sysctl: add and use base directory declarer and registration helper
[linux-block.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
8842d285 17#include <linux/bvec.h>
1da177e4 18#include <linux/cache.h>
56b17425 19#include <linux/rbtree.h>
51f3d02b 20#include <linux/socket.h>
c1d1b437 21#include <linux/refcount.h>
1da177e4 22
60063497 23#include <linux/atomic.h>
1da177e4
LT
24#include <asm/types.h>
25#include <linux/spinlock.h>
1da177e4 26#include <linux/net.h>
3fc7e8a6 27#include <linux/textsearch.h>
1da177e4 28#include <net/checksum.h>
a80958f4 29#include <linux/rcupdate.h>
b7aa0bf7 30#include <linux/hrtimer.h>
131ea667 31#include <linux/dma-mapping.h>
c8f44aff 32#include <linux/netdev_features.h>
363ec392 33#include <linux/sched.h>
e6017571 34#include <linux/sched/clock.h>
1bd758eb 35#include <net/flow_dissector.h>
a60e3cc7 36#include <linux/splice.h>
72b31f72 37#include <linux/in6.h>
8b10cab6 38#include <linux/if_packet.h>
f35f8219 39#include <linux/llist.h>
f70ea018 40#include <net/flow.h>
6a5bcd84 41#include <net/page_pool.h>
261db6c2
JS
42#if IS_ENABLED(CONFIG_NF_CONNTRACK)
43#include <linux/netfilter/nf_conntrack_common.h>
44#endif
1da177e4 45
7a6ae71b
TH
46/* The interface for checksum offload between the stack and networking drivers
47 * is as follows...
48 *
49 * A. IP checksum related features
50 *
51 * Drivers advertise checksum offload capabilities in the features of a device.
db1f00fb
DC
52 * From the stack's point of view these are capabilities offered by the driver.
53 * A driver typically only advertises features that it is capable of offloading
7a6ae71b
TH
54 * to its device.
55 *
56 * The checksum related features are:
57 *
58 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
59 * IP (one's complement) checksum for any combination
60 * of protocols or protocol layering. The checksum is
61 * computed and set in a packet per the CHECKSUM_PARTIAL
62 * interface (see below).
63 *
64 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
65 * TCP or UDP packets over IPv4. These are specifically
66 * unencapsulated packets of the form IPv4|TCP or
67 * IPv4|UDP where the Protocol field in the IPv4 header
db1f00fb 68 * is TCP or UDP. The IPv4 header may contain IP options.
7a6ae71b
TH
69 * This feature cannot be set in features for a device
70 * with NETIF_F_HW_CSUM also set. This feature is being
71 * DEPRECATED (see below).
72 *
73 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
74 * TCP or UDP packets over IPv6. These are specifically
75 * unencapsulated packets of the form IPv6|TCP or
645f0897 76 * IPv6|UDP where the Next Header field in the IPv6
7a6ae71b
TH
77 * header is either TCP or UDP. IPv6 extension headers
78 * are not supported with this feature. This feature
79 * cannot be set in features for a device with
80 * NETIF_F_HW_CSUM also set. This feature is being
81 * DEPRECATED (see below).
82 *
83 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
db1f00fb 84 * This flag is only used to disable the RX checksum
7a6ae71b
TH
85 * feature for a device. The stack will accept receive
86 * checksum indication in packets received on a device
87 * regardless of whether NETIF_F_RXCSUM is set.
88 *
89 * B. Checksumming of received packets by device. Indication of checksum
db1f00fb 90 * verification is set in skb->ip_summed. Possible values are:
78ea85f1
DB
91 *
92 * CHECKSUM_NONE:
93 *
7a6ae71b 94 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
95 * The packet contains full (though not verified) checksum in packet but
96 * not in skb->csum. Thus, skb->csum is undefined in this case.
97 *
98 * CHECKSUM_UNNECESSARY:
99 *
100 * The hardware you're dealing with doesn't calculate the full checksum
101 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
102 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
103 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
104 * though. A driver or device must never modify the checksum field in the
105 * packet even if checksum is verified.
77cffe23
TH
106 *
107 * CHECKSUM_UNNECESSARY is applicable to following protocols:
108 * TCP: IPv6 and IPv4.
109 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
110 * zero UDP checksum for either IPv4 or IPv6, the networking stack
111 * may perform further validation in this case.
112 * GRE: only if the checksum is present in the header.
113 * SCTP: indicates the CRC in SCTP header has been validated.
b4759dcd 114 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
115 *
116 * skb->csum_level indicates the number of consecutive checksums found in
117 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
118 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
119 * and a device is able to verify the checksums for UDP (possibly zero),
db1f00fb 120 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
77cffe23 121 * two. If the device were only able to verify the UDP checksum and not
db1f00fb 122 * GRE, either because it doesn't support GRE checksum or because GRE
77cffe23
TH
123 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
124 * not considered in this case).
78ea85f1
DB
125 *
126 * CHECKSUM_COMPLETE:
127 *
128 * This is the most generic way. The device supplied checksum of the _whole_
db1f00fb 129 * packet as seen by netif_rx() and fills in skb->csum. This means the
78ea85f1
DB
130 * hardware doesn't need to parse L3/L4 headers to implement this.
131 *
b4759dcd
DC
132 * Notes:
133 * - Even if device supports only some protocols, but is able to produce
134 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
135 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
136 *
137 * CHECKSUM_PARTIAL:
138 *
6edec0e6
TH
139 * A checksum is set up to be offloaded to a device as described in the
140 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 141 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
142 * on the same host, or it may be set in the input path in GRO or remote
143 * checksum offload. For the purposes of checksum verification, the checksum
144 * referred to by skb->csum_start + skb->csum_offset and any preceding
145 * checksums in the packet are considered verified. Any checksums in the
146 * packet that are after the checksum being offloaded are not considered to
147 * be verified.
78ea85f1 148 *
7a6ae71b
TH
149 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
150 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
151 *
152 * CHECKSUM_PARTIAL:
153 *
7a6ae71b 154 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 155 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
156 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
157 * csum_start and csum_offset values are valid values given the length and
db1f00fb
DC
158 * offset of the packet, but it should not attempt to validate that the
159 * checksum refers to a legitimate transport layer checksum -- it is the
7a6ae71b
TH
160 * purview of the stack to validate that csum_start and csum_offset are set
161 * correctly.
162 *
163 * When the stack requests checksum offload for a packet, the driver MUST
164 * ensure that the checksum is set correctly. A driver can either offload the
165 * checksum calculation to the device, or call skb_checksum_help (in the case
166 * that the device does not support offload for a particular checksum).
167 *
168 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
169 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
170 * checksum offload capability.
171 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
172 * on network device checksumming capabilities: if a packet does not match
173 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
174 * csum_not_inet, see item D.) is called to resolve the checksum.
78ea85f1 175 *
7a6ae71b 176 * CHECKSUM_NONE:
78ea85f1 177 *
7a6ae71b
TH
178 * The skb was already checksummed by the protocol, or a checksum is not
179 * required.
78ea85f1
DB
180 *
181 * CHECKSUM_UNNECESSARY:
182 *
db1f00fb 183 * This has the same meaning as CHECKSUM_NONE for checksum offload on
7a6ae71b 184 * output.
78ea85f1 185 *
7a6ae71b
TH
186 * CHECKSUM_COMPLETE:
187 * Not used in checksum output. If a driver observes a packet with this value
db1f00fb 188 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
7a6ae71b
TH
189 *
190 * D. Non-IP checksum (CRC) offloads
191 *
192 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
193 * offloading the SCTP CRC in a packet. To perform this offload the stack
db1f00fb 194 * will set csum_start and csum_offset accordingly, set ip_summed to
dba00306
DC
195 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
196 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
197 * A driver that supports both IP checksum offload and SCTP CRC32c offload
198 * must verify which offload is configured for a packet by testing the
199 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
200 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
201 *
202 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
203 * offloading the FCOE CRC in a packet. To perform this offload the stack
204 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
db1f00fb
DC
205 * accordingly. Note that there is no indication in the skbuff that the
206 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
7a6ae71b 207 * both IP checksum offload and FCOE CRC offload must verify which offload
db1f00fb 208 * is configured for a packet, presumably by inspecting packet headers.
7a6ae71b
TH
209 *
210 * E. Checksumming on output with GSO.
211 *
212 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
213 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
214 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
215 * part of the GSO operation is implied. If a checksum is being offloaded
db1f00fb
DC
216 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
217 * csum_offset are set to refer to the outermost checksum being offloaded
218 * (two offloaded checksums are possible with UDP encapsulation).
78ea85f1
DB
219 */
220
60476372 221/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
222#define CHECKSUM_NONE 0
223#define CHECKSUM_UNNECESSARY 1
224#define CHECKSUM_COMPLETE 2
225#define CHECKSUM_PARTIAL 3
1da177e4 226
77cffe23
TH
227/* Maximum value in skb->csum_level */
228#define SKB_MAX_CSUM_LEVEL 3
229
0bec8c88 230#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 231#define SKB_WITH_OVERHEAD(X) \
deea84b0 232 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
233#define SKB_MAX_ORDER(X, ORDER) \
234 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
235#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
236#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
237
87fb4b7b
ED
238/* return minimum truesize of one skb containing X bytes of data */
239#define SKB_TRUESIZE(X) ((X) + \
240 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
241 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
242
7999096f 243struct ahash_request;
1da177e4 244struct net_device;
716ea3a7 245struct scatterlist;
9c55e01c 246struct pipe_inode_info;
a8f820aa 247struct iov_iter;
fd11a83d 248struct napi_struct;
d58e468b
PP
249struct bpf_prog;
250union bpf_attr;
df5042f4 251struct skb_ext;
1da177e4 252
34666d46 253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 254struct nf_bridge_info {
3eaf4025
FW
255 enum {
256 BRNF_PROTO_UNCHANGED,
257 BRNF_PROTO_8021Q,
258 BRNF_PROTO_PPPOE
7fb48c5b 259 } orig_proto:8;
72b1e5e4
FW
260 u8 pkt_otherhost:1;
261 u8 in_prerouting:1;
262 u8 bridged_dnat:1;
411ffb4f 263 __u16 frag_max_size;
bf1ac5ca 264 struct net_device *physindev;
63cdbc06
FW
265
266 /* always valid & non-NULL from FORWARD on, for physdev match */
267 struct net_device *physoutdev;
7fb48c5b 268 union {
72b1e5e4 269 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
270 __be32 ipv4_daddr;
271 struct in6_addr ipv6_daddr;
72b1e5e4
FW
272
273 /* after prerouting + nat detected: store original source
274 * mac since neigh resolution overwrites it, only used while
275 * skb is out in neigh layer.
276 */
277 char neigh_header[8];
72b31f72 278 };
1da177e4
LT
279};
280#endif
281
95a7233c
PB
282#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
283/* Chain in tc_skb_ext will be used to share the tc chain with
284 * ovs recirc_id. It will be set to the current chain by tc
285 * and read by ovs to recirc_id.
286 */
287struct tc_skb_ext {
288 __u32 chain;
038ebb1a 289 __u16 mru;
635d448a 290 __u16 zone;
6f022c2d
PB
291 u8 post_ct:1;
292 u8 post_ct_snat:1;
293 u8 post_ct_dnat:1;
95a7233c
PB
294};
295#endif
296
1da177e4 297struct sk_buff_head {
1a2fb220
KC
298 /* These two members must be first to match sk_buff. */
299 struct_group_tagged(sk_buff_list, list,
300 struct sk_buff *next;
301 struct sk_buff *prev;
302 );
1da177e4
LT
303
304 __u32 qlen;
305 spinlock_t lock;
306};
307
308struct sk_buff;
309
c504e5c2
MD
310/* The reason of skb drop, which is used in kfree_skb_reason().
