bitops: Add non-atomic bitops for pointers
[linux-block.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
8842d285 17#include <linux/bvec.h>
1da177e4 18#include <linux/cache.h>
56b17425 19#include <linux/rbtree.h>
51f3d02b 20#include <linux/socket.h>
c1d1b437 21#include <linux/refcount.h>
1da177e4 22
60063497 23#include <linux/atomic.h>
1da177e4
LT
24#include <asm/types.h>
25#include <linux/spinlock.h>
1da177e4 26#include <linux/net.h>
3fc7e8a6 27#include <linux/textsearch.h>
1da177e4 28#include <net/checksum.h>
a80958f4 29#include <linux/rcupdate.h>
b7aa0bf7 30#include <linux/hrtimer.h>
131ea667 31#include <linux/dma-mapping.h>
c8f44aff 32#include <linux/netdev_features.h>
363ec392 33#include <linux/sched.h>
e6017571 34#include <linux/sched/clock.h>
1bd758eb 35#include <net/flow_dissector.h>
a60e3cc7 36#include <linux/splice.h>
72b31f72 37#include <linux/in6.h>
8b10cab6 38#include <linux/if_packet.h>
f70ea018 39#include <net/flow.h>
6a5bcd84 40#include <net/page_pool.h>
261db6c2
JS
41#if IS_ENABLED(CONFIG_NF_CONNTRACK)
42#include <linux/netfilter/nf_conntrack_common.h>
43#endif
1da177e4 44
7a6ae71b
TH
45/* The interface for checksum offload between the stack and networking drivers
46 * is as follows...
47 *
48 * A. IP checksum related features
49 *
50 * Drivers advertise checksum offload capabilities in the features of a device.
db1f00fb
DC
51 * From the stack's point of view these are capabilities offered by the driver.
52 * A driver typically only advertises features that it is capable of offloading
7a6ae71b
TH
53 * to its device.
54 *
55 * The checksum related features are:
56 *
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
62 *
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
db1f00fb 67 * is TCP or UDP. The IPv4 header may contain IP options.
7a6ae71b
TH
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
71 *
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
645f0897 75 * IPv6|UDP where the Next Header field in the IPv6
7a6ae71b
TH
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
81 *
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
db1f00fb 83 * This flag is only used to disable the RX checksum
7a6ae71b
TH
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
87 *
88 * B. Checksumming of received packets by device. Indication of checksum
db1f00fb 89 * verification is set in skb->ip_summed. Possible values are:
78ea85f1
DB
90 *
91 * CHECKSUM_NONE:
92 *
7a6ae71b 93 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
96 *
97 * CHECKSUM_UNNECESSARY:
98 *
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
77cffe23
TH
105 *
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
b4759dcd 113 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
114 *
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
db1f00fb 119 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
77cffe23 120 * two. If the device were only able to verify the UDP checksum and not
db1f00fb 121 * GRE, either because it doesn't support GRE checksum or because GRE
77cffe23
TH
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
78ea85f1
DB
124 *
125 * CHECKSUM_COMPLETE:
126 *
127 * This is the most generic way. The device supplied checksum of the _whole_
db1f00fb 128 * packet as seen by netif_rx() and fills in skb->csum. This means the
78ea85f1
DB
129 * hardware doesn't need to parse L3/L4 headers to implement this.
130 *
b4759dcd
DC
131 * Notes:
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
135 *
136 * CHECKSUM_PARTIAL:
137 *
6edec0e6
TH
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
146 * be verified.
78ea85f1 147 *
7a6ae71b
TH
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
150 *
151 * CHECKSUM_PARTIAL:
152 *
7a6ae71b 153 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 154 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
db1f00fb
DC
157 * offset of the packet, but it should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum -- it is the
7a6ae71b
TH
159 * purview of the stack to validate that csum_start and csum_offset are set
160 * correctly.
161 *
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
166 *
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
78ea85f1 174 *
7a6ae71b 175 * CHECKSUM_NONE:
78ea85f1 176 *
7a6ae71b
TH
177 * The skb was already checksummed by the protocol, or a checksum is not
178 * required.
78ea85f1
DB
179 *
180 * CHECKSUM_UNNECESSARY:
181 *
db1f00fb 182 * This has the same meaning as CHECKSUM_NONE for checksum offload on
7a6ae71b 183 * output.
78ea85f1 184 *
7a6ae71b
TH
185 * CHECKSUM_COMPLETE:
186 * Not used in checksum output. If a driver observes a packet with this value
db1f00fb 187 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
7a6ae71b
TH
188 *
189 * D. Non-IP checksum (CRC) offloads
190 *
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
db1f00fb 193 * will set csum_start and csum_offset accordingly, set ip_summed to
dba00306
DC
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
200 *
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
db1f00fb
DC
204 * accordingly. Note that there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
7a6ae71b 206 * both IP checksum offload and FCOE CRC offload must verify which offload
db1f00fb 207 * is configured for a packet, presumably by inspecting packet headers.
7a6ae71b
TH
208 *
209 * E. Checksumming on output with GSO.
210 *
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
db1f00fb
DC
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
216 * csum_offset are set to refer to the outermost checksum being offloaded
217 * (two offloaded checksums are possible with UDP encapsulation).
78ea85f1
DB
218 */
219
60476372 220/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
221#define CHECKSUM_NONE 0
222#define CHECKSUM_UNNECESSARY 1
223#define CHECKSUM_COMPLETE 2
224#define CHECKSUM_PARTIAL 3
1da177e4 225
77cffe23
TH
226/* Maximum value in skb->csum_level */
227#define SKB_MAX_CSUM_LEVEL 3
228
0bec8c88 229#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 230#define SKB_WITH_OVERHEAD(X) \
deea84b0 231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
232#define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
234#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
236
87fb4b7b
ED
237/* return minimum truesize of one skb containing X bytes of data */
238#define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
241
7999096f 242struct ahash_request;
1da177e4 243struct net_device;
716ea3a7 244struct scatterlist;
9c55e01c 245struct pipe_inode_info;
a8f820aa 246struct iov_iter;
fd11a83d 247struct napi_struct;
d58e468b
PP
248struct bpf_prog;
249union bpf_attr;
df5042f4 250struct skb_ext;
1da177e4 251
34666d46 252#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 253struct nf_bridge_info {
3eaf4025
FW
254 enum {
255 BRNF_PROTO_UNCHANGED,
256 BRNF_PROTO_8021Q,
257 BRNF_PROTO_PPPOE
7fb48c5b 258 } orig_proto:8;
72b1e5e4
FW
259 u8 pkt_otherhost:1;
260 u8 in_prerouting:1;
261 u8 bridged_dnat:1;
411ffb4f 262 __u16 frag_max_size;
bf1ac5ca 263 struct net_device *physindev;
63cdbc06
FW
264
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
7fb48c5b 267 union {
72b1e5e4 268 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
269 __be32 ipv4_daddr;
270 struct in6_addr ipv6_daddr;
72b1e5e4
FW
271
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
275 */
276 char neigh_header[8];
72b31f72 277 };
1da177e4
LT
278};
279#endif
280
95a7233c
PB
281#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
282/* Chain in tc_skb_ext will be used to share the tc chain with
283 * ovs recirc_id. It will be set to the current chain by tc
284 * and read by ovs to recirc_id.
285 */
286struct tc_skb_ext {
287 __u32 chain;
038ebb1a 288 __u16 mru;
d29334c1 289 bool post_ct;
95a7233c
PB
290};
291#endif
292
1da177e4
LT
293struct sk_buff_head {
294 /* These two members must be first. */
295 struct sk_buff *next;
296 struct sk_buff *prev;
297
298 __u32 qlen;
299 spinlock_t lock;
300};
301
302struct sk_buff;
303
9d4dde52
IC
304/* To allow 64K frame to be packed as single skb without frag_list we
305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
306 * buffers which do not start on a page boundary.
307 *
308 * Since GRO uses frags we allocate at least 16 regardless of page
309 * size.
a715dea3 310 */
9d4dde52 311#if (65536/PAGE_SIZE + 1) < 16
eec00954 312#define MAX_SKB_FRAGS 16UL
a715dea3 313#else
9d4dde52 314#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 315#endif
5f74f82e 316extern int sysctl_max_skb_frags;
1da177e4 317
3953c46c
MRL
318/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
319 * segment using its current segmentation instead.
320 */
321#define GSO_BY_FRAGS 0xFFFF
322
8842d285 323typedef struct bio_vec skb_frag_t;
1da177e4 324
161e6137 325/**
7240b60c 326 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
327 * @frag: skb fragment
328 */
9e903e08
ED
329static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330{
b8b576a1 331 return frag->bv_len;
9e903e08
ED
332}
333
161e6137 334/**
7240b60c 335 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
336 * @frag: skb fragment
337 * @size: size of fragment
338 */
9e903e08
ED
339static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
340{
b8b576a1 341 frag->bv_len = size;
9e903e08
ED
342}
343
161e6137 344/**
7240b60c 345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
346 * @frag: skb fragment
347 * @delta: value to add
348 */
9e903e08
ED
349static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
350{
b8b576a1 351 frag->bv_len += delta;
9e903e08
ED
352}
353
161e6137 354/**
7240b60c 355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
356 * @frag: skb fragment
357 * @delta: value to subtract
358 */
9e903e08
ED
359static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
360{
b8b576a1 361 frag->bv_len -= delta;
9e903e08
ED
362}
363
161e6137
PT
364/**
365 * skb_frag_must_loop - Test if %p is a high memory page
366 * @p: fragment's page
367 */
c613c209
WB
368static inline bool skb_frag_must_loop(struct page *p)
369{
370#if defined(CONFIG_HIGHMEM)
29766bcf 371 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
372 return true;
373#endif
374 return false;
375}
376
377/**
378 * skb_frag_foreach_page - loop over pages in a fragment
379 *
380 * @f: skb frag to operate on
1dfa5bd3 381 * @f_off: offset from start of f->bv_page
c613c209
WB
382 * @f_len: length from f_off to loop over
383 * @p: (temp var) current page
384 * @p_off: (temp var) offset from start of current page,
385 * non-zero only on first page.
386 * @p_len: (temp var) length in current page,
387 * < PAGE_SIZE only on first and last page.
388 * @copied: (temp var) length so far, excluding current p_len.
389 *
390 * A fragment can hold a compound page, in which case per-page
391 * operations, notably kmap_atomic, must be called for each
392 * regular page.
393 */
394#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
396 p_off = (f_off) & (PAGE_SIZE - 1), \
397 p_len = skb_frag_must_loop(p) ? \
398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
399 copied = 0; \
400 copied < f_len; \
401 copied += p_len, p++, p_off = 0, \
402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
403
ac45f602
PO
404#define HAVE_HW_TIME_STAMP
405
406/**
d3a21be8 407 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
408 * @hwtstamp: hardware time stamp transformed into duration
409 * since arbitrary point in time
ac45f602
PO
410 *
411 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 412 * skb->tstamp.
ac45f602
PO
413 *
414 * hwtstamps can only be compared against other hwtstamps from
415 * the same device.
416 *
417 * This structure is attached to packets as part of the
418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
419 */
420struct skb_shared_hwtstamps {
421 ktime_t hwtstamp;
ac45f602
PO
422};
423
2244d07b
OH
424/* Definitions for tx_flags in struct skb_shared_info */
425enum {
426 /* generate hardware time stamp */
427 SKBTX_HW_TSTAMP = 1 << 0,
428
e7fd2885 429 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
430 SKBTX_SW_TSTAMP = 1 << 1,
431
432 /* device driver is going to provide hardware time stamp */
433 SKBTX_IN_PROGRESS = 1 << 2,
434
6e3e939f 435 /* generate wifi status information (where possible) */
62b1a8ab 436 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 437
e7fd2885
WB
438 /* generate software time stamp when entering packet scheduling */
439 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
440};
441
e1c8a607 442#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 443 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
444#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
445
06b4feb3
JL
446/* Definitions for flags in struct skb_shared_info */
447enum {
448 /* use zcopy routines */
449 SKBFL_ZEROCOPY_ENABLE = BIT(0),
450
451 /* This indicates at least one fragment might be overwritten
452 * (as in vmsplice(), sendfile() ...)
453 * If we need to compute a TX checksum, we'll need to copy
454 * all frags to avoid possible bad checksum
455 */
456 SKBFL_SHARED_FRAG = BIT(1),
457};
458
459#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
460
a6686f2f
SM
461/*
462 * The callback notifies userspace to release buffers when skb DMA is done in
463 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
464 * The zerocopy_success argument is true if zero copy transmit occurred,
465 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
466 * The ctx field is used to track device context.
467 * The desc field is used to track userspace buffer index.
a6686f2f
SM
468 */
469struct ubuf_info {
36177832
JL
470 void (*callback)(struct sk_buff *, struct ubuf_info *,
471 bool zerocopy_success);
4ab6c99d
WB
472 union {
473 struct {
474 unsigned long desc;
475 void *ctx;
476 };
477 struct {
478 u32 id;
479 u16 len;
480 u16 zerocopy:1;
481 u32 bytelen;
482 };
483 };
c1d1b437 484 refcount_t refcnt;
04c2d33e 485 u8 flags;
a91dbff5
WB
486
487 struct mmpin {
488 struct user_struct *user;
489 unsigned int num_pg;
490 } mmp;
ac45f602
PO
491};
492
52267790
WB
493#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
494
6f89dbce
SV
495int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
496void mm_unaccount_pinned_pages(struct mmpin *mmp);
497
8c793822
JL
498struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
499struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
500 struct ubuf_info *uarg);
52267790 501
8c793822 502void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790 503
8c793822
JL
504void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
505 bool success);
52267790 506
b5947e5d 507int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
508int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
509 struct msghdr *msg, int len,
510 struct ubuf_info *uarg);
511
1da177e4
LT
512/* This data is invariant across clones and lives at
513 * the end of the header data, ie. at skb->end.
514 */
515struct skb_shared_info {
06b4feb3 516 __u8 flags;
de8f3a83
DB
517 __u8 meta_len;
518 __u8 nr_frags;
9f42f126 519 __u8 tx_flags;
7967168c
HX
520 unsigned short gso_size;
521 /* Warning: this field is not always filled in (UFO)! */
522 unsigned short gso_segs;
1da177e4 523 struct sk_buff *frag_list;
ac45f602 524 struct skb_shared_hwtstamps hwtstamps;
7f564528 525 unsigned int gso_type;
09c2d251 526 u32 tskey;
ec7d2f2c
ED
527
528 /*
529 * Warning : all fields before dataref are cleared in __alloc_skb()
530 */
531 atomic_t dataref;
532
69e3c75f
JB
533 /* Intermediate layers must ensure that destructor_arg
534 * remains valid until skb destructor */
535 void * destructor_arg;
a6686f2f 536
fed66381
ED
537 /* must be last field, see pskb_expand_head() */
538 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
539};
540
541/* We divide dataref into two halves. The higher 16 bits hold references
542 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
543 * the entire skb->data. A clone of a headerless skb holds the length of
544 * the header in skb->hdr_len.
1da177e4
LT
545 *
546 * All users must obey the rule that the skb->data reference count must be
547 * greater than or equal to the payload reference count.
548 *
549 * Holding a reference to the payload part means that the user does not
550 * care about modifications to the header part of skb->data.
551 */
552#define SKB_DATAREF_SHIFT 16
553#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
554
d179cd12
DM
555
556enum {
c8753d55
VS
557 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
558 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
559 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
560};
561
7967168c
HX
562enum {
563 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
564
565 /* This indicates the skb is from an untrusted source. */
d9d30adf 566 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
567
568 /* This indicates the tcp segment has CWR set. */
d9d30adf 569 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 570
d9d30adf 571 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 572
d9d30adf 573 SKB_GSO_TCPV6 = 1 << 4,
68c33163 574
d9d30adf 575 SKB_GSO_FCOE = 1 << 5,
73136267 576
d9d30adf 577 SKB_GSO_GRE = 1 << 6,
0d89d203 578
d9d30adf 579 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 580
d9d30adf 581 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 582
d9d30adf 583 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 584
d9d30adf 585 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 586
d9d30adf 587 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 588
d9d30adf 589 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 590
d9d30adf 591 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 592
d9d30adf 593 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 594
d9d30adf 595 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
596
597 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
598
599 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
600
601 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
602};
603
2e07fa9c
ACM
604#if BITS_PER_LONG > 32
605#define NET_SKBUFF_DATA_USES_OFFSET 1
606#endif
607
608#ifdef NET_SKBUFF_DATA_USES_OFFSET
609typedef unsigned int sk_buff_data_t;
610#else
611typedef unsigned char *sk_buff_data_t;
612#endif
613
161e6137 614/**
1da177e4
LT
615 * struct sk_buff - socket buffer
616 * @next: Next buffer in list
617 * @prev: Previous buffer in list
363ec392 618 * @tstamp: Time we arrived/left
d2f273f0
RD
619 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
620 * for retransmit timer
56b17425 621 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 622 * @list: queue head
d84e0bd7 623 * @sk: Socket we are owned by
d2f273f0
RD
624 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
625 * fragmentation management
1da177e4 626 * @dev: Device we arrived on/are leaving by
d2f273f0 627 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 628 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 629 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 630 * @sp: the security path, used for xfrm
1da177e4
LT
631 * @len: Length of actual data
632 * @data_len: Data length
633 * @mac_len: Length of link layer header
334a8132 634 * @hdr_len: writable header length of cloned skb
663ead3b
HX
635 * @csum: Checksum (must include start/offset pair)
636 * @csum_start: Offset from skb->head where checksumming should start
637 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 638 * @priority: Packet queueing priority
60ff7467 639 * @ignore_df: allow local fragmentation
1da177e4 640 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 641 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
642 * @nohdr: Payload reference only, must not modify header
643 * @pkt_type: Packet class
c83c2486 644 * @fclone: skbuff clone status
c83c2486 645 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
646 * @inner_protocol_type: whether the inner protocol is
647 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
648 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
649 * @offload_fwd_mark: Packet was L2-forwarded in hardware
650 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 651 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 652 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
653 * @redirected: packet was redirected by packet classifier
654 * @from_ingress: packet was redirected from the ingress path
31729363
RD
655 * @peeked: this packet has been seen already, so stats have been
656 * done for it, don't do them again
ba9dda3a 657 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
658 * @protocol: Packet protocol from driver
659 * @destructor: Destruct function
e2080072 660 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
6ed6e1c7 661 * @_sk_redir: socket redirection information for skmsg
a9e419dc 662 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 663 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 664 * @skb_iif: ifindex of device we arrived on
1da177e4 665 * @tc_index: Traffic control index
61b905da 666 * @hash: the packet hash
d84e0bd7 667 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
668 * @head_frag: skb was allocated from page fragments,
669 * not allocated by kmalloc() or vmalloc().
