bpf, net: Rework cookie generator as per-cpu one
[linux-block.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>
8842d285 17#include <linux/bvec.h>
1da177e4 18#include <linux/cache.h>
56b17425 19#include <linux/rbtree.h>
51f3d02b 20#include <linux/socket.h>
c1d1b437 21#include <linux/refcount.h>
1da177e4 22
60063497 23#include <linux/atomic.h>
1da177e4
LT
24#include <asm/types.h>
25#include <linux/spinlock.h>
1da177e4 26#include <linux/net.h>
3fc7e8a6 27#include <linux/textsearch.h>
1da177e4 28#include <net/checksum.h>
a80958f4 29#include <linux/rcupdate.h>
b7aa0bf7 30#include <linux/hrtimer.h>
131ea667 31#include <linux/dma-mapping.h>
c8f44aff 32#include <linux/netdev_features.h>
363ec392 33#include <linux/sched.h>
e6017571 34#include <linux/sched/clock.h>
1bd758eb 35#include <net/flow_dissector.h>
a60e3cc7 36#include <linux/splice.h>
72b31f72 37#include <linux/in6.h>
8b10cab6 38#include <linux/if_packet.h>
f70ea018 39#include <net/flow.h>
261db6c2
JS
40#if IS_ENABLED(CONFIG_NF_CONNTRACK)
41#include <linux/netfilter/nf_conntrack_common.h>
42#endif
1da177e4 43
7a6ae71b
TH
44/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
db1f00fb
DC
50 * From the stack's point of view these are capabilities offered by the driver.
51 * A driver typically only advertises features that it is capable of offloading
7a6ae71b
TH
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
db1f00fb 66 * is TCP or UDP. The IPv4 header may contain IP options.
7a6ae71b
TH
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
645f0897 74 * IPv6|UDP where the Next Header field in the IPv6
7a6ae71b
TH
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
db1f00fb 82 * This flag is only used to disable the RX checksum
7a6ae71b
TH
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
db1f00fb 88 * verification is set in skb->ip_summed. Possible values are:
78ea85f1
DB
89 *
90 * CHECKSUM_NONE:
91 *
7a6ae71b 92 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
77cffe23
TH
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
b4759dcd 112 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
db1f00fb 118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
77cffe23 119 * two. If the device were only able to verify the UDP checksum and not
db1f00fb 120 * GRE, either because it doesn't support GRE checksum or because GRE
77cffe23
TH
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
78ea85f1
DB
123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
db1f00fb 127 * packet as seen by netif_rx() and fills in skb->csum. This means the
78ea85f1
DB
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
b4759dcd
DC
130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
134 *
135 * CHECKSUM_PARTIAL:
136 *
6edec0e6
TH
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
78ea85f1 146 *
7a6ae71b
TH
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
149 *
150 * CHECKSUM_PARTIAL:
151 *
7a6ae71b 152 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 153 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
db1f00fb
DC
156 * offset of the packet, but it should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum -- it is the
7a6ae71b
TH
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
78ea85f1 173 *
7a6ae71b 174 * CHECKSUM_NONE:
78ea85f1 175 *
7a6ae71b
TH
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
78ea85f1
DB
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
db1f00fb 181 * This has the same meaning as CHECKSUM_NONE for checksum offload on
7a6ae71b 182 * output.
78ea85f1 183 *
7a6ae71b
TH
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
db1f00fb 186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
7a6ae71b
TH
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
db1f00fb 192 * will set csum_start and csum_offset accordingly, set ip_summed to
dba00306
DC
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
db1f00fb
DC
203 * accordingly. Note that there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
7a6ae71b 205 * both IP checksum offload and FCOE CRC offload must verify which offload
db1f00fb 206 * is configured for a packet, presumably by inspecting packet headers.
7a6ae71b
TH
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
db1f00fb
DC
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
215 * csum_offset are set to refer to the outermost checksum being offloaded
216 * (two offloaded checksums are possible with UDP encapsulation).
78ea85f1
DB
217 */
218
60476372 219/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
1da177e4 224
77cffe23
TH
225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
0bec8c88 228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 229#define SKB_WITH_OVERHEAD(X) \
deea84b0 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
87fb4b7b
ED
236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
7999096f 241struct ahash_request;
1da177e4 242struct net_device;
716ea3a7 243struct scatterlist;
9c55e01c 244struct pipe_inode_info;
a8f820aa 245struct iov_iter;
fd11a83d 246struct napi_struct;
d58e468b
PP
247struct bpf_prog;
248union bpf_attr;
df5042f4 249struct skb_ext;
1da177e4 250
34666d46 251#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 252struct nf_bridge_info {
3eaf4025
FW
253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
7fb48c5b 257 } orig_proto:8;
72b1e5e4
FW
258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
411ffb4f 261 __u16 frag_max_size;
bf1ac5ca 262 struct net_device *physindev;
63cdbc06
FW
263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
7fb48c5b 266 union {
72b1e5e4 267 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
72b1e5e4
FW
270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
72b31f72 276 };
1da177e4
LT
277};
278#endif
279
95a7233c
PB
280#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
281/* Chain in tc_skb_ext will be used to share the tc chain with
282 * ovs recirc_id. It will be set to the current chain by tc
283 * and read by ovs to recirc_id.
284 */
285struct tc_skb_ext {
286 __u32 chain;
038ebb1a 287 __u16 mru;
95a7233c
PB
288};
289#endif
290
1da177e4
LT
291struct sk_buff_head {
292 /* These two members must be first. */
293 struct sk_buff *next;
294 struct sk_buff *prev;
295
296 __u32 qlen;
297 spinlock_t lock;
298};
299
300struct sk_buff;
301
9d4dde52
IC
302/* To allow 64K frame to be packed as single skb without frag_list we
303 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
304 * buffers which do not start on a page boundary.
305 *
306 * Since GRO uses frags we allocate at least 16 regardless of page
307 * size.
a715dea3 308 */
9d4dde52 309#if (65536/PAGE_SIZE + 1) < 16
eec00954 310#define MAX_SKB_FRAGS 16UL
a715dea3 311#else
9d4dde52 312#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 313#endif
5f74f82e 314extern int sysctl_max_skb_frags;
1da177e4 315
3953c46c
MRL
316/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
317 * segment using its current segmentation instead.
318 */
319#define GSO_BY_FRAGS 0xFFFF
320
8842d285 321typedef struct bio_vec skb_frag_t;
1da177e4 322
161e6137 323/**
7240b60c 324 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
325 * @frag: skb fragment
326 */
9e903e08
ED
327static inline unsigned int skb_frag_size(const skb_frag_t *frag)
328{
b8b576a1 329 return frag->bv_len;
9e903e08
ED
330}
331
161e6137 332/**
7240b60c 333 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
334 * @frag: skb fragment
335 * @size: size of fragment
336 */
9e903e08
ED
337static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
338{
b8b576a1 339 frag->bv_len = size;
9e903e08
ED
340}
341
161e6137 342/**
7240b60c 343 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
344 * @frag: skb fragment
345 * @delta: value to add
346 */
9e903e08
ED
347static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
348{
b8b576a1 349 frag->bv_len += delta;
9e903e08
ED
350}
351
161e6137 352/**
7240b60c 353 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
354 * @frag: skb fragment
355 * @delta: value to subtract
356 */
9e903e08
ED
357static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
358{
b8b576a1 359 frag->bv_len -= delta;
9e903e08
ED
360}
361
161e6137
PT
362/**
363 * skb_frag_must_loop - Test if %p is a high memory page
364 * @p: fragment's page
365 */
c613c209
WB
366static inline bool skb_frag_must_loop(struct page *p)
367{
368#if defined(CONFIG_HIGHMEM)
369 if (PageHighMem(p))
370 return true;
371#endif
372 return false;
373}
374
375/**
376 * skb_frag_foreach_page - loop over pages in a fragment
377 *
378 * @f: skb frag to operate on
1dfa5bd3 379 * @f_off: offset from start of f->bv_page
c613c209
WB
380 * @f_len: length from f_off to loop over
381 * @p: (temp var) current page
382 * @p_off: (temp var) offset from start of current page,
383 * non-zero only on first page.
384 * @p_len: (temp var) length in current page,
385 * < PAGE_SIZE only on first and last page.
386 * @copied: (temp var) length so far, excluding current p_len.
387 *
388 * A fragment can hold a compound page, in which case per-page
389 * operations, notably kmap_atomic, must be called for each
390 * regular page.
391 */
392#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
393 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
394 p_off = (f_off) & (PAGE_SIZE - 1), \
395 p_len = skb_frag_must_loop(p) ? \
396 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
397 copied = 0; \
398 copied < f_len; \
399 copied += p_len, p++, p_off = 0, \
400 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
401
ac45f602
PO
402#define HAVE_HW_TIME_STAMP
403
404/**
d3a21be8 405 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
406 * @hwtstamp: hardware time stamp transformed into duration
407 * since arbitrary point in time
ac45f602
PO
408 *
409 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 410 * skb->tstamp.
ac45f602
PO
411 *
412 * hwtstamps can only be compared against other hwtstamps from
413 * the same device.
414 *
415 * This structure is attached to packets as part of the
416 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
417 */
418struct skb_shared_hwtstamps {
419 ktime_t hwtstamp;
ac45f602
PO
420};
421
2244d07b
OH
422/* Definitions for tx_flags in struct skb_shared_info */
423enum {
424 /* generate hardware time stamp */
425 SKBTX_HW_TSTAMP = 1 << 0,
426
e7fd2885 427 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
428 SKBTX_SW_TSTAMP = 1 << 1,
429
430 /* device driver is going to provide hardware time stamp */
431 SKBTX_IN_PROGRESS = 1 << 2,
432
a6686f2f 433 /* device driver supports TX zero-copy buffers */
62b1a8ab 434 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
435
436 /* generate wifi status information (where possible) */
62b1a8ab 437 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
438
439 /* This indicates at least one fragment might be overwritten
440 * (as in vmsplice(), sendfile() ...)
441 * If we need to compute a TX checksum, we'll need to copy
442 * all frags to avoid possible bad checksum
443 */
444 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
445
446 /* generate software time stamp when entering packet scheduling */
447 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
448};
449
52267790 450#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
e1c8a607 451#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 452 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
453#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
454
a6686f2f
SM
455/*
456 * The callback notifies userspace to release buffers when skb DMA is done in
457 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
458 * The zerocopy_success argument is true if zero copy transmit occurred,
459 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
460 * The ctx field is used to track device context.
461 * The desc field is used to track userspace buffer index.
a6686f2f
SM
462 */
463struct ubuf_info {
e19d6763 464 void (*callback)(struct ubuf_info *, bool zerocopy_success);
4ab6c99d
WB
465 union {
466 struct {
467 unsigned long desc;
468 void *ctx;
469 };
470 struct {
471 u32 id;
472 u16 len;
473 u16 zerocopy:1;
474 u32 bytelen;
475 };
476 };
c1d1b437 477 refcount_t refcnt;
a91dbff5
WB
478
479 struct mmpin {
480 struct user_struct *user;
481 unsigned int num_pg;
482 } mmp;
ac45f602
PO
483};
484
52267790
WB
485#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
486
6f89dbce
SV
487int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
488void mm_unaccount_pinned_pages(struct mmpin *mmp);
489
52267790 490struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
4ab6c99d
WB
491struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
492 struct ubuf_info *uarg);
52267790
WB
493
494static inline void sock_zerocopy_get(struct ubuf_info *uarg)
495{
c1d1b437 496 refcount_inc(&uarg->refcnt);
52267790
WB
497}
498
499void sock_zerocopy_put(struct ubuf_info *uarg);
52900d22 500void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790
WB
501
502void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
503
b5947e5d 504int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
505int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
506 struct msghdr *msg, int len,
507 struct ubuf_info *uarg);
508
1da177e4
LT
509/* This data is invariant across clones and lives at
510 * the end of the header data, ie. at skb->end.
511 */
512struct skb_shared_info {
de8f3a83
DB
513 __u8 __unused;
514 __u8 meta_len;
515 __u8 nr_frags;
9f42f126 516 __u8 tx_flags;
7967168c
HX
517 unsigned short gso_size;
518 /* Warning: this field is not always filled in (UFO)! */
519 unsigned short gso_segs;
1da177e4 520 struct sk_buff *frag_list;
ac45f602 521 struct skb_shared_hwtstamps hwtstamps;
7f564528 522 unsigned int gso_type;
09c2d251 523 u32 tskey;
ec7d2f2c
ED
524
525 /*
526 * Warning : all fields before dataref are cleared in __alloc_skb()
527 */
528 atomic_t dataref;
529
69e3c75f
JB
530 /* Intermediate layers must ensure that destructor_arg
531 * remains valid until skb destructor */
532 void * destructor_arg;
a6686f2f 533
fed66381
ED
534 /* must be last field, see pskb_expand_head() */
535 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
536};
537
538/* We divide dataref into two halves. The higher 16 bits hold references
539 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
540 * the entire skb->data. A clone of a headerless skb holds the length of
541 * the header in skb->hdr_len.
1da177e4
LT
542 *
543 * All users must obey the rule that the skb->data reference count must be
544 * greater than or equal to the payload reference count.
545 *
546 * Holding a reference to the payload part means that the user does not
547 * care about modifications to the header part of skb->data.
548 */
549#define SKB_DATAREF_SHIFT 16
550#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
551
d179cd12
DM
552
553enum {
c8753d55
VS
554 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
555 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
556 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
557};
558
7967168c
HX
559enum {
560 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
561
562 /* This indicates the skb is from an untrusted source. */
d9d30adf 563 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
564
565 /* This indicates the tcp segment has CWR set. */
d9d30adf 566 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 567
d9d30adf 568 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 569
d9d30adf 570 SKB_GSO_TCPV6 = 1 << 4,
68c33163 571
d9d30adf 572 SKB_GSO_FCOE = 1 << 5,
73136267 573
d9d30adf 574 SKB_GSO_GRE = 1 << 6,
0d89d203 575
d9d30adf 576 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 577
d9d30adf 578 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 579
d9d30adf 580 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 581
d9d30adf 582 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 583
d9d30adf 584 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 585
d9d30adf 586 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 587
d9d30adf 588 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 589
d9d30adf 590 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 591
d9d30adf 592 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
593
594 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
595
596 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
597
598 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
599};
600
2e07fa9c
ACM
601#if BITS_PER_LONG > 32
602#define NET_SKBUFF_DATA_USES_OFFSET 1
603#endif
604
605#ifdef NET_SKBUFF_DATA_USES_OFFSET
606typedef unsigned int sk_buff_data_t;
607#else
608typedef unsigned char *sk_buff_data_t;
609#endif
610
161e6137 611/**
1da177e4
LT
612 * struct sk_buff - socket buffer
613 * @next: Next buffer in list
614 * @prev: Previous buffer in list
363ec392 615 * @tstamp: Time we arrived/left
d2f273f0
RD
616 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
617 * for retransmit timer
56b17425 618 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 619 * @list: queue head
d84e0bd7 620 * @sk: Socket we are owned by
d2f273f0
RD
621 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
622 * fragmentation management
1da177e4 623 * @dev: Device we arrived on/are leaving by
d2f273f0 624 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 625 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 626 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 627 * @sp: the security path, used for xfrm
1da177e4
LT
628 * @len: Length of actual data
629 * @data_len: Data length
630 * @mac_len: Length of link layer header
334a8132 631 * @hdr_len: writable header length of cloned skb
663ead3b
HX
632 * @csum: Checksum (must include start/offset pair)
633 * @csum_start: Offset from skb->head where checksumming should start
634 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 635 * @priority: Packet queueing priority
60ff7467 636 * @ignore_df: allow local fragmentation
1da177e4 637 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 638 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
639 * @nohdr: Payload reference only, must not modify header
640 * @pkt_type: Packet class
c83c2486 641 * @fclone: skbuff clone status
c83c2486 642 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
643 * @inner_protocol_type: whether the inner protocol is
644 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
645 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
646 * @offload_fwd_mark: Packet was L2-forwarded in hardware
647 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 648 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 649 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
650 * @redirected: packet was redirected by packet classifier
651 * @from_ingress: packet was redirected from the ingress path
31729363
RD
652 * @peeked: this packet has been seen already, so stats have been
653 * done for it, don't do them again
ba9dda3a 654 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
655 * @protocol: Packet protocol from driver
656 * @destructor: Destruct function
e2080072 657 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 658 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 659 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 660 * @skb_iif: ifindex of device we arrived on
1da177e4 661 * @tc_index: Traffic control index
61b905da 662 * @hash: the packet hash
d84e0bd7 663 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
664 * @head_frag: skb was allocated from page fragments,
665 * not allocated by kmalloc() or vmalloc().
