e8883c 1032extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1033
7965bd4d
JP
1034void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1035bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1036 bool *fragstolen, int *delta_truesize);
bad43ca8 1037
7965bd4d
JP
1038struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1039 int node);
2ea2f62c 1040struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1041struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1042struct sk_buff *build_skb_around(struct sk_buff *skb,
1043 void *data, unsigned int frag_size);
161e6137
PT
1044
1045/**
1046 * alloc_skb - allocate a network buffer
1047 * @size: size to allocate
1048 * @priority: allocation mask
1049 *
1050 * This function is a convenient wrapper around __alloc_skb().
1051 */
d179cd12 1052static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1053 gfp_t priority)
d179cd12 1054{
564824b0 1055 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1056}
1057
2e4e4410
ED
1058struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1059 unsigned long data_len,
1060 int max_page_order,
1061 int *errcode,
1062 gfp_t gfp_mask);
da29e4b4 1063struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1064
d0bf4a9e
ED
1065/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1066struct sk_buff_fclones {
1067 struct sk_buff skb1;
1068
1069 struct sk_buff skb2;
1070
2638595a 1071 refcount_t fclone_ref;
d0bf4a9e
ED
1072};
1073
1074/**
1075 * skb_fclone_busy - check if fclone is busy
293de7de 1076 * @sk: socket
d0bf4a9e
ED
1077 * @skb: buffer
1078 *
bda13fed 1079 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1080 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1081 * so we also check that this didnt happen.
d0bf4a9e 1082 */
39bb5e62
ED
1083static inline bool skb_fclone_busy(const struct sock *sk,
1084 const struct sk_buff *skb)
d0bf4a9e
ED
1085{
1086 const struct sk_buff_fclones *fclones;
1087
1088 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1089
1090 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1091 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1092 fclones->skb2.sk == sk;
d0bf4a9e
ED
1093}
1094
161e6137
PT
1095/**
1096 * alloc_skb_fclone - allocate a network buffer from fclone cache
1097 * @size: size to allocate
1098 * @priority: allocation mask
1099 *
1100 * This function is a convenient wrapper around __alloc_skb().
1101 */
d179cd12 1102static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1103 gfp_t priority)
d179cd12 1104{
c93bdd0e 1105 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1106}
1107
7965bd4d 1108struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1109void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1110int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1111struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1112void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1113struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1114struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1115 gfp_t gfp_mask, bool fclone);
1116static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1117 gfp_t gfp_mask)
1118{
1119 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1120}
7965bd4d
JP
1121
1122int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1123struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1124 unsigned int headroom);
1125struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1126 int newtailroom, gfp_t priority);
48a1df65
JD
1127int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1128 int offset, int len);
1129int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1130 int offset, int len);
7965bd4d 1131int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1132int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1133
1134/**
1135 * skb_pad - zero pad the tail of an skb
1136 * @skb: buffer to pad
1137 * @pad: space to pad
1138 *
1139 * Ensure that a buffer is followed by a padding area that is zero
1140 * filled. Used by network drivers which may DMA or transfer data
1141 * beyond the buffer end onto the wire.
1142 *
1143 * May return error in out of memory cases. The skb is freed on error.
1144 */
1145static inline int skb_pad(struct sk_buff *skb, int pad)
1146{
1147 return __skb_pad(skb, pad, true);
1148}
ead2ceb0 1149#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1150
be12a1fe
HFS
1151int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1152 int offset, size_t size);
1153
d94d9fee 1154struct skb_seq_state {
677e90ed
TG
1155 __u32 lower_offset;
1156 __u32 upper_offset;
1157 __u32 frag_idx;
1158 __u32 stepped_offset;
1159 struct sk_buff *root_skb;
1160 struct sk_buff *cur_skb;
1161 __u8 *frag_data;
1162};
1163
7965bd4d
JP
1164void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1165 unsigned int to, struct skb_seq_state *st);
1166unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1167 struct skb_seq_state *st);
1168void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1169
7965bd4d 1170unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1171 unsigned int to, struct ts_config *config);
3fc7e8a6 1172
09323cc4
TH
1173/*
1174 * Packet hash types specify the type of hash in skb_set_hash.
1175 *
1176 * Hash types refer to the protocol layer addresses which are used to
1177 * construct a packet's hash. The hashes are used to differentiate or identify
1178 * flows of the protocol layer for the hash type. Hash types are either
1179 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1180 *
1181 * Properties of hashes:
1182 *
1183 * 1) Two packets in different flows have different hash values
1184 * 2) Two packets in the same flow should have the same hash value
1185 *
1186 * A hash at a higher layer is considered to be more specific. A driver should
1187 * set the most specific hash possible.
1188 *
1189 * A driver cannot indicate a more specific hash than the layer at which a hash
1190 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1191 *
1192 * A driver may indicate a hash level which is less specific than the
1193 * actual layer the hash was computed on. For instance, a hash computed
1194 * at L4 may be considered an L3 hash. This should only be done if the
1195 * driver can't unambiguously determine that the HW computed the hash at
1196 * the higher layer. Note that the "should" in the second property above
1197 * permits this.
1198 */
1199enum pkt_hash_types {
1200 PKT_HASH_TYPE_NONE, /* Undefined type */
1201 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1202 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1203 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1204};
1205
bcc83839 1206static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1207{
bcc83839 1208 skb->hash = 0;
a3b18ddb 1209 skb->sw_hash = 0;
bcc83839
TH
1210 skb->l4_hash = 0;
1211}
1212
1213static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1214{
1215 if (!skb->l4_hash)
1216 skb_clear_hash(skb);
1217}
1218
1219static inline void
1220__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1221{
1222 skb->l4_hash = is_l4;
1223 skb->sw_hash = is_sw;
61b905da 1224 skb->hash = hash;
09323cc4
TH
1225}
1226
bcc83839
TH
1227static inline void
1228skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1229{
1230 /* Used by drivers to set hash from HW */
1231 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1232}
1233
1234static inline void
1235__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1236{
1237 __skb_set_hash(skb, hash, true, is_l4);
1238}
1239
e5276937 1240void __skb_get_hash(struct sk_buff *skb);
b917783c 1241u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1242u32 skb_get_poff(const struct sk_buff *skb);
1243u32 __skb_get_poff(const struct sk_buff *skb, void *data,
72a338bc 1244 const struct flow_keys_basic *keys, int hlen);
e5276937
TH
1245__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1246 void *data, int hlen_proto);
1247
1248static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1249 int thoff, u8 ip_proto)
1250{
1251 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1252}
1253
1254void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1255 const struct flow_dissector_key *key,
1256 unsigned int key_count);
1257
2dfd184a 1258#ifdef CONFIG_NET
118c8e9a
SF
1259int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1260 union bpf_attr __user *uattr);
d58e468b
PP
1261int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1262 struct bpf_prog *prog);
1263
1264int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
2dfd184a 1265#else
118c8e9a
SF
1266static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1267 union bpf_attr __user *uattr)
1268{
1269 return -EOPNOTSUPP;
1270}
1271
2dfd184a
WB
1272static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1273 struct bpf_prog *prog)
1274{
1275 return -EOPNOTSUPP;
1276}
1277
1278static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1279{
1280 return -EOPNOTSUPP;
1281}
1282#endif
d58e468b 1283
089b19a9
SF
1284struct bpf_flow_dissector;
1285bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1286 __be16 proto, int nhoff, int hlen);
1287
3cbf4ffb
SF
1288bool __skb_flow_dissect(const struct net *net,
1289 const struct sk_buff *skb,
e5276937
TH
1290 struct flow_dissector *flow_dissector,
1291 void *target_container,
cd79a238
TH
1292 void *data, __be16 proto, int nhoff, int hlen,
1293 unsigned int flags);
e5276937
TH
1294
1295static inline bool skb_flow_dissect(const struct sk_buff *skb,
1296 struct flow_dissector *flow_dissector,
cd79a238 1297 void *target_container, unsigned int flags)
e5276937 1298{
3cbf4ffb
SF
1299 return __skb_flow_dissect(NULL, skb, flow_dissector,
1300 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1301}
1302
1303static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1304 struct flow_keys *flow,
1305 unsigned int flags)
e5276937
TH
1306{
1307 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1308 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1309 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1310}
1311
72a338bc 1312static inline bool
3cbf4ffb
SF
1313skb_flow_dissect_flow_keys_basic(const struct net *net,
1314 const struct sk_buff *skb,
72a338bc
PA
1315 struct flow_keys_basic *flow, void *data,
1316 __be16 proto, int nhoff, int hlen,
1317 unsigned int flags)
e5276937
TH
1318{
1319 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1320 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1321 data, proto, nhoff, hlen, flags);
e5276937
TH
1322}
1323
82828b88
JP
1324void skb_flow_dissect_meta(const struct sk_buff *skb,
1325 struct flow_dissector *flow_dissector,
1326 void *target_container);
1327
75a56758
PB
1328/* Gets a skb connection tracking info, ctinfo map should be a
1329 * a map of mapsize to translate enum ip_conntrack_info states
1330 * to user states.
1331 */
1332void
1333skb_flow_dissect_ct(const struct sk_buff *skb,
1334 struct flow_dissector *flow_dissector,
1335 void *target_container,
1336 u16 *ctinfo_map,
1337 size_t mapsize);
62b32379
SH
1338void
1339skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1340 struct flow_dissector *flow_dissector,
1341 void *target_container);
1342
3958afa1 1343static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1344{
a3b18ddb 1345 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1346 __skb_get_hash(skb);
bfb564e7 1347
61b905da 1348 return skb->hash;
bfb564e7
KK
1349}
1350
20a17bf6 1351static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1352{
c6cc1ca7
TH
1353 if (!skb->l4_hash && !skb->sw_hash) {
1354 struct flow_keys keys;
de4c1f8b 1355 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1356
de4c1f8b 1357 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1358 }
f70ea018
TH
1359
1360 return skb->hash;
1361}
1362
9d0b4021
ED
1363__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1364 const siphash_key_t *perturb);
50fb7992 1365
57bdf7f4
TH
1366static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1367{
61b905da 1368 return skb->hash;
57bdf7f4
TH
1369}
1370
3df7a74e
TH
1371static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1372{
61b905da 1373 to->hash = from->hash;
a3b18ddb 1374 to->sw_hash = from->sw_hash;
61b905da 1375 to->l4_hash = from->l4_hash;
3df7a74e
TH
1376};
1377
41477662
JK
1378static inline void skb_copy_decrypted(struct sk_buff *to,
1379 const struct sk_buff *from)
1380{
1381#ifdef CONFIG_TLS_DEVICE
1382 to->decrypted = from->decrypted;
1383#endif
1384}
1385
4305b541
ACM
1386#ifdef NET_SKBUFF_DATA_USES_OFFSET
1387static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1388{
1389 return skb->head + skb->end;
1390}
ec47ea82
AD
1391
1392static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1393{
1394 return skb->end;
1395}
4305b541
ACM
1396#else
1397static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1398{
1399 return skb->end;
1400}
ec47ea82
AD
1401
1402static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1403{
1404 return skb->end - skb->head;
1405}
4305b541
ACM
1406#endif
1407
1da177e4 1408/* Internal */
4305b541 1409#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1410
ac45f602
PO
1411static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1412{
1413 return &skb_shinfo(skb)->hwtstamps;
1414}
1415
52267790
WB
1416static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1417{
1418 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1419
1420 return is_zcopy ? skb_uarg(skb) : NULL;
1421}
1422
52900d22
WB
1423static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1424 bool *have_ref)
52267790
WB
1425{
1426 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1427 if (unlikely(have_ref && *have_ref))
1428 *have_ref = false;
1429 else
1430 sock_zerocopy_get(uarg);
52267790
WB
1431 skb_shinfo(skb)->destructor_arg = uarg;
1432 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1433 }
1434}
1435
5cd8d46e
WB
1436static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1437{
1438 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1439 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1440}
1441
1442static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1443{
1444 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1445}
1446
1447static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1448{
1449 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1450}
1451
52267790
WB
1452/* Release a reference on a zerocopy structure */
1453static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1454{
1455 struct ubuf_info *uarg = skb_zcopy(skb);
1456
1457 if (uarg) {
185ce5c3
WB
1458 if (skb_zcopy_is_nouarg(skb)) {
1459 /* no notification callback */
1460 } else if (uarg->callback == sock_zerocopy_callback) {
0a4a060b
WB
1461 uarg->zerocopy = uarg->zerocopy && zerocopy;
1462 sock_zerocopy_put(uarg);
185ce5c3 1463 } else {
0a4a060b
WB
1464 uarg->callback(uarg, zerocopy);
1465 }
1466
52267790
WB
1467 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1468 }
1469}
1470
1471/* Abort a zerocopy operation and revert zckey on error in send syscall */
1472static inline void skb_zcopy_abort(struct sk_buff *skb)
1473{
1474 struct ubuf_info *uarg = skb_zcopy(skb);
1475
1476 if (uarg) {
52900d22 1477 sock_zerocopy_put_abort(uarg, false);
52267790
WB
1478 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1479 }
1480}
1481
a8305bff
DM
1482static inline void skb_mark_not_on_list(struct sk_buff *skb)
1483{
1484 skb->next = NULL;
1485}
1486
992cba7e
DM
1487static inline void skb_list_del_init(struct sk_buff *skb)
1488{
1489 __list_del_entry(&skb->list);
1490 skb_mark_not_on_list(skb);
1491}
1492
1da177e4
LT
1493/**
1494 * skb_queue_empty - check if a queue is empty
1495 * @list: queue head
1496 *
1497 * Returns true if the queue is empty, false otherwise.