8b700862 670 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
6a5bcd84
IA
671 * @pp_recycle: mark the packet for recycling instead of freeing (implies
672 * page_pool support on driver)
df5042f4 673 * @active_extensions: active extensions (skb_ext_id types)
553a5672 674 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 675 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 676 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 677 * ports.
a3b18ddb 678 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
679 * @wifi_acked_valid: wifi_acked was set
680 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 681 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
682 * @encapsulation: indicates the inner headers in the skbuff are valid
683 * @encap_hdr_csum: software checksum is needed
684 * @csum_valid: checksum is already valid
dba00306 685 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
686 * @csum_complete_sw: checksum was completed by software
687 * @csum_level: indicates the number of consecutive checksums found in
688 * the packet minus one that have been verified as
689 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 690 * @dst_pending_confirm: need to confirm neighbour
a48d189e 691 * @decrypted: Decrypted SKB
161e6137 692 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 693 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 694 * @secmark: security marking
d84e0bd7 695 * @mark: Generic packet mark
d2f273f0
RD
696 * @reserved_tailroom: (aka @mark) number of bytes of free space available
697 * at the tail of an sk_buff
698 * @vlan_present: VLAN tag is present
86a9bad3 699 * @vlan_proto: vlan encapsulation protocol
6aa895b0 700 * @vlan_tci: vlan tag control information
0d89d203 701 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
702 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
703 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
704 * @inner_transport_header: Inner transport layer header (encapsulation)
705 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 706 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
707 * @transport_header: Transport layer header
708 * @network_header: Network layer header
709 * @mac_header: Link layer header
fa69ee5a 710 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
711 * @tail: Tail pointer
712 * @end: End pointer
713 * @head: Head of buffer
714 * @data: Data head pointer
715 * @truesize: Buffer size
716 * @users: User count - see {datagram,tcp}.c
df5042f4 717 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
718 */
719
720struct sk_buff {
363ec392 721 union {
56b17425
ED
722 struct {
723 /* These two members must be first. */
724 struct sk_buff *next;
725 struct sk_buff *prev;
726
727 union {
bffa72cf
ED
728 struct net_device *dev;
729 /* Some protocols might use this space to store information,
730 * while device pointer would be NULL.
731 * UDP receive path is one user.
732 */
733 unsigned long dev_scratch;
56b17425
ED
734 };
735 };
fa0f5273 736 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 737 struct list_head list;
363ec392 738 };
fa0f5273
PO
739
740 union {
741 struct sock *sk;
742 int ip_defrag_offset;
743 };
1da177e4 744
c84d9490 745 union {
bffa72cf 746 ktime_t tstamp;
d3edd06e 747 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 748 };
1da177e4
LT
749 /*
750 * This is the control buffer. It is free to use for every
751 * layer. Please put your private variables there. If you
752 * want to keep them across layers you have to do a skb_clone()
753 * first. This is owned by whoever has the skb queued ATM.
754 */
da3f5cf1 755 char cb[48] __aligned(8);
1da177e4 756
e2080072
ED
757 union {
758 struct {
759 unsigned long _skb_refdst;
760 void (*destructor)(struct sk_buff *skb);
761 };
762 struct list_head tcp_tsorted_anchor;
e3526bb9
CW
763#ifdef CONFIG_NET_SOCK_MSG
764 unsigned long _sk_redir;
765#endif
e2080072
ED
766 };
767
b1937227 768#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 769 unsigned long _nfct;
da3f5cf1 770#endif
1da177e4 771 unsigned int len,
334a8132
PM
772 data_len;
773 __u16 mac_len,
774 hdr_len;
b1937227
ED
775
776 /* Following fields are _not_ copied in __copy_skb_header()
777 * Note that queue_mapping is here mostly to fill a hole.
778 */
b1937227 779 __u16 queue_mapping;
36bbef52
DB
780
781/* if you move cloned around you also must adapt those constants */
782#ifdef __BIG_ENDIAN_BITFIELD
783#define CLONED_MASK (1 << 7)
784#else
785#define CLONED_MASK 1
786#endif
787#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
788
d2f273f0 789 /* private: */
36bbef52 790 __u8 __cloned_offset[0];
d2f273f0 791 /* public: */
b1937227 792 __u8 cloned:1,
6869c4d8 793 nohdr:1,
b84f4cc9 794 fclone:2,
a59322be 795 peeked:1,
b1937227 796 head_frag:1,
6a5bcd84
IA
797 pfmemalloc:1,
798 pp_recycle:1; /* page_pool recycle indicator */
df5042f4
FW
799#ifdef CONFIG_SKB_EXTENSIONS
800 __u8 active_extensions;
801#endif
6a5bcd84 802
b1937227
ED
803 /* fields enclosed in headers_start/headers_end are copied
804 * using a single memcpy() in __copy_skb_header()
805 */
ebcf34f3 806 /* private: */
b1937227 807 __u32 headers_start[0];
ebcf34f3 808 /* public: */
4031ae6e 809
233577a2
HFS
810/* if you move pkt_type around you also must adapt those constants */
811#ifdef __BIG_ENDIAN_BITFIELD
812#define PKT_TYPE_MAX (7 << 5)
813#else
814#define PKT_TYPE_MAX 7
1da177e4 815#endif
233577a2 816#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 817
d2f273f0 818 /* private: */
233577a2 819 __u8 __pkt_type_offset[0];
d2f273f0 820 /* public: */
b1937227 821 __u8 pkt_type:3;
b1937227 822 __u8 ignore_df:1;
b1937227
ED
823 __u8 nf_trace:1;
824 __u8 ip_summed:2;
3853b584 825 __u8 ooo_okay:1;
8b700862 826
61b905da 827 __u8 l4_hash:1;
a3b18ddb 828 __u8 sw_hash:1;
6e3e939f
JB
829 __u8 wifi_acked_valid:1;
830 __u8 wifi_acked:1;
3bdc0eba 831 __u8 no_fcs:1;
77cffe23 832 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 833 __u8 encapsulation:1;
7e2b10c1 834 __u8 encap_hdr_csum:1;
5d0c2b95 835 __u8 csum_valid:1;
8b700862 836
0c4b2d37
MM
837#ifdef __BIG_ENDIAN_BITFIELD
838#define PKT_VLAN_PRESENT_BIT 7
839#else
840#define PKT_VLAN_PRESENT_BIT 0
841#endif
842#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 843 /* private: */
0c4b2d37 844 __u8 __pkt_vlan_present_offset[0];
d2f273f0 845 /* public: */
0c4b2d37 846 __u8 vlan_present:1;
7e3cead5 847 __u8 csum_complete_sw:1;
b1937227 848 __u8 csum_level:2;
dba00306 849 __u8 csum_not_inet:1;
4ff06203 850 __u8 dst_pending_confirm:1;
b1937227
ED
851#ifdef CONFIG_IPV6_NDISC_NODETYPE
852 __u8 ndisc_nodetype:2;
853#endif
8b700862 854
0c4b2d37 855 __u8 ipvs_property:1;
8bce6d7d 856 __u8 inner_protocol_type:1;
e585f236 857 __u8 remcsum_offload:1;
6bc506b4
IS
858#ifdef CONFIG_NET_SWITCHDEV
859 __u8 offload_fwd_mark:1;
875e8939 860 __u8 offload_l3_fwd_mark:1;
6bc506b4 861#endif
e7246e12
WB
862#ifdef CONFIG_NET_CLS_ACT
863 __u8 tc_skip_classify:1;
8dc07fdb 864 __u8 tc_at_ingress:1;
2c64605b
PNA
865#endif
866#ifdef CONFIG_NET_REDIRECT
867 __u8 redirected:1;
868 __u8 from_ingress:1;
e7246e12 869#endif
a48d189e
SB
870#ifdef CONFIG_TLS_DEVICE
871 __u8 decrypted:1;
872#endif
b1937227
ED
873
874#ifdef CONFIG_NET_SCHED
875 __u16 tc_index; /* traffic control index */
b1937227 876#endif
fe55f6d5 877
b1937227
ED
878 union {
879 __wsum csum;
880 struct {
881 __u16 csum_start;
882 __u16 csum_offset;
883 };
884 };
885 __u32 priority;
886 int skb_iif;
887 __u32 hash;
888 __be16 vlan_proto;
889 __u16 vlan_tci;
2bd82484
ED
890#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
891 union {
892 unsigned int napi_id;
893 unsigned int sender_cpu;
894 };
97fc2f08 895#endif
984bc16c 896#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 897 __u32 secmark;
0c4f691f 898#endif
0c4f691f 899
3b885787
NH
900 union {
901 __u32 mark;
16fad69c 902 __u32 reserved_tailroom;
3b885787 903 };
1da177e4 904
8bce6d7d
TH
905 union {
906 __be16 inner_protocol;
907 __u8 inner_ipproto;
908 };
909
1a37e412
SH
910 __u16 inner_transport_header;
911 __u16 inner_network_header;
912 __u16 inner_mac_header;
b1937227
ED
913
914 __be16 protocol;
1a37e412
SH
915 __u16 transport_header;
916 __u16 network_header;
917 __u16 mac_header;
b1937227 918
fa69ee5a
ME
919#ifdef CONFIG_KCOV
920 u64 kcov_handle;
921#endif
922
ebcf34f3 923 /* private: */
b1937227 924 __u32 headers_end[0];
ebcf34f3 925 /* public: */
b1937227 926
1da177e4 927 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 928 sk_buff_data_t tail;
4305b541 929 sk_buff_data_t end;
1da177e4 930 unsigned char *head,
4305b541 931 *data;
27a884dc 932 unsigned int truesize;
63354797 933 refcount_t users;
df5042f4
FW
934
935#ifdef CONFIG_SKB_EXTENSIONS
936 /* only useable after checking ->active_extensions != 0 */
937 struct skb_ext *extensions;
938#endif
1da177e4
LT
939};
940
941#ifdef __KERNEL__
942/*
943 * Handling routines are only of interest to the kernel
944 */
1da177e4 945
c93bdd0e
MG
946#define SKB_ALLOC_FCLONE 0x01
947#define SKB_ALLOC_RX 0x02
fd11a83d 948#define SKB_ALLOC_NAPI 0x04
c93bdd0e 949
161e6137
PT
950/**
951 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
952 * @skb: buffer
953 */
c93bdd0e
MG
954static inline bool skb_pfmemalloc(const struct sk_buff *skb)
955{
956 return unlikely(skb->pfmemalloc);
957}
958
7fee226a
ED
959/*
960 * skb might have a dst pointer attached, refcounted or not.
961 * _skb_refdst low order bit is set if refcount was _not_ taken
962 */
963#define SKB_DST_NOREF 1UL
964#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
965
966/**
967 * skb_dst - returns skb dst_entry
968 * @skb: buffer
969 *
970 * Returns skb dst_entry, regardless of reference taken or not.
971 */
adf30907
ED
972static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
973{
161e6137 974 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
975 * rcu_read_lock section
976 */
977 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
978 !rcu_read_lock_held() &&
979 !rcu_read_lock_bh_held());
980 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
981}
982
7fee226a
ED
983/**
984 * skb_dst_set - sets skb dst
985 * @skb: buffer
986 * @dst: dst entry
987 *
988 * Sets skb dst, assuming a reference was taken on dst and should
989 * be released by skb_dst_drop()
990 */
adf30907
ED
991static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
992{
7fee226a
ED
993 skb->_skb_refdst = (unsigned long)dst;
994}
995
932bc4d7
JA
996/**
997 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
998 * @skb: buffer
999 * @dst: dst entry
1000 *
1001 * Sets skb dst, assuming a reference was not taken on dst.
1002 * If dst entry is cached, we do not take reference and dst_release
1003 * will be avoided by refdst_drop. If dst entry is not cached, we take
1004 * reference, so that last dst_release can destroy the dst immediately.
1005 */
1006static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1007{
dbfc4fb7
HFS
1008 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1009 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1010}
7fee226a
ED
1011
1012/**
25985edc 1013 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1014 * @skb: buffer
1015 */
1016static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1017{
1018 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1019}
1020
161e6137
PT
1021/**
1022 * skb_rtable - Returns the skb &rtable
1023 * @skb: buffer
1024 */
511c3f92
ED
1025static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1026{
adf30907 1027 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1028}
1029
8b10cab6
JHS
1030/* For mangling skb->pkt_type from user space side from applications
1031 * such as nft, tc, etc, we only allow a conservative subset of
1032 * possible pkt_types to be set.
1033*/
1034static inline bool skb_pkt_type_ok(u32 ptype)
1035{
1036 return ptype <= PACKET_OTHERHOST;
1037}
1038
161e6137
PT
1039/**
1040 * skb_napi_id - Returns the skb's NAPI id
1041 * @skb: buffer
1042 */
90b602f8
ML
1043static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1044{
1045#ifdef CONFIG_NET_RX_BUSY_POLL
1046 return skb->napi_id;
1047#else
1048 return 0;
1049#endif
1050}
1051
161e6137
PT
1052/**
1053 * skb_unref - decrement the skb's reference count
1054 * @skb: buffer
1055 *
1056 * Returns true if we can free the skb.
1057 */
3889a803
PA
1058static inline bool skb_unref(struct sk_buff *skb)
1059{
1060 if (unlikely(!skb))
1061 return false;
63354797 1062 if (likely(refcount_read(&skb->users) == 1))
3889a803 1063 smp_rmb();
63354797 1064 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1065 return false;
1066
1067 return true;
1068}
1069
0a463c78 1070void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1071void kfree_skb(struct sk_buff *skb);
1072void kfree_skb_list(struct sk_buff *segs);
6413139d 1073void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1074void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1075
1076#ifdef CONFIG_TRACEPOINTS
7965bd4d 1077void consume_skb(struct sk_buff *skb);
be769db2
HX
1078#else
1079static inline void consume_skb(struct sk_buff *skb)
1080{
1081 return kfree_skb(skb);
1082}
1083#endif
1084
ca2c1418 1085void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1086void __kfree_skb(struct sk_buff *skb);
d7e8883c 1087extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1088
7965bd4d
JP
1089void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1090bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1091 bool *fragstolen, int *delta_truesize);
bad43ca8 1092
7965bd4d
JP
1093struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1094 int node);
2ea2f62c 1095struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1096struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1097struct sk_buff *build_skb_around(struct sk_buff *skb,
1098 void *data, unsigned int frag_size);
161e6137 1099
f450d539
AL
1100struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1101
161e6137
PT
1102/**
1103 * alloc_skb - allocate a network buffer
1104 * @size: size to allocate
1105 * @priority: allocation mask
1106 *
1107 * This function is a convenient wrapper around __alloc_skb().
1108 */
d179cd12 1109static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1110 gfp_t priority)
d179cd12 1111{
564824b0 1112 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1113}
1114
2e4e4410
ED
1115struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1116 unsigned long data_len,
1117 int max_page_order,
1118 int *errcode,
1119 gfp_t gfp_mask);
da29e4b4 1120struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1121
d0bf4a9e
ED
1122/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1123struct sk_buff_fclones {
1124 struct sk_buff skb1;
1125
1126 struct sk_buff skb2;
1127
2638595a 1128 refcount_t fclone_ref;
d0bf4a9e
ED
1129};
1130
1131/**
1132 * skb_fclone_busy - check if fclone is busy
293de7de 1133 * @sk: socket
d0bf4a9e
ED
1134 * @skb: buffer
1135 *
bda13fed 1136 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1137 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1138 * so we also check that this didnt happen.
d0bf4a9e 1139 */
39bb5e62
ED
1140static inline bool skb_fclone_busy(const struct sock *sk,
1141 const struct sk_buff *skb)
d0bf4a9e
ED
1142{
1143 const struct sk_buff_fclones *fclones;
1144
1145 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1146
1147 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1148 refcount_read(&fclones->fclone_ref) > 1 &&
f4dae54e 1149 READ_ONCE(fclones->skb2.sk) == sk;
d0bf4a9e
ED
1150}
1151
161e6137
PT
1152/**
1153 * alloc_skb_fclone - allocate a network buffer from fclone cache
1154 * @size: size to allocate
1155 * @priority: allocation mask
1156 *
1157 * This function is a convenient wrapper around __alloc_skb().
