]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/linux/skbuff.h
mm: memcontrol: convert NR_ANON_THPS account to pages
[mirror_ubuntu-jammy-kernel.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
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)
29766bcf 369 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
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
6e3e939f 433 /* generate wifi status information (where possible) */
62b1a8ab 434 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 435
e7fd2885
WB
436 /* generate software time stamp when entering packet scheduling */
437 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
438};
439
e1c8a607 440#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 441 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
442#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
443
06b4feb3
JL
444/* Definitions for flags in struct skb_shared_info */
445enum {
446 /* use zcopy routines */
447 SKBFL_ZEROCOPY_ENABLE = BIT(0),
448
449 /* This indicates at least one fragment might be overwritten
450 * (as in vmsplice(), sendfile() ...)
451 * If we need to compute a TX checksum, we'll need to copy
452 * all frags to avoid possible bad checksum
453 */
454 SKBFL_SHARED_FRAG = BIT(1),
455};
456
457#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
458
a6686f2f
SM
459/*
460 * The callback notifies userspace to release buffers when skb DMA is done in
461 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
462 * The zerocopy_success argument is true if zero copy transmit occurred,
463 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
464 * The ctx field is used to track device context.
465 * The desc field is used to track userspace buffer index.
a6686f2f
SM
466 */
467struct ubuf_info {
36177832
JL
468 void (*callback)(struct sk_buff *, struct ubuf_info *,
469 bool zerocopy_success);
4ab6c99d
WB
470 union {
471 struct {
472 unsigned long desc;
473 void *ctx;
474 };
475 struct {
476 u32 id;
477 u16 len;
478 u16 zerocopy:1;
479 u32 bytelen;
480 };
481 };
c1d1b437 482 refcount_t refcnt;
04c2d33e 483 u8 flags;
a91dbff5
WB
484
485 struct mmpin {
486 struct user_struct *user;
487 unsigned int num_pg;
488 } mmp;
ac45f602
PO
489};
490
52267790
WB
491#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
492
6f89dbce
SV
493int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
494void mm_unaccount_pinned_pages(struct mmpin *mmp);
495
8c793822
JL
496struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
497struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
498 struct ubuf_info *uarg);
52267790 499
8c793822 500void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790 501
8c793822
JL
502void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
503 bool success);
52267790 504
b5947e5d 505int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
506int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
507 struct msghdr *msg, int len,
508 struct ubuf_info *uarg);
509
1da177e4
LT
510/* This data is invariant across clones and lives at
511 * the end of the header data, ie. at skb->end.
512 */
513struct skb_shared_info {
06b4feb3 514 __u8 flags;
de8f3a83
DB
515 __u8 meta_len;
516 __u8 nr_frags;
9f42f126 517 __u8 tx_flags;
7967168c
HX
518 unsigned short gso_size;
519 /* Warning: this field is not always filled in (UFO)! */
520 unsigned short gso_segs;
1da177e4 521 struct sk_buff *frag_list;
ac45f602 522 struct skb_shared_hwtstamps hwtstamps;
7f564528 523 unsigned int gso_type;
09c2d251 524 u32 tskey;
ec7d2f2c
ED
525
526 /*
527 * Warning : all fields before dataref are cleared in __alloc_skb()
528 */
529 atomic_t dataref;
530
69e3c75f
JB
531 /* Intermediate layers must ensure that destructor_arg
532 * remains valid until skb destructor */
533 void * destructor_arg;
a6686f2f 534
fed66381
ED
535 /* must be last field, see pskb_expand_head() */
536 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
537};
538
539/* We divide dataref into two halves. The higher 16 bits hold references
540 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
541 * the entire skb->data. A clone of a headerless skb holds the length of
542 * the header in skb->hdr_len.
1da177e4
LT
543 *
544 * All users must obey the rule that the skb->data reference count must be
545 * greater than or equal to the payload reference count.
546 *
547 * Holding a reference to the payload part means that the user does not
548 * care about modifications to the header part of skb->data.
549 */
550#define SKB_DATAREF_SHIFT 16
551#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
552
d179cd12
DM
553
554enum {
c8753d55
VS
555 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
556 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
557 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
558};
559
7967168c
HX
560enum {
561 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
562
563 /* This indicates the skb is from an untrusted source. */
d9d30adf 564 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
565
566 /* This indicates the tcp segment has CWR set. */
d9d30adf 567 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 568
d9d30adf 569 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 570
d9d30adf 571 SKB_GSO_TCPV6 = 1 << 4,
68c33163 572
d9d30adf 573 SKB_GSO_FCOE = 1 << 5,
73136267 574
d9d30adf 575 SKB_GSO_GRE = 1 << 6,
0d89d203 576
d9d30adf 577 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 578
d9d30adf 579 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 580
d9d30adf 581 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 582
d9d30adf 583 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 584
d9d30adf 585 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 586
d9d30adf 587 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 588
d9d30adf 589 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 590
d9d30adf 591 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 592
d9d30adf 593 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
594
595 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
596
597 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
598
599 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
600};
601
2e07fa9c
ACM
602#if BITS_PER_LONG > 32
603#define NET_SKBUFF_DATA_USES_OFFSET 1
604#endif
605
606#ifdef NET_SKBUFF_DATA_USES_OFFSET
607typedef unsigned int sk_buff_data_t;
608#else
609typedef unsigned char *sk_buff_data_t;
610#endif
611
161e6137 612/**
1da177e4
LT
613 * struct sk_buff - socket buffer
614 * @next: Next buffer in list
615 * @prev: Previous buffer in list
363ec392 616 * @tstamp: Time we arrived/left
d2f273f0
RD
617 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
618 * for retransmit timer
56b17425 619 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 620 * @list: queue head
d84e0bd7 621 * @sk: Socket we are owned by
d2f273f0
RD
622 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
623 * fragmentation management
1da177e4 624 * @dev: Device we arrived on/are leaving by
d2f273f0 625 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 626 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 627 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 628 * @sp: the security path, used for xfrm
1da177e4
LT
629 * @len: Length of actual data
630 * @data_len: Data length
631 * @mac_len: Length of link layer header
334a8132 632 * @hdr_len: writable header length of cloned skb
663ead3b
HX
633 * @csum: Checksum (must include start/offset pair)
634 * @csum_start: Offset from skb->head where checksumming should start
635 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 636 * @priority: Packet queueing priority
60ff7467 637 * @ignore_df: allow local fragmentation
1da177e4 638 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 639 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
640 * @nohdr: Payload reference only, must not modify header
641 * @pkt_type: Packet class
c83c2486 642 * @fclone: skbuff clone status
c83c2486 643 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
644 * @inner_protocol_type: whether the inner protocol is
645 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
646 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
647 * @offload_fwd_mark: Packet was L2-forwarded in hardware
648 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 649 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 650 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
651 * @redirected: packet was redirected by packet classifier
652 * @from_ingress: packet was redirected from the ingress path
31729363
RD
653 * @peeked: this packet has been seen already, so stats have been
654 * done for it, don't do them again
ba9dda3a 655 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
656 * @protocol: Packet protocol from driver
657 * @destructor: Destruct function
e2080072 658 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 659 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 660 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 661 * @skb_iif: ifindex of device we arrived on
1da177e4 662 * @tc_index: Traffic control index
61b905da 663 * @hash: the packet hash
d84e0bd7 664 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
665 * @head_frag: skb was allocated from page fragments,
666 * not allocated by kmalloc() or vmalloc().
8b700862 667 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
df5042f4 668 * @active_extensions: active extensions (skb_ext_id types)
553a5672 669 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 670 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 671 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 672 * ports.
a3b18ddb 673 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
674 * @wifi_acked_valid: wifi_acked was set
675 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 676 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
677 * @encapsulation: indicates the inner headers in the skbuff are valid
678 * @encap_hdr_csum: software checksum is needed
679 * @csum_valid: checksum is already valid
dba00306 680 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
681 * @csum_complete_sw: checksum was completed by software
682 * @csum_level: indicates the number of consecutive checksums found in
683 * the packet minus one that have been verified as
684 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 685 * @dst_pending_confirm: need to confirm neighbour
a48d189e 686 * @decrypted: Decrypted SKB
161e6137 687 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 688 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 689 * @secmark: security marking
d84e0bd7 690 * @mark: Generic packet mark
d2f273f0
RD
691 * @reserved_tailroom: (aka @mark) number of bytes of free space available
692 * at the tail of an sk_buff
693 * @vlan_present: VLAN tag is present
86a9bad3 694 * @vlan_proto: vlan encapsulation protocol
6aa895b0 695 * @vlan_tci: vlan tag control information
0d89d203 696 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
697 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
698 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
699 * @inner_transport_header: Inner transport layer header (encapsulation)
700 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 701 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
702 * @transport_header: Transport layer header
703 * @network_header: Network layer header
704 * @mac_header: Link layer header
fa69ee5a 705 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
706 * @tail: Tail pointer
707 * @end: End pointer
708 * @head: Head of buffer
709 * @data: Data head pointer
710 * @truesize: Buffer size
711 * @users: User count - see {datagram,tcp}.c
df5042f4 712 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
713 */
714
715struct sk_buff {
363ec392 716 union {
56b17425
ED
717 struct {
718 /* These two members must be first. */
719 struct sk_buff *next;
720 struct sk_buff *prev;
721
722 union {
bffa72cf
ED
723 struct net_device *dev;
724 /* Some protocols might use this space to store information,
725 * while device pointer would be NULL.
726 * UDP receive path is one user.
727 */
728 unsigned long dev_scratch;
56b17425
ED
729 };
730 };
fa0f5273 731 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 732 struct list_head list;
363ec392 733 };
fa0f5273
PO
734
735 union {
736 struct sock *sk;
737 int ip_defrag_offset;
738 };
1da177e4 739
c84d9490 740 union {
bffa72cf 741 ktime_t tstamp;
d3edd06e 742 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 743 };
1da177e4
LT
744 /*
745 * This is the control buffer. It is free to use for every
746 * layer. Please put your private variables there. If you
747 * want to keep them across layers you have to do a skb_clone()
748 * first. This is owned by whoever has the skb queued ATM.
749 */
da3f5cf1 750 char cb[48] __aligned(8);
1da177e4 751
e2080072
ED
752 union {
753 struct {
754 unsigned long _skb_refdst;
755 void (*destructor)(struct sk_buff *skb);
756 };
757 struct list_head tcp_tsorted_anchor;
758 };
759
b1937227 760#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 761 unsigned long _nfct;
da3f5cf1 762#endif
1da177e4 763 unsigned int len,
334a8132
PM
764 data_len;
765 __u16 mac_len,
766 hdr_len;
b1937227
ED
767
768 /* Following fields are _not_ copied in __copy_skb_header()
769 * Note that queue_mapping is here mostly to fill a hole.
770 */
b1937227 771 __u16 queue_mapping;
36bbef52
DB
772
773/* if you move cloned around you also must adapt those constants */
774#ifdef __BIG_ENDIAN_BITFIELD
775#define CLONED_MASK (1 << 7)
776#else
777#define CLONED_MASK 1
778#endif
779#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
780
d2f273f0 781 /* private: */
36bbef52 782 __u8 __cloned_offset[0];
d2f273f0 783 /* public: */
b1937227 784 __u8 cloned:1,
6869c4d8 785 nohdr:1,
b84f4cc9 786 fclone:2,
a59322be 787 peeked:1,
b1937227 788 head_frag:1,
8b700862 789 pfmemalloc:1;
df5042f4
FW
790#ifdef CONFIG_SKB_EXTENSIONS
791 __u8 active_extensions;
792#endif
b1937227
ED
793 /* fields enclosed in headers_start/headers_end are copied
794 * using a single memcpy() in __copy_skb_header()
795 */
ebcf34f3 796 /* private: */
b1937227 797 __u32 headers_start[0];
ebcf34f3 798 /* public: */
4031ae6e 799
233577a2
HFS
800/* if you move pkt_type around you also must adapt those constants */
801#ifdef __BIG_ENDIAN_BITFIELD
802#define PKT_TYPE_MAX (7 << 5)
803#else
804#define PKT_TYPE_MAX 7
1da177e4 805#endif
233577a2 806#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 807
d2f273f0 808 /* private: */
233577a2 809 __u8 __pkt_type_offset[0];
d2f273f0 810 /* public: */
b1937227 811 __u8 pkt_type:3;
b1937227 812 __u8 ignore_df:1;
b1937227
ED
813 __u8 nf_trace:1;
814 __u8 ip_summed:2;
3853b584 815 __u8 ooo_okay:1;
8b700862 816
61b905da 817 __u8 l4_hash:1;
a3b18ddb 818 __u8 sw_hash:1;
6e3e939f
JB
819 __u8 wifi_acked_valid:1;
820 __u8 wifi_acked:1;
3bdc0eba 821 __u8 no_fcs:1;
77cffe23 822 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 823 __u8 encapsulation:1;
7e2b10c1 824 __u8 encap_hdr_csum:1;
5d0c2b95 825 __u8 csum_valid:1;
8b700862 826
0c4b2d37
MM
827#ifdef __BIG_ENDIAN_BITFIELD
828#define PKT_VLAN_PRESENT_BIT 7
829#else
830#define PKT_VLAN_PRESENT_BIT 0
831#endif
832#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 833 /* private: */
0c4b2d37 834 __u8 __pkt_vlan_present_offset[0];
d2f273f0 835 /* public: */
0c4b2d37 836 __u8 vlan_present:1;
7e3cead5 837 __u8 csum_complete_sw:1;
b1937227 838 __u8 csum_level:2;
dba00306 839 __u8 csum_not_inet:1;
4ff06203 840 __u8 dst_pending_confirm:1;
b1937227
ED
841#ifdef CONFIG_IPV6_NDISC_NODETYPE
842 __u8 ndisc_nodetype:2;
843#endif
8b700862 844
0c4b2d37 845 __u8 ipvs_property:1;
8bce6d7d 846 __u8 inner_protocol_type:1;
e585f236 847 __u8 remcsum_offload:1;
6bc506b4
IS
848#ifdef CONFIG_NET_SWITCHDEV
849 __u8 offload_fwd_mark:1;
875e8939 850 __u8 offload_l3_fwd_mark:1;
6bc506b4 851#endif
e7246e12
WB
852#ifdef CONFIG_NET_CLS_ACT
853 __u8 tc_skip_classify:1;
8dc07fdb 854 __u8 tc_at_ingress:1;
2c64605b
PNA
855#endif
856#ifdef CONFIG_NET_REDIRECT
857 __u8 redirected:1;
858 __u8 from_ingress:1;
e7246e12 859#endif
a48d189e
SB
860#ifdef CONFIG_TLS_DEVICE
861 __u8 decrypted:1;
862#endif
b1937227
ED
863
864#ifdef CONFIG_NET_SCHED
865 __u16 tc_index; /* traffic control index */
b1937227 866#endif
fe55f6d5 867
b1937227
ED
868 union {
869 __wsum csum;
870 struct {
871 __u16 csum_start;
872 __u16 csum_offset;
873 };
874 };
875 __u32 priority;
876 int skb_iif;
877 __u32 hash;
878 __be16 vlan_proto;
879 __u16 vlan_tci;
2bd82484
ED
880#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
881 union {
882 unsigned int napi_id;
883 unsigned int sender_cpu;
884 };
97fc2f08 885#endif
984bc16c 886#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 887 __u32 secmark;
0c4f691f 888#endif
0c4f691f 889
3b885787
NH
890 union {
891 __u32 mark;
16fad69c 892 __u32 reserved_tailroom;
3b885787 893 };
1da177e4 894
8bce6d7d
TH
895 union {
896 __be16 inner_protocol;
897 __u8 inner_ipproto;
898 };
899
1a37e412
SH
900 __u16 inner_transport_header;
901 __u16 inner_network_header;
902 __u16 inner_mac_header;
b1937227
ED
903
904 __be16 protocol;
1a37e412
SH
905 __u16 transport_header;
906 __u16 network_header;
907 __u16 mac_header;
b1937227 908
fa69ee5a
ME
909#ifdef CONFIG_KCOV
910 u64 kcov_handle;
911#endif
912
ebcf34f3 913 /* private: */
b1937227 914 __u32 headers_end[0];
ebcf34f3 915 /* public: */
b1937227 916
1da177e4 917 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 918 sk_buff_data_t tail;
4305b541 919 sk_buff_data_t end;
1da177e4 920 unsigned char *head,
4305b541 921 *data;
27a884dc 922 unsigned int truesize;
63354797 923 refcount_t users;
df5042f4
FW
924
925#ifdef CONFIG_SKB_EXTENSIONS
926 /* only useable after checking ->active_extensions != 0 */
927 struct skb_ext *extensions;
928#endif
1da177e4
LT
929};
930
931#ifdef __KERNEL__
932/*
933 * Handling routines are only of interest to the kernel
934 */
1da177e4 935
c93bdd0e
MG
936#define SKB_ALLOC_FCLONE 0x01
937#define SKB_ALLOC_RX 0x02
fd11a83d 938#define SKB_ALLOC_NAPI 0x04
c93bdd0e 939
161e6137
PT
940/**
941 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
942 * @skb: buffer
943 */
c93bdd0e
MG
944static inline bool skb_pfmemalloc(const struct sk_buff *skb)
945{
946 return unlikely(skb->pfmemalloc);
947}
948
7fee226a
ED
949/*
950 * skb might have a dst pointer attached, refcounted or not.
