]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/linux/skbuff.h
net: preserve skb_end_offset() in skb_unclone_keeptruesize()
[mirror_ubuntu-jammy-kernel.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
8842d285 17#include <linux/bvec.h>
1da177e4 18#include <linux/cache.h>
56b17425 19#include <linux/rbtree.h>
51f3d02b 20#include <linux/socket.h>
c1d1b437 21#include <linux/refcount.h>
1da177e4 22
60063497 23#include <linux/atomic.h>
1da177e4
LT
24#include <asm/types.h>
25#include <linux/spinlock.h>
1da177e4 26#include <linux/net.h>
3fc7e8a6 27#include <linux/textsearch.h>
1da177e4 28#include <net/checksum.h>
a80958f4 29#include <linux/rcupdate.h>
b7aa0bf7 30#include <linux/hrtimer.h>
131ea667 31#include <linux/dma-mapping.h>
c8f44aff 32#include <linux/netdev_features.h>
363ec392 33#include <linux/sched.h>
e6017571 34#include <linux/sched/clock.h>
1bd758eb 35#include <net/flow_dissector.h>
a60e3cc7 36#include <linux/splice.h>
72b31f72 37#include <linux/in6.h>
8b10cab6 38#include <linux/if_packet.h>
f70ea018 39#include <net/flow.h>
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;
6b3d86da 289 __u16 zone;
55132848
PB
290 u8 post_ct:1;
291 u8 post_ct_snat:1;
292 u8 post_ct_dnat:1;
95a7233c
PB
293};
294#endif
295
1da177e4
LT
296struct sk_buff_head {
297 /* These two members must be first. */
298 struct sk_buff *next;
299 struct sk_buff *prev;
300
301 __u32 qlen;
302 spinlock_t lock;
303};
304
305struct sk_buff;
306
9d4dde52
IC
307/* To allow 64K frame to be packed as single skb without frag_list we
308 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
309 * buffers which do not start on a page boundary.
310 *
311 * Since GRO uses frags we allocate at least 16 regardless of page
312 * size.
a715dea3 313 */
9d4dde52 314#if (65536/PAGE_SIZE + 1) < 16
eec00954 315#define MAX_SKB_FRAGS 16UL
a715dea3 316#else
9d4dde52 317#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 318#endif
5f74f82e 319extern int sysctl_max_skb_frags;
1da177e4 320
3953c46c
MRL
321/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
322 * segment using its current segmentation instead.
323 */
324#define GSO_BY_FRAGS 0xFFFF
325
8842d285 326typedef struct bio_vec skb_frag_t;
1da177e4 327
161e6137 328/**
7240b60c 329 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
330 * @frag: skb fragment
331 */
9e903e08
ED
332static inline unsigned int skb_frag_size(const skb_frag_t *frag)
333{
b8b576a1 334 return frag->bv_len;
9e903e08
ED
335}
336
161e6137 337/**
7240b60c 338 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
339 * @frag: skb fragment
340 * @size: size of fragment
341 */
9e903e08
ED
342static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
343{
b8b576a1 344 frag->bv_len = size;
9e903e08
ED
345}
346
161e6137 347/**
7240b60c 348 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
349 * @frag: skb fragment
350 * @delta: value to add
351 */
9e903e08
ED
352static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
353{
b8b576a1 354 frag->bv_len += delta;
9e903e08
ED
355}
356
161e6137 357/**
7240b60c 358 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
359 * @frag: skb fragment
360 * @delta: value to subtract
361 */
9e903e08
ED
362static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
363{
b8b576a1 364 frag->bv_len -= delta;
9e903e08
ED
365}
366
161e6137
PT
367/**
368 * skb_frag_must_loop - Test if %p is a high memory page
369 * @p: fragment's page
370 */
c613c209
WB
371static inline bool skb_frag_must_loop(struct page *p)
372{
373#if defined(CONFIG_HIGHMEM)
29766bcf 374 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
375 return true;
376#endif
377 return false;
378}
379
380/**
381 * skb_frag_foreach_page - loop over pages in a fragment
382 *
383 * @f: skb frag to operate on
1dfa5bd3 384 * @f_off: offset from start of f->bv_page
c613c209
WB
385 * @f_len: length from f_off to loop over
386 * @p: (temp var) current page
387 * @p_off: (temp var) offset from start of current page,
388 * non-zero only on first page.
389 * @p_len: (temp var) length in current page,
390 * < PAGE_SIZE only on first and last page.
391 * @copied: (temp var) length so far, excluding current p_len.
392 *
393 * A fragment can hold a compound page, in which case per-page
394 * operations, notably kmap_atomic, must be called for each
395 * regular page.
396 */
397#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
398 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
399 p_off = (f_off) & (PAGE_SIZE - 1), \
400 p_len = skb_frag_must_loop(p) ? \
401 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
402 copied = 0; \
403 copied < f_len; \
404 copied += p_len, p++, p_off = 0, \
405 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
406
ac45f602
PO
407#define HAVE_HW_TIME_STAMP
408
409/**
d3a21be8 410 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
411 * @hwtstamp: hardware time stamp transformed into duration
412 * since arbitrary point in time
ac45f602
PO
413 *
414 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 415 * skb->tstamp.
ac45f602
PO
416 *
417 * hwtstamps can only be compared against other hwtstamps from
418 * the same device.
419 *
420 * This structure is attached to packets as part of the
421 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
422 */
423struct skb_shared_hwtstamps {
424 ktime_t hwtstamp;
ac45f602
PO
425};
426
2244d07b
OH
427/* Definitions for tx_flags in struct skb_shared_info */
428enum {
429 /* generate hardware time stamp */
430 SKBTX_HW_TSTAMP = 1 << 0,
431
e7fd2885 432 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
433 SKBTX_SW_TSTAMP = 1 << 1,
434
435 /* device driver is going to provide hardware time stamp */
436 SKBTX_IN_PROGRESS = 1 << 2,
437
6e3e939f 438 /* generate wifi status information (where possible) */
62b1a8ab 439 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 440
e7fd2885
WB
441 /* generate software time stamp when entering packet scheduling */
442 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
443};
444
e1c8a607 445#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 446 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
447#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
448
06b4feb3
JL
449/* Definitions for flags in struct skb_shared_info */
450enum {
451 /* use zcopy routines */
452 SKBFL_ZEROCOPY_ENABLE = BIT(0),
453
454 /* This indicates at least one fragment might be overwritten
455 * (as in vmsplice(), sendfile() ...)
456 * If we need to compute a TX checksum, we'll need to copy
457 * all frags to avoid possible bad checksum
458 */
459 SKBFL_SHARED_FRAG = BIT(1),
460};
461
462#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
463
a6686f2f
SM
464/*
465 * The callback notifies userspace to release buffers when skb DMA is done in
466 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
467 * The zerocopy_success argument is true if zero copy transmit occurred,
468 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
469 * The ctx field is used to track device context.
470 * The desc field is used to track userspace buffer index.
a6686f2f
SM
471 */
472struct ubuf_info {
36177832
JL
473 void (*callback)(struct sk_buff *, struct ubuf_info *,
474 bool zerocopy_success);
4ab6c99d
WB
475 union {
476 struct {
477 unsigned long desc;
478 void *ctx;
479 };
480 struct {
481 u32 id;
482 u16 len;
483 u16 zerocopy:1;
484 u32 bytelen;
485 };
486 };
c1d1b437 487 refcount_t refcnt;
04c2d33e 488 u8 flags;
a91dbff5
WB
489
490 struct mmpin {
491 struct user_struct *user;
492 unsigned int num_pg;
493 } mmp;
ac45f602
PO
494};
495
52267790
WB
496#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
497
6f89dbce
SV
498int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
499void mm_unaccount_pinned_pages(struct mmpin *mmp);
500
8c793822
JL
501struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
502struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
503 struct ubuf_info *uarg);
52267790 504
8c793822 505void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790 506
8c793822
JL
507void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
508 bool success);
52267790 509
b5947e5d 510int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
511int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
512 struct msghdr *msg, int len,
513 struct ubuf_info *uarg);
514
1da177e4
LT
515/* This data is invariant across clones and lives at
516 * the end of the header data, ie. at skb->end.
517 */
518struct skb_shared_info {
06b4feb3 519 __u8 flags;
de8f3a83
DB
520 __u8 meta_len;
521 __u8 nr_frags;
9f42f126 522 __u8 tx_flags;
7967168c
HX
523 unsigned short gso_size;
524 /* Warning: this field is not always filled in (UFO)! */
525 unsigned short gso_segs;
1da177e4 526 struct sk_buff *frag_list;
ac45f602 527 struct skb_shared_hwtstamps hwtstamps;
7f564528 528 unsigned int gso_type;
09c2d251 529 u32 tskey;
ec7d2f2c
ED
530
531 /*
532 * Warning : all fields before dataref are cleared in __alloc_skb()
533 */
534 atomic_t dataref;
535
69e3c75f
JB
536 /* Intermediate layers must ensure that destructor_arg
537 * remains valid until skb destructor */
538 void * destructor_arg;
a6686f2f 539
fed66381
ED
540 /* must be last field, see pskb_expand_head() */
541 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
542};
543
544/* We divide dataref into two halves. The higher 16 bits hold references
545 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
546 * the entire skb->data. A clone of a headerless skb holds the length of
547 * the header in skb->hdr_len.
1da177e4
LT
548 *
549 * All users must obey the rule that the skb->data reference count must be
550 * greater than or equal to the payload reference count.
551 *
552 * Holding a reference to the payload part means that the user does not
553 * care about modifications to the header part of skb->data.
554 */
555#define SKB_DATAREF_SHIFT 16
556#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
557
d179cd12
DM
558
559enum {
c8753d55
VS
560 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
561 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
562 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
563};
564
7967168c
HX
565enum {
566 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
567
568 /* This indicates the skb is from an untrusted source. */
d9d30adf 569 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
570
571 /* This indicates the tcp segment has CWR set. */
d9d30adf 572 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 573
d9d30adf 574 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 575
d9d30adf 576 SKB_GSO_TCPV6 = 1 << 4,
68c33163 577
d9d30adf 578 SKB_GSO_FCOE = 1 << 5,
73136267 579
d9d30adf 580 SKB_GSO_GRE = 1 << 6,
0d89d203 581
d9d30adf 582 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 583
d9d30adf 584 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 585
d9d30adf 586 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 587
d9d30adf 588 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 589
d9d30adf 590 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 591
d9d30adf 592 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 593
d9d30adf 594 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 595
d9d30adf 596 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 597
d9d30adf 598 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
599
600 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
601
602 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
603
604 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
605};
606
2e07fa9c
ACM
607#if BITS_PER_LONG > 32
608#define NET_SKBUFF_DATA_USES_OFFSET 1
609#endif
610
611#ifdef NET_SKBUFF_DATA_USES_OFFSET
612typedef unsigned int sk_buff_data_t;
613#else
614typedef unsigned char *sk_buff_data_t;
615#endif
616
161e6137 617/**
1da177e4
LT
618 * struct sk_buff - socket buffer
619 * @next: Next buffer in list
620 * @prev: Previous buffer in list
363ec392 621 * @tstamp: Time we arrived/left
d2f273f0
RD
622 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
623 * for retransmit timer
56b17425 624 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 625 * @list: queue head
d84e0bd7 626 * @sk: Socket we are owned by
d2f273f0
RD
627 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
628 * fragmentation management
1da177e4 629 * @dev: Device we arrived on/are leaving by
d2f273f0 630 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 631 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 632 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 633 * @sp: the security path, used for xfrm
1da177e4
LT
634 * @len: Length of actual data
635 * @data_len: Data length
636 * @mac_len: Length of link layer header
334a8132 637 * @hdr_len: writable header length of cloned skb
663ead3b
HX
638 * @csum: Checksum (must include start/offset pair)
639 * @csum_start: Offset from skb->head where checksumming should start
640 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 641 * @priority: Packet queueing priority
60ff7467 642 * @ignore_df: allow local fragmentation
1da177e4 643 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 644 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
645 * @nohdr: Payload reference only, must not modify header
646 * @pkt_type: Packet class
c83c2486 647 * @fclone: skbuff clone status
c83c2486 648 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
649 * @inner_protocol_type: whether the inner protocol is
650 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
651 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
652 * @offload_fwd_mark: Packet was L2-forwarded in hardware
653 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 654 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 655 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
656 * @redirected: packet was redirected by packet classifier
657 * @from_ingress: packet was redirected from the ingress path
31729363
RD
658 * @peeked: this packet has been seen already, so stats have been
659 * done for it, don't do them again
ba9dda3a 660 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
661 * @protocol: Packet protocol from driver
662 * @destructor: Destruct function
e2080072 663 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
6ed6e1c7 664 * @_sk_redir: socket redirection information for skmsg
a9e419dc 665 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 666 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 667 * @skb_iif: ifindex of device we arrived on
1da177e4 668 * @tc_index: Traffic control index
61b905da 669 * @hash: the packet hash
d84e0bd7 670 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
671 * @head_frag: skb was allocated from page fragments,
672 * not allocated by kmalloc() or vmalloc().
8b700862 673 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
6a5bcd84
IA
674 * @pp_recycle: mark the packet for recycling instead of freeing (implies
675 * page_pool support on driver)
df5042f4 676 * @active_extensions: active extensions (skb_ext_id types)
553a5672 677 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 678 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 679 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 680 * ports.
a3b18ddb 681 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
682 * @wifi_acked_valid: wifi_acked was set
683 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 684 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
685 * @encapsulation: indicates the inner headers in the skbuff are valid
686 * @encap_hdr_csum: software checksum is needed
687 * @csum_valid: checksum is already valid
dba00306 688 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
689 * @csum_complete_sw: checksum was completed by software
690 * @csum_level: indicates the number of consecutive checksums found in
691 * the packet minus one that have been verified as
692 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 693 * @dst_pending_confirm: need to confirm neighbour
a48d189e 694 * @decrypted: Decrypted SKB
5fc88f93 695 * @slow_gro: state present at GRO time, slower prepare step required
161e6137 696 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 697 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 698 * @secmark: security marking
d84e0bd7 699 * @mark: Generic packet mark
d2f273f0
RD
700 * @reserved_tailroom: (aka @mark) number of bytes of free space available
701 * at the tail of an sk_buff
702 * @vlan_present: VLAN tag is present
86a9bad3 703 * @vlan_proto: vlan encapsulation protocol
6aa895b0 704 * @vlan_tci: vlan tag control information
0d89d203 705 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
706 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
707 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
708 * @inner_transport_header: Inner transport layer header (encapsulation)
709 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 710 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
711 * @transport_header: Transport layer header
712 * @network_header: Network layer header
713 * @mac_header: Link layer header
fa69ee5a 714 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
715 * @tail: Tail pointer
716 * @end: End pointer
717 * @head: Head of buffer
718 * @data: Data head pointer
719 * @truesize: Buffer size
720 * @users: User count - see {datagram,tcp}.c
df5042f4 721 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
722 */
723
724struct sk_buff {
363ec392 725 union {
56b17425
ED
726 struct {
727 /* These two members must be first. */
728 struct sk_buff *next;
729 struct sk_buff *prev;
730
731 union {
bffa72cf
ED
732 struct net_device *dev;
733 /* Some protocols might use this space to store information,
734 * while device pointer would be NULL.
735 * UDP receive path is one user.
736 */
737 unsigned long dev_scratch;
56b17425
ED
738 };
739 };
fa0f5273 740 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 741 struct list_head list;
363ec392 742 };
fa0f5273
PO
743
744 union {
745 struct sock *sk;
746 int ip_defrag_offset;
747 };
1da177e4 748
c84d9490 749 union {
bffa72cf 750 ktime_t tstamp;
d3edd06e 751 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 752 };
1da177e4
LT
753 /*
754 * This is the control buffer. It is free to use for every
755 * layer. Please put your private variables there. If you
756 * want to keep them across layers you have to do a skb_clone()
757 * first. This is owned by whoever has the skb queued ATM.
758 */
da3f5cf1 759 char cb[48] __aligned(8);
1da177e4 760
e2080072
ED
761 union {
762 struct {
763 unsigned long _skb_refdst;
764 void (*destructor)(struct sk_buff *skb);
765 };
766 struct list_head tcp_tsorted_anchor;
e3526bb9
CW
767#ifdef CONFIG_NET_SOCK_MSG
768 unsigned long _sk_redir;
769#endif
e2080072
ED
770 };
771
b1937227 772#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 773 unsigned long _nfct;
da3f5cf1 774#endif
1da177e4 775 unsigned int len,
334a8132
PM
776 data_len;
777 __u16 mac_len,
778 hdr_len;
b1937227
ED
779
780 /* Following fields are _not_ copied in __copy_skb_header()
781 * Note that queue_mapping is here mostly to fill a hole.