1498 */
1499static inline int skb_queue_empty(const struct sk_buff_head *list)
1500{
fd44b93c 1501 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1502}
1503
a63dd641
ED
1504/**
1505 * skb_queue_empty_lockless - check if a queue is empty
1506 * @list: queue head
1507 *
1508 * Returns true if the queue is empty, false otherwise.
1509 * This variant can be used in lockless contexts.
1510 */
1511static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1512{
1513 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1514}
1515
1516
fc7ebb21
DM
1517/**
1518 * skb_queue_is_last - check if skb is the last entry in the queue
1519 * @list: queue head
1520 * @skb: buffer
1521 *
1522 * Returns true if @skb is the last buffer on the list.
1523 */
1524static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1525 const struct sk_buff *skb)
1526{
fd44b93c 1527 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1528}
1529
832d11c5
IJ
1530/**
1531 * skb_queue_is_first - check if skb is the first entry in the queue
1532 * @list: queue head
1533 * @skb: buffer
1534 *
1535 * Returns true if @skb is the first buffer on the list.
1536 */
1537static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1538 const struct sk_buff *skb)
1539{
fd44b93c 1540 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1541}
1542
249c8b42
DM
1543/**
1544 * skb_queue_next - return the next packet in the queue
1545 * @list: queue head
1546 * @skb: current buffer
1547 *
1548 * Return the next packet in @list after @skb. It is only valid to
1549 * call this if skb_queue_is_last() evaluates to false.
1550 */
1551static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1552 const struct sk_buff *skb)
1553{
1554 /* This BUG_ON may seem severe, but if we just return then we
1555 * are going to dereference garbage.
1556 */
1557 BUG_ON(skb_queue_is_last(list, skb));
1558 return skb->next;
1559}
1560
832d11c5
IJ
1561/**
1562 * skb_queue_prev - return the prev packet in the queue
1563 * @list: queue head
1564 * @skb: current buffer
1565 *
1566 * Return the prev packet in @list before @skb. It is only valid to
1567 * call this if skb_queue_is_first() evaluates to false.
1568 */
1569static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1570 const struct sk_buff *skb)
1571{
1572 /* This BUG_ON may seem severe, but if we just return then we
1573 * are going to dereference garbage.
1574 */
1575 BUG_ON(skb_queue_is_first(list, skb));
1576 return skb->prev;
1577}
1578
1da177e4
LT
1579/**
1580 * skb_get - reference buffer
1581 * @skb: buffer to reference
1582 *
1583 * Makes another reference to a socket buffer and returns a pointer
1584 * to the buffer.
1585 */
1586static inline struct sk_buff *skb_get(struct sk_buff *skb)
1587{
63354797 1588 refcount_inc(&skb->users);
1da177e4
LT
1589 return skb;
1590}
1591
1592/*
f8821f96 1593 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1594 */
1595
1da177e4
LT
1596/**
1597 * skb_cloned - is the buffer a clone
1598 * @skb: buffer to check
1599 *
1600 * Returns true if the buffer was generated with skb_clone() and is
1601 * one of multiple shared copies of the buffer. Cloned buffers are
1602 * shared data so must not be written to under normal circumstances.
1603 */
1604static inline int skb_cloned(const struct sk_buff *skb)
1605{
1606 return skb->cloned &&
1607 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1608}
1609
14bbd6a5
PS
1610static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1611{
d0164adc 1612 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1613
1614 if (skb_cloned(skb))
1615 return pskb_expand_head(skb, 0, 0, pri);
1616
1617 return 0;
1618}
1619
1da177e4
LT
1620/**
1621 * skb_header_cloned - is the header a clone
1622 * @skb: buffer to check
1623 *
1624 * Returns true if modifying the header part of the buffer requires
1625 * the data to be copied.
1626 */
1627static inline int skb_header_cloned(const struct sk_buff *skb)
1628{
1629 int dataref;
1630
1631 if (!skb->cloned)
1632 return 0;
1633
1634 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1635 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1636 return dataref != 1;
1637}
1638
9580bf2e
ED
1639static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1640{
1641 might_sleep_if(gfpflags_allow_blocking(pri));
1642
1643 if (skb_header_cloned(skb))
1644 return pskb_expand_head(skb, 0, 0, pri);
1645
1646 return 0;
1647}
1648
f4a775d1
ED
1649/**
1650 * __skb_header_release - release reference to header
1651 * @skb: buffer to operate on
f4a775d1
ED
1652 */
1653static inline void __skb_header_release(struct sk_buff *skb)
1654{
1655 skb->nohdr = 1;
1656 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1657}
1658
1659
1da177e4
LT
1660/**
1661 * skb_shared - is the buffer shared
1662 * @skb: buffer to check
1663 *
1664 * Returns true if more than one person has a reference to this
1665 * buffer.
1666 */
1667static inline int skb_shared(const struct sk_buff *skb)
1668{
63354797 1669 return refcount_read(&skb->users) != 1;
1da177e4
LT
1670}
1671
1672/**
1673 * skb_share_check - check if buffer is shared and if so clone it
1674 * @skb: buffer to check
1675 * @pri: priority for memory allocation
1676 *
1677 * If the buffer is shared the buffer is cloned and the old copy
1678 * drops a reference. A new clone with a single reference is returned.
1679 * If the buffer is not shared the original buffer is returned. When
1680 * being called from interrupt status or with spinlocks held pri must
1681 * be GFP_ATOMIC.
1682 *
1683 * NULL is returned on a memory allocation failure.
1684 */
47061bc4 1685static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1686{
d0164adc 1687 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1688 if (skb_shared(skb)) {
1689 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1690
1691 if (likely(nskb))
1692 consume_skb(skb);
1693 else
1694 kfree_skb(skb);
1da177e4
LT
1695 skb = nskb;
1696 }
1697 return skb;
1698}
1699
1700/*
1701 * Copy shared buffers into a new sk_buff. We effectively do COW on
1702 * packets to handle cases where we have a local reader and forward
1703 * and a couple of other messy ones. The normal one is tcpdumping
1704 * a packet thats being forwarded.
1705 */
1706
1707/**
1708 * skb_unshare - make a copy of a shared buffer
1709 * @skb: buffer to check
1710 * @pri: priority for memory allocation
1711 *
1712 * If the socket buffer is a clone then this function creates a new
1713 * copy of the data, drops a reference count on the old copy and returns
1714 * the new copy with the reference count at 1. If the buffer is not a clone
1715 * the original buffer is returned. When called with a spinlock held or
1716 * from interrupt state @pri must be %GFP_ATOMIC
1717 *
1718 * %NULL is returned on a memory allocation failure.
1719 */
e2bf521d 1720static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1721 gfp_t pri)
1da177e4 1722{
d0164adc 1723 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1724 if (skb_cloned(skb)) {
1725 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1726
1727 /* Free our shared copy */
1728 if (likely(nskb))
1729 consume_skb(skb);
1730 else
1731 kfree_skb(skb);
1da177e4
LT
1732 skb = nskb;
1733 }
1734 return skb;
1735}
1736
1737/**
1a5778aa 1738 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1739 * @list_: list to peek at
1740 *
1741 * Peek an &sk_buff. Unlike most other operations you _MUST_
1742 * be careful with this one. A peek leaves the buffer on the
1743 * list and someone else may run off with it. You must hold
1744 * the appropriate locks or have a private queue to do this.
1745 *
1746 * Returns %NULL for an empty list or a pointer to the head element.
1747 * The reference count is not incremented and the reference is therefore
1748 * volatile. Use with caution.
1749 */
05bdd2f1 1750static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1751{
18d07000
ED
1752 struct sk_buff *skb = list_->next;
1753
1754 if (skb == (struct sk_buff *)list_)
1755 skb = NULL;
1756 return skb;
1da177e4
LT
1757}
1758
8b69bd7d
DM
1759/**
1760 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1761 * @list_: list to peek at
1762 *
1763 * Like skb_peek(), but the caller knows that the list is not empty.
1764 */
1765static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1766{
1767 return list_->next;
1768}
1769
da5ef6e5
PE
1770/**
1771 * skb_peek_next - peek skb following the given one from a queue
1772 * @skb: skb to start from
1773 * @list_: list to peek at
1774 *
1775 * Returns %NULL when the end of the list is met or a pointer to the
1776 * next element. The reference count is not incremented and the
1777 * reference is therefore volatile. Use with caution.
1778 */
1779static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1780 const struct sk_buff_head *list_)
1781{
1782 struct sk_buff *next = skb->next;
18d07000 1783
da5ef6e5
PE
1784 if (next == (struct sk_buff *)list_)
1785 next = NULL;
1786 return next;
1787}
1788
1da177e4 1789/**
1a5778aa 1790 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1791 * @list_: list to peek at
1792 *
1793 * Peek an &sk_buff. Unlike most other operations you _MUST_
1794 * be careful with this one. A peek leaves the buffer on the
1795 * list and someone else may run off with it. You must hold
1796 * the appropriate locks or have a private queue to do this.
1797 *
1798 * Returns %NULL for an empty list or a pointer to the tail element.
1799 * The reference count is not incremented and the reference is therefore
1800 * volatile. Use with caution.
1801 */
05bdd2f1 1802static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1803{
18d07000
ED
1804 struct sk_buff *skb = list_->prev;
1805
1806 if (skb == (struct sk_buff *)list_)
1807 skb = NULL;
1808 return skb;
1809
1da177e4
LT
1810}
1811
1812/**
1813 * skb_queue_len - get queue length
1814 * @list_: list to measure
1815 *
1816 * Return the length of an &sk_buff queue.
1817 */
1818static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1819{
1820 return list_->qlen;
1821}
1822
67fed459
DM
1823/**
1824 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1825 * @list: queue to initialize
1826 *
1827 * This initializes only the list and queue length aspects of
1828 * an sk_buff_head object. This allows to initialize the list
1829 * aspects of an sk_buff_head without reinitializing things like
1830 * the spinlock. It can also be used for on-stack sk_buff_head
1831 * objects where the spinlock is known to not be used.
1832 */
1833static inline void __skb_queue_head_init(struct sk_buff_head *list)
1834{
1835 list->prev = list->next = (struct sk_buff *)list;
1836 list->qlen = 0;
1837}
1838
76f10ad0
AV
1839/*
1840 * This function creates a split out lock class for each invocation;
1841 * this is needed for now since a whole lot of users of the skb-queue
1842 * infrastructure in drivers have different locking usage (in hardirq)
1843 * than the networking core (in softirq only). In the long run either the
1844 * network layer or drivers should need annotation to consolidate the
1845 * main types of usage into 3 classes.
1846 */
1da177e4
LT
1847static inline void skb_queue_head_init(struct sk_buff_head *list)
1848{
1849 spin_lock_init(&list->lock);
67fed459 1850 __skb_queue_head_init(list);
1da177e4
LT
1851}
1852
c2ecba71
PE
1853static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1854 struct lock_class_key *class)
1855{
1856 skb_queue_head_init(list);
1857 lockdep_set_class(&list->lock, class);
1858}
1859
1da177e4 1860/*
bf299275 1861 * Insert an sk_buff on a list.
1da177e4
LT
1862 *
1863 * The "__skb_xxxx()" functions are the non-atomic ones that
1864 * can only be called with interrupts disabled.