951 * _skb_refdst low order bit is set if refcount was _not_ taken
952 */
953#define SKB_DST_NOREF 1UL
954#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
955
956/**
957 * skb_dst - returns skb dst_entry
958 * @skb: buffer
959 *
960 * Returns skb dst_entry, regardless of reference taken or not.
961 */
adf30907
ED
962static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
963{
161e6137 964 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
965 * rcu_read_lock section
966 */
967 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
968 !rcu_read_lock_held() &&
969 !rcu_read_lock_bh_held());
970 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
971}
972
7fee226a
ED
973/**
974 * skb_dst_set - sets skb dst
975 * @skb: buffer
976 * @dst: dst entry
977 *
978 * Sets skb dst, assuming a reference was taken on dst and should
979 * be released by skb_dst_drop()
980 */
adf30907
ED
981static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
982{
7fee226a
ED
983 skb->_skb_refdst = (unsigned long)dst;
984}
985
932bc4d7
JA
986/**
987 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
988 * @skb: buffer
989 * @dst: dst entry
990 *
991 * Sets skb dst, assuming a reference was not taken on dst.
992 * If dst entry is cached, we do not take reference and dst_release
993 * will be avoided by refdst_drop. If dst entry is not cached, we take
994 * reference, so that last dst_release can destroy the dst immediately.
995 */
996static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
997{
dbfc4fb7
HFS
998 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
999 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1000}
7fee226a
ED
1001
1002/**
25985edc 1003 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1004 * @skb: buffer
1005 */
1006static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1007{
1008 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1009}
1010
161e6137
PT
1011/**
1012 * skb_rtable - Returns the skb &rtable
1013 * @skb: buffer
1014 */
511c3f92
ED
1015static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1016{
adf30907 1017 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1018}
1019
8b10cab6
JHS
1020/* For mangling skb->pkt_type from user space side from applications
1021 * such as nft, tc, etc, we only allow a conservative subset of
1022 * possible pkt_types to be set.
1023*/
1024static inline bool skb_pkt_type_ok(u32 ptype)
1025{
1026 return ptype <= PACKET_OTHERHOST;
1027}
1028
161e6137
PT
1029/**
1030 * skb_napi_id - Returns the skb's NAPI id
1031 * @skb: buffer
1032 */
90b602f8
ML
1033static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1034{
1035#ifdef CONFIG_NET_RX_BUSY_POLL
1036 return skb->napi_id;
1037#else
1038 return 0;
1039#endif
1040}
1041
161e6137
PT
1042/**
1043 * skb_unref - decrement the skb's reference count
1044 * @skb: buffer
1045 *
1046 * Returns true if we can free the skb.
1047 */
3889a803
PA
1048static inline bool skb_unref(struct sk_buff *skb)
1049{
1050 if (unlikely(!skb))
1051 return false;
63354797 1052 if (likely(refcount_read(&skb->users) == 1))
3889a803 1053 smp_rmb();
63354797 1054 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1055 return false;
1056
1057 return true;
1058}
1059
0a463c78 1060void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1061void kfree_skb(struct sk_buff *skb);
1062void kfree_skb_list(struct sk_buff *segs);
6413139d 1063void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1064void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1065
1066#ifdef CONFIG_TRACEPOINTS
7965bd4d 1067void consume_skb(struct sk_buff *skb);
be769db2
HX
1068#else
1069static inline void consume_skb(struct sk_buff *skb)
1070{
1071 return kfree_skb(skb);
1072}
1073#endif
1074
ca2c1418 1075void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1076void __kfree_skb(struct sk_buff *skb);
d7e8883c 1077extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1078
7965bd4d
JP
1079void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1080bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1081 bool *fragstolen, int *delta_truesize);
bad43ca8 1082
7965bd4d
JP
1083struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1084 int node);
2ea2f62c 1085struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1086struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1087struct sk_buff *build_skb_around(struct sk_buff *skb,
1088 void *data, unsigned int frag_size);
161e6137 1089
f450d539
AL
1090struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1091
161e6137
PT
1092/**
1093 * alloc_skb - allocate a network buffer
1094 * @size: size to allocate
1095 * @priority: allocation mask
1096 *
1097 * This function is a convenient wrapper around __alloc_skb().
1098 */
d179cd12 1099static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1100 gfp_t priority)
d179cd12 1101{
564824b0 1102 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1103}
1104
2e4e4410
ED
1105struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1106 unsigned long data_len,
1107 int max_page_order,
1108 int *errcode,
1109 gfp_t gfp_mask);
da29e4b4 1110struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1111
d0bf4a9e
ED
1112/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1113struct sk_buff_fclones {
1114 struct sk_buff skb1;
1115
1116 struct sk_buff skb2;
1117
2638595a 1118 refcount_t fclone_ref;
d0bf4a9e
ED
1119};
1120
1121/**
1122 * skb_fclone_busy - check if fclone is busy
293de7de 1123 * @sk: socket
d0bf4a9e
ED
1124 * @skb: buffer
1125 *
bda13fed 1126 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1127 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1128 * so we also check that this didnt happen.
d0bf4a9e 1129 */
39bb5e62
ED
1130static inline bool skb_fclone_busy(const struct sock *sk,
1131 const struct sk_buff *skb)
d0bf4a9e
ED
1132{
1133 const struct sk_buff_fclones *fclones;
1134
1135 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1136
1137 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1138 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1139 fclones->skb2.sk == sk;
d0bf4a9e
ED
1140}
1141
161e6137
PT
1142/**
1143 * alloc_skb_fclone - allocate a network buffer from fclone cache
1144 * @size: size to allocate
1145 * @priority: allocation mask
1146 *
1147 * This function is a convenient wrapper around __alloc_skb().
1148 */
d179cd12 1149static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1150 gfp_t priority)
d179cd12 1151{
c93bdd0e 1152 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1153}
1154
7965bd4d 1155struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1156void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1157int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1158struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1159void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1160struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1161struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1162 gfp_t gfp_mask, bool fclone);
1163static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1164 gfp_t gfp_mask)
1165{
1166 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1167}
7965bd4d
JP
1168
1169int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1170struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1171 unsigned int headroom);
1172struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1173 int newtailroom, gfp_t priority);
48a1df65
JD
1174int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1175 int offset, int len);
1176int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1177 int offset, int len);
7965bd4d 1178int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1179int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1180
1181/**
1182 * skb_pad - zero pad the tail of an skb
1183 * @skb: buffer to pad
1184 * @pad: space to pad
1185 *
1186 * Ensure that a buffer is followed by a padding area that is zero
1187 * filled. Used by network drivers which may DMA or transfer data
1188 * beyond the buffer end onto the wire.
1189 *
1190 * May return error in out of memory cases. The skb is freed on error.
1191 */
1192static inline int skb_pad(struct sk_buff *skb, int pad)
1193{
1194 return __skb_pad(skb, pad, true);
1195}
ead2ceb0 1196#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1197
be12a1fe
HFS
1198int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1199 int offset, size_t size);
1200
d94d9fee 1201struct skb_seq_state {
677e90ed
TG
1202 __u32 lower_offset;
1203 __u32 upper_offset;
1204 __u32 frag_idx;
1205 __u32 stepped_offset;
1206 struct sk_buff *root_skb;
1207 struct sk_buff *cur_skb;
1208 __u8 *frag_data;
97550f6f 1209 __u32 frag_off;
677e90ed
TG
1210};
1211
7965bd4d
JP
1212void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1213 unsigned int to, struct skb_seq_state *st);
1214unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1215 struct skb_seq_state *st);
1216void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1217
7965bd4d 1218unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1219 unsigned int to, struct ts_config *config);
3fc7e8a6 1220
09323cc4
TH
1221/*
1222 * Packet hash types specify the type of hash in skb_set_hash.
1223 *
1224 * Hash types refer to the protocol layer addresses which are used to
1225 * construct a packet's hash. The hashes are used to differentiate or identify
1226 * flows of the protocol layer for the hash type. Hash types are either
1227 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1228 *
1229 * Properties of hashes:
1230 *
1231 * 1) Two packets in different flows have different hash values
1232 * 2) Two packets in the same flow should have the same hash value
1233 *
1234 * A hash at a higher layer is considered to be more specific. A driver should
1235 * set the most specific hash possible.
1236 *
1237 * A driver cannot indicate a more specific hash than the layer at which a hash
1238 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1239 *
1240 * A driver may indicate a hash level which is less specific than the
1241 * actual layer the hash was computed on. For instance, a hash computed
1242 * at L4 may be considered an L3 hash. This should only be done if the
1243 * driver can't unambiguously determine that the HW computed the hash at
1244 * the higher layer. Note that the "should" in the second property above
1245 * permits this.
1246 */
1247enum pkt_hash_types {
1248 PKT_HASH_TYPE_NONE, /* Undefined type */
1249 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1250 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1251 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1252};
1253
bcc83839 1254static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1255{
bcc83839 1256 skb->hash = 0;
a3b18ddb 1257 skb->sw_hash = 0;
bcc83839
TH
1258 skb->l4_hash = 0;
1259}
1260
1261static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1262{
1263 if (!skb->l4_hash)
1264 skb_clear_hash(skb);
1265}
1266
1267static inline void
1268__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1269{
1270 skb->l4_hash = is_l4;
1271 skb->sw_hash = is_sw;
61b905da 1272 skb->hash = hash;
09323cc4
TH
1273}
1274
bcc83839
TH
1275static inline void
1276skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1277{
1278 /* Used by drivers to set hash from HW */
1279 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1280}
1281
1282static inline void
1283__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1284{
1285 __skb_set_hash(skb, hash, true, is_l4);
1286}
1287
e5276937 1288void __skb_get_hash(struct sk_buff *skb);
b917783c 1289u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1290u32 skb_get_poff(const struct sk_buff *skb);
1291u32 __skb_get_poff(const struct sk_buff *skb, void *data,
72a338bc 1292 const struct flow_keys_basic *keys, int hlen);
e5276937
TH
1293__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1294 void *data, int hlen_proto);
1295
1296static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1297 int thoff, u8 ip_proto)
1298{
1299 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1300}
1301
1302void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1303 const struct flow_dissector_key *key,
1304 unsigned int key_count);
1305
089b19a9
SF
1306struct bpf_flow_dissector;
1307bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1308 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1309
3cbf4ffb
SF
1310bool __skb_flow_dissect(const struct net *net,
1311 const struct sk_buff *skb,
e5276937
TH
1312 struct flow_dissector *flow_dissector,
1313 void *target_container,
cd79a238
TH
1314 void *data, __be16 proto, int nhoff, int hlen,
1315 unsigned int flags);
e5276937
TH
1316
1317static inline bool skb_flow_dissect(const struct sk_buff *skb,
1318 struct flow_dissector *flow_dissector,
cd79a238 1319 void *target_container, unsigned int flags)
e5276937 1320{
3cbf4ffb
SF
1321 return __skb_flow_dissect(NULL, skb, flow_dissector,
1322 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1323}
1324
1325static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1326 struct flow_keys *flow,
1327 unsigned int flags)
e5276937
TH
1328{
1329 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1330 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1331 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1332}
1333
72a338bc 1334static inline bool
3cbf4ffb
SF
1335skb_flow_dissect_flow_keys_basic(const struct net *net,
1336 const struct sk_buff *skb,
72a338bc
PA
1337 struct flow_keys_basic *flow, void *data,
1338 __be16 proto, int nhoff, int hlen,
1339 unsigned int flags)
e5276937
TH
1340{
1341 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1342 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1343 data, proto, nhoff, hlen, flags);
e5276937
TH
1344}
1345
82828b88
JP
1346void skb_flow_dissect_meta(const struct sk_buff *skb,
1347 struct flow_dissector *flow_dissector,
1348 void *target_container);
1349
75a56758 1350/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1351 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1352 * to user states.