782 */
b1937227 783 __u16 queue_mapping;
36bbef52
DB
784
785/* if you move cloned around you also must adapt those constants */
786#ifdef __BIG_ENDIAN_BITFIELD
787#define CLONED_MASK (1 << 7)
788#else
789#define CLONED_MASK 1
790#endif
791#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
792
d2f273f0 793 /* private: */
36bbef52 794 __u8 __cloned_offset[0];
d2f273f0 795 /* public: */
b1937227 796 __u8 cloned:1,
6869c4d8 797 nohdr:1,
b84f4cc9 798 fclone:2,
a59322be 799 peeked:1,
b1937227 800 head_frag:1,
6a5bcd84
IA
801 pfmemalloc:1,
802 pp_recycle:1; /* page_pool recycle indicator */
df5042f4
FW
803#ifdef CONFIG_SKB_EXTENSIONS
804 __u8 active_extensions;
805#endif
6a5bcd84 806
b1937227
ED
807 /* fields enclosed in headers_start/headers_end are copied
808 * using a single memcpy() in __copy_skb_header()
809 */
ebcf34f3 810 /* private: */
b1937227 811 __u32 headers_start[0];
ebcf34f3 812 /* public: */
4031ae6e 813
233577a2
HFS
814/* if you move pkt_type around you also must adapt those constants */
815#ifdef __BIG_ENDIAN_BITFIELD
816#define PKT_TYPE_MAX (7 << 5)
817#else
818#define PKT_TYPE_MAX 7
1da177e4 819#endif
233577a2 820#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 821
d2f273f0 822 /* private: */
233577a2 823 __u8 __pkt_type_offset[0];
d2f273f0 824 /* public: */
b1937227 825 __u8 pkt_type:3;
b1937227 826 __u8 ignore_df:1;
b1937227
ED
827 __u8 nf_trace:1;
828 __u8 ip_summed:2;
3853b584 829 __u8 ooo_okay:1;
8b700862 830
61b905da 831 __u8 l4_hash:1;
a3b18ddb 832 __u8 sw_hash:1;
6e3e939f
JB
833 __u8 wifi_acked_valid:1;
834 __u8 wifi_acked:1;
3bdc0eba 835 __u8 no_fcs:1;
77cffe23 836 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 837 __u8 encapsulation:1;
7e2b10c1 838 __u8 encap_hdr_csum:1;
5d0c2b95 839 __u8 csum_valid:1;
8b700862 840
0c4b2d37
MM
841#ifdef __BIG_ENDIAN_BITFIELD
842#define PKT_VLAN_PRESENT_BIT 7
843#else
844#define PKT_VLAN_PRESENT_BIT 0
845#endif
846#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 847 /* private: */
0c4b2d37 848 __u8 __pkt_vlan_present_offset[0];
d2f273f0 849 /* public: */
0c4b2d37 850 __u8 vlan_present:1;
7e3cead5 851 __u8 csum_complete_sw:1;
b1937227 852 __u8 csum_level:2;
dba00306 853 __u8 csum_not_inet:1;
4ff06203 854 __u8 dst_pending_confirm:1;
b1937227
ED
855#ifdef CONFIG_IPV6_NDISC_NODETYPE
856 __u8 ndisc_nodetype:2;
857#endif
8b700862 858
0c4b2d37 859 __u8 ipvs_property:1;
8bce6d7d 860 __u8 inner_protocol_type:1;
e585f236 861 __u8 remcsum_offload:1;
6bc506b4
IS
862#ifdef CONFIG_NET_SWITCHDEV
863 __u8 offload_fwd_mark:1;
875e8939 864 __u8 offload_l3_fwd_mark:1;
6bc506b4 865#endif
e7246e12
WB
866#ifdef CONFIG_NET_CLS_ACT
867 __u8 tc_skip_classify:1;
8dc07fdb 868 __u8 tc_at_ingress:1;
2c64605b 869#endif
2c64605b 870 __u8 redirected:1;
11941f8a 871#ifdef CONFIG_NET_REDIRECT
2c64605b 872 __u8 from_ingress:1;
e7246e12 873#endif
a48d189e
SB
874#ifdef CONFIG_TLS_DEVICE
875 __u8 decrypted:1;
876#endif
5fc88f93 877 __u8 slow_gro:1;
b1937227
ED
878
879#ifdef CONFIG_NET_SCHED
880 __u16 tc_index; /* traffic control index */
b1937227 881#endif
fe55f6d5 882
b1937227
ED
883 union {
884 __wsum csum;
885 struct {
886 __u16 csum_start;
887 __u16 csum_offset;
888 };
889 };
890 __u32 priority;
891 int skb_iif;
892 __u32 hash;
893 __be16 vlan_proto;
894 __u16 vlan_tci;
2bd82484
ED
895#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
896 union {
897 unsigned int napi_id;
898 unsigned int sender_cpu;
899 };
97fc2f08 900#endif
984bc16c 901#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 902 __u32 secmark;
0c4f691f 903#endif
0c4f691f 904
3b885787
NH
905 union {
906 __u32 mark;
16fad69c 907 __u32 reserved_tailroom;
3b885787 908 };
1da177e4 909
8bce6d7d
TH
910 union {
911 __be16 inner_protocol;
912 __u8 inner_ipproto;
913 };
914
1a37e412
SH
915 __u16 inner_transport_header;
916 __u16 inner_network_header;
917 __u16 inner_mac_header;
b1937227
ED
918
919 __be16 protocol;
1a37e412
SH
920 __u16 transport_header;
921 __u16 network_header;
922 __u16 mac_header;
b1937227 923
fa69ee5a
ME
924#ifdef CONFIG_KCOV
925 u64 kcov_handle;
926#endif
927
ebcf34f3 928 /* private: */
b1937227 929 __u32 headers_end[0];
ebcf34f3 930 /* public: */
b1937227 931
1da177e4 932 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 933 sk_buff_data_t tail;
4305b541 934 sk_buff_data_t end;
1da177e4 935 unsigned char *head,
4305b541 936 *data;
27a884dc 937 unsigned int truesize;
63354797 938 refcount_t users;
df5042f4
FW
939
940#ifdef CONFIG_SKB_EXTENSIONS
941 /* only useable after checking ->active_extensions != 0 */
942 struct skb_ext *extensions;
943#endif
1da177e4
LT
944};
945
946#ifdef __KERNEL__
947/*
948 * Handling routines are only of interest to the kernel
949 */
1da177e4 950
c93bdd0e
MG
951#define SKB_ALLOC_FCLONE 0x01
952#define SKB_ALLOC_RX 0x02
fd11a83d 953#define SKB_ALLOC_NAPI 0x04
c93bdd0e 954
161e6137
PT
955/**
956 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
957 * @skb: buffer
958 */
c93bdd0e
MG
959static inline bool skb_pfmemalloc(const struct sk_buff *skb)
960{
961 return unlikely(skb->pfmemalloc);
962}
963
7fee226a
ED
964/*
965 * skb might have a dst pointer attached, refcounted or not.
966 * _skb_refdst low order bit is set if refcount was _not_ taken
967 */
968#define SKB_DST_NOREF 1UL
969#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
970
971/**
972 * skb_dst - returns skb dst_entry
973 * @skb: buffer
974 *
975 * Returns skb dst_entry, regardless of reference taken or not.
976 */
adf30907
ED
977static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
978{
161e6137 979 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
980 * rcu_read_lock section
981 */
982 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
983 !rcu_read_lock_held() &&
984 !rcu_read_lock_bh_held());
985 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
986}
987
7fee226a
ED
988/**
989 * skb_dst_set - sets skb dst
990 * @skb: buffer
991 * @dst: dst entry
992 *
993 * Sets skb dst, assuming a reference was taken on dst and should
994 * be released by skb_dst_drop()
995 */
adf30907
ED
996static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
997{
8a886b14 998 skb->slow_gro |= !!dst;
7fee226a
ED
999 skb->_skb_refdst = (unsigned long)dst;
1000}
1001
932bc4d7
JA
1002/**
1003 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1004 * @skb: buffer
1005 * @dst: dst entry
1006 *
1007 * Sets skb dst, assuming a reference was not taken on dst.
1008 * If dst entry is cached, we do not take reference and dst_release
1009 * will be avoided by refdst_drop. If dst entry is not cached, we take
1010 * reference, so that last dst_release can destroy the dst immediately.
1011 */
1012static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1013{
dbfc4fb7 1014 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
a432934a 1015 skb->slow_gro |= !!dst;
dbfc4fb7 1016 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1017}
7fee226a
ED
1018
1019/**
25985edc 1020 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1021 * @skb: buffer
1022 */
1023static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1024{
1025 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1026}
1027
161e6137
PT
1028/**
1029 * skb_rtable - Returns the skb &rtable
1030 * @skb: buffer
1031 */
511c3f92
ED
1032static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1033{
adf30907 1034 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1035}
1036
8b10cab6
JHS
1037/* For mangling skb->pkt_type from user space side from applications
1038 * such as nft, tc, etc, we only allow a conservative subset of
1039 * possible pkt_types to be set.
1040*/
1041static inline bool skb_pkt_type_ok(u32 ptype)
1042{
1043 return ptype <= PACKET_OTHERHOST;
1044}
1045
161e6137
PT
1046/**
1047 * skb_napi_id - Returns the skb's NAPI id
1048 * @skb: buffer
1049 */
90b602f8
ML
1050static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1051{
1052#ifdef CONFIG_NET_RX_BUSY_POLL
1053 return skb->napi_id;
1054#else
1055 return 0;
1056#endif
1057}
1058
161e6137
PT
1059/**
1060 * skb_unref - decrement the skb's reference count
1061 * @skb: buffer
1062 *
1063 * Returns true if we can free the skb.
1064 */
3889a803
PA
1065static inline bool skb_unref(struct sk_buff *skb)
1066{
1067 if (unlikely(!skb))
1068 return false;
63354797 1069 if (likely(refcount_read(&skb->users) == 1))
3889a803 1070 smp_rmb();
63354797 1071 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1072 return false;
1073
1074 return true;
1075}
1076
0a463c78 1077void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1078void kfree_skb(struct sk_buff *skb);
1079void kfree_skb_list(struct sk_buff *segs);
6413139d 1080void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1081void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1082
1083#ifdef CONFIG_TRACEPOINTS
7965bd4d 1084void consume_skb(struct sk_buff *skb);
be769db2
HX
1085#else
1086static inline void consume_skb(struct sk_buff *skb)
1087{
1088 return kfree_skb(skb);
1089}
1090#endif
1091
ca2c1418 1092void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1093void __kfree_skb(struct sk_buff *skb);
d7e8883c 1094extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1095
7965bd4d
JP
1096void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1097bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1098 bool *fragstolen, int *delta_truesize);
bad43ca8 1099
7965bd4d
JP
1100struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1101 int node);
2ea2f62c 1102struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1103struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1104struct sk_buff *build_skb_around(struct sk_buff *skb,
1105 void *data, unsigned int frag_size);
161e6137 1106
f450d539
AL
1107struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1108
161e6137
PT
1109/**
1110 * alloc_skb - allocate a network buffer
1111 * @size: size to allocate
1112 * @priority: allocation mask
1113 *
1114 * This function is a convenient wrapper around __alloc_skb().
1115 */
d179cd12 1116static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1117 gfp_t priority)
d179cd12 1118{
564824b0 1119 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1120}
1121
2e4e4410
ED
1122struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1123 unsigned long data_len,
1124 int max_page_order,
1125 int *errcode,
1126 gfp_t gfp_mask);
da29e4b4 1127struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1128
d0bf4a9e
ED
1129/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1130struct sk_buff_fclones {
1131 struct sk_buff skb1;
1132
1133 struct sk_buff skb2;
1134
2638595a 1135 refcount_t fclone_ref;
d0bf4a9e
ED
1136};
1137
1138/**
1139 * skb_fclone_busy - check if fclone is busy
293de7de 1140 * @sk: socket
d0bf4a9e
ED
1141 * @skb: buffer
1142 *
bda13fed 1143 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1144 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1145 * so we also check that this didnt happen.
d0bf4a9e 1146 */
39bb5e62
ED
1147static inline bool skb_fclone_busy(const struct sock *sk,
1148 const struct sk_buff *skb)
d0bf4a9e
ED
1149{
1150 const struct sk_buff_fclones *fclones;
1151
1152 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1153
1154 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1155 refcount_read(&fclones->fclone_ref) > 1 &&
f4dae54e 1156 READ_ONCE(fclones->skb2.sk) == sk;
d0bf4a9e
ED
1157}
1158
161e6137
PT
1159/**
1160 * alloc_skb_fclone - allocate a network buffer from fclone cache
1161 * @size: size to allocate
1162 * @priority: allocation mask
1163 *
1164 * This function is a convenient wrapper around __alloc_skb().
1165 */
d179cd12 1166static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1167 gfp_t priority)
d179cd12 1168{
c93bdd0e 1169 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1170}
1171
7965bd4d 1172struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1173void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1174int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1175struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1176void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1177struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1178struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1179 gfp_t gfp_mask, bool fclone);
1180static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1181 gfp_t gfp_mask)
1182{
1183 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1184}
7965bd4d
JP
1185
1186int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1187struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1188 unsigned int headroom);
f1260ff1 1189struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
7965bd4d
JP
1190struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1191 int newtailroom, gfp_t priority);
48a1df65
JD
1192int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1193 int offset, int len);
1194int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1195 int offset, int len);
7965bd4d 1196int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1197int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1198
1199/**
1200 * skb_pad - zero pad the tail of an skb
1201 * @skb: buffer to pad
1202 * @pad: space to pad
1203 *
1204 * Ensure that a buffer is followed by a padding area that is zero
1205 * filled. Used by network drivers which may DMA or transfer data
1206 * beyond the buffer end onto the wire.
1207 *
1208 * May return error in out of memory cases. The skb is freed on error.
1209 */
1210static inline int skb_pad(struct sk_buff *skb, int pad)
1211{
1212 return __skb_pad(skb, pad, true);
1213}
ead2ceb0 1214#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1215
be12a1fe
HFS
1216int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1217 int offset, size_t size);
1218
d94d9fee 1219struct skb_seq_state {
677e90ed
TG
1220 __u32 lower_offset;
1221 __u32 upper_offset;
1222 __u32 frag_idx;
1223 __u32 stepped_offset;
1224 struct sk_buff *root_skb;
1225 struct sk_buff *cur_skb;
1226 __u8 *frag_data;
97550f6f 1227 __u32 frag_off;
677e90ed
TG
1228};
1229
7965bd4d
JP
1230void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1231 unsigned int to, struct skb_seq_state *st);
1232unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1233 struct skb_seq_state *st);
1234void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1235
7965bd4d 1236unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1237 unsigned int to, struct ts_config *config);
3fc7e8a6 1238
09323cc4
TH
1239/*
1240 * Packet hash types specify the type of hash in skb_set_hash.
1241 *
1242 * Hash types refer to the protocol layer addresses which are used to
1243 * construct a packet's hash. The hashes are used to differentiate or identify
1244 * flows of the protocol layer for the hash type. Hash types are either
1245 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1246 *
1247 * Properties of hashes:
1248 *
1249 * 1) Two packets in different flows have different hash values
1250 * 2) Two packets in the same flow should have the same hash value
1251 *
1252 * A hash at a higher layer is considered to be more specific. A driver should
1253 * set the most specific hash possible.
1254 *
1255 * A driver cannot indicate a more specific hash than the layer at which a hash
1256 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1257 *
1258 * A driver may indicate a hash level which is less specific than the
1259 * actual layer the hash was computed on. For instance, a hash computed
1260 * at L4 may be considered an L3 hash. This should only be done if the
1261 * driver can't unambiguously determine that the HW computed the hash at
1262 * the higher layer. Note that the "should" in the second property above
1263 * permits this.