1158 */
d179cd12 1159static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1160 gfp_t priority)
d179cd12 1161{
c93bdd0e 1162 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1163}
1164
7965bd4d 1165struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1166void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1167int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1168struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1169void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1170struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1171struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1172 gfp_t gfp_mask, bool fclone);
1173static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1174 gfp_t gfp_mask)
1175{
1176 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1177}
7965bd4d
JP
1178
1179int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1180struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1181 unsigned int headroom);
1182struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1183 int newtailroom, gfp_t priority);
48a1df65
JD
1184int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1185 int offset, int len);
1186int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1187 int offset, int len);
7965bd4d 1188int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1189int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1190
1191/**
1192 * skb_pad - zero pad the tail of an skb
1193 * @skb: buffer to pad
1194 * @pad: space to pad
1195 *
1196 * Ensure that a buffer is followed by a padding area that is zero
1197 * filled. Used by network drivers which may DMA or transfer data
1198 * beyond the buffer end onto the wire.
1199 *
1200 * May return error in out of memory cases. The skb is freed on error.
1201 */
1202static inline int skb_pad(struct sk_buff *skb, int pad)
1203{
1204 return __skb_pad(skb, pad, true);
1205}
ead2ceb0 1206#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1207
be12a1fe
HFS
1208int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1209 int offset, size_t size);
1210
d94d9fee 1211struct skb_seq_state {
677e90ed
TG
1212 __u32 lower_offset;
1213 __u32 upper_offset;
1214 __u32 frag_idx;
1215 __u32 stepped_offset;
1216 struct sk_buff *root_skb;
1217 struct sk_buff *cur_skb;
1218 __u8 *frag_data;
97550f6f 1219 __u32 frag_off;
677e90ed
TG
1220};
1221
7965bd4d
JP
1222void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1223 unsigned int to, struct skb_seq_state *st);
1224unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1225 struct skb_seq_state *st);
1226void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1227
7965bd4d 1228unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1229 unsigned int to, struct ts_config *config);
3fc7e8a6 1230
09323cc4
TH
1231/*
1232 * Packet hash types specify the type of hash in skb_set_hash.
1233 *
1234 * Hash types refer to the protocol layer addresses which are used to
1235 * construct a packet's hash. The hashes are used to differentiate or identify
1236 * flows of the protocol layer for the hash type. Hash types are either
1237 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1238 *
1239 * Properties of hashes:
1240 *
1241 * 1) Two packets in different flows have different hash values
1242 * 2) Two packets in the same flow should have the same hash value
1243 *
1244 * A hash at a higher layer is considered to be more specific. A driver should
1245 * set the most specific hash possible.
1246 *
1247 * A driver cannot indicate a more specific hash than the layer at which a hash
1248 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1249 *
1250 * A driver may indicate a hash level which is less specific than the
1251 * actual layer the hash was computed on. For instance, a hash computed
1252 * at L4 may be considered an L3 hash. This should only be done if the
1253 * driver can't unambiguously determine that the HW computed the hash at
1254 * the higher layer. Note that the "should" in the second property above
1255 * permits this.
1256 */
1257enum pkt_hash_types {
1258 PKT_HASH_TYPE_NONE, /* Undefined type */
1259 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1260 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1261 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1262};
1263
bcc83839 1264static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1265{
bcc83839 1266 skb->hash = 0;
a3b18ddb 1267 skb->sw_hash = 0;
bcc83839
TH
1268 skb->l4_hash = 0;
1269}
1270
1271static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1272{
1273 if (!skb->l4_hash)
1274 skb_clear_hash(skb);
1275}
1276
1277static inline void
1278__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1279{
1280 skb->l4_hash = is_l4;
1281 skb->sw_hash = is_sw;
61b905da 1282 skb->hash = hash;
09323cc4
TH
1283}
1284
bcc83839
TH
1285static inline void
1286skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1287{
1288 /* Used by drivers to set hash from HW */
1289 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1290}
1291
1292static inline void
1293__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1294{
1295 __skb_set_hash(skb, hash, true, is_l4);
1296}
1297
e5276937 1298void __skb_get_hash(struct sk_buff *skb);
b917783c 1299u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937 1300u32 skb_get_poff(const struct sk_buff *skb);
f96533cd 1301u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
72a338bc 1302 const struct flow_keys_basic *keys, int hlen);
e5276937 1303__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
f96533cd 1304 const void *data, int hlen_proto);
e5276937
TH
1305
1306static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1307 int thoff, u8 ip_proto)
1308{
1309 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1310}
1311
1312void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1313 const struct flow_dissector_key *key,
1314 unsigned int key_count);
1315
089b19a9
SF
1316struct bpf_flow_dissector;
1317bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1318 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1319
3cbf4ffb
SF
1320bool __skb_flow_dissect(const struct net *net,
1321 const struct sk_buff *skb,
e5276937 1322 struct flow_dissector *flow_dissector,
f96533cd
AL
1323 void *target_container, const void *data,
1324 __be16 proto, int nhoff, int hlen, unsigned int flags);
e5276937
TH
1325
1326static inline bool skb_flow_dissect(const struct sk_buff *skb,
1327 struct flow_dissector *flow_dissector,
cd79a238 1328 void *target_container, unsigned int flags)
e5276937 1329{
3cbf4ffb
SF
1330 return __skb_flow_dissect(NULL, skb, flow_dissector,
1331 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1332}
1333
1334static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1335 struct flow_keys *flow,
1336 unsigned int flags)
e5276937
TH
1337{
1338 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1339 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1340 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1341}
1342
72a338bc 1343static inline bool
3cbf4ffb
SF
1344skb_flow_dissect_flow_keys_basic(const struct net *net,
1345 const struct sk_buff *skb,
f96533cd
AL
1346 struct flow_keys_basic *flow,
1347 const void *data, __be16 proto,
1348 int nhoff, int hlen, unsigned int flags)
e5276937
TH
1349{
1350 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1351 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1352 data, proto, nhoff, hlen, flags);
e5276937
TH
1353}
1354
82828b88
JP
1355void skb_flow_dissect_meta(const struct sk_buff *skb,
1356 struct flow_dissector *flow_dissector,
1357 void *target_container);
1358
75a56758 1359/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1360 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1361 * to user states.
1362 */
1363void
1364skb_flow_dissect_ct(const struct sk_buff *skb,
1365 struct flow_dissector *flow_dissector,
1366 void *target_container,
7baf2429 1367 u16 *ctinfo_map, size_t mapsize,
1368 bool post_ct);
62b32379
SH
1369void
1370skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1371 struct flow_dissector *flow_dissector,
1372 void *target_container);
1373
0cb09aff
AL
1374void skb_flow_dissect_hash(const struct sk_buff *skb,
1375 struct flow_dissector *flow_dissector,
1376 void *target_container);
1377
3958afa1 1378static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1379{
a3b18ddb 1380 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1381 __skb_get_hash(skb);
bfb564e7 1382
61b905da 1383 return skb->hash;
bfb564e7
KK
1384}
1385
20a17bf6 1386static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1387{
c6cc1ca7
TH
1388 if (!skb->l4_hash && !skb->sw_hash) {
1389 struct flow_keys keys;
de4c1f8b 1390 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1391
de4c1f8b 1392 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1393 }
f70ea018
TH
1394
1395 return skb->hash;
1396}
1397
55667441
ED
1398__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1399 const siphash_key_t *perturb);
50fb7992 1400
57bdf7f4
TH
1401static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1402{
61b905da 1403 return skb->hash;
57bdf7f4
TH
1404}
1405
3df7a74e
TH
1406static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1407{
61b905da 1408 to->hash = from->hash;
a3b18ddb 1409 to->sw_hash = from->sw_hash;
61b905da 1410 to->l4_hash = from->l4_hash;
3df7a74e
TH
1411};
1412
41477662
JK
1413static inline void skb_copy_decrypted(struct sk_buff *to,
1414 const struct sk_buff *from)
1415{
1416#ifdef CONFIG_TLS_DEVICE
1417 to->decrypted = from->decrypted;
1418#endif
1419}
1420
4305b541
ACM
1421#ifdef NET_SKBUFF_DATA_USES_OFFSET
1422static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1423{
1424 return skb->head + skb->end;
1425}
ec47ea82
AD
1426
1427static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1428{
1429 return skb->end;
1430}
4305b541
ACM
1431#else
1432static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1433{
1434 return skb->end;
1435}
ec47ea82
AD
1436
1437static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1438{
1439 return skb->end - skb->head;
1440}
4305b541
ACM
1441#endif
1442
1da177e4 1443/* Internal */
4305b541 1444#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1445
ac45f602
PO
1446static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1447{
1448 return &skb_shinfo(skb)->hwtstamps;
1449}
1450
52267790
WB
1451static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1452{
06b4feb3 1453 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1454
1455 return is_zcopy ? skb_uarg(skb) : NULL;
1456}
1457
8e044917 1458static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1459{
1460 refcount_inc(&uarg->refcnt);
1461}
1462
9ee5e5ad
JL
1463static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1464{
1465 skb_shinfo(skb)->destructor_arg = uarg;
1466 skb_shinfo(skb)->flags |= uarg->flags;
1467}
1468
52900d22
WB
1469static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1470 bool *have_ref)
52267790
WB
1471{
1472 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1473 if (unlikely(have_ref && *have_ref))
1474 *have_ref = false;
1475 else
8e044917 1476 net_zcopy_get(uarg);
9ee5e5ad 1477 skb_zcopy_init(skb, uarg);
52267790
WB
1478 }
1479}
1480
5cd8d46e
WB
1481static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1482{
1483 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1484 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1485}
1486
1487static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1488{
1489 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1490}
1491
1492static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1493{
1494 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1495}
1496
8e044917 1497static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1498{
1499 if (uarg)
36177832 1500 uarg->callback(NULL, uarg, true);
59776362
JL
1501}
1502
8e044917 1503static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1504{
1505 if (uarg) {
8c793822
JL
1506 if (uarg->callback == msg_zerocopy_callback)
1507 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1508 else if (have_uref)
8e044917 1509 net_zcopy_put(uarg);
236a6b1c
JL
1510 }
1511}
1512
52267790 1513/* Release a reference on a zerocopy structure */
36177832 1514static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1515{
1516 struct ubuf_info *uarg = skb_zcopy(skb);
1517
1518 if (uarg) {
36177832
JL
1519 if (!skb_zcopy_is_nouarg(skb))
1520 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1521
06b4feb3 1522 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
52267790
WB
1523 }
1524}
1525
a8305bff
DM
1526static inline void skb_mark_not_on_list(struct sk_buff *skb)
1527{
1528 skb->next = NULL;
1529}
1530
dcfea72e 1531/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1532#define skb_list_walk_safe(first, skb, next_skb) \
1533 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1534 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1535
992cba7e
DM
1536static inline void skb_list_del_init(struct sk_buff *skb)
1537{
1538 __list_del_entry(&skb->list);
1539 skb_mark_not_on_list(skb);
1540}
1541
1da177e4
LT
1542/**
1543 * skb_queue_empty - check if a queue is empty
1544 * @list: queue head
1545 *
1546 * Returns true if the queue is empty, false otherwise.
1547 */
1548static inline int skb_queue_empty(const struct sk_buff_head *list)
1549{
fd44b93c 1550 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1551}
1552
d7d16a89
ED
1553/**
1554 * skb_queue_empty_lockless - check if a queue is empty
1555 * @list: queue head
1556 *
1557 * Returns true if the queue is empty, false otherwise.
1558 * This variant can be used in lockless contexts.
1559 */
1560static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1561{
1562 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1563}
1564
1565
fc7ebb21
DM
1566/**
1567 * skb_queue_is_last - check if skb is the last entry in the queue
1568 * @list: queue head
1569 * @skb: buffer
1570 *
1571 * Returns true if @skb is the last buffer on the list.
1572 */
1573static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1574 const struct sk_buff *skb)
1575{
fd44b93c 1576 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1577}
1578
832d11c5
IJ
1579/**
1580 * skb_queue_is_first - check if skb is the first entry in the queue
1581 * @list: queue head
1582 * @skb: buffer
1583 *
1584 * Returns true if @skb is the first buffer on the list.
1585 */
1586static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1587 const struct sk_buff *skb)
1588{
fd44b93c 1589 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1590}
1591
249c8b42
DM
1592/**
1593 * skb_queue_next - return the next packet in the queue
1594 * @list: queue head
1595 * @skb: current buffer
1596 *
1597 * Return the next packet in @list after @skb. It is only valid to
1598 * call this if skb_queue_is_last() evaluates to false.
1599 */
1600static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1601 const struct sk_buff *skb)
1602{
1603 /* This BUG_ON may seem severe, but if we just return then we
1604 * are going to dereference garbage.
1605 */
1606 BUG_ON(skb_queue_is_last(list, skb));
1607 return skb->next;
1608}
1609
832d11c5
IJ
1610/**
1611 * skb_queue_prev - return the prev packet in the queue
1612 * @list: queue head
1613 * @skb: current buffer
1614 *
1615 * Return the prev packet in @list before @skb. It is only valid to
1616 * call this if skb_queue_is_first() evaluates to false.
1617 */
1618static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1619 const struct sk_buff *skb)
1620{
1621 /* This BUG_ON may seem severe, but if we just return then we
1622 * are going to dereference garbage.
1623 */
1624 BUG_ON(skb_queue_is_first(list, skb));
1625 return skb->prev;
1626}
1627
1da177e4
LT
1628/**
1629 * skb_get - reference buffer
1630 * @skb: buffer to reference
1631 *
1632 * Makes another reference to a socket buffer and returns a pointer
1633 * to the buffer.
1634 */
1635static inline struct sk_buff *skb_get(struct sk_buff *skb)
1636{
63354797 1637 refcount_inc(&skb->users);
1da177e4
LT
1638 return skb;
1639}
1640
1641/*
f8821f96 1642 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1643 */
1644
1da177e4
LT
1645/**
1646 * skb_cloned - is the buffer a clone
1647 * @skb: buffer to check
1648 *
1649 * Returns true if the buffer was generated with skb_clone() and is
1650 * one of multiple shared copies of the buffer. Cloned buffers are
1651 * shared data so must not be written to under normal circumstances.
1652 */
1653static inline int skb_cloned(const struct sk_buff *skb)
1654{
1655 return skb->cloned &&
1656 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1657}
1658
14bbd6a5
PS
1659static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1660{
d0164adc 1661 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1662
1663 if (skb_cloned(skb))
1664 return pskb_expand_head(skb, 0, 0, pri);
1665
1666 return 0;
1667}
1668
1da177e4
LT
1669/**
1670 * skb_header_cloned - is the header a clone
1671 * @skb: buffer to check
1672 *
1673 * Returns true if modifying the header part of the buffer requires
1674 * the data to be copied.
1675 */
1676static inline int skb_header_cloned(const struct sk_buff *skb)
1677{
1678 int dataref;
1679
1680 if (!skb->cloned)
1681 return 0;
1682
1683 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1684 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1685 return dataref != 1;
1686}
1687
9580bf2e
ED
1688static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1689{
1690 might_sleep_if(gfpflags_allow_blocking(pri));
1691
1692 if (skb_header_cloned(skb))
1693 return pskb_expand_head(skb, 0, 0, pri);
1694
1695 return 0;
1696}
1697
f4a775d1
ED
1698/**
1699 * __skb_header_release - release reference to header
1700 * @skb: buffer to operate on
f4a775d1
ED
1701 */
1702static inline void __skb_header_release(struct sk_buff *skb)
1703{
1704 skb->nohdr = 1;
1705 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1706}
1707
1708
1da177e4
LT
1709/**
1710 * skb_shared - is the buffer shared
1711 * @skb: buffer to check
1712 *
1713 * Returns true if more than one person has a reference to this
1714 * buffer.
1715 */
1716static inline int skb_shared(const struct sk_buff *skb)
1717{
63354797 1718 return refcount_read(&skb->users) != 1;
1da177e4
LT
1719}
1720
1721/**
1722 * skb_share_check - check if buffer is shared and if so clone it
1723 * @skb: buffer to check
1724 * @pri: priority for memory allocation
1725 *
1726 * If the buffer is shared the buffer is cloned and the old copy
1727 * drops a reference. A new clone with a single reference is returned.
1728 * If the buffer is not shared the original buffer is returned. When
1729 * being called from interrupt status or with spinlocks held pri must
1730 * be GFP_ATOMIC.
1731 *
1732 * NULL is returned on a memory allocation failure.
1733 */
47061bc4 1734static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1735{
d0164adc 1736 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1737 if (skb_shared(skb)) {
1738 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1739
1740 if (likely(nskb))
1741 consume_skb(skb);
1742 else
1743 kfree_skb(skb);
1da177e4
LT
1744 skb = nskb;
1745 }
1746 return skb;
1747}
1748
1749/*
1750 * Copy shared buffers into a new sk_buff. We effectively do COW on
1751 * packets to handle cases where we have a local reader and forward
1752 * and a couple of other messy ones. The normal one is tcpdumping
1753 * a packet thats being forwarded.