311 * en...maybe they should be splited by group?
312 *
313 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
314 * used to translate the reason to string.
315 */
316enum skb_drop_reason {
317 SKB_DROP_REASON_NOT_SPECIFIED,
85125597
MD
318 SKB_DROP_REASON_NO_SOCKET,
319 SKB_DROP_REASON_PKT_TOO_SMALL,
320 SKB_DROP_REASON_TCP_CSUM,
321 SKB_DROP_REASON_TCP_FILTER,
1c7fab70 322 SKB_DROP_REASON_UDP_CSUM,
c504e5c2
MD
323 SKB_DROP_REASON_MAX,
324};
325
9d4dde52
IC
326/* To allow 64K frame to be packed as single skb without frag_list we
327 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
328 * buffers which do not start on a page boundary.
329 *
330 * Since GRO uses frags we allocate at least 16 regardless of page
331 * size.
a715dea3 332 */
9d4dde52 333#if (65536/PAGE_SIZE + 1) < 16
eec00954 334#define MAX_SKB_FRAGS 16UL
a715dea3 335#else
9d4dde52 336#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 337#endif
5f74f82e 338extern int sysctl_max_skb_frags;
1da177e4 339
3953c46c
MRL
340/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
341 * segment using its current segmentation instead.
342 */
343#define GSO_BY_FRAGS 0xFFFF
344
8842d285 345typedef struct bio_vec skb_frag_t;
1da177e4 346
161e6137 347/**
7240b60c 348 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
349 * @frag: skb fragment
350 */
9e903e08
ED
351static inline unsigned int skb_frag_size(const skb_frag_t *frag)
352{
b8b576a1 353 return frag->bv_len;
9e903e08
ED
354}
355
161e6137 356/**
7240b60c 357 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
358 * @frag: skb fragment
359 * @size: size of fragment
360 */
9e903e08
ED
361static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
362{
b8b576a1 363 frag->bv_len = size;
9e903e08
ED
364}
365
161e6137 366/**
7240b60c 367 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
368 * @frag: skb fragment
369 * @delta: value to add
370 */
9e903e08
ED
371static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
372{
b8b576a1 373 frag->bv_len += delta;
9e903e08
ED
374}
375
161e6137 376/**
7240b60c 377 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
378 * @frag: skb fragment
379 * @delta: value to subtract
380 */
9e903e08
ED
381static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
382{
b8b576a1 383 frag->bv_len -= delta;
9e903e08
ED
384}
385
161e6137
PT
386/**
387 * skb_frag_must_loop - Test if %p is a high memory page
388 * @p: fragment's page
389 */
c613c209
WB
390static inline bool skb_frag_must_loop(struct page *p)
391{
392#if defined(CONFIG_HIGHMEM)
29766bcf 393 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
394 return true;
395#endif
396 return false;
397}
398
399/**
400 * skb_frag_foreach_page - loop over pages in a fragment
401 *
402 * @f: skb frag to operate on
1dfa5bd3 403 * @f_off: offset from start of f->bv_page
c613c209
WB
404 * @f_len: length from f_off to loop over
405 * @p: (temp var) current page
406 * @p_off: (temp var) offset from start of current page,
407 * non-zero only on first page.
408 * @p_len: (temp var) length in current page,
409 * < PAGE_SIZE only on first and last page.
410 * @copied: (temp var) length so far, excluding current p_len.
411 *
412 * A fragment can hold a compound page, in which case per-page
413 * operations, notably kmap_atomic, must be called for each
414 * regular page.
415 */
416#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
417 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
418 p_off = (f_off) & (PAGE_SIZE - 1), \
419 p_len = skb_frag_must_loop(p) ? \
420 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
421 copied = 0; \
422 copied < f_len; \
423 copied += p_len, p++, p_off = 0, \
424 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
425
ac45f602
PO
426#define HAVE_HW_TIME_STAMP
427
428/**
d3a21be8 429 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
430 * @hwtstamp: hardware time stamp transformed into duration
431 * since arbitrary point in time
ac45f602
PO
432 *
433 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 434 * skb->tstamp.
ac45f602
PO
435 *
436 * hwtstamps can only be compared against other hwtstamps from
437 * the same device.
438 *
439 * This structure is attached to packets as part of the
440 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
441 */
442struct skb_shared_hwtstamps {
443 ktime_t hwtstamp;
ac45f602
PO
444};
445
2244d07b
OH
446/* Definitions for tx_flags in struct skb_shared_info */
447enum {
448 /* generate hardware time stamp */
449 SKBTX_HW_TSTAMP = 1 << 0,
450
e7fd2885 451 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
452 SKBTX_SW_TSTAMP = 1 << 1,
453
454 /* device driver is going to provide hardware time stamp */
455 SKBTX_IN_PROGRESS = 1 << 2,
456
6e3e939f 457 /* generate wifi status information (where possible) */
62b1a8ab 458 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 459
e7fd2885
WB
460 /* generate software time stamp when entering packet scheduling */
461 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
462};
463
e1c8a607 464#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 465 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
466#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
467
06b4feb3
JL
468/* Definitions for flags in struct skb_shared_info */
469enum {
470 /* use zcopy routines */
471 SKBFL_ZEROCOPY_ENABLE = BIT(0),
472
473 /* This indicates at least one fragment might be overwritten
474 * (as in vmsplice(), sendfile() ...)
475 * If we need to compute a TX checksum, we'll need to copy
476 * all frags to avoid possible bad checksum
477 */
478 SKBFL_SHARED_FRAG = BIT(1),
9b65b17d
TA
479
480 /* segment contains only zerocopy data and should not be
481 * charged to the kernel memory.
482 */
483 SKBFL_PURE_ZEROCOPY = BIT(2),
06b4feb3
JL
484};
485
486#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
9b65b17d 487#define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY)
06b4feb3 488
a6686f2f
SM
489/*
490 * The callback notifies userspace to release buffers when skb DMA is done in
491 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
492 * The zerocopy_success argument is true if zero copy transmit occurred,
493 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
494 * The ctx field is used to track device context.
495 * The desc field is used to track userspace buffer index.
a6686f2f
SM
496 */
497struct ubuf_info {
36177832
JL
498 void (*callback)(struct sk_buff *, struct ubuf_info *,
499 bool zerocopy_success);
4ab6c99d
WB
500 union {
501 struct {
502 unsigned long desc;
503 void *ctx;
504 };
505 struct {
506 u32 id;
507 u16 len;
508 u16 zerocopy:1;
509 u32 bytelen;
510 };
511 };
c1d1b437 512 refcount_t refcnt;
04c2d33e 513 u8 flags;
a91dbff5
WB
514
515 struct mmpin {
516 struct user_struct *user;
517 unsigned int num_pg;
518 } mmp;
ac45f602
PO
519};
520
52267790
WB
521#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
522
6f89dbce
SV
523int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
524void mm_unaccount_pinned_pages(struct mmpin *mmp);
525
8c793822
JL
526struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
527struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
528 struct ubuf_info *uarg);
52267790 529
8c793822 530void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790 531
8c793822
JL
532void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
533 bool success);
52267790 534
b5947e5d 535int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
536int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
537 struct msghdr *msg, int len,
538 struct ubuf_info *uarg);
539
1da177e4
LT
540/* This data is invariant across clones and lives at
541 * the end of the header data, ie. at skb->end.
542 */
543struct skb_shared_info {
06b4feb3 544 __u8 flags;
de8f3a83
DB
545 __u8 meta_len;
546 __u8 nr_frags;
9f42f126 547 __u8 tx_flags;
7967168c
HX
548 unsigned short gso_size;
549 /* Warning: this field is not always filled in (UFO)! */
550 unsigned short gso_segs;
1da177e4 551 struct sk_buff *frag_list;
ac45f602 552 struct skb_shared_hwtstamps hwtstamps;
7f564528 553 unsigned int gso_type;
09c2d251 554 u32 tskey;
ec7d2f2c
ED
555
556 /*
557 * Warning : all fields before dataref are cleared in __alloc_skb()
558 */
559 atomic_t dataref;
560
69e3c75f
JB
561 /* Intermediate layers must ensure that destructor_arg
562 * remains valid until skb destructor */
563 void * destructor_arg;
a6686f2f 564
fed66381
ED
565 /* must be last field, see pskb_expand_head() */
566 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
567};
568
569/* We divide dataref into two halves. The higher 16 bits hold references
570 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
571 * the entire skb->data. A clone of a headerless skb holds the length of
572 * the header in skb->hdr_len.
1da177e4
LT
573 *
574 * All users must obey the rule that the skb->data reference count must be
575 * greater than or equal to the payload reference count.
576 *
577 * Holding a reference to the payload part means that the user does not
578 * care about modifications to the header part of skb->data.
579 */
580#define SKB_DATAREF_SHIFT 16
581#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
582
d179cd12
DM
583
584enum {
c8753d55
VS
585 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
586 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
587 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
588};
589
7967168c
HX
590enum {
591 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
592
593 /* This indicates the skb is from an untrusted source. */
d9d30adf 594 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
595
596 /* This indicates the tcp segment has CWR set. */
d9d30adf 597 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 598
d9d30adf 599 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 600
d9d30adf 601 SKB_GSO_TCPV6 = 1 << 4,
68c33163 602
d9d30adf 603 SKB_GSO_FCOE = 1 << 5,
73136267 604
d9d30adf 605 SKB_GSO_GRE = 1 << 6,
0d89d203 606
d9d30adf 607 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 608
d9d30adf 609 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 610
d9d30adf 611 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 612
d9d30adf 613 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 614
d9d30adf 615 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 616
d9d30adf 617 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 618
d9d30adf 619 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 620
d9d30adf 621 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 622
d9d30adf 623 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
624
625 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
626
627 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
628
629 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
630};
631
2e07fa9c
ACM
632#if BITS_PER_LONG > 32
633#define NET_SKBUFF_DATA_USES_OFFSET 1
634#endif
635
636#ifdef NET_SKBUFF_DATA_USES_OFFSET
637typedef unsigned int sk_buff_data_t;
638#else
639typedef unsigned char *sk_buff_data_t;
640#endif
641
161e6137 642/**
1da177e4
LT
643 * struct sk_buff - socket buffer
644 * @next: Next buffer in list
645 * @prev: Previous buffer in list
363ec392 646 * @tstamp: Time we arrived/left
d2f273f0
RD
647 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
648 * for retransmit timer
56b17425 649 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 650 * @list: queue head
df6160de 651 * @ll_node: anchor in an llist (eg socket defer_list)
d84e0bd7 652 * @sk: Socket we are owned by
d2f273f0
RD
653 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
654 * fragmentation management
1da177e4 655 * @dev: Device we arrived on/are leaving by
d2f273f0 656 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 657 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 658 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 659 * @sp: the security path, used for xfrm
1da177e4
LT
660 * @len: Length of actual data
661 * @data_len: Data length
662 * @mac_len: Length of link layer header
334a8132 663 * @hdr_len: writable header length of cloned skb
663ead3b
HX
664 * @csum: Checksum (must include start/offset pair)
665 * @csum_start: Offset from skb->head where checksumming should start
666 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 667 * @priority: Packet queueing priority
60ff7467 668 * @ignore_df: allow local fragmentation
1da177e4 669 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 670 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
671 * @nohdr: Payload reference only, must not modify header
672 * @pkt_type: Packet class
c83c2486 673 * @fclone: skbuff clone status
c83c2486 674 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
675 * @inner_protocol_type: whether the inner protocol is
676 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
677 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
678 * @offload_fwd_mark: Packet was L2-forwarded in hardware
679 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 680 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 681 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
682 * @redirected: packet was redirected by packet classifier
683 * @from_ingress: packet was redirected from the ingress path
42df6e1d 684 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
31729363
RD
685 * @peeked: this packet has been seen already, so stats have been
686 * done for it, don't do them again
ba9dda3a 687 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
688 * @protocol: Packet protocol from driver
689 * @destructor: Destruct function
e2080072 690 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
6ed6e1c7 691 * @_sk_redir: socket redirection information for skmsg
a9e419dc 692 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 693 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 694 * @skb_iif: ifindex of device we arrived on
1da177e4 695 * @tc_index: Traffic control index
61b905da 696 * @hash: the packet hash
d84e0bd7 697 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
698 * @head_frag: skb was allocated from page fragments,
699 * not allocated by kmalloc() or vmalloc().