1865 */
bf299275
GR
1866static inline void __skb_insert(struct sk_buff *newsk,
1867 struct sk_buff *prev, struct sk_buff *next,
1868 struct sk_buff_head *list)
1869{
a63dd641
ED
1870 /* see skb_queue_empty_lockless() for the opposite READ_ONCE() */
1871 WRITE_ONCE(newsk->next, next);
1872 WRITE_ONCE(newsk->prev, prev);
1873 WRITE_ONCE(next->prev, newsk);
1874 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1875 list->qlen++;
1876}
1da177e4 1877
67fed459
DM
1878static inline void __skb_queue_splice(const struct sk_buff_head *list,
1879 struct sk_buff *prev,
1880 struct sk_buff *next)
1881{
1882 struct sk_buff *first = list->next;
1883 struct sk_buff *last = list->prev;
1884
a63dd641
ED
1885 WRITE_ONCE(first->prev, prev);
1886 WRITE_ONCE(prev->next, first);
67fed459 1887
a63dd641
ED
1888 WRITE_ONCE(last->next, next);
1889 WRITE_ONCE(next->prev, last);
67fed459
DM
1890}
1891
1892/**
1893 * skb_queue_splice - join two skb lists, this is designed for stacks
1894 * @list: the new list to add
1895 * @head: the place to add it in the first list
1896 */
1897static inline void skb_queue_splice(const struct sk_buff_head *list,
1898 struct sk_buff_head *head)
1899{
1900 if (!skb_queue_empty(list)) {
1901 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1902 head->qlen += list->qlen;
67fed459
DM
1903 }
1904}
1905
1906/**
d9619496 1907 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1908 * @list: the new list to add
1909 * @head: the place to add it in the first list
1910 *
1911 * The list at @list is reinitialised
1912 */
1913static inline void skb_queue_splice_init(struct sk_buff_head *list,
1914 struct sk_buff_head *head)
1915{
1916 if (!skb_queue_empty(list)) {
1917 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1918 head->qlen += list->qlen;
67fed459
DM
1919 __skb_queue_head_init(list);
1920 }
1921}
1922
1923/**
1924 * skb_queue_splice_tail - join two skb lists, each list being a queue
1925 * @list: the new list to add
1926 * @head: the place to add it in the first list
1927 */
1928static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1929 struct sk_buff_head *head)
1930{
1931 if (!skb_queue_empty(list)) {
1932 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1933 head->qlen += list->qlen;
67fed459
DM
1934 }
1935}
1936
1937/**
d9619496 1938 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1939 * @list: the new list to add
1940 * @head: the place to add it in the first list
1941 *
1942 * Each of the lists is a queue.
1943 * The list at @list is reinitialised
1944 */
1945static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1946 struct sk_buff_head *head)
1947{
1948 if (!skb_queue_empty(list)) {
1949 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1950 head->qlen += list->qlen;
67fed459
DM
1951 __skb_queue_head_init(list);
1952 }
1953}
1954
1da177e4 1955/**
300ce174 1956 * __skb_queue_after - queue a buffer at the list head
1da177e4 1957 * @list: list to use
300ce174 1958 * @prev: place after this buffer
1da177e4
LT
1959 * @newsk: buffer to queue
1960 *
300ce174 1961 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1962 * and you must therefore hold required locks before calling it.
1963 *
1964 * A buffer cannot be placed on two lists at the same time.
1965 */
300ce174
SH
1966static inline void __skb_queue_after(struct sk_buff_head *list,
1967 struct sk_buff *prev,
1968 struct sk_buff *newsk)
1da177e4 1969{
bf299275 1970 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1971}
1972
7965bd4d
JP
1973void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1974 struct sk_buff_head *list);
7de6c033 1975
f5572855
GR
1976static inline void __skb_queue_before(struct sk_buff_head *list,
1977 struct sk_buff *next,
1978 struct sk_buff *newsk)
1979{
1980 __skb_insert(newsk, next->prev, next, list);
1981}
1982
300ce174
SH
1983/**
1984 * __skb_queue_head - queue a buffer at the list head
1985 * @list: list to use
1986 * @newsk: buffer to queue
1987 *
1988 * Queue a buffer at the start of a list. This function takes no locks
1989 * and you must therefore hold required locks before calling it.
1990 *
1991 * A buffer cannot be placed on two lists at the same time.
1992 */
300ce174
SH
1993static inline void __skb_queue_head(struct sk_buff_head *list,
1994 struct sk_buff *newsk)
1995{
1996 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1997}
4ea7b0cf 1998void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 1999
1da177e4
LT
2000/**
2001 * __skb_queue_tail - queue a buffer at the list tail
2002 * @list: list to use
2003 * @newsk: buffer to queue
2004 *
2005 * Queue a buffer at the end of a list. This function takes no locks
2006 * and you must therefore hold required locks before calling it.
2007 *
2008 * A buffer cannot be placed on two lists at the same time.
2009 */
1da177e4
LT
2010static inline void __skb_queue_tail(struct sk_buff_head *list,
2011 struct sk_buff *newsk)
2012{
f5572855 2013 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2014}
4ea7b0cf 2015void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2016
1da177e4
LT
2017/*
2018 * remove sk_buff from list. _Must_ be called atomically, and with
2019 * the list known..
2020 */
7965bd4d 2021void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2022static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2023{
2024 struct sk_buff *next, *prev;
2025
2026 list->qlen--;
2027 next = skb->next;
2028 prev = skb->prev;
2029 skb->next = skb->prev = NULL;
a63dd641
ED
2030 WRITE_ONCE(next->prev, prev);
2031 WRITE_ONCE(prev->next, next);
1da177e4
LT
2032}
2033
f525c06d
GR
2034/**
2035 * __skb_dequeue - remove from the head of the queue
2036 * @list: list to dequeue from
2037 *
2038 * Remove the head of the list. This function does not take any locks
2039 * so must be used with appropriate locks held only. The head item is
2040 * returned or %NULL if the list is empty.
2041 */
f525c06d
GR
2042static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2043{
2044 struct sk_buff *skb = skb_peek(list);
2045 if (skb)
2046 __skb_unlink(skb, list);
2047 return skb;
2048}
4ea7b0cf 2049struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2050
2051/**
2052 * __skb_dequeue_tail - remove from the tail of the queue
2053 * @list: list to dequeue from
2054 *
2055 * Remove the tail of the list. This function does not take any locks
2056 * so must be used with appropriate locks held only. The tail item is
2057 * returned or %NULL if the list is empty.
2058 */
1da177e4
LT
2059static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2060{
2061 struct sk_buff *skb = skb_peek_tail(list);
2062 if (skb)
2063 __skb_unlink(skb, list);
2064 return skb;
2065}
4ea7b0cf 2066struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2067
2068
bdcc0924 2069static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2070{
2071 return skb->data_len;
2072}
2073
2074static inline unsigned int skb_headlen(const struct sk_buff *skb)
2075{
2076 return skb->len - skb->data_len;
2077}
2078
3ece7826 2079static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2080{
c72d8cda 2081 unsigned int i, len = 0;
1da177e4 2082
c72d8cda 2083 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2084 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2085 return len;
2086}
2087
2088static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2089{
2090 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2091}
2092
131ea667
IC
2093/**
2094 * __skb_fill_page_desc - initialise a paged fragment in an skb
2095 * @skb: buffer containing fragment to be initialised
2096 * @i: paged fragment index to initialise
2097 * @page: the page to use for this fragment
2098 * @off: the offset to the data with @page
2099 * @size: the length of the data
2100 *
2101 * Initialises the @i'th fragment of @skb to point to &size bytes at
2102 * offset @off within @page.
2103 *
2104 * Does not take any additional reference on the fragment.
2105 */
2106static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2107 struct page *page, int off, int size)
1da177e4
LT
2108{
2109 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2110
c48a11c7 2111 /*
2f064f34
MH
2112 * Propagate page pfmemalloc to the skb if we can. The problem is
2113 * that not all callers have unique ownership of the page but rely
2114 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2115 */
a8605c60 2116 frag->page.p = page;
1da177e4 2117 frag->page_offset = off;
9e903e08 2118 skb_frag_size_set(frag, size);
cca7af38
PE
2119
2120 page = compound_head(page);
2f064f34 2121 if (page_is_pfmemalloc(page))
cca7af38 2122 skb->pfmemalloc = true;
131ea667
IC
2123}
2124
2125/**
2126 * skb_fill_page_desc - initialise a paged fragment in an skb
2127 * @skb: buffer containing fragment to be initialised
2128 * @i: paged fragment index to initialise
2129 * @page: the page to use for this fragment
2130 * @off: the offset to the data with @page
2131 * @size: the length of the data
2132 *
2133 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2134 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2135 * addition updates @skb such that @i is the last fragment.
2136 *
2137 * Does not take any additional reference on the fragment.
2138 */
2139static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2140 struct page *page, int off, int size)
2141{
2142 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2143 skb_shinfo(skb)->nr_frags = i + 1;
2144}
2145
7965bd4d
JP
2146void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2147 int size, unsigned int truesize);
654bed16 2148
f8e617e1
JW
2149void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2150 unsigned int truesize);
2151
1da177e4
LT
2152#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2153
27a884dc
ACM
2154#ifdef NET_SKBUFF_DATA_USES_OFFSET
2155static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2156{
2157 return skb->head + skb->tail;
2158}
2159
2160static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2161{
2162 skb->tail = skb->data - skb->head;
2163}
2164
2165static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2166{
2167 skb_reset_tail_pointer(skb);
2168 skb->tail += offset;
2169}
7cc46190 2170
27a884dc
ACM
2171#else /* NET_SKBUFF_DATA_USES_OFFSET */
2172static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2173{
2174 return skb->tail;
2175}
2176
2177static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2178{
2179 skb->tail = skb->data;
2180}
2181
2182static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2183{
2184 skb->tail = skb->data + offset;
2185}
4305b541 2186
27a884dc
ACM
2187#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2188
1da177e4
LT
2189/*
2190 * Add data to an sk_buff
2191 */
4df864c1
JB
2192void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2193void *skb_put(struct sk_buff *skb, unsigned int len);
2194static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2195{
4df864c1 2196 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2197 SKB_LINEAR_ASSERT(skb);
2198 skb->tail += len;
2199 skb->len += len;
2200 return tmp;
2201}
2202
de77b966 2203static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2204{
2205 void *tmp = __skb_put(skb, len);
2206
2207 memset(tmp, 0, len);
2208 return tmp;
2209}
2210
2211static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2212 unsigned int len)
2213{
2214 void *tmp = __skb_put(skb, len);
2215
2216 memcpy(tmp, data, len);
2217 return tmp;
2218}
2219
2220static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2221{
2222 *(u8 *)__skb_put(skb, 1) = val;
2223}
2224
83ad357d 2225static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2226{
83ad357d 2227 void *tmp = skb_put(skb, len);
e45a79da
JB
2228
2229 memset(tmp, 0, len);
2230
2231 return tmp;
2232}
2233
59ae1d12
JB
2234static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2235 unsigned int len)
2236{
2237 void *tmp = skb_put(skb, len);
2238
2239 memcpy(tmp, data, len);
2240
2241 return tmp;
2242}
2243
634fef61
JB
2244static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2245{
2246 *(u8 *)skb_put(skb, 1) = val;
2247}
2248
d58ff351
JB
2249void *skb_push(struct sk_buff *skb, unsigned int len);
2250static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2251{
2252 skb->data -= len;
2253 skb->len += len;
2254 return skb->data;
2255}
2256
af72868b
JB
2257void *skb_pull(struct sk_buff *skb, unsigned int len);
2258static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2259{
2260 skb->len -= len;
2261 BUG_ON(skb->len < skb->data_len);
2262 return skb->data += len;
2263}
2264
af72868b 2265static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2266{
2267 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2268}
2269
af72868b 2270void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2271
af72868b 2272static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2273{
2274 if (len > skb_headlen(skb) &&
987c402a 2275 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2276 return NULL;
2277 skb->len -= len;
2278 return skb->data += len;
2279}
2280
af72868b 2281static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2282{
2283 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2284}
2285
2286static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2287{
2288 if (likely(len <= skb_headlen(skb)))
2289 return 1;
2290 if (unlikely(len > skb->len))
2291 return 0;
987c402a 2292 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2293}
2294
c8c8b127
ED
2295void skb_condense(struct sk_buff *skb);
2296
1da177e4
LT
2297/**
2298 * skb_headroom - bytes at buffer head
2299 * @skb: buffer to check
2300 *
2301 * Return the number of bytes of free space at the head of an &sk_buff.