1754 */
1755
1756/**
1757 * skb_unshare - make a copy of a shared buffer
1758 * @skb: buffer to check
1759 * @pri: priority for memory allocation
1760 *
1761 * If the socket buffer is a clone then this function creates a new
1762 * copy of the data, drops a reference count on the old copy and returns
1763 * the new copy with the reference count at 1. If the buffer is not a clone
1764 * the original buffer is returned. When called with a spinlock held or
1765 * from interrupt state @pri must be %GFP_ATOMIC
1766 *
1767 * %NULL is returned on a memory allocation failure.
1768 */
e2bf521d 1769static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1770 gfp_t pri)
1da177e4 1771{
d0164adc 1772 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1773 if (skb_cloned(skb)) {
1774 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1775
1776 /* Free our shared copy */
1777 if (likely(nskb))
1778 consume_skb(skb);
1779 else
1780 kfree_skb(skb);
1da177e4
LT
1781 skb = nskb;
1782 }
1783 return skb;
1784}
1785
1786/**
1a5778aa 1787 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1788 * @list_: list to peek at
1789 *
1790 * Peek an &sk_buff. Unlike most other operations you _MUST_
1791 * be careful with this one. A peek leaves the buffer on the
1792 * list and someone else may run off with it. You must hold
1793 * the appropriate locks or have a private queue to do this.
1794 *
1795 * Returns %NULL for an empty list or a pointer to the head element.
1796 * The reference count is not incremented and the reference is therefore
1797 * volatile. Use with caution.
1798 */
05bdd2f1 1799static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1800{
18d07000
ED
1801 struct sk_buff *skb = list_->next;
1802
1803 if (skb == (struct sk_buff *)list_)
1804 skb = NULL;
1805 return skb;
1da177e4
LT
1806}
1807
8b69bd7d
DM
1808/**
1809 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1810 * @list_: list to peek at
1811 *
1812 * Like skb_peek(), but the caller knows that the list is not empty.
1813 */
1814static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1815{
1816 return list_->next;
1817}
1818
da5ef6e5
PE
1819/**
1820 * skb_peek_next - peek skb following the given one from a queue
1821 * @skb: skb to start from
1822 * @list_: list to peek at
1823 *
1824 * Returns %NULL when the end of the list is met or a pointer to the
1825 * next element. The reference count is not incremented and the
1826 * reference is therefore volatile. Use with caution.
1827 */
1828static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1829 const struct sk_buff_head *list_)
1830{
1831 struct sk_buff *next = skb->next;
18d07000 1832
da5ef6e5
PE
1833 if (next == (struct sk_buff *)list_)
1834 next = NULL;
1835 return next;
1836}
1837
1da177e4 1838/**
1a5778aa 1839 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1840 * @list_: list to peek at
1841 *
1842 * Peek an &sk_buff. Unlike most other operations you _MUST_
1843 * be careful with this one. A peek leaves the buffer on the
1844 * list and someone else may run off with it. You must hold
1845 * the appropriate locks or have a private queue to do this.
1846 *
1847 * Returns %NULL for an empty list or a pointer to the tail element.
1848 * The reference count is not incremented and the reference is therefore
1849 * volatile. Use with caution.
1850 */
05bdd2f1 1851static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1852{
f8cc62ca 1853 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1854
1855 if (skb == (struct sk_buff *)list_)
1856 skb = NULL;
1857 return skb;
1858
1da177e4
LT
1859}
1860
1861/**
1862 * skb_queue_len - get queue length
1863 * @list_: list to measure
1864 *
1865 * Return the length of an &sk_buff queue.
1866 */
1867static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1868{
1869 return list_->qlen;
1870}
1871
86b18aaa
QC
1872/**
1873 * skb_queue_len_lockless - get queue length
1874 * @list_: list to measure
1875 *
1876 * Return the length of an &sk_buff queue.
1877 * This variant can be used in lockless contexts.
1878 */
1879static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1880{
1881 return READ_ONCE(list_->qlen);
1882}
1883
67fed459
DM
1884/**
1885 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1886 * @list: queue to initialize
1887 *
1888 * This initializes only the list and queue length aspects of
1889 * an sk_buff_head object. This allows to initialize the list
1890 * aspects of an sk_buff_head without reinitializing things like
1891 * the spinlock. It can also be used for on-stack sk_buff_head
1892 * objects where the spinlock is known to not be used.
1893 */
1894static inline void __skb_queue_head_init(struct sk_buff_head *list)
1895{
1896 list->prev = list->next = (struct sk_buff *)list;
1897 list->qlen = 0;
1898}
1899
76f10ad0
AV
1900/*
1901 * This function creates a split out lock class for each invocation;
1902 * this is needed for now since a whole lot of users of the skb-queue
1903 * infrastructure in drivers have different locking usage (in hardirq)
1904 * than the networking core (in softirq only). In the long run either the
1905 * network layer or drivers should need annotation to consolidate the
1906 * main types of usage into 3 classes.
1907 */
1da177e4
LT
1908static inline void skb_queue_head_init(struct sk_buff_head *list)
1909{
1910 spin_lock_init(&list->lock);
67fed459 1911 __skb_queue_head_init(list);
1da177e4
LT
1912}
1913
c2ecba71
PE
1914static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1915 struct lock_class_key *class)
1916{
1917 skb_queue_head_init(list);
1918 lockdep_set_class(&list->lock, class);
1919}
1920
1da177e4 1921/*
bf299275 1922 * Insert an sk_buff on a list.
1da177e4
LT
1923 *
1924 * The "__skb_xxxx()" functions are the non-atomic ones that
1925 * can only be called with interrupts disabled.
1926 */
bf299275
GR
1927static inline void __skb_insert(struct sk_buff *newsk,
1928 struct sk_buff *prev, struct sk_buff *next,
1929 struct sk_buff_head *list)
1930{
f8cc62ca
ED
1931 /* See skb_queue_empty_lockless() and skb_peek_tail()
1932 * for the opposite READ_ONCE()
1933 */
d7d16a89
ED
1934 WRITE_ONCE(newsk->next, next);
1935 WRITE_ONCE(newsk->prev, prev);
1936 WRITE_ONCE(next->prev, newsk);
1937 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1938 list->qlen++;
1939}
1da177e4 1940
67fed459
DM
1941static inline void __skb_queue_splice(const struct sk_buff_head *list,
1942 struct sk_buff *prev,
1943 struct sk_buff *next)
1944{
1945 struct sk_buff *first = list->next;
1946 struct sk_buff *last = list->prev;
1947
d7d16a89
ED
1948 WRITE_ONCE(first->prev, prev);
1949 WRITE_ONCE(prev->next, first);
67fed459 1950
d7d16a89
ED
1951 WRITE_ONCE(last->next, next);
1952 WRITE_ONCE(next->prev, last);
67fed459
DM
1953}
1954
1955/**
1956 * skb_queue_splice - join two skb lists, this is designed for stacks
1957 * @list: the new list to add
1958 * @head: the place to add it in the first list
1959 */
1960static inline void skb_queue_splice(const struct sk_buff_head *list,
1961 struct sk_buff_head *head)
1962{
1963 if (!skb_queue_empty(list)) {
1964 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1965 head->qlen += list->qlen;
67fed459
DM
1966 }
1967}
1968
1969/**
d9619496 1970 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1971 * @list: the new list to add
1972 * @head: the place to add it in the first list
1973 *
1974 * The list at @list is reinitialised
1975 */
1976static inline void skb_queue_splice_init(struct sk_buff_head *list,
1977 struct sk_buff_head *head)
1978{
1979 if (!skb_queue_empty(list)) {
1980 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1981 head->qlen += list->qlen;
67fed459
DM
1982 __skb_queue_head_init(list);
1983 }
1984}
1985
1986/**
1987 * skb_queue_splice_tail - join two skb lists, each list being a queue
1988 * @list: the new list to add
1989 * @head: the place to add it in the first list
1990 */
1991static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1992 struct sk_buff_head *head)
1993{
1994 if (!skb_queue_empty(list)) {
1995 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1996 head->qlen += list->qlen;
67fed459
DM
1997 }
1998}
1999
2000/**
d9619496 2001 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2002 * @list: the new list to add
2003 * @head: the place to add it in the first list
2004 *
2005 * Each of the lists is a queue.
2006 * The list at @list is reinitialised
2007 */
2008static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2009 struct sk_buff_head *head)
2010{
2011 if (!skb_queue_empty(list)) {
2012 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2013 head->qlen += list->qlen;
67fed459
DM
2014 __skb_queue_head_init(list);
2015 }
2016}
2017
1da177e4 2018/**
300ce174 2019 * __skb_queue_after - queue a buffer at the list head
1da177e4 2020 * @list: list to use
300ce174 2021 * @prev: place after this buffer
1da177e4
LT
2022 * @newsk: buffer to queue
2023 *
300ce174 2024 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2025 * and you must therefore hold required locks before calling it.
2026 *
2027 * A buffer cannot be placed on two lists at the same time.
2028 */
300ce174
SH
2029static inline void __skb_queue_after(struct sk_buff_head *list,
2030 struct sk_buff *prev,
2031 struct sk_buff *newsk)
1da177e4 2032{
bf299275 2033 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
2034}
2035
7965bd4d
JP
2036void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2037 struct sk_buff_head *list);
7de6c033 2038
f5572855
GR
2039static inline void __skb_queue_before(struct sk_buff_head *list,
2040 struct sk_buff *next,
2041 struct sk_buff *newsk)
2042{
2043 __skb_insert(newsk, next->prev, next, list);
2044}
2045
300ce174
SH
2046/**
2047 * __skb_queue_head - queue a buffer at the list head
2048 * @list: list to use
2049 * @newsk: buffer to queue
2050 *
2051 * Queue a buffer at the start of a list. This function takes no locks
2052 * and you must therefore hold required locks before calling it.
2053 *
2054 * A buffer cannot be placed on two lists at the same time.
2055 */
300ce174
SH
2056static inline void __skb_queue_head(struct sk_buff_head *list,
2057 struct sk_buff *newsk)
2058{
2059 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2060}
4ea7b0cf 2061void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2062
1da177e4
LT
2063/**
2064 * __skb_queue_tail - queue a buffer at the list tail
2065 * @list: list to use
2066 * @newsk: buffer to queue
2067 *
2068 * Queue a buffer at the end of a list. This function takes no locks
2069 * and you must therefore hold required locks before calling it.
2070 *
2071 * A buffer cannot be placed on two lists at the same time.
2072 */
1da177e4
LT
2073static inline void __skb_queue_tail(struct sk_buff_head *list,
2074 struct sk_buff *newsk)
2075{
f5572855 2076 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2077}
4ea7b0cf 2078void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2079
1da177e4
LT
2080/*
2081 * remove sk_buff from list. _Must_ be called atomically, and with
2082 * the list known..
2083 */
7965bd4d 2084void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2085static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2086{
2087 struct sk_buff *next, *prev;
2088
86b18aaa 2089 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2090 next = skb->next;
2091 prev = skb->prev;
2092 skb->next = skb->prev = NULL;
d7d16a89
ED
2093 WRITE_ONCE(next->prev, prev);
2094 WRITE_ONCE(prev->next, next);
1da177e4
LT
2095}
2096
f525c06d
GR
2097/**
2098 * __skb_dequeue - remove from the head of the queue
2099 * @list: list to dequeue from
2100 *
2101 * Remove the head of the list. This function does not take any locks
2102 * so must be used with appropriate locks held only. The head item is
2103 * returned or %NULL if the list is empty.
2104 */
f525c06d
GR
2105static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2106{
2107 struct sk_buff *skb = skb_peek(list);
2108 if (skb)
2109 __skb_unlink(skb, list);
2110 return skb;
2111}
4ea7b0cf 2112struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2113
2114/**
2115 * __skb_dequeue_tail - remove from the tail of the queue
2116 * @list: list to dequeue from
2117 *
2118 * Remove the tail of the list. This function does not take any locks
2119 * so must be used with appropriate locks held only. The tail item is
2120 * returned or %NULL if the list is empty.
2121 */
1da177e4
LT
2122static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2123{
2124 struct sk_buff *skb = skb_peek_tail(list);
2125 if (skb)
2126 __skb_unlink(skb, list);
2127 return skb;
2128}
4ea7b0cf 2129struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2130
2131
bdcc0924 2132static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2133{
2134 return skb->data_len;
2135}
2136
2137static inline unsigned int skb_headlen(const struct sk_buff *skb)
2138{
2139 return skb->len - skb->data_len;
2140}
2141
3ece7826 2142static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2143{
c72d8cda 2144 unsigned int i, len = 0;
1da177e4 2145
c72d8cda 2146 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2147 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2148 return len;
2149}
2150
2151static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2152{
2153 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2154}
2155
131ea667
IC
2156/**
2157 * __skb_fill_page_desc - initialise a paged fragment in an skb
2158 * @skb: buffer containing fragment to be initialised
2159 * @i: paged fragment index to initialise
2160 * @page: the page to use for this fragment
2161 * @off: the offset to the data with @page
2162 * @size: the length of the data
2163 *
2164 * Initialises the @i'th fragment of @skb to point to &size bytes at
2165 * offset @off within @page.
2166 *
2167 * Does not take any additional reference on the fragment.
2168 */
2169static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2170 struct page *page, int off, int size)
1da177e4
LT
2171{
2172 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2173
c48a11c7 2174 /*
2f064f34
MH
2175 * Propagate page pfmemalloc to the skb if we can. The problem is
2176 * that not all callers have unique ownership of the page but rely
2177 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2178 */
1dfa5bd3 2179 frag->bv_page = page;
65c84f14 2180 frag->bv_offset = off;
9e903e08 2181 skb_frag_size_set(frag, size);
cca7af38
PE
2182
2183 page = compound_head(page);
2f064f34 2184 if (page_is_pfmemalloc(page))
cca7af38 2185 skb->pfmemalloc = true;
131ea667
IC
2186}
2187
2188/**
2189 * skb_fill_page_desc - initialise a paged fragment in an skb
2190 * @skb: buffer containing fragment to be initialised
2191 * @i: paged fragment index to initialise
2192 * @page: the page to use for this fragment
2193 * @off: the offset to the data with @page
2194 * @size: the length of the data
2195 *
2196 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2197 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2198 * addition updates @skb such that @i is the last fragment.
2199 *
2200 * Does not take any additional reference on the fragment.
2201 */
2202static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2203 struct page *page, int off, int size)
2204{
2205 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2206 skb_shinfo(skb)->nr_frags = i + 1;
2207}
2208
7965bd4d
JP
2209void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2210 int size, unsigned int truesize);
654bed16 2211
f8e617e1
JW
2212void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2213 unsigned int truesize);
2214
1da177e4
LT
2215#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2216
27a884dc
ACM
2217#ifdef NET_SKBUFF_DATA_USES_OFFSET
2218static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2219{
2220 return skb->head + skb->tail;
2221}
2222
2223static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2224{
2225 skb->tail = skb->data - skb->head;
2226}
2227
2228static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2229{
2230 skb_reset_tail_pointer(skb);
2231 skb->tail += offset;
2232}
7cc46190 2233
27a884dc
ACM
2234#else /* NET_SKBUFF_DATA_USES_OFFSET */
2235static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2236{
2237 return skb->tail;
2238}
2239
2240static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2241{
2242 skb->tail = skb->data;
2243}
2244
2245static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2246{
2247 skb->tail = skb->data + offset;
2248}
4305b541 2249
27a884dc
ACM
2250#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2251
1da177e4
LT
2252/*
2253 * Add data to an sk_buff
2254 */
4df864c1
JB
2255void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2256void *skb_put(struct sk_buff *skb, unsigned int len);
2257static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2258{
4df864c1 2259 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2260 SKB_LINEAR_ASSERT(skb);
2261 skb->tail += len;
2262 skb->len += len;
2263 return tmp;
2264}
2265
de77b966 2266static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2267{
2268 void *tmp = __skb_put(skb, len);
2269
2270 memset(tmp, 0, len);
2271 return tmp;
2272}
2273
2274static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2275 unsigned int len)
2276{
2277 void *tmp = __skb_put(skb, len);
2278
2279 memcpy(tmp, data, len);
2280 return tmp;
2281}
2282
2283static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2284{
2285 *(u8 *)__skb_put(skb, 1) = val;
2286}
2287
83ad357d 2288static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2289{
83ad357d 2290 void *tmp = skb_put(skb, len);
e45a79da
JB
2291
2292 memset(tmp, 0, len);
2293
2294 return tmp;
2295}
2296
59ae1d12
JB
2297static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2298 unsigned int len)
2299{
2300 void *tmp = skb_put(skb, len);
2301
2302 memcpy(tmp, data, len);
2303
2304 return tmp;
2305}
2306
634fef61
JB
2307static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2308{
2309 *(u8 *)skb_put(skb, 1) = val;
2310}
2311
d58ff351
JB
2312void *skb_push(struct sk_buff *skb, unsigned int len);
2313static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2314{
2315 skb->data -= len;
2316 skb->len += len;
2317 return skb->data;
2318}
2319
af72868b
JB
2320void *skb_pull(struct sk_buff *skb, unsigned int len);
2321static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2322{
2323 skb->len -= len;
2324 BUG_ON(skb->len < skb->data_len);
2325 return skb->data += len;
2326}
2327
af72868b 2328static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2329{
2330 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2331}
2332
af72868b 2333void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2334
af72868b 2335static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2336{
2337 if (len > skb_headlen(skb) &&
987c402a 2338 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2339 return NULL;
2340 skb->len -= len;
2341 return skb->data += len;
2342}
2343
af72868b 2344static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2345{
2346 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2347}
2348
b9df4fd7 2349static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2350{
2351 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2352 return true;
1da177e4 2353 if (unlikely(len > skb->len))
b9df4fd7 2354 return false;
987c402a 2355 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2356}
2357
c8c8b127
ED
2358void skb_condense(struct sk_buff *skb);
2359
1da177e4
LT
2360/**
2361 * skb_headroom - bytes at buffer head
2362 * @skb: buffer to check
2363 *
2364 * Return the number of bytes of free space at the head of an &sk_buff.