1353 */
1354void
1355skb_flow_dissect_ct(const struct sk_buff *skb,
1356 struct flow_dissector *flow_dissector,
1357 void *target_container,
7baf2429 1358 u16 *ctinfo_map, size_t mapsize,
1359 bool post_ct);
62b32379
SH
1360void
1361skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1362 struct flow_dissector *flow_dissector,
1363 void *target_container);
1364
0cb09aff
AL
1365void skb_flow_dissect_hash(const struct sk_buff *skb,
1366 struct flow_dissector *flow_dissector,
1367 void *target_container);
1368
3958afa1 1369static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1370{
a3b18ddb 1371 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1372 __skb_get_hash(skb);
bfb564e7 1373
61b905da 1374 return skb->hash;
bfb564e7
KK
1375}
1376
20a17bf6 1377static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1378{
c6cc1ca7
TH
1379 if (!skb->l4_hash && !skb->sw_hash) {
1380 struct flow_keys keys;
de4c1f8b 1381 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1382
de4c1f8b 1383 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1384 }
f70ea018
TH
1385
1386 return skb->hash;
1387}
1388
55667441
ED
1389__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1390 const siphash_key_t *perturb);
50fb7992 1391
57bdf7f4
TH
1392static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1393{
61b905da 1394 return skb->hash;
57bdf7f4
TH
1395}
1396
3df7a74e
TH
1397static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1398{
61b905da 1399 to->hash = from->hash;
a3b18ddb 1400 to->sw_hash = from->sw_hash;
61b905da 1401 to->l4_hash = from->l4_hash;
3df7a74e
TH
1402};
1403
41477662
JK
1404static inline void skb_copy_decrypted(struct sk_buff *to,
1405 const struct sk_buff *from)
1406{
1407#ifdef CONFIG_TLS_DEVICE
1408 to->decrypted = from->decrypted;
1409#endif
1410}
1411
4305b541
ACM
1412#ifdef NET_SKBUFF_DATA_USES_OFFSET
1413static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1414{
1415 return skb->head + skb->end;
1416}
ec47ea82
AD
1417
1418static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1419{
1420 return skb->end;
1421}
4305b541
ACM
1422#else
1423static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1424{
1425 return skb->end;
1426}
ec47ea82
AD
1427
1428static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1429{
1430 return skb->end - skb->head;
1431}
4305b541
ACM
1432#endif
1433
1da177e4 1434/* Internal */
4305b541 1435#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1436
ac45f602
PO
1437static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1438{
1439 return &skb_shinfo(skb)->hwtstamps;
1440}
1441
52267790
WB
1442static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1443{
06b4feb3 1444 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1445
1446 return is_zcopy ? skb_uarg(skb) : NULL;
1447}
1448
8e044917 1449static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1450{
1451 refcount_inc(&uarg->refcnt);
1452}
1453
9ee5e5ad
JL
1454static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1455{
1456 skb_shinfo(skb)->destructor_arg = uarg;
1457 skb_shinfo(skb)->flags |= uarg->flags;
1458}
1459
52900d22
WB
1460static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1461 bool *have_ref)
52267790
WB
1462{
1463 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1464 if (unlikely(have_ref && *have_ref))
1465 *have_ref = false;
1466 else
8e044917 1467 net_zcopy_get(uarg);
9ee5e5ad 1468 skb_zcopy_init(skb, uarg);
52267790
WB
1469 }
1470}
1471
5cd8d46e
WB
1472static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1473{
1474 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1475 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1476}
1477
1478static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1479{
1480 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1481}
1482
1483static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1484{
1485 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1486}
1487
8e044917 1488static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1489{
1490 if (uarg)
36177832 1491 uarg->callback(NULL, uarg, true);
59776362
JL
1492}
1493
8e044917 1494static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1495{
1496 if (uarg) {
8c793822
JL
1497 if (uarg->callback == msg_zerocopy_callback)
1498 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1499 else if (have_uref)
8e044917 1500 net_zcopy_put(uarg);
236a6b1c
JL
1501 }
1502}
1503
52267790 1504/* Release a reference on a zerocopy structure */
36177832 1505static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1506{
1507 struct ubuf_info *uarg = skb_zcopy(skb);
1508
1509 if (uarg) {
36177832
JL
1510 if (!skb_zcopy_is_nouarg(skb))
1511 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1512
06b4feb3 1513 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
52267790
WB
1514 }
1515}
1516
a8305bff
DM
1517static inline void skb_mark_not_on_list(struct sk_buff *skb)
1518{
1519 skb->next = NULL;
1520}
1521
dcfea72e 1522/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1523#define skb_list_walk_safe(first, skb, next_skb) \
1524 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1525 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1526
992cba7e
DM
1527static inline void skb_list_del_init(struct sk_buff *skb)
1528{
1529 __list_del_entry(&skb->list);
1530 skb_mark_not_on_list(skb);
1531}
1532
1da177e4
LT
1533/**
1534 * skb_queue_empty - check if a queue is empty
1535 * @list: queue head
1536 *
1537 * Returns true if the queue is empty, false otherwise.
1538 */
1539static inline int skb_queue_empty(const struct sk_buff_head *list)
1540{
fd44b93c 1541 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1542}
1543
d7d16a89
ED
1544/**
1545 * skb_queue_empty_lockless - check if a queue is empty
1546 * @list: queue head
1547 *
1548 * Returns true if the queue is empty, false otherwise.
1549 * This variant can be used in lockless contexts.
1550 */
1551static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1552{
1553 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1554}
1555
1556
fc7ebb21
DM
1557/**
1558 * skb_queue_is_last - check if skb is the last entry in the queue
1559 * @list: queue head
1560 * @skb: buffer
1561 *
1562 * Returns true if @skb is the last buffer on the list.
1563 */
1564static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1565 const struct sk_buff *skb)
1566{
fd44b93c 1567 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1568}
1569
832d11c5
IJ
1570/**
1571 * skb_queue_is_first - check if skb is the first entry in the queue
1572 * @list: queue head
1573 * @skb: buffer
1574 *
1575 * Returns true if @skb is the first buffer on the list.
1576 */
1577static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1578 const struct sk_buff *skb)
1579{
fd44b93c 1580 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1581}
1582
249c8b42
DM
1583/**
1584 * skb_queue_next - return the next packet in the queue
1585 * @list: queue head
1586 * @skb: current buffer
1587 *
1588 * Return the next packet in @list after @skb. It is only valid to
1589 * call this if skb_queue_is_last() evaluates to false.
1590 */
1591static inline struct sk_buff *skb_queue_next(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_last(list, skb));
1598 return skb->next;
1599}
1600
832d11c5
IJ
1601/**
1602 * skb_queue_prev - return the prev packet in the queue
1603 * @list: queue head
1604 * @skb: current buffer
1605 *
1606 * Return the prev packet in @list before @skb. It is only valid to
1607 * call this if skb_queue_is_first() evaluates to false.
1608 */
1609static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1610 const struct sk_buff *skb)
1611{
1612 /* This BUG_ON may seem severe, but if we just return then we
1613 * are going to dereference garbage.
1614 */
1615 BUG_ON(skb_queue_is_first(list, skb));
1616 return skb->prev;
1617}
1618
1da177e4
LT
1619/**
1620 * skb_get - reference buffer
1621 * @skb: buffer to reference
1622 *
1623 * Makes another reference to a socket buffer and returns a pointer
1624 * to the buffer.
1625 */
1626static inline struct sk_buff *skb_get(struct sk_buff *skb)
1627{
63354797 1628 refcount_inc(&skb->users);
1da177e4
LT
1629 return skb;
1630}
1631
1632/*
f8821f96 1633 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1634 */
1635
1da177e4
LT
1636/**
1637 * skb_cloned - is the buffer a clone
1638 * @skb: buffer to check
1639 *
1640 * Returns true if the buffer was generated with skb_clone() and is
1641 * one of multiple shared copies of the buffer. Cloned buffers are
1642 * shared data so must not be written to under normal circumstances.
1643 */
1644static inline int skb_cloned(const struct sk_buff *skb)
1645{
1646 return skb->cloned &&
1647 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1648}
1649
14bbd6a5
PS
1650static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1651{
d0164adc 1652 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1653
1654 if (skb_cloned(skb))
1655 return pskb_expand_head(skb, 0, 0, pri);
1656
1657 return 0;
1658}
1659
1da177e4
LT
1660/**
1661 * skb_header_cloned - is the header a clone
1662 * @skb: buffer to check
1663 *
1664 * Returns true if modifying the header part of the buffer requires
1665 * the data to be copied.
1666 */
1667static inline int skb_header_cloned(const struct sk_buff *skb)
1668{
1669 int dataref;
1670
1671 if (!skb->cloned)
1672 return 0;
1673
1674 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1675 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1676 return dataref != 1;
1677}
1678
9580bf2e
ED
1679static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1680{
1681 might_sleep_if(gfpflags_allow_blocking(pri));
1682
1683 if (skb_header_cloned(skb))
1684 return pskb_expand_head(skb, 0, 0, pri);
1685
1686 return 0;
1687}
1688
f4a775d1
ED
1689/**
1690 * __skb_header_release - release reference to header
1691 * @skb: buffer to operate on
f4a775d1
ED
1692 */
1693static inline void __skb_header_release(struct sk_buff *skb)
1694{
1695 skb->nohdr = 1;
1696 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1697}
1698
1699
1da177e4
LT
1700/**
1701 * skb_shared - is the buffer shared
1702 * @skb: buffer to check
1703 *
1704 * Returns true if more than one person has a reference to this
1705 * buffer.
1706 */
1707static inline int skb_shared(const struct sk_buff *skb)
1708{
63354797 1709 return refcount_read(&skb->users) != 1;
1da177e4
LT
1710}
1711
1712/**
1713 * skb_share_check - check if buffer is shared and if so clone it
1714 * @skb: buffer to check
1715 * @pri: priority for memory allocation
1716 *
1717 * If the buffer is shared the buffer is cloned and the old copy
1718 * drops a reference. A new clone with a single reference is returned.
1719 * If the buffer is not shared the original buffer is returned. When
1720 * being called from interrupt status or with spinlocks held pri must
1721 * be GFP_ATOMIC.
1722 *
1723 * NULL is returned on a memory allocation failure.
1724 */
47061bc4 1725static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1726{
d0164adc 1727 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1728 if (skb_shared(skb)) {
1729 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1730
1731 if (likely(nskb))
1732 consume_skb(skb);
1733 else
1734 kfree_skb(skb);
1da177e4
LT
1735 skb = nskb;
1736 }
1737 return skb;
1738}
1739
1740/*
1741 * Copy shared buffers into a new sk_buff. We effectively do COW on
1742 * packets to handle cases where we have a local reader and forward
1743 * and a couple of other messy ones. The normal one is tcpdumping
1744 * a packet thats being forwarded.
1745 */
1746
1747/**
1748 * skb_unshare - make a copy of a shared buffer
1749 * @skb: buffer to check
1750 * @pri: priority for memory allocation
1751 *
1752 * If the socket buffer is a clone then this function creates a new
1753 * copy of the data, drops a reference count on the old copy and returns
1754 * the new copy with the reference count at 1. If the buffer is not a clone
1755 * the original buffer is returned. When called with a spinlock held or
1756 * from interrupt state @pri must be %GFP_ATOMIC
1757 *
1758 * %NULL is returned on a memory allocation failure.
1759 */
e2bf521d 1760static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1761 gfp_t pri)
1da177e4 1762{
d0164adc 1763 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1764 if (skb_cloned(skb)) {
1765 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1766
1767 /* Free our shared copy */
1768 if (likely(nskb))
1769 consume_skb(skb);
1770 else
1771 kfree_skb(skb);
1da177e4
LT
1772 skb = nskb;
1773 }
1774 return skb;
1775}
1776
1777/**
1a5778aa 1778 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1779 * @list_: list to peek at
1780 *
1781 * Peek an &sk_buff. Unlike most other operations you _MUST_
1782 * be careful with this one. A peek leaves the buffer on the
1783 * list and someone else may run off with it. You must hold
1784 * the appropriate locks or have a private queue to do this.
1785 *
1786 * Returns %NULL for an empty list or a pointer to the head element.
1787 * The reference count is not incremented and the reference is therefore
1788 * volatile. Use with caution.
1789 */
05bdd2f1 1790static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1791{
18d07000
ED
1792 struct sk_buff *skb = list_->next;
1793
1794 if (skb == (struct sk_buff *)list_)
1795 skb = NULL;
1796 return skb;
1da177e4
LT
1797}
1798
8b69bd7d
DM
1799/**
1800 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1801 * @list_: list to peek at
1802 *
1803 * Like skb_peek(), but the caller knows that the list is not empty.
1804 */
1805static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1806{
1807 return list_->next;
1808}
1809
da5ef6e5
PE
1810/**
1811 * skb_peek_next - peek skb following the given one from a queue
1812 * @skb: skb to start from
1813 * @list_: list to peek at
1814 *
1815 * Returns %NULL when the end of the list is met or a pointer to the
1816 * next element. The reference count is not incremented and the
1817 * reference is therefore volatile. Use with caution.
1818 */
1819static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1820 const struct sk_buff_head *list_)
1821{
1822 struct sk_buff *next = skb->next;
18d07000 1823
da5ef6e5
PE
1824 if (next == (struct sk_buff *)list_)
1825 next = NULL;
1826 return next;
1827}
1828
1da177e4 1829/**
1a5778aa 1830 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1831 * @list_: list to peek at
1832 *
1833 * Peek an &sk_buff. Unlike most other operations you _MUST_
1834 * be careful with this one. A peek leaves the buffer on the
1835 * list and someone else may run off with it. You must hold
1836 * the appropriate locks or have a private queue to do this.
1837 *
1838 * Returns %NULL for an empty list or a pointer to the tail element.
1839 * The reference count is not incremented and the reference is therefore
1840 * volatile. Use with caution.
1841 */
05bdd2f1 1842static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1843{
f8cc62ca 1844 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1845
1846 if (skb == (struct sk_buff *)list_)
1847 skb = NULL;
1848 return skb;
1849
1da177e4
LT
1850}
1851
1852/**
1853 * skb_queue_len - get queue length
1854 * @list_: list to measure
1855 *
1856 * Return the length of an &sk_buff queue.
1857 */
1858static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1859{
1860 return list_->qlen;
1861}
1862
86b18aaa
QC
1863/**
1864 * skb_queue_len_lockless - get queue length
1865 * @list_: list to measure
1866 *
1867 * Return the length of an &sk_buff queue.
1868 * This variant can be used in lockless contexts.
1869 */
1870static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1871{
1872 return READ_ONCE(list_->qlen);
1873}
1874
67fed459
DM
1875/**
1876 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1877 * @list: queue to initialize
1878 *
1879 * This initializes only the list and queue length aspects of
1880 * an sk_buff_head object. This allows to initialize the list
1881 * aspects of an sk_buff_head without reinitializing things like
1882 * the spinlock. It can also be used for on-stack sk_buff_head
1883 * objects where the spinlock is known to not be used.
1884 */
1885static inline void __skb_queue_head_init(struct sk_buff_head *list)
1886{
1887 list->prev = list->next = (struct sk_buff *)list;
1888 list->qlen = 0;
1889}
1890
76f10ad0
AV
1891/*
1892 * This function creates a split out lock class for each invocation;
1893 * this is needed for now since a whole lot of users of the skb-queue
1894 * infrastructure in drivers have different locking usage (in hardirq)
1895 * than the networking core (in softirq only). In the long run either the
1896 * network layer or drivers should need annotation to consolidate the
1897 * main types of usage into 3 classes.
1898 */
1da177e4
LT
1899static inline void skb_queue_head_init(struct sk_buff_head *list)
1900{
1901 spin_lock_init(&list->lock);
67fed459 1902 __skb_queue_head_init(list);
1da177e4
LT
1903}
1904
c2ecba71
PE
1905static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1906 struct lock_class_key *class)
1907{
1908 skb_queue_head_init(list);
1909 lockdep_set_class(&list->lock, class);
1910}
1911
1da177e4 1912/*
bf299275 1913 * Insert an sk_buff on a list.
1da177e4
LT
1914 *
1915 * The "__skb_xxxx()" functions are the non-atomic ones that
1916 * can only be called with interrupts disabled.
1917 */
bf299275
GR
1918static inline void __skb_insert(struct sk_buff *newsk,
1919 struct sk_buff *prev, struct sk_buff *next,
1920 struct sk_buff_head *list)
1921{
f8cc62ca
ED
1922 /* See skb_queue_empty_lockless() and skb_peek_tail()
1923 * for the opposite READ_ONCE()
1924 */
d7d16a89
ED
1925 WRITE_ONCE(newsk->next, next);
1926 WRITE_ONCE(newsk->prev, prev);
1927 WRITE_ONCE(next->prev, newsk);
1928 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1929 list->qlen++;
1930}
1da177e4 1931
67fed459
DM
1932static inline void __skb_queue_splice(const struct sk_buff_head *list,
1933 struct sk_buff *prev,
1934 struct sk_buff *next)
1935{
1936 struct sk_buff *first = list->next;
1937 struct sk_buff *last = list->prev;
1938
d7d16a89
ED
1939 WRITE_ONCE(first->prev, prev);
1940 WRITE_ONCE(prev->next, first);
67fed459 1941
d7d16a89
ED
1942 WRITE_ONCE(last->next, next);
1943 WRITE_ONCE(next->prev, last);
67fed459
DM
1944}
1945
1946/**
1947 * skb_queue_splice - join two skb lists, this is designed for stacks
1948 * @list: the new list to add
1949 * @head: the place to add it in the first list
1950 */
1951static inline void skb_queue_splice(const struct sk_buff_head *list,
1952 struct sk_buff_head *head)
1953{
1954 if (!skb_queue_empty(list)) {
1955 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1956 head->qlen += list->qlen;
67fed459
DM
1957 }
1958}
1959
1960/**
d9619496 1961 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1962 * @list: the new list to add
1963 * @head: the place to add it in the first list
1964 *
1965 * The list at @list is reinitialised
1966 */
1967static inline void skb_queue_splice_init(struct sk_buff_head *list,
1968 struct sk_buff_head *head)
1969{
1970 if (!skb_queue_empty(list)) {
1971 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1972 head->qlen += list->qlen;
67fed459
DM
1973 __skb_queue_head_init(list);
1974 }
1975}
1976
1977/**
1978 * skb_queue_splice_tail - join two skb lists, each list being a queue
1979 * @list: the new list to add
1980 * @head: the place to add it in the first list
1981 */
1982static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1983 struct sk_buff_head *head)
1984{
1985 if (!skb_queue_empty(list)) {
1986 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1987 head->qlen += list->qlen;
67fed459
DM
1988 }
1989}
1990
1991/**
d9619496 1992 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1993 * @list: the new list to add
1994 * @head: the place to add it in the first list
1995 *
1996 * Each of the lists is a queue.