1264 */
1265enum pkt_hash_types {
1266 PKT_HASH_TYPE_NONE, /* Undefined type */
1267 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1268 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1269 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1270};
1271
bcc83839 1272static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1273{
bcc83839 1274 skb->hash = 0;
a3b18ddb 1275 skb->sw_hash = 0;
bcc83839
TH
1276 skb->l4_hash = 0;
1277}
1278
1279static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1280{
1281 if (!skb->l4_hash)
1282 skb_clear_hash(skb);
1283}
1284
1285static inline void
1286__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1287{
1288 skb->l4_hash = is_l4;
1289 skb->sw_hash = is_sw;
61b905da 1290 skb->hash = hash;
09323cc4
TH
1291}
1292
bcc83839
TH
1293static inline void
1294skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1295{
1296 /* Used by drivers to set hash from HW */
1297 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1298}
1299
1300static inline void
1301__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1302{
1303 __skb_set_hash(skb, hash, true, is_l4);
1304}
1305
e5276937 1306void __skb_get_hash(struct sk_buff *skb);
b917783c 1307u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937 1308u32 skb_get_poff(const struct sk_buff *skb);
f96533cd 1309u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
72a338bc 1310 const struct flow_keys_basic *keys, int hlen);
e5276937 1311__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
f96533cd 1312 const void *data, int hlen_proto);
e5276937
TH
1313
1314static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1315 int thoff, u8 ip_proto)
1316{
1317 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1318}
1319
1320void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1321 const struct flow_dissector_key *key,
1322 unsigned int key_count);
1323
089b19a9
SF
1324struct bpf_flow_dissector;
1325bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1326 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1327
3cbf4ffb
SF
1328bool __skb_flow_dissect(const struct net *net,
1329 const struct sk_buff *skb,
e5276937 1330 struct flow_dissector *flow_dissector,
f96533cd
AL
1331 void *target_container, const void *data,
1332 __be16 proto, int nhoff, int hlen, unsigned int flags);
e5276937
TH
1333
1334static inline bool skb_flow_dissect(const struct sk_buff *skb,
1335 struct flow_dissector *flow_dissector,
cd79a238 1336 void *target_container, unsigned int flags)
e5276937 1337{
3cbf4ffb
SF
1338 return __skb_flow_dissect(NULL, skb, flow_dissector,
1339 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1340}
1341
1342static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1343 struct flow_keys *flow,
1344 unsigned int flags)
e5276937
TH
1345{
1346 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1347 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1348 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1349}
1350
72a338bc 1351static inline bool
3cbf4ffb
SF
1352skb_flow_dissect_flow_keys_basic(const struct net *net,
1353 const struct sk_buff *skb,
f96533cd
AL
1354 struct flow_keys_basic *flow,
1355 const void *data, __be16 proto,
1356 int nhoff, int hlen, unsigned int flags)
e5276937
TH
1357{
1358 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1359 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1360 data, proto, nhoff, hlen, flags);
e5276937
TH
1361}
1362
82828b88
JP
1363void skb_flow_dissect_meta(const struct sk_buff *skb,
1364 struct flow_dissector *flow_dissector,
1365 void *target_container);
1366
75a56758 1367/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1368 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1369 * to user states.
1370 */
1371void
1372skb_flow_dissect_ct(const struct sk_buff *skb,
1373 struct flow_dissector *flow_dissector,
1374 void *target_container,
7baf2429 1375 u16 *ctinfo_map, size_t mapsize,
9dd8e692 1376 bool post_ct, u16 zone);
62b32379
SH
1377void
1378skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1379 struct flow_dissector *flow_dissector,
1380 void *target_container);
1381
0cb09aff
AL
1382void skb_flow_dissect_hash(const struct sk_buff *skb,
1383 struct flow_dissector *flow_dissector,
1384 void *target_container);
1385
3958afa1 1386static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1387{
a3b18ddb 1388 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1389 __skb_get_hash(skb);
bfb564e7 1390
61b905da 1391 return skb->hash;
bfb564e7
KK
1392}
1393
20a17bf6 1394static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1395{
c6cc1ca7
TH
1396 if (!skb->l4_hash && !skb->sw_hash) {
1397 struct flow_keys keys;
de4c1f8b 1398 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1399
de4c1f8b 1400 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1401 }
f70ea018
TH
1402
1403 return skb->hash;
1404}
1405
55667441
ED
1406__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1407 const siphash_key_t *perturb);
50fb7992 1408
57bdf7f4
TH
1409static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1410{
61b905da 1411 return skb->hash;
57bdf7f4
TH
1412}
1413
3df7a74e
TH
1414static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1415{
61b905da 1416 to->hash = from->hash;
a3b18ddb 1417 to->sw_hash = from->sw_hash;
61b905da 1418 to->l4_hash = from->l4_hash;
3df7a74e
TH
1419};
1420
41477662
JK
1421static inline void skb_copy_decrypted(struct sk_buff *to,
1422 const struct sk_buff *from)
1423{
1424#ifdef CONFIG_TLS_DEVICE
1425 to->decrypted = from->decrypted;
1426#endif
1427}
1428
4305b541
ACM
1429#ifdef NET_SKBUFF_DATA_USES_OFFSET
1430static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1431{
1432 return skb->head + skb->end;
1433}
ec47ea82
AD
1434
1435static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1436{
1437 return skb->end;
1438}
704d34ae
ED
1439
1440static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1441{
1442 skb->end = offset;
1443}
4305b541
ACM
1444#else
1445static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1446{
1447 return skb->end;
1448}
ec47ea82
AD
1449
1450static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1451{
1452 return skb->end - skb->head;
1453}
704d34ae
ED
1454
1455static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1456{
1457 skb->end = skb->head + offset;
1458}
4305b541
ACM
1459#endif
1460
1da177e4 1461/* Internal */
4305b541 1462#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1463
ac45f602
PO
1464static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1465{
1466 return &skb_shinfo(skb)->hwtstamps;
1467}
1468
52267790
WB
1469static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1470{
06b4feb3 1471 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1472
1473 return is_zcopy ? skb_uarg(skb) : NULL;
1474}
1475
8e044917 1476static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1477{
1478 refcount_inc(&uarg->refcnt);
1479}
1480
9ee5e5ad
JL
1481static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1482{
1483 skb_shinfo(skb)->destructor_arg = uarg;
1484 skb_shinfo(skb)->flags |= uarg->flags;
1485}
1486
52900d22
WB
1487static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1488 bool *have_ref)
52267790
WB
1489{
1490 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1491 if (unlikely(have_ref && *have_ref))
1492 *have_ref = false;
1493 else
8e044917 1494 net_zcopy_get(uarg);
9ee5e5ad 1495 skb_zcopy_init(skb, uarg);
52267790
WB
1496 }
1497}
1498
5cd8d46e
WB
1499static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1500{
1501 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1502 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1503}
1504
1505static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1506{
1507 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1508}
1509
1510static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1511{
1512 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1513}
1514
8e044917 1515static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1516{
1517 if (uarg)
36177832 1518 uarg->callback(NULL, uarg, true);
59776362
JL
1519}
1520
8e044917 1521static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1522{
1523 if (uarg) {
8c793822
JL
1524 if (uarg->callback == msg_zerocopy_callback)
1525 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1526 else if (have_uref)
8e044917 1527 net_zcopy_put(uarg);
236a6b1c
JL
1528 }
1529}
1530
52267790 1531/* Release a reference on a zerocopy structure */
36177832 1532static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1533{
1534 struct ubuf_info *uarg = skb_zcopy(skb);
1535
1536 if (uarg) {
36177832
JL
1537 if (!skb_zcopy_is_nouarg(skb))
1538 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1539
06b4feb3 1540 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
52267790
WB
1541 }
1542}
1543
a8305bff
DM
1544static inline void skb_mark_not_on_list(struct sk_buff *skb)
1545{
1546 skb->next = NULL;
1547}
1548
dcfea72e 1549/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1550#define skb_list_walk_safe(first, skb, next_skb) \
1551 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1552 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1553
992cba7e
DM
1554static inline void skb_list_del_init(struct sk_buff *skb)
1555{
1556 __list_del_entry(&skb->list);
1557 skb_mark_not_on_list(skb);
1558}
1559
1da177e4
LT
1560/**
1561 * skb_queue_empty - check if a queue is empty
1562 * @list: queue head
1563 *
1564 * Returns true if the queue is empty, false otherwise.
1565 */
1566static inline int skb_queue_empty(const struct sk_buff_head *list)
1567{
fd44b93c 1568 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1569}
1570
d7d16a89
ED
1571/**
1572 * skb_queue_empty_lockless - check if a queue is empty
1573 * @list: queue head
1574 *
1575 * Returns true if the queue is empty, false otherwise.
1576 * This variant can be used in lockless contexts.
1577 */
1578static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1579{
1580 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1581}
1582
1583
fc7ebb21
DM
1584/**
1585 * skb_queue_is_last - check if skb is the last entry in the queue
1586 * @list: queue head
1587 * @skb: buffer
1588 *
1589 * Returns true if @skb is the last buffer on the list.
1590 */
1591static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1592 const struct sk_buff *skb)
1593{
fd44b93c 1594 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1595}
1596
832d11c5
IJ
1597/**
1598 * skb_queue_is_first - check if skb is the first entry in the queue
1599 * @list: queue head
1600 * @skb: buffer
1601 *
1602 * Returns true if @skb is the first buffer on the list.
1603 */
1604static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1605 const struct sk_buff *skb)
1606{
fd44b93c 1607 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1608}
1609
249c8b42
DM
1610/**
1611 * skb_queue_next - return the next packet in the queue
1612 * @list: queue head
1613 * @skb: current buffer
1614 *
1615 * Return the next packet in @list after @skb. It is only valid to
1616 * call this if skb_queue_is_last() evaluates to false.
1617 */
1618static inline struct sk_buff *skb_queue_next(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_last(list, skb));
1625 return skb->next;
1626}
1627
832d11c5
IJ
1628/**
1629 * skb_queue_prev - return the prev packet in the queue
1630 * @list: queue head
1631 * @skb: current buffer
1632 *
1633 * Return the prev packet in @list before @skb. It is only valid to
1634 * call this if skb_queue_is_first() evaluates to false.
1635 */
1636static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1637 const struct sk_buff *skb)
1638{
1639 /* This BUG_ON may seem severe, but if we just return then we
1640 * are going to dereference garbage.
1641 */
1642 BUG_ON(skb_queue_is_first(list, skb));
1643 return skb->prev;
1644}
1645
1da177e4
LT
1646/**
1647 * skb_get - reference buffer
1648 * @skb: buffer to reference
1649 *
1650 * Makes another reference to a socket buffer and returns a pointer
1651 * to the buffer.
1652 */
1653static inline struct sk_buff *skb_get(struct sk_buff *skb)
1654{
63354797 1655 refcount_inc(&skb->users);
1da177e4
LT
1656 return skb;
1657}
1658
1659/*
f8821f96 1660 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1661 */
1662
1da177e4
LT
1663/**
1664 * skb_cloned - is the buffer a clone
1665 * @skb: buffer to check
1666 *
1667 * Returns true if the buffer was generated with skb_clone() and is
1668 * one of multiple shared copies of the buffer. Cloned buffers are
1669 * shared data so must not be written to under normal circumstances.
1670 */
1671static inline int skb_cloned(const struct sk_buff *skb)
1672{
1673 return skb->cloned &&
1674 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1675}
1676
14bbd6a5
PS
1677static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1678{
d0164adc 1679 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1680
1681 if (skb_cloned(skb))
1682 return pskb_expand_head(skb, 0, 0, pri);
1683
1684 return 0;
1685}
1686
04900f03
ED
1687/* This variant of skb_unclone() makes sure skb->truesize
1688 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1689 *
1690 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1691 * when various debugging features are in place.
1692 */
1693int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
08ff09fd
ED
1694static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1695{
1696 might_sleep_if(gfpflags_allow_blocking(pri));
1697
04900f03
ED
1698 if (skb_cloned(skb))
1699 return __skb_unclone_keeptruesize(skb, pri);
08ff09fd
ED
1700 return 0;
1701}
1702
1da177e4
LT
1703/**
1704 * skb_header_cloned - is the header a clone
1705 * @skb: buffer to check
1706 *
1707 * Returns true if modifying the header part of the buffer requires
1708 * the data to be copied.
1709 */
1710static inline int skb_header_cloned(const struct sk_buff *skb)
1711{
1712 int dataref;
1713
1714 if (!skb->cloned)
1715 return 0;
1716
1717 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1718 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1719 return dataref != 1;
1720}
1721
9580bf2e
ED
1722static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1723{
1724 might_sleep_if(gfpflags_allow_blocking(pri));
1725
1726 if (skb_header_cloned(skb))
1727 return pskb_expand_head(skb, 0, 0, pri);
1728
1729 return 0;
1730}
1731
f4a775d1
ED
1732/**
1733 * __skb_header_release - release reference to header
1734 * @skb: buffer to operate on
f4a775d1
ED
1735 */
1736static inline void __skb_header_release(struct sk_buff *skb)
1737{
1738 skb->nohdr = 1;
1739 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1740}
1741
1742
1da177e4
LT
1743/**
1744 * skb_shared - is the buffer shared
1745 * @skb: buffer to check
1746 *
1747 * Returns true if more than one person has a reference to this
1748 * buffer.
1749 */
1750static inline int skb_shared(const struct sk_buff *skb)
1751{
63354797 1752 return refcount_read(&skb->users) != 1;
1da177e4
LT
1753}
1754
1755/**
1756 * skb_share_check - check if buffer is shared and if so clone it
1757 * @skb: buffer to check
1758 * @pri: priority for memory allocation
1759 *
1760 * If the buffer is shared the buffer is cloned and the old copy
1761 * drops a reference. A new clone with a single reference is returned.
1762 * If the buffer is not shared the original buffer is returned. When
1763 * being called from interrupt status or with spinlocks held pri must
1764 * be GFP_ATOMIC.
1765 *
1766 * NULL is returned on a memory allocation failure.
1767 */
47061bc4 1768static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1769{
d0164adc 1770 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1771 if (skb_shared(skb)) {
1772 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1773
1774 if (likely(nskb))
1775 consume_skb(skb);
1776 else
1777 kfree_skb(skb);
1da177e4
LT
1778 skb = nskb;
1779 }
1780 return skb;
1781}
1782
1783/*
1784 * Copy shared buffers into a new sk_buff. We effectively do COW on
1785 * packets to handle cases where we have a local reader and forward
1786 * and a couple of other messy ones. The normal one is tcpdumping
1787 * a packet thats being forwarded.
1788 */
1789
1790/**
1791 * skb_unshare - make a copy of a shared buffer
1792 * @skb: buffer to check
1793 * @pri: priority for memory allocation
1794 *
1795 * If the socket buffer is a clone then this function creates a new
1796 * copy of the data, drops a reference count on the old copy and returns
1797 * the new copy with the reference count at 1. If the buffer is not a clone
1798 * the original buffer is returned. When called with a spinlock held or
1799 * from interrupt state @pri must be %GFP_ATOMIC
1800 *
1801 * %NULL is returned on a memory allocation failure.
1802 */
e2bf521d 1803static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1804 gfp_t pri)
1da177e4 1805{
d0164adc 1806 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1807 if (skb_cloned(skb)) {
1808 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1809
1810 /* Free our shared copy */
1811 if (likely(nskb))
1812 consume_skb(skb);
1813 else
1814 kfree_skb(skb);
1da177e4
LT
1815 skb = nskb;
1816 }
1817 return skb;
1818}
1819
1820/**
1a5778aa 1821 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1822 * @list_: list to peek at
1823 *
1824 * Peek an &sk_buff. Unlike most other operations you _MUST_
1825 * be careful with this one. A peek leaves the buffer on the
1826 * list and someone else may run off with it. You must hold
1827 * the appropriate locks or have a private queue to do this.
1828 *
1829 * Returns %NULL for an empty list or a pointer to the head element.
1830 * The reference count is not incremented and the reference is therefore
1831 * volatile. Use with caution.
1832 */
05bdd2f1 1833static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1834{
18d07000
ED
1835 struct sk_buff *skb = list_->next;
1836
1837 if (skb == (struct sk_buff *)list_)
1838 skb = NULL;
1839 return skb;
1da177e4
LT
1840}
1841
8b69bd7d
DM
1842/**
1843 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1844 * @list_: list to peek at
1845 *
1846 * Like skb_peek(), but the caller knows that the list is not empty.
1847 */
1848static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1849{
1850 return list_->next;
1851}
1852
da5ef6e5
PE
1853/**
1854 * skb_peek_next - peek skb following the given one from a queue
1855 * @skb: skb to start from
1856 * @list_: list to peek at
1857 *
1858 * Returns %NULL when the end of the list is met or a pointer to the
1859 * next element. The reference count is not incremented and the
1860 * reference is therefore volatile. Use with caution.
1861 */
1862static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1863 const struct sk_buff_head *list_)
1864{
1865 struct sk_buff *next = skb->next;
18d07000 1866
da5ef6e5
PE
1867 if (next == (struct sk_buff *)list_)
1868 next = NULL;
1869 return next;
1870}
1871
1da177e4 1872/**
1a5778aa 1873 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1874 * @list_: list to peek at
1875 *
1876 * Peek an &sk_buff. Unlike most other operations you _MUST_
1877 * be careful with this one. A peek leaves the buffer on the
1878 * list and someone else may run off with it. You must hold
1879 * the appropriate locks or have a private queue to do this.
1880 *
1881 * Returns %NULL for an empty list or a pointer to the tail element.