2302 */
c2636b4d 2303static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2304{
2305 return skb->data - skb->head;
2306}
2307
2308/**
2309 * skb_tailroom - bytes at buffer end
2310 * @skb: buffer to check
2311 *
2312 * Return the number of bytes of free space at the tail of an sk_buff
2313 */
2314static inline int skb_tailroom(const struct sk_buff *skb)
2315{
4305b541 2316 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2317}
2318
a21d4572
ED
2319/**
2320 * skb_availroom - bytes at buffer end
2321 * @skb: buffer to check
2322 *
2323 * Return the number of bytes of free space at the tail of an sk_buff
2324 * allocated by sk_stream_alloc()
2325 */
2326static inline int skb_availroom(const struct sk_buff *skb)
2327{
16fad69c
ED
2328 if (skb_is_nonlinear(skb))
2329 return 0;
2330
2331 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2332}
2333
1da177e4
LT
2334/**
2335 * skb_reserve - adjust headroom
2336 * @skb: buffer to alter
2337 * @len: bytes to move
2338 *
2339 * Increase the headroom of an empty &sk_buff by reducing the tail
2340 * room. This is only allowed for an empty buffer.
2341 */
8243126c 2342static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2343{
2344 skb->data += len;
2345 skb->tail += len;
2346}
2347
1837b2e2
BP
2348/**
2349 * skb_tailroom_reserve - adjust reserved_tailroom
2350 * @skb: buffer to alter
2351 * @mtu: maximum amount of headlen permitted
2352 * @needed_tailroom: minimum amount of reserved_tailroom
2353 *
2354 * Set reserved_tailroom so that headlen can be as large as possible but
2355 * not larger than mtu and tailroom cannot be smaller than
2356 * needed_tailroom.
2357 * The required headroom should already have been reserved before using
2358 * this function.
2359 */
2360static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2361 unsigned int needed_tailroom)
2362{
2363 SKB_LINEAR_ASSERT(skb);
2364 if (mtu < skb_tailroom(skb) - needed_tailroom)
2365 /* use at most mtu */
2366 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2367 else
2368 /* use up to all available space */
2369 skb->reserved_tailroom = needed_tailroom;
2370}
2371
8bce6d7d
TH
2372#define ENCAP_TYPE_ETHER 0
2373#define ENCAP_TYPE_IPPROTO 1
2374
2375static inline void skb_set_inner_protocol(struct sk_buff *skb,
2376 __be16 protocol)
2377{
2378 skb->inner_protocol = protocol;
2379 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2380}
2381
2382static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2383 __u8 ipproto)
2384{
2385 skb->inner_ipproto = ipproto;
2386 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2387}
2388
6a674e9c
JG
2389static inline void skb_reset_inner_headers(struct sk_buff *skb)
2390{
aefbd2b3 2391 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2392 skb->inner_network_header = skb->network_header;
2393 skb->inner_transport_header = skb->transport_header;
2394}
2395
0b5c9db1
JP
2396static inline void skb_reset_mac_len(struct sk_buff *skb)
2397{
2398 skb->mac_len = skb->network_header - skb->mac_header;
2399}
2400
6a674e9c
JG
2401static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2402 *skb)
2403{
2404 return skb->head + skb->inner_transport_header;
2405}
2406
55dc5a9f
TH
2407static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2408{
2409 return skb_inner_transport_header(skb) - skb->data;
2410}
2411
6a674e9c
JG
2412static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2413{
2414 skb->inner_transport_header = skb->data - skb->head;
2415}
2416
2417static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2418 const int offset)
2419{
2420 skb_reset_inner_transport_header(skb);
2421 skb->inner_transport_header += offset;
2422}
2423
2424static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2425{
2426 return skb->head + skb->inner_network_header;
2427}
2428
2429static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2430{
2431 skb->inner_network_header = skb->data - skb->head;
2432}
2433
2434static inline void skb_set_inner_network_header(struct sk_buff *skb,
2435 const int offset)
2436{
2437 skb_reset_inner_network_header(skb);
2438 skb->inner_network_header += offset;
2439}
2440
aefbd2b3
PS
2441static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2442{
2443 return skb->head + skb->inner_mac_header;
2444}
2445
2446static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2447{
2448 skb->inner_mac_header = skb->data - skb->head;
2449}
2450
2451static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2452 const int offset)
2453{
2454 skb_reset_inner_mac_header(skb);
2455 skb->inner_mac_header += offset;
2456}
fda55eca
ED
2457static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2458{
35d04610 2459 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2460}
2461
9c70220b
ACM
2462static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2463{
2e07fa9c 2464 return skb->head + skb->transport_header;
9c70220b
ACM
2465}
2466
badff6d0
ACM
2467static inline void skb_reset_transport_header(struct sk_buff *skb)
2468{
2e07fa9c 2469 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2470}
2471
967b05f6
ACM
2472static inline void skb_set_transport_header(struct sk_buff *skb,
2473 const int offset)
2474{
2e07fa9c
ACM
2475 skb_reset_transport_header(skb);
2476 skb->transport_header += offset;
ea2ae17d
ACM
2477}
2478
d56f90a7
ACM
2479static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2480{
2e07fa9c 2481 return skb->head + skb->network_header;
d56f90a7
ACM
2482}
2483
c1d2bbe1
ACM
2484static inline void skb_reset_network_header(struct sk_buff *skb)
2485{
2e07fa9c 2486 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2487}
2488
c14d2450
ACM
2489static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2490{
2e07fa9c
ACM
2491 skb_reset_network_header(skb);
2492 skb->network_header += offset;
c14d2450
ACM
2493}
2494
2e07fa9c 2495static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2496{
2e07fa9c 2497 return skb->head + skb->mac_header;
bbe735e4
ACM
2498}
2499
ea6da4fd
AV
2500static inline int skb_mac_offset(const struct sk_buff *skb)
2501{
2502 return skb_mac_header(skb) - skb->data;
2503}
2504
0daf4349
DB
2505static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2506{
2507 return skb->network_header - skb->mac_header;
2508}
2509
2e07fa9c 2510static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2511{
35d04610 2512 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2513}
2514
2515static inline void skb_reset_mac_header(struct sk_buff *skb)
2516{
2517 skb->mac_header = skb->data - skb->head;
2518}
2519
2520static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2521{
2522 skb_reset_mac_header(skb);
2523 skb->mac_header += offset;
2524}
2525
0e3da5bb
TT
2526static inline void skb_pop_mac_header(struct sk_buff *skb)
2527{
2528 skb->mac_header = skb->network_header;
2529}
2530
d2aa125d 2531static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2532{
72a338bc 2533 struct flow_keys_basic keys;
fbbdb8f0
YX
2534
2535 if (skb_transport_header_was_set(skb))
2536 return;
72a338bc 2537
3cbf4ffb
SF
2538 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2539 NULL, 0, 0, 0, 0))
42aecaa9 2540 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2541}
2542
03606895
ED
2543static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2544{
2545 if (skb_mac_header_was_set(skb)) {
2546 const unsigned char *old_mac = skb_mac_header(skb);
2547
2548 skb_set_mac_header(skb, -skb->mac_len);
2549 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2550 }
2551}
2552
04fb451e
MM
2553static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2554{
2555 return skb->csum_start - skb_headroom(skb);
2556}
2557
08b64fcc
AD
2558static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2559{
2560 return skb->head + skb->csum_start;
2561}
2562
2e07fa9c
ACM
2563static inline int skb_transport_offset(const struct sk_buff *skb)
2564{
2565 return skb_transport_header(skb) - skb->data;
2566}
2567
2568static inline u32 skb_network_header_len(const struct sk_buff *skb)
2569{
2570 return skb->transport_header - skb->network_header;
2571}
2572
6a674e9c
JG
2573static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2574{
2575 return skb->inner_transport_header - skb->inner_network_header;
2576}
2577
2e07fa9c
ACM
2578static inline int skb_network_offset(const struct sk_buff *skb)
2579{
2580 return skb_network_header(skb) - skb->data;
2581}
48d49d0c 2582
6a674e9c
JG
2583static inline int skb_inner_network_offset(const struct sk_buff *skb)
2584{
2585 return skb_inner_network_header(skb) - skb->data;
2586}
2587
f9599ce1
CG
2588static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2589{
2590 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2591}
2592
1da177e4
LT
2593/*
2594 * CPUs often take a performance hit when accessing unaligned memory
2595 * locations. The actual performance hit varies, it can be small if the
2596 * hardware handles it or large if we have to take an exception and fix it
2597 * in software.
2598 *
2599 * Since an ethernet header is 14 bytes network drivers often end up with
2600 * the IP header at an unaligned offset. The IP header can be aligned by
2601 * shifting the start of the packet by 2 bytes. Drivers should do this
2602 * with:
2603 *
8660c124 2604 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2605 *
2606 * The downside to this alignment of the IP header is that the DMA is now
2607 * unaligned. On some architectures the cost of an unaligned DMA is high
2608 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2609 *
1da177e4
LT
2610 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2611 * to be overridden.
2612 */
2613#ifndef NET_IP_ALIGN
2614#define NET_IP_ALIGN 2
2615#endif
2616
025be81e
AB
2617/*
2618 * The networking layer reserves some headroom in skb data (via
2619 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2620 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2621 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2622 *
2623 * Unfortunately this headroom changes the DMA alignment of the resulting
2624 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2625 * on some architectures. An architecture can override this value,
2626 * perhaps setting it to a cacheline in size (since that will maintain
2627 * cacheline alignment of the DMA). It must be a power of 2.
2628 *
d6301d3d 2629 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2630 * headroom, you should not reduce this.
5933dd2f
ED
2631 *
2632 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2633 * to reduce average number of cache lines per packet.
2634 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2635 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2636 */
2637#ifndef NET_SKB_PAD
5933dd2f 2638#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2639#endif
2640
7965bd4d 2641int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2642
5293efe6 2643static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2644{
5e1abdc3 2645 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2646 return;
27a884dc
ACM
2647 skb->len = len;
2648 skb_set_tail_pointer(skb, len);
1da177e4
LT
2649}
2650
5293efe6
DB
2651static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2652{
2653 __skb_set_length(skb, len);
2654}
2655
7965bd4d 2656void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2657
2658static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2659{
3cc0e873
HX
2660 if (skb->data_len)
2661 return ___pskb_trim(skb, len);
2662 __skb_trim(skb, len);
2663 return 0;
1da177e4
LT
2664}
2665
2666static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2667{
2668 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2669}
2670
e9fa4f7b
HX
2671/**
2672 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2673 * @skb: buffer to alter
2674 * @len: new length
2675 *
2676 * This is identical to pskb_trim except that the caller knows that
2677 * the skb is not cloned so we should never get an error due to out-
2678 * of-memory.
2679 */
2680static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2681{
2682 int err = pskb_trim(skb, len);
2683 BUG_ON(err);
2684}
2685
5293efe6
DB
2686static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2687{
2688 unsigned int diff = len - skb->len;
2689
2690 if (skb_tailroom(skb) < diff) {
2691 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2692 GFP_ATOMIC);
2693 if (ret)
2694 return ret;
2695 }
2696 __skb_set_length(skb, len);
2697 return 0;
2698}
2699
1da177e4
LT
2700/**
2701 * skb_orphan - orphan a buffer
2702 * @skb: buffer to orphan
2703 *
2704 * If a buffer currently has an owner then we call the owner's
2705 * destructor function and make the @skb unowned. The buffer continues
2706 * to exist but is no longer charged to its former owner.
2707 */
2708static inline void skb_orphan(struct sk_buff *skb)
2709{
c34a7612 2710 if (skb->destructor) {
1da177e4 2711 skb->destructor(skb);
c34a7612
ED
2712 skb->destructor = NULL;
2713 skb->sk = NULL;
376c7311
ED
2714 } else {
2715 BUG_ON(skb->sk);
c34a7612 2716 }
1da177e4
LT
2717}
2718
a353e0ce
MT
2719/**
2720 * skb_orphan_frags - orphan the frags contained in a buffer
2721 * @skb: buffer to orphan frags from
2722 * @gfp_mask: allocation mask for replacement pages
2723 *
2724 * For each frag in the SKB which needs a destructor (i.e. has an
2725 * owner) create a copy of that frag and release the original
2726 * page by calling the destructor.