8b700862 700 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
6a5bcd84
IA
701 * @pp_recycle: mark the packet for recycling instead of freeing (implies
702 * page_pool support on driver)
df5042f4 703 * @active_extensions: active extensions (skb_ext_id types)
553a5672 704 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 705 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 706 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 707 * ports.
a3b18ddb 708 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
709 * @wifi_acked_valid: wifi_acked was set
710 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 711 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
712 * @encapsulation: indicates the inner headers in the skbuff are valid
713 * @encap_hdr_csum: software checksum is needed
714 * @csum_valid: checksum is already valid
dba00306 715 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
716 * @csum_complete_sw: checksum was completed by software
717 * @csum_level: indicates the number of consecutive checksums found in
718 * the packet minus one that have been verified as
719 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 720 * @dst_pending_confirm: need to confirm neighbour
a48d189e 721 * @decrypted: Decrypted SKB
5fc88f93 722 * @slow_gro: state present at GRO time, slower prepare step required
161e6137 723 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 724 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 725 * @secmark: security marking
d84e0bd7 726 * @mark: Generic packet mark
d2f273f0
RD
727 * @reserved_tailroom: (aka @mark) number of bytes of free space available
728 * at the tail of an sk_buff
729 * @vlan_present: VLAN tag is present
86a9bad3 730 * @vlan_proto: vlan encapsulation protocol
6aa895b0 731 * @vlan_tci: vlan tag control information
0d89d203 732 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
733 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
734 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
735 * @inner_transport_header: Inner transport layer header (encapsulation)
736 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 737 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
738 * @transport_header: Transport layer header
739 * @network_header: Network layer header
740 * @mac_header: Link layer header
fa69ee5a 741 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
742 * @tail: Tail pointer
743 * @end: End pointer
744 * @head: Head of buffer
745 * @data: Data head pointer
746 * @truesize: Buffer size
747 * @users: User count - see {datagram,tcp}.c
df5042f4 748 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
749 */
750
751struct sk_buff {
363ec392 752 union {
56b17425 753 struct {
1a2fb220 754 /* These two members must be first to match sk_buff_head. */
56b17425
ED
755 struct sk_buff *next;
756 struct sk_buff *prev;
757
758 union {
bffa72cf
ED
759 struct net_device *dev;
760 /* Some protocols might use this space to store information,
761 * while device pointer would be NULL.
762 * UDP receive path is one user.
763 */
764 unsigned long dev_scratch;
56b17425
ED
765 };
766 };
fa0f5273 767 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 768 struct list_head list;
f35f8219 769 struct llist_node ll_node;
363ec392 770 };
fa0f5273
PO
771
772 union {
773 struct sock *sk;
774 int ip_defrag_offset;
775 };
1da177e4 776
c84d9490 777 union {
bffa72cf 778 ktime_t tstamp;
d3edd06e 779 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 780 };
1da177e4
LT
781 /*
782 * This is the control buffer. It is free to use for every
783 * layer. Please put your private variables there. If you
784 * want to keep them across layers you have to do a skb_clone()
785 * first. This is owned by whoever has the skb queued ATM.
786 */
da3f5cf1 787 char cb[48] __aligned(8);
1da177e4 788
e2080072
ED
789 union {
790 struct {
791 unsigned long _skb_refdst;
792 void (*destructor)(struct sk_buff *skb);
793 };
794 struct list_head tcp_tsorted_anchor;
e3526bb9
CW
795#ifdef CONFIG_NET_SOCK_MSG
796 unsigned long _sk_redir;
797#endif
e2080072
ED
798 };
799
b1937227 800#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 801 unsigned long _nfct;
da3f5cf1 802#endif
1da177e4 803 unsigned int len,
334a8132
PM
804 data_len;
805 __u16 mac_len,
806 hdr_len;
b1937227
ED
807
808 /* Following fields are _not_ copied in __copy_skb_header()
809 * Note that queue_mapping is here mostly to fill a hole.
810 */
b1937227 811 __u16 queue_mapping;
36bbef52
DB
812
813/* if you move cloned around you also must adapt those constants */
814#ifdef __BIG_ENDIAN_BITFIELD
815#define CLONED_MASK (1 << 7)
816#else
817#define CLONED_MASK 1
818#endif
fba84957 819#define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
36bbef52 820
d2f273f0 821 /* private: */
36bbef52 822 __u8 __cloned_offset[0];
d2f273f0 823 /* public: */
b1937227 824 __u8 cloned:1,
6869c4d8 825 nohdr:1,
b84f4cc9 826 fclone:2,
a59322be 827 peeked:1,
b1937227 828 head_frag:1,
6a5bcd84
IA
829 pfmemalloc:1,
830 pp_recycle:1; /* page_pool recycle indicator */
df5042f4
FW
831#ifdef CONFIG_SKB_EXTENSIONS
832 __u8 active_extensions;
833#endif
6a5bcd84 834
03f61041 835 /* Fields enclosed in headers group are copied
b1937227
ED
836 * using a single memcpy() in __copy_skb_header()
837 */
03f61041 838 struct_group(headers,
4031ae6e 839
d2f273f0 840 /* private: */
233577a2 841 __u8 __pkt_type_offset[0];
d2f273f0 842 /* public: */
fba84957 843 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
b1937227 844 __u8 ignore_df:1;
b1937227
ED
845 __u8 nf_trace:1;
846 __u8 ip_summed:2;
3853b584 847 __u8 ooo_okay:1;
8b700862 848
61b905da 849 __u8 l4_hash:1;
a3b18ddb 850 __u8 sw_hash:1;
6e3e939f
JB
851 __u8 wifi_acked_valid:1;
852 __u8 wifi_acked:1;
3bdc0eba 853 __u8 no_fcs:1;
77cffe23 854 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 855 __u8 encapsulation:1;
7e2b10c1 856 __u8 encap_hdr_csum:1;
5d0c2b95 857 __u8 csum_valid:1;
8b700862 858
d2f273f0 859 /* private: */
0c4b2d37 860 __u8 __pkt_vlan_present_offset[0];
d2f273f0 861 /* public: */
fba84957 862 __u8 vlan_present:1; /* See PKT_VLAN_PRESENT_BIT */
7e3cead5 863 __u8 csum_complete_sw:1;
b1937227 864 __u8 csum_level:2;
dba00306 865 __u8 csum_not_inet:1;
4ff06203 866 __u8 dst_pending_confirm:1;
b1937227
ED
867#ifdef CONFIG_IPV6_NDISC_NODETYPE
868 __u8 ndisc_nodetype:2;
869#endif
8b700862 870
0c4b2d37 871 __u8 ipvs_property:1;
8bce6d7d 872 __u8 inner_protocol_type:1;
e585f236 873 __u8 remcsum_offload:1;
6bc506b4
IS
874#ifdef CONFIG_NET_SWITCHDEV
875 __u8 offload_fwd_mark:1;
875e8939 876 __u8 offload_l3_fwd_mark:1;
6bc506b4 877#endif
e7246e12
WB
878#ifdef CONFIG_NET_CLS_ACT
879 __u8 tc_skip_classify:1;
8dc07fdb 880 __u8 tc_at_ingress:1;
2c64605b 881#endif
2c64605b 882 __u8 redirected:1;
11941f8a 883#ifdef CONFIG_NET_REDIRECT
2c64605b 884 __u8 from_ingress:1;
e7246e12 885#endif
42df6e1d
LW
886#ifdef CONFIG_NETFILTER_SKIP_EGRESS
887 __u8 nf_skip_egress:1;
888#endif
a48d189e
SB
889#ifdef CONFIG_TLS_DEVICE
890 __u8 decrypted:1;
891#endif
5fc88f93 892 __u8 slow_gro:1;
b1937227
ED
893
894#ifdef CONFIG_NET_SCHED
895 __u16 tc_index; /* traffic control index */
b1937227 896#endif
fe55f6d5 897
b1937227
ED
898 union {
899 __wsum csum;
900 struct {
901 __u16 csum_start;
902 __u16 csum_offset;
903 };
904 };
905 __u32 priority;
906 int skb_iif;
907 __u32 hash;
908 __be16 vlan_proto;
909 __u16 vlan_tci;
2bd82484
ED
910#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
911 union {
912 unsigned int napi_id;
913 unsigned int sender_cpu;
914 };
97fc2f08 915#endif
984bc16c 916#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 917 __u32 secmark;
0c4f691f 918#endif
0c4f691f 919
3b885787
NH
920 union {
921 __u32 mark;
16fad69c 922 __u32 reserved_tailroom;
3b885787 923 };
1da177e4 924
8bce6d7d
TH
925 union {
926 __be16 inner_protocol;
927 __u8 inner_ipproto;
928 };
929
1a37e412
SH
930 __u16 inner_transport_header;
931 __u16 inner_network_header;
932 __u16 inner_mac_header;
b1937227
ED
933
934 __be16 protocol;
1a37e412
SH
935 __u16 transport_header;
936 __u16 network_header;
937 __u16 mac_header;
b1937227 938
fa69ee5a
ME
939#ifdef CONFIG_KCOV
940 u64 kcov_handle;
941#endif
942
03f61041 943 ); /* end headers group */
b1937227 944
1da177e4 945 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 946 sk_buff_data_t tail;
4305b541 947 sk_buff_data_t end;
1da177e4 948 unsigned char *head,
4305b541 949 *data;
27a884dc 950 unsigned int truesize;
63354797 951 refcount_t users;
df5042f4
FW
952
953#ifdef CONFIG_SKB_EXTENSIONS
954 /* only useable after checking ->active_extensions != 0 */
955 struct skb_ext *extensions;
956#endif
1da177e4
LT
957};
958
fba84957
KC
959/* if you move pkt_type around you also must adapt those constants */
960#ifdef __BIG_ENDIAN_BITFIELD
961#define PKT_TYPE_MAX (7 << 5)
962#else
963#define PKT_TYPE_MAX 7
964#endif
965#define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset)
966
967/* if you move pkt_vlan_present around you also must adapt these constants */
968#ifdef __BIG_ENDIAN_BITFIELD
969#define PKT_VLAN_PRESENT_BIT 7
970#else
971#define PKT_VLAN_PRESENT_BIT 0
972#endif
973#define PKT_VLAN_PRESENT_OFFSET offsetof(struct sk_buff, __pkt_vlan_present_offset)
974
1da177e4
LT
975#ifdef __KERNEL__
976/*
977 * Handling routines are only of interest to the kernel
978 */
1da177e4 979
c93bdd0e
MG
980#define SKB_ALLOC_FCLONE 0x01
981#define SKB_ALLOC_RX 0x02
fd11a83d 982#define SKB_ALLOC_NAPI 0x04
c93bdd0e 983
161e6137
PT
984/**
985 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
986 * @skb: buffer
987 */
c93bdd0e
MG
988static inline bool skb_pfmemalloc(const struct sk_buff *skb)
989{
990 return unlikely(skb->pfmemalloc);
991}
992
7fee226a
ED
993/*
994 * skb might have a dst pointer attached, refcounted or not.