1882 * The reference count is not incremented and the reference is therefore
1883 * volatile. Use with caution.
1884 */
05bdd2f1 1885static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1886{
f8cc62ca 1887 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1888
1889 if (skb == (struct sk_buff *)list_)
1890 skb = NULL;
1891 return skb;
1892
1da177e4
LT
1893}
1894
1895/**
1896 * skb_queue_len - get queue length
1897 * @list_: list to measure
1898 *
1899 * Return the length of an &sk_buff queue.
1900 */
1901static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1902{
1903 return list_->qlen;
1904}
1905
86b18aaa
QC
1906/**
1907 * skb_queue_len_lockless - get queue length
1908 * @list_: list to measure
1909 *
1910 * Return the length of an &sk_buff queue.
1911 * This variant can be used in lockless contexts.
1912 */
1913static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1914{
1915 return READ_ONCE(list_->qlen);
1916}
1917
67fed459
DM
1918/**
1919 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1920 * @list: queue to initialize
1921 *
1922 * This initializes only the list and queue length aspects of
1923 * an sk_buff_head object. This allows to initialize the list
1924 * aspects of an sk_buff_head without reinitializing things like
1925 * the spinlock. It can also be used for on-stack sk_buff_head
1926 * objects where the spinlock is known to not be used.
1927 */
1928static inline void __skb_queue_head_init(struct sk_buff_head *list)
1929{
1930 list->prev = list->next = (struct sk_buff *)list;
1931 list->qlen = 0;
1932}
1933
76f10ad0
AV
1934/*
1935 * This function creates a split out lock class for each invocation;
1936 * this is needed for now since a whole lot of users of the skb-queue
1937 * infrastructure in drivers have different locking usage (in hardirq)
1938 * than the networking core (in softirq only). In the long run either the
1939 * network layer or drivers should need annotation to consolidate the
1940 * main types of usage into 3 classes.
1941 */
1da177e4
LT
1942static inline void skb_queue_head_init(struct sk_buff_head *list)
1943{
1944 spin_lock_init(&list->lock);
67fed459 1945 __skb_queue_head_init(list);
1da177e4
LT
1946}
1947
c2ecba71
PE
1948static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1949 struct lock_class_key *class)
1950{
1951 skb_queue_head_init(list);
1952 lockdep_set_class(&list->lock, class);
1953}
1954
1da177e4 1955/*
bf299275 1956 * Insert an sk_buff on a list.
1da177e4
LT
1957 *
1958 * The "__skb_xxxx()" functions are the non-atomic ones that
1959 * can only be called with interrupts disabled.
1960 */
bf299275
GR
1961static inline void __skb_insert(struct sk_buff *newsk,
1962 struct sk_buff *prev, struct sk_buff *next,
1963 struct sk_buff_head *list)
1964{
f8cc62ca
ED
1965 /* See skb_queue_empty_lockless() and skb_peek_tail()
1966 * for the opposite READ_ONCE()
1967 */
d7d16a89
ED
1968 WRITE_ONCE(newsk->next, next);
1969 WRITE_ONCE(newsk->prev, prev);
1970 WRITE_ONCE(next->prev, newsk);
1971 WRITE_ONCE(prev->next, newsk);
04f08eb4 1972 WRITE_ONCE(list->qlen, list->qlen + 1);
bf299275 1973}
1da177e4 1974
67fed459
DM
1975static inline void __skb_queue_splice(const struct sk_buff_head *list,
1976 struct sk_buff *prev,
1977 struct sk_buff *next)
1978{
1979 struct sk_buff *first = list->next;
1980 struct sk_buff *last = list->prev;
1981
d7d16a89
ED
1982 WRITE_ONCE(first->prev, prev);
1983 WRITE_ONCE(prev->next, first);
67fed459 1984
d7d16a89
ED
1985 WRITE_ONCE(last->next, next);
1986 WRITE_ONCE(next->prev, last);
67fed459
DM
1987}
1988
1989/**
1990 * skb_queue_splice - join two skb lists, this is designed for stacks
1991 * @list: the new list to add
1992 * @head: the place to add it in the first list
1993 */
1994static inline void skb_queue_splice(const struct sk_buff_head *list,
1995 struct sk_buff_head *head)
1996{
1997 if (!skb_queue_empty(list)) {
1998 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1999 head->qlen += list->qlen;
67fed459
DM
2000 }
2001}
2002
2003/**
d9619496 2004 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2005 * @list: the new list to add
2006 * @head: the place to add it in the first list
2007 *
2008 * The list at @list is reinitialised
2009 */
2010static inline void skb_queue_splice_init(struct sk_buff_head *list,
2011 struct sk_buff_head *head)
2012{
2013 if (!skb_queue_empty(list)) {
2014 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 2015 head->qlen += list->qlen;
67fed459
DM
2016 __skb_queue_head_init(list);
2017 }
2018}
2019
2020/**
2021 * skb_queue_splice_tail - join two skb lists, each list being a queue
2022 * @list: the new list to add
2023 * @head: the place to add it in the first list
2024 */
2025static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2026 struct sk_buff_head *head)
2027{
2028 if (!skb_queue_empty(list)) {
2029 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2030 head->qlen += list->qlen;
67fed459
DM
2031 }
2032}
2033
2034/**
d9619496 2035 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2036 * @list: the new list to add
2037 * @head: the place to add it in the first list
2038 *
2039 * Each of the lists is a queue.
2040 * The list at @list is reinitialised
2041 */
2042static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2043 struct sk_buff_head *head)
2044{
2045 if (!skb_queue_empty(list)) {
2046 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2047 head->qlen += list->qlen;
67fed459
DM
2048 __skb_queue_head_init(list);
2049 }
2050}
2051
1da177e4 2052/**
300ce174 2053 * __skb_queue_after - queue a buffer at the list head
1da177e4 2054 * @list: list to use
300ce174 2055 * @prev: place after this buffer
1da177e4
LT
2056 * @newsk: buffer to queue
2057 *
300ce174 2058 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2059 * and you must therefore hold required locks before calling it.
2060 *
2061 * A buffer cannot be placed on two lists at the same time.
2062 */
300ce174
SH
2063static inline void __skb_queue_after(struct sk_buff_head *list,
2064 struct sk_buff *prev,
2065 struct sk_buff *newsk)
1da177e4 2066{
bf299275 2067 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
2068}
2069
7965bd4d
JP
2070void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2071 struct sk_buff_head *list);
7de6c033 2072
f5572855
GR
2073static inline void __skb_queue_before(struct sk_buff_head *list,
2074 struct sk_buff *next,
2075 struct sk_buff *newsk)
2076{
2077 __skb_insert(newsk, next->prev, next, list);
2078}
2079
300ce174
SH
2080/**
2081 * __skb_queue_head - queue a buffer at the list head
2082 * @list: list to use
2083 * @newsk: buffer to queue
2084 *
2085 * Queue a buffer at the start of a list. This function takes no locks
2086 * and you must therefore hold required locks before calling it.
2087 *
2088 * A buffer cannot be placed on two lists at the same time.
2089 */
300ce174
SH
2090static inline void __skb_queue_head(struct sk_buff_head *list,
2091 struct sk_buff *newsk)
2092{
2093 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2094}
4ea7b0cf 2095void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2096
1da177e4
LT
2097/**
2098 * __skb_queue_tail - queue a buffer at the list tail
2099 * @list: list to use
2100 * @newsk: buffer to queue
2101 *
2102 * Queue a buffer at the end of a list. This function takes no locks
2103 * and you must therefore hold required locks before calling it.
2104 *
2105 * A buffer cannot be placed on two lists at the same time.
2106 */
1da177e4
LT
2107static inline void __skb_queue_tail(struct sk_buff_head *list,
2108 struct sk_buff *newsk)
2109{
f5572855 2110 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2111}
4ea7b0cf 2112void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2113
1da177e4
LT
2114/*
2115 * remove sk_buff from list. _Must_ be called atomically, and with
2116 * the list known..
2117 */
7965bd4d 2118void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2119static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2120{
2121 struct sk_buff *next, *prev;
2122
86b18aaa 2123 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2124 next = skb->next;
2125 prev = skb->prev;
2126 skb->next = skb->prev = NULL;
d7d16a89
ED
2127 WRITE_ONCE(next->prev, prev);
2128 WRITE_ONCE(prev->next, next);
1da177e4
LT
2129}
2130
f525c06d
GR
2131/**
2132 * __skb_dequeue - remove from the head of the queue
2133 * @list: list to dequeue from
2134 *
2135 * Remove the head of the list. This function does not take any locks
2136 * so must be used with appropriate locks held only. The head item is
2137 * returned or %NULL if the list is empty.
2138 */
f525c06d
GR
2139static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2140{
2141 struct sk_buff *skb = skb_peek(list);
2142 if (skb)
2143 __skb_unlink(skb, list);
2144 return skb;
2145}
4ea7b0cf 2146struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2147
2148/**
2149 * __skb_dequeue_tail - remove from the tail of the queue
2150 * @list: list to dequeue from
2151 *
2152 * Remove the tail of the list. This function does not take any locks
2153 * so must be used with appropriate locks held only. The tail item is
2154 * returned or %NULL if the list is empty.
2155 */
1da177e4
LT
2156static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2157{
2158 struct sk_buff *skb = skb_peek_tail(list);
2159 if (skb)
2160 __skb_unlink(skb, list);
2161 return skb;
2162}
4ea7b0cf 2163struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2164
2165
bdcc0924 2166static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2167{
2168 return skb->data_len;
2169}
2170
2171static inline unsigned int skb_headlen(const struct sk_buff *skb)
2172{
2173 return skb->len - skb->data_len;
2174}
2175
3ece7826 2176static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2177{
c72d8cda 2178 unsigned int i, len = 0;
1da177e4 2179
c72d8cda 2180 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2181 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2182 return len;
2183}
2184
2185static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2186{
2187 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2188}
2189
131ea667
IC
2190/**
2191 * __skb_fill_page_desc - initialise a paged fragment in an skb
2192 * @skb: buffer containing fragment to be initialised
2193 * @i: paged fragment index to initialise
2194 * @page: the page to use for this fragment
2195 * @off: the offset to the data with @page
2196 * @size: the length of the data
2197 *
2198 * Initialises the @i'th fragment of @skb to point to &size bytes at
2199 * offset @off within @page.
2200 *
2201 * Does not take any additional reference on the fragment.
2202 */
2203static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2204 struct page *page, int off, int size)
1da177e4
LT
2205{
2206 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2207
c48a11c7 2208 /*
2f064f34
MH
2209 * Propagate page pfmemalloc to the skb if we can. The problem is
2210 * that not all callers have unique ownership of the page but rely
2211 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2212 */
1dfa5bd3 2213 frag->bv_page = page;
65c84f14 2214 frag->bv_offset = off;
9e903e08 2215 skb_frag_size_set(frag, size);
cca7af38
PE
2216
2217 page = compound_head(page);
2f064f34 2218 if (page_is_pfmemalloc(page))
cca7af38 2219 skb->pfmemalloc = true;
131ea667
IC
2220}
2221
2222/**
2223 * skb_fill_page_desc - initialise a paged fragment in an skb
2224 * @skb: buffer containing fragment to be initialised
2225 * @i: paged fragment index to initialise
2226 * @page: the page to use for this fragment
2227 * @off: the offset to the data with @page
2228 * @size: the length of the data
2229 *
2230 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2231 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2232 * addition updates @skb such that @i is the last fragment.
2233 *
2234 * Does not take any additional reference on the fragment.
2235 */
2236static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2237 struct page *page, int off, int size)
2238{
2239 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2240 skb_shinfo(skb)->nr_frags = i + 1;
2241}
2242
7965bd4d
JP
2243void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2244 int size, unsigned int truesize);
654bed16 2245
f8e617e1
JW
2246void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2247 unsigned int truesize);
2248
1da177e4
LT
2249#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2250
27a884dc
ACM
2251#ifdef NET_SKBUFF_DATA_USES_OFFSET
2252static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2253{
2254 return skb->head + skb->tail;
2255}
2256
2257static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2258{
2259 skb->tail = skb->data - skb->head;
2260}
2261
2262static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2263{
2264 skb_reset_tail_pointer(skb);
2265 skb->tail += offset;
2266}
7cc46190 2267
27a884dc
ACM
2268#else /* NET_SKBUFF_DATA_USES_OFFSET */
2269static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2270{
2271 return skb->tail;
2272}
2273
2274static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2275{
2276 skb->tail = skb->data;
2277}
2278
2279static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2280{
2281 skb->tail = skb->data + offset;
2282}
4305b541 2283
27a884dc
ACM
2284#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2285
1da177e4
LT
2286/*
2287 * Add data to an sk_buff
2288 */
4df864c1
JB
2289void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2290void *skb_put(struct sk_buff *skb, unsigned int len);
2291static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2292{
4df864c1 2293 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2294 SKB_LINEAR_ASSERT(skb);
2295 skb->tail += len;
2296 skb->len += len;
2297 return tmp;
2298}
2299
de77b966 2300static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2301{
2302 void *tmp = __skb_put(skb, len);
2303
2304 memset(tmp, 0, len);
2305 return tmp;
2306}
2307
2308static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2309 unsigned int len)
2310{
2311 void *tmp = __skb_put(skb, len);
2312
2313 memcpy(tmp, data, len);
2314 return tmp;
2315}
2316
2317static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2318{
2319 *(u8 *)__skb_put(skb, 1) = val;
2320}
2321
83ad357d 2322static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2323{
83ad357d 2324 void *tmp = skb_put(skb, len);
e45a79da
JB
2325
2326 memset(tmp, 0, len);
2327
2328 return tmp;
2329}
2330
59ae1d12
JB
2331static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2332 unsigned int len)
2333{
2334 void *tmp = skb_put(skb, len);
2335
2336 memcpy(tmp, data, len);
2337
2338 return tmp;
2339}
2340
634fef61
JB
2341static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2342{
2343 *(u8 *)skb_put(skb, 1) = val;
2344}
2345
d58ff351
JB
2346void *skb_push(struct sk_buff *skb, unsigned int len);
2347static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2348{
2349 skb->data -= len;
2350 skb->len += len;
2351 return skb->data;
2352}
2353
af72868b
JB
2354void *skb_pull(struct sk_buff *skb, unsigned int len);
2355static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2356{
2357 skb->len -= len;
2358 BUG_ON(skb->len < skb->data_len);
2359 return skb->data += len;
2360}
2361
af72868b 2362static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2363{
2364 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2365}
2366
af72868b 2367void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2368
af72868b 2369static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2370{
2371 if (len > skb_headlen(skb) &&
987c402a 2372 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2373 return NULL;
2374 skb->len -= len;
2375 return skb->data += len;
2376}
2377
af72868b 2378static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2379{
2380 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2381}
2382
b9df4fd7 2383static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2384{
2385 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2386 return true;
1da177e4 2387 if (unlikely(len > skb->len))
b9df4fd7 2388 return false;
987c402a 2389 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2390}
2391
c8c8b127
ED
2392void skb_condense(struct sk_buff *skb);
2393
1da177e4
LT
2394/**
2395 * skb_headroom - bytes at buffer head
2396 * @skb: buffer to check
2397 *
2398 * Return the number of bytes of free space at the head of an &sk_buff.
2399 */
c2636b4d 2400static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2401{
2402 return skb->data - skb->head;
2403}
2404
2405/**
2406 * skb_tailroom - bytes at buffer end
2407 * @skb: buffer to check
2408 *
2409 * Return the number of bytes of free space at the tail of an sk_buff
2410 */
2411static inline int skb_tailroom(const struct sk_buff *skb)
2412{
4305b541 2413 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2414}
2415
a21d4572
ED
2416/**
2417 * skb_availroom - bytes at buffer end
2418 * @skb: buffer to check
2419 *
2420 * Return the number of bytes of free space at the tail of an sk_buff
2421 * allocated by sk_stream_alloc()
2422 */
2423static inline int skb_availroom(const struct sk_buff *skb)
2424{
16fad69c
ED
2425 if (skb_is_nonlinear(skb))
2426 return 0;
2427
2428 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2429}
2430
1da177e4
LT
2431/**
2432 * skb_reserve - adjust headroom
2433 * @skb: buffer to alter
2434 * @len: bytes to move
2435 *
2436 * Increase the headroom of an empty &sk_buff by reducing the tail
2437 * room. This is only allowed for an empty buffer.
2438 */
8243126c 2439static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2440{
2441 skb->data += len;
2442 skb->tail += len;
2443}
2444
1837b2e2
BP
2445/**
2446 * skb_tailroom_reserve - adjust reserved_tailroom
2447 * @skb: buffer to alter
2448 * @mtu: maximum amount of headlen permitted
2449 * @needed_tailroom: minimum amount of reserved_tailroom
2450 *
2451 * Set reserved_tailroom so that headlen can be as large as possible but
2452 * not larger than mtu and tailroom cannot be smaller than
2453 * needed_tailroom.