8b700862 666 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
df5042f4 667 * @active_extensions: active extensions (skb_ext_id types)
553a5672 668 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 669 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 670 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 671 * ports.
a3b18ddb 672 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
673 * @wifi_acked_valid: wifi_acked was set
674 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 675 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
676 * @encapsulation: indicates the inner headers in the skbuff are valid
677 * @encap_hdr_csum: software checksum is needed
678 * @csum_valid: checksum is already valid
dba00306 679 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
680 * @csum_complete_sw: checksum was completed by software
681 * @csum_level: indicates the number of consecutive checksums found in
682 * the packet minus one that have been verified as
683 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 684 * @dst_pending_confirm: need to confirm neighbour
a48d189e 685 * @decrypted: Decrypted SKB
161e6137 686 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 687 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 688 * @secmark: security marking
d84e0bd7 689 * @mark: Generic packet mark
d2f273f0
RD
690 * @reserved_tailroom: (aka @mark) number of bytes of free space available
691 * at the tail of an sk_buff
692 * @vlan_present: VLAN tag is present
86a9bad3 693 * @vlan_proto: vlan encapsulation protocol
6aa895b0 694 * @vlan_tci: vlan tag control information
0d89d203 695 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
696 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
697 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
698 * @inner_transport_header: Inner transport layer header (encapsulation)
699 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 700 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
701 * @transport_header: Transport layer header
702 * @network_header: Network layer header
703 * @mac_header: Link layer header
704 * @tail: Tail pointer
705 * @end: End pointer
706 * @head: Head of buffer
707 * @data: Data head pointer
708 * @truesize: Buffer size
709 * @users: User count - see {datagram,tcp}.c
df5042f4 710 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
711 */
712
713struct sk_buff {
363ec392 714 union {
56b17425
ED
715 struct {
716 /* These two members must be first. */
717 struct sk_buff *next;
718 struct sk_buff *prev;
719
720 union {
bffa72cf
ED
721 struct net_device *dev;
722 /* Some protocols might use this space to store information,
723 * while device pointer would be NULL.
724 * UDP receive path is one user.
725 */
726 unsigned long dev_scratch;
56b17425
ED
727 };
728 };
fa0f5273 729 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 730 struct list_head list;
363ec392 731 };
fa0f5273
PO
732
733 union {
734 struct sock *sk;
735 int ip_defrag_offset;
736 };
1da177e4 737
c84d9490 738 union {
bffa72cf 739 ktime_t tstamp;
d3edd06e 740 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 741 };
1da177e4
LT
742 /*
743 * This is the control buffer. It is free to use for every
744 * layer. Please put your private variables there. If you
745 * want to keep them across layers you have to do a skb_clone()
746 * first. This is owned by whoever has the skb queued ATM.
747 */
da3f5cf1 748 char cb[48] __aligned(8);
1da177e4 749
e2080072
ED
750 union {
751 struct {
752 unsigned long _skb_refdst;
753 void (*destructor)(struct sk_buff *skb);
754 };
755 struct list_head tcp_tsorted_anchor;
756 };
757
b1937227 758#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 759 unsigned long _nfct;
da3f5cf1 760#endif
1da177e4 761 unsigned int len,
334a8132
PM
762 data_len;
763 __u16 mac_len,
764 hdr_len;
b1937227
ED
765
766 /* Following fields are _not_ copied in __copy_skb_header()
767 * Note that queue_mapping is here mostly to fill a hole.
768 */
b1937227 769 __u16 queue_mapping;
36bbef52
DB
770
771/* if you move cloned around you also must adapt those constants */
772#ifdef __BIG_ENDIAN_BITFIELD
773#define CLONED_MASK (1 << 7)
774#else
775#define CLONED_MASK 1
776#endif
777#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
778
d2f273f0 779 /* private: */
36bbef52 780 __u8 __cloned_offset[0];
d2f273f0 781 /* public: */
b1937227 782 __u8 cloned:1,
6869c4d8 783 nohdr:1,
b84f4cc9 784 fclone:2,
a59322be 785 peeked:1,
b1937227 786 head_frag:1,
8b700862 787 pfmemalloc:1;
df5042f4
FW
788#ifdef CONFIG_SKB_EXTENSIONS
789 __u8 active_extensions;
790#endif
b1937227
ED
791 /* fields enclosed in headers_start/headers_end are copied
792 * using a single memcpy() in __copy_skb_header()
793 */
ebcf34f3 794 /* private: */
b1937227 795 __u32 headers_start[0];
ebcf34f3 796 /* public: */
4031ae6e 797
233577a2
HFS
798/* if you move pkt_type around you also must adapt those constants */
799#ifdef __BIG_ENDIAN_BITFIELD
800#define PKT_TYPE_MAX (7 << 5)
801#else
802#define PKT_TYPE_MAX 7
1da177e4 803#endif
233577a2 804#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 805
d2f273f0 806 /* private: */
233577a2 807 __u8 __pkt_type_offset[0];
d2f273f0 808 /* public: */
b1937227 809 __u8 pkt_type:3;
b1937227 810 __u8 ignore_df:1;
b1937227
ED
811 __u8 nf_trace:1;
812 __u8 ip_summed:2;
3853b584 813 __u8 ooo_okay:1;
8b700862 814
61b905da 815 __u8 l4_hash:1;
a3b18ddb 816 __u8 sw_hash:1;
6e3e939f
JB
817 __u8 wifi_acked_valid:1;
818 __u8 wifi_acked:1;
3bdc0eba 819 __u8 no_fcs:1;
77cffe23 820 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 821 __u8 encapsulation:1;
7e2b10c1 822 __u8 encap_hdr_csum:1;
5d0c2b95 823 __u8 csum_valid:1;
8b700862 824
0c4b2d37
MM
825#ifdef __BIG_ENDIAN_BITFIELD
826#define PKT_VLAN_PRESENT_BIT 7
827#else
828#define PKT_VLAN_PRESENT_BIT 0
829#endif
830#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 831 /* private: */
0c4b2d37 832 __u8 __pkt_vlan_present_offset[0];
d2f273f0 833 /* public: */
0c4b2d37 834 __u8 vlan_present:1;
7e3cead5 835 __u8 csum_complete_sw:1;
b1937227 836 __u8 csum_level:2;
dba00306 837 __u8 csum_not_inet:1;
4ff06203 838 __u8 dst_pending_confirm:1;
b1937227
ED
839#ifdef CONFIG_IPV6_NDISC_NODETYPE
840 __u8 ndisc_nodetype:2;
841#endif
8b700862 842
0c4b2d37 843 __u8 ipvs_property:1;
8bce6d7d 844 __u8 inner_protocol_type:1;
e585f236 845 __u8 remcsum_offload:1;
6bc506b4
IS
846#ifdef CONFIG_NET_SWITCHDEV
847 __u8 offload_fwd_mark:1;
875e8939 848 __u8 offload_l3_fwd_mark:1;
6bc506b4 849#endif
e7246e12
WB
850#ifdef CONFIG_NET_CLS_ACT
851 __u8 tc_skip_classify:1;
8dc07fdb 852 __u8 tc_at_ingress:1;
2c64605b
PNA
853#endif
854#ifdef CONFIG_NET_REDIRECT
855 __u8 redirected:1;
856 __u8 from_ingress:1;
e7246e12 857#endif
a48d189e
SB
858#ifdef CONFIG_TLS_DEVICE
859 __u8 decrypted:1;
860#endif
b1937227
ED
861
862#ifdef CONFIG_NET_SCHED
863 __u16 tc_index; /* traffic control index */
b1937227 864#endif
fe55f6d5 865
b1937227
ED
866 union {
867 __wsum csum;
868 struct {
869 __u16 csum_start;
870 __u16 csum_offset;
871 };
872 };
873 __u32 priority;
874 int skb_iif;
875 __u32 hash;
876 __be16 vlan_proto;
877 __u16 vlan_tci;
2bd82484
ED
878#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
879 union {
880 unsigned int napi_id;
881 unsigned int sender_cpu;
882 };
97fc2f08 883#endif
984bc16c 884#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 885 __u32 secmark;
0c4f691f 886#endif
0c4f691f 887
3b885787
NH
888 union {
889 __u32 mark;
16fad69c 890 __u32 reserved_tailroom;
3b885787 891 };
1da177e4 892
8bce6d7d
TH
893 union {
894 __be16 inner_protocol;
895 __u8 inner_ipproto;
896 };
897
1a37e412
SH
898 __u16 inner_transport_header;
899 __u16 inner_network_header;
900 __u16 inner_mac_header;
b1937227
ED
901
902 __be16 protocol;
1a37e412
SH
903 __u16 transport_header;
904 __u16 network_header;
905 __u16 mac_header;
b1937227 906
ebcf34f3 907 /* private: */
b1937227 908 __u32 headers_end[0];
ebcf34f3 909 /* public: */
b1937227 910
1da177e4 911 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 912 sk_buff_data_t tail;
4305b541 913 sk_buff_data_t end;
1da177e4 914 unsigned char *head,
4305b541 915 *data;
27a884dc 916 unsigned int truesize;
63354797 917 refcount_t users;
df5042f4
FW
918
919#ifdef CONFIG_SKB_EXTENSIONS
920 /* only useable after checking ->active_extensions != 0 */
921 struct skb_ext *extensions;
922#endif
1da177e4
LT
923};
924
925#ifdef __KERNEL__
926/*
927 * Handling routines are only of interest to the kernel
928 */
1da177e4 929
c93bdd0e
MG
930#define SKB_ALLOC_FCLONE 0x01
931#define SKB_ALLOC_RX 0x02
fd11a83d 932#define SKB_ALLOC_NAPI 0x04
c93bdd0e 933
161e6137
PT
934/**
935 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
936 * @skb: buffer
937 */
c93bdd0e
MG
938static inline bool skb_pfmemalloc(const struct sk_buff *skb)
939{
940 return unlikely(skb->pfmemalloc);
941}
942
7fee226a
ED
943/*
944 * skb might have a dst pointer attached, refcounted or not.
945 * _skb_refdst low order bit is set if refcount was _not_ taken
946 */
947#define SKB_DST_NOREF 1UL
948#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
949
950/**
951 * skb_dst - returns skb dst_entry
952 * @skb: buffer
953 *
954 * Returns skb dst_entry, regardless of reference taken or not.
955 */
adf30907
ED
956static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
957{
161e6137 958 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
959 * rcu_read_lock section
960 */
961 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
962 !rcu_read_lock_held() &&
963 !rcu_read_lock_bh_held());
964 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
965}
966
7fee226a
ED
967/**
968 * skb_dst_set - sets skb dst
969 * @skb: buffer
970 * @dst: dst entry
971 *
972 * Sets skb dst, assuming a reference was taken on dst and should
973 * be released by skb_dst_drop()
974 */
adf30907
ED
975static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
976{
7fee226a
ED
977 skb->_skb_refdst = (unsigned long)dst;
978}
979
932bc4d7
JA
980/**
981 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
982 * @skb: buffer
983 * @dst: dst entry
984 *
985 * Sets skb dst, assuming a reference was not taken on dst.
986 * If dst entry is cached, we do not take reference and dst_release
987 * will be avoided by refdst_drop. If dst entry is not cached, we take
988 * reference, so that last dst_release can destroy the dst immediately.
989 */
990static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
991{
dbfc4fb7
HFS
992 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
993 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 994}
7fee226a
ED
995
996/**
25985edc 997 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
998 * @skb: buffer
999 */
1000static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1001{
1002 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1003}
1004
161e6137
PT
1005/**
1006 * skb_rtable - Returns the skb &rtable
1007 * @skb: buffer
1008 */
511c3f92
ED
1009static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1010{
adf30907 1011 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1012}
1013
8b10cab6
JHS
1014/* For mangling skb->pkt_type from user space side from applications
1015 * such as nft, tc, etc, we only allow a conservative subset of
1016 * possible pkt_types to be set.
1017*/
1018static inline bool skb_pkt_type_ok(u32 ptype)
1019{
1020 return ptype <= PACKET_OTHERHOST;
1021}
1022
161e6137
PT
1023/**
1024 * skb_napi_id - Returns the skb's NAPI id
1025 * @skb: buffer
1026 */
90b602f8
ML
1027static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1028{
1029#ifdef CONFIG_NET_RX_BUSY_POLL
1030 return skb->napi_id;
1031#else
1032 return 0;
1033#endif
1034}
1035
161e6137
PT
1036/**
1037 * skb_unref - decrement the skb's reference count
1038 * @skb: buffer
1039 *
1040 * Returns true if we can free the skb.
1041 */
3889a803
PA
1042static inline bool skb_unref(struct sk_buff *skb)
1043{
1044 if (unlikely(!skb))
1045 return false;
63354797 1046 if (likely(refcount_read(&skb->users) == 1))
3889a803 1047 smp_rmb();
63354797 1048 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1049 return false;
1050
1051 return true;
1052}
1053
0a463c78 1054void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1055void kfree_skb(struct sk_buff *skb);
1056void kfree_skb_list(struct sk_buff *segs);
6413139d 1057void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1058void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1059
1060#ifdef CONFIG_TRACEPOINTS
7965bd4d 1061void consume_skb(struct sk_buff *skb);
be769db2
HX
1062#else
1063static inline void consume_skb(struct sk_buff *skb)
1064{
1065 return kfree_skb(skb);
1066}
1067#endif
1068
ca2c1418 1069void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1070void __kfree_skb(struct sk_buff *skb);
d7e8883c 1071extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1072
7965bd4d
JP
1073void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1074bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1075 bool *fragstolen, int *delta_truesize);
bad43ca8 1076
7965bd4d
JP
1077struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1078 int node);
2ea2f62c 1079struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1080struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1081struct sk_buff *build_skb_around(struct sk_buff *skb,
1082 void *data, unsigned int frag_size);
161e6137
PT
1083
1084/**
1085 * alloc_skb - allocate a network buffer
1086 * @size: size to allocate
1087 * @priority: allocation mask
1088 *
1089 * This function is a convenient wrapper around __alloc_skb().
1090 */
d179cd12 1091static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1092 gfp_t priority)
d179cd12 1093{
564824b0 1094 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1095}
1096
2e4e4410
ED
1097struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1098 unsigned long data_len,
1099 int max_page_order,
1100 int *errcode,
1101 gfp_t gfp_mask);
da29e4b4 1102struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1103
d0bf4a9e
ED
1104/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1105struct sk_buff_fclones {
1106 struct sk_buff skb1;
1107
1108 struct sk_buff skb2;
1109
2638595a 1110 refcount_t fclone_ref;
d0bf4a9e
ED
1111};
1112
1113/**
1114 * skb_fclone_busy - check if fclone is busy
293de7de 1115 * @sk: socket
d0bf4a9e
ED
1116 * @skb: buffer
1117 *
bda13fed 1118 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1119 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1120 * so we also check that this didnt happen.
d0bf4a9e 1121 */
39bb5e62
ED
1122static inline bool skb_fclone_busy(const struct sock *sk,
1123 const struct sk_buff *skb)
d0bf4a9e
ED
1124{
1125 const struct sk_buff_fclones *fclones;
1126
1127 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1128
1129 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1130 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1131 fclones->skb2.sk == sk;
d0bf4a9e
ED
1132}
1133
161e6137
PT
1134/**
1135 * alloc_skb_fclone - allocate a network buffer from fclone cache
1136 * @size: size to allocate
1137 * @priority: allocation mask
1138 *
1139 * This function is a convenient wrapper around __alloc_skb().
1140 */
d179cd12 1141static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1142 gfp_t priority)
d179cd12 1143{
c93bdd0e 1144 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1145}
1146
7965bd4d 1147struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1148void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1149int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1150struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1151void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1152struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1153struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1154 gfp_t gfp_mask, bool fclone);
1155static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1156 gfp_t gfp_mask)
1157{
1158 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1159}
7965bd4d
JP
1160
1161int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1162struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1163 unsigned int headroom);
1164struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1165 int newtailroom, gfp_t priority);
48a1df65
JD
1166int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1167 int offset, int len);
1168int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1169 int offset, int len);
7965bd4d 1170int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1171int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1172
1173/**
1174 * skb_pad - zero pad the tail of an skb
1175 * @skb: buffer to pad
1176 * @pad: space to pad
1177 *
1178 * Ensure that a buffer is followed by a padding area that is zero
1179 * filled. Used by network drivers which may DMA or transfer data
1180 * beyond the buffer end onto the wire.