1997 * The list at @list is reinitialised
1998 */
1999static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2000 struct sk_buff_head *head)
2001{
2002 if (!skb_queue_empty(list)) {
2003 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2004 head->qlen += list->qlen;
67fed459
DM
2005 __skb_queue_head_init(list);
2006 }
2007}
2008
1da177e4 2009/**
300ce174 2010 * __skb_queue_after - queue a buffer at the list head
1da177e4 2011 * @list: list to use
300ce174 2012 * @prev: place after this buffer
1da177e4
LT
2013 * @newsk: buffer to queue
2014 *
300ce174 2015 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2016 * and you must therefore hold required locks before calling it.
2017 *
2018 * A buffer cannot be placed on two lists at the same time.
2019 */
300ce174
SH
2020static inline void __skb_queue_after(struct sk_buff_head *list,
2021 struct sk_buff *prev,
2022 struct sk_buff *newsk)
1da177e4 2023{
bf299275 2024 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
2025}
2026
7965bd4d
JP
2027void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2028 struct sk_buff_head *list);
7de6c033 2029
f5572855
GR
2030static inline void __skb_queue_before(struct sk_buff_head *list,
2031 struct sk_buff *next,
2032 struct sk_buff *newsk)
2033{
2034 __skb_insert(newsk, next->prev, next, list);
2035}
2036
300ce174
SH
2037/**
2038 * __skb_queue_head - queue a buffer at the list head
2039 * @list: list to use
2040 * @newsk: buffer to queue
2041 *
2042 * Queue a buffer at the start of a list. This function takes no locks
2043 * and you must therefore hold required locks before calling it.
2044 *
2045 * A buffer cannot be placed on two lists at the same time.
2046 */
300ce174
SH
2047static inline void __skb_queue_head(struct sk_buff_head *list,
2048 struct sk_buff *newsk)
2049{
2050 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2051}
4ea7b0cf 2052void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2053
1da177e4
LT
2054/**
2055 * __skb_queue_tail - queue a buffer at the list tail
2056 * @list: list to use
2057 * @newsk: buffer to queue
2058 *
2059 * Queue a buffer at the end of a list. This function takes no locks
2060 * and you must therefore hold required locks before calling it.
2061 *
2062 * A buffer cannot be placed on two lists at the same time.
2063 */
1da177e4
LT
2064static inline void __skb_queue_tail(struct sk_buff_head *list,
2065 struct sk_buff *newsk)
2066{
f5572855 2067 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2068}
4ea7b0cf 2069void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2070
1da177e4
LT
2071/*
2072 * remove sk_buff from list. _Must_ be called atomically, and with
2073 * the list known..
2074 */
7965bd4d 2075void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2076static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2077{
2078 struct sk_buff *next, *prev;
2079
86b18aaa 2080 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2081 next = skb->next;
2082 prev = skb->prev;
2083 skb->next = skb->prev = NULL;
d7d16a89
ED
2084 WRITE_ONCE(next->prev, prev);
2085 WRITE_ONCE(prev->next, next);
1da177e4
LT
2086}
2087
f525c06d
GR
2088/**
2089 * __skb_dequeue - remove from the head of the queue
2090 * @list: list to dequeue from
2091 *
2092 * Remove the head of the list. This function does not take any locks
2093 * so must be used with appropriate locks held only. The head item is
2094 * returned or %NULL if the list is empty.
2095 */
f525c06d
GR
2096static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2097{
2098 struct sk_buff *skb = skb_peek(list);
2099 if (skb)
2100 __skb_unlink(skb, list);
2101 return skb;
2102}
4ea7b0cf 2103struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2104
2105/**
2106 * __skb_dequeue_tail - remove from the tail of the queue
2107 * @list: list to dequeue from
2108 *
2109 * Remove the tail of the list. This function does not take any locks
2110 * so must be used with appropriate locks held only. The tail item is
2111 * returned or %NULL if the list is empty.
2112 */
1da177e4
LT
2113static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2114{
2115 struct sk_buff *skb = skb_peek_tail(list);
2116 if (skb)
2117 __skb_unlink(skb, list);
2118 return skb;
2119}
4ea7b0cf 2120struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2121
2122
bdcc0924 2123static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2124{
2125 return skb->data_len;
2126}
2127
2128static inline unsigned int skb_headlen(const struct sk_buff *skb)
2129{
2130 return skb->len - skb->data_len;
2131}
2132
3ece7826 2133static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2134{
c72d8cda 2135 unsigned int i, len = 0;
1da177e4 2136
c72d8cda 2137 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2138 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2139 return len;
2140}
2141
2142static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2143{
2144 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2145}
2146
131ea667
IC
2147/**
2148 * __skb_fill_page_desc - initialise a paged fragment in an skb
2149 * @skb: buffer containing fragment to be initialised
2150 * @i: paged fragment index to initialise
2151 * @page: the page to use for this fragment
2152 * @off: the offset to the data with @page
2153 * @size: the length of the data
2154 *
2155 * Initialises the @i'th fragment of @skb to point to &size bytes at
2156 * offset @off within @page.
2157 *
2158 * Does not take any additional reference on the fragment.
2159 */
2160static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2161 struct page *page, int off, int size)
1da177e4
LT
2162{
2163 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2164
c48a11c7 2165 /*
2f064f34
MH
2166 * Propagate page pfmemalloc to the skb if we can. The problem is
2167 * that not all callers have unique ownership of the page but rely
2168 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2169 */
1dfa5bd3 2170 frag->bv_page = page;
65c84f14 2171 frag->bv_offset = off;
9e903e08 2172 skb_frag_size_set(frag, size);
cca7af38
PE
2173
2174 page = compound_head(page);
2f064f34 2175 if (page_is_pfmemalloc(page))
cca7af38 2176 skb->pfmemalloc = true;
131ea667
IC
2177}
2178
2179/**
2180 * skb_fill_page_desc - initialise a paged fragment in an skb
2181 * @skb: buffer containing fragment to be initialised
2182 * @i: paged fragment index to initialise
2183 * @page: the page to use for this fragment
2184 * @off: the offset to the data with @page
2185 * @size: the length of the data
2186 *
2187 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2188 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2189 * addition updates @skb such that @i is the last fragment.
2190 *
2191 * Does not take any additional reference on the fragment.
2192 */
2193static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2194 struct page *page, int off, int size)
2195{
2196 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2197 skb_shinfo(skb)->nr_frags = i + 1;
2198}
2199
7965bd4d
JP
2200void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2201 int size, unsigned int truesize);
654bed16 2202
f8e617e1
JW
2203void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2204 unsigned int truesize);
2205
1da177e4
LT
2206#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2207
27a884dc
ACM
2208#ifdef NET_SKBUFF_DATA_USES_OFFSET
2209static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2210{
2211 return skb->head + skb->tail;
2212}
2213
2214static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2215{
2216 skb->tail = skb->data - skb->head;
2217}
2218
2219static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2220{
2221 skb_reset_tail_pointer(skb);
2222 skb->tail += offset;
2223}
7cc46190 2224
27a884dc
ACM
2225#else /* NET_SKBUFF_DATA_USES_OFFSET */
2226static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2227{
2228 return skb->tail;
2229}
2230
2231static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2232{
2233 skb->tail = skb->data;
2234}
2235
2236static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2237{
2238 skb->tail = skb->data + offset;
2239}
4305b541 2240
27a884dc
ACM
2241#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2242
1da177e4
LT
2243/*
2244 * Add data to an sk_buff
2245 */
4df864c1
JB
2246void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2247void *skb_put(struct sk_buff *skb, unsigned int len);
2248static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2249{
4df864c1 2250 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2251 SKB_LINEAR_ASSERT(skb);
2252 skb->tail += len;
2253 skb->len += len;
2254 return tmp;
2255}
2256
de77b966 2257static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2258{
2259 void *tmp = __skb_put(skb, len);
2260
2261 memset(tmp, 0, len);
2262 return tmp;
2263}
2264
2265static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2266 unsigned int len)
2267{
2268 void *tmp = __skb_put(skb, len);
2269
2270 memcpy(tmp, data, len);
2271 return tmp;
2272}
2273
2274static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2275{
2276 *(u8 *)__skb_put(skb, 1) = val;
2277}
2278
83ad357d 2279static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2280{
83ad357d 2281 void *tmp = skb_put(skb, len);
e45a79da
JB
2282
2283 memset(tmp, 0, len);
2284
2285 return tmp;
2286}
2287
59ae1d12
JB
2288static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2289 unsigned int len)
2290{
2291 void *tmp = skb_put(skb, len);
2292
2293 memcpy(tmp, data, len);
2294
2295 return tmp;
2296}
2297
634fef61
JB
2298static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2299{
2300 *(u8 *)skb_put(skb, 1) = val;
2301}
2302
d58ff351
JB
2303void *skb_push(struct sk_buff *skb, unsigned int len);
2304static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2305{
2306 skb->data -= len;
2307 skb->len += len;
2308 return skb->data;
2309}
2310
af72868b
JB
2311void *skb_pull(struct sk_buff *skb, unsigned int len);
2312static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2313{
2314 skb->len -= len;
2315 BUG_ON(skb->len < skb->data_len);
2316 return skb->data += len;
2317}
2318
af72868b 2319static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2320{
2321 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2322}
2323
af72868b 2324void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2325
af72868b 2326static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2327{
2328 if (len > skb_headlen(skb) &&
987c402a 2329 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2330 return NULL;
2331 skb->len -= len;
2332 return skb->data += len;
2333}
2334
af72868b 2335static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2336{
2337 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2338}
2339
b9df4fd7 2340static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2341{
2342 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2343 return true;
1da177e4 2344 if (unlikely(len > skb->len))
b9df4fd7 2345 return false;
987c402a 2346 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2347}
2348
c8c8b127
ED
2349void skb_condense(struct sk_buff *skb);
2350
1da177e4
LT
2351/**
2352 * skb_headroom - bytes at buffer head
2353 * @skb: buffer to check
2354 *
2355 * Return the number of bytes of free space at the head of an &sk_buff.
2356 */
c2636b4d 2357static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2358{
2359 return skb->data - skb->head;
2360}
2361
2362/**
2363 * skb_tailroom - bytes at buffer end
2364 * @skb: buffer to check
2365 *
2366 * Return the number of bytes of free space at the tail of an sk_buff
2367 */
2368static inline int skb_tailroom(const struct sk_buff *skb)
2369{
4305b541 2370 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2371}
2372
a21d4572
ED
2373/**
2374 * skb_availroom - bytes at buffer end
2375 * @skb: buffer to check
2376 *
2377 * Return the number of bytes of free space at the tail of an sk_buff
2378 * allocated by sk_stream_alloc()
2379 */
2380static inline int skb_availroom(const struct sk_buff *skb)
2381{
16fad69c
ED
2382 if (skb_is_nonlinear(skb))
2383 return 0;
2384
2385 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2386}
2387
1da177e4
LT
2388/**
2389 * skb_reserve - adjust headroom
2390 * @skb: buffer to alter
2391 * @len: bytes to move
2392 *
2393 * Increase the headroom of an empty &sk_buff by reducing the tail
2394 * room. This is only allowed for an empty buffer.
2395 */
8243126c 2396static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2397{
2398 skb->data += len;
2399 skb->tail += len;
2400}
2401
1837b2e2
BP
2402/**
2403 * skb_tailroom_reserve - adjust reserved_tailroom
2404 * @skb: buffer to alter
2405 * @mtu: maximum amount of headlen permitted
2406 * @needed_tailroom: minimum amount of reserved_tailroom
2407 *
2408 * Set reserved_tailroom so that headlen can be as large as possible but
2409 * not larger than mtu and tailroom cannot be smaller than
2410 * needed_tailroom.
2411 * The required headroom should already have been reserved before using
2412 * this function.