2454 * The required headroom should already have been reserved before using
2455 * this function.
2456 */
2457static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2458 unsigned int needed_tailroom)
2459{
2460 SKB_LINEAR_ASSERT(skb);
2461 if (mtu < skb_tailroom(skb) - needed_tailroom)
2462 /* use at most mtu */
2463 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2464 else
2465 /* use up to all available space */
2466 skb->reserved_tailroom = needed_tailroom;
2467}
2468
8bce6d7d
TH
2469#define ENCAP_TYPE_ETHER 0
2470#define ENCAP_TYPE_IPPROTO 1
2471
2472static inline void skb_set_inner_protocol(struct sk_buff *skb,
2473 __be16 protocol)
2474{
2475 skb->inner_protocol = protocol;
2476 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2477}
2478
2479static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2480 __u8 ipproto)
2481{
2482 skb->inner_ipproto = ipproto;
2483 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2484}
2485
6a674e9c
JG
2486static inline void skb_reset_inner_headers(struct sk_buff *skb)
2487{
aefbd2b3 2488 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2489 skb->inner_network_header = skb->network_header;
2490 skb->inner_transport_header = skb->transport_header;
2491}
2492
0b5c9db1
JP
2493static inline void skb_reset_mac_len(struct sk_buff *skb)
2494{
2495 skb->mac_len = skb->network_header - skb->mac_header;
2496}
2497
6a674e9c
JG
2498static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2499 *skb)
2500{
2501 return skb->head + skb->inner_transport_header;
2502}
2503
55dc5a9f
TH
2504static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2505{
2506 return skb_inner_transport_header(skb) - skb->data;
2507}
2508
6a674e9c
JG
2509static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2510{
2511 skb->inner_transport_header = skb->data - skb->head;
2512}
2513
2514static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2515 const int offset)
2516{
2517 skb_reset_inner_transport_header(skb);
2518 skb->inner_transport_header += offset;
2519}
2520
2521static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2522{
2523 return skb->head + skb->inner_network_header;
2524}
2525
2526static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2527{
2528 skb->inner_network_header = skb->data - skb->head;
2529}
2530
2531static inline void skb_set_inner_network_header(struct sk_buff *skb,
2532 const int offset)
2533{
2534 skb_reset_inner_network_header(skb);
2535 skb->inner_network_header += offset;
2536}
2537
aefbd2b3
PS
2538static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2539{
2540 return skb->head + skb->inner_mac_header;
2541}
2542
2543static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2544{
2545 skb->inner_mac_header = skb->data - skb->head;
2546}
2547
2548static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2549 const int offset)
2550{
2551 skb_reset_inner_mac_header(skb);
2552 skb->inner_mac_header += offset;
2553}
fda55eca
ED
2554static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2555{
35d04610 2556 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2557}
2558
9c70220b
ACM
2559static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2560{
2e07fa9c 2561 return skb->head + skb->transport_header;
9c70220b
ACM
2562}
2563
badff6d0
ACM
2564static inline void skb_reset_transport_header(struct sk_buff *skb)
2565{
2e07fa9c 2566 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2567}
2568
967b05f6
ACM
2569static inline void skb_set_transport_header(struct sk_buff *skb,
2570 const int offset)
2571{
2e07fa9c
ACM
2572 skb_reset_transport_header(skb);
2573 skb->transport_header += offset;
ea2ae17d
ACM
2574}
2575
d56f90a7
ACM
2576static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2577{
2e07fa9c 2578 return skb->head + skb->network_header;
d56f90a7
ACM
2579}
2580
c1d2bbe1
ACM
2581static inline void skb_reset_network_header(struct sk_buff *skb)
2582{
2e07fa9c 2583 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2584}
2585
c14d2450
ACM
2586static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2587{
2e07fa9c
ACM
2588 skb_reset_network_header(skb);
2589 skb->network_header += offset;
c14d2450
ACM
2590}
2591
2e07fa9c 2592static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2593{
2e07fa9c 2594 return skb->head + skb->mac_header;
bbe735e4
ACM
2595}
2596
ea6da4fd
AV
2597static inline int skb_mac_offset(const struct sk_buff *skb)
2598{
2599 return skb_mac_header(skb) - skb->data;
2600}
2601
0daf4349
DB
2602static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2603{
2604 return skb->network_header - skb->mac_header;
2605}
2606
2e07fa9c 2607static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2608{
35d04610 2609 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2610}
2611
b4ab3141
DB
2612static inline void skb_unset_mac_header(struct sk_buff *skb)
2613{
2614 skb->mac_header = (typeof(skb->mac_header))~0U;
2615}
2616
2e07fa9c
ACM
2617static inline void skb_reset_mac_header(struct sk_buff *skb)
2618{
2619 skb->mac_header = skb->data - skb->head;
2620}
2621
2622static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2623{
2624 skb_reset_mac_header(skb);
2625 skb->mac_header += offset;
2626}
2627
0e3da5bb
TT
2628static inline void skb_pop_mac_header(struct sk_buff *skb)
2629{
2630 skb->mac_header = skb->network_header;
2631}
2632
d2aa125d 2633static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2634{
72a338bc 2635 struct flow_keys_basic keys;
fbbdb8f0
YX
2636
2637 if (skb_transport_header_was_set(skb))
2638 return;
72a338bc 2639
3cbf4ffb
SF
2640 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2641 NULL, 0, 0, 0, 0))
42aecaa9 2642 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2643}
2644
03606895
ED
2645static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2646{
2647 if (skb_mac_header_was_set(skb)) {
2648 const unsigned char *old_mac = skb_mac_header(skb);
2649
2650 skb_set_mac_header(skb, -skb->mac_len);
2651 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2652 }
2653}
2654
04fb451e
MM
2655static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2656{
2657 return skb->csum_start - skb_headroom(skb);
2658}
2659
08b64fcc
AD
2660static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2661{
2662 return skb->head + skb->csum_start;
2663}
2664
2e07fa9c
ACM
2665static inline int skb_transport_offset(const struct sk_buff *skb)
2666{
2667 return skb_transport_header(skb) - skb->data;
2668}
2669
2670static inline u32 skb_network_header_len(const struct sk_buff *skb)
2671{
2672 return skb->transport_header - skb->network_header;
2673}
2674
6a674e9c
JG
2675static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2676{
2677 return skb->inner_transport_header - skb->inner_network_header;
2678}
2679
2e07fa9c
ACM
2680static inline int skb_network_offset(const struct sk_buff *skb)
2681{
2682 return skb_network_header(skb) - skb->data;
2683}
48d49d0c 2684
6a674e9c
JG
2685static inline int skb_inner_network_offset(const struct sk_buff *skb)
2686{
2687 return skb_inner_network_header(skb) - skb->data;
2688}
2689
f9599ce1
CG
2690static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2691{
2692 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2693}
2694
1da177e4
LT
2695/*
2696 * CPUs often take a performance hit when accessing unaligned memory
2697 * locations. The actual performance hit varies, it can be small if the
2698 * hardware handles it or large if we have to take an exception and fix it
2699 * in software.
2700 *
2701 * Since an ethernet header is 14 bytes network drivers often end up with
2702 * the IP header at an unaligned offset. The IP header can be aligned by
2703 * shifting the start of the packet by 2 bytes. Drivers should do this
2704 * with:
2705 *
8660c124 2706 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2707 *
2708 * The downside to this alignment of the IP header is that the DMA is now
2709 * unaligned. On some architectures the cost of an unaligned DMA is high
2710 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2711 *
1da177e4
LT
2712 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2713 * to be overridden.
2714 */
2715#ifndef NET_IP_ALIGN
2716#define NET_IP_ALIGN 2
2717#endif
2718
025be81e
AB
2719/*
2720 * The networking layer reserves some headroom in skb data (via
2721 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2722 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2723 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2724 *
2725 * Unfortunately this headroom changes the DMA alignment of the resulting
2726 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2727 * on some architectures. An architecture can override this value,
2728 * perhaps setting it to a cacheline in size (since that will maintain
2729 * cacheline alignment of the DMA). It must be a power of 2.
2730 *
d6301d3d 2731 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2732 * headroom, you should not reduce this.
5933dd2f
ED
2733 *
2734 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2735 * to reduce average number of cache lines per packet.
645f0897 2736 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2737 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2738 */
2739#ifndef NET_SKB_PAD
5933dd2f 2740#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2741#endif
2742
7965bd4d 2743int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2744
5293efe6 2745static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2746{
5e1abdc3 2747 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2748 return;
27a884dc
ACM
2749 skb->len = len;
2750 skb_set_tail_pointer(skb, len);
1da177e4
LT
2751}
2752
5293efe6
DB
2753static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2754{
2755 __skb_set_length(skb, len);
2756}
2757
7965bd4d 2758void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2759
2760static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2761{
3cc0e873
HX
2762 if (skb->data_len)
2763 return ___pskb_trim(skb, len);
2764 __skb_trim(skb, len);
2765 return 0;
1da177e4
LT
2766}
2767
2768static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2769{
2770 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2771}
2772
e9fa4f7b
HX
2773/**
2774 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2775 * @skb: buffer to alter
2776 * @len: new length
2777 *
2778 * This is identical to pskb_trim except that the caller knows that
2779 * the skb is not cloned so we should never get an error due to out-
2780 * of-memory.
2781 */
2782static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2783{
2784 int err = pskb_trim(skb, len);
2785 BUG_ON(err);
2786}
2787
5293efe6
DB
2788static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2789{
2790 unsigned int diff = len - skb->len;
2791
2792 if (skb_tailroom(skb) < diff) {
2793 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2794 GFP_ATOMIC);
2795 if (ret)
2796 return ret;
2797 }
2798 __skb_set_length(skb, len);
2799 return 0;
2800}
2801
1da177e4
LT
2802/**
2803 * skb_orphan - orphan a buffer
2804 * @skb: buffer to orphan
2805 *
2806 * If a buffer currently has an owner then we call the owner's
2807 * destructor function and make the @skb unowned. The buffer continues
2808 * to exist but is no longer charged to its former owner.
2809 */
2810static inline void skb_orphan(struct sk_buff *skb)
2811{
c34a7612 2812 if (skb->destructor) {
1da177e4 2813 skb->destructor(skb);
c34a7612
ED
2814 skb->destructor = NULL;
2815 skb->sk = NULL;
376c7311
ED
2816 } else {
2817 BUG_ON(skb->sk);
c34a7612 2818 }
1da177e4
LT
2819}
2820
a353e0ce
MT
2821/**
2822 * skb_orphan_frags - orphan the frags contained in a buffer
2823 * @skb: buffer to orphan frags from
2824 * @gfp_mask: allocation mask for replacement pages
2825 *
2826 * For each frag in the SKB which needs a destructor (i.e. has an
2827 * owner) create a copy of that frag and release the original
2828 * page by calling the destructor.
2829 */
2830static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2831{
1f8b977a
WB
2832 if (likely(!skb_zcopy(skb)))
2833 return 0;
185ce5c3 2834 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 2835 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
2836 return 0;
2837 return skb_copy_ubufs(skb, gfp_mask);
2838}
2839
2840/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2841static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2842{
2843 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2844 return 0;
2845 return skb_copy_ubufs(skb, gfp_mask);
2846}
2847
1da177e4
LT
2848/**
2849 * __skb_queue_purge - empty a list
2850 * @list: list to empty
2851 *
2852 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2853 * the list and one reference dropped. This function does not take the
2854 * list lock and the caller must hold the relevant locks to use it.
2855 */
1da177e4
LT
2856static inline void __skb_queue_purge(struct sk_buff_head *list)
2857{
2858 struct sk_buff *skb;
2859 while ((skb = __skb_dequeue(list)) != NULL)
2860 kfree_skb(skb);
2861}
4ea7b0cf 2862void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2863
385114de 2864unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2865
3f6e687d
KH
2866void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2867
2868/**
2869 * netdev_alloc_frag - allocate a page fragment
2870 * @fragsz: fragment size
2871 *
2872 * Allocates a frag from a page for receive buffer.
2873 * Uses GFP_ATOMIC allocations.
2874 */
2875static inline void *netdev_alloc_frag(unsigned int fragsz)
2876{
2877 return __netdev_alloc_frag_align(fragsz, ~0u);
2878}
2879
2880static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2881 unsigned int align)
2882{
2883 WARN_ON_ONCE(!is_power_of_2(align));
2884 return __netdev_alloc_frag_align(fragsz, -align);
2885}
1da177e4 2886
7965bd4d
JP
2887struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2888 gfp_t gfp_mask);
8af27456
CH
2889
2890/**
2891 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2892 * @dev: network device to receive on
2893 * @length: length to allocate
2894 *
2895 * Allocate a new &sk_buff and assign it a usage count of one. The
2896 * buffer has unspecified headroom built in. Users should allocate
2897 * the headroom they think they need without accounting for the
2898 * built in space. The built in space is used for optimisations.
2899 *
2900 * %NULL is returned if there is no free memory. Although this function
2901 * allocates memory it can be called from an interrupt.
2902 */
2903static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2904 unsigned int length)
8af27456
CH
2905{
2906 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2907}
2908
6f532612
ED
2909/* legacy helper around __netdev_alloc_skb() */
2910static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2911 gfp_t gfp_mask)
2912{
2913 return __netdev_alloc_skb(NULL, length, gfp_mask);
2914}
2915
2916/* legacy helper around netdev_alloc_skb() */
2917static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2918{
2919 return netdev_alloc_skb(NULL, length);
2920}
2921
2922
4915a0de
ED
2923static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2924 unsigned int length, gfp_t gfp)
61321bbd 2925{
4915a0de 2926 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2927
2928 if (NET_IP_ALIGN && skb)
2929 skb_reserve(skb, NET_IP_ALIGN);
2930 return skb;
2931}
2932
4915a0de
ED
2933static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2934 unsigned int length)
2935{
2936 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2937}
2938
181edb2b
AD
2939static inline void skb_free_frag(void *addr)
2940{
8c2dd3e4 2941 page_frag_free(addr);
181edb2b
AD
2942}
2943
3f6e687d
KH
2944void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2945
2946static inline void *napi_alloc_frag(unsigned int fragsz)
2947{
2948 return __napi_alloc_frag_align(fragsz, ~0u);
2949}
2950
2951static inline void *napi_alloc_frag_align(unsigned int fragsz,
2952 unsigned int align)
2953{
2954 WARN_ON_ONCE(!is_power_of_2(align));
2955 return __napi_alloc_frag_align(fragsz, -align);
2956}
2957
fd11a83d
AD
2958struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2959 unsigned int length, gfp_t gfp_mask);
2960static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2961 unsigned int length)
2962{
2963 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2964}
795bb1c0
JDB
2965void napi_consume_skb(struct sk_buff *skb, int budget);
2966
9243adfc 2967void napi_skb_free_stolen_head(struct sk_buff *skb);
15fad714 2968void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2969
71dfda58
AD
2970/**
2971 * __dev_alloc_pages - allocate page for network Rx
2972 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2973 * @order: size of the allocation
2974 *
2975 * Allocate a new page.
2976 *
2977 * %NULL is returned if there is no free memory.
2978*/
2979static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2980 unsigned int order)
2981{
2982 /* This piece of code contains several assumptions.
2983 * 1. This is for device Rx, therefor a cold page is preferred.
2984 * 2. The expectation is the user wants a compound page.
2985 * 3. If requesting a order 0 page it will not be compound
2986 * due to the check to see if order has a value in prep_new_page
2987 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2988 * code in gfp_to_alloc_flags that should be enforcing this.
2989 */
453f85d4 2990 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2991
2992 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2993}
2994
2995static inline struct page *dev_alloc_pages(unsigned int order)
2996{
95829b3a 2997 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2998}
2999
3000/**
3001 * __dev_alloc_page - allocate a page for network Rx
3002 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3003 *
3004 * Allocate a new page.
3005 *
3006 * %NULL is returned if there is no free memory.
3007 */
3008static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3009{
3010 return __dev_alloc_pages(gfp_mask, 0);
3011}
3012
3013static inline struct page *dev_alloc_page(void)
3014{
95829b3a 3015 return dev_alloc_pages(0);
71dfda58
AD
3016}
3017
bc38f30f
AL
3018/**
3019 * dev_page_is_reusable - check whether a page can be reused for network Rx
3020 * @page: the page to test
3021 *
3022 * A page shouldn't be considered for reusing/recycling if it was allocated
3023 * under memory pressure or at a distant memory node.
3024 *
3025 * Returns false if this page should be returned to page allocator, true
3026 * otherwise.