2727 */
2728static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2729{
1f8b977a
WB
2730 if (likely(!skb_zcopy(skb)))
2731 return 0;
185ce5c3
WB
2732 if (!skb_zcopy_is_nouarg(skb) &&
2733 skb_uarg(skb)->callback == sock_zerocopy_callback)
1f8b977a
WB
2734 return 0;
2735 return skb_copy_ubufs(skb, gfp_mask);
2736}
2737
2738/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2739static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2740{
2741 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2742 return 0;
2743 return skb_copy_ubufs(skb, gfp_mask);
2744}
2745
1da177e4
LT
2746/**
2747 * __skb_queue_purge - empty a list
2748 * @list: list to empty
2749 *
2750 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2751 * the list and one reference dropped. This function does not take the
2752 * list lock and the caller must hold the relevant locks to use it.
2753 */
1da177e4
LT
2754static inline void __skb_queue_purge(struct sk_buff_head *list)
2755{
2756 struct sk_buff *skb;
2757 while ((skb = __skb_dequeue(list)) != NULL)
2758 kfree_skb(skb);
2759}
4ea7b0cf 2760void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2761
385114de 2762unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2763
7965bd4d 2764void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2765
7965bd4d
JP
2766struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2767 gfp_t gfp_mask);
8af27456
CH
2768
2769/**
2770 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2771 * @dev: network device to receive on
2772 * @length: length to allocate
2773 *
2774 * Allocate a new &sk_buff and assign it a usage count of one. The
2775 * buffer has unspecified headroom built in. Users should allocate
2776 * the headroom they think they need without accounting for the
2777 * built in space. The built in space is used for optimisations.
2778 *
2779 * %NULL is returned if there is no free memory. Although this function
2780 * allocates memory it can be called from an interrupt.
2781 */
2782static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2783 unsigned int length)
8af27456
CH
2784{
2785 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2786}
2787
6f532612
ED
2788/* legacy helper around __netdev_alloc_skb() */
2789static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2790 gfp_t gfp_mask)
2791{
2792 return __netdev_alloc_skb(NULL, length, gfp_mask);
2793}
2794
2795/* legacy helper around netdev_alloc_skb() */
2796static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2797{
2798 return netdev_alloc_skb(NULL, length);
2799}
2800
2801
4915a0de
ED
2802static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2803 unsigned int length, gfp_t gfp)
61321bbd 2804{
4915a0de 2805 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2806
2807 if (NET_IP_ALIGN && skb)
2808 skb_reserve(skb, NET_IP_ALIGN);
2809 return skb;
2810}
2811
4915a0de
ED
2812static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2813 unsigned int length)
2814{
2815 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2816}
2817
181edb2b
AD
2818static inline void skb_free_frag(void *addr)
2819{
8c2dd3e4 2820 page_frag_free(addr);
181edb2b
AD
2821}
2822
ffde7328 2823void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2824struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2825 unsigned int length, gfp_t gfp_mask);
2826static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2827 unsigned int length)
2828{
2829 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2830}
795bb1c0
JDB
2831void napi_consume_skb(struct sk_buff *skb, int budget);
2832
2833void __kfree_skb_flush(void);
15fad714 2834void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2835
71dfda58
AD
2836/**
2837 * __dev_alloc_pages - allocate page for network Rx
2838 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2839 * @order: size of the allocation
2840 *
2841 * Allocate a new page.
2842 *
2843 * %NULL is returned if there is no free memory.
2844*/
2845static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2846 unsigned int order)
2847{
2848 /* This piece of code contains several assumptions.
2849 * 1. This is for device Rx, therefor a cold page is preferred.
2850 * 2. The expectation is the user wants a compound page.
2851 * 3. If requesting a order 0 page it will not be compound
2852 * due to the check to see if order has a value in prep_new_page
2853 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2854 * code in gfp_to_alloc_flags that should be enforcing this.
2855 */
453f85d4 2856 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2857
2858 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2859}
2860
2861static inline struct page *dev_alloc_pages(unsigned int order)
2862{
95829b3a 2863 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2864}
2865
2866/**
2867 * __dev_alloc_page - allocate a page for network Rx
2868 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2869 *
2870 * Allocate a new page.
2871 *
2872 * %NULL is returned if there is no free memory.
2873 */
2874static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2875{
2876 return __dev_alloc_pages(gfp_mask, 0);
2877}
2878
2879static inline struct page *dev_alloc_page(void)
2880{
95829b3a 2881 return dev_alloc_pages(0);
71dfda58
AD
2882}
2883
0614002b
MG
2884/**
2885 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2886 * @page: The page that was allocated from skb_alloc_page
2887 * @skb: The skb that may need pfmemalloc set
2888 */
2889static inline void skb_propagate_pfmemalloc(struct page *page,
2890 struct sk_buff *skb)
2891{
2f064f34 2892 if (page_is_pfmemalloc(page))
0614002b
MG
2893 skb->pfmemalloc = true;
2894}
2895
131ea667 2896/**
e227867f 2897 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2898 * @frag: the paged fragment
2899 *
2900 * Returns the &struct page associated with @frag.
2901 */
2902static inline struct page *skb_frag_page(const skb_frag_t *frag)
2903{
a8605c60 2904 return frag->page.p;
131ea667
IC
2905}
2906
2907/**
2908 * __skb_frag_ref - take an addition reference on a paged fragment.
2909 * @frag: the paged fragment
2910 *
2911 * Takes an additional reference on the paged fragment @frag.
2912 */
2913static inline void __skb_frag_ref(skb_frag_t *frag)
2914{
2915 get_page(skb_frag_page(frag));
2916}
2917
2918/**
2919 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2920 * @skb: the buffer
2921 * @f: the fragment offset.
2922 *
2923 * Takes an additional reference on the @f'th paged fragment of @skb.
2924 */
2925static inline void skb_frag_ref(struct sk_buff *skb, int f)
2926{
2927 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2928}
2929
2930/**
2931 * __skb_frag_unref - release a reference on a paged fragment.
2932 * @frag: the paged fragment
2933 *
2934 * Releases a reference on the paged fragment @frag.
2935 */
2936static inline void __skb_frag_unref(skb_frag_t *frag)
2937{
2938 put_page(skb_frag_page(frag));
2939}
2940
2941/**
2942 * skb_frag_unref - release a reference on a paged fragment of an skb.
2943 * @skb: the buffer
2944 * @f: the fragment offset
2945 *
2946 * Releases a reference on the @f'th paged fragment of @skb.
2947 */
2948static inline void skb_frag_unref(struct sk_buff *skb, int f)
2949{
2950 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2951}
2952
2953/**
2954 * skb_frag_address - gets the address of the data contained in a paged fragment
2955 * @frag: the paged fragment buffer
2956 *
2957 * Returns the address of the data within @frag. The page must already
2958 * be mapped.
2959 */
2960static inline void *skb_frag_address(const skb_frag_t *frag)
2961{
2962 return page_address(skb_frag_page(frag)) + frag->page_offset;
2963}
2964
2965/**
2966 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2967 * @frag: the paged fragment buffer
2968 *
2969 * Returns the address of the data within @frag. Checks that the page
2970 * is mapped and returns %NULL otherwise.
2971 */
2972static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2973{
2974 void *ptr = page_address(skb_frag_page(frag));
2975 if (unlikely(!ptr))
2976 return NULL;
2977
2978 return ptr + frag->page_offset;
2979}
2980
2981/**
2982 * __skb_frag_set_page - sets the page contained in a paged fragment
2983 * @frag: the paged fragment
2984 * @page: the page to set
2985 *
2986 * Sets the fragment @frag to contain @page.
2987 */
2988static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2989{
a8605c60 2990 frag->page.p = page;
131ea667
IC
2991}
2992
2993/**
2994 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2995 * @skb: the buffer
2996 * @f: the fragment offset
2997 * @page: the page to set
2998 *
2999 * Sets the @f'th fragment of @skb to contain @page.
3000 */
3001static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3002 struct page *page)
3003{
3004 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3005}
3006
400dfd3a
ED
3007bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3008
131ea667
IC
3009/**
3010 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3011 * @dev: the device to map the fragment to
131ea667
IC
3012 * @frag: the paged fragment to map
3013 * @offset: the offset within the fragment (starting at the
3014 * fragment's own offset)
3015 * @size: the number of bytes to map
771b00a8 3016 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3017 *
3018 * Maps the page associated with @frag to @device.
3019 */
3020static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3021 const skb_frag_t *frag,
3022 size_t offset, size_t size,
3023 enum dma_data_direction dir)
3024{
3025 return dma_map_page(dev, skb_frag_page(frag),
3026 frag->page_offset + offset, size, dir);
3027}
3028
117632e6
ED
3029static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3030 gfp_t gfp_mask)
3031{
3032 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3033}
3034
bad93e9d
OP
3035
3036static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3037 gfp_t gfp_mask)
3038{
3039 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3040}
3041
3042
334a8132
PM
3043/**
3044 * skb_clone_writable - is the header of a clone writable
3045 * @skb: buffer to check
3046 * @len: length up to which to write
3047 *
3048 * Returns true if modifying the header part of the cloned buffer
3049 * does not requires the data to be copied.
3050 */
05bdd2f1 3051static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3052{
3053 return !skb_header_cloned(skb) &&
3054 skb_headroom(skb) + len <= skb->hdr_len;
3055}
3056
3697649f
DB
3057static inline int skb_try_make_writable(struct sk_buff *skb,
3058 unsigned int write_len)
3059{
3060 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3061 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3062}
3063
d9cc2048
HX
3064static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3065 int cloned)
3066{
3067 int delta = 0;
3068
d9cc2048
HX
3069 if (headroom > skb_headroom(skb))
3070 delta = headroom - skb_headroom(skb);
3071
3072 if (delta || cloned)
3073 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3074 GFP_ATOMIC);
3075 return 0;
3076}
3077
1da177e4
LT
3078/**
3079 * skb_cow - copy header of skb when it is required
3080 * @skb: buffer to cow
3081 * @headroom: needed headroom
3082 *
3083 * If the skb passed lacks sufficient headroom or its data part
3084 * is shared, data is reallocated. If reallocation fails, an error
3085 * is returned and original skb is not changed.
3086 *
3087 * The result is skb with writable area skb->head...skb->tail
3088 * and at least @headroom of space at head.
3089 */
3090static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3091{
d9cc2048
HX
3092 return __skb_cow(skb, headroom, skb_cloned(skb));
3093}
1da177e4 3094
d9cc2048
HX
3095/**
3096 * skb_cow_head - skb_cow but only making the head writable
3097 * @skb: buffer to cow
3098 * @headroom: needed headroom
3099 *
3100 * This function is identical to skb_cow except that we replace the
3101 * skb_cloned check by skb_header_cloned. It should be used when
3102 * you only need to push on some header and do not need to modify
3103 * the data.
3104 */
3105static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3106{
3107 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3108}
3109
3110/**
3111 * skb_padto - pad an skbuff up to a minimal size
3112 * @skb: buffer to pad
3113 * @len: minimal length
3114 *
3115 * Pads up a buffer to ensure the trailing bytes exist and are
3116 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3117 * is untouched. Otherwise it is extended. Returns zero on
3118 * success. The skb is freed on error.
1da177e4 3119 */
5b057c6b 3120static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3121{
3122 unsigned int size = skb->len;
3123 if (likely(size >= len))
5b057c6b 3124 return 0;
987c402a 3125 return skb_pad(skb, len - size);
1da177e4
LT
3126}
3127
9c0c1124 3128/**
4ea7b0cf 3129 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3130 * @skb: buffer to pad
3131 * @len: minimal length
cd0a137a 3132 * @free_on_error: free buffer on error
9c0c1124
AD
3133 *
3134 * Pads up a buffer to ensure the trailing bytes exist and are
3135 * blanked. If the buffer already contains sufficient data it
3136 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3137 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3138 */
cd0a137a
FF
3139static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3140 bool free_on_error)
9c0c1124
AD
3141{
3142 unsigned int size = skb->len;
3143
3144 if (unlikely(size < len)) {
3145 len -= size;
cd0a137a 3146 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3147 return -ENOMEM;
3148 __skb_put(skb, len);
3149 }
3150 return 0;
3151}
3152
cd0a137a
FF
3153/**
3154 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3155 * @skb: buffer to pad
3156 * @len: minimal length
3157 *
3158 * Pads up a buffer to ensure the trailing bytes exist and are
3159 * blanked. If the buffer already contains sufficient data it
3160 * is untouched. Otherwise it is extended. Returns zero on
3161 * success. The skb is freed on error.