1181 *
1182 * May return error in out of memory cases. The skb is freed on error.
1183 */
1184static inline int skb_pad(struct sk_buff *skb, int pad)
1185{
1186 return __skb_pad(skb, pad, true);
1187}
ead2ceb0 1188#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1189
be12a1fe
HFS
1190int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1191 int offset, size_t size);
1192
d94d9fee 1193struct skb_seq_state {
677e90ed
TG
1194 __u32 lower_offset;
1195 __u32 upper_offset;
1196 __u32 frag_idx;
1197 __u32 stepped_offset;
1198 struct sk_buff *root_skb;
1199 struct sk_buff *cur_skb;
1200 __u8 *frag_data;
1201};
1202
7965bd4d
JP
1203void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1204 unsigned int to, struct skb_seq_state *st);
1205unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1206 struct skb_seq_state *st);
1207void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1208
7965bd4d 1209unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1210 unsigned int to, struct ts_config *config);
3fc7e8a6 1211
09323cc4
TH
1212/*
1213 * Packet hash types specify the type of hash in skb_set_hash.
1214 *
1215 * Hash types refer to the protocol layer addresses which are used to
1216 * construct a packet's hash. The hashes are used to differentiate or identify
1217 * flows of the protocol layer for the hash type. Hash types are either
1218 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1219 *
1220 * Properties of hashes:
1221 *
1222 * 1) Two packets in different flows have different hash values
1223 * 2) Two packets in the same flow should have the same hash value
1224 *
1225 * A hash at a higher layer is considered to be more specific. A driver should
1226 * set the most specific hash possible.
1227 *
1228 * A driver cannot indicate a more specific hash than the layer at which a hash
1229 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1230 *
1231 * A driver may indicate a hash level which is less specific than the
1232 * actual layer the hash was computed on. For instance, a hash computed
1233 * at L4 may be considered an L3 hash. This should only be done if the
1234 * driver can't unambiguously determine that the HW computed the hash at
1235 * the higher layer. Note that the "should" in the second property above
1236 * permits this.
1237 */
1238enum pkt_hash_types {
1239 PKT_HASH_TYPE_NONE, /* Undefined type */
1240 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1241 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1242 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1243};
1244
bcc83839 1245static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1246{
bcc83839 1247 skb->hash = 0;
a3b18ddb 1248 skb->sw_hash = 0;
bcc83839
TH
1249 skb->l4_hash = 0;
1250}
1251
1252static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1253{
1254 if (!skb->l4_hash)
1255 skb_clear_hash(skb);
1256}
1257
1258static inline void
1259__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1260{
1261 skb->l4_hash = is_l4;
1262 skb->sw_hash = is_sw;
61b905da 1263 skb->hash = hash;
09323cc4
TH
1264}
1265
bcc83839
TH
1266static inline void
1267skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1268{
1269 /* Used by drivers to set hash from HW */
1270 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1271}
1272
1273static inline void
1274__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1275{
1276 __skb_set_hash(skb, hash, true, is_l4);
1277}
1278
e5276937 1279void __skb_get_hash(struct sk_buff *skb);
b917783c 1280u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1281u32 skb_get_poff(const struct sk_buff *skb);
1282u32 __skb_get_poff(const struct sk_buff *skb, void *data,
72a338bc 1283 const struct flow_keys_basic *keys, int hlen);
e5276937
TH
1284__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1285 void *data, int hlen_proto);
1286
1287static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1288 int thoff, u8 ip_proto)
1289{
1290 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1291}
1292
1293void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1294 const struct flow_dissector_key *key,
1295 unsigned int key_count);
1296
089b19a9
SF
1297struct bpf_flow_dissector;
1298bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1299 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1300
3cbf4ffb
SF
1301bool __skb_flow_dissect(const struct net *net,
1302 const struct sk_buff *skb,
e5276937
TH
1303 struct flow_dissector *flow_dissector,
1304 void *target_container,
cd79a238
TH
1305 void *data, __be16 proto, int nhoff, int hlen,
1306 unsigned int flags);
e5276937
TH
1307
1308static inline bool skb_flow_dissect(const struct sk_buff *skb,
1309 struct flow_dissector *flow_dissector,
cd79a238 1310 void *target_container, unsigned int flags)
e5276937 1311{
3cbf4ffb
SF
1312 return __skb_flow_dissect(NULL, skb, flow_dissector,
1313 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1314}
1315
1316static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1317 struct flow_keys *flow,
1318 unsigned int flags)
e5276937
TH
1319{
1320 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1321 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1322 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1323}
1324
72a338bc 1325static inline bool
3cbf4ffb
SF
1326skb_flow_dissect_flow_keys_basic(const struct net *net,
1327 const struct sk_buff *skb,
72a338bc
PA
1328 struct flow_keys_basic *flow, void *data,
1329 __be16 proto, int nhoff, int hlen,
1330 unsigned int flags)
e5276937
TH
1331{
1332 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1333 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1334 data, proto, nhoff, hlen, flags);
e5276937
TH
1335}
1336
82828b88
JP
1337void skb_flow_dissect_meta(const struct sk_buff *skb,
1338 struct flow_dissector *flow_dissector,
1339 void *target_container);
1340
75a56758 1341/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1342 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1343 * to user states.
1344 */
1345void
1346skb_flow_dissect_ct(const struct sk_buff *skb,
1347 struct flow_dissector *flow_dissector,
1348 void *target_container,
1349 u16 *ctinfo_map,
1350 size_t mapsize);
62b32379
SH
1351void
1352skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1353 struct flow_dissector *flow_dissector,
1354 void *target_container);
1355
0cb09aff
AL
1356void skb_flow_dissect_hash(const struct sk_buff *skb,
1357 struct flow_dissector *flow_dissector,
1358 void *target_container);
1359
3958afa1 1360static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1361{
a3b18ddb 1362 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1363 __skb_get_hash(skb);
bfb564e7 1364
61b905da 1365 return skb->hash;
bfb564e7
KK
1366}
1367
20a17bf6 1368static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1369{
c6cc1ca7
TH
1370 if (!skb->l4_hash && !skb->sw_hash) {
1371 struct flow_keys keys;
de4c1f8b 1372 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1373
de4c1f8b 1374 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1375 }
f70ea018
TH
1376
1377 return skb->hash;
1378}
1379
55667441
ED
1380__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1381 const siphash_key_t *perturb);
50fb7992 1382
57bdf7f4
TH
1383static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1384{
61b905da 1385 return skb->hash;
57bdf7f4
TH
1386}
1387
3df7a74e
TH
1388static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1389{
61b905da 1390 to->hash = from->hash;
a3b18ddb 1391 to->sw_hash = from->sw_hash;
61b905da 1392 to->l4_hash = from->l4_hash;
3df7a74e
TH
1393};
1394
41477662
JK
1395static inline void skb_copy_decrypted(struct sk_buff *to,
1396 const struct sk_buff *from)
1397{
1398#ifdef CONFIG_TLS_DEVICE
1399 to->decrypted = from->decrypted;
1400#endif
1401}
1402
4305b541
ACM
1403#ifdef NET_SKBUFF_DATA_USES_OFFSET
1404static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1405{
1406 return skb->head + skb->end;
1407}
ec47ea82
AD
1408
1409static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1410{
1411 return skb->end;
1412}
4305b541
ACM
1413#else
1414static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1415{
1416 return skb->end;
1417}
ec47ea82
AD
1418
1419static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1420{
1421 return skb->end - skb->head;
1422}
4305b541
ACM
1423#endif
1424
1da177e4 1425/* Internal */
4305b541 1426#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1427
ac45f602
PO
1428static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1429{
1430 return &skb_shinfo(skb)->hwtstamps;
1431}
1432
52267790
WB
1433static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1434{
1435 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1436
1437 return is_zcopy ? skb_uarg(skb) : NULL;
1438}
1439
52900d22
WB
1440static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1441 bool *have_ref)
52267790
WB
1442{
1443 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1444 if (unlikely(have_ref && *have_ref))
1445 *have_ref = false;
1446 else
1447 sock_zerocopy_get(uarg);
52267790
WB
1448 skb_shinfo(skb)->destructor_arg = uarg;
1449 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1450 }
1451}
1452
5cd8d46e
WB
1453static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1454{
1455 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1456 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1457}
1458
1459static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1460{
1461 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1462}
1463
1464static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1465{
1466 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1467}
1468
52267790
WB
1469/* Release a reference on a zerocopy structure */
1470static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1471{
1472 struct ubuf_info *uarg = skb_zcopy(skb);
1473
1474 if (uarg) {
185ce5c3
WB
1475 if (skb_zcopy_is_nouarg(skb)) {
1476 /* no notification callback */
1477 } else if (uarg->callback == sock_zerocopy_callback) {
0a4a060b
WB
1478 uarg->zerocopy = uarg->zerocopy && zerocopy;
1479 sock_zerocopy_put(uarg);
185ce5c3 1480 } else {
0a4a060b
WB
1481 uarg->callback(uarg, zerocopy);
1482 }
1483
52267790
WB
1484 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1485 }
1486}
1487
1488/* Abort a zerocopy operation and revert zckey on error in send syscall */
1489static inline void skb_zcopy_abort(struct sk_buff *skb)
1490{
1491 struct ubuf_info *uarg = skb_zcopy(skb);
1492
1493 if (uarg) {
52900d22 1494 sock_zerocopy_put_abort(uarg, false);
52267790
WB
1495 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1496 }
1497}
1498
a8305bff
DM
1499static inline void skb_mark_not_on_list(struct sk_buff *skb)
1500{
1501 skb->next = NULL;
1502}
1503
dcfea72e 1504/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1505#define skb_list_walk_safe(first, skb, next_skb) \
1506 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1507 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1508
992cba7e
DM
1509static inline void skb_list_del_init(struct sk_buff *skb)
1510{
1511 __list_del_entry(&skb->list);
1512 skb_mark_not_on_list(skb);
1513}
1514
1da177e4
LT
1515/**
1516 * skb_queue_empty - check if a queue is empty
1517 * @list: queue head
1518 *
1519 * Returns true if the queue is empty, false otherwise.
1520 */
1521static inline int skb_queue_empty(const struct sk_buff_head *list)
1522{
fd44b93c 1523 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1524}
1525
d7d16a89
ED
1526/**
1527 * skb_queue_empty_lockless - check if a queue is empty
1528 * @list: queue head
1529 *
1530 * Returns true if the queue is empty, false otherwise.
1531 * This variant can be used in lockless contexts.
1532 */
1533static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1534{
1535 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1536}
1537
1538
fc7ebb21
DM
1539/**
1540 * skb_queue_is_last - check if skb is the last entry in the queue
1541 * @list: queue head
1542 * @skb: buffer
1543 *
1544 * Returns true if @skb is the last buffer on the list.
1545 */
1546static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1547 const struct sk_buff *skb)
1548{
fd44b93c 1549 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1550}
1551
832d11c5
IJ
1552/**
1553 * skb_queue_is_first - check if skb is the first entry in the queue
1554 * @list: queue head
1555 * @skb: buffer
1556 *
1557 * Returns true if @skb is the first buffer on the list.
1558 */
1559static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1560 const struct sk_buff *skb)
1561{
fd44b93c 1562 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1563}
1564
249c8b42
DM
1565/**
1566 * skb_queue_next - return the next packet in the queue
1567 * @list: queue head
1568 * @skb: current buffer
1569 *
1570 * Return the next packet in @list after @skb. It is only valid to
1571 * call this if skb_queue_is_last() evaluates to false.
1572 */
1573static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1574 const struct sk_buff *skb)
1575{
1576 /* This BUG_ON may seem severe, but if we just return then we
1577 * are going to dereference garbage.
1578 */
1579 BUG_ON(skb_queue_is_last(list, skb));
1580 return skb->next;
1581}
1582
832d11c5
IJ
1583/**
1584 * skb_queue_prev - return the prev packet in the queue
1585 * @list: queue head
1586 * @skb: current buffer
1587 *
1588 * Return the prev packet in @list before @skb. It is only valid to
1589 * call this if skb_queue_is_first() evaluates to false.
1590 */
1591static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1592 const struct sk_buff *skb)
1593{
1594 /* This BUG_ON may seem severe, but if we just return then we
1595 * are going to dereference garbage.
1596 */
1597 BUG_ON(skb_queue_is_first(list, skb));
1598 return skb->prev;
1599}
1600
1da177e4
LT
1601/**
1602 * skb_get - reference buffer
1603 * @skb: buffer to reference
1604 *
1605 * Makes another reference to a socket buffer and returns a pointer
1606 * to the buffer.
1607 */
1608static inline struct sk_buff *skb_get(struct sk_buff *skb)
1609{
63354797 1610 refcount_inc(&skb->users);
1da177e4
LT
1611 return skb;
1612}
1613
1614/*
f8821f96 1615 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1616 */
1617
1da177e4
LT
1618/**
1619 * skb_cloned - is the buffer a clone
1620 * @skb: buffer to check
1621 *
1622 * Returns true if the buffer was generated with skb_clone() and is
1623 * one of multiple shared copies of the buffer. Cloned buffers are
1624 * shared data so must not be written to under normal circumstances.
1625 */
1626static inline int skb_cloned(const struct sk_buff *skb)
1627{
1628 return skb->cloned &&
1629 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1630}
1631
14bbd6a5
PS
1632static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1633{
d0164adc 1634 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1635
1636 if (skb_cloned(skb))
1637 return pskb_expand_head(skb, 0, 0, pri);
1638
1639 return 0;
1640}
1641
1da177e4
LT
1642/**
1643 * skb_header_cloned - is the header a clone
1644 * @skb: buffer to check
1645 *
1646 * Returns true if modifying the header part of the buffer requires
1647 * the data to be copied.
1648 */
1649static inline int skb_header_cloned(const struct sk_buff *skb)
1650{
1651 int dataref;
1652
1653 if (!skb->cloned)
1654 return 0;
1655
1656 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1657 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1658 return dataref != 1;
1659}
1660
9580bf2e
ED
1661static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1662{
1663 might_sleep_if(gfpflags_allow_blocking(pri));
1664
1665 if (skb_header_cloned(skb))
1666 return pskb_expand_head(skb, 0, 0, pri);
1667
1668 return 0;
1669}
1670
f4a775d1
ED
1671/**
1672 * __skb_header_release - release reference to header
1673 * @skb: buffer to operate on
f4a775d1
ED
1674 */
1675static inline void __skb_header_release(struct sk_buff *skb)
1676{
1677 skb->nohdr = 1;
1678 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1679}
1680
1681
1da177e4
LT
1682/**
1683 * skb_shared - is the buffer shared
1684 * @skb: buffer to check
1685 *
1686 * Returns true if more than one person has a reference to this
1687 * buffer.
1688 */
1689static inline int skb_shared(const struct sk_buff *skb)
1690{
63354797 1691 return refcount_read(&skb->users) != 1;
1da177e4
LT
1692}
1693
1694/**
1695 * skb_share_check - check if buffer is shared and if so clone it
1696 * @skb: buffer to check
1697 * @pri: priority for memory allocation
1698 *
1699 * If the buffer is shared the buffer is cloned and the old copy
1700 * drops a reference. A new clone with a single reference is returned.
1701 * If the buffer is not shared the original buffer is returned. When
1702 * being called from interrupt status or with spinlocks held pri must
1703 * be GFP_ATOMIC.
1704 *
1705 * NULL is returned on a memory allocation failure.
1706 */
47061bc4 1707static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1708{
d0164adc 1709 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1710 if (skb_shared(skb)) {
1711 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1712
1713 if (likely(nskb))
1714 consume_skb(skb);
1715 else
1716 kfree_skb(skb);
1da177e4
LT
1717 skb = nskb;
1718 }
1719 return skb;
1720}
1721
1722/*
1723 * Copy shared buffers into a new sk_buff. We effectively do COW on
1724 * packets to handle cases where we have a local reader and forward
1725 * and a couple of other messy ones. The normal one is tcpdumping
1726 * a packet thats being forwarded.
1727 */
1728
1729/**
1730 * skb_unshare - make a copy of a shared buffer
1731 * @skb: buffer to check
1732 * @pri: priority for memory allocation
1733 *
1734 * If the socket buffer is a clone then this function creates a new
1735 * copy of the data, drops a reference count on the old copy and returns
1736 * the new copy with the reference count at 1. If the buffer is not a clone
1737 * the original buffer is returned. When called with a spinlock held or
1738 * from interrupt state @pri must be %GFP_ATOMIC
1739 *
1740 * %NULL is returned on a memory allocation failure.