2365 */
c2636b4d 2366static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2367{
2368 return skb->data - skb->head;
2369}
2370
2371/**
2372 * skb_tailroom - bytes at buffer end
2373 * @skb: buffer to check
2374 *
2375 * Return the number of bytes of free space at the tail of an sk_buff
2376 */
2377static inline int skb_tailroom(const struct sk_buff *skb)
2378{
4305b541 2379 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2380}
2381
a21d4572
ED
2382/**
2383 * skb_availroom - bytes at buffer end
2384 * @skb: buffer to check
2385 *
2386 * Return the number of bytes of free space at the tail of an sk_buff
2387 * allocated by sk_stream_alloc()
2388 */
2389static inline int skb_availroom(const struct sk_buff *skb)
2390{
16fad69c
ED
2391 if (skb_is_nonlinear(skb))
2392 return 0;
2393
2394 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2395}
2396
1da177e4
LT
2397/**
2398 * skb_reserve - adjust headroom
2399 * @skb: buffer to alter
2400 * @len: bytes to move
2401 *
2402 * Increase the headroom of an empty &sk_buff by reducing the tail
2403 * room. This is only allowed for an empty buffer.
2404 */
8243126c 2405static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2406{
2407 skb->data += len;
2408 skb->tail += len;
2409}
2410
1837b2e2
BP
2411/**
2412 * skb_tailroom_reserve - adjust reserved_tailroom
2413 * @skb: buffer to alter
2414 * @mtu: maximum amount of headlen permitted
2415 * @needed_tailroom: minimum amount of reserved_tailroom
2416 *
2417 * Set reserved_tailroom so that headlen can be as large as possible but
2418 * not larger than mtu and tailroom cannot be smaller than
2419 * needed_tailroom.
2420 * The required headroom should already have been reserved before using
2421 * this function.
2422 */
2423static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2424 unsigned int needed_tailroom)
2425{
2426 SKB_LINEAR_ASSERT(skb);
2427 if (mtu < skb_tailroom(skb) - needed_tailroom)
2428 /* use at most mtu */
2429 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2430 else
2431 /* use up to all available space */
2432 skb->reserved_tailroom = needed_tailroom;
2433}
2434
8bce6d7d
TH
2435#define ENCAP_TYPE_ETHER 0
2436#define ENCAP_TYPE_IPPROTO 1
2437
2438static inline void skb_set_inner_protocol(struct sk_buff *skb,
2439 __be16 protocol)
2440{
2441 skb->inner_protocol = protocol;
2442 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2443}
2444
2445static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2446 __u8 ipproto)
2447{
2448 skb->inner_ipproto = ipproto;
2449 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2450}
2451
6a674e9c
JG
2452static inline void skb_reset_inner_headers(struct sk_buff *skb)
2453{
aefbd2b3 2454 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2455 skb->inner_network_header = skb->network_header;
2456 skb->inner_transport_header = skb->transport_header;
2457}
2458
0b5c9db1
JP
2459static inline void skb_reset_mac_len(struct sk_buff *skb)
2460{
2461 skb->mac_len = skb->network_header - skb->mac_header;
2462}
2463
6a674e9c
JG
2464static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2465 *skb)
2466{
2467 return skb->head + skb->inner_transport_header;
2468}
2469
55dc5a9f
TH
2470static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2471{
2472 return skb_inner_transport_header(skb) - skb->data;
2473}
2474
6a674e9c
JG
2475static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2476{
2477 skb->inner_transport_header = skb->data - skb->head;
2478}
2479
2480static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2481 const int offset)
2482{
2483 skb_reset_inner_transport_header(skb);
2484 skb->inner_transport_header += offset;
2485}
2486
2487static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2488{
2489 return skb->head + skb->inner_network_header;
2490}
2491
2492static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2493{
2494 skb->inner_network_header = skb->data - skb->head;
2495}
2496
2497static inline void skb_set_inner_network_header(struct sk_buff *skb,
2498 const int offset)
2499{
2500 skb_reset_inner_network_header(skb);
2501 skb->inner_network_header += offset;
2502}
2503
aefbd2b3
PS
2504static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2505{
2506 return skb->head + skb->inner_mac_header;
2507}
2508
2509static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2510{
2511 skb->inner_mac_header = skb->data - skb->head;
2512}
2513
2514static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2515 const int offset)
2516{
2517 skb_reset_inner_mac_header(skb);
2518 skb->inner_mac_header += offset;
2519}
fda55eca
ED
2520static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2521{
35d04610 2522 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2523}
2524
9c70220b
ACM
2525static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2526{
2e07fa9c 2527 return skb->head + skb->transport_header;
9c70220b
ACM
2528}
2529
badff6d0
ACM
2530static inline void skb_reset_transport_header(struct sk_buff *skb)
2531{
2e07fa9c 2532 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2533}
2534
967b05f6
ACM
2535static inline void skb_set_transport_header(struct sk_buff *skb,
2536 const int offset)
2537{
2e07fa9c
ACM
2538 skb_reset_transport_header(skb);
2539 skb->transport_header += offset;
ea2ae17d
ACM
2540}
2541
d56f90a7
ACM
2542static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2543{
2e07fa9c 2544 return skb->head + skb->network_header;
d56f90a7
ACM
2545}
2546
c1d2bbe1
ACM
2547static inline void skb_reset_network_header(struct sk_buff *skb)
2548{
2e07fa9c 2549 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2550}
2551
c14d2450
ACM
2552static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2553{
2e07fa9c
ACM
2554 skb_reset_network_header(skb);
2555 skb->network_header += offset;
c14d2450
ACM
2556}
2557
2e07fa9c 2558static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2559{
2e07fa9c 2560 return skb->head + skb->mac_header;
bbe735e4
ACM
2561}
2562
ea6da4fd
AV
2563static inline int skb_mac_offset(const struct sk_buff *skb)
2564{
2565 return skb_mac_header(skb) - skb->data;
2566}
2567
0daf4349
DB
2568static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2569{
2570 return skb->network_header - skb->mac_header;
2571}
2572
2e07fa9c 2573static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2574{
35d04610 2575 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2576}
2577
b4ab3141
DB
2578static inline void skb_unset_mac_header(struct sk_buff *skb)
2579{
2580 skb->mac_header = (typeof(skb->mac_header))~0U;
2581}
2582
2e07fa9c
ACM
2583static inline void skb_reset_mac_header(struct sk_buff *skb)
2584{
2585 skb->mac_header = skb->data - skb->head;
2586}
2587
2588static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2589{
2590 skb_reset_mac_header(skb);
2591 skb->mac_header += offset;
2592}
2593
0e3da5bb
TT
2594static inline void skb_pop_mac_header(struct sk_buff *skb)
2595{
2596 skb->mac_header = skb->network_header;
2597}
2598
d2aa125d 2599static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2600{
72a338bc 2601 struct flow_keys_basic keys;
fbbdb8f0
YX
2602
2603 if (skb_transport_header_was_set(skb))
2604 return;
72a338bc 2605
3cbf4ffb
SF
2606 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2607 NULL, 0, 0, 0, 0))
42aecaa9 2608 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2609}
2610
03606895
ED
2611static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2612{
2613 if (skb_mac_header_was_set(skb)) {
2614 const unsigned char *old_mac = skb_mac_header(skb);
2615
2616 skb_set_mac_header(skb, -skb->mac_len);
2617 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2618 }
2619}
2620
04fb451e
MM
2621static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2622{
2623 return skb->csum_start - skb_headroom(skb);
2624}
2625
08b64fcc
AD
2626static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2627{
2628 return skb->head + skb->csum_start;
2629}
2630
2e07fa9c
ACM
2631static inline int skb_transport_offset(const struct sk_buff *skb)
2632{
2633 return skb_transport_header(skb) - skb->data;
2634}
2635
2636static inline u32 skb_network_header_len(const struct sk_buff *skb)
2637{
2638 return skb->transport_header - skb->network_header;
2639}
2640
6a674e9c
JG
2641static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2642{
2643 return skb->inner_transport_header - skb->inner_network_header;
2644}
2645
2e07fa9c
ACM
2646static inline int skb_network_offset(const struct sk_buff *skb)
2647{
2648 return skb_network_header(skb) - skb->data;
2649}
48d49d0c 2650
6a674e9c
JG
2651static inline int skb_inner_network_offset(const struct sk_buff *skb)
2652{
2653 return skb_inner_network_header(skb) - skb->data;
2654}
2655
f9599ce1
CG
2656static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2657{
2658 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2659}
2660
1da177e4
LT
2661/*
2662 * CPUs often take a performance hit when accessing unaligned memory
2663 * locations. The actual performance hit varies, it can be small if the
2664 * hardware handles it or large if we have to take an exception and fix it
2665 * in software.
2666 *
2667 * Since an ethernet header is 14 bytes network drivers often end up with
2668 * the IP header at an unaligned offset. The IP header can be aligned by
2669 * shifting the start of the packet by 2 bytes. Drivers should do this
2670 * with:
2671 *
8660c124 2672 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2673 *
2674 * The downside to this alignment of the IP header is that the DMA is now
2675 * unaligned. On some architectures the cost of an unaligned DMA is high
2676 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2677 *
1da177e4
LT
2678 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2679 * to be overridden.
2680 */
2681#ifndef NET_IP_ALIGN
2682#define NET_IP_ALIGN 2
2683#endif
2684
025be81e
AB
2685/*
2686 * The networking layer reserves some headroom in skb data (via
2687 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2688 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2689 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2690 *
2691 * Unfortunately this headroom changes the DMA alignment of the resulting
2692 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2693 * on some architectures. An architecture can override this value,
2694 * perhaps setting it to a cacheline in size (since that will maintain
2695 * cacheline alignment of the DMA). It must be a power of 2.
2696 *
d6301d3d 2697 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2698 * headroom, you should not reduce this.
5933dd2f
ED
2699 *
2700 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2701 * to reduce average number of cache lines per packet.
645f0897 2702 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2703 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2704 */
2705#ifndef NET_SKB_PAD
5933dd2f 2706#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2707#endif
2708
7965bd4d 2709int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2710
5293efe6 2711static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2712{
5e1abdc3 2713 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2714 return;
27a884dc
ACM
2715 skb->len = len;
2716 skb_set_tail_pointer(skb, len);
1da177e4
LT
2717}
2718
5293efe6
DB
2719static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2720{
2721 __skb_set_length(skb, len);
2722}
2723
7965bd4d 2724void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2725
2726static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2727{
3cc0e873
HX
2728 if (skb->data_len)
2729 return ___pskb_trim(skb, len);
2730 __skb_trim(skb, len);
2731 return 0;
1da177e4
LT
2732}
2733
2734static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2735{
2736 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2737}
2738
e9fa4f7b
HX
2739/**
2740 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2741 * @skb: buffer to alter
2742 * @len: new length
2743 *
2744 * This is identical to pskb_trim except that the caller knows that
2745 * the skb is not cloned so we should never get an error due to out-
2746 * of-memory.
2747 */
2748static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2749{
2750 int err = pskb_trim(skb, len);
2751 BUG_ON(err);
2752}
2753
5293efe6
DB
2754static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2755{
2756 unsigned int diff = len - skb->len;
2757
2758 if (skb_tailroom(skb) < diff) {
2759 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2760 GFP_ATOMIC);
2761 if (ret)
2762 return ret;
2763 }
2764 __skb_set_length(skb, len);
2765 return 0;
2766}
2767
1da177e4
LT
2768/**
2769 * skb_orphan - orphan a buffer
2770 * @skb: buffer to orphan
2771 *
2772 * If a buffer currently has an owner then we call the owner's
2773 * destructor function and make the @skb unowned. The buffer continues
2774 * to exist but is no longer charged to its former owner.
2775 */
2776static inline void skb_orphan(struct sk_buff *skb)
2777{
c34a7612 2778 if (skb->destructor) {
1da177e4 2779 skb->destructor(skb);
c34a7612
ED
2780 skb->destructor = NULL;
2781 skb->sk = NULL;
376c7311
ED
2782 } else {
2783 BUG_ON(skb->sk);
c34a7612 2784 }
1da177e4
LT
2785}
2786
a353e0ce
MT
2787/**
2788 * skb_orphan_frags - orphan the frags contained in a buffer
2789 * @skb: buffer to orphan frags from
2790 * @gfp_mask: allocation mask for replacement pages
2791 *
2792 * For each frag in the SKB which needs a destructor (i.e. has an
2793 * owner) create a copy of that frag and release the original
2794 * page by calling the destructor.
2795 */
2796static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2797{
1f8b977a
WB
2798 if (likely(!skb_zcopy(skb)))
2799 return 0;
185ce5c3 2800 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 2801 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
2802 return 0;
2803 return skb_copy_ubufs(skb, gfp_mask);
2804}
2805
2806/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2807static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2808{
2809 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2810 return 0;
2811 return skb_copy_ubufs(skb, gfp_mask);
2812}
2813
1da177e4
LT
2814/**
2815 * __skb_queue_purge - empty a list
2816 * @list: list to empty
2817 *
2818 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2819 * the list and one reference dropped. This function does not take the
2820 * list lock and the caller must hold the relevant locks to use it.
2821 */
1da177e4
LT
2822static inline void __skb_queue_purge(struct sk_buff_head *list)
2823{
2824 struct sk_buff *skb;
2825 while ((skb = __skb_dequeue(list)) != NULL)
2826 kfree_skb(skb);
2827}
4ea7b0cf 2828void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2829
385114de 2830unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2831
3f6e687d
KH
2832void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2833
2834/**
2835 * netdev_alloc_frag - allocate a page fragment
2836 * @fragsz: fragment size
2837 *
2838 * Allocates a frag from a page for receive buffer.
2839 * Uses GFP_ATOMIC allocations.
2840 */
2841static inline void *netdev_alloc_frag(unsigned int fragsz)
2842{
2843 return __netdev_alloc_frag_align(fragsz, ~0u);
2844}
2845
2846static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2847 unsigned int align)
2848{
2849 WARN_ON_ONCE(!is_power_of_2(align));
2850 return __netdev_alloc_frag_align(fragsz, -align);
2851}
1da177e4 2852
7965bd4d
JP
2853struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2854 gfp_t gfp_mask);
8af27456
CH
2855
2856/**
2857 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2858 * @dev: network device to receive on
2859 * @length: length to allocate
2860 *
2861 * Allocate a new &sk_buff and assign it a usage count of one. The
2862 * buffer has unspecified headroom built in. Users should allocate
2863 * the headroom they think they need without accounting for the
2864 * built in space. The built in space is used for optimisations.
2865 *
2866 * %NULL is returned if there is no free memory. Although this function
2867 * allocates memory it can be called from an interrupt.
2868 */
2869static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2870 unsigned int length)
8af27456
CH
2871{
2872 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2873}
2874
6f532612
ED
2875/* legacy helper around __netdev_alloc_skb() */
2876static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2877 gfp_t gfp_mask)
2878{
2879 return __netdev_alloc_skb(NULL, length, gfp_mask);
2880}
2881
2882/* legacy helper around netdev_alloc_skb() */
2883static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2884{
2885 return netdev_alloc_skb(NULL, length);
2886}
2887
2888
4915a0de
ED
2889static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2890 unsigned int length, gfp_t gfp)
61321bbd 2891{
4915a0de 2892 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2893
2894 if (NET_IP_ALIGN && skb)
2895 skb_reserve(skb, NET_IP_ALIGN);
2896 return skb;
2897}
2898
4915a0de
ED
2899static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2900 unsigned int length)
2901{
2902 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2903}
2904
181edb2b
AD
2905static inline void skb_free_frag(void *addr)
2906{
8c2dd3e4 2907 page_frag_free(addr);
181edb2b
AD
2908}
2909
3f6e687d
KH
2910void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2911
2912static inline void *napi_alloc_frag(unsigned int fragsz)
2913{
2914 return __napi_alloc_frag_align(fragsz, ~0u);
2915}
2916
2917static inline void *napi_alloc_frag_align(unsigned int fragsz,
2918 unsigned int align)
2919{
2920 WARN_ON_ONCE(!is_power_of_2(align));
2921 return __napi_alloc_frag_align(fragsz, -align);
2922}
2923
fd11a83d
AD
2924struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2925 unsigned int length, gfp_t gfp_mask);
2926static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2927 unsigned int length)
2928{
2929 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2930}
795bb1c0
JDB
2931void napi_consume_skb(struct sk_buff *skb, int budget);
2932
9243adfc 2933void napi_skb_free_stolen_head(struct sk_buff *skb);
15fad714 2934void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2935
71dfda58
AD
2936/**
2937 * __dev_alloc_pages - allocate page for network Rx
2938 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2939 * @order: size of the allocation
2940 *
2941 * Allocate a new page.