3162 */
3163static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3164{
3165 return __skb_put_padto(skb, len, true);
3166}
3167
1da177e4 3168static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3169 struct iov_iter *from, int copy)
1da177e4
LT
3170{
3171 const int off = skb->len;
3172
3173 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3174 __wsum csum = 0;
15e6cb46
AV
3175 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3176 &csum, from)) {
1da177e4
LT
3177 skb->csum = csum_block_add(skb->csum, csum, off);
3178 return 0;
3179 }
15e6cb46 3180 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3181 return 0;
3182
3183 __skb_trim(skb, off);
3184 return -EFAULT;
3185}
3186
38ba0a65
ED
3187static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3188 const struct page *page, int off)
1da177e4 3189{
1f8b977a
WB
3190 if (skb_zcopy(skb))
3191 return false;
1da177e4 3192 if (i) {
9e903e08 3193 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3194
ea2ab693 3195 return page == skb_frag_page(frag) &&
9e903e08 3196 off == frag->page_offset + skb_frag_size(frag);
1da177e4 3197 }
38ba0a65 3198 return false;
1da177e4
LT
3199}
3200
364c6bad
HX
3201static inline int __skb_linearize(struct sk_buff *skb)
3202{
3203 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3204}
3205
1da177e4
LT
3206/**
3207 * skb_linearize - convert paged skb to linear one
3208 * @skb: buffer to linarize
1da177e4
LT
3209 *
3210 * If there is no free memory -ENOMEM is returned, otherwise zero
3211 * is returned and the old skb data released.
3212 */
364c6bad
HX
3213static inline int skb_linearize(struct sk_buff *skb)
3214{
3215 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3216}
3217
cef401de
ED
3218/**
3219 * skb_has_shared_frag - can any frag be overwritten
3220 * @skb: buffer to test
3221 *
3222 * Return true if the skb has at least one frag that might be modified
3223 * by an external entity (as in vmsplice()/sendfile())
3224 */
3225static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3226{
c9af6db4
PS
3227 return skb_is_nonlinear(skb) &&
3228 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
3229}
3230
364c6bad
HX
3231/**
3232 * skb_linearize_cow - make sure skb is linear and writable
3233 * @skb: buffer to process
3234 *
3235 * If there is no free memory -ENOMEM is returned, otherwise zero
3236 * is returned and the old skb data released.
3237 */
3238static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3239{
364c6bad
HX
3240 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3241 __skb_linearize(skb) : 0;
1da177e4
LT
3242}
3243
479ffccc
DB
3244static __always_inline void
3245__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3246 unsigned int off)
3247{
3248 if (skb->ip_summed == CHECKSUM_COMPLETE)
3249 skb->csum = csum_block_sub(skb->csum,
3250 csum_partial(start, len, 0), off);
3251 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3252 skb_checksum_start_offset(skb) < 0)
3253 skb->ip_summed = CHECKSUM_NONE;
3254}
3255
1da177e4
LT
3256/**
3257 * skb_postpull_rcsum - update checksum for received skb after pull
3258 * @skb: buffer to update
3259 * @start: start of data before pull
3260 * @len: length of data pulled
3261 *
3262 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3263 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3264 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3265 */
1da177e4 3266static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3267 const void *start, unsigned int len)
1da177e4 3268{
479ffccc 3269 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3270}
3271
479ffccc
DB
3272static __always_inline void
3273__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3274 unsigned int off)
3275{
3276 if (skb->ip_summed == CHECKSUM_COMPLETE)
3277 skb->csum = csum_block_add(skb->csum,
3278 csum_partial(start, len, 0), off);
3279}
cbb042f9 3280
479ffccc
DB
3281/**
3282 * skb_postpush_rcsum - update checksum for received skb after push
3283 * @skb: buffer to update
3284 * @start: start of data after push
3285 * @len: length of data pushed
3286 *
3287 * After doing a push on a received packet, you need to call this to
3288 * update the CHECKSUM_COMPLETE checksum.
3289 */
f8ffad69
DB
3290static inline void skb_postpush_rcsum(struct sk_buff *skb,
3291 const void *start, unsigned int len)
3292{
479ffccc 3293 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3294}
3295
af72868b 3296void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3297
82a31b92
WC
3298/**
3299 * skb_push_rcsum - push skb and update receive checksum
3300 * @skb: buffer to update
3301 * @len: length of data pulled
3302 *
3303 * This function performs an skb_push on the packet and updates
3304 * the CHECKSUM_COMPLETE checksum. It should be used on
3305 * receive path processing instead of skb_push unless you know
3306 * that the checksum difference is zero (e.g., a valid IP header)
3307 * or you are setting ip_summed to CHECKSUM_NONE.
3308 */
d58ff351 3309static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3310{
3311 skb_push(skb, len);
3312 skb_postpush_rcsum(skb, skb->data, len);
3313 return skb->data;
3314}
3315
88078d98 3316int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3317/**
3318 * pskb_trim_rcsum - trim received skb and update checksum
3319 * @skb: buffer to trim
3320 * @len: new length
3321 *
3322 * This is exactly the same as pskb_trim except that it ensures the
3323 * checksum of received packets are still valid after the operation.
6c57f045 3324 * It can change skb pointers.
7ce5a27f
DM
3325 */
3326
3327static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3328{
3329 if (likely(len >= skb->len))
3330 return 0;
88078d98 3331 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3332}
3333
5293efe6
DB
3334static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3335{
3336 if (skb->ip_summed == CHECKSUM_COMPLETE)
3337 skb->ip_summed = CHECKSUM_NONE;
3338 __skb_trim(skb, len);
3339 return 0;
3340}
3341
3342static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3343{
3344 if (skb->ip_summed == CHECKSUM_COMPLETE)
3345 skb->ip_summed = CHECKSUM_NONE;
3346 return __skb_grow(skb, len);
3347}
3348
18a4c0ea
ED
3349#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3350#define skb_rb_first(root) rb_to_skb(rb_first(root))
3351#define skb_rb_last(root) rb_to_skb(rb_last(root))
3352#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3353#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3354
1da177e4
LT
3355#define skb_queue_walk(queue, skb) \
3356 for (skb = (queue)->next; \
a1e4891f 3357 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3358 skb = skb->next)
3359
46f8914e
JC
3360#define skb_queue_walk_safe(queue, skb, tmp) \
3361 for (skb = (queue)->next, tmp = skb->next; \
3362 skb != (struct sk_buff *)(queue); \
3363 skb = tmp, tmp = skb->next)
3364
1164f52a 3365#define skb_queue_walk_from(queue, skb) \
a1e4891f 3366 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3367 skb = skb->next)
3368
18a4c0ea
ED
3369#define skb_rbtree_walk(skb, root) \
3370 for (skb = skb_rb_first(root); skb != NULL; \
3371 skb = skb_rb_next(skb))
3372
3373#define skb_rbtree_walk_from(skb) \
3374 for (; skb != NULL; \
3375 skb = skb_rb_next(skb))
3376
3377#define skb_rbtree_walk_from_safe(skb, tmp) \
3378 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3379 skb = tmp)
3380
1164f52a
DM
3381#define skb_queue_walk_from_safe(queue, skb, tmp) \
3382 for (tmp = skb->next; \
3383 skb != (struct sk_buff *)(queue); \
3384 skb = tmp, tmp = skb->next)
3385
300ce174
SH
3386#define skb_queue_reverse_walk(queue, skb) \
3387 for (skb = (queue)->prev; \
a1e4891f 3388 skb != (struct sk_buff *)(queue); \
300ce174
SH
3389 skb = skb->prev)
3390
686a2955
DM
3391#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3392 for (skb = (queue)->prev, tmp = skb->prev; \
3393 skb != (struct sk_buff *)(queue); \
3394 skb = tmp, tmp = skb->prev)
3395
3396#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3397 for (tmp = skb->prev; \
3398 skb != (struct sk_buff *)(queue); \
3399 skb = tmp, tmp = skb->prev)
1da177e4 3400
21dc3301 3401static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3402{
3403 return skb_shinfo(skb)->frag_list != NULL;
3404}
3405
3406static inline void skb_frag_list_init(struct sk_buff *skb)
3407{
3408 skb_shinfo(skb)->frag_list = NULL;
3409}
3410
ee039871
DM
3411#define skb_walk_frags(skb, iter) \
3412 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3413
ea3793ee
RW
3414
3415int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3416 const struct sk_buff *skb);
65101aec
PA
3417struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3418 struct sk_buff_head *queue,
3419 unsigned int flags,
3420 void (*destructor)(struct sock *sk,
3421 struct sk_buff *skb),
fd69c399 3422 int *off, int *err,
65101aec 3423 struct sk_buff **last);
ea3793ee 3424struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3425 void (*destructor)(struct sock *sk,
3426 struct sk_buff *skb),
fd69c399 3427 int *off, int *err,
ea3793ee 3428 struct sk_buff **last);
7965bd4d 3429struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3430 void (*destructor)(struct sock *sk,
3431 struct sk_buff *skb),
fd69c399 3432 int *off, int *err);
7965bd4d
JP
3433struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3434 int *err);
a11e1d43
LT
3435__poll_t datagram_poll(struct file *file, struct socket *sock,
3436 struct poll_table_struct *wait);
c0371da6
AV
3437int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3438 struct iov_iter *to, int size);
51f3d02b
DM
3439static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3440 struct msghdr *msg, int size)
3441{
e5a4b0bb 3442 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3443}
e5a4b0bb
AV
3444int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3445 struct msghdr *msg);
65d69e25
SG
3446int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3447 struct iov_iter *to, int len,
3448 struct ahash_request *hash);
3a654f97
AV
3449int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3450 struct iov_iter *from, int len);
3a654f97 3451int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3452void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3453void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3454static inline void skb_free_datagram_locked(struct sock *sk,
3455 struct sk_buff *skb)
3456{
3457 __skb_free_datagram_locked(sk, skb, 0);
3458}
7965bd4d 3459int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3460int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3461int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3462__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3463 int len, __wsum csum);
a60e3cc7 3464int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3465 struct pipe_inode_info *pipe, unsigned int len,
25869262 3466 unsigned int flags);
20bf50de
TH
3467int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3468 int len);
7965bd4d 3469void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3470unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3471int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3472 int len, int hlen);
7965bd4d
JP
3473void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3474int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3475void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3476bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3477bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3478struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3479struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3480int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3481int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3482int skb_vlan_pop(struct sk_buff *skb);
3483int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
77913a4d
DC
3484int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3485 int mac_len);
3486int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len);
d27cf5c5 3487int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3488int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3489struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3490 gfp_t gfp);
20380731 3491
6ce8e9ce
AV
3492static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3493{
3073f070 3494 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3495}
3496
7eab8d9e
AV
3497static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3498{
e5a4b0bb 3499 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3500}
3501
2817a336
DB
3502struct skb_checksum_ops {
3503 __wsum (*update)(const void *mem, int len, __wsum wsum);
3504 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3505};
3506
9617813d
DC
3507extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3508
2817a336
DB
3509__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3510 __wsum csum, const struct skb_checksum_ops *ops);
3511__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3512 __wsum csum);
3513
1e98a0f0
ED
3514static inline void * __must_check
3515__skb_header_pointer(const struct sk_buff *skb, int offset,
3516 int len, void *data, int hlen, void *buffer)
1da177e4 3517{
55820ee2 3518 if (hlen - offset >= len)
690e36e7 3519 return data + offset;
1da177e4 3520
690e36e7
DM
3521 if (!skb ||
3522 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3523 return NULL;
3524
3525 return buffer;
3526}
3527
1e98a0f0
ED
3528static inline void * __must_check
3529skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3530{
3531 return __skb_header_pointer(skb, offset, len, skb->data,
3532 skb_headlen(skb), buffer);
3533}
3534
4262e5cc
DB
3535/**
3536 * skb_needs_linearize - check if we need to linearize a given skb
3537 * depending on the given device features.
3538 * @skb: socket buffer to check
3539 * @features: net device features
3540 *
3541 * Returns true if either:
3542 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3543 * 2. skb is fragmented and the device does not support SG.
3544 */
3545static inline bool skb_needs_linearize(struct sk_buff *skb,
3546 netdev_features_t features)
3547{
3548 return skb_is_nonlinear(skb) &&
3549 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3550 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3551}
3552
d626f62b
ACM
3553static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3554 void *to,
3555 const unsigned int len)
3556{
3557 memcpy(to, skb->data, len);
3558}
3559
3560static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3561 const int offset, void *to,
3562 const unsigned int len)
3563{
3564 memcpy(to, skb->data + offset, len);
3565}
3566
27d7ff46
ACM
3567static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3568 const void *from,
3569 const unsigned int len)
3570{
3571 memcpy(skb->data, from, len);
3572}
3573
3574static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3575 const int offset,
3576 const void *from,
3577 const unsigned int len)
3578{
3579 memcpy(skb->data + offset, from, len);
3580}
3581
7965bd4d 3582void skb_init(void);
1da177e4 3583
ac45f602
PO
3584static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3585{
3586 return skb->tstamp;
3587}
3588
a61bbcf2
PM
3589/**
3590 * skb_get_timestamp - get timestamp from a skb
3591 * @skb: skb to get stamp from
13c6ee2a 3592 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3593 *
3594 * Timestamps are stored in the skb as offsets to a base timestamp.