3027 */
3028static inline bool dev_page_is_reusable(const struct page *page)
3029{
3030 return likely(page_to_nid(page) == numa_mem_id() &&
3031 !page_is_pfmemalloc(page));
3032}
3033
0614002b
MG
3034/**
3035 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3036 * @page: The page that was allocated from skb_alloc_page
3037 * @skb: The skb that may need pfmemalloc set
3038 */
48f971c9
AL
3039static inline void skb_propagate_pfmemalloc(const struct page *page,
3040 struct sk_buff *skb)
0614002b 3041{
2f064f34 3042 if (page_is_pfmemalloc(page))
0614002b
MG
3043 skb->pfmemalloc = true;
3044}
3045
7240b60c
JL
3046/**
3047 * skb_frag_off() - Returns the offset of a skb fragment
3048 * @frag: the paged fragment
3049 */
3050static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3051{
65c84f14 3052 return frag->bv_offset;
7240b60c
JL
3053}
3054
3055/**
3056 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3057 * @frag: skb fragment
3058 * @delta: value to add
3059 */
3060static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3061{
65c84f14 3062 frag->bv_offset += delta;
7240b60c
JL
3063}
3064
3065/**
3066 * skb_frag_off_set() - Sets the offset of a skb fragment
3067 * @frag: skb fragment
3068 * @offset: offset of fragment
3069 */
3070static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3071{
65c84f14 3072 frag->bv_offset = offset;
7240b60c
JL
3073}
3074
3075/**
3076 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3077 * @fragto: skb fragment where offset is set
3078 * @fragfrom: skb fragment offset is copied from
3079 */
3080static inline void skb_frag_off_copy(skb_frag_t *fragto,
3081 const skb_frag_t *fragfrom)
3082{
65c84f14 3083 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
3084}
3085
131ea667 3086/**
e227867f 3087 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
3088 * @frag: the paged fragment
3089 *
3090 * Returns the &struct page associated with @frag.
3091 */
3092static inline struct page *skb_frag_page(const skb_frag_t *frag)
3093{
1dfa5bd3 3094 return frag->bv_page;
131ea667
IC
3095}
3096
3097/**
3098 * __skb_frag_ref - take an addition reference on a paged fragment.
3099 * @frag: the paged fragment
3100 *
3101 * Takes an additional reference on the paged fragment @frag.
3102 */
3103static inline void __skb_frag_ref(skb_frag_t *frag)
3104{
3105 get_page(skb_frag_page(frag));
3106}
3107
3108/**
3109 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3110 * @skb: the buffer
3111 * @f: the fragment offset.
3112 *
3113 * Takes an additional reference on the @f'th paged fragment of @skb.
3114 */
3115static inline void skb_frag_ref(struct sk_buff *skb, int f)
3116{
3117 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3118}
3119
3120/**
3121 * __skb_frag_unref - release a reference on a paged fragment.
3122 * @frag: the paged fragment
c420c989 3123 * @recycle: recycle the page if allocated via page_pool
131ea667 3124 *
c420c989
MC
3125 * Releases a reference on the paged fragment @frag
3126 * or recycles the page via the page_pool API.
131ea667 3127 */
c420c989 3128static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
131ea667 3129{
6a5bcd84
IA
3130 struct page *page = skb_frag_page(frag);
3131
3132#ifdef CONFIG_PAGE_POOL
3133 if (recycle && page_pool_return_skb_page(page))
3134 return;
3135#endif
3136 put_page(page);
131ea667
IC
3137}
3138
3139/**
3140 * skb_frag_unref - release a reference on a paged fragment of an skb.
3141 * @skb: the buffer
3142 * @f: the fragment offset
3143 *
3144 * Releases a reference on the @f'th paged fragment of @skb.
3145 */
3146static inline void skb_frag_unref(struct sk_buff *skb, int f)
3147{
6a5bcd84 3148 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
131ea667
IC
3149}
3150
3151/**
3152 * skb_frag_address - gets the address of the data contained in a paged fragment
3153 * @frag: the paged fragment buffer
3154 *
3155 * Returns the address of the data within @frag. The page must already
3156 * be mapped.
3157 */
3158static inline void *skb_frag_address(const skb_frag_t *frag)
3159{
7240b60c 3160 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3161}
3162
3163/**
3164 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3165 * @frag: the paged fragment buffer
3166 *
3167 * Returns the address of the data within @frag. Checks that the page
3168 * is mapped and returns %NULL otherwise.
3169 */
3170static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3171{
3172 void *ptr = page_address(skb_frag_page(frag));
3173 if (unlikely(!ptr))
3174 return NULL;
3175
7240b60c
JL
3176 return ptr + skb_frag_off(frag);
3177}
3178
3179/**
3180 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3181 * @fragto: skb fragment where page is set
3182 * @fragfrom: skb fragment page is copied from
3183 */
3184static inline void skb_frag_page_copy(skb_frag_t *fragto,
3185 const skb_frag_t *fragfrom)
3186{
3187 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3188}
3189
3190/**
3191 * __skb_frag_set_page - sets the page contained in a paged fragment
3192 * @frag: the paged fragment
3193 * @page: the page to set
3194 *
3195 * Sets the fragment @frag to contain @page.
3196 */
3197static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3198{
1dfa5bd3 3199 frag->bv_page = page;
131ea667
IC
3200}
3201
3202/**
3203 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3204 * @skb: the buffer
3205 * @f: the fragment offset
3206 * @page: the page to set
3207 *
3208 * Sets the @f'th fragment of @skb to contain @page.
3209 */
3210static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3211 struct page *page)
3212{
3213 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3214}
3215
400dfd3a
ED
3216bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3217
131ea667
IC
3218/**
3219 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3220 * @dev: the device to map the fragment to
131ea667
IC
3221 * @frag: the paged fragment to map
3222 * @offset: the offset within the fragment (starting at the
3223 * fragment's own offset)
3224 * @size: the number of bytes to map
771b00a8 3225 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3226 *
3227 * Maps the page associated with @frag to @device.
3228 */
3229static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3230 const skb_frag_t *frag,
3231 size_t offset, size_t size,
3232 enum dma_data_direction dir)
3233{
3234 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3235 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3236}
3237
117632e6
ED
3238static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3239 gfp_t gfp_mask)
3240{
3241 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3242}
3243
bad93e9d
OP
3244
3245static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3246 gfp_t gfp_mask)
3247{
3248 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3249}
3250
3251
334a8132
PM
3252/**
3253 * skb_clone_writable - is the header of a clone writable
3254 * @skb: buffer to check
3255 * @len: length up to which to write
3256 *
3257 * Returns true if modifying the header part of the cloned buffer
3258 * does not requires the data to be copied.
3259 */
05bdd2f1 3260static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3261{
3262 return !skb_header_cloned(skb) &&
3263 skb_headroom(skb) + len <= skb->hdr_len;
3264}
3265
3697649f
DB
3266static inline int skb_try_make_writable(struct sk_buff *skb,
3267 unsigned int write_len)
3268{
3269 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3270 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3271}
3272
d9cc2048
HX
3273static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3274 int cloned)
3275{
3276 int delta = 0;
3277
d9cc2048
HX
3278 if (headroom > skb_headroom(skb))
3279 delta = headroom - skb_headroom(skb);
3280
3281 if (delta || cloned)
3282 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3283 GFP_ATOMIC);
3284 return 0;
3285}
3286
1da177e4
LT
3287/**
3288 * skb_cow - copy header of skb when it is required
3289 * @skb: buffer to cow
3290 * @headroom: needed headroom
3291 *
3292 * If the skb passed lacks sufficient headroom or its data part
3293 * is shared, data is reallocated. If reallocation fails, an error
3294 * is returned and original skb is not changed.
3295 *
3296 * The result is skb with writable area skb->head...skb->tail
3297 * and at least @headroom of space at head.
3298 */
3299static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3300{
d9cc2048
HX
3301 return __skb_cow(skb, headroom, skb_cloned(skb));
3302}
1da177e4 3303
d9cc2048
HX
3304/**
3305 * skb_cow_head - skb_cow but only making the head writable
3306 * @skb: buffer to cow
3307 * @headroom: needed headroom
3308 *
3309 * This function is identical to skb_cow except that we replace the
3310 * skb_cloned check by skb_header_cloned. It should be used when
3311 * you only need to push on some header and do not need to modify
3312 * the data.
3313 */
3314static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3315{
3316 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3317}
3318
3319/**
3320 * skb_padto - pad an skbuff up to a minimal size
3321 * @skb: buffer to pad
3322 * @len: minimal length
3323 *
3324 * Pads up a buffer to ensure the trailing bytes exist and are
3325 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3326 * is untouched. Otherwise it is extended. Returns zero on
3327 * success. The skb is freed on error.
1da177e4 3328 */
5b057c6b 3329static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3330{
3331 unsigned int size = skb->len;
3332 if (likely(size >= len))
5b057c6b 3333 return 0;
987c402a 3334 return skb_pad(skb, len - size);
1da177e4
LT
3335}
3336
9c0c1124 3337/**
4ea7b0cf 3338 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3339 * @skb: buffer to pad
3340 * @len: minimal length
cd0a137a 3341 * @free_on_error: free buffer on error
9c0c1124
AD
3342 *
3343 * Pads up a buffer to ensure the trailing bytes exist and are
3344 * blanked. If the buffer already contains sufficient data it
3345 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3346 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3347 */
4a009cb0
ED
3348static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3349 unsigned int len,
3350 bool free_on_error)
9c0c1124
AD
3351{
3352 unsigned int size = skb->len;
3353
3354 if (unlikely(size < len)) {
3355 len -= size;
cd0a137a 3356 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3357 return -ENOMEM;
3358 __skb_put(skb, len);
3359 }
3360 return 0;
3361}
3362
cd0a137a
FF
3363/**
3364 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3365 * @skb: buffer to pad
3366 * @len: minimal length
3367 *
3368 * Pads up a buffer to ensure the trailing bytes exist and are
3369 * blanked. If the buffer already contains sufficient data it
3370 * is untouched. Otherwise it is extended. Returns zero on
3371 * success. The skb is freed on error.
3372 */
4a009cb0 3373static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3374{
3375 return __skb_put_padto(skb, len, true);
3376}
3377
1da177e4 3378static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3379 struct iov_iter *from, int copy)
1da177e4
LT
3380{
3381 const int off = skb->len;
3382
3383 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3384 __wsum csum = 0;
15e6cb46
AV
3385 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3386 &csum, from)) {
1da177e4
LT
3387 skb->csum = csum_block_add(skb->csum, csum, off);
3388 return 0;
3389 }
15e6cb46 3390 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3391 return 0;
3392
3393 __skb_trim(skb, off);
3394 return -EFAULT;
3395}
3396
38ba0a65
ED
3397static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3398 const struct page *page, int off)
1da177e4 3399{
1f8b977a
WB
3400 if (skb_zcopy(skb))
3401 return false;
1da177e4 3402 if (i) {
d8e18a51 3403 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3404
ea2ab693 3405 return page == skb_frag_page(frag) &&
7240b60c 3406 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3407 }
38ba0a65 3408 return false;
1da177e4
LT
3409}
3410
364c6bad
HX
3411static inline int __skb_linearize(struct sk_buff *skb)
3412{
3413 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3414}
3415
1da177e4
LT
3416/**
3417 * skb_linearize - convert paged skb to linear one
3418 * @skb: buffer to linarize
1da177e4
LT
3419 *
3420 * If there is no free memory -ENOMEM is returned, otherwise zero
3421 * is returned and the old skb data released.
3422 */
364c6bad
HX
3423static inline int skb_linearize(struct sk_buff *skb)
3424{
3425 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3426}
3427
cef401de
ED
3428/**
3429 * skb_has_shared_frag - can any frag be overwritten
3430 * @skb: buffer to test
3431 *
3432 * Return true if the skb has at least one frag that might be modified
3433 * by an external entity (as in vmsplice()/sendfile())
3434 */
3435static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3436{
c9af6db4 3437 return skb_is_nonlinear(skb) &&
06b4feb3 3438 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3439}
3440
364c6bad
HX
3441/**
3442 * skb_linearize_cow - make sure skb is linear and writable
3443 * @skb: buffer to process
3444 *
3445 * If there is no free memory -ENOMEM is returned, otherwise zero
3446 * is returned and the old skb data released.
3447 */
3448static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3449{
364c6bad
HX
3450 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3451 __skb_linearize(skb) : 0;
1da177e4
LT
3452}
3453
479ffccc
DB
3454static __always_inline void
3455__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3456 unsigned int off)
3457{
3458 if (skb->ip_summed == CHECKSUM_COMPLETE)
3459 skb->csum = csum_block_sub(skb->csum,
3460 csum_partial(start, len, 0), off);
3461 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3462 skb_checksum_start_offset(skb) < 0)
3463 skb->ip_summed = CHECKSUM_NONE;
3464}
3465
1da177e4
LT
3466/**
3467 * skb_postpull_rcsum - update checksum for received skb after pull
3468 * @skb: buffer to update
3469 * @start: start of data before pull
3470 * @len: length of data pulled
3471 *
3472 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3473 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3474 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3475 */
1da177e4 3476static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3477 const void *start, unsigned int len)
1da177e4 3478{
479ffccc 3479 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3480}
3481
479ffccc
DB
3482static __always_inline void
3483__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3484 unsigned int off)
3485{
3486 if (skb->ip_summed == CHECKSUM_COMPLETE)
3487 skb->csum = csum_block_add(skb->csum,
3488 csum_partial(start, len, 0), off);
3489}
cbb042f9 3490
479ffccc
DB
3491/**
3492 * skb_postpush_rcsum - update checksum for received skb after push
3493 * @skb: buffer to update
3494 * @start: start of data after push
3495 * @len: length of data pushed
3496 *
3497 * After doing a push on a received packet, you need to call this to
3498 * update the CHECKSUM_COMPLETE checksum.
3499 */
f8ffad69
DB
3500static inline void skb_postpush_rcsum(struct sk_buff *skb,
3501 const void *start, unsigned int len)
3502{
479ffccc 3503 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3504}
3505
af72868b 3506void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3507
82a31b92
WC
3508/**
3509 * skb_push_rcsum - push skb and update receive checksum
3510 * @skb: buffer to update
3511 * @len: length of data pulled
3512 *
3513 * This function performs an skb_push on the packet and updates
3514 * the CHECKSUM_COMPLETE checksum. It should be used on
3515 * receive path processing instead of skb_push unless you know
3516 * that the checksum difference is zero (e.g., a valid IP header)
3517 * or you are setting ip_summed to CHECKSUM_NONE.
3518 */
d58ff351 3519static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3520{
3521 skb_push(skb, len);
3522 skb_postpush_rcsum(skb, skb->data, len);
3523 return skb->data;
3524}
3525
88078d98 3526int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3527/**
3528 * pskb_trim_rcsum - trim received skb and update checksum
3529 * @skb: buffer to trim
3530 * @len: new length
3531 *
3532 * This is exactly the same as pskb_trim except that it ensures the
3533 * checksum of received packets are still valid after the operation.
6c57f045 3534 * It can change skb pointers.