2942 *
2943 * %NULL is returned if there is no free memory.
2944*/
2945static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2946 unsigned int order)
2947{
2948 /* This piece of code contains several assumptions.
2949 * 1. This is for device Rx, therefor a cold page is preferred.
2950 * 2. The expectation is the user wants a compound page.
2951 * 3. If requesting a order 0 page it will not be compound
2952 * due to the check to see if order has a value in prep_new_page
2953 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2954 * code in gfp_to_alloc_flags that should be enforcing this.
2955 */
453f85d4 2956 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2957
2958 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2959}
2960
2961static inline struct page *dev_alloc_pages(unsigned int order)
2962{
95829b3a 2963 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2964}
2965
2966/**
2967 * __dev_alloc_page - allocate a page for network Rx
2968 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2969 *
2970 * Allocate a new page.
2971 *
2972 * %NULL is returned if there is no free memory.
2973 */
2974static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2975{
2976 return __dev_alloc_pages(gfp_mask, 0);
2977}
2978
2979static inline struct page *dev_alloc_page(void)
2980{
95829b3a 2981 return dev_alloc_pages(0);
71dfda58
AD
2982}
2983
bc38f30f
AL
2984/**
2985 * dev_page_is_reusable - check whether a page can be reused for network Rx
2986 * @page: the page to test
2987 *
2988 * A page shouldn't be considered for reusing/recycling if it was allocated
2989 * under memory pressure or at a distant memory node.
2990 *
2991 * Returns false if this page should be returned to page allocator, true
2992 * otherwise.
2993 */
2994static inline bool dev_page_is_reusable(const struct page *page)
2995{
2996 return likely(page_to_nid(page) == numa_mem_id() &&
2997 !page_is_pfmemalloc(page));
2998}
2999
0614002b
MG
3000/**
3001 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3002 * @page: The page that was allocated from skb_alloc_page
3003 * @skb: The skb that may need pfmemalloc set
3004 */
48f971c9
AL
3005static inline void skb_propagate_pfmemalloc(const struct page *page,
3006 struct sk_buff *skb)
0614002b 3007{
2f064f34 3008 if (page_is_pfmemalloc(page))
0614002b
MG
3009 skb->pfmemalloc = true;
3010}
3011
7240b60c
JL
3012/**
3013 * skb_frag_off() - Returns the offset of a skb fragment
3014 * @frag: the paged fragment
3015 */
3016static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3017{
65c84f14 3018 return frag->bv_offset;
7240b60c
JL
3019}
3020
3021/**
3022 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3023 * @frag: skb fragment
3024 * @delta: value to add
3025 */
3026static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3027{
65c84f14 3028 frag->bv_offset += delta;
7240b60c
JL
3029}
3030
3031/**
3032 * skb_frag_off_set() - Sets the offset of a skb fragment
3033 * @frag: skb fragment
3034 * @offset: offset of fragment
3035 */
3036static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3037{
65c84f14 3038 frag->bv_offset = offset;
7240b60c
JL
3039}
3040
3041/**
3042 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3043 * @fragto: skb fragment where offset is set
3044 * @fragfrom: skb fragment offset is copied from
3045 */
3046static inline void skb_frag_off_copy(skb_frag_t *fragto,
3047 const skb_frag_t *fragfrom)
3048{
65c84f14 3049 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
3050}
3051
131ea667 3052/**
e227867f 3053 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
3054 * @frag: the paged fragment
3055 *
3056 * Returns the &struct page associated with @frag.
3057 */
3058static inline struct page *skb_frag_page(const skb_frag_t *frag)
3059{
1dfa5bd3 3060 return frag->bv_page;
131ea667
IC
3061}
3062
3063/**
3064 * __skb_frag_ref - take an addition reference on a paged fragment.
3065 * @frag: the paged fragment
3066 *
3067 * Takes an additional reference on the paged fragment @frag.
3068 */
3069static inline void __skb_frag_ref(skb_frag_t *frag)
3070{
3071 get_page(skb_frag_page(frag));
3072}
3073
3074/**
3075 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3076 * @skb: the buffer
3077 * @f: the fragment offset.
3078 *
3079 * Takes an additional reference on the @f'th paged fragment of @skb.
3080 */
3081static inline void skb_frag_ref(struct sk_buff *skb, int f)
3082{
3083 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3084}
3085
3086/**
3087 * __skb_frag_unref - release a reference on a paged fragment.
3088 * @frag: the paged fragment
c420c989 3089 * @recycle: recycle the page if allocated via page_pool
131ea667 3090 *
c420c989
MC
3091 * Releases a reference on the paged fragment @frag
3092 * or recycles the page via the page_pool API.
131ea667 3093 */
c420c989 3094static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
131ea667 3095{
6a5bcd84
IA
3096 struct page *page = skb_frag_page(frag);
3097
3098#ifdef CONFIG_PAGE_POOL
3099 if (recycle && page_pool_return_skb_page(page))
3100 return;
3101#endif
3102 put_page(page);
131ea667
IC
3103}
3104
3105/**
3106 * skb_frag_unref - release a reference on a paged fragment of an skb.
3107 * @skb: the buffer
3108 * @f: the fragment offset
3109 *
3110 * Releases a reference on the @f'th paged fragment of @skb.
3111 */
3112static inline void skb_frag_unref(struct sk_buff *skb, int f)
3113{
6a5bcd84 3114 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
131ea667
IC
3115}
3116
3117/**
3118 * skb_frag_address - gets the address of the data contained in a paged fragment
3119 * @frag: the paged fragment buffer
3120 *
3121 * Returns the address of the data within @frag. The page must already
3122 * be mapped.
3123 */
3124static inline void *skb_frag_address(const skb_frag_t *frag)
3125{
7240b60c 3126 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3127}
3128
3129/**
3130 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3131 * @frag: the paged fragment buffer
3132 *
3133 * Returns the address of the data within @frag. Checks that the page
3134 * is mapped and returns %NULL otherwise.
3135 */
3136static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3137{
3138 void *ptr = page_address(skb_frag_page(frag));
3139 if (unlikely(!ptr))
3140 return NULL;
3141
7240b60c
JL
3142 return ptr + skb_frag_off(frag);
3143}
3144
3145/**
3146 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3147 * @fragto: skb fragment where page is set
3148 * @fragfrom: skb fragment page is copied from
3149 */
3150static inline void skb_frag_page_copy(skb_frag_t *fragto,
3151 const skb_frag_t *fragfrom)
3152{
3153 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3154}
3155
3156/**
3157 * __skb_frag_set_page - sets the page contained in a paged fragment
3158 * @frag: the paged fragment
3159 * @page: the page to set
3160 *
3161 * Sets the fragment @frag to contain @page.
3162 */
3163static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3164{
1dfa5bd3 3165 frag->bv_page = page;
131ea667
IC
3166}
3167
3168/**
3169 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3170 * @skb: the buffer
3171 * @f: the fragment offset
3172 * @page: the page to set
3173 *
3174 * Sets the @f'th fragment of @skb to contain @page.
3175 */
3176static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3177 struct page *page)
3178{
3179 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3180}
3181
400dfd3a
ED
3182bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3183
131ea667
IC
3184/**
3185 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3186 * @dev: the device to map the fragment to
131ea667
IC
3187 * @frag: the paged fragment to map
3188 * @offset: the offset within the fragment (starting at the
3189 * fragment's own offset)
3190 * @size: the number of bytes to map
771b00a8 3191 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3192 *
3193 * Maps the page associated with @frag to @device.
3194 */
3195static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3196 const skb_frag_t *frag,
3197 size_t offset, size_t size,
3198 enum dma_data_direction dir)
3199{
3200 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3201 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3202}
3203
117632e6
ED
3204static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3205 gfp_t gfp_mask)
3206{
3207 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3208}
3209
bad93e9d
OP
3210
3211static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3212 gfp_t gfp_mask)
3213{
3214 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3215}
3216
3217
334a8132
PM
3218/**
3219 * skb_clone_writable - is the header of a clone writable
3220 * @skb: buffer to check
3221 * @len: length up to which to write
3222 *
3223 * Returns true if modifying the header part of the cloned buffer
3224 * does not requires the data to be copied.
3225 */
05bdd2f1 3226static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3227{
3228 return !skb_header_cloned(skb) &&
3229 skb_headroom(skb) + len <= skb->hdr_len;
3230}
3231
3697649f
DB
3232static inline int skb_try_make_writable(struct sk_buff *skb,
3233 unsigned int write_len)
3234{
3235 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3236 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3237}
3238
d9cc2048
HX
3239static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3240 int cloned)
3241{
3242 int delta = 0;
3243
d9cc2048
HX
3244 if (headroom > skb_headroom(skb))
3245 delta = headroom - skb_headroom(skb);
3246
3247 if (delta || cloned)
3248 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3249 GFP_ATOMIC);
3250 return 0;
3251}
3252
1da177e4
LT
3253/**
3254 * skb_cow - copy header of skb when it is required
3255 * @skb: buffer to cow
3256 * @headroom: needed headroom
3257 *
3258 * If the skb passed lacks sufficient headroom or its data part
3259 * is shared, data is reallocated. If reallocation fails, an error
3260 * is returned and original skb is not changed.
3261 *
3262 * The result is skb with writable area skb->head...skb->tail
3263 * and at least @headroom of space at head.
3264 */
3265static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3266{
d9cc2048
HX
3267 return __skb_cow(skb, headroom, skb_cloned(skb));
3268}
1da177e4 3269
d9cc2048
HX
3270/**
3271 * skb_cow_head - skb_cow but only making the head writable
3272 * @skb: buffer to cow
3273 * @headroom: needed headroom
3274 *
3275 * This function is identical to skb_cow except that we replace the
3276 * skb_cloned check by skb_header_cloned. It should be used when
3277 * you only need to push on some header and do not need to modify
3278 * the data.
3279 */
3280static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3281{
3282 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3283}
3284
3285/**
3286 * skb_padto - pad an skbuff up to a minimal size
3287 * @skb: buffer to pad
3288 * @len: minimal length
3289 *
3290 * Pads up a buffer to ensure the trailing bytes exist and are
3291 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3292 * is untouched. Otherwise it is extended. Returns zero on
3293 * success. The skb is freed on error.
1da177e4 3294 */
5b057c6b 3295static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3296{
3297 unsigned int size = skb->len;
3298 if (likely(size >= len))
5b057c6b 3299 return 0;
987c402a 3300 return skb_pad(skb, len - size);
1da177e4
LT
3301}
3302
9c0c1124 3303/**
4ea7b0cf 3304 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3305 * @skb: buffer to pad
3306 * @len: minimal length
cd0a137a 3307 * @free_on_error: free buffer on error
9c0c1124
AD
3308 *
3309 * Pads up a buffer to ensure the trailing bytes exist and are
3310 * blanked. If the buffer already contains sufficient data it
3311 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3312 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3313 */
4a009cb0
ED
3314static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3315 unsigned int len,
3316 bool free_on_error)
9c0c1124
AD
3317{
3318 unsigned int size = skb->len;
3319
3320 if (unlikely(size < len)) {
3321 len -= size;
cd0a137a 3322 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3323 return -ENOMEM;
3324 __skb_put(skb, len);
3325 }
3326 return 0;
3327}
3328
cd0a137a
FF
3329/**
3330 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3331 * @skb: buffer to pad
3332 * @len: minimal length
3333 *
3334 * Pads up a buffer to ensure the trailing bytes exist and are
3335 * blanked. If the buffer already contains sufficient data it
3336 * is untouched. Otherwise it is extended. Returns zero on
3337 * success. The skb is freed on error.
3338 */
4a009cb0 3339static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3340{
3341 return __skb_put_padto(skb, len, true);
3342}
3343
1da177e4 3344static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3345 struct iov_iter *from, int copy)
1da177e4
LT
3346{
3347 const int off = skb->len;
3348
3349 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3350 __wsum csum = 0;
15e6cb46
AV
3351 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3352 &csum, from)) {
1da177e4
LT
3353 skb->csum = csum_block_add(skb->csum, csum, off);
3354 return 0;
3355 }
15e6cb46 3356 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3357 return 0;
3358
3359 __skb_trim(skb, off);
3360 return -EFAULT;
3361}
3362
38ba0a65
ED
3363static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3364 const struct page *page, int off)
1da177e4 3365{
1f8b977a
WB
3366 if (skb_zcopy(skb))
3367 return false;
1da177e4 3368 if (i) {
d8e18a51 3369 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3370
ea2ab693 3371 return page == skb_frag_page(frag) &&
7240b60c 3372 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3373 }
38ba0a65 3374 return false;
1da177e4
LT
3375}
3376
364c6bad
HX
3377static inline int __skb_linearize(struct sk_buff *skb)
3378{
3379 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3380}
3381
1da177e4
LT
3382/**
3383 * skb_linearize - convert paged skb to linear one
3384 * @skb: buffer to linarize
1da177e4
LT
3385 *
3386 * If there is no free memory -ENOMEM is returned, otherwise zero
3387 * is returned and the old skb data released.
3388 */
364c6bad
HX
3389static inline int skb_linearize(struct sk_buff *skb)
3390{
3391 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3392}
3393
cef401de
ED
3394/**
3395 * skb_has_shared_frag - can any frag be overwritten
3396 * @skb: buffer to test
3397 *
3398 * Return true if the skb has at least one frag that might be modified
3399 * by an external entity (as in vmsplice()/sendfile())
3400 */
3401static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3402{
c9af6db4 3403 return skb_is_nonlinear(skb) &&
06b4feb3 3404 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3405}
3406
364c6bad
HX
3407/**
3408 * skb_linearize_cow - make sure skb is linear and writable
3409 * @skb: buffer to process
3410 *
3411 * If there is no free memory -ENOMEM is returned, otherwise zero
3412 * is returned and the old skb data released.
3413 */
3414static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3415{
364c6bad
HX
3416 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3417 __skb_linearize(skb) : 0;
1da177e4
LT
3418}
3419
479ffccc
DB
3420static __always_inline void
3421__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3422 unsigned int off)
3423{
3424 if (skb->ip_summed == CHECKSUM_COMPLETE)
3425 skb->csum = csum_block_sub(skb->csum,
3426 csum_partial(start, len, 0), off);
3427 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3428 skb_checksum_start_offset(skb) < 0)
3429 skb->ip_summed = CHECKSUM_NONE;
3430}
3431
1da177e4
LT
3432/**
3433 * skb_postpull_rcsum - update checksum for received skb after pull
3434 * @skb: buffer to update
3435 * @start: start of data before pull
3436 * @len: length of data pulled
3437 *
3438 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3439 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3440 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3441 */
1da177e4 3442static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3443 const void *start, unsigned int len)
1da177e4 3444{
479ffccc 3445 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3446}
3447
479ffccc
DB
3448static __always_inline void
3449__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3450 unsigned int off)
3451{
3452 if (skb->ip_summed == CHECKSUM_COMPLETE)
3453 skb->csum = csum_block_add(skb->csum,
3454 csum_partial(start, len, 0), off);
3455}
cbb042f9 3456
479ffccc
DB
3457/**
3458 * skb_postpush_rcsum - update checksum for received skb after push
3459 * @skb: buffer to update
3460 * @start: start of data after push
3461 * @len: length of data pushed
3462 *
3463 * After doing a push on a received packet, you need to call this to
3464 * update the CHECKSUM_COMPLETE checksum.
3465 */
f8ffad69
DB
3466static inline void skb_postpush_rcsum(struct sk_buff *skb,
3467 const void *start, unsigned int len)
3468{
479ffccc 3469 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3470}
3471
af72868b 3472void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3473
82a31b92
WC
3474/**
3475 * skb_push_rcsum - push skb and update receive checksum
3476 * @skb: buffer to update
3477 * @len: length of data pulled
3478 *
3479 * This function performs an skb_push on the packet and updates
3480 * the CHECKSUM_COMPLETE checksum. It should be used on
3481 * receive path processing instead of skb_push unless you know
3482 * that the checksum difference is zero (e.g., a valid IP header)
3483 * or you are setting ip_summed to CHECKSUM_NONE.
3484 */
d58ff351 3485static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3486{
3487 skb_push(skb, len);
3488 skb_postpush_rcsum(skb, skb->data, len);
3489 return skb->data;
3490}
3491
88078d98 3492int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3493/**
3494 * pskb_trim_rcsum - trim received skb and update checksum
3495 * @skb: buffer to trim
3496 * @len: new length
3497 *
3498 * This is exactly the same as pskb_trim except that it ensures the
3499 * checksum of received packets are still valid after the operation.
6c57f045 3500 * It can change skb pointers.