2413 */
2414static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2415 unsigned int needed_tailroom)
2416{
2417 SKB_LINEAR_ASSERT(skb);
2418 if (mtu < skb_tailroom(skb) - needed_tailroom)
2419 /* use at most mtu */
2420 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2421 else
2422 /* use up to all available space */
2423 skb->reserved_tailroom = needed_tailroom;
2424}
2425
8bce6d7d
TH
2426#define ENCAP_TYPE_ETHER 0
2427#define ENCAP_TYPE_IPPROTO 1
2428
2429static inline void skb_set_inner_protocol(struct sk_buff *skb,
2430 __be16 protocol)
2431{
2432 skb->inner_protocol = protocol;
2433 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2434}
2435
2436static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2437 __u8 ipproto)
2438{
2439 skb->inner_ipproto = ipproto;
2440 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2441}
2442
6a674e9c
JG
2443static inline void skb_reset_inner_headers(struct sk_buff *skb)
2444{
aefbd2b3 2445 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2446 skb->inner_network_header = skb->network_header;
2447 skb->inner_transport_header = skb->transport_header;
2448}
2449
0b5c9db1
JP
2450static inline void skb_reset_mac_len(struct sk_buff *skb)
2451{
2452 skb->mac_len = skb->network_header - skb->mac_header;
2453}
2454
6a674e9c
JG
2455static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2456 *skb)
2457{
2458 return skb->head + skb->inner_transport_header;
2459}
2460
55dc5a9f
TH
2461static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2462{
2463 return skb_inner_transport_header(skb) - skb->data;
2464}
2465
6a674e9c
JG
2466static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2467{
2468 skb->inner_transport_header = skb->data - skb->head;
2469}
2470
2471static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2472 const int offset)
2473{
2474 skb_reset_inner_transport_header(skb);
2475 skb->inner_transport_header += offset;
2476}
2477
2478static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2479{
2480 return skb->head + skb->inner_network_header;
2481}
2482
2483static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2484{
2485 skb->inner_network_header = skb->data - skb->head;
2486}
2487
2488static inline void skb_set_inner_network_header(struct sk_buff *skb,
2489 const int offset)
2490{
2491 skb_reset_inner_network_header(skb);
2492 skb->inner_network_header += offset;
2493}
2494
aefbd2b3
PS
2495static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2496{
2497 return skb->head + skb->inner_mac_header;
2498}
2499
2500static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2501{
2502 skb->inner_mac_header = skb->data - skb->head;
2503}
2504
2505static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2506 const int offset)
2507{
2508 skb_reset_inner_mac_header(skb);
2509 skb->inner_mac_header += offset;
2510}
fda55eca
ED
2511static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2512{
35d04610 2513 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2514}
2515
9c70220b
ACM
2516static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2517{
2e07fa9c 2518 return skb->head + skb->transport_header;
9c70220b
ACM
2519}
2520
badff6d0
ACM
2521static inline void skb_reset_transport_header(struct sk_buff *skb)
2522{
2e07fa9c 2523 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2524}
2525
967b05f6
ACM
2526static inline void skb_set_transport_header(struct sk_buff *skb,
2527 const int offset)
2528{
2e07fa9c
ACM
2529 skb_reset_transport_header(skb);
2530 skb->transport_header += offset;
ea2ae17d
ACM
2531}
2532
d56f90a7
ACM
2533static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2534{
2e07fa9c 2535 return skb->head + skb->network_header;
d56f90a7
ACM
2536}
2537
c1d2bbe1
ACM
2538static inline void skb_reset_network_header(struct sk_buff *skb)
2539{
2e07fa9c 2540 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2541}
2542
c14d2450
ACM
2543static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2544{
2e07fa9c
ACM
2545 skb_reset_network_header(skb);
2546 skb->network_header += offset;
c14d2450
ACM
2547}
2548
2e07fa9c 2549static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2550{
2e07fa9c 2551 return skb->head + skb->mac_header;
bbe735e4
ACM
2552}
2553
ea6da4fd
AV
2554static inline int skb_mac_offset(const struct sk_buff *skb)
2555{
2556 return skb_mac_header(skb) - skb->data;
2557}
2558
0daf4349
DB
2559static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2560{
2561 return skb->network_header - skb->mac_header;
2562}
2563
2e07fa9c 2564static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2565{
35d04610 2566 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2567}
2568
b4ab3141
DB
2569static inline void skb_unset_mac_header(struct sk_buff *skb)
2570{
2571 skb->mac_header = (typeof(skb->mac_header))~0U;
2572}
2573
2e07fa9c
ACM
2574static inline void skb_reset_mac_header(struct sk_buff *skb)
2575{
2576 skb->mac_header = skb->data - skb->head;
2577}
2578
2579static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2580{
2581 skb_reset_mac_header(skb);
2582 skb->mac_header += offset;
2583}
2584
0e3da5bb
TT
2585static inline void skb_pop_mac_header(struct sk_buff *skb)
2586{
2587 skb->mac_header = skb->network_header;
2588}
2589
d2aa125d 2590static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2591{
72a338bc 2592 struct flow_keys_basic keys;
fbbdb8f0
YX
2593
2594 if (skb_transport_header_was_set(skb))
2595 return;
72a338bc 2596
3cbf4ffb
SF
2597 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2598 NULL, 0, 0, 0, 0))
42aecaa9 2599 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2600}
2601
03606895
ED
2602static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2603{
2604 if (skb_mac_header_was_set(skb)) {
2605 const unsigned char *old_mac = skb_mac_header(skb);
2606
2607 skb_set_mac_header(skb, -skb->mac_len);
2608 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2609 }
2610}
2611
04fb451e
MM
2612static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2613{
2614 return skb->csum_start - skb_headroom(skb);
2615}
2616
08b64fcc
AD
2617static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2618{
2619 return skb->head + skb->csum_start;
2620}
2621
2e07fa9c
ACM
2622static inline int skb_transport_offset(const struct sk_buff *skb)
2623{
2624 return skb_transport_header(skb) - skb->data;
2625}
2626
2627static inline u32 skb_network_header_len(const struct sk_buff *skb)
2628{
2629 return skb->transport_header - skb->network_header;
2630}
2631
6a674e9c
JG
2632static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2633{
2634 return skb->inner_transport_header - skb->inner_network_header;
2635}
2636
2e07fa9c
ACM
2637static inline int skb_network_offset(const struct sk_buff *skb)
2638{
2639 return skb_network_header(skb) - skb->data;
2640}
48d49d0c 2641
6a674e9c
JG
2642static inline int skb_inner_network_offset(const struct sk_buff *skb)
2643{
2644 return skb_inner_network_header(skb) - skb->data;
2645}
2646
f9599ce1
CG
2647static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2648{
2649 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2650}
2651
1da177e4
LT
2652/*
2653 * CPUs often take a performance hit when accessing unaligned memory
2654 * locations. The actual performance hit varies, it can be small if the
2655 * hardware handles it or large if we have to take an exception and fix it
2656 * in software.
2657 *
2658 * Since an ethernet header is 14 bytes network drivers often end up with
2659 * the IP header at an unaligned offset. The IP header can be aligned by
2660 * shifting the start of the packet by 2 bytes. Drivers should do this
2661 * with:
2662 *
8660c124 2663 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2664 *
2665 * The downside to this alignment of the IP header is that the DMA is now
2666 * unaligned. On some architectures the cost of an unaligned DMA is high
2667 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2668 *
1da177e4
LT
2669 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2670 * to be overridden.
2671 */
2672#ifndef NET_IP_ALIGN
2673#define NET_IP_ALIGN 2
2674#endif
2675
025be81e
AB
2676/*
2677 * The networking layer reserves some headroom in skb data (via
2678 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2679 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2680 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2681 *
2682 * Unfortunately this headroom changes the DMA alignment of the resulting
2683 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2684 * on some architectures. An architecture can override this value,
2685 * perhaps setting it to a cacheline in size (since that will maintain
2686 * cacheline alignment of the DMA). It must be a power of 2.
2687 *
d6301d3d 2688 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2689 * headroom, you should not reduce this.
5933dd2f
ED
2690 *
2691 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2692 * to reduce average number of cache lines per packet.
645f0897 2693 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2694 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2695 */
2696#ifndef NET_SKB_PAD
5933dd2f 2697#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2698#endif
2699
7965bd4d 2700int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2701
5293efe6 2702static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2703{
5e1abdc3 2704 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2705 return;
27a884dc
ACM
2706 skb->len = len;
2707 skb_set_tail_pointer(skb, len);
1da177e4
LT
2708}
2709
5293efe6
DB
2710static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2711{
2712 __skb_set_length(skb, len);
2713}
2714
7965bd4d 2715void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2716
2717static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2718{
3cc0e873
HX
2719 if (skb->data_len)
2720 return ___pskb_trim(skb, len);
2721 __skb_trim(skb, len);
2722 return 0;
1da177e4
LT
2723}
2724
2725static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2726{
2727 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2728}
2729
e9fa4f7b
HX
2730/**
2731 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2732 * @skb: buffer to alter
2733 * @len: new length
2734 *
2735 * This is identical to pskb_trim except that the caller knows that
2736 * the skb is not cloned so we should never get an error due to out-
2737 * of-memory.
2738 */
2739static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2740{
2741 int err = pskb_trim(skb, len);
2742 BUG_ON(err);
2743}
2744
5293efe6
DB
2745static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2746{
2747 unsigned int diff = len - skb->len;
2748
2749 if (skb_tailroom(skb) < diff) {
2750 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2751 GFP_ATOMIC);
2752 if (ret)
2753 return ret;
2754 }
2755 __skb_set_length(skb, len);
2756 return 0;
2757}
2758
1da177e4
LT
2759/**
2760 * skb_orphan - orphan a buffer
2761 * @skb: buffer to orphan
2762 *
2763 * If a buffer currently has an owner then we call the owner's
2764 * destructor function and make the @skb unowned. The buffer continues
2765 * to exist but is no longer charged to its former owner.
2766 */
2767static inline void skb_orphan(struct sk_buff *skb)
2768{
c34a7612 2769 if (skb->destructor) {
1da177e4 2770 skb->destructor(skb);
c34a7612
ED
2771 skb->destructor = NULL;
2772 skb->sk = NULL;
376c7311
ED
2773 } else {
2774 BUG_ON(skb->sk);
c34a7612 2775 }
1da177e4
LT
2776}
2777
a353e0ce
MT
2778/**
2779 * skb_orphan_frags - orphan the frags contained in a buffer
2780 * @skb: buffer to orphan frags from
2781 * @gfp_mask: allocation mask for replacement pages
2782 *
2783 * For each frag in the SKB which needs a destructor (i.e. has an
2784 * owner) create a copy of that frag and release the original
2785 * page by calling the destructor.
2786 */
2787static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2788{
1f8b977a
WB
2789 if (likely(!skb_zcopy(skb)))
2790 return 0;
185ce5c3 2791 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 2792 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
2793 return 0;
2794 return skb_copy_ubufs(skb, gfp_mask);
2795}
2796
2797/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2798static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2799{
2800 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2801 return 0;
2802 return skb_copy_ubufs(skb, gfp_mask);
2803}
2804
1da177e4
LT
2805/**
2806 * __skb_queue_purge - empty a list
2807 * @list: list to empty
2808 *
2809 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2810 * the list and one reference dropped. This function does not take the
2811 * list lock and the caller must hold the relevant locks to use it.
2812 */
1da177e4
LT
2813static inline void __skb_queue_purge(struct sk_buff_head *list)
2814{
2815 struct sk_buff *skb;
2816 while ((skb = __skb_dequeue(list)) != NULL)
2817 kfree_skb(skb);
2818}
4ea7b0cf 2819void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2820
385114de 2821unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2822
3f6e687d
KH
2823void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2824
2825/**
2826 * netdev_alloc_frag - allocate a page fragment
2827 * @fragsz: fragment size
2828 *
2829 * Allocates a frag from a page for receive buffer.
2830 * Uses GFP_ATOMIC allocations.
2831 */
2832static inline void *netdev_alloc_frag(unsigned int fragsz)
2833{
2834 return __netdev_alloc_frag_align(fragsz, ~0u);
2835}
2836
2837static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2838 unsigned int align)
2839{
2840 WARN_ON_ONCE(!is_power_of_2(align));
2841 return __netdev_alloc_frag_align(fragsz, -align);
2842}
1da177e4 2843
7965bd4d
JP
2844struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2845 gfp_t gfp_mask);
8af27456
CH
2846
2847/**
2848 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2849 * @dev: network device to receive on
2850 * @length: length to allocate
2851 *
2852 * Allocate a new &sk_buff and assign it a usage count of one. The
2853 * buffer has unspecified headroom built in. Users should allocate
2854 * the headroom they think they need without accounting for the
2855 * built in space. The built in space is used for optimisations.
2856 *
2857 * %NULL is returned if there is no free memory. Although this function
2858 * allocates memory it can be called from an interrupt.
2859 */
2860static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2861 unsigned int length)
8af27456
CH
2862{
2863 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2864}
2865
6f532612
ED
2866/* legacy helper around __netdev_alloc_skb() */
2867static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2868 gfp_t gfp_mask)
2869{
2870 return __netdev_alloc_skb(NULL, length, gfp_mask);
2871}
2872
2873/* legacy helper around netdev_alloc_skb() */
2874static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2875{
2876 return netdev_alloc_skb(NULL, length);
2877}
2878
2879
4915a0de
ED
2880static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2881 unsigned int length, gfp_t gfp)
61321bbd 2882{
4915a0de 2883 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2884
2885 if (NET_IP_ALIGN && skb)
2886 skb_reserve(skb, NET_IP_ALIGN);
2887 return skb;
2888}
2889
4915a0de
ED
2890static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2891 unsigned int length)
2892{
2893 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2894}
2895
181edb2b
AD
2896static inline void skb_free_frag(void *addr)
2897{
8c2dd3e4 2898 page_frag_free(addr);
181edb2b
AD
2899}
2900
3f6e687d
KH
2901void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2902
2903static inline void *napi_alloc_frag(unsigned int fragsz)
2904{
2905 return __napi_alloc_frag_align(fragsz, ~0u);
2906}
2907
2908static inline void *napi_alloc_frag_align(unsigned int fragsz,
2909 unsigned int align)
2910{
2911 WARN_ON_ONCE(!is_power_of_2(align));
2912 return __napi_alloc_frag_align(fragsz, -align);
2913}
2914
fd11a83d
AD
2915struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2916 unsigned int length, gfp_t gfp_mask);
2917static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2918 unsigned int length)
2919{
2920 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2921}
795bb1c0
JDB
2922void napi_consume_skb(struct sk_buff *skb, int budget);
2923
9243adfc 2924void napi_skb_free_stolen_head(struct sk_buff *skb);
15fad714 2925void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2926
71dfda58
AD
2927/**
2928 * __dev_alloc_pages - allocate page for network Rx
2929 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2930 * @order: size of the allocation
2931 *
2932 * Allocate a new page.
2933 *
2934 * %NULL is returned if there is no free memory.
2935*/
2936static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2937 unsigned int order)
2938{
2939 /* This piece of code contains several assumptions.
2940 * 1. This is for device Rx, therefor a cold page is preferred.
2941 * 2. The expectation is the user wants a compound page.
2942 * 3. If requesting a order 0 page it will not be compound
2943 * due to the check to see if order has a value in prep_new_page
2944 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2945 * code in gfp_to_alloc_flags that should be enforcing this.
2946 */
453f85d4 2947 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2948
2949 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2950}
2951
2952static inline struct page *dev_alloc_pages(unsigned int order)
2953{
95829b3a 2954 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2955}
2956
2957/**
2958 * __dev_alloc_page - allocate a page for network Rx
2959 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2960 *
2961 * Allocate a new page.
2962 *
2963 * %NULL is returned if there is no free memory.
2964 */
2965static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2966{
2967 return __dev_alloc_pages(gfp_mask, 0);
2968}
2969
2970static inline struct page *dev_alloc_page(void)
2971{
95829b3a 2972 return dev_alloc_pages(0);
71dfda58
AD
2973}
2974
bc38f30f
AL
2975/**
2976 * dev_page_is_reusable - check whether a page can be reused for network Rx
2977 * @page: the page to test
2978 *
2979 * A page shouldn't be considered for reusing/recycling if it was allocated
2980 * under memory pressure or at a distant memory node.
2981 *
2982 * Returns false if this page should be returned to page allocator, true
2983 * otherwise.
2984 */
2985static inline bool dev_page_is_reusable(const struct page *page)
2986{
2987 return likely(page_to_nid(page) == numa_mem_id() &&
2988 !page_is_pfmemalloc(page));
2989}
2990
0614002b
MG
2991/**
2992 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2993 * @page: The page that was allocated from skb_alloc_page
2994 * @skb: The skb that may need pfmemalloc set
2995 */
48f971c9
AL
2996static inline void skb_propagate_pfmemalloc(const struct page *page,
2997 struct sk_buff *skb)
0614002b 2998{
2f064f34 2999 if (page_is_pfmemalloc(page))
0614002b
MG
3000 skb->pfmemalloc = true;
3001}
3002
7240b60c
JL
3003/**
3004 * skb_frag_off() - Returns the offset of a skb fragment
3005 * @frag: the paged fragment
3006 */
3007static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3008{
65c84f14 3009 return frag->bv_offset;
7240b60c
JL
3010}
3011
3012/**
3013 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3014 * @frag: skb fragment
3015 * @delta: value to add
3016 */
3017static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3018{
65c84f14 3019 frag->bv_offset += delta;
7240b60c
JL
3020}
3021
3022/**
3023 * skb_frag_off_set() - Sets the offset of a skb fragment
3024 * @frag: skb fragment
3025 * @offset: offset of fragment
3026 */
3027static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3028{
65c84f14 3029 frag->bv_offset = offset;
7240b60c
JL
3030}
3031
3032/**
3033 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3034 * @fragto: skb fragment where offset is set
3035 * @fragfrom: skb fragment offset is copied from
3036 */
3037static inline void skb_frag_off_copy(skb_frag_t *fragto,
3038 const skb_frag_t *fragfrom)
3039{
65c84f14 3040 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
3041}
3042
131ea667 3043/**
e227867f 3044 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
3045 * @frag: the paged fragment
3046 *
3047 * Returns the &struct page associated with @frag.
3048 */
3049static inline struct page *skb_frag_page(const skb_frag_t *frag)
3050{
1dfa5bd3 3051 return frag->bv_page;
131ea667
IC
3052}
3053
3054/**
3055 * __skb_frag_ref - take an addition reference on a paged fragment.