995 * _skb_refdst low order bit is set if refcount was _not_ taken
996 */
997#define SKB_DST_NOREF 1UL
998#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
999
1000/**
1001 * skb_dst - returns skb dst_entry
1002 * @skb: buffer
1003 *
1004 * Returns skb dst_entry, regardless of reference taken or not.
1005 */
adf30907
ED
1006static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1007{
161e6137 1008 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
1009 * rcu_read_lock section
1010 */
1011 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1012 !rcu_read_lock_held() &&
1013 !rcu_read_lock_bh_held());
1014 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
1015}
1016
7fee226a
ED
1017/**
1018 * skb_dst_set - sets skb dst
1019 * @skb: buffer
1020 * @dst: dst entry
1021 *
1022 * Sets skb dst, assuming a reference was taken on dst and should
1023 * be released by skb_dst_drop()
1024 */
adf30907
ED
1025static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1026{
8a886b14 1027 skb->slow_gro |= !!dst;
7fee226a
ED
1028 skb->_skb_refdst = (unsigned long)dst;
1029}
1030
932bc4d7
JA
1031/**
1032 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1033 * @skb: buffer
1034 * @dst: dst entry
1035 *
1036 * Sets skb dst, assuming a reference was not taken on dst.
1037 * If dst entry is cached, we do not take reference and dst_release
1038 * will be avoided by refdst_drop. If dst entry is not cached, we take
1039 * reference, so that last dst_release can destroy the dst immediately.
1040 */
1041static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1042{
dbfc4fb7 1043 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
a432934a 1044 skb->slow_gro |= !!dst;
dbfc4fb7 1045 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1046}
7fee226a
ED
1047
1048/**
25985edc 1049 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1050 * @skb: buffer
1051 */
1052static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1053{
1054 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1055}
1056
161e6137
PT
1057/**
1058 * skb_rtable - Returns the skb &rtable
1059 * @skb: buffer
1060 */
511c3f92
ED
1061static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1062{
adf30907 1063 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1064}
1065
8b10cab6
JHS
1066/* For mangling skb->pkt_type from user space side from applications
1067 * such as nft, tc, etc, we only allow a conservative subset of
1068 * possible pkt_types to be set.
1069*/
1070static inline bool skb_pkt_type_ok(u32 ptype)
1071{
1072 return ptype <= PACKET_OTHERHOST;
1073}
1074
161e6137
PT
1075/**
1076 * skb_napi_id - Returns the skb's NAPI id
1077 * @skb: buffer
1078 */
90b602f8
ML
1079static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1080{
1081#ifdef CONFIG_NET_RX_BUSY_POLL
1082 return skb->napi_id;
1083#else
1084 return 0;
1085#endif
1086}
1087
161e6137
PT
1088/**
1089 * skb_unref - decrement the skb's reference count
1090 * @skb: buffer
1091 *
1092 * Returns true if we can free the skb.
1093 */
3889a803
PA
1094static inline bool skb_unref(struct sk_buff *skb)
1095{
1096 if (unlikely(!skb))
1097 return false;
63354797 1098 if (likely(refcount_read(&skb->users) == 1))
3889a803 1099 smp_rmb();
63354797 1100 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1101 return false;
1102
1103 return true;
1104}
1105
c504e5c2
MD
1106void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1107
1108/**
1109 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1110 * @skb: buffer to free
1111 */
1112static inline void kfree_skb(struct sk_buff *skb)
1113{
1114 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1115}
1116
0a463c78 1117void skb_release_head_state(struct sk_buff *skb);
7965bd4d 1118void kfree_skb_list(struct sk_buff *segs);
6413139d 1119void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1120void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1121
1122#ifdef CONFIG_TRACEPOINTS
7965bd4d 1123void consume_skb(struct sk_buff *skb);
be769db2
HX
1124#else
1125static inline void consume_skb(struct sk_buff *skb)
1126{
1127 return kfree_skb(skb);
1128}
1129#endif
1130
ca2c1418 1131void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1132void __kfree_skb(struct sk_buff *skb);
d7e8883c 1133extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1134
7965bd4d
JP
1135void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1136bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1137 bool *fragstolen, int *delta_truesize);
bad43ca8 1138
7965bd4d
JP
1139struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1140 int node);
2ea2f62c 1141struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1142struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1143struct sk_buff *build_skb_around(struct sk_buff *skb,
1144 void *data, unsigned int frag_size);
161e6137 1145
f450d539
AL
1146struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1147
161e6137
PT
1148/**
1149 * alloc_skb - allocate a network buffer
1150 * @size: size to allocate
1151 * @priority: allocation mask
1152 *
1153 * This function is a convenient wrapper around __alloc_skb().
1154 */
d179cd12 1155static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1156 gfp_t priority)
d179cd12 1157{
564824b0 1158 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1159}
1160
2e4e4410
ED
1161struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1162 unsigned long data_len,
1163 int max_page_order,
1164 int *errcode,
1165 gfp_t gfp_mask);
da29e4b4 1166struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1167
d0bf4a9e
ED
1168/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1169struct sk_buff_fclones {
1170 struct sk_buff skb1;
1171
1172 struct sk_buff skb2;
1173
2638595a 1174 refcount_t fclone_ref;
d0bf4a9e
ED
1175};
1176
1177/**
1178 * skb_fclone_busy - check if fclone is busy
293de7de 1179 * @sk: socket
d0bf4a9e
ED
1180 * @skb: buffer
1181 *
bda13fed 1182 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1183 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1184 * so we also check that this didnt happen.
d0bf4a9e 1185 */
39bb5e62
ED
1186static inline bool skb_fclone_busy(const struct sock *sk,
1187 const struct sk_buff *skb)
d0bf4a9e
ED
1188{
1189 const struct sk_buff_fclones *fclones;
1190
1191 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1192
1193 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1194 refcount_read(&fclones->fclone_ref) > 1 &&
f4dae54e 1195 READ_ONCE(fclones->skb2.sk) == sk;
d0bf4a9e
ED
1196}
1197
161e6137
PT
1198/**
1199 * alloc_skb_fclone - allocate a network buffer from fclone cache
1200 * @size: size to allocate
1201 * @priority: allocation mask
1202 *
1203 * This function is a convenient wrapper around __alloc_skb().
1204 */
d179cd12 1205static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1206 gfp_t priority)
d179cd12 1207{
c93bdd0e 1208 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1209}
1210
7965bd4d 1211struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1212void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1213int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1214struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1215void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1216struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1217struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1218 gfp_t gfp_mask, bool fclone);
1219static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1220 gfp_t gfp_mask)
1221{
1222 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1223}
7965bd4d
JP
1224
1225int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1226struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1227 unsigned int headroom);
f1260ff1 1228struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
7965bd4d
JP
1229struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1230 int newtailroom, gfp_t priority);
48a1df65
JD
1231int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1232 int offset, int len);
1233int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1234 int offset, int len);
7965bd4d 1235int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1236int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1237
1238/**
1239 * skb_pad - zero pad the tail of an skb
1240 * @skb: buffer to pad
1241 * @pad: space to pad
1242 *
1243 * Ensure that a buffer is followed by a padding area that is zero
1244 * filled. Used by network drivers which may DMA or transfer data
1245 * beyond the buffer end onto the wire.
1246 *
1247 * May return error in out of memory cases. The skb is freed on error.
1248 */
1249static inline int skb_pad(struct sk_buff *skb, int pad)
1250{
1251 return __skb_pad(skb, pad, true);
1252}
ead2ceb0 1253#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1254
be12a1fe
HFS
1255int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1256 int offset, size_t size);
1257
d94d9fee 1258struct skb_seq_state {
677e90ed
TG
1259 __u32 lower_offset;
1260 __u32 upper_offset;
1261 __u32 frag_idx;
1262 __u32 stepped_offset;
1263 struct sk_buff *root_skb;
1264 struct sk_buff *cur_skb;
1265 __u8 *frag_data;
97550f6f 1266 __u32 frag_off;
677e90ed
TG
1267};
1268
7965bd4d
JP
1269void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1270 unsigned int to, struct skb_seq_state *st);
1271unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1272 struct skb_seq_state *st);
1273void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1274
7965bd4d 1275unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1276 unsigned int to, struct ts_config *config);
3fc7e8a6 1277
09323cc4
TH
1278/*
1279 * Packet hash types specify the type of hash in skb_set_hash.
1280 *
1281 * Hash types refer to the protocol layer addresses which are used to
1282 * construct a packet's hash. The hashes are used to differentiate or identify
1283 * flows of the protocol layer for the hash type. Hash types are either
1284 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1285 *
1286 * Properties of hashes:
1287 *
1288 * 1) Two packets in different flows have different hash values
1289 * 2) Two packets in the same flow should have the same hash value
1290 *
1291 * A hash at a higher layer is considered to be more specific. A driver should
1292 * set the most specific hash possible.
1293 *
1294 * A driver cannot indicate a more specific hash than the layer at which a hash
1295 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1296 *
1297 * A driver may indicate a hash level which is less specific than the
1298 * actual layer the hash was computed on. For instance, a hash computed
1299 * at L4 may be considered an L3 hash. This should only be done if the
1300 * driver can't unambiguously determine that the HW computed the hash at
1301 * the higher layer. Note that the "should" in the second property above
1302 * permits this.
1303 */
1304enum pkt_hash_types {
1305 PKT_HASH_TYPE_NONE, /* Undefined type */
1306 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1307 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1308 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1309};
1310
bcc83839 1311static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1312{
bcc83839 1313 skb->hash = 0;
a3b18ddb 1314 skb->sw_hash = 0;
bcc83839
TH
1315 skb->l4_hash = 0;
1316}
1317
1318static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1319{
1320 if (!skb->l4_hash)
1321 skb_clear_hash(skb);
1322}
1323
1324static inline void
1325__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1326{
1327 skb->l4_hash = is_l4;
1328 skb->sw_hash = is_sw;
61b905da 1329 skb->hash = hash;
09323cc4
TH
1330}
1331
bcc83839
TH
1332static inline void
1333skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1334{
1335 /* Used by drivers to set hash from HW */
1336 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1337}
1338
1339static inline void
1340__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1341{
1342 __skb_set_hash(skb, hash, true, is_l4);
1343}
1344
e5276937 1345void __skb_get_hash(struct sk_buff *skb);
b917783c 1346u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937 1347u32 skb_get_poff(const struct sk_buff *skb);
f96533cd 1348u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
72a338bc 1349 const struct flow_keys_basic *keys, int hlen);
e5276937 1350__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
f96533cd 1351 const void *data, int hlen_proto);
e5276937
TH
1352
1353static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1354 int thoff, u8 ip_proto)
1355{
1356 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1357}
1358
1359void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1360 const struct flow_dissector_key *key,
1361 unsigned int key_count);
1362
089b19a9
SF
1363struct bpf_flow_dissector;
1364bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1365 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1366
3cbf4ffb
SF
1367bool __skb_flow_dissect(const struct net *net,
1368 const struct sk_buff *skb,
e5276937 1369 struct flow_dissector *flow_dissector,
f96533cd
AL
1370 void *target_container, const void *data,
1371 __be16 proto, int nhoff, int hlen, unsigned int flags);
e5276937
TH
1372
1373static inline bool skb_flow_dissect(const struct sk_buff *skb,
1374 struct flow_dissector *flow_dissector,
cd79a238 1375 void *target_container, unsigned int flags)
e5276937 1376{
3cbf4ffb
SF
1377 return __skb_flow_dissect(NULL, skb, flow_dissector,
1378 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1379}
1380
1381static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1382 struct flow_keys *flow,
1383 unsigned int flags)
e5276937
TH
1384{
1385 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1386 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1387 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1388}
1389
72a338bc 1390static inline bool
3cbf4ffb
SF
1391skb_flow_dissect_flow_keys_basic(const struct net *net,
1392 const struct sk_buff *skb,
f96533cd
AL
1393 struct flow_keys_basic *flow,
1394 const void *data, __be16 proto,
1395 int nhoff, int hlen, unsigned int flags)
e5276937
TH
1396{
1397 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1398 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1399 data, proto, nhoff, hlen, flags);
e5276937
TH
1400}
1401
82828b88
JP
1402void skb_flow_dissect_meta(const struct sk_buff *skb,
1403 struct flow_dissector *flow_dissector,
1404 void *target_container);
1405
75a56758 1406/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1407 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1408 * to user states.