7ce5a27f
DM
3501 */
3502
3503static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3504{
3505 if (likely(len >= skb->len))
3506 return 0;
88078d98 3507 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3508}
3509
5293efe6
DB
3510static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3511{
3512 if (skb->ip_summed == CHECKSUM_COMPLETE)
3513 skb->ip_summed = CHECKSUM_NONE;
3514 __skb_trim(skb, len);
3515 return 0;
3516}
3517
3518static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3519{
3520 if (skb->ip_summed == CHECKSUM_COMPLETE)
3521 skb->ip_summed = CHECKSUM_NONE;
3522 return __skb_grow(skb, len);
3523}
3524
18a4c0ea
ED
3525#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3526#define skb_rb_first(root) rb_to_skb(rb_first(root))
3527#define skb_rb_last(root) rb_to_skb(rb_last(root))
3528#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3529#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3530
1da177e4
LT
3531#define skb_queue_walk(queue, skb) \
3532 for (skb = (queue)->next; \
a1e4891f 3533 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3534 skb = skb->next)
3535
46f8914e
JC
3536#define skb_queue_walk_safe(queue, skb, tmp) \
3537 for (skb = (queue)->next, tmp = skb->next; \
3538 skb != (struct sk_buff *)(queue); \
3539 skb = tmp, tmp = skb->next)
3540
1164f52a 3541#define skb_queue_walk_from(queue, skb) \
a1e4891f 3542 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3543 skb = skb->next)
3544
18a4c0ea
ED
3545#define skb_rbtree_walk(skb, root) \
3546 for (skb = skb_rb_first(root); skb != NULL; \
3547 skb = skb_rb_next(skb))
3548
3549#define skb_rbtree_walk_from(skb) \
3550 for (; skb != NULL; \
3551 skb = skb_rb_next(skb))
3552
3553#define skb_rbtree_walk_from_safe(skb, tmp) \
3554 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3555 skb = tmp)
3556
1164f52a
DM
3557#define skb_queue_walk_from_safe(queue, skb, tmp) \
3558 for (tmp = skb->next; \
3559 skb != (struct sk_buff *)(queue); \
3560 skb = tmp, tmp = skb->next)
3561
300ce174
SH
3562#define skb_queue_reverse_walk(queue, skb) \
3563 for (skb = (queue)->prev; \
a1e4891f 3564 skb != (struct sk_buff *)(queue); \
300ce174
SH
3565 skb = skb->prev)
3566
686a2955
DM
3567#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3568 for (skb = (queue)->prev, tmp = skb->prev; \
3569 skb != (struct sk_buff *)(queue); \
3570 skb = tmp, tmp = skb->prev)
3571
3572#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3573 for (tmp = skb->prev; \
3574 skb != (struct sk_buff *)(queue); \
3575 skb = tmp, tmp = skb->prev)
1da177e4 3576
21dc3301 3577static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3578{
3579 return skb_shinfo(skb)->frag_list != NULL;
3580}
3581
3582static inline void skb_frag_list_init(struct sk_buff *skb)
3583{
3584 skb_shinfo(skb)->frag_list = NULL;
3585}
3586
ee039871
DM
3587#define skb_walk_frags(skb, iter) \
3588 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3589
ea3793ee 3590
b50b0580
SD
3591int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3592 int *err, long *timeo_p,
ea3793ee 3593 const struct sk_buff *skb);
65101aec
PA
3594struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3595 struct sk_buff_head *queue,
3596 unsigned int flags,
fd69c399 3597 int *off, int *err,
65101aec 3598 struct sk_buff **last);
b50b0580
SD
3599struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3600 struct sk_buff_head *queue,
e427cad6 3601 unsigned int flags, int *off, int *err,
ea3793ee 3602 struct sk_buff **last);
b50b0580
SD
3603struct sk_buff *__skb_recv_datagram(struct sock *sk,
3604 struct sk_buff_head *sk_queue,
e427cad6 3605 unsigned int flags, int *off, int *err);
7965bd4d
JP
3606struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3607 int *err);
a11e1d43
LT
3608__poll_t datagram_poll(struct file *file, struct socket *sock,
3609 struct poll_table_struct *wait);
c0371da6
AV
3610int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3611 struct iov_iter *to, int size);
51f3d02b
DM
3612static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3613 struct msghdr *msg, int size)
3614{
e5a4b0bb 3615 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3616}
e5a4b0bb
AV
3617int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3618 struct msghdr *msg);
65d69e25
SG
3619int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3620 struct iov_iter *to, int len,
3621 struct ahash_request *hash);
3a654f97
AV
3622int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3623 struct iov_iter *from, int len);
3a654f97 3624int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3625void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3626void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3627static inline void skb_free_datagram_locked(struct sock *sk,
3628 struct sk_buff *skb)
3629{
3630 __skb_free_datagram_locked(sk, skb, 0);
3631}
7965bd4d 3632int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3633int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3634int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3635__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 3636 int len);
a60e3cc7 3637int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3638 struct pipe_inode_info *pipe, unsigned int len,
25869262 3639 unsigned int flags);
20bf50de
TH
3640int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3641 int len);
0739cd28 3642int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
7965bd4d 3643void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3644unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3645int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3646 int len, int hlen);
7965bd4d
JP
3647void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3648int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3649void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3650bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3651bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3652struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3653struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3654 unsigned int offset);
0d5501c1 3655struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3656int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3657int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3658int skb_vlan_pop(struct sk_buff *skb);
3659int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
3660int skb_eth_pop(struct sk_buff *skb);
3661int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3662 const unsigned char *src);
fa4e0f88 3663int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3664 int mac_len, bool ethernet);
040b5cfb
MV
3665int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3666 bool ethernet);
d27cf5c5 3667int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3668int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3669struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3670 gfp_t gfp);
20380731 3671
6ce8e9ce
AV
3672static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3673{
3073f070 3674 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3675}
3676
7eab8d9e
AV
3677static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3678{
e5a4b0bb 3679 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3680}
3681
2817a336
DB
3682struct skb_checksum_ops {
3683 __wsum (*update)(const void *mem, int len, __wsum wsum);
3684 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3685};
3686
9617813d
DC
3687extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3688
2817a336
DB
3689__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3690 __wsum csum, const struct skb_checksum_ops *ops);
3691__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3692 __wsum csum);
3693
1e98a0f0 3694static inline void * __must_check
e3305138
AL
3695__skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3696 const void *data, int hlen, void *buffer)
1da177e4 3697{
d206121f 3698 if (likely(hlen - offset >= len))
e3305138 3699 return (void *)data + offset;
1da177e4 3700
d206121f 3701 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
1da177e4
LT
3702 return NULL;
3703
3704 return buffer;
3705}
3706
1e98a0f0
ED
3707static inline void * __must_check
3708skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3709{
3710 return __skb_header_pointer(skb, offset, len, skb->data,
3711 skb_headlen(skb), buffer);
3712}
3713
4262e5cc
DB
3714/**
3715 * skb_needs_linearize - check if we need to linearize a given skb
3716 * depending on the given device features.
3717 * @skb: socket buffer to check
3718 * @features: net device features
3719 *
3720 * Returns true if either:
3721 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3722 * 2. skb is fragmented and the device does not support SG.
3723 */
3724static inline bool skb_needs_linearize(struct sk_buff *skb,
3725 netdev_features_t features)
3726{
3727 return skb_is_nonlinear(skb) &&
3728 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3729 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3730}
3731
d626f62b
ACM
3732static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3733 void *to,
3734 const unsigned int len)
3735{
3736 memcpy(to, skb->data, len);
3737}
3738
3739static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3740 const int offset, void *to,
3741 const unsigned int len)
3742{
3743 memcpy(to, skb->data + offset, len);
3744}
3745
27d7ff46
ACM
3746static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3747 const void *from,
3748 const unsigned int len)
3749{
3750 memcpy(skb->data, from, len);
3751}
3752
3753static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3754 const int offset,
3755 const void *from,
3756 const unsigned int len)
3757{
3758 memcpy(skb->data + offset, from, len);
3759}
3760
7965bd4d 3761void skb_init(void);
1da177e4 3762
ac45f602
PO
3763static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3764{
3765 return skb->tstamp;
3766}
3767
a61bbcf2
PM
3768/**
3769 * skb_get_timestamp - get timestamp from a skb
3770 * @skb: skb to get stamp from
13c6ee2a 3771 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3772 *
3773 * Timestamps are stored in the skb as offsets to a base timestamp.
3774 * This function converts the offset back to a struct timeval and stores
3775 * it in stamp.
3776 */
ac45f602 3777static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3778 struct __kernel_old_timeval *stamp)
a61bbcf2 3779{
13c6ee2a 3780 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3781}
3782
887feae3
DD
3783static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3784 struct __kernel_sock_timeval *stamp)
3785{
3786 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3787
3788 stamp->tv_sec = ts.tv_sec;
3789 stamp->tv_usec = ts.tv_nsec / 1000;
3790}
3791
ac45f602 3792static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3793 struct __kernel_old_timespec *stamp)
ac45f602 3794{
df1b4ba9
AB
3795 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3796
3797 stamp->tv_sec = ts.tv_sec;
3798 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3799}
3800
887feae3
DD
3801static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3802 struct __kernel_timespec *stamp)
3803{
3804 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3805
3806 stamp->tv_sec = ts.tv_sec;
3807 stamp->tv_nsec = ts.tv_nsec;
3808}
3809
b7aa0bf7 3810static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3811{
b7aa0bf7 3812 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3813}
3814
164891aa
SH
3815static inline ktime_t net_timedelta(ktime_t t)
3816{
3817 return ktime_sub(ktime_get_real(), t);
3818}
3819
b9ce204f
IJ
3820static inline ktime_t net_invalid_timestamp(void)
3821{
8b0e1953 3822 return 0;
b9ce204f 3823}
a61bbcf2 3824
de8f3a83
DB
3825static inline u8 skb_metadata_len(const struct sk_buff *skb)
3826{
3827 return skb_shinfo(skb)->meta_len;
3828}
3829
3830static inline void *skb_metadata_end(const struct sk_buff *skb)
3831{
3832 return skb_mac_header(skb);
3833}
3834
3835static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3836 const struct sk_buff *skb_b,
3837 u8 meta_len)
3838{
3839 const void *a = skb_metadata_end(skb_a);
3840 const void *b = skb_metadata_end(skb_b);
3841 /* Using more efficient varaiant than plain call to memcmp(). */
3842#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3843 u64 diffs = 0;
3844
3845 switch (meta_len) {
3846#define __it(x, op) (x -= sizeof(u##op))
3847#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3848 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3849 fallthrough;
de8f3a83 3850 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3851 fallthrough;
de8f3a83 3852 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3853 fallthrough;
de8f3a83
DB
3854 case 8: diffs |= __it_diff(a, b, 64);
3855 break;
3856 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3857 fallthrough;
de8f3a83 3858 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3859 fallthrough;
de8f3a83 3860 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3861 fallthrough;
de8f3a83
DB
3862 case 4: diffs |= __it_diff(a, b, 32);
3863 break;
3864 }
3865 return diffs;
3866#else
3867 return memcmp(a - meta_len, b - meta_len, meta_len);
3868#endif
3869}
3870
3871static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3872 const struct sk_buff *skb_b)
3873{
3874 u8 len_a = skb_metadata_len(skb_a);
3875 u8 len_b = skb_metadata_len(skb_b);
3876
3877 if (!(len_a | len_b))
3878 return false;
3879
3880 return len_a != len_b ?
3881 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3882}
3883
3884static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3885{
3886 skb_shinfo(skb)->meta_len = meta_len;
3887}
3888
3889static inline void skb_metadata_clear(struct sk_buff *skb)
3890{
3891 skb_metadata_set(skb, 0);
3892}
3893
62bccb8c
AD
3894struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3895
c1f19b51
RC
3896#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3897
7965bd4d
JP
3898void skb_clone_tx_timestamp(struct sk_buff *skb);
3899bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3900
3901#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3902
3903static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3904{
3905}
3906
3907static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3908{
3909 return false;
3910}
3911
3912#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3913
3914/**
3915 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3916 *
da92b194
RC
3917 * PHY drivers may accept clones of transmitted packets for
3918 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3919 * must call this function to return the skb back to the stack with a
3920 * timestamp.
da92b194 3921 *
2ff17117 3922 * @skb: clone of the original outgoing packet
7a76a021 3923 * @hwtstamps: hardware time stamps
c1f19b51
RC
3924 *
3925 */
3926void skb_complete_tx_timestamp(struct sk_buff *skb,
3927 struct skb_shared_hwtstamps *hwtstamps);
3928
e7ed11ee 3929void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
3930 struct skb_shared_hwtstamps *hwtstamps,
3931 struct sock *sk, int tstype);
3932
ac45f602
PO
3933/**
3934 * skb_tstamp_tx - queue clone of skb with send time stamps
3935 * @orig_skb: the original outgoing packet
3936 * @hwtstamps: hardware time stamps, may be NULL if not available
3937 *
3938 * If the skb has a socket associated, then this function clones the
3939 * skb (thus sharing the actual data and optional structures), stores
3940 * the optional hardware time stamping information (if non NULL) or
3941 * generates a software time stamp (otherwise), then queues the clone
3942 * to the error queue of the socket. Errors are silently ignored.
3943 */
7965bd4d
JP
3944void skb_tstamp_tx(struct sk_buff *orig_skb,
3945 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3946
4507a715
RC
3947/**
3948 * skb_tx_timestamp() - Driver hook for transmit timestamping
3949 *
3950 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3951 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3952 *
73409f3b
DM
3953 * Specifically, one should make absolutely sure that this function is
3954 * called before TX completion of this packet can trigger. Otherwise
3955 * the packet could potentially already be freed.
3956 *
4507a715
RC
3957 * @skb: A socket buffer.
3958 */
3959static inline void skb_tx_timestamp(struct sk_buff *skb)
3960{
c1f19b51 3961 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3962 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3963 skb_tstamp_tx(skb, NULL);
4507a715
RC
3964}
3965
6e3e939f
JB
3966/**
3967 * skb_complete_wifi_ack - deliver skb with wifi status
3968 *
3969 * @skb: the original outgoing packet
3970 * @acked: ack status
3971 *
3972 */
3973void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3974
7965bd4d
JP
3975__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3976__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3977
60476372
HX
3978static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3979{
6edec0e6
TH
3980 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3981 skb->csum_valid ||
3982 (skb->ip_summed == CHECKSUM_PARTIAL &&
3983 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3984}
3985
fb286bb2
HX
3986/**
3987 * skb_checksum_complete - Calculate checksum of an entire packet
3988 * @skb: packet to process
3989 *
3990 * This function calculates the checksum over the entire packet plus
3991 * the value of skb->csum. The latter can be used to supply the
3992 * checksum of a pseudo header as used by TCP/UDP. It returns the
3993 * checksum.
3994 *
3995 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3996 * this function can be used to verify that checksum on received
3997 * packets. In that case the function should return zero if the
3998 * checksum is correct. In particular, this function will return zero
3999 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4000 * hardware has already verified the correctness of the checksum.
4001 */
4381ca3c 4002static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 4003{
60476372
HX
4004 return skb_csum_unnecessary(skb) ?
4005 0 : __skb_checksum_complete(skb);
fb286bb2
HX
4006}
4007
77cffe23
TH
4008static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4009{
4010 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4011 if (skb->csum_level == 0)
4012 skb->ip_summed = CHECKSUM_NONE;
4013 else
4014 skb->csum_level--;
4015 }
4016}
4017
4018static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4019{
4020 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4021 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4022 skb->csum_level++;
4023 } else if (skb->ip_summed == CHECKSUM_NONE) {
4024 skb->ip_summed = CHECKSUM_UNNECESSARY;
4025 skb->csum_level = 0;
4026 }
4027}
4028
836e66c2
DB
4029static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4030{
4031 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4032 skb->ip_summed = CHECKSUM_NONE;
4033 skb->csum_level = 0;
4034 }
4035}
4036
76ba0aae
TH
4037/* Check if we need to perform checksum complete validation.
4038 *
4039 * Returns true if checksum complete is needed, false otherwise
4040 * (either checksum is unnecessary or zero checksum is allowed).
4041 */
4042static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4043 bool zero_okay,
4044 __sum16 check)
4045{
5d0c2b95
TH
4046 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4047 skb->csum_valid = 1;
77cffe23 4048 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
4049 return false;
4050 }
4051
4052 return true;
4053}
4054
da279887 4055/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
4056 * in checksum_init.
4057 */
4058#define CHECKSUM_BREAK 76
4059
4e18b9ad
TH
4060/* Unset checksum-complete
4061 *
4062 * Unset checksum complete can be done when packet is being modified
4063 * (uncompressed for instance) and checksum-complete value is
4064 * invalidated.
4065 */
4066static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4067{
4068 if (skb->ip_summed == CHECKSUM_COMPLETE)
4069 skb->ip_summed = CHECKSUM_NONE;
4070}
4071
76ba0aae
TH
4072/* Validate (init) checksum based on checksum complete.
4073 *
4074 * Return values:
4075 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4076 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4077 * checksum is stored in skb->csum for use in __skb_checksum_complete
4078 * non-zero: value of invalid checksum
4079 *
4080 */
4081static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4082 bool complete,
4083 __wsum psum)
4084{
4085 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4086 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4087 skb->csum_valid = 1;
76ba0aae
TH
4088 return 0;
4089 }
4090 }
4091
4092 skb->csum = psum;
4093
5d0c2b95
TH
4094 if (complete || skb->len <= CHECKSUM_BREAK) {
4095 __sum16 csum;
4096
4097 csum = __skb_checksum_complete(skb);
4098 skb->csum_valid = !csum;
4099 return csum;
4100 }
76ba0aae
TH
4101
4102 return 0;
4103}
4104
4105static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4106{
4107 return 0;
4108}
4109
4110/* Perform checksum validate (init). Note that this is a macro since we only
4111 * want to calculate the pseudo header which is an input function if necessary.