3056 * @frag: the paged fragment
3057 *
3058 * Takes an additional reference on the paged fragment @frag.
3059 */
3060static inline void __skb_frag_ref(skb_frag_t *frag)
3061{
3062 get_page(skb_frag_page(frag));
3063}
3064
3065/**
3066 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3067 * @skb: the buffer
3068 * @f: the fragment offset.
3069 *
3070 * Takes an additional reference on the @f'th paged fragment of @skb.
3071 */
3072static inline void skb_frag_ref(struct sk_buff *skb, int f)
3073{
3074 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3075}
3076
3077/**
3078 * __skb_frag_unref - release a reference on a paged fragment.
3079 * @frag: the paged fragment
3080 *
3081 * Releases a reference on the paged fragment @frag.
3082 */
3083static inline void __skb_frag_unref(skb_frag_t *frag)
3084{
3085 put_page(skb_frag_page(frag));
3086}
3087
3088/**
3089 * skb_frag_unref - release a reference on a paged fragment of an skb.
3090 * @skb: the buffer
3091 * @f: the fragment offset
3092 *
3093 * Releases a reference on the @f'th paged fragment of @skb.
3094 */
3095static inline void skb_frag_unref(struct sk_buff *skb, int f)
3096{
3097 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3098}
3099
3100/**
3101 * skb_frag_address - gets the address of the data contained in a paged fragment
3102 * @frag: the paged fragment buffer
3103 *
3104 * Returns the address of the data within @frag. The page must already
3105 * be mapped.
3106 */
3107static inline void *skb_frag_address(const skb_frag_t *frag)
3108{
7240b60c 3109 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3110}
3111
3112/**
3113 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3114 * @frag: the paged fragment buffer
3115 *
3116 * Returns the address of the data within @frag. Checks that the page
3117 * is mapped and returns %NULL otherwise.
3118 */
3119static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3120{
3121 void *ptr = page_address(skb_frag_page(frag));
3122 if (unlikely(!ptr))
3123 return NULL;
3124
7240b60c
JL
3125 return ptr + skb_frag_off(frag);
3126}
3127
3128/**
3129 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3130 * @fragto: skb fragment where page is set
3131 * @fragfrom: skb fragment page is copied from
3132 */
3133static inline void skb_frag_page_copy(skb_frag_t *fragto,
3134 const skb_frag_t *fragfrom)
3135{
3136 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3137}
3138
3139/**
3140 * __skb_frag_set_page - sets the page contained in a paged fragment
3141 * @frag: the paged fragment
3142 * @page: the page to set
3143 *
3144 * Sets the fragment @frag to contain @page.
3145 */
3146static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3147{
1dfa5bd3 3148 frag->bv_page = page;
131ea667
IC
3149}
3150
3151/**
3152 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3153 * @skb: the buffer
3154 * @f: the fragment offset
3155 * @page: the page to set
3156 *
3157 * Sets the @f'th fragment of @skb to contain @page.
3158 */
3159static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3160 struct page *page)
3161{
3162 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3163}
3164
400dfd3a
ED
3165bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3166
131ea667
IC
3167/**
3168 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3169 * @dev: the device to map the fragment to
131ea667
IC
3170 * @frag: the paged fragment to map
3171 * @offset: the offset within the fragment (starting at the
3172 * fragment's own offset)
3173 * @size: the number of bytes to map
771b00a8 3174 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3175 *
3176 * Maps the page associated with @frag to @device.
3177 */
3178static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3179 const skb_frag_t *frag,
3180 size_t offset, size_t size,
3181 enum dma_data_direction dir)
3182{
3183 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3184 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3185}
3186
117632e6
ED
3187static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3188 gfp_t gfp_mask)
3189{
3190 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3191}
3192
bad93e9d
OP
3193
3194static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3195 gfp_t gfp_mask)
3196{
3197 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3198}
3199
3200
334a8132
PM
3201/**
3202 * skb_clone_writable - is the header of a clone writable
3203 * @skb: buffer to check
3204 * @len: length up to which to write
3205 *
3206 * Returns true if modifying the header part of the cloned buffer
3207 * does not requires the data to be copied.
3208 */
05bdd2f1 3209static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3210{
3211 return !skb_header_cloned(skb) &&
3212 skb_headroom(skb) + len <= skb->hdr_len;
3213}
3214
3697649f
DB
3215static inline int skb_try_make_writable(struct sk_buff *skb,
3216 unsigned int write_len)
3217{
3218 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3219 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3220}
3221
d9cc2048
HX
3222static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3223 int cloned)
3224{
3225 int delta = 0;
3226
d9cc2048
HX
3227 if (headroom > skb_headroom(skb))
3228 delta = headroom - skb_headroom(skb);
3229
3230 if (delta || cloned)
3231 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3232 GFP_ATOMIC);
3233 return 0;
3234}
3235
1da177e4
LT
3236/**
3237 * skb_cow - copy header of skb when it is required
3238 * @skb: buffer to cow
3239 * @headroom: needed headroom
3240 *
3241 * If the skb passed lacks sufficient headroom or its data part
3242 * is shared, data is reallocated. If reallocation fails, an error
3243 * is returned and original skb is not changed.
3244 *
3245 * The result is skb with writable area skb->head...skb->tail
3246 * and at least @headroom of space at head.
3247 */
3248static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3249{
d9cc2048
HX
3250 return __skb_cow(skb, headroom, skb_cloned(skb));
3251}
1da177e4 3252
d9cc2048
HX
3253/**
3254 * skb_cow_head - skb_cow but only making the head writable
3255 * @skb: buffer to cow
3256 * @headroom: needed headroom
3257 *
3258 * This function is identical to skb_cow except that we replace the
3259 * skb_cloned check by skb_header_cloned. It should be used when
3260 * you only need to push on some header and do not need to modify
3261 * the data.
3262 */
3263static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3264{
3265 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3266}
3267
3268/**
3269 * skb_padto - pad an skbuff up to a minimal size
3270 * @skb: buffer to pad
3271 * @len: minimal length
3272 *
3273 * Pads up a buffer to ensure the trailing bytes exist and are
3274 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3275 * is untouched. Otherwise it is extended. Returns zero on
3276 * success. The skb is freed on error.
1da177e4 3277 */
5b057c6b 3278static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3279{
3280 unsigned int size = skb->len;
3281 if (likely(size >= len))
5b057c6b 3282 return 0;
987c402a 3283 return skb_pad(skb, len - size);
1da177e4
LT
3284}
3285
9c0c1124 3286/**
4ea7b0cf 3287 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3288 * @skb: buffer to pad
3289 * @len: minimal length
cd0a137a 3290 * @free_on_error: free buffer on error
9c0c1124
AD
3291 *
3292 * Pads up a buffer to ensure the trailing bytes exist and are
3293 * blanked. If the buffer already contains sufficient data it
3294 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3295 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3296 */
4a009cb0
ED
3297static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3298 unsigned int len,
3299 bool free_on_error)
9c0c1124
AD
3300{
3301 unsigned int size = skb->len;
3302
3303 if (unlikely(size < len)) {
3304 len -= size;
cd0a137a 3305 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3306 return -ENOMEM;
3307 __skb_put(skb, len);
3308 }
3309 return 0;
3310}
3311
cd0a137a
FF
3312/**
3313 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3314 * @skb: buffer to pad
3315 * @len: minimal length
3316 *
3317 * Pads up a buffer to ensure the trailing bytes exist and are
3318 * blanked. If the buffer already contains sufficient data it
3319 * is untouched. Otherwise it is extended. Returns zero on
3320 * success. The skb is freed on error.
3321 */
4a009cb0 3322static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3323{
3324 return __skb_put_padto(skb, len, true);
3325}
3326
1da177e4 3327static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3328 struct iov_iter *from, int copy)
1da177e4
LT
3329{
3330 const int off = skb->len;
3331
3332 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3333 __wsum csum = 0;
15e6cb46
AV
3334 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3335 &csum, from)) {
1da177e4
LT
3336 skb->csum = csum_block_add(skb->csum, csum, off);
3337 return 0;
3338 }
15e6cb46 3339 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3340 return 0;
3341
3342 __skb_trim(skb, off);
3343 return -EFAULT;
3344}
3345
38ba0a65
ED
3346static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3347 const struct page *page, int off)
1da177e4 3348{
1f8b977a
WB
3349 if (skb_zcopy(skb))
3350 return false;
1da177e4 3351 if (i) {
d8e18a51 3352 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3353
ea2ab693 3354 return page == skb_frag_page(frag) &&
7240b60c 3355 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3356 }
38ba0a65 3357 return false;
1da177e4
LT
3358}
3359
364c6bad
HX
3360static inline int __skb_linearize(struct sk_buff *skb)
3361{
3362 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3363}
3364
1da177e4
LT
3365/**
3366 * skb_linearize - convert paged skb to linear one
3367 * @skb: buffer to linarize
1da177e4
LT
3368 *
3369 * If there is no free memory -ENOMEM is returned, otherwise zero
3370 * is returned and the old skb data released.
3371 */
364c6bad
HX
3372static inline int skb_linearize(struct sk_buff *skb)
3373{
3374 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3375}
3376
cef401de
ED
3377/**
3378 * skb_has_shared_frag - can any frag be overwritten
3379 * @skb: buffer to test
3380 *
3381 * Return true if the skb has at least one frag that might be modified
3382 * by an external entity (as in vmsplice()/sendfile())
3383 */
3384static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3385{
c9af6db4 3386 return skb_is_nonlinear(skb) &&
06b4feb3 3387 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3388}
3389
364c6bad
HX
3390/**
3391 * skb_linearize_cow - make sure skb is linear and writable
3392 * @skb: buffer to process
3393 *
3394 * If there is no free memory -ENOMEM is returned, otherwise zero
3395 * is returned and the old skb data released.
3396 */
3397static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3398{
364c6bad
HX
3399 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3400 __skb_linearize(skb) : 0;
1da177e4
LT
3401}
3402
479ffccc
DB
3403static __always_inline void
3404__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3405 unsigned int off)
3406{
3407 if (skb->ip_summed == CHECKSUM_COMPLETE)
3408 skb->csum = csum_block_sub(skb->csum,
3409 csum_partial(start, len, 0), off);
3410 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3411 skb_checksum_start_offset(skb) < 0)
3412 skb->ip_summed = CHECKSUM_NONE;
3413}
3414
1da177e4
LT
3415/**
3416 * skb_postpull_rcsum - update checksum for received skb after pull
3417 * @skb: buffer to update
3418 * @start: start of data before pull
3419 * @len: length of data pulled
3420 *
3421 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3422 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3423 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3424 */
1da177e4 3425static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3426 const void *start, unsigned int len)
1da177e4 3427{
479ffccc 3428 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3429}
3430
479ffccc
DB
3431static __always_inline void
3432__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3433 unsigned int off)
3434{
3435 if (skb->ip_summed == CHECKSUM_COMPLETE)
3436 skb->csum = csum_block_add(skb->csum,
3437 csum_partial(start, len, 0), off);
3438}
cbb042f9 3439
479ffccc
DB
3440/**
3441 * skb_postpush_rcsum - update checksum for received skb after push
3442 * @skb: buffer to update
3443 * @start: start of data after push
3444 * @len: length of data pushed
3445 *
3446 * After doing a push on a received packet, you need to call this to
3447 * update the CHECKSUM_COMPLETE checksum.
3448 */
f8ffad69
DB
3449static inline void skb_postpush_rcsum(struct sk_buff *skb,
3450 const void *start, unsigned int len)
3451{
479ffccc 3452 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3453}
3454
af72868b 3455void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3456
82a31b92
WC
3457/**
3458 * skb_push_rcsum - push skb and update receive checksum
3459 * @skb: buffer to update
3460 * @len: length of data pulled
3461 *
3462 * This function performs an skb_push on the packet and updates
3463 * the CHECKSUM_COMPLETE checksum. It should be used on
3464 * receive path processing instead of skb_push unless you know
3465 * that the checksum difference is zero (e.g., a valid IP header)
3466 * or you are setting ip_summed to CHECKSUM_NONE.
3467 */
d58ff351 3468static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3469{
3470 skb_push(skb, len);
3471 skb_postpush_rcsum(skb, skb->data, len);
3472 return skb->data;
3473}
3474
88078d98 3475int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3476/**
3477 * pskb_trim_rcsum - trim received skb and update checksum
3478 * @skb: buffer to trim
3479 * @len: new length
3480 *
3481 * This is exactly the same as pskb_trim except that it ensures the
3482 * checksum of received packets are still valid after the operation.
6c57f045 3483 * It can change skb pointers.