1741 */
e2bf521d 1742static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1743 gfp_t pri)
1da177e4 1744{
d0164adc 1745 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1746 if (skb_cloned(skb)) {
1747 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1748
1749 /* Free our shared copy */
1750 if (likely(nskb))
1751 consume_skb(skb);
1752 else
1753 kfree_skb(skb);
1da177e4
LT
1754 skb = nskb;
1755 }
1756 return skb;
1757}
1758
1759/**
1a5778aa 1760 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1761 * @list_: list to peek at
1762 *
1763 * Peek an &sk_buff. Unlike most other operations you _MUST_
1764 * be careful with this one. A peek leaves the buffer on the
1765 * list and someone else may run off with it. You must hold
1766 * the appropriate locks or have a private queue to do this.
1767 *
1768 * Returns %NULL for an empty list or a pointer to the head element.
1769 * The reference count is not incremented and the reference is therefore
1770 * volatile. Use with caution.
1771 */
05bdd2f1 1772static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1773{
18d07000
ED
1774 struct sk_buff *skb = list_->next;
1775
1776 if (skb == (struct sk_buff *)list_)
1777 skb = NULL;
1778 return skb;
1da177e4
LT
1779}
1780
8b69bd7d
DM
1781/**
1782 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1783 * @list_: list to peek at
1784 *
1785 * Like skb_peek(), but the caller knows that the list is not empty.
1786 */
1787static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1788{
1789 return list_->next;
1790}
1791
da5ef6e5
PE
1792/**
1793 * skb_peek_next - peek skb following the given one from a queue
1794 * @skb: skb to start from
1795 * @list_: list to peek at
1796 *
1797 * Returns %NULL when the end of the list is met or a pointer to the
1798 * next element. The reference count is not incremented and the
1799 * reference is therefore volatile. Use with caution.
1800 */
1801static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1802 const struct sk_buff_head *list_)
1803{
1804 struct sk_buff *next = skb->next;
18d07000 1805
da5ef6e5
PE
1806 if (next == (struct sk_buff *)list_)
1807 next = NULL;
1808 return next;
1809}
1810
1da177e4 1811/**
1a5778aa 1812 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1813 * @list_: list to peek at
1814 *
1815 * Peek an &sk_buff. Unlike most other operations you _MUST_
1816 * be careful with this one. A peek leaves the buffer on the
1817 * list and someone else may run off with it. You must hold
1818 * the appropriate locks or have a private queue to do this.
1819 *
1820 * Returns %NULL for an empty list or a pointer to the tail element.
1821 * The reference count is not incremented and the reference is therefore
1822 * volatile. Use with caution.
1823 */
05bdd2f1 1824static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1825{
f8cc62ca 1826 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1827
1828 if (skb == (struct sk_buff *)list_)
1829 skb = NULL;
1830 return skb;
1831
1da177e4
LT
1832}
1833
1834/**
1835 * skb_queue_len - get queue length
1836 * @list_: list to measure
1837 *
1838 * Return the length of an &sk_buff queue.
1839 */
1840static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1841{
1842 return list_->qlen;
1843}
1844
86b18aaa
QC
1845/**
1846 * skb_queue_len_lockless - get queue length
1847 * @list_: list to measure
1848 *
1849 * Return the length of an &sk_buff queue.
1850 * This variant can be used in lockless contexts.
1851 */
1852static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1853{
1854 return READ_ONCE(list_->qlen);
1855}
1856
67fed459
DM
1857/**
1858 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1859 * @list: queue to initialize
1860 *
1861 * This initializes only the list and queue length aspects of
1862 * an sk_buff_head object. This allows to initialize the list
1863 * aspects of an sk_buff_head without reinitializing things like
1864 * the spinlock. It can also be used for on-stack sk_buff_head
1865 * objects where the spinlock is known to not be used.
1866 */
1867static inline void __skb_queue_head_init(struct sk_buff_head *list)
1868{
1869 list->prev = list->next = (struct sk_buff *)list;
1870 list->qlen = 0;
1871}
1872
76f10ad0
AV
1873/*
1874 * This function creates a split out lock class for each invocation;
1875 * this is needed for now since a whole lot of users of the skb-queue
1876 * infrastructure in drivers have different locking usage (in hardirq)
1877 * than the networking core (in softirq only). In the long run either the
1878 * network layer or drivers should need annotation to consolidate the
1879 * main types of usage into 3 classes.
1880 */
1da177e4
LT
1881static inline void skb_queue_head_init(struct sk_buff_head *list)
1882{
1883 spin_lock_init(&list->lock);
67fed459 1884 __skb_queue_head_init(list);
1da177e4
LT
1885}
1886
c2ecba71
PE
1887static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1888 struct lock_class_key *class)
1889{
1890 skb_queue_head_init(list);
1891 lockdep_set_class(&list->lock, class);
1892}
1893
1da177e4 1894/*
bf299275 1895 * Insert an sk_buff on a list.
1da177e4
LT
1896 *
1897 * The "__skb_xxxx()" functions are the non-atomic ones that
1898 * can only be called with interrupts disabled.
1899 */
bf299275
GR
1900static inline void __skb_insert(struct sk_buff *newsk,
1901 struct sk_buff *prev, struct sk_buff *next,
1902 struct sk_buff_head *list)
1903{
f8cc62ca
ED
1904 /* See skb_queue_empty_lockless() and skb_peek_tail()
1905 * for the opposite READ_ONCE()
1906 */
d7d16a89
ED
1907 WRITE_ONCE(newsk->next, next);
1908 WRITE_ONCE(newsk->prev, prev);
1909 WRITE_ONCE(next->prev, newsk);
1910 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1911 list->qlen++;
1912}
1da177e4 1913
67fed459
DM
1914static inline void __skb_queue_splice(const struct sk_buff_head *list,
1915 struct sk_buff *prev,
1916 struct sk_buff *next)
1917{
1918 struct sk_buff *first = list->next;
1919 struct sk_buff *last = list->prev;
1920
d7d16a89
ED
1921 WRITE_ONCE(first->prev, prev);
1922 WRITE_ONCE(prev->next, first);
67fed459 1923
d7d16a89
ED
1924 WRITE_ONCE(last->next, next);
1925 WRITE_ONCE(next->prev, last);
67fed459
DM
1926}
1927
1928/**
1929 * skb_queue_splice - join two skb lists, this is designed for stacks
1930 * @list: the new list to add
1931 * @head: the place to add it in the first list
1932 */
1933static inline void skb_queue_splice(const struct sk_buff_head *list,
1934 struct sk_buff_head *head)
1935{
1936 if (!skb_queue_empty(list)) {
1937 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1938 head->qlen += list->qlen;
67fed459
DM
1939 }
1940}
1941
1942/**
d9619496 1943 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1944 * @list: the new list to add
1945 * @head: the place to add it in the first list
1946 *
1947 * The list at @list is reinitialised
1948 */
1949static inline void skb_queue_splice_init(struct sk_buff_head *list,
1950 struct sk_buff_head *head)
1951{
1952 if (!skb_queue_empty(list)) {
1953 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1954 head->qlen += list->qlen;
67fed459
DM
1955 __skb_queue_head_init(list);
1956 }
1957}
1958
1959/**
1960 * skb_queue_splice_tail - join two skb lists, each list being a queue
1961 * @list: the new list to add
1962 * @head: the place to add it in the first list
1963 */
1964static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1965 struct sk_buff_head *head)
1966{
1967 if (!skb_queue_empty(list)) {
1968 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1969 head->qlen += list->qlen;
67fed459
DM
1970 }
1971}
1972
1973/**
d9619496 1974 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1975 * @list: the new list to add
1976 * @head: the place to add it in the first list
1977 *
1978 * Each of the lists is a queue.
1979 * The list at @list is reinitialised
1980 */
1981static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1982 struct sk_buff_head *head)
1983{
1984 if (!skb_queue_empty(list)) {
1985 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1986 head->qlen += list->qlen;
67fed459
DM
1987 __skb_queue_head_init(list);
1988 }
1989}
1990
1da177e4 1991/**
300ce174 1992 * __skb_queue_after - queue a buffer at the list head
1da177e4 1993 * @list: list to use
300ce174 1994 * @prev: place after this buffer
1da177e4
LT
1995 * @newsk: buffer to queue
1996 *
300ce174 1997 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1998 * and you must therefore hold required locks before calling it.
1999 *
2000 * A buffer cannot be placed on two lists at the same time.
2001 */
300ce174
SH
2002static inline void __skb_queue_after(struct sk_buff_head *list,
2003 struct sk_buff *prev,
2004 struct sk_buff *newsk)
1da177e4 2005{
bf299275 2006 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
2007}
2008
7965bd4d
JP
2009void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2010 struct sk_buff_head *list);
7de6c033 2011
f5572855
GR
2012static inline void __skb_queue_before(struct sk_buff_head *list,
2013 struct sk_buff *next,
2014 struct sk_buff *newsk)
2015{
2016 __skb_insert(newsk, next->prev, next, list);
2017}
2018
300ce174
SH
2019/**
2020 * __skb_queue_head - queue a buffer at the list head
2021 * @list: list to use
2022 * @newsk: buffer to queue
2023 *
2024 * Queue a buffer at the start of a list. This function takes no locks
2025 * and you must therefore hold required locks before calling it.
2026 *
2027 * A buffer cannot be placed on two lists at the same time.
2028 */
300ce174
SH
2029static inline void __skb_queue_head(struct sk_buff_head *list,
2030 struct sk_buff *newsk)
2031{
2032 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2033}
4ea7b0cf 2034void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2035
1da177e4
LT
2036/**
2037 * __skb_queue_tail - queue a buffer at the list tail
2038 * @list: list to use
2039 * @newsk: buffer to queue
2040 *
2041 * Queue a buffer at the end of a list. This function takes no locks
2042 * and you must therefore hold required locks before calling it.
2043 *
2044 * A buffer cannot be placed on two lists at the same time.
2045 */
1da177e4
LT
2046static inline void __skb_queue_tail(struct sk_buff_head *list,
2047 struct sk_buff *newsk)
2048{
f5572855 2049 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2050}
4ea7b0cf 2051void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2052
1da177e4
LT
2053/*
2054 * remove sk_buff from list. _Must_ be called atomically, and with
2055 * the list known..
2056 */
7965bd4d 2057void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2058static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2059{
2060 struct sk_buff *next, *prev;
2061
86b18aaa 2062 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2063 next = skb->next;
2064 prev = skb->prev;
2065 skb->next = skb->prev = NULL;
d7d16a89
ED
2066 WRITE_ONCE(next->prev, prev);
2067 WRITE_ONCE(prev->next, next);
1da177e4
LT
2068}
2069
f525c06d
GR
2070/**
2071 * __skb_dequeue - remove from the head of the queue
2072 * @list: list to dequeue from
2073 *
2074 * Remove the head of the list. This function does not take any locks
2075 * so must be used with appropriate locks held only. The head item is
2076 * returned or %NULL if the list is empty.
2077 */
f525c06d
GR
2078static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2079{
2080 struct sk_buff *skb = skb_peek(list);
2081 if (skb)
2082 __skb_unlink(skb, list);
2083 return skb;
2084}
4ea7b0cf 2085struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2086
2087/**
2088 * __skb_dequeue_tail - remove from the tail of the queue
2089 * @list: list to dequeue from
2090 *
2091 * Remove the tail of the list. This function does not take any locks
2092 * so must be used with appropriate locks held only. The tail item is
2093 * returned or %NULL if the list is empty.
2094 */
1da177e4
LT
2095static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2096{
2097 struct sk_buff *skb = skb_peek_tail(list);
2098 if (skb)
2099 __skb_unlink(skb, list);
2100 return skb;
2101}
4ea7b0cf 2102struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2103
2104
bdcc0924 2105static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2106{
2107 return skb->data_len;
2108}
2109
2110static inline unsigned int skb_headlen(const struct sk_buff *skb)
2111{
2112 return skb->len - skb->data_len;
2113}
2114
3ece7826 2115static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2116{
c72d8cda 2117 unsigned int i, len = 0;
1da177e4 2118
c72d8cda 2119 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2120 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2121 return len;
2122}
2123
2124static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2125{
2126 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2127}
2128
131ea667
IC
2129/**
2130 * __skb_fill_page_desc - initialise a paged fragment in an skb
2131 * @skb: buffer containing fragment to be initialised
2132 * @i: paged fragment index to initialise
2133 * @page: the page to use for this fragment
2134 * @off: the offset to the data with @page
2135 * @size: the length of the data
2136 *
2137 * Initialises the @i'th fragment of @skb to point to &size bytes at
2138 * offset @off within @page.
2139 *
2140 * Does not take any additional reference on the fragment.
2141 */
2142static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2143 struct page *page, int off, int size)
1da177e4
LT
2144{
2145 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2146
c48a11c7 2147 /*
2f064f34
MH
2148 * Propagate page pfmemalloc to the skb if we can. The problem is
2149 * that not all callers have unique ownership of the page but rely
2150 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2151 */
1dfa5bd3 2152 frag->bv_page = page;
65c84f14 2153 frag->bv_offset = off;
9e903e08 2154 skb_frag_size_set(frag, size);
cca7af38
PE
2155
2156 page = compound_head(page);
2f064f34 2157 if (page_is_pfmemalloc(page))
cca7af38 2158 skb->pfmemalloc = true;
131ea667
IC
2159}
2160
2161/**
2162 * skb_fill_page_desc - initialise a paged fragment in an skb
2163 * @skb: buffer containing fragment to be initialised
2164 * @i: paged fragment index to initialise
2165 * @page: the page to use for this fragment
2166 * @off: the offset to the data with @page
2167 * @size: the length of the data
2168 *
2169 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2170 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2171 * addition updates @skb such that @i is the last fragment.
2172 *
2173 * Does not take any additional reference on the fragment.
2174 */
2175static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2176 struct page *page, int off, int size)
2177{
2178 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2179 skb_shinfo(skb)->nr_frags = i + 1;
2180}
2181
7965bd4d
JP
2182void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2183 int size, unsigned int truesize);
654bed16 2184
f8e617e1
JW
2185void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2186 unsigned int truesize);
2187
1da177e4
LT
2188#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2189
27a884dc
ACM
2190#ifdef NET_SKBUFF_DATA_USES_OFFSET
2191static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2192{
2193 return skb->head + skb->tail;
2194}
2195
2196static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2197{
2198 skb->tail = skb->data - skb->head;
2199}
2200
2201static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2202{
2203 skb_reset_tail_pointer(skb);
2204 skb->tail += offset;
2205}
7cc46190 2206
27a884dc
ACM
2207#else /* NET_SKBUFF_DATA_USES_OFFSET */
2208static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2209{
2210 return skb->tail;
2211}
2212
2213static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2214{
2215 skb->tail = skb->data;
2216}
2217
2218static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2219{
2220 skb->tail = skb->data + offset;
2221}
4305b541 2222
27a884dc
ACM
2223#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2224
1da177e4
LT
2225/*
2226 * Add data to an sk_buff
2227 */
4df864c1
JB
2228void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2229void *skb_put(struct sk_buff *skb, unsigned int len);
2230static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2231{
4df864c1 2232 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2233 SKB_LINEAR_ASSERT(skb);
2234 skb->tail += len;
2235 skb->len += len;
2236 return tmp;
2237}
2238
de77b966 2239static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2240{
2241 void *tmp = __skb_put(skb, len);
2242
2243 memset(tmp, 0, len);
2244 return tmp;
2245}
2246
2247static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2248 unsigned int len)
2249{
2250 void *tmp = __skb_put(skb, len);
2251
2252 memcpy(tmp, data, len);
2253 return tmp;
2254}
2255
2256static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2257{
2258 *(u8 *)__skb_put(skb, 1) = val;
2259}
2260
83ad357d 2261static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2262{
83ad357d 2263 void *tmp = skb_put(skb, len);
e45a79da
JB
2264
2265 memset(tmp, 0, len);
2266
2267 return tmp;
2268}
2269
59ae1d12
JB
2270static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2271 unsigned int len)
2272{
2273 void *tmp = skb_put(skb, len);
2274
2275 memcpy(tmp, data, len);
2276
2277 return tmp;
2278}
2279
634fef61
JB
2280static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2281{
2282 *(u8 *)skb_put(skb, 1) = val;
2283}
2284
d58ff351
JB
2285void *skb_push(struct sk_buff *skb, unsigned int len);
2286static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2287{
2288 skb->data -= len;
2289 skb->len += len;
2290 return skb->data;
2291}
2292
af72868b
JB
2293void *skb_pull(struct sk_buff *skb, unsigned int len);
2294static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2295{
2296 skb->len -= len;
2297 BUG_ON(skb->len < skb->data_len);
2298 return skb->data += len;
2299}
2300
af72868b 2301static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2302{
2303 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2304}
2305
af72868b 2306void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2307
af72868b 2308static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2309{
2310 if (len > skb_headlen(skb) &&
987c402a 2311 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2312 return NULL;
2313 skb->len -= len;
2314 return skb->data += len;
2315}
2316
af72868b 2317static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2318{
2319 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2320}
2321
b9df4fd7 2322static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2323{
2324 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2325 return true;
1da177e4 2326 if (unlikely(len > skb->len))
b9df4fd7 2327 return false;
987c402a 2328 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2329}
2330
c8c8b127
ED
2331void skb_condense(struct sk_buff *skb);
2332
1da177e4
LT
2333/**
2334 * skb_headroom - bytes at buffer head
2335 * @skb: buffer to check
2336 *
2337 * Return the number of bytes of free space at the head of an &sk_buff.