1409 */
1410void
1411skb_flow_dissect_ct(const struct sk_buff *skb,
1412 struct flow_dissector *flow_dissector,
1413 void *target_container,
7baf2429 1414 u16 *ctinfo_map, size_t mapsize,
38495958 1415 bool post_ct, u16 zone);
62b32379
SH
1416void
1417skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1418 struct flow_dissector *flow_dissector,
1419 void *target_container);
1420
0cb09aff
AL
1421void skb_flow_dissect_hash(const struct sk_buff *skb,
1422 struct flow_dissector *flow_dissector,
1423 void *target_container);
1424
3958afa1 1425static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1426{
a3b18ddb 1427 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1428 __skb_get_hash(skb);
bfb564e7 1429
61b905da 1430 return skb->hash;
bfb564e7
KK
1431}
1432
20a17bf6 1433static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1434{
c6cc1ca7
TH
1435 if (!skb->l4_hash && !skb->sw_hash) {
1436 struct flow_keys keys;
de4c1f8b 1437 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1438
de4c1f8b 1439 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1440 }
f70ea018
TH
1441
1442 return skb->hash;
1443}
1444
55667441
ED
1445__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1446 const siphash_key_t *perturb);
50fb7992 1447
57bdf7f4
TH
1448static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1449{
61b905da 1450 return skb->hash;
57bdf7f4
TH
1451}
1452
3df7a74e
TH
1453static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1454{
61b905da 1455 to->hash = from->hash;
a3b18ddb 1456 to->sw_hash = from->sw_hash;
61b905da 1457 to->l4_hash = from->l4_hash;
3df7a74e
TH
1458};
1459
41477662
JK
1460static inline void skb_copy_decrypted(struct sk_buff *to,
1461 const struct sk_buff *from)
1462{
1463#ifdef CONFIG_TLS_DEVICE
1464 to->decrypted = from->decrypted;
1465#endif
1466}
1467
4305b541
ACM
1468#ifdef NET_SKBUFF_DATA_USES_OFFSET
1469static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1470{
1471 return skb->head + skb->end;
1472}
ec47ea82
AD
1473
1474static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1475{
1476 return skb->end;
1477}
4305b541
ACM
1478#else
1479static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1480{
1481 return skb->end;
1482}
ec47ea82
AD
1483
1484static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1485{
1486 return skb->end - skb->head;
1487}
4305b541
ACM
1488#endif
1489
1da177e4 1490/* Internal */
4305b541 1491#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1492
ac45f602
PO
1493static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1494{
1495 return &skb_shinfo(skb)->hwtstamps;
1496}
1497
52267790
WB
1498static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1499{
06b4feb3 1500 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1501
1502 return is_zcopy ? skb_uarg(skb) : NULL;
1503}
1504
9b65b17d
TA
1505static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1506{
1507 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1508}
1509
1510static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1511 const struct sk_buff *skb2)
1512{
1513 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1514}
1515
8e044917 1516static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1517{
1518 refcount_inc(&uarg->refcnt);
1519}
1520
9ee5e5ad
JL
1521static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1522{
1523 skb_shinfo(skb)->destructor_arg = uarg;
1524 skb_shinfo(skb)->flags |= uarg->flags;
1525}
1526
52900d22
WB
1527static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1528 bool *have_ref)
52267790
WB
1529{
1530 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1531 if (unlikely(have_ref && *have_ref))
1532 *have_ref = false;
1533 else
8e044917 1534 net_zcopy_get(uarg);
9ee5e5ad 1535 skb_zcopy_init(skb, uarg);
52267790
WB
1536 }
1537}
1538
5cd8d46e
WB
1539static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1540{
1541 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1542 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1543}
1544
1545static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1546{
1547 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1548}
1549
1550static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1551{
1552 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1553}
1554
8e044917 1555static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1556{
1557 if (uarg)
36177832 1558 uarg->callback(NULL, uarg, true);
59776362
JL
1559}
1560
8e044917 1561static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1562{
1563 if (uarg) {
8c793822
JL
1564 if (uarg->callback == msg_zerocopy_callback)
1565 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1566 else if (have_uref)
8e044917 1567 net_zcopy_put(uarg);
236a6b1c
JL
1568 }
1569}
1570
52267790 1571/* Release a reference on a zerocopy structure */
36177832 1572static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1573{
1574 struct ubuf_info *uarg = skb_zcopy(skb);
1575
1576 if (uarg) {
36177832
JL
1577 if (!skb_zcopy_is_nouarg(skb))
1578 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1579
9b65b17d 1580 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
52267790
WB
1581 }
1582}
1583
a8305bff
DM
1584static inline void skb_mark_not_on_list(struct sk_buff *skb)
1585{
1586 skb->next = NULL;
1587}
1588
dcfea72e 1589/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1590#define skb_list_walk_safe(first, skb, next_skb) \
1591 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1592 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1593
992cba7e
DM
1594static inline void skb_list_del_init(struct sk_buff *skb)
1595{
1596 __list_del_entry(&skb->list);
1597 skb_mark_not_on_list(skb);
1598}
1599
1da177e4
LT
1600/**
1601 * skb_queue_empty - check if a queue is empty
1602 * @list: queue head
1603 *
1604 * Returns true if the queue is empty, false otherwise.
1605 */
1606static inline int skb_queue_empty(const struct sk_buff_head *list)
1607{
fd44b93c 1608 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1609}
1610
d7d16a89
ED
1611/**
1612 * skb_queue_empty_lockless - check if a queue is empty
1613 * @list: queue head
1614 *
1615 * Returns true if the queue is empty, false otherwise.
1616 * This variant can be used in lockless contexts.
1617 */
1618static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1619{
1620 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1621}
1622
1623
fc7ebb21
DM
1624/**
1625 * skb_queue_is_last - check if skb is the last entry in the queue
1626 * @list: queue head
1627 * @skb: buffer
1628 *
1629 * Returns true if @skb is the last buffer on the list.
1630 */
1631static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1632 const struct sk_buff *skb)
1633{
fd44b93c 1634 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1635}
1636
832d11c5
IJ
1637/**
1638 * skb_queue_is_first - check if skb is the first entry in the queue
1639 * @list: queue head
1640 * @skb: buffer
1641 *
1642 * Returns true if @skb is the first buffer on the list.
1643 */
1644static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1645 const struct sk_buff *skb)
1646{
fd44b93c 1647 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1648}
1649
249c8b42
DM
1650/**
1651 * skb_queue_next - return the next packet in the queue
1652 * @list: queue head
1653 * @skb: current buffer
1654 *
1655 * Return the next packet in @list after @skb. It is only valid to
1656 * call this if skb_queue_is_last() evaluates to false.
1657 */
1658static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1659 const struct sk_buff *skb)
1660{
1661 /* This BUG_ON may seem severe, but if we just return then we
1662 * are going to dereference garbage.
1663 */
1664 BUG_ON(skb_queue_is_last(list, skb));
1665 return skb->next;
1666}
1667
832d11c5
IJ
1668/**
1669 * skb_queue_prev - return the prev packet in the queue
1670 * @list: queue head
1671 * @skb: current buffer
1672 *
1673 * Return the prev packet in @list before @skb. It is only valid to
1674 * call this if skb_queue_is_first() evaluates to false.
1675 */
1676static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1677 const struct sk_buff *skb)
1678{
1679 /* This BUG_ON may seem severe, but if we just return then we
1680 * are going to dereference garbage.
1681 */
1682 BUG_ON(skb_queue_is_first(list, skb));
1683 return skb->prev;
1684}
1685
1da177e4
LT
1686/**
1687 * skb_get - reference buffer
1688 * @skb: buffer to reference
1689 *
1690 * Makes another reference to a socket buffer and returns a pointer
1691 * to the buffer.
1692 */
1693static inline struct sk_buff *skb_get(struct sk_buff *skb)
1694{
63354797 1695 refcount_inc(&skb->users);
1da177e4
LT
1696 return skb;
1697}
1698
1699/*
f8821f96 1700 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1701 */
1702
1da177e4
LT
1703/**
1704 * skb_cloned - is the buffer a clone
1705 * @skb: buffer to check
1706 *
1707 * Returns true if the buffer was generated with skb_clone() and is
1708 * one of multiple shared copies of the buffer. Cloned buffers are
1709 * shared data so must not be written to under normal circumstances.
1710 */
1711static inline int skb_cloned(const struct sk_buff *skb)
1712{
1713 return skb->cloned &&
1714 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1715}
1716
14bbd6a5
PS
1717static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1718{
d0164adc 1719 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1720
1721 if (skb_cloned(skb))
1722 return pskb_expand_head(skb, 0, 0, pri);
1723
1724 return 0;
1725}
1726
c4777efa
ED
1727/* This variant of skb_unclone() makes sure skb->truesize is not changed */
1728static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1729{
1730 might_sleep_if(gfpflags_allow_blocking(pri));
1731
1732 if (skb_cloned(skb)) {
1733 unsigned int save = skb->truesize;
1734 int res;
1735
1736 res = pskb_expand_head(skb, 0, 0, pri);
1737 skb->truesize = save;
1738 return res;
1739 }
1740 return 0;
1741}
1742
1da177e4
LT
1743/**
1744 * skb_header_cloned - is the header a clone
1745 * @skb: buffer to check
1746 *
1747 * Returns true if modifying the header part of the buffer requires
1748 * the data to be copied.
1749 */
1750static inline int skb_header_cloned(const struct sk_buff *skb)
1751{
1752 int dataref;
1753
1754 if (!skb->cloned)
1755 return 0;
1756
1757 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1758 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1759 return dataref != 1;
1760}
1761
9580bf2e
ED
1762static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1763{
1764 might_sleep_if(gfpflags_allow_blocking(pri));
1765
1766 if (skb_header_cloned(skb))
1767 return pskb_expand_head(skb, 0, 0, pri);
1768
1769 return 0;
1770}
1771
f4a775d1
ED
1772/**
1773 * __skb_header_release - release reference to header
1774 * @skb: buffer to operate on
f4a775d1
ED
1775 */
1776static inline void __skb_header_release(struct sk_buff *skb)
1777{
1778 skb->nohdr = 1;
1779 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1780}
1781
1782
1da177e4
LT
1783/**
1784 * skb_shared - is the buffer shared
1785 * @skb: buffer to check
1786 *
1787 * Returns true if more than one person has a reference to this
1788 * buffer.