7ce5a27f
DM
3484 */
3485
3486static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3487{
3488 if (likely(len >= skb->len))
3489 return 0;
88078d98 3490 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3491}
3492
5293efe6
DB
3493static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3494{
3495 if (skb->ip_summed == CHECKSUM_COMPLETE)
3496 skb->ip_summed = CHECKSUM_NONE;
3497 __skb_trim(skb, len);
3498 return 0;
3499}
3500
3501static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3502{
3503 if (skb->ip_summed == CHECKSUM_COMPLETE)
3504 skb->ip_summed = CHECKSUM_NONE;
3505 return __skb_grow(skb, len);
3506}
3507
18a4c0ea
ED
3508#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3509#define skb_rb_first(root) rb_to_skb(rb_first(root))
3510#define skb_rb_last(root) rb_to_skb(rb_last(root))
3511#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3512#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3513
1da177e4
LT
3514#define skb_queue_walk(queue, skb) \
3515 for (skb = (queue)->next; \
a1e4891f 3516 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3517 skb = skb->next)
3518
46f8914e
JC
3519#define skb_queue_walk_safe(queue, skb, tmp) \
3520 for (skb = (queue)->next, tmp = skb->next; \
3521 skb != (struct sk_buff *)(queue); \
3522 skb = tmp, tmp = skb->next)
3523
1164f52a 3524#define skb_queue_walk_from(queue, skb) \
a1e4891f 3525 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3526 skb = skb->next)
3527
18a4c0ea
ED
3528#define skb_rbtree_walk(skb, root) \
3529 for (skb = skb_rb_first(root); skb != NULL; \
3530 skb = skb_rb_next(skb))
3531
3532#define skb_rbtree_walk_from(skb) \
3533 for (; skb != NULL; \
3534 skb = skb_rb_next(skb))
3535
3536#define skb_rbtree_walk_from_safe(skb, tmp) \
3537 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3538 skb = tmp)
3539
1164f52a
DM
3540#define skb_queue_walk_from_safe(queue, skb, tmp) \
3541 for (tmp = skb->next; \
3542 skb != (struct sk_buff *)(queue); \
3543 skb = tmp, tmp = skb->next)
3544
300ce174
SH
3545#define skb_queue_reverse_walk(queue, skb) \
3546 for (skb = (queue)->prev; \
a1e4891f 3547 skb != (struct sk_buff *)(queue); \
300ce174
SH
3548 skb = skb->prev)
3549
686a2955
DM
3550#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3551 for (skb = (queue)->prev, tmp = skb->prev; \
3552 skb != (struct sk_buff *)(queue); \
3553 skb = tmp, tmp = skb->prev)
3554
3555#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3556 for (tmp = skb->prev; \
3557 skb != (struct sk_buff *)(queue); \
3558 skb = tmp, tmp = skb->prev)
1da177e4 3559
21dc3301 3560static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3561{
3562 return skb_shinfo(skb)->frag_list != NULL;
3563}
3564
3565static inline void skb_frag_list_init(struct sk_buff *skb)
3566{
3567 skb_shinfo(skb)->frag_list = NULL;
3568}
3569
ee039871
DM
3570#define skb_walk_frags(skb, iter) \
3571 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3572
ea3793ee 3573
b50b0580
SD
3574int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3575 int *err, long *timeo_p,
ea3793ee 3576 const struct sk_buff *skb);
65101aec
PA
3577struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3578 struct sk_buff_head *queue,
3579 unsigned int flags,
fd69c399 3580 int *off, int *err,
65101aec 3581 struct sk_buff **last);
b50b0580
SD
3582struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3583 struct sk_buff_head *queue,
e427cad6 3584 unsigned int flags, int *off, int *err,
ea3793ee 3585 struct sk_buff **last);
b50b0580
SD
3586struct sk_buff *__skb_recv_datagram(struct sock *sk,
3587 struct sk_buff_head *sk_queue,
e427cad6 3588 unsigned int flags, int *off, int *err);
7965bd4d
JP
3589struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3590 int *err);
a11e1d43
LT
3591__poll_t datagram_poll(struct file *file, struct socket *sock,
3592 struct poll_table_struct *wait);
c0371da6
AV
3593int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3594 struct iov_iter *to, int size);
51f3d02b
DM
3595static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3596 struct msghdr *msg, int size)
3597{
e5a4b0bb 3598 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3599}
e5a4b0bb
AV
3600int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3601 struct msghdr *msg);
65d69e25
SG
3602int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3603 struct iov_iter *to, int len,
3604 struct ahash_request *hash);
3a654f97
AV
3605int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3606 struct iov_iter *from, int len);
3a654f97 3607int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3608void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3609void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3610static inline void skb_free_datagram_locked(struct sock *sk,
3611 struct sk_buff *skb)
3612{
3613 __skb_free_datagram_locked(sk, skb, 0);
3614}
7965bd4d 3615int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3616int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3617int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3618__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 3619 int len);
a60e3cc7 3620int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3621 struct pipe_inode_info *pipe, unsigned int len,
25869262 3622 unsigned int flags);
20bf50de
TH
3623int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3624 int len);
7965bd4d 3625void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3626unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3627int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3628 int len, int hlen);
7965bd4d
JP
3629void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3630int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3631void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3632bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3633bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3634struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3635struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3636 unsigned int offset);
0d5501c1 3637struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3638int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3639int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3640int skb_vlan_pop(struct sk_buff *skb);
3641int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
3642int skb_eth_pop(struct sk_buff *skb);
3643int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3644 const unsigned char *src);
fa4e0f88 3645int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3646 int mac_len, bool ethernet);
040b5cfb
MV
3647int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3648 bool ethernet);
d27cf5c5 3649int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3650int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3651struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3652 gfp_t gfp);
20380731 3653
6ce8e9ce
AV
3654static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3655{
3073f070 3656 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3657}
3658
7eab8d9e
AV
3659static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3660{
e5a4b0bb 3661 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3662}
3663
2817a336
DB
3664struct skb_checksum_ops {
3665 __wsum (*update)(const void *mem, int len, __wsum wsum);
3666 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3667};
3668
9617813d
DC
3669extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3670
2817a336
DB
3671__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3672 __wsum csum, const struct skb_checksum_ops *ops);
3673__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3674 __wsum csum);
3675
1e98a0f0
ED
3676static inline void * __must_check
3677__skb_header_pointer(const struct sk_buff *skb, int offset,
3678 int len, void *data, int hlen, void *buffer)
1da177e4 3679{
55820ee2 3680 if (hlen - offset >= len)
690e36e7 3681 return data + offset;
1da177e4 3682
690e36e7
DM
3683 if (!skb ||
3684 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3685 return NULL;
3686
3687 return buffer;
3688}
3689
1e98a0f0
ED
3690static inline void * __must_check
3691skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3692{
3693 return __skb_header_pointer(skb, offset, len, skb->data,
3694 skb_headlen(skb), buffer);
3695}
3696
4262e5cc
DB
3697/**
3698 * skb_needs_linearize - check if we need to linearize a given skb
3699 * depending on the given device features.
3700 * @skb: socket buffer to check
3701 * @features: net device features
3702 *
3703 * Returns true if either:
3704 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3705 * 2. skb is fragmented and the device does not support SG.
3706 */
3707static inline bool skb_needs_linearize(struct sk_buff *skb,
3708 netdev_features_t features)
3709{
3710 return skb_is_nonlinear(skb) &&
3711 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3712 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3713}
3714
d626f62b
ACM
3715static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3716 void *to,
3717 const unsigned int len)
3718{
3719 memcpy(to, skb->data, len);
3720}
3721
3722static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3723 const int offset, void *to,
3724 const unsigned int len)
3725{
3726 memcpy(to, skb->data + offset, len);
3727}
3728
27d7ff46
ACM
3729static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3730 const void *from,
3731 const unsigned int len)
3732{
3733 memcpy(skb->data, from, len);
3734}
3735
3736static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3737 const int offset,
3738 const void *from,
3739 const unsigned int len)
3740{
3741 memcpy(skb->data + offset, from, len);
3742}
3743
7965bd4d 3744void skb_init(void);
1da177e4 3745
ac45f602
PO
3746static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3747{
3748 return skb->tstamp;
3749}
3750
a61bbcf2
PM
3751/**
3752 * skb_get_timestamp - get timestamp from a skb
3753 * @skb: skb to get stamp from
13c6ee2a 3754 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3755 *
3756 * Timestamps are stored in the skb as offsets to a base timestamp.
3757 * This function converts the offset back to a struct timeval and stores
3758 * it in stamp.
3759 */
ac45f602 3760static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3761 struct __kernel_old_timeval *stamp)
a61bbcf2 3762{
13c6ee2a 3763 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3764}
3765
887feae3
DD
3766static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3767 struct __kernel_sock_timeval *stamp)
3768{
3769 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3770
3771 stamp->tv_sec = ts.tv_sec;
3772 stamp->tv_usec = ts.tv_nsec / 1000;
3773}
3774
ac45f602 3775static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3776 struct __kernel_old_timespec *stamp)
ac45f602 3777{
df1b4ba9
AB
3778 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3779
3780 stamp->tv_sec = ts.tv_sec;
3781 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3782}
3783
887feae3
DD
3784static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3785 struct __kernel_timespec *stamp)
3786{
3787 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3788
3789 stamp->tv_sec = ts.tv_sec;
3790 stamp->tv_nsec = ts.tv_nsec;
3791}
3792
b7aa0bf7 3793static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3794{
b7aa0bf7 3795 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3796}
3797
164891aa
SH
3798static inline ktime_t net_timedelta(ktime_t t)
3799{
3800 return ktime_sub(ktime_get_real(), t);
3801}
3802
b9ce204f
IJ
3803static inline ktime_t net_invalid_timestamp(void)
3804{
8b0e1953 3805 return 0;
b9ce204f 3806}
a61bbcf2 3807
de8f3a83
DB
3808static inline u8 skb_metadata_len(const struct sk_buff *skb)
3809{
3810 return skb_shinfo(skb)->meta_len;
3811}
3812
3813static inline void *skb_metadata_end(const struct sk_buff *skb)
3814{
3815 return skb_mac_header(skb);
3816}
3817
3818static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3819 const struct sk_buff *skb_b,
3820 u8 meta_len)
3821{
3822 const void *a = skb_metadata_end(skb_a);
3823 const void *b = skb_metadata_end(skb_b);
3824 /* Using more efficient varaiant than plain call to memcmp(). */
3825#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3826 u64 diffs = 0;
3827
3828 switch (meta_len) {
3829#define __it(x, op) (x -= sizeof(u##op))
3830#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3831 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3832 fallthrough;
de8f3a83 3833 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3834 fallthrough;
de8f3a83 3835 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3836 fallthrough;
de8f3a83
DB
3837 case 8: diffs |= __it_diff(a, b, 64);
3838 break;
3839 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3840 fallthrough;
de8f3a83 3841 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3842 fallthrough;
de8f3a83 3843 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3844 fallthrough;
de8f3a83
DB
3845 case 4: diffs |= __it_diff(a, b, 32);
3846 break;
3847 }
3848 return diffs;
3849#else
3850 return memcmp(a - meta_len, b - meta_len, meta_len);
3851#endif
3852}
3853
3854static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3855 const struct sk_buff *skb_b)
3856{
3857 u8 len_a = skb_metadata_len(skb_a);
3858 u8 len_b = skb_metadata_len(skb_b);
3859
3860 if (!(len_a | len_b))
3861 return false;
3862
3863 return len_a != len_b ?
3864 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3865}
3866
3867static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3868{
3869 skb_shinfo(skb)->meta_len = meta_len;
3870}
3871
3872static inline void skb_metadata_clear(struct sk_buff *skb)
3873{
3874 skb_metadata_set(skb, 0);
3875}
3876
62bccb8c
AD
3877struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3878
c1f19b51
RC
3879#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3880
7965bd4d
JP
3881void skb_clone_tx_timestamp(struct sk_buff *skb);
3882bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3883
3884#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3885
3886static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3887{
3888}
3889
3890static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3891{
3892 return false;
3893}
3894
3895#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3896
3897/**
3898 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3899 *
da92b194
RC
3900 * PHY drivers may accept clones of transmitted packets for
3901 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3902 * must call this function to return the skb back to the stack with a
3903 * timestamp.
da92b194 3904 *
2ff17117 3905 * @skb: clone of the original outgoing packet
7a76a021 3906 * @hwtstamps: hardware time stamps
c1f19b51
RC
3907 *
3908 */
3909void skb_complete_tx_timestamp(struct sk_buff *skb,
3910 struct skb_shared_hwtstamps *hwtstamps);
3911
e7ed11ee 3912void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
3913 struct skb_shared_hwtstamps *hwtstamps,
3914 struct sock *sk, int tstype);
3915
ac45f602
PO
3916/**
3917 * skb_tstamp_tx - queue clone of skb with send time stamps
3918 * @orig_skb: the original outgoing packet
3919 * @hwtstamps: hardware time stamps, may be NULL if not available
3920 *
3921 * If the skb has a socket associated, then this function clones the
3922 * skb (thus sharing the actual data and optional structures), stores
3923 * the optional hardware time stamping information (if non NULL) or
3924 * generates a software time stamp (otherwise), then queues the clone
3925 * to the error queue of the socket. Errors are silently ignored.
3926 */
7965bd4d
JP
3927void skb_tstamp_tx(struct sk_buff *orig_skb,
3928 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3929
4507a715
RC
3930/**
3931 * skb_tx_timestamp() - Driver hook for transmit timestamping
3932 *
3933 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3934 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3935 *
73409f3b
DM
3936 * Specifically, one should make absolutely sure that this function is
3937 * called before TX completion of this packet can trigger. Otherwise
3938 * the packet could potentially already be freed.
3939 *
4507a715
RC
3940 * @skb: A socket buffer.
3941 */
3942static inline void skb_tx_timestamp(struct sk_buff *skb)
3943{
c1f19b51 3944 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3945 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3946 skb_tstamp_tx(skb, NULL);
4507a715
RC
3947}
3948
6e3e939f
JB
3949/**
3950 * skb_complete_wifi_ack - deliver skb with wifi status
3951 *
3952 * @skb: the original outgoing packet
3953 * @acked: ack status
3954 *
3955 */
3956void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3957
7965bd4d
JP
3958__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3959__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3960
60476372
HX
3961static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3962{
6edec0e6
TH
3963 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3964 skb->csum_valid ||
3965 (skb->ip_summed == CHECKSUM_PARTIAL &&
3966 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3967}
3968
fb286bb2
HX
3969/**
3970 * skb_checksum_complete - Calculate checksum of an entire packet
3971 * @skb: packet to process
3972 *
3973 * This function calculates the checksum over the entire packet plus
3974 * the value of skb->csum. The latter can be used to supply the
3975 * checksum of a pseudo header as used by TCP/UDP. It returns the
3976 * checksum.
3977 *
3978 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3979 * this function can be used to verify that checksum on received
3980 * packets. In that case the function should return zero if the
3981 * checksum is correct. In particular, this function will return zero
3982 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3983 * hardware has already verified the correctness of the checksum.
3984 */
4381ca3c 3985static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3986{
60476372
HX
3987 return skb_csum_unnecessary(skb) ?
3988 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3989}
3990
77cffe23
TH
3991static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3992{
3993 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3994 if (skb->csum_level == 0)
3995 skb->ip_summed = CHECKSUM_NONE;
3996 else
3997 skb->csum_level--;
3998 }
3999}
4000
4001static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4002{
4003 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4004 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4005 skb->csum_level++;
4006 } else if (skb->ip_summed == CHECKSUM_NONE) {
4007 skb->ip_summed = CHECKSUM_UNNECESSARY;
4008 skb->csum_level = 0;
4009 }
4010}
4011
836e66c2
DB
4012static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4013{
4014 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4015 skb->ip_summed = CHECKSUM_NONE;
4016 skb->csum_level = 0;
4017 }
4018}
4019
76ba0aae
TH
4020/* Check if we need to perform checksum complete validation.
4021 *
4022 * Returns true if checksum complete is needed, false otherwise
4023 * (either checksum is unnecessary or zero checksum is allowed).
4024 */
4025static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4026 bool zero_okay,
4027 __sum16 check)
4028{
5d0c2b95
TH
4029 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4030 skb->csum_valid = 1;
77cffe23 4031 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
4032 return false;
4033 }
4034
4035 return true;
4036}
4037
da279887 4038/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
4039 * in checksum_init.
4040 */
4041#define CHECKSUM_BREAK 76
4042
4e18b9ad
TH
4043/* Unset checksum-complete
4044 *
4045 * Unset checksum complete can be done when packet is being modified
4046 * (uncompressed for instance) and checksum-complete value is
4047 * invalidated.
4048 */
4049static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4050{
4051 if (skb->ip_summed == CHECKSUM_COMPLETE)
4052 skb->ip_summed = CHECKSUM_NONE;
4053}
4054
76ba0aae
TH
4055/* Validate (init) checksum based on checksum complete.
4056 *
4057 * Return values:
4058 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4059 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4060 * checksum is stored in skb->csum for use in __skb_checksum_complete
4061 * non-zero: value of invalid checksum
4062 *
4063 */
4064static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4065 bool complete,
4066 __wsum psum)
4067{
4068 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4069 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4070 skb->csum_valid = 1;
76ba0aae
TH
4071 return 0;
4072 }
4073 }
4074
4075 skb->csum = psum;
4076
5d0c2b95
TH
4077 if (complete || skb->len <= CHECKSUM_BREAK) {
4078 __sum16 csum;
4079
4080 csum = __skb_checksum_complete(skb);
4081 skb->csum_valid = !csum;
4082 return csum;
4083 }
76ba0aae
TH
4084
4085 return 0;
4086}
4087
4088static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4089{
4090 return 0;
4091}
4092
4093/* Perform checksum validate (init). Note that this is a macro since we only
4094 * want to calculate the pseudo header which is an input function if necessary.
4095 * First we try to validate without any computation (checksum unnecessary) and
4096 * then calculate based on checksum complete calling the function to compute
4097 * pseudo header.