7ce5a27f
DM
3535 */
3536
3537static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3538{
3539 if (likely(len >= skb->len))
3540 return 0;
88078d98 3541 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3542}
3543
5293efe6
DB
3544static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3545{
3546 if (skb->ip_summed == CHECKSUM_COMPLETE)
3547 skb->ip_summed = CHECKSUM_NONE;
3548 __skb_trim(skb, len);
3549 return 0;
3550}
3551
3552static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3553{
3554 if (skb->ip_summed == CHECKSUM_COMPLETE)
3555 skb->ip_summed = CHECKSUM_NONE;
3556 return __skb_grow(skb, len);
3557}
3558
18a4c0ea
ED
3559#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3560#define skb_rb_first(root) rb_to_skb(rb_first(root))
3561#define skb_rb_last(root) rb_to_skb(rb_last(root))
3562#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3563#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3564
1da177e4
LT
3565#define skb_queue_walk(queue, skb) \
3566 for (skb = (queue)->next; \
a1e4891f 3567 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3568 skb = skb->next)
3569
46f8914e
JC
3570#define skb_queue_walk_safe(queue, skb, tmp) \
3571 for (skb = (queue)->next, tmp = skb->next; \
3572 skb != (struct sk_buff *)(queue); \
3573 skb = tmp, tmp = skb->next)
3574
1164f52a 3575#define skb_queue_walk_from(queue, skb) \
a1e4891f 3576 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3577 skb = skb->next)
3578
18a4c0ea
ED
3579#define skb_rbtree_walk(skb, root) \
3580 for (skb = skb_rb_first(root); skb != NULL; \
3581 skb = skb_rb_next(skb))
3582
3583#define skb_rbtree_walk_from(skb) \
3584 for (; skb != NULL; \
3585 skb = skb_rb_next(skb))
3586
3587#define skb_rbtree_walk_from_safe(skb, tmp) \
3588 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3589 skb = tmp)
3590
1164f52a
DM
3591#define skb_queue_walk_from_safe(queue, skb, tmp) \
3592 for (tmp = skb->next; \
3593 skb != (struct sk_buff *)(queue); \
3594 skb = tmp, tmp = skb->next)
3595
300ce174
SH
3596#define skb_queue_reverse_walk(queue, skb) \
3597 for (skb = (queue)->prev; \
a1e4891f 3598 skb != (struct sk_buff *)(queue); \
300ce174
SH
3599 skb = skb->prev)
3600
686a2955
DM
3601#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3602 for (skb = (queue)->prev, tmp = skb->prev; \
3603 skb != (struct sk_buff *)(queue); \
3604 skb = tmp, tmp = skb->prev)
3605
3606#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3607 for (tmp = skb->prev; \
3608 skb != (struct sk_buff *)(queue); \
3609 skb = tmp, tmp = skb->prev)
1da177e4 3610
21dc3301 3611static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3612{
3613 return skb_shinfo(skb)->frag_list != NULL;
3614}
3615
3616static inline void skb_frag_list_init(struct sk_buff *skb)
3617{
3618 skb_shinfo(skb)->frag_list = NULL;
3619}
3620
ee039871
DM
3621#define skb_walk_frags(skb, iter) \
3622 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3623
ea3793ee 3624
b50b0580
SD
3625int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3626 int *err, long *timeo_p,
ea3793ee 3627 const struct sk_buff *skb);
65101aec
PA
3628struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3629 struct sk_buff_head *queue,
3630 unsigned int flags,
fd69c399 3631 int *off, int *err,
65101aec 3632 struct sk_buff **last);
b50b0580
SD
3633struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3634 struct sk_buff_head *queue,
e427cad6 3635 unsigned int flags, int *off, int *err,
ea3793ee 3636 struct sk_buff **last);
b50b0580
SD
3637struct sk_buff *__skb_recv_datagram(struct sock *sk,
3638 struct sk_buff_head *sk_queue,
e427cad6 3639 unsigned int flags, int *off, int *err);
7965bd4d
JP
3640struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3641 int *err);
a11e1d43
LT
3642__poll_t datagram_poll(struct file *file, struct socket *sock,
3643 struct poll_table_struct *wait);
c0371da6
AV
3644int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3645 struct iov_iter *to, int size);
51f3d02b
DM
3646static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3647 struct msghdr *msg, int size)
3648{
e5a4b0bb 3649 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3650}
e5a4b0bb
AV
3651int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3652 struct msghdr *msg);
65d69e25
SG
3653int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3654 struct iov_iter *to, int len,
3655 struct ahash_request *hash);
3a654f97
AV
3656int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3657 struct iov_iter *from, int len);
3a654f97 3658int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3659void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3660void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3661static inline void skb_free_datagram_locked(struct sock *sk,
3662 struct sk_buff *skb)
3663{
3664 __skb_free_datagram_locked(sk, skb, 0);
3665}
7965bd4d 3666int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3667int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3668int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3669__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 3670 int len);
a60e3cc7 3671int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3672 struct pipe_inode_info *pipe, unsigned int len,
25869262 3673 unsigned int flags);
20bf50de
TH
3674int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3675 int len);
0739cd28 3676int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
7965bd4d 3677void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3678unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3679int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3680 int len, int hlen);
7965bd4d
JP
3681void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3682int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3683void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3684bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3685bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3686struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3687struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3688 unsigned int offset);
0d5501c1 3689struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3690int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3691int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3692int skb_vlan_pop(struct sk_buff *skb);
3693int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
3694int skb_eth_pop(struct sk_buff *skb);
3695int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3696 const unsigned char *src);
fa4e0f88 3697int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3698 int mac_len, bool ethernet);
040b5cfb
MV
3699int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3700 bool ethernet);
d27cf5c5 3701int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3702int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3703struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3704 gfp_t gfp);
20380731 3705
6ce8e9ce
AV
3706static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3707{
3073f070 3708 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3709}
3710
7eab8d9e
AV
3711static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3712{
e5a4b0bb 3713 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3714}
3715
2817a336
DB
3716struct skb_checksum_ops {
3717 __wsum (*update)(const void *mem, int len, __wsum wsum);
3718 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3719};
3720
9617813d
DC
3721extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3722
2817a336
DB
3723__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3724 __wsum csum, const struct skb_checksum_ops *ops);
3725__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3726 __wsum csum);
3727
1e98a0f0 3728static inline void * __must_check
e3305138
AL
3729__skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3730 const void *data, int hlen, void *buffer)
1da177e4 3731{
d206121f 3732 if (likely(hlen - offset >= len))
e3305138 3733 return (void *)data + offset;
1da177e4 3734
d206121f 3735 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
1da177e4
LT
3736 return NULL;
3737
3738 return buffer;
3739}
3740
1e98a0f0
ED
3741static inline void * __must_check
3742skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3743{
3744 return __skb_header_pointer(skb, offset, len, skb->data,
3745 skb_headlen(skb), buffer);
3746}
3747
4262e5cc
DB
3748/**
3749 * skb_needs_linearize - check if we need to linearize a given skb
3750 * depending on the given device features.
3751 * @skb: socket buffer to check
3752 * @features: net device features
3753 *
3754 * Returns true if either:
3755 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3756 * 2. skb is fragmented and the device does not support SG.
3757 */
3758static inline bool skb_needs_linearize(struct sk_buff *skb,
3759 netdev_features_t features)
3760{
3761 return skb_is_nonlinear(skb) &&
3762 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3763 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3764}
3765
d626f62b
ACM
3766static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3767 void *to,
3768 const unsigned int len)
3769{
3770 memcpy(to, skb->data, len);
3771}
3772
3773static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3774 const int offset, void *to,
3775 const unsigned int len)
3776{
3777 memcpy(to, skb->data + offset, len);
3778}
3779
27d7ff46
ACM
3780static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3781 const void *from,
3782 const unsigned int len)
3783{
3784 memcpy(skb->data, from, len);
3785}
3786
3787static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3788 const int offset,
3789 const void *from,
3790 const unsigned int len)
3791{
3792 memcpy(skb->data + offset, from, len);
3793}
3794
7965bd4d 3795void skb_init(void);
1da177e4 3796
ac45f602
PO
3797static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3798{
3799 return skb->tstamp;
3800}
3801
a61bbcf2
PM
3802/**
3803 * skb_get_timestamp - get timestamp from a skb
3804 * @skb: skb to get stamp from
13c6ee2a 3805 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3806 *
3807 * Timestamps are stored in the skb as offsets to a base timestamp.
3808 * This function converts the offset back to a struct timeval and stores
3809 * it in stamp.
3810 */
ac45f602 3811static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3812 struct __kernel_old_timeval *stamp)
a61bbcf2 3813{
13c6ee2a 3814 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3815}
3816
887feae3
DD
3817static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3818 struct __kernel_sock_timeval *stamp)
3819{
3820 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3821
3822 stamp->tv_sec = ts.tv_sec;
3823 stamp->tv_usec = ts.tv_nsec / 1000;
3824}
3825
ac45f602 3826static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3827 struct __kernel_old_timespec *stamp)
ac45f602 3828{
df1b4ba9
AB
3829 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3830
3831 stamp->tv_sec = ts.tv_sec;
3832 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3833}
3834
887feae3
DD
3835static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3836 struct __kernel_timespec *stamp)
3837{
3838 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3839
3840 stamp->tv_sec = ts.tv_sec;
3841 stamp->tv_nsec = ts.tv_nsec;
3842}
3843
b7aa0bf7 3844static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3845{
b7aa0bf7 3846 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3847}
3848
164891aa
SH
3849static inline ktime_t net_timedelta(ktime_t t)
3850{
3851 return ktime_sub(ktime_get_real(), t);
3852}
3853
b9ce204f
IJ
3854static inline ktime_t net_invalid_timestamp(void)
3855{
8b0e1953 3856 return 0;
b9ce204f 3857}
a61bbcf2 3858
de8f3a83
DB
3859static inline u8 skb_metadata_len(const struct sk_buff *skb)
3860{
3861 return skb_shinfo(skb)->meta_len;
3862}
3863
3864static inline void *skb_metadata_end(const struct sk_buff *skb)
3865{
3866 return skb_mac_header(skb);
3867}
3868
3869static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3870 const struct sk_buff *skb_b,
3871 u8 meta_len)
3872{
3873 const void *a = skb_metadata_end(skb_a);
3874 const void *b = skb_metadata_end(skb_b);
3875 /* Using more efficient varaiant than plain call to memcmp(). */
3876#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3877 u64 diffs = 0;
3878
3879 switch (meta_len) {
3880#define __it(x, op) (x -= sizeof(u##op))
3881#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3882 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3883 fallthrough;
de8f3a83 3884 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3885 fallthrough;
de8f3a83 3886 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3887 fallthrough;
de8f3a83
DB
3888 case 8: diffs |= __it_diff(a, b, 64);
3889 break;
3890 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3891 fallthrough;
de8f3a83 3892 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3893 fallthrough;
de8f3a83 3894 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3895 fallthrough;
de8f3a83
DB
3896 case 4: diffs |= __it_diff(a, b, 32);
3897 break;
3898 }
3899 return diffs;
3900#else
3901 return memcmp(a - meta_len, b - meta_len, meta_len);
3902#endif
3903}
3904
3905static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3906 const struct sk_buff *skb_b)
3907{
3908 u8 len_a = skb_metadata_len(skb_a);
3909 u8 len_b = skb_metadata_len(skb_b);
3910
3911 if (!(len_a | len_b))
3912 return false;
3913
3914 return len_a != len_b ?
3915 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3916}
3917
3918static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3919{
3920 skb_shinfo(skb)->meta_len = meta_len;
3921}
3922
3923static inline void skb_metadata_clear(struct sk_buff *skb)
3924{
3925 skb_metadata_set(skb, 0);
3926}
3927
62bccb8c
AD
3928struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3929
c1f19b51
RC
3930#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3931
7965bd4d
JP
3932void skb_clone_tx_timestamp(struct sk_buff *skb);
3933bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3934
3935#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3936
3937static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3938{
3939}
3940
3941static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3942{
3943 return false;
3944}
3945
3946#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3947
3948/**
3949 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3950 *
da92b194
RC
3951 * PHY drivers may accept clones of transmitted packets for
3952 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3953 * must call this function to return the skb back to the stack with a
3954 * timestamp.
da92b194 3955 *
2ff17117 3956 * @skb: clone of the original outgoing packet
7a76a021 3957 * @hwtstamps: hardware time stamps
c1f19b51
RC
3958 *
3959 */
3960void skb_complete_tx_timestamp(struct sk_buff *skb,
3961 struct skb_shared_hwtstamps *hwtstamps);
3962
e7ed11ee 3963void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
3964 struct skb_shared_hwtstamps *hwtstamps,
3965 struct sock *sk, int tstype);
3966
ac45f602
PO
3967/**
3968 * skb_tstamp_tx - queue clone of skb with send time stamps
3969 * @orig_skb: the original outgoing packet
3970 * @hwtstamps: hardware time stamps, may be NULL if not available
3971 *
3972 * If the skb has a socket associated, then this function clones the
3973 * skb (thus sharing the actual data and optional structures), stores
3974 * the optional hardware time stamping information (if non NULL) or
3975 * generates a software time stamp (otherwise), then queues the clone
3976 * to the error queue of the socket. Errors are silently ignored.
3977 */
7965bd4d
JP
3978void skb_tstamp_tx(struct sk_buff *orig_skb,
3979 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3980
4507a715
RC
3981/**
3982 * skb_tx_timestamp() - Driver hook for transmit timestamping
3983 *
3984 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3985 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3986 *
73409f3b
DM
3987 * Specifically, one should make absolutely sure that this function is
3988 * called before TX completion of this packet can trigger. Otherwise
3989 * the packet could potentially already be freed.
3990 *
4507a715
RC
3991 * @skb: A socket buffer.
3992 */
3993static inline void skb_tx_timestamp(struct sk_buff *skb)
3994{
c1f19b51 3995 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3996 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3997 skb_tstamp_tx(skb, NULL);
4507a715
RC
3998}
3999
6e3e939f
JB
4000/**
4001 * skb_complete_wifi_ack - deliver skb with wifi status
4002 *
4003 * @skb: the original outgoing packet
4004 * @acked: ack status
4005 *
4006 */
4007void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4008
7965bd4d
JP
4009__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4010__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 4011
60476372
HX
4012static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4013{
6edec0e6
TH
4014 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4015 skb->csum_valid ||
4016 (skb->ip_summed == CHECKSUM_PARTIAL &&
4017 skb_checksum_start_offset(skb) >= 0));
60476372
HX
4018}
4019
fb286bb2
HX
4020/**
4021 * skb_checksum_complete - Calculate checksum of an entire packet
4022 * @skb: packet to process
4023 *
4024 * This function calculates the checksum over the entire packet plus
4025 * the value of skb->csum. The latter can be used to supply the
4026 * checksum of a pseudo header as used by TCP/UDP. It returns the
4027 * checksum.
4028 *
4029 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4030 * this function can be used to verify that checksum on received
4031 * packets. In that case the function should return zero if the
4032 * checksum is correct. In particular, this function will return zero
4033 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4034 * hardware has already verified the correctness of the checksum.
4035 */
4381ca3c 4036static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 4037{
60476372
HX
4038 return skb_csum_unnecessary(skb) ?
4039 0 : __skb_checksum_complete(skb);
fb286bb2
HX
4040}
4041
77cffe23
TH
4042static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4043{
4044 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4045 if (skb->csum_level == 0)
4046 skb->ip_summed = CHECKSUM_NONE;
4047 else
4048 skb->csum_level--;
4049 }
4050}
4051
4052static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4053{
4054 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4055 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4056 skb->csum_level++;
4057 } else if (skb->ip_summed == CHECKSUM_NONE) {
4058 skb->ip_summed = CHECKSUM_UNNECESSARY;
4059 skb->csum_level = 0;
4060 }
4061}
4062
836e66c2
DB
4063static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4064{
4065 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4066 skb->ip_summed = CHECKSUM_NONE;
4067 skb->csum_level = 0;
4068 }
4069}
4070
76ba0aae
TH
4071/* Check if we need to perform checksum complete validation.
4072 *
4073 * Returns true if checksum complete is needed, false otherwise
4074 * (either checksum is unnecessary or zero checksum is allowed).
4075 */
4076static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4077 bool zero_okay,
4078 __sum16 check)
4079{
5d0c2b95
TH
4080 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4081 skb->csum_valid = 1;
77cffe23 4082 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
4083 return false;
4084 }
4085
4086 return true;
4087}
4088
da279887 4089/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
4090 * in checksum_init.
4091 */
4092#define CHECKSUM_BREAK 76
4093
4e18b9ad
TH
4094/* Unset checksum-complete
4095 *
4096 * Unset checksum complete can be done when packet is being modified
4097 * (uncompressed for instance) and checksum-complete value is
4098 * invalidated.
4099 */
4100static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4101{
4102 if (skb->ip_summed == CHECKSUM_COMPLETE)
4103 skb->ip_summed = CHECKSUM_NONE;
4104}
4105
76ba0aae
TH
4106/* Validate (init) checksum based on checksum complete.
4107 *
4108 * Return values:
4109 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4110 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4111 * checksum is stored in skb->csum for use in __skb_checksum_complete
4112 * non-zero: value of invalid checksum
4113 *
4114 */
4115static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4116 bool complete,
4117 __wsum psum)
4118{
4119 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4120 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4121 skb->csum_valid = 1;
76ba0aae
TH
4122 return 0;
4123 }
4124 }
4125
4126 skb->csum = psum;
4127
5d0c2b95
TH
4128 if (complete || skb->len <= CHECKSUM_BREAK) {
4129 __sum16 csum;
4130
4131 csum = __skb_checksum_complete(skb);
4132 skb->csum_valid = !csum;
4133 return csum;
4134 }
76ba0aae
TH
4135
4136 return 0;
4137}
4138
4139static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4140{
4141 return 0;
4142}
4143
4144/* Perform checksum validate (init). Note that this is a macro since we only
4145 * want to calculate the pseudo header which is an input function if necessary.
4146 * First we try to validate without any computation (checksum unnecessary) and
4147 * then calculate based on checksum complete calling the function to compute
4148 * pseudo header.