1789 */
1790static inline int skb_shared(const struct sk_buff *skb)
1791{
63354797 1792 return refcount_read(&skb->users) != 1;
1da177e4
LT
1793}
1794
1795/**
1796 * skb_share_check - check if buffer is shared and if so clone it
1797 * @skb: buffer to check
1798 * @pri: priority for memory allocation
1799 *
1800 * If the buffer is shared the buffer is cloned and the old copy
1801 * drops a reference. A new clone with a single reference is returned.
1802 * If the buffer is not shared the original buffer is returned. When
1803 * being called from interrupt status or with spinlocks held pri must
1804 * be GFP_ATOMIC.
1805 *
1806 * NULL is returned on a memory allocation failure.
1807 */
47061bc4 1808static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1809{
d0164adc 1810 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1811 if (skb_shared(skb)) {
1812 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1813
1814 if (likely(nskb))
1815 consume_skb(skb);
1816 else
1817 kfree_skb(skb);
1da177e4
LT
1818 skb = nskb;
1819 }
1820 return skb;
1821}
1822
1823/*
1824 * Copy shared buffers into a new sk_buff. We effectively do COW on
1825 * packets to handle cases where we have a local reader and forward
1826 * and a couple of other messy ones. The normal one is tcpdumping
1827 * a packet thats being forwarded.
1828 */
1829
1830/**
1831 * skb_unshare - make a copy of a shared buffer
1832 * @skb: buffer to check
1833 * @pri: priority for memory allocation
1834 *
1835 * If the socket buffer is a clone then this function creates a new
1836 * copy of the data, drops a reference count on the old copy and returns
1837 * the new copy with the reference count at 1. If the buffer is not a clone
1838 * the original buffer is returned. When called with a spinlock held or
1839 * from interrupt state @pri must be %GFP_ATOMIC
1840 *
1841 * %NULL is returned on a memory allocation failure.
1842 */
e2bf521d 1843static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1844 gfp_t pri)
1da177e4 1845{
d0164adc 1846 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1847 if (skb_cloned(skb)) {
1848 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1849
1850 /* Free our shared copy */
1851 if (likely(nskb))
1852 consume_skb(skb);
1853 else
1854 kfree_skb(skb);
1da177e4
LT
1855 skb = nskb;
1856 }
1857 return skb;
1858}
1859
1860/**
1a5778aa 1861 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1862 * @list_: list to peek at
1863 *
1864 * Peek an &sk_buff. Unlike most other operations you _MUST_
1865 * be careful with this one. A peek leaves the buffer on the
1866 * list and someone else may run off with it. You must hold
1867 * the appropriate locks or have a private queue to do this.
1868 *
1869 * Returns %NULL for an empty list or a pointer to the head element.
1870 * The reference count is not incremented and the reference is therefore
1871 * volatile. Use with caution.
1872 */
05bdd2f1 1873static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1874{
18d07000
ED
1875 struct sk_buff *skb = list_->next;
1876
1877 if (skb == (struct sk_buff *)list_)
1878 skb = NULL;
1879 return skb;
1da177e4
LT
1880}
1881
8b69bd7d
DM
1882/**
1883 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1884 * @list_: list to peek at
1885 *
1886 * Like skb_peek(), but the caller knows that the list is not empty.
1887 */
1888static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1889{
1890 return list_->next;
1891}
1892
da5ef6e5
PE
1893/**
1894 * skb_peek_next - peek skb following the given one from a queue
1895 * @skb: skb to start from
1896 * @list_: list to peek at
1897 *
1898 * Returns %NULL when the end of the list is met or a pointer to the
1899 * next element. The reference count is not incremented and the
1900 * reference is therefore volatile. Use with caution.
1901 */
1902static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1903 const struct sk_buff_head *list_)
1904{
1905 struct sk_buff *next = skb->next;
18d07000 1906
da5ef6e5
PE
1907 if (next == (struct sk_buff *)list_)
1908 next = NULL;
1909 return next;
1910}
1911
1da177e4 1912/**
1a5778aa 1913 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1914 * @list_: list to peek at
1915 *
1916 * Peek an &sk_buff. Unlike most other operations you _MUST_
1917 * be careful with this one. A peek leaves the buffer on the
1918 * list and someone else may run off with it. You must hold
1919 * the appropriate locks or have a private queue to do this.
1920 *
1921 * Returns %NULL for an empty list or a pointer to the tail element.
1922 * The reference count is not incremented and the reference is therefore
1923 * volatile. Use with caution.
1924 */
05bdd2f1 1925static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1926{
f8cc62ca 1927 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1928
1929 if (skb == (struct sk_buff *)list_)
1930 skb = NULL;
1931 return skb;
1932
1da177e4
LT
1933}
1934
1935/**
1936 * skb_queue_len - get queue length
1937 * @list_: list to measure
1938 *
1939 * Return the length of an &sk_buff queue.
1940 */
1941static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1942{
1943 return list_->qlen;
1944}
1945
86b18aaa
QC
1946/**
1947 * skb_queue_len_lockless - get queue length
1948 * @list_: list to measure
1949 *
1950 * Return the length of an &sk_buff queue.
1951 * This variant can be used in lockless contexts.
1952 */
1953static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1954{
1955 return READ_ONCE(list_->qlen);
1956}
1957
67fed459
DM
1958/**
1959 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1960 * @list: queue to initialize
1961 *
1962 * This initializes only the list and queue length aspects of
1963 * an sk_buff_head object. This allows to initialize the list
1964 * aspects of an sk_buff_head without reinitializing things like
1965 * the spinlock. It can also be used for on-stack sk_buff_head
1966 * objects where the spinlock is known to not be used.
1967 */
1968static inline void __skb_queue_head_init(struct sk_buff_head *list)
1969{
1970 list->prev = list->next = (struct sk_buff *)list;
1971 list->qlen = 0;
1972}
1973
76f10ad0
AV
1974/*
1975 * This function creates a split out lock class for each invocation;
1976 * this is needed for now since a whole lot of users of the skb-queue
1977 * infrastructure in drivers have different locking usage (in hardirq)
1978 * than the networking core (in softirq only). In the long run either the
1979 * network layer or drivers should need annotation to consolidate the
1980 * main types of usage into 3 classes.
1981 */
1da177e4
LT
1982static inline void skb_queue_head_init(struct sk_buff_head *list)
1983{
1984 spin_lock_init(&list->lock);
67fed459 1985 __skb_queue_head_init(list);
1da177e4
LT
1986}
1987
c2ecba71
PE
1988static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1989 struct lock_class_key *class)
1990{
1991 skb_queue_head_init(list);
1992 lockdep_set_class(&list->lock, class);
1993}
1994
1da177e4 1995/*
bf299275 1996 * Insert an sk_buff on a list.
1da177e4
LT
1997 *
1998 * The "__skb_xxxx()" functions are the non-atomic ones that
1999 * can only be called with interrupts disabled.
2000 */
bf299275
GR
2001static inline void __skb_insert(struct sk_buff *newsk,
2002 struct sk_buff *prev, struct sk_buff *next,
2003 struct sk_buff_head *list)
2004{
f8cc62ca
ED
2005 /* See skb_queue_empty_lockless() and skb_peek_tail()
2006 * for the opposite READ_ONCE()
2007 */
d7d16a89
ED
2008 WRITE_ONCE(newsk->next, next);
2009 WRITE_ONCE(newsk->prev, prev);
1a2fb220
KC
2010 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2011 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
04f08eb4 2012 WRITE_ONCE(list->qlen, list->qlen + 1);
bf299275 2013}
1da177e4 2014
67fed459
DM
2015static inline void __skb_queue_splice(const struct sk_buff_head *list,
2016 struct sk_buff *prev,
2017 struct sk_buff *next)
2018{
2019 struct sk_buff *first = list->next;
2020 struct sk_buff *last = list->prev;
2021
d7d16a89
ED
2022 WRITE_ONCE(first->prev, prev);
2023 WRITE_ONCE(prev->next, first);
67fed459 2024
d7d16a89
ED
2025 WRITE_ONCE(last->next, next);
2026 WRITE_ONCE(next->prev, last);
67fed459
DM
2027}
2028
2029/**
2030 * skb_queue_splice - join two skb lists, this is designed for stacks
2031 * @list: the new list to add
2032 * @head: the place to add it in the first list
2033 */
2034static inline void skb_queue_splice(const struct sk_buff_head *list,
2035 struct sk_buff_head *head)
2036{
2037 if (!skb_queue_empty(list)) {
2038 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 2039 head->qlen += list->qlen;
67fed459
DM
2040 }
2041}
2042
2043/**
d9619496 2044 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2045 * @list: the new list to add
2046 * @head: the place to add it in the first list
2047 *
2048 * The list at @list is reinitialised
2049 */
2050static inline void skb_queue_splice_init(struct sk_buff_head *list,
2051 struct sk_buff_head *head)
2052{
2053 if (!skb_queue_empty(list)) {
2054 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 2055 head->qlen += list->qlen;
67fed459
DM
2056 __skb_queue_head_init(list);
2057 }
2058}
2059
2060/**
2061 * skb_queue_splice_tail - join two skb lists, each list being a queue
2062 * @list: the new list to add
2063 * @head: the place to add it in the first list
2064 */
2065static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2066 struct sk_buff_head *head)
2067{
2068 if (!skb_queue_empty(list)) {
2069 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2070 head->qlen += list->qlen;
67fed459
DM
2071 }
2072}
2073
2074/**
d9619496 2075 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2076 * @list: the new list to add
2077 * @head: the place to add it in the first list
2078 *
2079 * Each of the lists is a queue.
2080 * The list at @list is reinitialised
2081 */
2082static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2083 struct sk_buff_head *head)
2084{
2085 if (!skb_queue_empty(list)) {
2086 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2087 head->qlen += list->qlen;
67fed459
DM
2088 __skb_queue_head_init(list);
2089 }
2090}
2091
1da177e4 2092/**
300ce174 2093 * __skb_queue_after - queue a buffer at the list head
1da177e4 2094 * @list: list to use
300ce174 2095 * @prev: place after this buffer
1da177e4
LT
2096 * @newsk: buffer to queue
2097 *
300ce174 2098 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2099 * and you must therefore hold required locks before calling it.
2100 *
2101 * A buffer cannot be placed on two lists at the same time.
2102 */
300ce174
SH
2103static inline void __skb_queue_after(struct sk_buff_head *list,
2104 struct sk_buff *prev,
2105 struct sk_buff *newsk)
1da177e4 2106{
1a2fb220 2107 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
1da177e4
LT
2108}
2109
7965bd4d
JP
2110void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2111 struct sk_buff_head *list);
7de6c033 2112
f5572855
GR
2113static inline void __skb_queue_before(struct sk_buff_head *list,
2114 struct sk_buff *next,
2115 struct sk_buff *newsk)
2116{
1a2fb220 2117 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
f5572855
GR
2118}
2119
300ce174
SH
2120/**
2121 * __skb_queue_head - queue a buffer at the list head
2122 * @list: list to use
2123 * @newsk: buffer to queue
2124 *
2125 * Queue a buffer at the start of a list. This function takes no locks
2126 * and you must therefore hold required locks before calling it.
2127 *
2128 * A buffer cannot be placed on two lists at the same time.
2129 */
300ce174
SH
2130static inline void __skb_queue_head(struct sk_buff_head *list,
2131 struct sk_buff *newsk)
2132{
2133 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2134}
4ea7b0cf 2135void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2136
1da177e4
LT
2137/**
2138 * __skb_queue_tail - queue a buffer at the list tail
2139 * @list: list to use
2140 * @newsk: buffer to queue
2141 *
2142 * Queue a buffer at the end of a list. This function takes no locks
2143 * and you must therefore hold required locks before calling it.