2338 */
c2636b4d 2339static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2340{
2341 return skb->data - skb->head;
2342}
2343
2344/**
2345 * skb_tailroom - bytes at buffer end
2346 * @skb: buffer to check
2347 *
2348 * Return the number of bytes of free space at the tail of an sk_buff
2349 */
2350static inline int skb_tailroom(const struct sk_buff *skb)
2351{
4305b541 2352 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2353}
2354
a21d4572
ED
2355/**
2356 * skb_availroom - bytes at buffer end
2357 * @skb: buffer to check
2358 *
2359 * Return the number of bytes of free space at the tail of an sk_buff
2360 * allocated by sk_stream_alloc()
2361 */
2362static inline int skb_availroom(const struct sk_buff *skb)
2363{
16fad69c
ED
2364 if (skb_is_nonlinear(skb))
2365 return 0;
2366
2367 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2368}
2369
1da177e4
LT
2370/**
2371 * skb_reserve - adjust headroom
2372 * @skb: buffer to alter
2373 * @len: bytes to move
2374 *
2375 * Increase the headroom of an empty &sk_buff by reducing the tail
2376 * room. This is only allowed for an empty buffer.
2377 */
8243126c 2378static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2379{
2380 skb->data += len;
2381 skb->tail += len;
2382}
2383
1837b2e2
BP
2384/**
2385 * skb_tailroom_reserve - adjust reserved_tailroom
2386 * @skb: buffer to alter
2387 * @mtu: maximum amount of headlen permitted
2388 * @needed_tailroom: minimum amount of reserved_tailroom
2389 *
2390 * Set reserved_tailroom so that headlen can be as large as possible but
2391 * not larger than mtu and tailroom cannot be smaller than
2392 * needed_tailroom.
2393 * The required headroom should already have been reserved before using
2394 * this function.
2395 */
2396static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2397 unsigned int needed_tailroom)
2398{
2399 SKB_LINEAR_ASSERT(skb);
2400 if (mtu < skb_tailroom(skb) - needed_tailroom)
2401 /* use at most mtu */
2402 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2403 else
2404 /* use up to all available space */
2405 skb->reserved_tailroom = needed_tailroom;
2406}
2407
8bce6d7d
TH
2408#define ENCAP_TYPE_ETHER 0
2409#define ENCAP_TYPE_IPPROTO 1
2410
2411static inline void skb_set_inner_protocol(struct sk_buff *skb,
2412 __be16 protocol)
2413{
2414 skb->inner_protocol = protocol;
2415 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2416}
2417
2418static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2419 __u8 ipproto)
2420{
2421 skb->inner_ipproto = ipproto;
2422 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2423}
2424
6a674e9c
JG
2425static inline void skb_reset_inner_headers(struct sk_buff *skb)
2426{
aefbd2b3 2427 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2428 skb->inner_network_header = skb->network_header;
2429 skb->inner_transport_header = skb->transport_header;
2430}
2431
0b5c9db1
JP
2432static inline void skb_reset_mac_len(struct sk_buff *skb)
2433{
2434 skb->mac_len = skb->network_header - skb->mac_header;
2435}
2436
6a674e9c
JG
2437static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2438 *skb)
2439{
2440 return skb->head + skb->inner_transport_header;
2441}
2442
55dc5a9f
TH
2443static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2444{
2445 return skb_inner_transport_header(skb) - skb->data;
2446}
2447
6a674e9c
JG
2448static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2449{
2450 skb->inner_transport_header = skb->data - skb->head;
2451}
2452
2453static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2454 const int offset)
2455{
2456 skb_reset_inner_transport_header(skb);
2457 skb->inner_transport_header += offset;
2458}
2459
2460static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2461{
2462 return skb->head + skb->inner_network_header;
2463}
2464
2465static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2466{
2467 skb->inner_network_header = skb->data - skb->head;
2468}
2469
2470static inline void skb_set_inner_network_header(struct sk_buff *skb,
2471 const int offset)
2472{
2473 skb_reset_inner_network_header(skb);
2474 skb->inner_network_header += offset;
2475}
2476
aefbd2b3
PS
2477static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2478{
2479 return skb->head + skb->inner_mac_header;
2480}
2481
2482static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2483{
2484 skb->inner_mac_header = skb->data - skb->head;
2485}
2486
2487static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2488 const int offset)
2489{
2490 skb_reset_inner_mac_header(skb);
2491 skb->inner_mac_header += offset;
2492}
fda55eca
ED
2493static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2494{
35d04610 2495 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2496}
2497
9c70220b
ACM
2498static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2499{
2e07fa9c 2500 return skb->head + skb->transport_header;
9c70220b
ACM
2501}
2502
badff6d0
ACM
2503static inline void skb_reset_transport_header(struct sk_buff *skb)
2504{
2e07fa9c 2505 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2506}
2507
967b05f6
ACM
2508static inline void skb_set_transport_header(struct sk_buff *skb,
2509 const int offset)
2510{
2e07fa9c
ACM
2511 skb_reset_transport_header(skb);
2512 skb->transport_header += offset;
ea2ae17d
ACM
2513}
2514
d56f90a7
ACM
2515static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2516{
2e07fa9c 2517 return skb->head + skb->network_header;
d56f90a7
ACM
2518}
2519
c1d2bbe1
ACM
2520static inline void skb_reset_network_header(struct sk_buff *skb)
2521{
2e07fa9c 2522 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2523}
2524
c14d2450
ACM
2525static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2526{
2e07fa9c
ACM
2527 skb_reset_network_header(skb);
2528 skb->network_header += offset;
c14d2450
ACM
2529}
2530
2e07fa9c 2531static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2532{
2e07fa9c 2533 return skb->head + skb->mac_header;
bbe735e4
ACM
2534}
2535
ea6da4fd
AV
2536static inline int skb_mac_offset(const struct sk_buff *skb)
2537{
2538 return skb_mac_header(skb) - skb->data;
2539}
2540
0daf4349
DB
2541static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2542{
2543 return skb->network_header - skb->mac_header;
2544}
2545
2e07fa9c 2546static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2547{
35d04610 2548 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2549}
2550
2551static inline void skb_reset_mac_header(struct sk_buff *skb)
2552{
2553 skb->mac_header = skb->data - skb->head;
2554}
2555
2556static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2557{
2558 skb_reset_mac_header(skb);
2559 skb->mac_header += offset;
2560}
2561
0e3da5bb
TT
2562static inline void skb_pop_mac_header(struct sk_buff *skb)
2563{
2564 skb->mac_header = skb->network_header;
2565}
2566
d2aa125d 2567static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2568{
72a338bc 2569 struct flow_keys_basic keys;
fbbdb8f0
YX
2570
2571 if (skb_transport_header_was_set(skb))
2572 return;
72a338bc 2573
3cbf4ffb
SF
2574 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2575 NULL, 0, 0, 0, 0))
42aecaa9 2576 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2577}
2578
03606895
ED
2579static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2580{
2581 if (skb_mac_header_was_set(skb)) {
2582 const unsigned char *old_mac = skb_mac_header(skb);
2583
2584 skb_set_mac_header(skb, -skb->mac_len);
2585 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2586 }
2587}
2588
04fb451e
MM
2589static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2590{
2591 return skb->csum_start - skb_headroom(skb);
2592}
2593
08b64fcc
AD
2594static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2595{
2596 return skb->head + skb->csum_start;
2597}
2598
2e07fa9c
ACM
2599static inline int skb_transport_offset(const struct sk_buff *skb)
2600{
2601 return skb_transport_header(skb) - skb->data;
2602}
2603
2604static inline u32 skb_network_header_len(const struct sk_buff *skb)
2605{
2606 return skb->transport_header - skb->network_header;
2607}
2608
6a674e9c
JG
2609static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2610{
2611 return skb->inner_transport_header - skb->inner_network_header;
2612}
2613
2e07fa9c
ACM
2614static inline int skb_network_offset(const struct sk_buff *skb)
2615{
2616 return skb_network_header(skb) - skb->data;
2617}
48d49d0c 2618
6a674e9c
JG
2619static inline int skb_inner_network_offset(const struct sk_buff *skb)
2620{
2621 return skb_inner_network_header(skb) - skb->data;
2622}
2623
f9599ce1
CG
2624static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2625{
2626 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2627}
2628
1da177e4
LT
2629/*
2630 * CPUs often take a performance hit when accessing unaligned memory
2631 * locations. The actual performance hit varies, it can be small if the
2632 * hardware handles it or large if we have to take an exception and fix it
2633 * in software.
2634 *
2635 * Since an ethernet header is 14 bytes network drivers often end up with
2636 * the IP header at an unaligned offset. The IP header can be aligned by
2637 * shifting the start of the packet by 2 bytes. Drivers should do this
2638 * with:
2639 *
8660c124 2640 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2641 *
2642 * The downside to this alignment of the IP header is that the DMA is now
2643 * unaligned. On some architectures the cost of an unaligned DMA is high
2644 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2645 *
1da177e4
LT
2646 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2647 * to be overridden.
2648 */
2649#ifndef NET_IP_ALIGN
2650#define NET_IP_ALIGN 2
2651#endif
2652
025be81e
AB
2653/*
2654 * The networking layer reserves some headroom in skb data (via
2655 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2656 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2657 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2658 *
2659 * Unfortunately this headroom changes the DMA alignment of the resulting
2660 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2661 * on some architectures. An architecture can override this value,
2662 * perhaps setting it to a cacheline in size (since that will maintain
2663 * cacheline alignment of the DMA). It must be a power of 2.
2664 *
d6301d3d 2665 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2666 * headroom, you should not reduce this.
5933dd2f
ED
2667 *
2668 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2669 * to reduce average number of cache lines per packet.
645f0897 2670 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2671 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2672 */
2673#ifndef NET_SKB_PAD
5933dd2f 2674#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2675#endif
2676
7965bd4d 2677int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2678
5293efe6 2679static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2680{
5e1abdc3 2681 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2682 return;
27a884dc
ACM
2683 skb->len = len;
2684 skb_set_tail_pointer(skb, len);
1da177e4
LT
2685}
2686
5293efe6
DB
2687static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2688{
2689 __skb_set_length(skb, len);
2690}
2691
7965bd4d 2692void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2693
2694static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2695{
3cc0e873
HX
2696 if (skb->data_len)
2697 return ___pskb_trim(skb, len);
2698 __skb_trim(skb, len);
2699 return 0;
1da177e4
LT
2700}
2701
2702static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2703{
2704 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2705}
2706
e9fa4f7b
HX
2707/**
2708 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2709 * @skb: buffer to alter
2710 * @len: new length
2711 *
2712 * This is identical to pskb_trim except that the caller knows that
2713 * the skb is not cloned so we should never get an error due to out-
2714 * of-memory.
2715 */
2716static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2717{
2718 int err = pskb_trim(skb, len);
2719 BUG_ON(err);
2720}
2721
5293efe6
DB
2722static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2723{
2724 unsigned int diff = len - skb->len;
2725
2726 if (skb_tailroom(skb) < diff) {
2727 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2728 GFP_ATOMIC);
2729 if (ret)
2730 return ret;
2731 }
2732 __skb_set_length(skb, len);
2733 return 0;
2734}
2735
1da177e4
LT
2736/**
2737 * skb_orphan - orphan a buffer
2738 * @skb: buffer to orphan
2739 *
2740 * If a buffer currently has an owner then we call the owner's
2741 * destructor function and make the @skb unowned. The buffer continues
2742 * to exist but is no longer charged to its former owner.
2743 */
2744static inline void skb_orphan(struct sk_buff *skb)
2745{
c34a7612 2746 if (skb->destructor) {
1da177e4 2747 skb->destructor(skb);
c34a7612
ED
2748 skb->destructor = NULL;
2749 skb->sk = NULL;
376c7311
ED
2750 } else {
2751 BUG_ON(skb->sk);
c34a7612 2752 }
1da177e4
LT
2753}
2754
a353e0ce
MT
2755/**
2756 * skb_orphan_frags - orphan the frags contained in a buffer
2757 * @skb: buffer to orphan frags from
2758 * @gfp_mask: allocation mask for replacement pages
2759 *
2760 * For each frag in the SKB which needs a destructor (i.e. has an
2761 * owner) create a copy of that frag and release the original
2762 * page by calling the destructor.
2763 */
2764static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2765{
1f8b977a
WB
2766 if (likely(!skb_zcopy(skb)))
2767 return 0;
185ce5c3
WB
2768 if (!skb_zcopy_is_nouarg(skb) &&
2769 skb_uarg(skb)->callback == sock_zerocopy_callback)
1f8b977a
WB
2770 return 0;
2771 return skb_copy_ubufs(skb, gfp_mask);
2772}
2773
2774/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2775static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2776{
2777 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2778 return 0;
2779 return skb_copy_ubufs(skb, gfp_mask);
2780}
2781
1da177e4
LT
2782/**
2783 * __skb_queue_purge - empty a list
2784 * @list: list to empty
2785 *
2786 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2787 * the list and one reference dropped. This function does not take the
2788 * list lock and the caller must hold the relevant locks to use it.
2789 */
1da177e4
LT
2790static inline void __skb_queue_purge(struct sk_buff_head *list)
2791{
2792 struct sk_buff *skb;
2793 while ((skb = __skb_dequeue(list)) != NULL)
2794 kfree_skb(skb);
2795}
4ea7b0cf 2796void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2797
385114de 2798unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2799
7965bd4d 2800void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2801
7965bd4d
JP
2802struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2803 gfp_t gfp_mask);
8af27456
CH
2804
2805/**
2806 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2807 * @dev: network device to receive on
2808 * @length: length to allocate
2809 *
2810 * Allocate a new &sk_buff and assign it a usage count of one. The
2811 * buffer has unspecified headroom built in. Users should allocate
2812 * the headroom they think they need without accounting for the
2813 * built in space. The built in space is used for optimisations.
2814 *
2815 * %NULL is returned if there is no free memory. Although this function
2816 * allocates memory it can be called from an interrupt.
2817 */
2818static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2819 unsigned int length)
8af27456
CH
2820{
2821 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2822}
2823
6f532612
ED
2824/* legacy helper around __netdev_alloc_skb() */
2825static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2826 gfp_t gfp_mask)
2827{
2828 return __netdev_alloc_skb(NULL, length, gfp_mask);
2829}
2830
2831/* legacy helper around netdev_alloc_skb() */
2832static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2833{
2834 return netdev_alloc_skb(NULL, length);
2835}
2836
2837
4915a0de
ED
2838static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2839 unsigned int length, gfp_t gfp)
61321bbd 2840{
4915a0de 2841 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2842
2843 if (NET_IP_ALIGN && skb)
2844 skb_reserve(skb, NET_IP_ALIGN);
2845 return skb;
2846}
2847
4915a0de
ED
2848static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2849 unsigned int length)
2850{
2851 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2852}
2853
181edb2b
AD
2854static inline void skb_free_frag(void *addr)
2855{
8c2dd3e4 2856 page_frag_free(addr);
181edb2b
AD
2857}
2858
ffde7328 2859void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2860struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2861 unsigned int length, gfp_t gfp_mask);
2862static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2863 unsigned int length)
2864{
2865 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2866}
795bb1c0
JDB
2867void napi_consume_skb(struct sk_buff *skb, int budget);
2868
2869void __kfree_skb_flush(void);
15fad714 2870void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2871
71dfda58
AD
2872/**
2873 * __dev_alloc_pages - allocate page for network Rx
2874 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2875 * @order: size of the allocation
2876 *
2877 * Allocate a new page.