4149 *
4150 * Return values:
4151 * 0: checksum is validated or try to in skb_checksum_complete
4152 * non-zero: value of invalid checksum
4153 */
4154#define __skb_checksum_validate(skb, proto, complete, \
4155 zero_okay, check, compute_pseudo) \
4156({ \
4157 __sum16 __ret = 0; \
5d0c2b95 4158 skb->csum_valid = 0; \
76ba0aae
TH
4159 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4160 __ret = __skb_checksum_validate_complete(skb, \
4161 complete, compute_pseudo(skb, proto)); \
4162 __ret; \
4163})
4164
4165#define skb_checksum_init(skb, proto, compute_pseudo) \
4166 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4167
4168#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4169 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4170
4171#define skb_checksum_validate(skb, proto, compute_pseudo) \
4172 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4173
4174#define skb_checksum_validate_zero_check(skb, proto, check, \
4175 compute_pseudo) \
096a4cfa 4176 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4177
4178#define skb_checksum_simple_validate(skb) \
4179 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4180
d96535a1
TH
4181static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4182{
219f1d79 4183 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4184}
4185
e4aa33ad 4186static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4187{
4188 skb->csum = ~pseudo;
4189 skb->ip_summed = CHECKSUM_COMPLETE;
4190}
4191
e4aa33ad 4192#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4193do { \
4194 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4195 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4196} while (0)
4197
15e2396d
TH
4198static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4199 u16 start, u16 offset)
4200{
4201 skb->ip_summed = CHECKSUM_PARTIAL;
4202 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4203 skb->csum_offset = offset - start;
4204}
4205
dcdc8994
TH
4206/* Update skbuf and packet to reflect the remote checksum offload operation.
4207 * When called, ptr indicates the starting point for skb->csum when
4208 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4209 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4210 */
4211static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4212 int start, int offset, bool nopartial)
dcdc8994
TH
4213{
4214 __wsum delta;
4215
15e2396d
TH
4216 if (!nopartial) {
4217 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4218 return;
4219 }
4220
dcdc8994
TH
4221 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4222 __skb_checksum_complete(skb);
4223 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4224 }
4225
4226 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4227
4228 /* Adjust skb->csum since we changed the packet */
4229 skb->csum = csum_add(skb->csum, delta);
4230}
4231
cb9c6836
FW
4232static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4233{
4234#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4235 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4236#else
4237 return NULL;
4238#endif
4239}
4240
261db6c2 4241static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4242{
261db6c2
JS
4243#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4244 return skb->_nfct;
4245#else
4246 return 0UL;
4247#endif
1da177e4 4248}
261db6c2
JS
4249
4250static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4251{
261db6c2 4252#if IS_ENABLED(CONFIG_NF_CONNTRACK)
5fc88f93 4253 skb->slow_gro |= !!nfct;
261db6c2 4254 skb->_nfct = nfct;
2fc72c7b 4255#endif
261db6c2 4256}
df5042f4
FW
4257
4258#ifdef CONFIG_SKB_EXTENSIONS
4259enum skb_ext_id {
4260#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4261 SKB_EXT_BRIDGE_NF,
4165079b
FW
4262#endif
4263#ifdef CONFIG_XFRM
4264 SKB_EXT_SEC_PATH,
95a7233c
PB
4265#endif
4266#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4267 TC_SKB_EXT,
3ee17bc7
MM
4268#endif
4269#if IS_ENABLED(CONFIG_MPTCP)
4270 SKB_EXT_MPTCP,
df5042f4
FW
4271#endif
4272 SKB_EXT_NUM, /* must be last */
4273};
4274
4275/**
4276 * struct skb_ext - sk_buff extensions
4277 * @refcnt: 1 on allocation, deallocated on 0
4278 * @offset: offset to add to @data to obtain extension address
4279 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4280 * @data: start of extension data, variable sized
4281 *
4282 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4283 * to use 'u8' types while allowing up to 2kb worth of extension data.
4284 */
4285struct skb_ext {
4286 refcount_t refcnt;
4287 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4288 u8 chunks; /* same */
5c91aa1d 4289 char data[] __aligned(8);
df5042f4
FW
4290};
4291
4930f483 4292struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4293void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4294 struct skb_ext *ext);
df5042f4
FW
4295void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4296void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4297void __skb_ext_put(struct skb_ext *ext);
4298
4299static inline void skb_ext_put(struct sk_buff *skb)
4300{
4301 if (skb->active_extensions)
4302 __skb_ext_put(skb->extensions);
4303}
4304
df5042f4
FW
4305static inline void __skb_ext_copy(struct sk_buff *dst,
4306 const struct sk_buff *src)
4307{
4308 dst->active_extensions = src->active_extensions;
4309
4310 if (src->active_extensions) {
4311 struct skb_ext *ext = src->extensions;
4312
4313 refcount_inc(&ext->refcnt);
4314 dst->extensions = ext;
4315 }
4316}
4317
4318static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4319{
4320 skb_ext_put(dst);
4321 __skb_ext_copy(dst, src);
4322}
4323
4324static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4325{
4326 return !!ext->offset[i];
4327}
4328
4329static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4330{
4331 return skb->active_extensions & (1 << id);
4332}
4333
4334static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4335{
4336 if (skb_ext_exist(skb, id))
4337 __skb_ext_del(skb, id);
4338}
4339
4340static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4341{
4342 if (skb_ext_exist(skb, id)) {
4343 struct skb_ext *ext = skb->extensions;
4344
4345 return (void *)ext + (ext->offset[id] << 3);
4346 }
4347
4348 return NULL;
4349}
174e2381
FW
4350
4351static inline void skb_ext_reset(struct sk_buff *skb)
4352{
4353 if (unlikely(skb->active_extensions)) {
4354 __skb_ext_put(skb->extensions);
4355 skb->active_extensions = 0;
4356 }
4357}
677bf08c
FW
4358
4359static inline bool skb_has_extensions(struct sk_buff *skb)
4360{
4361 return unlikely(skb->active_extensions);
4362}
df5042f4
FW
4363#else
4364static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4365static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4366static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4367static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4368static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4369static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4370#endif /* CONFIG_SKB_EXTENSIONS */
4371
895b5c9f 4372static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4373{
5f79e0f9 4374#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4375 nf_conntrack_put(skb_nfct(skb));
4376 skb->_nfct = 0;
2fc72c7b 4377#endif
a193a4ab
PM
4378}
4379
124dff01
PM
4380static inline void nf_reset_trace(struct sk_buff *skb)
4381{
478b360a 4382#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4383 skb->nf_trace = 0;
4384#endif
a193a4ab
PM
4385}
4386
2b5ec1a5
YY
4387static inline void ipvs_reset(struct sk_buff *skb)
4388{
4389#if IS_ENABLED(CONFIG_IP_VS)
4390 skb->ipvs_property = 0;
4391#endif
4392}
4393
de8bda1d 4394/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4395static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4396 bool copy)
edda553c 4397{
5f79e0f9 4398#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4399 dst->_nfct = src->_nfct;
4400 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4401#endif
478b360a 4402#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4403 if (copy)
4404 dst->nf_trace = src->nf_trace;
478b360a 4405#endif
edda553c
YK
4406}
4407
e7ac05f3
YK
4408static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4409{
e7ac05f3 4410#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4411 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4412#endif
5fc88f93 4413 dst->slow_gro = src->slow_gro;
b1937227 4414 __nf_copy(dst, src, true);
e7ac05f3
YK
4415}
4416
984bc16c
JM
4417#ifdef CONFIG_NETWORK_SECMARK
4418static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4419{
4420 to->secmark = from->secmark;
4421}
4422
4423static inline void skb_init_secmark(struct sk_buff *skb)
4424{
4425 skb->secmark = 0;
4426}
4427#else
4428static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4429{ }
4430
4431static inline void skb_init_secmark(struct sk_buff *skb)
4432{ }
4433#endif
4434
7af8f4ca
FW
4435static inline int secpath_exists(const struct sk_buff *skb)
4436{
4437#ifdef CONFIG_XFRM
4165079b 4438 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4439#else
4440 return 0;
4441#endif
4442}
4443
574f7194
EB
4444static inline bool skb_irq_freeable(const struct sk_buff *skb)
4445{
4446 return !skb->destructor &&
7af8f4ca 4447 !secpath_exists(skb) &&
cb9c6836 4448 !skb_nfct(skb) &&
574f7194
EB
4449 !skb->_skb_refdst &&
4450 !skb_has_frag_list(skb);
4451}
4452
f25f4e44
PWJ
4453static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4454{
f25f4e44 4455 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4456}
4457
9247744e 4458static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4459{
4e3ab47a 4460 return skb->queue_mapping;
4e3ab47a
PE
4461}
4462
f25f4e44
PWJ
4463static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4464{
f25f4e44 4465 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4466}
4467
d5a9e24a
DM
4468static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4469{
4470 skb->queue_mapping = rx_queue + 1;
4471}
4472
9247744e 4473static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4474{
4475 return skb->queue_mapping - 1;
4476}
4477
9247744e 4478static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4479{
a02cec21 4480 return skb->queue_mapping != 0;
d5a9e24a
DM
4481}
4482
4ff06203
JA
4483static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4484{
4485 skb->dst_pending_confirm = val;
4486}
4487
4488static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4489{
4490 return skb->dst_pending_confirm != 0;
4491}
4492
2294be0f 4493static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4494{
0b3d8e08 4495#ifdef CONFIG_XFRM
4165079b 4496 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4497#else
def8b4fa 4498 return NULL;
def8b4fa 4499#endif
0b3d8e08 4500}
def8b4fa 4501
68c33163
PS
4502/* Keeps track of mac header offset relative to skb->head.
4503 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4504 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4505 * tunnel skb it points to outer mac header.
4506 * Keeps track of level of encapsulation of network headers.
4507 */
68c33163 4508struct skb_gso_cb {
802ab55a
AD
4509 union {
4510 int mac_offset;
4511 int data_offset;
4512 };
3347c960 4513 int encap_level;
76443456 4514 __wsum csum;
7e2b10c1 4515 __u16 csum_start;
68c33163 4516};
a08e7fd9
CZ
4517#define SKB_GSO_CB_OFFSET 32
4518#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4519
4520static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4521{
4522 return (skb_mac_header(inner_skb) - inner_skb->head) -
4523 SKB_GSO_CB(inner_skb)->mac_offset;
4524}
4525
1e2bd517
PS
4526static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4527{
4528 int new_headroom, headroom;
4529 int ret;
4530
4531 headroom = skb_headroom(skb);
4532 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4533 if (ret)
4534 return ret;
4535
4536 new_headroom = skb_headroom(skb);
4537 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4538 return 0;
4539}
4540
08b64fcc
AD
4541static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4542{
4543 /* Do not update partial checksums if remote checksum is enabled. */
4544 if (skb->remcsum_offload)
4545 return;
4546
4547 SKB_GSO_CB(skb)->csum = res;
4548 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4549}
4550
7e2b10c1
TH
4551/* Compute the checksum for a gso segment. First compute the checksum value
4552 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4553 * then add in skb->csum (checksum from csum_start to end of packet).
4554 * skb->csum and csum_start are then updated to reflect the checksum of the
4555 * resultant packet starting from the transport header-- the resultant checksum
4556 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4557 * header.
4558 */
4559static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4560{
76443456
AD
4561 unsigned char *csum_start = skb_transport_header(skb);
4562 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4563 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4564
76443456
AD
4565 SKB_GSO_CB(skb)->csum = res;
4566 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4567
76443456 4568 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4569}
4570
bdcc0924 4571static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4572{
4573 return skb_shinfo(skb)->gso_size;
4574}
4575
36a8f39e 4576/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4577static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4578{
4579 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4580}
4581
d02f51cb
DA
4582/* Note: Should be called only if skb_is_gso(skb) is true */
4583static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4584{
4585 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4586}
4587
4c3024de 4588/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4589static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4590{
4c3024de 4591 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4592}
4593
5293efe6
DB
4594static inline void skb_gso_reset(struct sk_buff *skb)
4595{
4596 skb_shinfo(skb)->gso_size = 0;
4597 skb_shinfo(skb)->gso_segs = 0;
4598 skb_shinfo(skb)->gso_type = 0;
4599}
4600
d02f51cb
DA
4601static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4602 u16 increment)
4603{
4604 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4605 return;
4606 shinfo->gso_size += increment;
4607}
4608
4609static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4610 u16 decrement)
4611{
4612 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4613 return;
4614 shinfo->gso_size -= decrement;
4615}
4616
7965bd4d 4617void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4618
4619static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4620{
4621 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4622 * wanted then gso_type will be set. */
05bdd2f1
ED
4623 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4624
b78462eb
AD
4625 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4626 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4627 __skb_warn_lro_forwarding(skb);
4628 return true;
4629 }
4630 return false;
4631}
4632
35fc92a9
HX
4633static inline void skb_forward_csum(struct sk_buff *skb)
4634{
4635 /* Unfortunately we don't support this one. Any brave souls? */
4636 if (skb->ip_summed == CHECKSUM_COMPLETE)
4637 skb->ip_summed = CHECKSUM_NONE;
4638}
4639
bc8acf2c
ED
4640/**
4641 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4642 * @skb: skb to check
4643 *
4644 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4645 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4646 * use this helper, to document places where we make this assertion.
4647 */
05bdd2f1 4648static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4649{
4650#ifdef DEBUG
4651 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4652#endif
4653}
4654
f35d9d8a 4655bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4656
ed1f50c3 4657int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4658struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4659 unsigned int transport_len,
4660 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4661
3a7c1ee4
AD
4662/**
4663 * skb_head_is_locked - Determine if the skb->head is locked down
4664 * @skb: skb to check
4665 *
4666 * The head on skbs build around a head frag can be removed if they are
4667 * not cloned. This function returns true if the skb head is locked down
4668 * due to either being allocated via kmalloc, or by being a clone with
4669 * multiple references to the head.
4670 */
4671static inline bool skb_head_is_locked(const struct sk_buff *skb)
4672{
4673 return !skb->head_frag || skb_cloned(skb);
4674}
fe6cc55f 4675
179bc67f
EC
4676/* Local Checksum Offload.
4677 * Compute outer checksum based on the assumption that the
4678 * inner checksum will be offloaded later.
d0dcde64 4679 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4680 * explanation of how this works.
179bc67f
EC
4681 * Fill in outer checksum adjustment (e.g. with sum of outer
4682 * pseudo-header) before calling.
4683 * Also ensure that inner checksum is in linear data area.
4684 */
4685static inline __wsum lco_csum(struct sk_buff *skb)
4686{
9e74a6da
AD
4687 unsigned char *csum_start = skb_checksum_start(skb);
4688 unsigned char *l4_hdr = skb_transport_header(skb);
4689 __wsum partial;
179bc67f
EC
4690
4691 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4692 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4693 skb->csum_offset));
4694
179bc67f 4695 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4696 * adjustment filled in by caller) and return result.
179bc67f 4697 */
9e74a6da 4698 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4699}
4700
2c64605b
PNA
4701static inline bool skb_is_redirected(const struct sk_buff *skb)
4702{
2c64605b 4703 return skb->redirected;
2c64605b
PNA
4704}
4705
4706static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4707{
2c64605b 4708 skb->redirected = 1;
11941f8a 4709#ifdef CONFIG_NET_REDIRECT
2c64605b
PNA
4710 skb->from_ingress = from_ingress;
4711 if (skb->from_ingress)
4712 skb->tstamp = 0;
4713#endif
4714}
4715
4716static inline void skb_reset_redirect(struct sk_buff *skb)
4717{
2c64605b 4718 skb->redirected = 0;
2c64605b
PNA
4719}
4720
fa821170
XL
4721static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4722{
4723 return skb->csum_not_inet;
4724}
4725
6370cc3b
AN
4726static inline void skb_set_kcov_handle(struct sk_buff *skb,
4727 const u64 kcov_handle)
4728{
fa69ee5a
ME
4729#ifdef CONFIG_KCOV
4730 skb->kcov_handle = kcov_handle;
4731#endif
6370cc3b
AN
4732}
4733
4734static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4735{
fa69ee5a
ME
4736#ifdef CONFIG_KCOV
4737 return skb->kcov_handle;
6370cc3b 4738#else
fa69ee5a
ME
4739 return 0;
4740#endif
4741}
6370cc3b 4742
6a5bcd84 4743#ifdef CONFIG_PAGE_POOL
57f05bc2 4744static inline void skb_mark_for_recycle(struct sk_buff *skb)
6a5bcd84
IA
4745{
4746 skb->pp_recycle = 1;
6a5bcd84
IA
4747}
4748#endif
4749
4750static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4751{
4752 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4753 return false;
4754 return page_pool_return_skb_page(virt_to_page(data));
4755}
4756
1da177e4
LT
4757#endif /* __KERNEL__ */
4758#endif /* _LINUX_SKBUFF_H */