2144 *
2145 * A buffer cannot be placed on two lists at the same time.
2146 */
1da177e4
LT
2147static inline void __skb_queue_tail(struct sk_buff_head *list,
2148 struct sk_buff *newsk)
2149{
f5572855 2150 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2151}
4ea7b0cf 2152void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2153
1da177e4
LT
2154/*
2155 * remove sk_buff from list. _Must_ be called atomically, and with
2156 * the list known..
2157 */
7965bd4d 2158void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2159static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2160{
2161 struct sk_buff *next, *prev;
2162
86b18aaa 2163 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2164 next = skb->next;
2165 prev = skb->prev;
2166 skb->next = skb->prev = NULL;
d7d16a89
ED
2167 WRITE_ONCE(next->prev, prev);
2168 WRITE_ONCE(prev->next, next);
1da177e4
LT
2169}
2170
f525c06d
GR
2171/**
2172 * __skb_dequeue - remove from the head of the queue
2173 * @list: list to dequeue from
2174 *
2175 * Remove the head of the list. This function does not take any locks
2176 * so must be used with appropriate locks held only. The head item is
2177 * returned or %NULL if the list is empty.
2178 */
f525c06d
GR
2179static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2180{
2181 struct sk_buff *skb = skb_peek(list);
2182 if (skb)
2183 __skb_unlink(skb, list);
2184 return skb;
2185}
4ea7b0cf 2186struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2187
2188/**
2189 * __skb_dequeue_tail - remove from the tail of the queue
2190 * @list: list to dequeue from
2191 *
2192 * Remove the tail of the list. This function does not take any locks
2193 * so must be used with appropriate locks held only. The tail item is
2194 * returned or %NULL if the list is empty.
2195 */
1da177e4
LT
2196static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2197{
2198 struct sk_buff *skb = skb_peek_tail(list);
2199 if (skb)
2200 __skb_unlink(skb, list);
2201 return skb;
2202}
4ea7b0cf 2203struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2204
2205
bdcc0924 2206static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2207{
2208 return skb->data_len;
2209}
2210
2211static inline unsigned int skb_headlen(const struct sk_buff *skb)
2212{
2213 return skb->len - skb->data_len;
2214}
2215
3ece7826 2216static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2217{
c72d8cda 2218 unsigned int i, len = 0;
1da177e4 2219
c72d8cda 2220 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2221 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2222 return len;
2223}
2224
2225static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2226{
2227 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2228}
2229
131ea667
IC
2230/**
2231 * __skb_fill_page_desc - initialise a paged fragment in an skb
2232 * @skb: buffer containing fragment to be initialised
2233 * @i: paged fragment index to initialise
2234 * @page: the page to use for this fragment
2235 * @off: the offset to the data with @page
2236 * @size: the length of the data
2237 *
2238 * Initialises the @i'th fragment of @skb to point to &size bytes at
2239 * offset @off within @page.
2240 *
2241 * Does not take any additional reference on the fragment.
2242 */
2243static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2244 struct page *page, int off, int size)
1da177e4
LT
2245{
2246 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2247
c48a11c7 2248 /*
2f064f34
MH
2249 * Propagate page pfmemalloc to the skb if we can. The problem is
2250 * that not all callers have unique ownership of the page but rely
2251 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2252 */
1dfa5bd3 2253 frag->bv_page = page;
65c84f14 2254 frag->bv_offset = off;
9e903e08 2255 skb_frag_size_set(frag, size);
cca7af38
PE
2256
2257 page = compound_head(page);
2f064f34 2258 if (page_is_pfmemalloc(page))
cca7af38 2259 skb->pfmemalloc = true;
131ea667
IC
2260}
2261
2262/**
2263 * skb_fill_page_desc - initialise a paged fragment in an skb
2264 * @skb: buffer containing fragment to be initialised
2265 * @i: paged fragment index to initialise
2266 * @page: the page to use for this fragment
2267 * @off: the offset to the data with @page
2268 * @size: the length of the data
2269 *
2270 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2271 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2272 * addition updates @skb such that @i is the last fragment.
2273 *
2274 * Does not take any additional reference on the fragment.
2275 */
2276static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2277 struct page *page, int off, int size)
2278{
2279 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2280 skb_shinfo(skb)->nr_frags = i + 1;
2281}
2282
7965bd4d
JP
2283void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2284 int size, unsigned int truesize);
654bed16 2285
f8e617e1
JW
2286void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2287 unsigned int truesize);
2288
1da177e4
LT
2289#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2290
27a884dc
ACM
2291#ifdef NET_SKBUFF_DATA_USES_OFFSET
2292static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2293{
2294 return skb->head + skb->tail;
2295}
2296
2297static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2298{
2299 skb->tail = skb->data - skb->head;
2300}
2301
2302static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2303{
2304 skb_reset_tail_pointer(skb);
2305 skb->tail += offset;
2306}
7cc46190 2307
27a884dc
ACM
2308#else /* NET_SKBUFF_DATA_USES_OFFSET */
2309static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2310{
2311 return skb->tail;
2312}
2313
2314static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2315{
2316 skb->tail = skb->data;
2317}
2318
2319static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2320{
2321 skb->tail = skb->data + offset;
2322}
4305b541 2323
27a884dc
ACM
2324#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2325
1da177e4
LT
2326/*
2327 * Add data to an sk_buff
2328 */
4df864c1
JB
2329void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2330void *skb_put(struct sk_buff *skb, unsigned int len);
2331static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2332{
4df864c1 2333 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2334 SKB_LINEAR_ASSERT(skb);
2335 skb->tail += len;
2336 skb->len += len;
2337 return tmp;
2338}
2339
de77b966 2340static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2341{
2342 void *tmp = __skb_put(skb, len);
2343
2344 memset(tmp, 0, len);
2345 return tmp;
2346}
2347
2348static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2349 unsigned int len)
2350{
2351 void *tmp = __skb_put(skb, len);
2352
2353 memcpy(tmp, data, len);
2354 return tmp;
2355}
2356
2357static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2358{
2359 *(u8 *)__skb_put(skb, 1) = val;
2360}
2361
83ad357d 2362static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2363{
83ad357d 2364 void *tmp = skb_put(skb, len);
e45a79da
JB
2365
2366 memset(tmp, 0, len);
2367
2368 return tmp;
2369}
2370
59ae1d12
JB
2371static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2372 unsigned int len)
2373{
2374 void *tmp = skb_put(skb, len);
2375
2376 memcpy(tmp, data, len);
2377
2378 return tmp;
2379}
2380
634fef61
JB
2381static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2382{
2383 *(u8 *)skb_put(skb, 1) = val;
2384}
2385
d58ff351
JB
2386void *skb_push(struct sk_buff *skb, unsigned int len);
2387static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2388{
2389 skb->data -= len;
2390 skb->len += len;
2391 return skb->data;
2392}
2393
af72868b
JB
2394void *skb_pull(struct sk_buff *skb, unsigned int len);
2395static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2396{
2397 skb->len -= len;
2398 BUG_ON(skb->len < skb->data_len);
2399 return skb->data += len;
2400}
2401
af72868b 2402static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2403{
2404 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2405}
2406
13244ccc
LAD
2407void *skb_pull_data(struct sk_buff *skb, size_t len);
2408
af72868b 2409void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2410
af72868b 2411static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2412{
2413 if (len > skb_headlen(skb) &&
987c402a 2414 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2415 return NULL;
2416 skb->len -= len;
2417 return skb->data += len;
2418}
2419
af72868b 2420static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2421{
2422 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2423}
2424
b9df4fd7 2425static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2426{
2427 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2428 return true;
1da177e4 2429 if (unlikely(len > skb->len))
b9df4fd7 2430 return false;
987c402a 2431 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2432}
2433
c8c8b127
ED
2434void skb_condense(struct sk_buff *skb);
2435
1da177e4
LT
2436/**
2437 * skb_headroom - bytes at buffer head
2438 * @skb: buffer to check
2439 *
2440 * Return the number of bytes of free space at the head of an &sk_buff.
2441 */
c2636b4d 2442static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2443{
2444 return skb->data - skb->head;
2445}
2446
2447/**
2448 * skb_tailroom - bytes at buffer end
2449 * @skb: buffer to check
2450 *
2451 * Return the number of bytes of free space at the tail of an sk_buff
2452 */
2453static inline int skb_tailroom(const struct sk_buff *skb)
2454{
4305b541 2455 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2456}
2457
a21d4572
ED
2458/**
2459 * skb_availroom - bytes at buffer end
2460 * @skb: buffer to check
2461 *
2462 * Return the number of bytes of free space at the tail of an sk_buff
2463 * allocated by sk_stream_alloc()
2464 */
2465static inline int skb_availroom(const struct sk_buff *skb)
2466{
16fad69c
ED
2467 if (skb_is_nonlinear(skb))
2468 return 0;
2469
2470 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2471}
2472
1da177e4
LT
2473/**
2474 * skb_reserve - adjust headroom
2475 * @skb: buffer to alter
2476 * @len: bytes to move
2477 *
2478 * Increase the headroom of an empty &sk_buff by reducing the tail
2479 * room. This is only allowed for an empty buffer.
2480 */
8243126c 2481static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2482{
2483 skb->data += len;
2484 skb->tail += len;
2485}
2486
1837b2e2
BP
2487/**
2488 * skb_tailroom_reserve - adjust reserved_tailroom
2489 * @skb: buffer to alter
2490 * @mtu: maximum amount of headlen permitted
2491 * @needed_tailroom: minimum amount of reserved_tailroom
2492 *
2493 * Set reserved_tailroom so that headlen can be as large as possible but
2494 * not larger than mtu and tailroom cannot be smaller than
2495 * needed_tailroom.
2496 * The required headroom should already have been reserved before using
2497 * this function.