2878 *
2879 * %NULL is returned if there is no free memory.
2880*/
2881static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2882 unsigned int order)
2883{
2884 /* This piece of code contains several assumptions.
2885 * 1. This is for device Rx, therefor a cold page is preferred.
2886 * 2. The expectation is the user wants a compound page.
2887 * 3. If requesting a order 0 page it will not be compound
2888 * due to the check to see if order has a value in prep_new_page
2889 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2890 * code in gfp_to_alloc_flags that should be enforcing this.
2891 */
453f85d4 2892 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2893
2894 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2895}
2896
2897static inline struct page *dev_alloc_pages(unsigned int order)
2898{
95829b3a 2899 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2900}
2901
2902/**
2903 * __dev_alloc_page - allocate a page for network Rx
2904 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2905 *
2906 * Allocate a new page.
2907 *
2908 * %NULL is returned if there is no free memory.
2909 */
2910static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2911{
2912 return __dev_alloc_pages(gfp_mask, 0);
2913}
2914
2915static inline struct page *dev_alloc_page(void)
2916{
95829b3a 2917 return dev_alloc_pages(0);
71dfda58
AD
2918}
2919
0614002b
MG
2920/**
2921 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2922 * @page: The page that was allocated from skb_alloc_page
2923 * @skb: The skb that may need pfmemalloc set
2924 */
2925static inline void skb_propagate_pfmemalloc(struct page *page,
2926 struct sk_buff *skb)
2927{
2f064f34 2928 if (page_is_pfmemalloc(page))
0614002b
MG
2929 skb->pfmemalloc = true;
2930}
2931
7240b60c
JL
2932/**
2933 * skb_frag_off() - Returns the offset of a skb fragment
2934 * @frag: the paged fragment
2935 */
2936static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2937{
65c84f14 2938 return frag->bv_offset;
7240b60c
JL
2939}
2940
2941/**
2942 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2943 * @frag: skb fragment
2944 * @delta: value to add
2945 */
2946static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2947{
65c84f14 2948 frag->bv_offset += delta;
7240b60c
JL
2949}
2950
2951/**
2952 * skb_frag_off_set() - Sets the offset of a skb fragment
2953 * @frag: skb fragment
2954 * @offset: offset of fragment
2955 */
2956static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2957{
65c84f14 2958 frag->bv_offset = offset;
7240b60c
JL
2959}
2960
2961/**
2962 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2963 * @fragto: skb fragment where offset is set
2964 * @fragfrom: skb fragment offset is copied from
2965 */
2966static inline void skb_frag_off_copy(skb_frag_t *fragto,
2967 const skb_frag_t *fragfrom)
2968{
65c84f14 2969 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
2970}
2971
131ea667 2972/**
e227867f 2973 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2974 * @frag: the paged fragment
2975 *
2976 * Returns the &struct page associated with @frag.
2977 */
2978static inline struct page *skb_frag_page(const skb_frag_t *frag)
2979{
1dfa5bd3 2980 return frag->bv_page;
131ea667
IC
2981}
2982
2983/**
2984 * __skb_frag_ref - take an addition reference on a paged fragment.
2985 * @frag: the paged fragment
2986 *
2987 * Takes an additional reference on the paged fragment @frag.
2988 */
2989static inline void __skb_frag_ref(skb_frag_t *frag)
2990{
2991 get_page(skb_frag_page(frag));
2992}
2993
2994/**
2995 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2996 * @skb: the buffer
2997 * @f: the fragment offset.
2998 *
2999 * Takes an additional reference on the @f'th paged fragment of @skb.
3000 */
3001static inline void skb_frag_ref(struct sk_buff *skb, int f)
3002{
3003 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3004}
3005
3006/**
3007 * __skb_frag_unref - release a reference on a paged fragment.
3008 * @frag: the paged fragment
3009 *
3010 * Releases a reference on the paged fragment @frag.
3011 */
3012static inline void __skb_frag_unref(skb_frag_t *frag)
3013{
3014 put_page(skb_frag_page(frag));
3015}
3016
3017/**
3018 * skb_frag_unref - release a reference on a paged fragment of an skb.
3019 * @skb: the buffer
3020 * @f: the fragment offset
3021 *
3022 * Releases a reference on the @f'th paged fragment of @skb.
3023 */
3024static inline void skb_frag_unref(struct sk_buff *skb, int f)
3025{
3026 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3027}
3028
3029/**
3030 * skb_frag_address - gets the address of the data contained in a paged fragment
3031 * @frag: the paged fragment buffer
3032 *
3033 * Returns the address of the data within @frag. The page must already
3034 * be mapped.
3035 */
3036static inline void *skb_frag_address(const skb_frag_t *frag)
3037{
7240b60c 3038 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3039}
3040
3041/**
3042 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3043 * @frag: the paged fragment buffer
3044 *
3045 * Returns the address of the data within @frag. Checks that the page
3046 * is mapped and returns %NULL otherwise.
3047 */
3048static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3049{
3050 void *ptr = page_address(skb_frag_page(frag));
3051 if (unlikely(!ptr))
3052 return NULL;
3053
7240b60c
JL
3054 return ptr + skb_frag_off(frag);
3055}
3056
3057/**
3058 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3059 * @fragto: skb fragment where page is set
3060 * @fragfrom: skb fragment page is copied from
3061 */
3062static inline void skb_frag_page_copy(skb_frag_t *fragto,
3063 const skb_frag_t *fragfrom)
3064{
3065 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3066}
3067
3068/**
3069 * __skb_frag_set_page - sets the page contained in a paged fragment
3070 * @frag: the paged fragment
3071 * @page: the page to set
3072 *
3073 * Sets the fragment @frag to contain @page.
3074 */
3075static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3076{
1dfa5bd3 3077 frag->bv_page = page;
131ea667
IC
3078}
3079
3080/**
3081 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3082 * @skb: the buffer
3083 * @f: the fragment offset
3084 * @page: the page to set
3085 *
3086 * Sets the @f'th fragment of @skb to contain @page.
3087 */
3088static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3089 struct page *page)
3090{
3091 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3092}
3093
400dfd3a
ED
3094bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3095
131ea667
IC
3096/**
3097 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3098 * @dev: the device to map the fragment to
131ea667
IC
3099 * @frag: the paged fragment to map
3100 * @offset: the offset within the fragment (starting at the
3101 * fragment's own offset)
3102 * @size: the number of bytes to map
771b00a8 3103 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3104 *
3105 * Maps the page associated with @frag to @device.
3106 */
3107static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3108 const skb_frag_t *frag,
3109 size_t offset, size_t size,
3110 enum dma_data_direction dir)
3111{
3112 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3113 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3114}
3115
117632e6
ED
3116static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3117 gfp_t gfp_mask)
3118{
3119 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3120}
3121
bad93e9d
OP
3122
3123static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3124 gfp_t gfp_mask)
3125{
3126 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3127}
3128
3129
334a8132
PM
3130/**
3131 * skb_clone_writable - is the header of a clone writable
3132 * @skb: buffer to check
3133 * @len: length up to which to write
3134 *
3135 * Returns true if modifying the header part of the cloned buffer
3136 * does not requires the data to be copied.
3137 */
05bdd2f1 3138static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3139{
3140 return !skb_header_cloned(skb) &&
3141 skb_headroom(skb) + len <= skb->hdr_len;
3142}
3143
3697649f
DB
3144static inline int skb_try_make_writable(struct sk_buff *skb,
3145 unsigned int write_len)
3146{
3147 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3148 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3149}
3150
d9cc2048
HX
3151static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3152 int cloned)
3153{
3154 int delta = 0;
3155
d9cc2048
HX
3156 if (headroom > skb_headroom(skb))
3157 delta = headroom - skb_headroom(skb);
3158
3159 if (delta || cloned)
3160 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3161 GFP_ATOMIC);
3162 return 0;
3163}
3164
1da177e4
LT
3165/**
3166 * skb_cow - copy header of skb when it is required
3167 * @skb: buffer to cow
3168 * @headroom: needed headroom
3169 *
3170 * If the skb passed lacks sufficient headroom or its data part
3171 * is shared, data is reallocated. If reallocation fails, an error
3172 * is returned and original skb is not changed.
3173 *
3174 * The result is skb with writable area skb->head...skb->tail
3175 * and at least @headroom of space at head.
3176 */
3177static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3178{
d9cc2048
HX
3179 return __skb_cow(skb, headroom, skb_cloned(skb));
3180}
1da177e4 3181
d9cc2048
HX
3182/**
3183 * skb_cow_head - skb_cow but only making the head writable
3184 * @skb: buffer to cow
3185 * @headroom: needed headroom
3186 *
3187 * This function is identical to skb_cow except that we replace the
3188 * skb_cloned check by skb_header_cloned. It should be used when
3189 * you only need to push on some header and do not need to modify
3190 * the data.
3191 */
3192static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3193{
3194 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3195}
3196
3197/**
3198 * skb_padto - pad an skbuff up to a minimal size
3199 * @skb: buffer to pad
3200 * @len: minimal length
3201 *
3202 * Pads up a buffer to ensure the trailing bytes exist and are
3203 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3204 * is untouched. Otherwise it is extended. Returns zero on
3205 * success. The skb is freed on error.
1da177e4 3206 */
5b057c6b 3207static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3208{
3209 unsigned int size = skb->len;
3210 if (likely(size >= len))
5b057c6b 3211 return 0;
987c402a 3212 return skb_pad(skb, len - size);
1da177e4
LT
3213}
3214
9c0c1124 3215/**
4ea7b0cf 3216 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3217 * @skb: buffer to pad
3218 * @len: minimal length
cd0a137a 3219 * @free_on_error: free buffer on error
9c0c1124
AD
3220 *
3221 * Pads up a buffer to ensure the trailing bytes exist and are
3222 * blanked. If the buffer already contains sufficient data it
3223 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3224 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3225 */
4a009cb0
ED
3226static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3227 unsigned int len,
3228 bool free_on_error)
9c0c1124
AD
3229{
3230 unsigned int size = skb->len;
3231
3232 if (unlikely(size < len)) {
3233 len -= size;
cd0a137a 3234 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3235 return -ENOMEM;
3236 __skb_put(skb, len);
3237 }
3238 return 0;
3239}
3240
cd0a137a
FF
3241/**
3242 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3243 * @skb: buffer to pad
3244 * @len: minimal length
3245 *
3246 * Pads up a buffer to ensure the trailing bytes exist and are
3247 * blanked. If the buffer already contains sufficient data it
3248 * is untouched. Otherwise it is extended. Returns zero on
3249 * success. The skb is freed on error.
3250 */
4a009cb0 3251static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3252{
3253 return __skb_put_padto(skb, len, true);
3254}
3255
1da177e4 3256static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3257 struct iov_iter *from, int copy)
1da177e4
LT
3258{
3259 const int off = skb->len;
3260
3261 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3262 __wsum csum = 0;
15e6cb46
AV
3263 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3264 &csum, from)) {
1da177e4
LT
3265 skb->csum = csum_block_add(skb->csum, csum, off);
3266 return 0;
3267 }
15e6cb46 3268 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3269 return 0;
3270
3271 __skb_trim(skb, off);
3272 return -EFAULT;
3273}
3274
38ba0a65
ED
3275static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3276 const struct page *page, int off)
1da177e4 3277{
1f8b977a
WB
3278 if (skb_zcopy(skb))
3279 return false;
1da177e4 3280 if (i) {
d8e18a51 3281 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3282
ea2ab693 3283 return page == skb_frag_page(frag) &&
7240b60c 3284 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3285 }
38ba0a65 3286 return false;
1da177e4
LT
3287}
3288
364c6bad
HX
3289static inline int __skb_linearize(struct sk_buff *skb)
3290{
3291 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3292}
3293
1da177e4
LT
3294/**
3295 * skb_linearize - convert paged skb to linear one
3296 * @skb: buffer to linarize
1da177e4
LT
3297 *
3298 * If there is no free memory -ENOMEM is returned, otherwise zero
3299 * is returned and the old skb data released.
3300 */
364c6bad
HX
3301static inline int skb_linearize(struct sk_buff *skb)
3302{
3303 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3304}
3305
cef401de
ED
3306/**
3307 * skb_has_shared_frag - can any frag be overwritten
3308 * @skb: buffer to test
3309 *
3310 * Return true if the skb has at least one frag that might be modified
3311 * by an external entity (as in vmsplice()/sendfile())
3312 */
3313static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3314{
c9af6db4
PS
3315 return skb_is_nonlinear(skb) &&
3316 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
3317}
3318
364c6bad
HX
3319/**
3320 * skb_linearize_cow - make sure skb is linear and writable
3321 * @skb: buffer to process
3322 *
3323 * If there is no free memory -ENOMEM is returned, otherwise zero
3324 * is returned and the old skb data released.
3325 */
3326static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3327{
364c6bad
HX
3328 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3329 __skb_linearize(skb) : 0;
1da177e4
LT
3330}
3331
479ffccc
DB
3332static __always_inline void
3333__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3334 unsigned int off)
3335{
3336 if (skb->ip_summed == CHECKSUM_COMPLETE)
3337 skb->csum = csum_block_sub(skb->csum,
3338 csum_partial(start, len, 0), off);
3339 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3340 skb_checksum_start_offset(skb) < 0)
3341 skb->ip_summed = CHECKSUM_NONE;
3342}
3343
1da177e4
LT
3344/**
3345 * skb_postpull_rcsum - update checksum for received skb after pull
3346 * @skb: buffer to update
3347 * @start: start of data before pull
3348 * @len: length of data pulled
3349 *
3350 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3351 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3352 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3353 */
1da177e4 3354static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3355 const void *start, unsigned int len)
1da177e4 3356{
479ffccc 3357 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3358}
3359
479ffccc
DB
3360static __always_inline void
3361__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3362 unsigned int off)
3363{
3364 if (skb->ip_summed == CHECKSUM_COMPLETE)
3365 skb->csum = csum_block_add(skb->csum,
3366 csum_partial(start, len, 0), off);
3367}
cbb042f9 3368
479ffccc
DB
3369/**
3370 * skb_postpush_rcsum - update checksum for received skb after push
3371 * @skb: buffer to update
3372 * @start: start of data after push
3373 * @len: length of data pushed
3374 *
3375 * After doing a push on a received packet, you need to call this to
3376 * update the CHECKSUM_COMPLETE checksum.
3377 */
f8ffad69
DB
3378static inline void skb_postpush_rcsum(struct sk_buff *skb,
3379 const void *start, unsigned int len)
3380{
479ffccc 3381 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3382}
3383
af72868b 3384void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3385
82a31b92
WC
3386/**
3387 * skb_push_rcsum - push skb and update receive checksum
3388 * @skb: buffer to update
3389 * @len: length of data pulled
3390 *
3391 * This function performs an skb_push on the packet and updates
3392 * the CHECKSUM_COMPLETE checksum. It should be used on
3393 * receive path processing instead of skb_push unless you know
3394 * that the checksum difference is zero (e.g., a valid IP header)
3395 * or you are setting ip_summed to CHECKSUM_NONE.
3396 */
d58ff351 3397static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3398{
3399 skb_push(skb, len);
3400 skb_postpush_rcsum(skb, skb->data, len);
3401 return skb->data;
3402}
3403
88078d98 3404int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3405/**
3406 * pskb_trim_rcsum - trim received skb and update checksum
3407 * @skb: buffer to trim
3408 * @len: new length
3409 *
3410 * This is exactly the same as pskb_trim except that it ensures the
3411 * checksum of received packets are still valid after the operation.
6c57f045 3412 * It can change skb pointers.