3595 * This function converts the offset back to a struct timeval and stores
3596 * it in stamp.
3597 */
ac45f602 3598static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3599 struct __kernel_old_timeval *stamp)
a61bbcf2 3600{
13c6ee2a 3601 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3602}
3603
887feae3
DD
3604static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3605 struct __kernel_sock_timeval *stamp)
3606{
3607 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3608
3609 stamp->tv_sec = ts.tv_sec;
3610 stamp->tv_usec = ts.tv_nsec / 1000;
3611}
3612
ac45f602
PO
3613static inline void skb_get_timestampns(const struct sk_buff *skb,
3614 struct timespec *stamp)
3615{
3616 *stamp = ktime_to_timespec(skb->tstamp);
3617}
3618
887feae3
DD
3619static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3620 struct __kernel_timespec *stamp)
3621{
3622 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3623
3624 stamp->tv_sec = ts.tv_sec;
3625 stamp->tv_nsec = ts.tv_nsec;
3626}
3627
b7aa0bf7 3628static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3629{
b7aa0bf7 3630 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3631}
3632
164891aa
SH
3633static inline ktime_t net_timedelta(ktime_t t)
3634{
3635 return ktime_sub(ktime_get_real(), t);
3636}
3637
b9ce204f
IJ
3638static inline ktime_t net_invalid_timestamp(void)
3639{
8b0e1953 3640 return 0;
b9ce204f 3641}
a61bbcf2 3642
de8f3a83
DB
3643static inline u8 skb_metadata_len(const struct sk_buff *skb)
3644{
3645 return skb_shinfo(skb)->meta_len;
3646}
3647
3648static inline void *skb_metadata_end(const struct sk_buff *skb)
3649{
3650 return skb_mac_header(skb);
3651}
3652
3653static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3654 const struct sk_buff *skb_b,
3655 u8 meta_len)
3656{
3657 const void *a = skb_metadata_end(skb_a);
3658 const void *b = skb_metadata_end(skb_b);
3659 /* Using more efficient varaiant than plain call to memcmp(). */
3660#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3661 u64 diffs = 0;
3662
3663 switch (meta_len) {
3664#define __it(x, op) (x -= sizeof(u##op))
3665#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3666 case 32: diffs |= __it_diff(a, b, 64);
82385b0d 3667 /* fall through */
de8f3a83 3668 case 24: diffs |= __it_diff(a, b, 64);
82385b0d 3669 /* fall through */
de8f3a83 3670 case 16: diffs |= __it_diff(a, b, 64);
82385b0d 3671 /* fall through */
de8f3a83
DB
3672 case 8: diffs |= __it_diff(a, b, 64);
3673 break;
3674 case 28: diffs |= __it_diff(a, b, 64);
82385b0d 3675 /* fall through */
de8f3a83 3676 case 20: diffs |= __it_diff(a, b, 64);
82385b0d 3677 /* fall through */
de8f3a83 3678 case 12: diffs |= __it_diff(a, b, 64);
82385b0d 3679 /* fall through */
de8f3a83
DB
3680 case 4: diffs |= __it_diff(a, b, 32);
3681 break;
3682 }
3683 return diffs;
3684#else
3685 return memcmp(a - meta_len, b - meta_len, meta_len);
3686#endif
3687}
3688
3689static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3690 const struct sk_buff *skb_b)
3691{
3692 u8 len_a = skb_metadata_len(skb_a);
3693 u8 len_b = skb_metadata_len(skb_b);
3694
3695 if (!(len_a | len_b))
3696 return false;
3697
3698 return len_a != len_b ?
3699 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3700}
3701
3702static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3703{
3704 skb_shinfo(skb)->meta_len = meta_len;
3705}
3706
3707static inline void skb_metadata_clear(struct sk_buff *skb)
3708{
3709 skb_metadata_set(skb, 0);
3710}
3711
62bccb8c
AD
3712struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3713
c1f19b51
RC
3714#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3715
7965bd4d
JP
3716void skb_clone_tx_timestamp(struct sk_buff *skb);
3717bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3718
3719#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3720
3721static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3722{
3723}
3724
3725static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3726{
3727 return false;
3728}
3729
3730#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3731
3732/**
3733 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3734 *
da92b194
RC
3735 * PHY drivers may accept clones of transmitted packets for
3736 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3737 * must call this function to return the skb back to the stack with a
3738 * timestamp.
da92b194 3739 *
c1f19b51 3740 * @skb: clone of the the original outgoing packet
7a76a021 3741 * @hwtstamps: hardware time stamps
c1f19b51
RC
3742 *
3743 */
3744void skb_complete_tx_timestamp(struct sk_buff *skb,
3745 struct skb_shared_hwtstamps *hwtstamps);
3746
e7fd2885
WB
3747void __skb_tstamp_tx(struct sk_buff *orig_skb,
3748 struct skb_shared_hwtstamps *hwtstamps,
3749 struct sock *sk, int tstype);
3750
ac45f602
PO
3751/**
3752 * skb_tstamp_tx - queue clone of skb with send time stamps
3753 * @orig_skb: the original outgoing packet
3754 * @hwtstamps: hardware time stamps, may be NULL if not available
3755 *
3756 * If the skb has a socket associated, then this function clones the
3757 * skb (thus sharing the actual data and optional structures), stores
3758 * the optional hardware time stamping information (if non NULL) or
3759 * generates a software time stamp (otherwise), then queues the clone
3760 * to the error queue of the socket. Errors are silently ignored.
3761 */
7965bd4d
JP
3762void skb_tstamp_tx(struct sk_buff *orig_skb,
3763 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3764
4507a715
RC
3765/**
3766 * skb_tx_timestamp() - Driver hook for transmit timestamping
3767 *
3768 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3769 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3770 *
73409f3b
DM
3771 * Specifically, one should make absolutely sure that this function is
3772 * called before TX completion of this packet can trigger. Otherwise
3773 * the packet could potentially already be freed.
3774 *
4507a715
RC
3775 * @skb: A socket buffer.
3776 */
3777static inline void skb_tx_timestamp(struct sk_buff *skb)
3778{
c1f19b51 3779 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3780 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3781 skb_tstamp_tx(skb, NULL);
4507a715
RC
3782}
3783
6e3e939f
JB
3784/**
3785 * skb_complete_wifi_ack - deliver skb with wifi status
3786 *
3787 * @skb: the original outgoing packet
3788 * @acked: ack status
3789 *
3790 */
3791void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3792
7965bd4d
JP
3793__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3794__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3795
60476372
HX
3796static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3797{
6edec0e6
TH
3798 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3799 skb->csum_valid ||
3800 (skb->ip_summed == CHECKSUM_PARTIAL &&
3801 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3802}
3803
fb286bb2
HX
3804/**
3805 * skb_checksum_complete - Calculate checksum of an entire packet
3806 * @skb: packet to process
3807 *
3808 * This function calculates the checksum over the entire packet plus
3809 * the value of skb->csum. The latter can be used to supply the
3810 * checksum of a pseudo header as used by TCP/UDP. It returns the
3811 * checksum.
3812 *
3813 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3814 * this function can be used to verify that checksum on received
3815 * packets. In that case the function should return zero if the
3816 * checksum is correct. In particular, this function will return zero
3817 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3818 * hardware has already verified the correctness of the checksum.
3819 */
4381ca3c 3820static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3821{
60476372
HX
3822 return skb_csum_unnecessary(skb) ?
3823 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3824}
3825
77cffe23
TH
3826static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3827{
3828 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3829 if (skb->csum_level == 0)
3830 skb->ip_summed = CHECKSUM_NONE;
3831 else
3832 skb->csum_level--;
3833 }
3834}
3835
3836static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3837{
3838 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3839 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3840 skb->csum_level++;
3841 } else if (skb->ip_summed == CHECKSUM_NONE) {
3842 skb->ip_summed = CHECKSUM_UNNECESSARY;
3843 skb->csum_level = 0;
3844 }
3845}
3846
76ba0aae
TH
3847/* Check if we need to perform checksum complete validation.
3848 *
3849 * Returns true if checksum complete is needed, false otherwise
3850 * (either checksum is unnecessary or zero checksum is allowed).
3851 */
3852static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3853 bool zero_okay,
3854 __sum16 check)
3855{
5d0c2b95
TH
3856 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3857 skb->csum_valid = 1;
77cffe23 3858 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3859 return false;
3860 }
3861
3862 return true;
3863}
3864
da279887 3865/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
3866 * in checksum_init.
3867 */
3868#define CHECKSUM_BREAK 76
3869
4e18b9ad
TH
3870/* Unset checksum-complete
3871 *
3872 * Unset checksum complete can be done when packet is being modified
3873 * (uncompressed for instance) and checksum-complete value is
3874 * invalidated.
3875 */
3876static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3877{
3878 if (skb->ip_summed == CHECKSUM_COMPLETE)
3879 skb->ip_summed = CHECKSUM_NONE;
3880}
3881
76ba0aae
TH
3882/* Validate (init) checksum based on checksum complete.
3883 *
3884 * Return values:
3885 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3886 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3887 * checksum is stored in skb->csum for use in __skb_checksum_complete
3888 * non-zero: value of invalid checksum
3889 *
3890 */
3891static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3892 bool complete,
3893 __wsum psum)
3894{
3895 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3896 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3897 skb->csum_valid = 1;
76ba0aae
TH
3898 return 0;
3899 }
3900 }
3901
3902 skb->csum = psum;
3903
5d0c2b95
TH
3904 if (complete || skb->len <= CHECKSUM_BREAK) {
3905 __sum16 csum;
3906
3907 csum = __skb_checksum_complete(skb);
3908 skb->csum_valid = !csum;
3909 return csum;
3910 }
76ba0aae
TH
3911
3912 return 0;
3913}
3914
3915static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3916{
3917 return 0;
3918}
3919
3920/* Perform checksum validate (init). Note that this is a macro since we only
3921 * want to calculate the pseudo header which is an input function if necessary.
3922 * First we try to validate without any computation (checksum unnecessary) and
3923 * then calculate based on checksum complete calling the function to compute
3924 * pseudo header.
3925 *
3926 * Return values:
3927 * 0: checksum is validated or try to in skb_checksum_complete
3928 * non-zero: value of invalid checksum
3929 */
3930#define __skb_checksum_validate(skb, proto, complete, \
3931 zero_okay, check, compute_pseudo) \
3932({ \
3933 __sum16 __ret = 0; \
5d0c2b95 3934 skb->csum_valid = 0; \
76ba0aae
TH
3935 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3936 __ret = __skb_checksum_validate_complete(skb, \
3937 complete, compute_pseudo(skb, proto)); \
3938 __ret; \
3939})
3940
3941#define skb_checksum_init(skb, proto, compute_pseudo) \
3942 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3943
3944#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3945 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3946
3947#define skb_checksum_validate(skb, proto, compute_pseudo) \
3948 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3949
3950#define skb_checksum_validate_zero_check(skb, proto, check, \
3951 compute_pseudo) \
096a4cfa 3952 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3953
3954#define skb_checksum_simple_validate(skb) \
3955 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3956
d96535a1
TH
3957static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3958{
219f1d79 3959 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
3960}
3961
e4aa33ad 3962static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
3963{
3964 skb->csum = ~pseudo;
3965 skb->ip_summed = CHECKSUM_COMPLETE;
3966}
3967
e4aa33ad 3968#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
3969do { \
3970 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 3971 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
3972} while (0)
3973
15e2396d
TH
3974static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3975 u16 start, u16 offset)
3976{
3977 skb->ip_summed = CHECKSUM_PARTIAL;
3978 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3979 skb->csum_offset = offset - start;
3980}
3981
dcdc8994
TH
3982/* Update skbuf and packet to reflect the remote checksum offload operation.