4112 * First we try to validate without any computation (checksum unnecessary) and
4113 * then calculate based on checksum complete calling the function to compute
4114 * pseudo header.
4115 *
4116 * Return values:
4117 * 0: checksum is validated or try to in skb_checksum_complete
4118 * non-zero: value of invalid checksum
4119 */
4120#define __skb_checksum_validate(skb, proto, complete, \
4121 zero_okay, check, compute_pseudo) \
4122({ \
4123 __sum16 __ret = 0; \
5d0c2b95 4124 skb->csum_valid = 0; \
76ba0aae
TH
4125 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4126 __ret = __skb_checksum_validate_complete(skb, \
4127 complete, compute_pseudo(skb, proto)); \
4128 __ret; \
4129})
4130
4131#define skb_checksum_init(skb, proto, compute_pseudo) \
4132 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4133
4134#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4135 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4136
4137#define skb_checksum_validate(skb, proto, compute_pseudo) \
4138 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4139
4140#define skb_checksum_validate_zero_check(skb, proto, check, \
4141 compute_pseudo) \
096a4cfa 4142 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4143
4144#define skb_checksum_simple_validate(skb) \
4145 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4146
d96535a1
TH
4147static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4148{
219f1d79 4149 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4150}
4151
e4aa33ad 4152static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4153{
4154 skb->csum = ~pseudo;
4155 skb->ip_summed = CHECKSUM_COMPLETE;
4156}
4157
e4aa33ad 4158#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4159do { \
4160 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4161 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4162} while (0)
4163
15e2396d
TH
4164static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4165 u16 start, u16 offset)
4166{
4167 skb->ip_summed = CHECKSUM_PARTIAL;
4168 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4169 skb->csum_offset = offset - start;
4170}
4171
dcdc8994
TH
4172/* Update skbuf and packet to reflect the remote checksum offload operation.
4173 * When called, ptr indicates the starting point for skb->csum when
4174 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4175 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4176 */
4177static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4178 int start, int offset, bool nopartial)
dcdc8994
TH
4179{
4180 __wsum delta;
4181
15e2396d
TH
4182 if (!nopartial) {
4183 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4184 return;
4185 }
4186
dcdc8994
TH
4187 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4188 __skb_checksum_complete(skb);
4189 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4190 }
4191
4192 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4193
4194 /* Adjust skb->csum since we changed the packet */
4195 skb->csum = csum_add(skb->csum, delta);
4196}
4197
cb9c6836
FW
4198static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4199{
4200#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4201 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4202#else
4203 return NULL;
4204#endif
4205}
4206
261db6c2 4207static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4208{
261db6c2
JS
4209#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4210 return skb->_nfct;
4211#else
4212 return 0UL;
4213#endif
1da177e4 4214}
261db6c2
JS
4215
4216static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4217{
261db6c2
JS
4218#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4219 skb->_nfct = nfct;
2fc72c7b 4220#endif
261db6c2 4221}
df5042f4
FW
4222
4223#ifdef CONFIG_SKB_EXTENSIONS
4224enum skb_ext_id {
4225#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4226 SKB_EXT_BRIDGE_NF,
4165079b
FW
4227#endif
4228#ifdef CONFIG_XFRM
4229 SKB_EXT_SEC_PATH,
95a7233c
PB
4230#endif
4231#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4232 TC_SKB_EXT,
3ee17bc7
MM
4233#endif
4234#if IS_ENABLED(CONFIG_MPTCP)
4235 SKB_EXT_MPTCP,
df5042f4
FW
4236#endif
4237 SKB_EXT_NUM, /* must be last */
4238};
4239
4240/**
4241 * struct skb_ext - sk_buff extensions
4242 * @refcnt: 1 on allocation, deallocated on 0
4243 * @offset: offset to add to @data to obtain extension address
4244 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4245 * @data: start of extension data, variable sized
4246 *
4247 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4248 * to use 'u8' types while allowing up to 2kb worth of extension data.
4249 */
4250struct skb_ext {
4251 refcount_t refcnt;
4252 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4253 u8 chunks; /* same */
5c91aa1d 4254 char data[] __aligned(8);
df5042f4
FW
4255};
4256
4930f483 4257struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4258void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4259 struct skb_ext *ext);
df5042f4
FW
4260void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4261void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4262void __skb_ext_put(struct skb_ext *ext);
4263
4264static inline void skb_ext_put(struct sk_buff *skb)
4265{
4266 if (skb->active_extensions)
4267 __skb_ext_put(skb->extensions);
4268}
4269
df5042f4
FW
4270static inline void __skb_ext_copy(struct sk_buff *dst,
4271 const struct sk_buff *src)
4272{
4273 dst->active_extensions = src->active_extensions;
4274
4275 if (src->active_extensions) {
4276 struct skb_ext *ext = src->extensions;
4277
4278 refcount_inc(&ext->refcnt);
4279 dst->extensions = ext;
4280 }
4281}
4282
4283static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4284{
4285 skb_ext_put(dst);
4286 __skb_ext_copy(dst, src);
4287}
4288
4289static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4290{
4291 return !!ext->offset[i];
4292}
4293
4294static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4295{
4296 return skb->active_extensions & (1 << id);
4297}
4298
4299static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4300{
4301 if (skb_ext_exist(skb, id))
4302 __skb_ext_del(skb, id);
4303}
4304
4305static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4306{
4307 if (skb_ext_exist(skb, id)) {
4308 struct skb_ext *ext = skb->extensions;
4309
4310 return (void *)ext + (ext->offset[id] << 3);
4311 }
4312
4313 return NULL;
4314}
174e2381
FW
4315
4316static inline void skb_ext_reset(struct sk_buff *skb)
4317{
4318 if (unlikely(skb->active_extensions)) {
4319 __skb_ext_put(skb->extensions);
4320 skb->active_extensions = 0;
4321 }
4322}
677bf08c
FW
4323
4324static inline bool skb_has_extensions(struct sk_buff *skb)
4325{
4326 return unlikely(skb->active_extensions);
4327}
df5042f4
FW
4328#else
4329static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4330static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4331static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4332static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4333static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4334static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4335#endif /* CONFIG_SKB_EXTENSIONS */
4336
895b5c9f 4337static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4338{
5f79e0f9 4339#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4340 nf_conntrack_put(skb_nfct(skb));
4341 skb->_nfct = 0;
2fc72c7b 4342#endif
a193a4ab
PM
4343}
4344
124dff01
PM
4345static inline void nf_reset_trace(struct sk_buff *skb)
4346{
478b360a 4347#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4348 skb->nf_trace = 0;
4349#endif
a193a4ab
PM
4350}
4351
2b5ec1a5
YY
4352static inline void ipvs_reset(struct sk_buff *skb)
4353{
4354#if IS_ENABLED(CONFIG_IP_VS)
4355 skb->ipvs_property = 0;
4356#endif
4357}
4358
de8bda1d 4359/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4360static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4361 bool copy)
edda553c 4362{
5f79e0f9 4363#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4364 dst->_nfct = src->_nfct;
4365 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4366#endif
478b360a 4367#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4368 if (copy)
4369 dst->nf_trace = src->nf_trace;
478b360a 4370#endif
edda553c
YK
4371}
4372
e7ac05f3
YK
4373static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4374{
e7ac05f3 4375#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4376 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4377#endif
b1937227 4378 __nf_copy(dst, src, true);
e7ac05f3
YK
4379}
4380
984bc16c
JM
4381#ifdef CONFIG_NETWORK_SECMARK
4382static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4383{
4384 to->secmark = from->secmark;
4385}
4386
4387static inline void skb_init_secmark(struct sk_buff *skb)
4388{
4389 skb->secmark = 0;
4390}
4391#else
4392static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4393{ }
4394
4395static inline void skb_init_secmark(struct sk_buff *skb)
4396{ }
4397#endif
4398
7af8f4ca
FW
4399static inline int secpath_exists(const struct sk_buff *skb)
4400{
4401#ifdef CONFIG_XFRM
4165079b 4402 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4403#else
4404 return 0;
4405#endif
4406}
4407
574f7194
EB
4408static inline bool skb_irq_freeable(const struct sk_buff *skb)
4409{
4410 return !skb->destructor &&
7af8f4ca 4411 !secpath_exists(skb) &&
cb9c6836 4412 !skb_nfct(skb) &&
574f7194
EB
4413 !skb->_skb_refdst &&
4414 !skb_has_frag_list(skb);
4415}
4416
f25f4e44
PWJ
4417static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4418{
f25f4e44 4419 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4420}
4421
9247744e 4422static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4423{
4e3ab47a 4424 return skb->queue_mapping;
4e3ab47a
PE
4425}
4426
f25f4e44
PWJ
4427static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4428{
f25f4e44 4429 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4430}
4431
d5a9e24a
DM
4432static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4433{
4434 skb->queue_mapping = rx_queue + 1;
4435}
4436
9247744e 4437static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4438{
4439 return skb->queue_mapping - 1;
4440}
4441
9247744e 4442static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4443{
a02cec21 4444 return skb->queue_mapping != 0;
d5a9e24a
DM
4445}
4446
4ff06203
JA
4447static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4448{
4449 skb->dst_pending_confirm = val;
4450}
4451
4452static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4453{
4454 return skb->dst_pending_confirm != 0;
4455}
4456
2294be0f 4457static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4458{
0b3d8e08 4459#ifdef CONFIG_XFRM
4165079b 4460 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4461#else
def8b4fa 4462 return NULL;
def8b4fa 4463#endif
0b3d8e08 4464}
def8b4fa 4465
68c33163
PS
4466/* Keeps track of mac header offset relative to skb->head.
4467 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4468 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4469 * tunnel skb it points to outer mac header.
4470 * Keeps track of level of encapsulation of network headers.
4471 */
68c33163 4472struct skb_gso_cb {
802ab55a
AD
4473 union {
4474 int mac_offset;
4475 int data_offset;
4476 };
3347c960 4477 int encap_level;
76443456 4478 __wsum csum;
7e2b10c1 4479 __u16 csum_start;
68c33163 4480};
a08e7fd9
CZ
4481#define SKB_GSO_CB_OFFSET 32
4482#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4483
4484static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4485{
4486 return (skb_mac_header(inner_skb) - inner_skb->head) -
4487 SKB_GSO_CB(inner_skb)->mac_offset;
4488}
4489
1e2bd517
PS
4490static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4491{
4492 int new_headroom, headroom;
4493 int ret;
4494
4495 headroom = skb_headroom(skb);
4496 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4497 if (ret)
4498 return ret;
4499
4500 new_headroom = skb_headroom(skb);
4501 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4502 return 0;
4503}
4504
08b64fcc
AD
4505static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4506{
4507 /* Do not update partial checksums if remote checksum is enabled. */
4508 if (skb->remcsum_offload)
4509 return;
4510
4511 SKB_GSO_CB(skb)->csum = res;
4512 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4513}
4514
7e2b10c1
TH
4515/* Compute the checksum for a gso segment. First compute the checksum value
4516 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4517 * then add in skb->csum (checksum from csum_start to end of packet).
4518 * skb->csum and csum_start are then updated to reflect the checksum of the
4519 * resultant packet starting from the transport header-- the resultant checksum
4520 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4521 * header.
4522 */
4523static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4524{
76443456
AD
4525 unsigned char *csum_start = skb_transport_header(skb);
4526 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4527 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4528
76443456
AD
4529 SKB_GSO_CB(skb)->csum = res;
4530 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4531
76443456 4532 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4533}
4534
bdcc0924 4535static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4536{
4537 return skb_shinfo(skb)->gso_size;
4538}
4539
36a8f39e 4540/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4541static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4542{
4543 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4544}
4545
d02f51cb
DA
4546/* Note: Should be called only if skb_is_gso(skb) is true */
4547static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4548{
4549 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4550}
4551
4c3024de 4552/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4553static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4554{
4c3024de 4555 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4556}
4557
5293efe6
DB
4558static inline void skb_gso_reset(struct sk_buff *skb)
4559{
4560 skb_shinfo(skb)->gso_size = 0;
4561 skb_shinfo(skb)->gso_segs = 0;
4562 skb_shinfo(skb)->gso_type = 0;
4563}
4564
d02f51cb
DA
4565static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4566 u16 increment)
4567{
4568 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4569 return;
4570 shinfo->gso_size += increment;
4571}
4572
4573static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4574 u16 decrement)
4575{
4576 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4577 return;
4578 shinfo->gso_size -= decrement;
4579}
4580
7965bd4d 4581void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4582
4583static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4584{
4585 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4586 * wanted then gso_type will be set. */
05bdd2f1
ED
4587 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4588
b78462eb
AD
4589 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4590 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4591 __skb_warn_lro_forwarding(skb);
4592 return true;
4593 }
4594 return false;
4595}
4596
35fc92a9
HX
4597static inline void skb_forward_csum(struct sk_buff *skb)
4598{
4599 /* Unfortunately we don't support this one. Any brave souls? */
4600 if (skb->ip_summed == CHECKSUM_COMPLETE)
4601 skb->ip_summed = CHECKSUM_NONE;
4602}
4603
bc8acf2c
ED
4604/**
4605 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4606 * @skb: skb to check
4607 *
4608 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4609 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4610 * use this helper, to document places where we make this assertion.
4611 */
05bdd2f1 4612static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4613{
4614#ifdef DEBUG
4615 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4616#endif
4617}
4618
f35d9d8a 4619bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4620
ed1f50c3 4621int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4622struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4623 unsigned int transport_len,
4624 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4625
3a7c1ee4
AD
4626/**
4627 * skb_head_is_locked - Determine if the skb->head is locked down
4628 * @skb: skb to check
4629 *
4630 * The head on skbs build around a head frag can be removed if they are
4631 * not cloned. This function returns true if the skb head is locked down
4632 * due to either being allocated via kmalloc, or by being a clone with
4633 * multiple references to the head.
4634 */
4635static inline bool skb_head_is_locked(const struct sk_buff *skb)
4636{
4637 return !skb->head_frag || skb_cloned(skb);
4638}
fe6cc55f 4639
179bc67f
EC
4640/* Local Checksum Offload.
4641 * Compute outer checksum based on the assumption that the
4642 * inner checksum will be offloaded later.
d0dcde64 4643 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4644 * explanation of how this works.
179bc67f
EC
4645 * Fill in outer checksum adjustment (e.g. with sum of outer
4646 * pseudo-header) before calling.
4647 * Also ensure that inner checksum is in linear data area.
4648 */
4649static inline __wsum lco_csum(struct sk_buff *skb)
4650{
9e74a6da
AD
4651 unsigned char *csum_start = skb_checksum_start(skb);
4652 unsigned char *l4_hdr = skb_transport_header(skb);
4653 __wsum partial;
179bc67f
EC
4654
4655 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4656 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4657 skb->csum_offset));
4658
179bc67f 4659 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4660 * adjustment filled in by caller) and return result.
179bc67f 4661 */
9e74a6da 4662 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4663}
4664
2c64605b
PNA
4665static inline bool skb_is_redirected(const struct sk_buff *skb)
4666{
4667#ifdef CONFIG_NET_REDIRECT
4668 return skb->redirected;
4669#else
4670 return false;
4671#endif
4672}
4673
4674static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4675{
4676#ifdef CONFIG_NET_REDIRECT
4677 skb->redirected = 1;
4678 skb->from_ingress = from_ingress;
4679 if (skb->from_ingress)
4680 skb->tstamp = 0;
4681#endif
4682}
4683
4684static inline void skb_reset_redirect(struct sk_buff *skb)
4685{
4686#ifdef CONFIG_NET_REDIRECT
4687 skb->redirected = 0;
4688#endif
4689}
4690
fa821170
XL
4691static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4692{
4693 return skb->csum_not_inet;
4694}
4695
6370cc3b
AN
4696static inline void skb_set_kcov_handle(struct sk_buff *skb,
4697 const u64 kcov_handle)
4698{
fa69ee5a
ME
4699#ifdef CONFIG_KCOV
4700 skb->kcov_handle = kcov_handle;
4701#endif
6370cc3b
AN
4702}
4703
4704static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4705{
fa69ee5a
ME
4706#ifdef CONFIG_KCOV
4707 return skb->kcov_handle;
6370cc3b 4708#else
fa69ee5a
ME
4709 return 0;
4710#endif
4711}
6370cc3b 4712
6a5bcd84
IA
4713#ifdef CONFIG_PAGE_POOL
4714static inline void skb_mark_for_recycle(struct sk_buff *skb, struct page *page,
4715 struct page_pool *pp)
4716{
4717 skb->pp_recycle = 1;
4718 page_pool_store_mem_info(page, pp);
4719}
4720#endif
4721
4722static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4723{
4724 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4725 return false;
4726 return page_pool_return_skb_page(virt_to_page(data));
4727}
4728
1da177e4
LT
4729#endif /* __KERNEL__ */
4730#endif /* _LINUX_SKBUFF_H */