7ce5a27f
DM
3413 */
3414
3415static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3416{
3417 if (likely(len >= skb->len))
3418 return 0;
88078d98 3419 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3420}
3421
5293efe6
DB
3422static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3423{
3424 if (skb->ip_summed == CHECKSUM_COMPLETE)
3425 skb->ip_summed = CHECKSUM_NONE;
3426 __skb_trim(skb, len);
3427 return 0;
3428}
3429
3430static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3431{
3432 if (skb->ip_summed == CHECKSUM_COMPLETE)
3433 skb->ip_summed = CHECKSUM_NONE;
3434 return __skb_grow(skb, len);
3435}
3436
18a4c0ea
ED
3437#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3438#define skb_rb_first(root) rb_to_skb(rb_first(root))
3439#define skb_rb_last(root) rb_to_skb(rb_last(root))
3440#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3441#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3442
1da177e4
LT
3443#define skb_queue_walk(queue, skb) \
3444 for (skb = (queue)->next; \
a1e4891f 3445 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3446 skb = skb->next)
3447
46f8914e
JC
3448#define skb_queue_walk_safe(queue, skb, tmp) \
3449 for (skb = (queue)->next, tmp = skb->next; \
3450 skb != (struct sk_buff *)(queue); \
3451 skb = tmp, tmp = skb->next)
3452
1164f52a 3453#define skb_queue_walk_from(queue, skb) \
a1e4891f 3454 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3455 skb = skb->next)
3456
18a4c0ea
ED
3457#define skb_rbtree_walk(skb, root) \
3458 for (skb = skb_rb_first(root); skb != NULL; \
3459 skb = skb_rb_next(skb))
3460
3461#define skb_rbtree_walk_from(skb) \
3462 for (; skb != NULL; \
3463 skb = skb_rb_next(skb))
3464
3465#define skb_rbtree_walk_from_safe(skb, tmp) \
3466 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3467 skb = tmp)
3468
1164f52a
DM
3469#define skb_queue_walk_from_safe(queue, skb, tmp) \
3470 for (tmp = skb->next; \
3471 skb != (struct sk_buff *)(queue); \
3472 skb = tmp, tmp = skb->next)
3473
300ce174
SH
3474#define skb_queue_reverse_walk(queue, skb) \
3475 for (skb = (queue)->prev; \
a1e4891f 3476 skb != (struct sk_buff *)(queue); \
300ce174
SH
3477 skb = skb->prev)
3478
686a2955
DM
3479#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3480 for (skb = (queue)->prev, tmp = skb->prev; \
3481 skb != (struct sk_buff *)(queue); \
3482 skb = tmp, tmp = skb->prev)
3483
3484#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3485 for (tmp = skb->prev; \
3486 skb != (struct sk_buff *)(queue); \
3487 skb = tmp, tmp = skb->prev)
1da177e4 3488
21dc3301 3489static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3490{
3491 return skb_shinfo(skb)->frag_list != NULL;
3492}
3493
3494static inline void skb_frag_list_init(struct sk_buff *skb)
3495{
3496 skb_shinfo(skb)->frag_list = NULL;
3497}
3498
ee039871
DM
3499#define skb_walk_frags(skb, iter) \
3500 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3501
ea3793ee 3502
b50b0580
SD
3503int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3504 int *err, long *timeo_p,
ea3793ee 3505 const struct sk_buff *skb);
65101aec
PA
3506struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3507 struct sk_buff_head *queue,
3508 unsigned int flags,
fd69c399 3509 int *off, int *err,
65101aec 3510 struct sk_buff **last);
b50b0580
SD
3511struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3512 struct sk_buff_head *queue,
e427cad6 3513 unsigned int flags, int *off, int *err,
ea3793ee 3514 struct sk_buff **last);
b50b0580
SD
3515struct sk_buff *__skb_recv_datagram(struct sock *sk,
3516 struct sk_buff_head *sk_queue,
e427cad6 3517 unsigned int flags, int *off, int *err);
7965bd4d
JP
3518struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3519 int *err);
a11e1d43
LT
3520__poll_t datagram_poll(struct file *file, struct socket *sock,
3521 struct poll_table_struct *wait);
c0371da6
AV
3522int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3523 struct iov_iter *to, int size);
51f3d02b
DM
3524static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3525 struct msghdr *msg, int size)
3526{
e5a4b0bb 3527 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3528}
e5a4b0bb
AV
3529int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3530 struct msghdr *msg);
65d69e25
SG
3531int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3532 struct iov_iter *to, int len,
3533 struct ahash_request *hash);
3a654f97
AV
3534int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3535 struct iov_iter *from, int len);
3a654f97 3536int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3537void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3538void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3539static inline void skb_free_datagram_locked(struct sock *sk,
3540 struct sk_buff *skb)
3541{
3542 __skb_free_datagram_locked(sk, skb, 0);
3543}
7965bd4d 3544int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3545int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3546int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3547__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3548 int len, __wsum csum);
a60e3cc7 3549int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3550 struct pipe_inode_info *pipe, unsigned int len,
25869262 3551 unsigned int flags);
20bf50de
TH
3552int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3553 int len);
7965bd4d 3554void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3555unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3556int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3557 int len, int hlen);
7965bd4d
JP
3558void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3559int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3560void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3561bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3562bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3563struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3564struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3565 unsigned int offset);
0d5501c1 3566struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3567int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3568int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3569int skb_vlan_pop(struct sk_buff *skb);
3570int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
fa4e0f88 3571int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3572 int mac_len, bool ethernet);
040b5cfb
MV
3573int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3574 bool ethernet);
d27cf5c5 3575int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3576int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3577struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3578 gfp_t gfp);
20380731 3579
6ce8e9ce
AV
3580static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3581{
3073f070 3582 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3583}
3584
7eab8d9e
AV
3585static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3586{
e5a4b0bb 3587 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3588}
3589
2817a336
DB
3590struct skb_checksum_ops {
3591 __wsum (*update)(const void *mem, int len, __wsum wsum);
3592 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3593};
3594
9617813d
DC
3595extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3596
2817a336
DB
3597__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3598 __wsum csum, const struct skb_checksum_ops *ops);
3599__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3600 __wsum csum);
3601
1e98a0f0
ED
3602static inline void * __must_check
3603__skb_header_pointer(const struct sk_buff *skb, int offset,
3604 int len, void *data, int hlen, void *buffer)
1da177e4 3605{
55820ee2 3606 if (hlen - offset >= len)
690e36e7 3607 return data + offset;
1da177e4 3608
690e36e7
DM
3609 if (!skb ||
3610 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3611 return NULL;
3612
3613 return buffer;
3614}
3615
1e98a0f0
ED
3616static inline void * __must_check
3617skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3618{
3619 return __skb_header_pointer(skb, offset, len, skb->data,
3620 skb_headlen(skb), buffer);
3621}
3622
4262e5cc
DB
3623/**
3624 * skb_needs_linearize - check if we need to linearize a given skb
3625 * depending on the given device features.
3626 * @skb: socket buffer to check
3627 * @features: net device features
3628 *
3629 * Returns true if either:
3630 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3631 * 2. skb is fragmented and the device does not support SG.
3632 */
3633static inline bool skb_needs_linearize(struct sk_buff *skb,
3634 netdev_features_t features)
3635{
3636 return skb_is_nonlinear(skb) &&
3637 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3638 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3639}
3640
d626f62b
ACM
3641static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3642 void *to,
3643 const unsigned int len)
3644{
3645 memcpy(to, skb->data, len);
3646}
3647
3648static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3649 const int offset, void *to,
3650 const unsigned int len)
3651{
3652 memcpy(to, skb->data + offset, len);
3653}
3654
27d7ff46
ACM
3655static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3656 const void *from,
3657 const unsigned int len)
3658{
3659 memcpy(skb->data, from, len);
3660}
3661
3662static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3663 const int offset,
3664 const void *from,
3665 const unsigned int len)
3666{
3667 memcpy(skb->data + offset, from, len);
3668}
3669
7965bd4d 3670void skb_init(void);
1da177e4 3671
ac45f602
PO
3672static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3673{
3674 return skb->tstamp;
3675}
3676
a61bbcf2
PM
3677/**
3678 * skb_get_timestamp - get timestamp from a skb
3679 * @skb: skb to get stamp from
13c6ee2a 3680 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3681 *
3682 * Timestamps are stored in the skb as offsets to a base timestamp.
3683 * This function converts the offset back to a struct timeval and stores
3684 * it in stamp.
3685 */
ac45f602 3686static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3687 struct __kernel_old_timeval *stamp)
a61bbcf2 3688{
13c6ee2a 3689 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3690}
3691
887feae3
DD
3692static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3693 struct __kernel_sock_timeval *stamp)
3694{
3695 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3696
3697 stamp->tv_sec = ts.tv_sec;
3698 stamp->tv_usec = ts.tv_nsec / 1000;
3699}
3700
ac45f602 3701static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3702 struct __kernel_old_timespec *stamp)
ac45f602 3703{
df1b4ba9
AB
3704 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3705
3706 stamp->tv_sec = ts.tv_sec;
3707 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3708}
3709
887feae3
DD
3710static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3711 struct __kernel_timespec *stamp)
3712{
3713 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3714
3715 stamp->tv_sec = ts.tv_sec;
3716 stamp->tv_nsec = ts.tv_nsec;
3717}
3718
b7aa0bf7 3719static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3720{
b7aa0bf7 3721 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3722}
3723
164891aa
SH
3724static inline ktime_t net_timedelta(ktime_t t)
3725{
3726 return ktime_sub(ktime_get_real(), t);
3727}
3728
b9ce204f
IJ
3729static inline ktime_t net_invalid_timestamp(void)
3730{
8b0e1953 3731 return 0;
b9ce204f 3732}
a61bbcf2 3733
de8f3a83
DB
3734static inline u8 skb_metadata_len(const struct sk_buff *skb)
3735{
3736 return skb_shinfo(skb)->meta_len;
3737}
3738
3739static inline void *skb_metadata_end(const struct sk_buff *skb)
3740{
3741 return skb_mac_header(skb);
3742}
3743
3744static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3745 const struct sk_buff *skb_b,
3746 u8 meta_len)
3747{
3748 const void *a = skb_metadata_end(skb_a);
3749 const void *b = skb_metadata_end(skb_b);
3750 /* Using more efficient varaiant than plain call to memcmp(). */
3751#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3752 u64 diffs = 0;
3753
3754 switch (meta_len) {
3755#define __it(x, op) (x -= sizeof(u##op))
3756#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3757 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3758 fallthrough;
de8f3a83 3759 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3760 fallthrough;
de8f3a83 3761 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3762 fallthrough;
de8f3a83
DB
3763 case 8: diffs |= __it_diff(a, b, 64);
3764 break;
3765 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3766 fallthrough;
de8f3a83 3767 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3768 fallthrough;
de8f3a83 3769 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3770 fallthrough;
de8f3a83
DB
3771 case 4: diffs |= __it_diff(a, b, 32);
3772 break;
3773 }
3774 return diffs;
3775#else
3776 return memcmp(a - meta_len, b - meta_len, meta_len);
3777#endif
3778}
3779
3780static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3781 const struct sk_buff *skb_b)
3782{
3783 u8 len_a = skb_metadata_len(skb_a);
3784 u8 len_b = skb_metadata_len(skb_b);
3785
3786 if (!(len_a | len_b))
3787 return false;
3788
3789 return len_a != len_b ?
3790 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3791}
3792
3793static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3794{
3795 skb_shinfo(skb)->meta_len = meta_len;
3796}
3797
3798static inline void skb_metadata_clear(struct sk_buff *skb)
3799{
3800 skb_metadata_set(skb, 0);
3801}
3802
62bccb8c
AD
3803struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3804
c1f19b51
RC
3805#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3806
7965bd4d
JP
3807void skb_clone_tx_timestamp(struct sk_buff *skb);
3808bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3809
3810#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3811
3812static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3813{
3814}
3815
3816static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3817{
3818 return false;
3819}
3820
3821#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3822
3823/**
3824 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3825 *
da92b194
RC
3826 * PHY drivers may accept clones of transmitted packets for
3827 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3828 * must call this function to return the skb back to the stack with a
3829 * timestamp.
da92b194 3830 *
2ff17117 3831 * @skb: clone of the original outgoing packet
7a76a021 3832 * @hwtstamps: hardware time stamps
c1f19b51
RC
3833 *
3834 */
3835void skb_complete_tx_timestamp(struct sk_buff *skb,
3836 struct skb_shared_hwtstamps *hwtstamps);
3837
e7fd2885
WB
3838void __skb_tstamp_tx(struct sk_buff *orig_skb,
3839 struct skb_shared_hwtstamps *hwtstamps,
3840 struct sock *sk, int tstype);
3841
ac45f602
PO
3842/**
3843 * skb_tstamp_tx - queue clone of skb with send time stamps
3844 * @orig_skb: the original outgoing packet
3845 * @hwtstamps: hardware time stamps, may be NULL if not available
3846 *
3847 * If the skb has a socket associated, then this function clones the
3848 * skb (thus sharing the actual data and optional structures), stores
3849 * the optional hardware time stamping information (if non NULL) or
3850 * generates a software time stamp (otherwise), then queues the clone
3851 * to the error queue of the socket. Errors are silently ignored.
3852 */
7965bd4d
JP
3853void skb_tstamp_tx(struct sk_buff *orig_skb,
3854 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3855
4507a715
RC
3856/**
3857 * skb_tx_timestamp() - Driver hook for transmit timestamping
3858 *
3859 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3860 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3861 *
73409f3b
DM
3862 * Specifically, one should make absolutely sure that this function is
3863 * called before TX completion of this packet can trigger. Otherwise
3864 * the packet could potentially already be freed.
3865 *
4507a715
RC
3866 * @skb: A socket buffer.
3867 */
3868static inline void skb_tx_timestamp(struct sk_buff *skb)
3869{
c1f19b51 3870 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3871 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3872 skb_tstamp_tx(skb, NULL);
4507a715
RC
3873}
3874
6e3e939f
JB
3875/**
3876 * skb_complete_wifi_ack - deliver skb with wifi status
3877 *
3878 * @skb: the original outgoing packet
3879 * @acked: ack status
3880 *
3881 */
3882void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3883
7965bd4d
JP
3884__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3885__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3886
60476372
HX
3887static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3888{
6edec0e6
TH
3889 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3890 skb->csum_valid ||
3891 (skb->ip_summed == CHECKSUM_PARTIAL &&
3892 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3893}
3894
fb286bb2
HX
3895/**
3896 * skb_checksum_complete - Calculate checksum of an entire packet
3897 * @skb: packet to process
3898 *
3899 * This function calculates the checksum over the entire packet plus
3900 * the value of skb->csum. The latter can be used to supply the
3901 * checksum of a pseudo header as used by TCP/UDP. It returns the
3902 * checksum.
3903 *
3904 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3905 * this function can be used to verify that checksum on received
3906 * packets. In that case the function should return zero if the
3907 * checksum is correct. In particular, this function will return zero
3908 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3909 * hardware has already verified the correctness of the checksum.
3910 */
4381ca3c 3911static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3912{
60476372
HX
3913 return skb_csum_unnecessary(skb) ?
3914 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3915}
3916
77cffe23
TH
3917static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3918{
3919 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3920 if (skb->csum_level == 0)
3921 skb->ip_summed = CHECKSUM_NONE;
3922 else
3923 skb->csum_level--;
3924 }
3925}
3926
3927static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3928{
3929 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3930 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3931 skb->csum_level++;
3932 } else if (skb->ip_summed == CHECKSUM_NONE) {
3933 skb->ip_summed = CHECKSUM_UNNECESSARY;
3934 skb->csum_level = 0;
3935 }
3936}
3937
836e66c2
DB
3938static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3939{
3940 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3941 skb->ip_summed = CHECKSUM_NONE;
3942 skb->csum_level = 0;
3943 }
3944}
3945
76ba0aae
TH
3946/* Check if we need to perform checksum complete validation.
3947 *
3948 * Returns true if checksum complete is needed, false otherwise
3949 * (either checksum is unnecessary or zero checksum is allowed).
3950 */
3951static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3952 bool zero_okay,
3953 __sum16 check)
3954{
5d0c2b95
TH
3955 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3956 skb->csum_valid = 1;
77cffe23 3957 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3958 return false;
3959 }
3960
3961 return true;
3962}
3963
da279887 3964/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
3965 * in checksum_init.
3966 */
3967#define CHECKSUM_BREAK 76
3968
4e18b9ad
TH
3969/* Unset checksum-complete
3970 *
3971 * Unset checksum complete can be done when packet is being modified
3972 * (uncompressed for instance) and checksum-complete value is
3973 * invalidated.
3974 */
3975static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3976{
3977 if (skb->ip_summed == CHECKSUM_COMPLETE)
3978 skb->ip_summed = CHECKSUM_NONE;
3979}
3980
76ba0aae
TH
3981/* Validate (init) checksum based on checksum complete.