3983 * When called, ptr indicates the starting point for skb->csum when
3984 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3985 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3986 */
3987static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3988 int start, int offset, bool nopartial)
dcdc8994
TH
3989{
3990 __wsum delta;
3991
15e2396d
TH
3992 if (!nopartial) {
3993 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3994 return;
3995 }
3996
dcdc8994
TH
3997 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3998 __skb_checksum_complete(skb);
3999 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4000 }
4001
4002 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4003
4004 /* Adjust skb->csum since we changed the packet */
4005 skb->csum = csum_add(skb->csum, delta);
4006}
4007
cb9c6836
FW
4008static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4009{
4010#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 4011 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
4012#else
4013 return NULL;
4014#endif
4015}
4016
5f79e0f9 4017#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 4018void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
4019static inline void nf_conntrack_put(struct nf_conntrack *nfct)
4020{
4021 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 4022 nf_conntrack_destroy(nfct);
1da177e4
LT
4023}
4024static inline void nf_conntrack_get(struct nf_conntrack *nfct)
4025{
4026 if (nfct)
4027 atomic_inc(&nfct->use);
4028}
2fc72c7b 4029#endif
df5042f4
FW
4030
4031#ifdef CONFIG_SKB_EXTENSIONS
4032enum skb_ext_id {
4033#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4034 SKB_EXT_BRIDGE_NF,
4165079b
FW
4035#endif
4036#ifdef CONFIG_XFRM
4037 SKB_EXT_SEC_PATH,
df5042f4
FW
4038#endif
4039 SKB_EXT_NUM, /* must be last */
4040};
4041
4042/**
4043 * struct skb_ext - sk_buff extensions
4044 * @refcnt: 1 on allocation, deallocated on 0
4045 * @offset: offset to add to @data to obtain extension address
4046 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4047 * @data: start of extension data, variable sized
4048 *
4049 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4050 * to use 'u8' types while allowing up to 2kb worth of extension data.
4051 */
4052struct skb_ext {
4053 refcount_t refcnt;
4054 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4055 u8 chunks; /* same */
4056 char data[0] __aligned(8);
4057};
4058
4059void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4060void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4061void __skb_ext_put(struct skb_ext *ext);
4062
4063static inline void skb_ext_put(struct sk_buff *skb)
4064{
4065 if (skb->active_extensions)
4066 __skb_ext_put(skb->extensions);
4067}
4068
df5042f4
FW
4069static inline void __skb_ext_copy(struct sk_buff *dst,
4070 const struct sk_buff *src)
4071{
4072 dst->active_extensions = src->active_extensions;
4073
4074 if (src->active_extensions) {
4075 struct skb_ext *ext = src->extensions;
4076
4077 refcount_inc(&ext->refcnt);
4078 dst->extensions = ext;
4079 }
4080}
4081
4082static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4083{
4084 skb_ext_put(dst);
4085 __skb_ext_copy(dst, src);
4086}
4087
4088static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4089{
4090 return !!ext->offset[i];
4091}
4092
4093static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4094{
4095 return skb->active_extensions & (1 << id);
4096}
4097
4098static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4099{
4100 if (skb_ext_exist(skb, id))
4101 __skb_ext_del(skb, id);
4102}
4103
4104static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4105{
4106 if (skb_ext_exist(skb, id)) {
4107 struct skb_ext *ext = skb->extensions;
4108
4109 return (void *)ext + (ext->offset[id] << 3);
4110 }
4111
4112 return NULL;
4113}
4114#else
4115static inline void skb_ext_put(struct sk_buff *skb) {}
df5042f4
FW
4116static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4117static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4118static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4119#endif /* CONFIG_SKB_EXTENSIONS */
4120
a193a4ab
PM
4121static inline void nf_reset(struct sk_buff *skb)
4122{
5f79e0f9 4123#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4124 nf_conntrack_put(skb_nfct(skb));
4125 skb->_nfct = 0;
2fc72c7b 4126#endif
34666d46 4127#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
de8bda1d 4128 skb_ext_del(skb, SKB_EXT_BRIDGE_NF);
a193a4ab
PM
4129#endif
4130}
4131
124dff01
PM
4132static inline void nf_reset_trace(struct sk_buff *skb)
4133{
478b360a 4134#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4135 skb->nf_trace = 0;
4136#endif
a193a4ab
PM
4137}
4138
2b5ec1a5
YY
4139static inline void ipvs_reset(struct sk_buff *skb)
4140{
4141#if IS_ENABLED(CONFIG_IP_VS)
4142 skb->ipvs_property = 0;
4143#endif
4144}
4145
de8bda1d 4146/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4147static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4148 bool copy)
edda553c 4149{
5f79e0f9 4150#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4151 dst->_nfct = src->_nfct;
4152 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4153#endif
478b360a 4154#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4155 if (copy)
4156 dst->nf_trace = src->nf_trace;
478b360a 4157#endif
edda553c
YK
4158}
4159
e7ac05f3
YK
4160static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4161{
e7ac05f3 4162#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4163 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4164#endif
b1937227 4165 __nf_copy(dst, src, true);
e7ac05f3
YK
4166}
4167
984bc16c
JM
4168#ifdef CONFIG_NETWORK_SECMARK
4169static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4170{
4171 to->secmark = from->secmark;
4172}
4173
4174static inline void skb_init_secmark(struct sk_buff *skb)
4175{
4176 skb->secmark = 0;
4177}
4178#else
4179static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4180{ }
4181
4182static inline void skb_init_secmark(struct sk_buff *skb)
4183{ }
4184#endif
4185
7af8f4ca
FW
4186static inline int secpath_exists(const struct sk_buff *skb)
4187{
4188#ifdef CONFIG_XFRM
4165079b 4189 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4190#else
4191 return 0;
4192#endif
4193}
4194
574f7194
EB
4195static inline bool skb_irq_freeable(const struct sk_buff *skb)
4196{
4197 return !skb->destructor &&
7af8f4ca 4198 !secpath_exists(skb) &&
cb9c6836 4199 !skb_nfct(skb) &&
574f7194
EB
4200 !skb->_skb_refdst &&
4201 !skb_has_frag_list(skb);
4202}
4203
f25f4e44
PWJ
4204static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4205{
f25f4e44 4206 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4207}
4208
9247744e 4209static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4210{
4e3ab47a 4211 return skb->queue_mapping;
4e3ab47a
PE
4212}
4213
f25f4e44
PWJ
4214static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4215{
f25f4e44 4216 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4217}
4218
d5a9e24a
DM
4219static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4220{
4221 skb->queue_mapping = rx_queue + 1;
4222}
4223
9247744e 4224static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4225{
4226 return skb->queue_mapping - 1;
4227}
4228
9247744e 4229static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4230{
a02cec21 4231 return skb->queue_mapping != 0;
d5a9e24a
DM
4232}
4233
4ff06203
JA
4234static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4235{
4236 skb->dst_pending_confirm = val;
4237}
4238
4239static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4240{
4241 return skb->dst_pending_confirm != 0;
4242}
4243
2294be0f 4244static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4245{
0b3d8e08 4246#ifdef CONFIG_XFRM
4165079b 4247 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4248#else
def8b4fa 4249 return NULL;
def8b4fa 4250#endif
0b3d8e08 4251}
def8b4fa 4252
68c33163
PS
4253/* Keeps track of mac header offset relative to skb->head.
4254 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4255 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4256 * tunnel skb it points to outer mac header.
4257 * Keeps track of level of encapsulation of network headers.
4258 */
68c33163 4259struct skb_gso_cb {
802ab55a
AD
4260 union {
4261 int mac_offset;
4262 int data_offset;
4263 };
3347c960 4264 int encap_level;
76443456 4265 __wsum csum;
7e2b10c1 4266 __u16 csum_start;
68c33163 4267};
9207f9d4
KK
4268#define SKB_SGO_CB_OFFSET 32
4269#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
4270
4271static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4272{
4273 return (skb_mac_header(inner_skb) - inner_skb->head) -
4274 SKB_GSO_CB(inner_skb)->mac_offset;
4275}
4276
1e2bd517
PS
4277static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4278{
4279 int new_headroom, headroom;
4280 int ret;
4281
4282 headroom = skb_headroom(skb);
4283 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4284 if (ret)
4285 return ret;
4286
4287 new_headroom = skb_headroom(skb);
4288 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4289 return 0;
4290}
4291
08b64fcc
AD
4292static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4293{
4294 /* Do not update partial checksums if remote checksum is enabled. */
4295 if (skb->remcsum_offload)
4296 return;
4297
4298 SKB_GSO_CB(skb)->csum = res;
4299 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4300}
4301
7e2b10c1
TH
4302/* Compute the checksum for a gso segment. First compute the checksum value
4303 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4304 * then add in skb->csum (checksum from csum_start to end of packet).
4305 * skb->csum and csum_start are then updated to reflect the checksum of the
4306 * resultant packet starting from the transport header-- the resultant checksum
4307 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4308 * header.
4309 */
4310static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4311{
76443456
AD
4312 unsigned char *csum_start = skb_transport_header(skb);
4313 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4314 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4315
76443456
AD
4316 SKB_GSO_CB(skb)->csum = res;
4317 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4318
76443456 4319 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4320}
4321
bdcc0924 4322static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4323{
4324 return skb_shinfo(skb)->gso_size;
4325}
4326
36a8f39e 4327/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4328static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4329{
4330 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4331}
4332
d02f51cb
DA
4333/* Note: Should be called only if skb_is_gso(skb) is true */
4334static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4335{
4336 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4337}
4338
4c3024de 4339/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4340static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4341{
4c3024de 4342 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4343}
4344
5293efe6
DB
4345static inline void skb_gso_reset(struct sk_buff *skb)
4346{
4347 skb_shinfo(skb)->gso_size = 0;
4348 skb_shinfo(skb)->gso_segs = 0;
4349 skb_shinfo(skb)->gso_type = 0;
4350}
4351
d02f51cb
DA
4352static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4353 u16 increment)
4354{
4355 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4356 return;
4357 shinfo->gso_size += increment;
4358}
4359
4360static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4361 u16 decrement)
4362{
4363 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4364 return;
4365 shinfo->gso_size -= decrement;
4366}
4367
7965bd4d 4368void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4369
4370static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4371{
4372 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4373 * wanted then gso_type will be set. */
05bdd2f1
ED
4374 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4375
b78462eb
AD
4376 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4377 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4378 __skb_warn_lro_forwarding(skb);
4379 return true;
4380 }
4381 return false;
4382}
4383
35fc92a9
HX
4384static inline void skb_forward_csum(struct sk_buff *skb)
4385{
4386 /* Unfortunately we don't support this one. Any brave souls? */
4387 if (skb->ip_summed == CHECKSUM_COMPLETE)
4388 skb->ip_summed = CHECKSUM_NONE;
4389}
4390
bc8acf2c
ED
4391/**
4392 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4393 * @skb: skb to check
4394 *
4395 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4396 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4397 * use this helper, to document places where we make this assertion.
4398 */
05bdd2f1 4399static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4400{
4401#ifdef DEBUG
4402 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4403#endif
4404}
4405
f35d9d8a 4406bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4407
ed1f50c3 4408int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4409struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4410 unsigned int transport_len,
4411 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4412
3a7c1ee4
AD
4413/**
4414 * skb_head_is_locked - Determine if the skb->head is locked down
4415 * @skb: skb to check
4416 *
4417 * The head on skbs build around a head frag can be removed if they are
4418 * not cloned. This function returns true if the skb head is locked down
4419 * due to either being allocated via kmalloc, or by being a clone with
4420 * multiple references to the head.
4421 */
4422static inline bool skb_head_is_locked(const struct sk_buff *skb)
4423{
4424 return !skb->head_frag || skb_cloned(skb);
4425}
fe6cc55f 4426
179bc67f
EC
4427/* Local Checksum Offload.
4428 * Compute outer checksum based on the assumption that the
4429 * inner checksum will be offloaded later.
d0dcde64 4430 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4431 * explanation of how this works.
179bc67f
EC
4432 * Fill in outer checksum adjustment (e.g. with sum of outer
4433 * pseudo-header) before calling.
4434 * Also ensure that inner checksum is in linear data area.
4435 */
4436static inline __wsum lco_csum(struct sk_buff *skb)
4437{
9e74a6da
AD
4438 unsigned char *csum_start = skb_checksum_start(skb);
4439 unsigned char *l4_hdr = skb_transport_header(skb);
4440 __wsum partial;
179bc67f
EC
4441
4442 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4443 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4444 skb->csum_offset));
4445
179bc67f 4446 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4447 * adjustment filled in by caller) and return result.
179bc67f 4448 */
9e74a6da 4449 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4450}
4451
1da177e4
LT
4452#endif /* __KERNEL__ */
4453#endif /* _LINUX_SKBUFF_H */