4098 *
4099 * Return values:
4100 * 0: checksum is validated or try to in skb_checksum_complete
4101 * non-zero: value of invalid checksum
4102 */
4103#define __skb_checksum_validate(skb, proto, complete, \
4104 zero_okay, check, compute_pseudo) \
4105({ \
4106 __sum16 __ret = 0; \
5d0c2b95 4107 skb->csum_valid = 0; \
76ba0aae
TH
4108 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4109 __ret = __skb_checksum_validate_complete(skb, \
4110 complete, compute_pseudo(skb, proto)); \
4111 __ret; \
4112})
4113
4114#define skb_checksum_init(skb, proto, compute_pseudo) \
4115 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4116
4117#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4118 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4119
4120#define skb_checksum_validate(skb, proto, compute_pseudo) \
4121 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4122
4123#define skb_checksum_validate_zero_check(skb, proto, check, \
4124 compute_pseudo) \
096a4cfa 4125 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4126
4127#define skb_checksum_simple_validate(skb) \
4128 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4129
d96535a1
TH
4130static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4131{
219f1d79 4132 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4133}
4134
e4aa33ad 4135static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4136{
4137 skb->csum = ~pseudo;
4138 skb->ip_summed = CHECKSUM_COMPLETE;
4139}
4140
e4aa33ad 4141#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4142do { \
4143 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4144 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4145} while (0)
4146
15e2396d
TH
4147static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4148 u16 start, u16 offset)
4149{
4150 skb->ip_summed = CHECKSUM_PARTIAL;
4151 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4152 skb->csum_offset = offset - start;
4153}
4154
dcdc8994
TH
4155/* Update skbuf and packet to reflect the remote checksum offload operation.
4156 * When called, ptr indicates the starting point for skb->csum when
4157 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4158 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4159 */
4160static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4161 int start, int offset, bool nopartial)
dcdc8994
TH
4162{
4163 __wsum delta;
4164
15e2396d
TH
4165 if (!nopartial) {
4166 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4167 return;
4168 }
4169
dcdc8994
TH
4170 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4171 __skb_checksum_complete(skb);
4172 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4173 }
4174
4175 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4176
4177 /* Adjust skb->csum since we changed the packet */
4178 skb->csum = csum_add(skb->csum, delta);
4179}
4180
cb9c6836
FW
4181static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4182{
4183#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4184 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4185#else
4186 return NULL;
4187#endif
4188}
4189
261db6c2 4190static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4191{
261db6c2
JS
4192#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4193 return skb->_nfct;
4194#else
4195 return 0UL;
4196#endif
1da177e4 4197}
261db6c2
JS
4198
4199static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4200{
261db6c2
JS
4201#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4202 skb->_nfct = nfct;
2fc72c7b 4203#endif
261db6c2 4204}
df5042f4
FW
4205
4206#ifdef CONFIG_SKB_EXTENSIONS
4207enum skb_ext_id {
4208#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4209 SKB_EXT_BRIDGE_NF,
4165079b
FW
4210#endif
4211#ifdef CONFIG_XFRM
4212 SKB_EXT_SEC_PATH,
95a7233c
PB
4213#endif
4214#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4215 TC_SKB_EXT,
3ee17bc7
MM
4216#endif
4217#if IS_ENABLED(CONFIG_MPTCP)
4218 SKB_EXT_MPTCP,
df5042f4
FW
4219#endif
4220 SKB_EXT_NUM, /* must be last */
4221};
4222
4223/**
4224 * struct skb_ext - sk_buff extensions
4225 * @refcnt: 1 on allocation, deallocated on 0
4226 * @offset: offset to add to @data to obtain extension address
4227 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4228 * @data: start of extension data, variable sized
4229 *
4230 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4231 * to use 'u8' types while allowing up to 2kb worth of extension data.
4232 */
4233struct skb_ext {
4234 refcount_t refcnt;
4235 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4236 u8 chunks; /* same */
5c91aa1d 4237 char data[] __aligned(8);
df5042f4
FW
4238};
4239
4930f483 4240struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4241void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4242 struct skb_ext *ext);
df5042f4
FW
4243void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4244void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4245void __skb_ext_put(struct skb_ext *ext);
4246
4247static inline void skb_ext_put(struct sk_buff *skb)
4248{
4249 if (skb->active_extensions)
4250 __skb_ext_put(skb->extensions);
4251}
4252
df5042f4
FW
4253static inline void __skb_ext_copy(struct sk_buff *dst,
4254 const struct sk_buff *src)
4255{
4256 dst->active_extensions = src->active_extensions;
4257
4258 if (src->active_extensions) {
4259 struct skb_ext *ext = src->extensions;
4260
4261 refcount_inc(&ext->refcnt);
4262 dst->extensions = ext;
4263 }
4264}
4265
4266static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4267{
4268 skb_ext_put(dst);
4269 __skb_ext_copy(dst, src);
4270}
4271
4272static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4273{
4274 return !!ext->offset[i];
4275}
4276
4277static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4278{
4279 return skb->active_extensions & (1 << id);
4280}
4281
4282static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4283{
4284 if (skb_ext_exist(skb, id))
4285 __skb_ext_del(skb, id);
4286}
4287
4288static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4289{
4290 if (skb_ext_exist(skb, id)) {
4291 struct skb_ext *ext = skb->extensions;
4292
4293 return (void *)ext + (ext->offset[id] << 3);
4294 }
4295
4296 return NULL;
4297}
174e2381
FW
4298
4299static inline void skb_ext_reset(struct sk_buff *skb)
4300{
4301 if (unlikely(skb->active_extensions)) {
4302 __skb_ext_put(skb->extensions);
4303 skb->active_extensions = 0;
4304 }
4305}
677bf08c
FW
4306
4307static inline bool skb_has_extensions(struct sk_buff *skb)
4308{
4309 return unlikely(skb->active_extensions);
4310}
df5042f4
FW
4311#else
4312static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4313static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4314static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4315static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4316static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4317static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4318#endif /* CONFIG_SKB_EXTENSIONS */
4319
895b5c9f 4320static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4321{
5f79e0f9 4322#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4323 nf_conntrack_put(skb_nfct(skb));
4324 skb->_nfct = 0;
2fc72c7b 4325#endif
a193a4ab
PM
4326}
4327
124dff01
PM
4328static inline void nf_reset_trace(struct sk_buff *skb)
4329{
478b360a 4330#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4331 skb->nf_trace = 0;
4332#endif
a193a4ab
PM
4333}
4334
2b5ec1a5
YY
4335static inline void ipvs_reset(struct sk_buff *skb)
4336{
4337#if IS_ENABLED(CONFIG_IP_VS)
4338 skb->ipvs_property = 0;
4339#endif
4340}
4341
de8bda1d 4342/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4343static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4344 bool copy)
edda553c 4345{
5f79e0f9 4346#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4347 dst->_nfct = src->_nfct;
4348 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4349#endif
478b360a 4350#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4351 if (copy)
4352 dst->nf_trace = src->nf_trace;
478b360a 4353#endif
edda553c
YK
4354}
4355
e7ac05f3
YK
4356static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4357{
e7ac05f3 4358#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4359 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4360#endif
b1937227 4361 __nf_copy(dst, src, true);
e7ac05f3
YK
4362}
4363
984bc16c
JM
4364#ifdef CONFIG_NETWORK_SECMARK
4365static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4366{
4367 to->secmark = from->secmark;
4368}
4369
4370static inline void skb_init_secmark(struct sk_buff *skb)
4371{
4372 skb->secmark = 0;
4373}
4374#else
4375static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4376{ }
4377
4378static inline void skb_init_secmark(struct sk_buff *skb)
4379{ }
4380#endif
4381
7af8f4ca
FW
4382static inline int secpath_exists(const struct sk_buff *skb)
4383{
4384#ifdef CONFIG_XFRM
4165079b 4385 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4386#else
4387 return 0;
4388#endif
4389}
4390
574f7194
EB
4391static inline bool skb_irq_freeable(const struct sk_buff *skb)
4392{
4393 return !skb->destructor &&
7af8f4ca 4394 !secpath_exists(skb) &&
cb9c6836 4395 !skb_nfct(skb) &&
574f7194
EB
4396 !skb->_skb_refdst &&
4397 !skb_has_frag_list(skb);
4398}
4399
f25f4e44
PWJ
4400static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4401{
f25f4e44 4402 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4403}
4404
9247744e 4405static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4406{
4e3ab47a 4407 return skb->queue_mapping;
4e3ab47a
PE
4408}
4409
f25f4e44
PWJ
4410static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4411{
f25f4e44 4412 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4413}
4414
d5a9e24a
DM
4415static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4416{
4417 skb->queue_mapping = rx_queue + 1;
4418}
4419
9247744e 4420static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4421{
4422 return skb->queue_mapping - 1;
4423}
4424
9247744e 4425static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4426{
a02cec21 4427 return skb->queue_mapping != 0;
d5a9e24a
DM
4428}
4429
4ff06203
JA
4430static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4431{
4432 skb->dst_pending_confirm = val;
4433}
4434
4435static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4436{
4437 return skb->dst_pending_confirm != 0;
4438}
4439
2294be0f 4440static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4441{
0b3d8e08 4442#ifdef CONFIG_XFRM
4165079b 4443 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4444#else
def8b4fa 4445 return NULL;
def8b4fa 4446#endif
0b3d8e08 4447}
def8b4fa 4448
68c33163
PS
4449/* Keeps track of mac header offset relative to skb->head.
4450 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4451 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4452 * tunnel skb it points to outer mac header.
4453 * Keeps track of level of encapsulation of network headers.
4454 */
68c33163 4455struct skb_gso_cb {
802ab55a
AD
4456 union {
4457 int mac_offset;
4458 int data_offset;
4459 };
3347c960 4460 int encap_level;
76443456 4461 __wsum csum;
7e2b10c1 4462 __u16 csum_start;
68c33163 4463};
a08e7fd9
CZ
4464#define SKB_GSO_CB_OFFSET 32
4465#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4466
4467static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4468{
4469 return (skb_mac_header(inner_skb) - inner_skb->head) -
4470 SKB_GSO_CB(inner_skb)->mac_offset;
4471}
4472
1e2bd517
PS
4473static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4474{
4475 int new_headroom, headroom;
4476 int ret;
4477
4478 headroom = skb_headroom(skb);
4479 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4480 if (ret)
4481 return ret;
4482
4483 new_headroom = skb_headroom(skb);
4484 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4485 return 0;
4486}
4487
08b64fcc
AD
4488static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4489{
4490 /* Do not update partial checksums if remote checksum is enabled. */
4491 if (skb->remcsum_offload)
4492 return;
4493
4494 SKB_GSO_CB(skb)->csum = res;
4495 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4496}
4497
7e2b10c1
TH
4498/* Compute the checksum for a gso segment. First compute the checksum value
4499 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4500 * then add in skb->csum (checksum from csum_start to end of packet).
4501 * skb->csum and csum_start are then updated to reflect the checksum of the
4502 * resultant packet starting from the transport header-- the resultant checksum
4503 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4504 * header.
4505 */
4506static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4507{
76443456
AD
4508 unsigned char *csum_start = skb_transport_header(skb);
4509 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4510 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4511
76443456
AD
4512 SKB_GSO_CB(skb)->csum = res;
4513 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4514
76443456 4515 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4516}
4517
bdcc0924 4518static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4519{
4520 return skb_shinfo(skb)->gso_size;
4521}
4522
36a8f39e 4523/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4524static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4525{
4526 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4527}
4528
d02f51cb
DA
4529/* Note: Should be called only if skb_is_gso(skb) is true */
4530static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4531{
4532 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4533}
4534
4c3024de 4535/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4536static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4537{
4c3024de 4538 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4539}
4540
5293efe6
DB
4541static inline void skb_gso_reset(struct sk_buff *skb)
4542{
4543 skb_shinfo(skb)->gso_size = 0;
4544 skb_shinfo(skb)->gso_segs = 0;
4545 skb_shinfo(skb)->gso_type = 0;
4546}
4547
d02f51cb
DA
4548static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4549 u16 increment)
4550{
4551 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4552 return;
4553 shinfo->gso_size += increment;
4554}
4555
4556static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4557 u16 decrement)
4558{
4559 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4560 return;
4561 shinfo->gso_size -= decrement;
4562}
4563
7965bd4d 4564void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4565
4566static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4567{
4568 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4569 * wanted then gso_type will be set. */
05bdd2f1
ED
4570 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4571
b78462eb
AD
4572 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4573 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4574 __skb_warn_lro_forwarding(skb);
4575 return true;
4576 }
4577 return false;
4578}
4579
35fc92a9
HX
4580static inline void skb_forward_csum(struct sk_buff *skb)
4581{
4582 /* Unfortunately we don't support this one. Any brave souls? */
4583 if (skb->ip_summed == CHECKSUM_COMPLETE)
4584 skb->ip_summed = CHECKSUM_NONE;
4585}
4586
bc8acf2c
ED
4587/**
4588 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4589 * @skb: skb to check
4590 *
4591 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4592 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4593 * use this helper, to document places where we make this assertion.
4594 */
05bdd2f1 4595static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4596{
4597#ifdef DEBUG
4598 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4599#endif
4600}
4601
f35d9d8a 4602bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4603
ed1f50c3 4604int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4605struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4606 unsigned int transport_len,
4607 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4608
3a7c1ee4
AD
4609/**
4610 * skb_head_is_locked - Determine if the skb->head is locked down
4611 * @skb: skb to check
4612 *
4613 * The head on skbs build around a head frag can be removed if they are
4614 * not cloned. This function returns true if the skb head is locked down
4615 * due to either being allocated via kmalloc, or by being a clone with
4616 * multiple references to the head.
4617 */
4618static inline bool skb_head_is_locked(const struct sk_buff *skb)
4619{
4620 return !skb->head_frag || skb_cloned(skb);
4621}
fe6cc55f 4622
179bc67f
EC
4623/* Local Checksum Offload.
4624 * Compute outer checksum based on the assumption that the
4625 * inner checksum will be offloaded later.
d0dcde64 4626 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4627 * explanation of how this works.
179bc67f
EC
4628 * Fill in outer checksum adjustment (e.g. with sum of outer
4629 * pseudo-header) before calling.
4630 * Also ensure that inner checksum is in linear data area.
4631 */
4632static inline __wsum lco_csum(struct sk_buff *skb)
4633{
9e74a6da
AD
4634 unsigned char *csum_start = skb_checksum_start(skb);
4635 unsigned char *l4_hdr = skb_transport_header(skb);
4636 __wsum partial;
179bc67f
EC
4637
4638 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4639 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4640 skb->csum_offset));
4641
179bc67f 4642 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4643 * adjustment filled in by caller) and return result.
179bc67f 4644 */
9e74a6da 4645 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4646}
4647
2c64605b
PNA
4648static inline bool skb_is_redirected(const struct sk_buff *skb)
4649{
4650#ifdef CONFIG_NET_REDIRECT
4651 return skb->redirected;
4652#else
4653 return false;
4654#endif
4655}
4656
4657static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4658{
4659#ifdef CONFIG_NET_REDIRECT
4660 skb->redirected = 1;
4661 skb->from_ingress = from_ingress;
4662 if (skb->from_ingress)
4663 skb->tstamp = 0;
4664#endif
4665}
4666
4667static inline void skb_reset_redirect(struct sk_buff *skb)
4668{
4669#ifdef CONFIG_NET_REDIRECT
4670 skb->redirected = 0;
4671#endif
4672}
4673
fa821170
XL
4674static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4675{
4676 return skb->csum_not_inet;
4677}
4678
6370cc3b
AN
4679static inline void skb_set_kcov_handle(struct sk_buff *skb,
4680 const u64 kcov_handle)
4681{
fa69ee5a
ME
4682#ifdef CONFIG_KCOV
4683 skb->kcov_handle = kcov_handle;
4684#endif
6370cc3b
AN
4685}
4686
4687static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4688{
fa69ee5a
ME
4689#ifdef CONFIG_KCOV
4690 return skb->kcov_handle;
6370cc3b 4691#else
fa69ee5a
ME
4692 return 0;
4693#endif
4694}
6370cc3b 4695
1da177e4
LT
4696#endif /* __KERNEL__ */
4697#endif /* _LINUX_SKBUFF_H */