3982 *
3983 * Return values:
3984 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3985 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3986 * checksum is stored in skb->csum for use in __skb_checksum_complete
3987 * non-zero: value of invalid checksum
3988 *
3989 */
3990static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3991 bool complete,
3992 __wsum psum)
3993{
3994 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3995 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3996 skb->csum_valid = 1;
76ba0aae
TH
3997 return 0;
3998 }
3999 }
4000
4001 skb->csum = psum;
4002
5d0c2b95
TH
4003 if (complete || skb->len <= CHECKSUM_BREAK) {
4004 __sum16 csum;
4005
4006 csum = __skb_checksum_complete(skb);
4007 skb->csum_valid = !csum;
4008 return csum;
4009 }
76ba0aae
TH
4010
4011 return 0;
4012}
4013
4014static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4015{
4016 return 0;
4017}
4018
4019/* Perform checksum validate (init). Note that this is a macro since we only
4020 * want to calculate the pseudo header which is an input function if necessary.
4021 * First we try to validate without any computation (checksum unnecessary) and
4022 * then calculate based on checksum complete calling the function to compute
4023 * pseudo header.
4024 *
4025 * Return values:
4026 * 0: checksum is validated or try to in skb_checksum_complete
4027 * non-zero: value of invalid checksum
4028 */
4029#define __skb_checksum_validate(skb, proto, complete, \
4030 zero_okay, check, compute_pseudo) \
4031({ \
4032 __sum16 __ret = 0; \
5d0c2b95 4033 skb->csum_valid = 0; \
76ba0aae
TH
4034 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4035 __ret = __skb_checksum_validate_complete(skb, \
4036 complete, compute_pseudo(skb, proto)); \
4037 __ret; \
4038})
4039
4040#define skb_checksum_init(skb, proto, compute_pseudo) \
4041 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4042
4043#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4044 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4045
4046#define skb_checksum_validate(skb, proto, compute_pseudo) \
4047 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4048
4049#define skb_checksum_validate_zero_check(skb, proto, check, \
4050 compute_pseudo) \
096a4cfa 4051 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4052
4053#define skb_checksum_simple_validate(skb) \
4054 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4055
d96535a1
TH
4056static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4057{
219f1d79 4058 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4059}
4060
e4aa33ad 4061static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4062{
4063 skb->csum = ~pseudo;
4064 skb->ip_summed = CHECKSUM_COMPLETE;
4065}
4066
e4aa33ad 4067#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4068do { \
4069 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4070 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4071} while (0)
4072
15e2396d
TH
4073static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4074 u16 start, u16 offset)
4075{
4076 skb->ip_summed = CHECKSUM_PARTIAL;
4077 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4078 skb->csum_offset = offset - start;
4079}
4080
dcdc8994
TH
4081/* Update skbuf and packet to reflect the remote checksum offload operation.
4082 * When called, ptr indicates the starting point for skb->csum when
4083 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4084 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4085 */
4086static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4087 int start, int offset, bool nopartial)
dcdc8994
TH
4088{
4089 __wsum delta;
4090
15e2396d
TH
4091 if (!nopartial) {
4092 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4093 return;
4094 }
4095
dcdc8994
TH
4096 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4097 __skb_checksum_complete(skb);
4098 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4099 }
4100
4101 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4102
4103 /* Adjust skb->csum since we changed the packet */
4104 skb->csum = csum_add(skb->csum, delta);
4105}
4106
cb9c6836
FW
4107static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4108{
4109#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4110 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4111#else
4112 return NULL;
4113#endif
4114}
4115
261db6c2 4116static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4117{
261db6c2
JS
4118#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4119 return skb->_nfct;
4120#else
4121 return 0UL;
4122#endif
1da177e4 4123}
261db6c2
JS
4124
4125static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4126{
261db6c2
JS
4127#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4128 skb->_nfct = nfct;
2fc72c7b 4129#endif
261db6c2 4130}
df5042f4
FW
4131
4132#ifdef CONFIG_SKB_EXTENSIONS
4133enum skb_ext_id {
4134#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4135 SKB_EXT_BRIDGE_NF,
4165079b
FW
4136#endif
4137#ifdef CONFIG_XFRM
4138 SKB_EXT_SEC_PATH,
95a7233c
PB
4139#endif
4140#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4141 TC_SKB_EXT,
3ee17bc7
MM
4142#endif
4143#if IS_ENABLED(CONFIG_MPTCP)
4144 SKB_EXT_MPTCP,
df5042f4
FW
4145#endif
4146 SKB_EXT_NUM, /* must be last */
4147};
4148
4149/**
4150 * struct skb_ext - sk_buff extensions
4151 * @refcnt: 1 on allocation, deallocated on 0
4152 * @offset: offset to add to @data to obtain extension address
4153 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4154 * @data: start of extension data, variable sized
4155 *
4156 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4157 * to use 'u8' types while allowing up to 2kb worth of extension data.
4158 */
4159struct skb_ext {
4160 refcount_t refcnt;
4161 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4162 u8 chunks; /* same */
5c91aa1d 4163 char data[] __aligned(8);
df5042f4
FW
4164};
4165
4930f483 4166struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4167void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4168 struct skb_ext *ext);
df5042f4
FW
4169void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4170void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4171void __skb_ext_put(struct skb_ext *ext);
4172
4173static inline void skb_ext_put(struct sk_buff *skb)
4174{
4175 if (skb->active_extensions)
4176 __skb_ext_put(skb->extensions);
4177}
4178
df5042f4
FW
4179static inline void __skb_ext_copy(struct sk_buff *dst,
4180 const struct sk_buff *src)
4181{
4182 dst->active_extensions = src->active_extensions;
4183
4184 if (src->active_extensions) {
4185 struct skb_ext *ext = src->extensions;
4186
4187 refcount_inc(&ext->refcnt);
4188 dst->extensions = ext;
4189 }
4190}
4191
4192static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4193{
4194 skb_ext_put(dst);
4195 __skb_ext_copy(dst, src);
4196}
4197
4198static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4199{
4200 return !!ext->offset[i];
4201}
4202
4203static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4204{
4205 return skb->active_extensions & (1 << id);
4206}
4207
4208static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4209{
4210 if (skb_ext_exist(skb, id))
4211 __skb_ext_del(skb, id);
4212}
4213
4214static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4215{
4216 if (skb_ext_exist(skb, id)) {
4217 struct skb_ext *ext = skb->extensions;
4218
4219 return (void *)ext + (ext->offset[id] << 3);
4220 }
4221
4222 return NULL;
4223}
174e2381
FW
4224
4225static inline void skb_ext_reset(struct sk_buff *skb)
4226{
4227 if (unlikely(skb->active_extensions)) {
4228 __skb_ext_put(skb->extensions);
4229 skb->active_extensions = 0;
4230 }
4231}
677bf08c
FW
4232
4233static inline bool skb_has_extensions(struct sk_buff *skb)
4234{
4235 return unlikely(skb->active_extensions);
4236}
df5042f4
FW
4237#else
4238static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4239static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4240static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4241static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4242static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4243static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4244#endif /* CONFIG_SKB_EXTENSIONS */
4245
895b5c9f 4246static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4247{
5f79e0f9 4248#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4249 nf_conntrack_put(skb_nfct(skb));
4250 skb->_nfct = 0;
2fc72c7b 4251#endif
a193a4ab
PM
4252}
4253
124dff01
PM
4254static inline void nf_reset_trace(struct sk_buff *skb)
4255{
478b360a 4256#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4257 skb->nf_trace = 0;
4258#endif
a193a4ab
PM
4259}
4260
2b5ec1a5
YY
4261static inline void ipvs_reset(struct sk_buff *skb)
4262{
4263#if IS_ENABLED(CONFIG_IP_VS)
4264 skb->ipvs_property = 0;
4265#endif
4266}
4267
de8bda1d 4268/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4269static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4270 bool copy)
edda553c 4271{
5f79e0f9 4272#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4273 dst->_nfct = src->_nfct;
4274 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4275#endif
478b360a 4276#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4277 if (copy)
4278 dst->nf_trace = src->nf_trace;
478b360a 4279#endif
edda553c
YK
4280}
4281
e7ac05f3
YK
4282static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4283{
e7ac05f3 4284#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4285 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4286#endif
b1937227 4287 __nf_copy(dst, src, true);
e7ac05f3
YK
4288}
4289
984bc16c
JM
4290#ifdef CONFIG_NETWORK_SECMARK
4291static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4292{
4293 to->secmark = from->secmark;
4294}
4295
4296static inline void skb_init_secmark(struct sk_buff *skb)
4297{
4298 skb->secmark = 0;
4299}
4300#else
4301static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4302{ }
4303
4304static inline void skb_init_secmark(struct sk_buff *skb)
4305{ }
4306#endif
4307
7af8f4ca
FW
4308static inline int secpath_exists(const struct sk_buff *skb)
4309{
4310#ifdef CONFIG_XFRM
4165079b 4311 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4312#else
4313 return 0;
4314#endif
4315}
4316
574f7194
EB
4317static inline bool skb_irq_freeable(const struct sk_buff *skb)
4318{
4319 return !skb->destructor &&
7af8f4ca 4320 !secpath_exists(skb) &&
cb9c6836 4321 !skb_nfct(skb) &&
574f7194
EB
4322 !skb->_skb_refdst &&
4323 !skb_has_frag_list(skb);
4324}
4325
f25f4e44
PWJ
4326static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4327{
f25f4e44 4328 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4329}
4330
9247744e 4331static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4332{
4e3ab47a 4333 return skb->queue_mapping;
4e3ab47a
PE
4334}
4335
f25f4e44
PWJ
4336static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4337{
f25f4e44 4338 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4339}
4340
d5a9e24a
DM
4341static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4342{
4343 skb->queue_mapping = rx_queue + 1;
4344}
4345
9247744e 4346static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4347{
4348 return skb->queue_mapping - 1;
4349}
4350
9247744e 4351static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4352{
a02cec21 4353 return skb->queue_mapping != 0;
d5a9e24a
DM
4354}
4355
4ff06203
JA
4356static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4357{
4358 skb->dst_pending_confirm = val;
4359}
4360
4361static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4362{
4363 return skb->dst_pending_confirm != 0;
4364}
4365
2294be0f 4366static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4367{
0b3d8e08 4368#ifdef CONFIG_XFRM
4165079b 4369 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4370#else
def8b4fa 4371 return NULL;
def8b4fa 4372#endif
0b3d8e08 4373}
def8b4fa 4374
68c33163
PS
4375/* Keeps track of mac header offset relative to skb->head.
4376 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4377 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4378 * tunnel skb it points to outer mac header.
4379 * Keeps track of level of encapsulation of network headers.
4380 */
68c33163 4381struct skb_gso_cb {
802ab55a
AD
4382 union {
4383 int mac_offset;
4384 int data_offset;
4385 };
3347c960 4386 int encap_level;
76443456 4387 __wsum csum;
7e2b10c1 4388 __u16 csum_start;
68c33163 4389};
a08e7fd9
CZ
4390#define SKB_GSO_CB_OFFSET 32
4391#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4392
4393static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4394{
4395 return (skb_mac_header(inner_skb) - inner_skb->head) -
4396 SKB_GSO_CB(inner_skb)->mac_offset;
4397}
4398
1e2bd517
PS
4399static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4400{
4401 int new_headroom, headroom;
4402 int ret;
4403
4404 headroom = skb_headroom(skb);
4405 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4406 if (ret)
4407 return ret;
4408
4409 new_headroom = skb_headroom(skb);
4410 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4411 return 0;
4412}
4413
08b64fcc
AD
4414static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4415{
4416 /* Do not update partial checksums if remote checksum is enabled. */
4417 if (skb->remcsum_offload)
4418 return;
4419
4420 SKB_GSO_CB(skb)->csum = res;
4421 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4422}
4423
7e2b10c1
TH
4424/* Compute the checksum for a gso segment. First compute the checksum value
4425 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4426 * then add in skb->csum (checksum from csum_start to end of packet).
4427 * skb->csum and csum_start are then updated to reflect the checksum of the
4428 * resultant packet starting from the transport header-- the resultant checksum
4429 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4430 * header.
4431 */
4432static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4433{
76443456
AD
4434 unsigned char *csum_start = skb_transport_header(skb);
4435 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4436 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4437
76443456
AD
4438 SKB_GSO_CB(skb)->csum = res;
4439 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4440
76443456 4441 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4442}
4443
bdcc0924 4444static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4445{
4446 return skb_shinfo(skb)->gso_size;
4447}
4448
36a8f39e 4449/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4450static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4451{
4452 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4453}
4454
d02f51cb
DA
4455/* Note: Should be called only if skb_is_gso(skb) is true */
4456static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4457{
4458 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4459}
4460
4c3024de 4461/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4462static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4463{
4c3024de 4464 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4465}
4466
5293efe6
DB
4467static inline void skb_gso_reset(struct sk_buff *skb)
4468{
4469 skb_shinfo(skb)->gso_size = 0;
4470 skb_shinfo(skb)->gso_segs = 0;
4471 skb_shinfo(skb)->gso_type = 0;
4472}
4473
d02f51cb
DA
4474static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4475 u16 increment)
4476{
4477 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4478 return;
4479 shinfo->gso_size += increment;
4480}
4481
4482static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4483 u16 decrement)
4484{
4485 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4486 return;
4487 shinfo->gso_size -= decrement;
4488}
4489
7965bd4d 4490void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4491
4492static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4493{
4494 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4495 * wanted then gso_type will be set. */
05bdd2f1
ED
4496 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4497
b78462eb
AD
4498 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4499 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4500 __skb_warn_lro_forwarding(skb);
4501 return true;
4502 }
4503 return false;
4504}
4505
35fc92a9
HX
4506static inline void skb_forward_csum(struct sk_buff *skb)
4507{
4508 /* Unfortunately we don't support this one. Any brave souls? */
4509 if (skb->ip_summed == CHECKSUM_COMPLETE)
4510 skb->ip_summed = CHECKSUM_NONE;
4511}
4512
bc8acf2c
ED
4513/**
4514 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4515 * @skb: skb to check
4516 *
4517 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4518 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4519 * use this helper, to document places where we make this assertion.
4520 */
05bdd2f1 4521static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4522{
4523#ifdef DEBUG
4524 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4525#endif
4526}
4527
f35d9d8a 4528bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4529
ed1f50c3 4530int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4531struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4532 unsigned int transport_len,
4533 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4534
3a7c1ee4
AD
4535/**
4536 * skb_head_is_locked - Determine if the skb->head is locked down
4537 * @skb: skb to check
4538 *
4539 * The head on skbs build around a head frag can be removed if they are
4540 * not cloned. This function returns true if the skb head is locked down
4541 * due to either being allocated via kmalloc, or by being a clone with
4542 * multiple references to the head.
4543 */
4544static inline bool skb_head_is_locked(const struct sk_buff *skb)
4545{
4546 return !skb->head_frag || skb_cloned(skb);
4547}
fe6cc55f 4548
179bc67f
EC
4549/* Local Checksum Offload.
4550 * Compute outer checksum based on the assumption that the
4551 * inner checksum will be offloaded later.
d0dcde64 4552 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4553 * explanation of how this works.
179bc67f
EC
4554 * Fill in outer checksum adjustment (e.g. with sum of outer
4555 * pseudo-header) before calling.
4556 * Also ensure that inner checksum is in linear data area.
4557 */
4558static inline __wsum lco_csum(struct sk_buff *skb)
4559{
9e74a6da
AD
4560 unsigned char *csum_start = skb_checksum_start(skb);
4561 unsigned char *l4_hdr = skb_transport_header(skb);
4562 __wsum partial;
179bc67f
EC
4563
4564 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4565 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4566 skb->csum_offset));
4567
179bc67f 4568 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4569 * adjustment filled in by caller) and return result.
179bc67f 4570 */
9e74a6da 4571 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4572}
4573
2c64605b
PNA
4574static inline bool skb_is_redirected(const struct sk_buff *skb)
4575{
4576#ifdef CONFIG_NET_REDIRECT
4577 return skb->redirected;
4578#else
4579 return false;
4580#endif
4581}
4582
4583static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4584{
4585#ifdef CONFIG_NET_REDIRECT
4586 skb->redirected = 1;
4587 skb->from_ingress = from_ingress;
4588 if (skb->from_ingress)
4589 skb->tstamp = 0;
4590#endif
4591}
4592
4593static inline void skb_reset_redirect(struct sk_buff *skb)
4594{
4595#ifdef CONFIG_NET_REDIRECT
4596 skb->redirected = 0;
4597#endif
4598}
4599
1da177e4
LT
4600#endif /* __KERNEL__ */
4601#endif /* _LINUX_SKBUFF_H */