2498 */
2499static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2500 unsigned int needed_tailroom)
2501{
2502 SKB_LINEAR_ASSERT(skb);
2503 if (mtu < skb_tailroom(skb) - needed_tailroom)
2504 /* use at most mtu */
2505 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2506 else
2507 /* use up to all available space */
2508 skb->reserved_tailroom = needed_tailroom;
2509}
2510
8bce6d7d
TH
2511#define ENCAP_TYPE_ETHER 0
2512#define ENCAP_TYPE_IPPROTO 1
2513
2514static inline void skb_set_inner_protocol(struct sk_buff *skb,
2515 __be16 protocol)
2516{
2517 skb->inner_protocol = protocol;
2518 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2519}
2520
2521static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2522 __u8 ipproto)
2523{
2524 skb->inner_ipproto = ipproto;
2525 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2526}
2527
6a674e9c
JG
2528static inline void skb_reset_inner_headers(struct sk_buff *skb)
2529{
aefbd2b3 2530 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2531 skb->inner_network_header = skb->network_header;
2532 skb->inner_transport_header = skb->transport_header;
2533}
2534
0b5c9db1
JP
2535static inline void skb_reset_mac_len(struct sk_buff *skb)
2536{
2537 skb->mac_len = skb->network_header - skb->mac_header;
2538}
2539
6a674e9c
JG
2540static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2541 *skb)
2542{
2543 return skb->head + skb->inner_transport_header;
2544}
2545
55dc5a9f
TH
2546static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2547{
2548 return skb_inner_transport_header(skb) - skb->data;
2549}
2550
6a674e9c
JG
2551static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2552{
2553 skb->inner_transport_header = skb->data - skb->head;
2554}
2555
2556static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2557 const int offset)
2558{
2559 skb_reset_inner_transport_header(skb);
2560 skb->inner_transport_header += offset;
2561}
2562
2563static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2564{
2565 return skb->head + skb->inner_network_header;
2566}
2567
2568static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2569{
2570 skb->inner_network_header = skb->data - skb->head;
2571}
2572
2573static inline void skb_set_inner_network_header(struct sk_buff *skb,
2574 const int offset)
2575{
2576 skb_reset_inner_network_header(skb);
2577 skb->inner_network_header += offset;
2578}
2579
aefbd2b3
PS
2580static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2581{
2582 return skb->head + skb->inner_mac_header;
2583}
2584
2585static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2586{
2587 skb->inner_mac_header = skb->data - skb->head;
2588}
2589
2590static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2591 const int offset)
2592{
2593 skb_reset_inner_mac_header(skb);
2594 skb->inner_mac_header += offset;
2595}
fda55eca
ED
2596static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2597{
35d04610 2598 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2599}
2600
9c70220b
ACM
2601static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2602{
2e07fa9c 2603 return skb->head + skb->transport_header;
9c70220b
ACM
2604}
2605
badff6d0
ACM
2606static inline void skb_reset_transport_header(struct sk_buff *skb)
2607{
2e07fa9c 2608 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2609}
2610
967b05f6
ACM
2611static inline void skb_set_transport_header(struct sk_buff *skb,
2612 const int offset)
2613{
2e07fa9c
ACM
2614 skb_reset_transport_header(skb);
2615 skb->transport_header += offset;
ea2ae17d
ACM
2616}
2617
d56f90a7
ACM
2618static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2619{
2e07fa9c 2620 return skb->head + skb->network_header;
d56f90a7
ACM
2621}
2622
c1d2bbe1
ACM
2623static inline void skb_reset_network_header(struct sk_buff *skb)
2624{
2e07fa9c 2625 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2626}
2627
c14d2450
ACM
2628static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2629{
2e07fa9c
ACM
2630 skb_reset_network_header(skb);
2631 skb->network_header += offset;
c14d2450
ACM
2632}
2633
2e07fa9c 2634static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2635{
2e07fa9c 2636 return skb->head + skb->mac_header;
bbe735e4
ACM
2637}
2638
ea6da4fd
AV
2639static inline int skb_mac_offset(const struct sk_buff *skb)
2640{
2641 return skb_mac_header(skb) - skb->data;
2642}
2643
0daf4349
DB
2644static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2645{
2646 return skb->network_header - skb->mac_header;
2647}
2648
2e07fa9c 2649static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2650{
35d04610 2651 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2652}
2653
b4ab3141
DB
2654static inline void skb_unset_mac_header(struct sk_buff *skb)
2655{
2656 skb->mac_header = (typeof(skb->mac_header))~0U;
2657}
2658
2e07fa9c
ACM
2659static inline void skb_reset_mac_header(struct sk_buff *skb)
2660{
2661 skb->mac_header = skb->data - skb->head;
2662}
2663
2664static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2665{
2666 skb_reset_mac_header(skb);
2667 skb->mac_header += offset;
2668}
2669
0e3da5bb
TT
2670static inline void skb_pop_mac_header(struct sk_buff *skb)
2671{
2672 skb->mac_header = skb->network_header;
2673}
2674
d2aa125d 2675static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2676{
72a338bc 2677 struct flow_keys_basic keys;
fbbdb8f0
YX
2678
2679 if (skb_transport_header_was_set(skb))
2680 return;
72a338bc 2681
3cbf4ffb
SF
2682 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2683 NULL, 0, 0, 0, 0))
42aecaa9 2684 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2685}
2686
03606895
ED
2687static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2688{
2689 if (skb_mac_header_was_set(skb)) {
2690 const unsigned char *old_mac = skb_mac_header(skb);
2691
2692 skb_set_mac_header(skb, -skb->mac_len);
2693 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2694 }
2695}
2696
04fb451e
MM
2697static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2698{
2699 return skb->csum_start - skb_headroom(skb);
2700}
2701
08b64fcc
AD
2702static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2703{
2704 return skb->head + skb->csum_start;
2705}
2706
2e07fa9c
ACM
2707static inline int skb_transport_offset(const struct sk_buff *skb)
2708{
2709 return skb_transport_header(skb) - skb->data;
2710}
2711
2712static inline u32 skb_network_header_len(const struct sk_buff *skb)
2713{
2714 return skb->transport_header - skb->network_header;
2715}
2716
6a674e9c
JG
2717static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2718{
2719 return skb->inner_transport_header - skb->inner_network_header;
2720}
2721
2e07fa9c
ACM
2722static inline int skb_network_offset(const struct sk_buff *skb)
2723{
2724 return skb_network_header(skb) - skb->data;
2725}
48d49d0c 2726
6a674e9c
JG
2727static inline int skb_inner_network_offset(const struct sk_buff *skb)
2728{
2729 return skb_inner_network_header(skb) - skb->data;
2730}
2731
f9599ce1
CG
2732static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2733{
2734 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2735}
2736
1da177e4
LT
2737/*
2738 * CPUs often take a performance hit when accessing unaligned memory
2739 * locations. The actual performance hit varies, it can be small if the
2740 * hardware handles it or large if we have to take an exception and fix it
2741 * in software.
2742 *
2743 * Since an ethernet header is 14 bytes network drivers often end up with
2744 * the IP header at an unaligned offset. The IP header can be aligned by
2745 * shifting the start of the packet by 2 bytes. Drivers should do this
2746 * with:
2747 *
8660c124 2748 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2749 *
2750 * The downside to this alignment of the IP header is that the DMA is now
2751 * unaligned. On some architectures the cost of an unaligned DMA is high
2752 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2753 *
1da177e4
LT
2754 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2755 * to be overridden.
2756 */
2757#ifndef NET_IP_ALIGN
2758#define NET_IP_ALIGN 2
2759#endif
2760
025be81e
AB
2761/*
2762 * The networking layer reserves some headroom in skb data (via
2763 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2764 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2765 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2766 *
2767 * Unfortunately this headroom changes the DMA alignment of the resulting
2768 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2769 * on some architectures. An architecture can override this value,
2770 * perhaps setting it to a cacheline in size (since that will maintain
2771 * cacheline alignment of the DMA). It must be a power of 2.
2772 *
d6301d3d 2773 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2774 * headroom, you should not reduce this.
5933dd2f
ED
2775 *
2776 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2777 * to reduce average number of cache lines per packet.
645f0897 2778 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2779 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2780 */
2781#ifndef NET_SKB_PAD
5933dd2f 2782#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2783#endif
2784
7965bd4d 2785int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2786
5293efe6 2787static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2788{
5e1abdc3 2789 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2790 return;
27a884dc
ACM
2791 skb->len = len;
2792 skb_set_tail_pointer(skb, len);
1da177e4
LT
2793}
2794
5293efe6
DB
2795static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2796{
2797 __skb_set_length(skb, len);
2798}
2799
7965bd4d 2800void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2801
2802static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2803{
3cc0e873
HX
2804 if (skb->data_len)
2805 return ___pskb_trim(skb, len);
2806 __skb_trim(skb, len);
2807 return 0;
1da177e4
LT
2808}
2809
2810static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2811{
2812 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2813}
2814
e9fa4f7b
HX
2815/**
2816 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2817 * @skb: buffer to alter
2818 * @len: new length
2819 *
2820 * This is identical to pskb_trim except that the caller knows that
2821 * the skb is not cloned so we should never get an error due to out-
2822 * of-memory.
2823 */
2824static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2825{
2826 int err = pskb_trim(skb, len);
2827 BUG_ON(err);
2828}
2829
5293efe6
DB
2830static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2831{
2832 unsigned int diff = len - skb->len;
2833
2834 if (skb_tailroom(skb) < diff) {
2835 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2836 GFP_ATOMIC);
2837 if (ret)
2838 return ret;
2839 }
2840 __skb_set_length(skb, len);
2841 return 0;
2842}
2843
1da177e4
LT
2844/**
2845 * skb_orphan - orphan a buffer
2846 * @skb: buffer to orphan
2847 *
2848 * If a buffer currently has an owner then we call the owner's
2849 * destructor function and make the @skb unowned. The buffer continues
2850 * to exist but is no longer charged to its former owner.
2851 */
2852static inline void skb_orphan(struct sk_buff *skb)
2853{
c34a7612 2854 if (skb->destructor) {
1da177e4 2855 skb->destructor(skb);
c34a7612
ED
2856 skb->destructor = NULL;
2857 skb->sk = NULL;
376c7311
ED
2858 } else {
2859 BUG_ON(skb->sk);
c34a7612 2860 }
1da177e4
LT
2861}
2862
a353e0ce
MT
2863/**
2864 * skb_orphan_frags - orphan the frags contained in a buffer
2865 * @skb: buffer to orphan frags from
2866 * @gfp_mask: allocation mask for replacement pages
2867 *
2868 * For each frag in the SKB which needs a destructor (i.e. has an
2869 * owner) create a copy of that frag and release the original
2870 * page by calling the destructor.
2871 */
2872static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2873{
1f8b977a
WB
2874 if (likely(!skb_zcopy(skb)))
2875 return 0;
185ce5c3 2876 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 2877 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
2878 return 0;
2879 return skb_copy_ubufs(skb, gfp_mask);
2880}
2881
2882/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2883static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2884{
2885 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2886 return 0;
2887 return skb_copy_ubufs(skb, gfp_mask);
2888}
2889
1da177e4
LT
2890/**
2891 * __skb_queue_purge - empty a list
2892 * @list: list to empty
2893 *
2894 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2895 * the list and one reference dropped. This function does not take the
2896 * list lock and the caller must hold the relevant locks to use it.
2897 */
1da177e4
LT
2898static inline void __skb_queue_purge(struct sk_buff_head *list)
2899{
2900 struct sk_buff *skb;
2901 while ((skb = __skb_dequeue(list)) != NULL)
2902 kfree_skb(skb);
2903}
4ea7b0cf 2904void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2905
385114de 2906unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2907
3f6e687d
KH
2908void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2909
2910/**
2911 * netdev_alloc_frag - allocate a page fragment
2912 * @fragsz: fragment size
2913 *
2914 * Allocates a frag from a page for receive buffer.
2915 * Uses GFP_ATOMIC allocations.
2916 */
2917static inline void *netdev_alloc_frag(unsigned int fragsz)
2918{
2919 return __netdev_alloc_frag_align(fragsz, ~0u);
2920}
2921
2922static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2923 unsigned int align)
2924{
2925 WARN_ON_ONCE(!is_power_of_2(align));
2926 return __netdev_alloc_frag_align(fragsz, -align);
2927}
1da177e4 2928
7965bd4d
JP
2929struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2930 gfp_t gfp_mask);
8af27456
CH
2931
2932/**
2933 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2934 * @dev: network device to receive on
2935 * @length: length to allocate
2936 *
2937 * Allocate a new &sk_buff and assign it a usage count of one. The
2938 * buffer has unspecified headroom built in. Users should allocate
2939 * the headroom they think they need without accounting for the
2940 * built in space. The built in space is used for optimisations.
2941 *
2942 * %NULL is returned if there is no free memory. Although this function
2943 * allocates memory it can be called from an interrupt.
2944 */
2945static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2946 unsigned int length)
8af27456
CH
2947{
2948 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2949}
2950
6f532612
ED
2951/* legacy helper around __netdev_alloc_skb() */
2952static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2953 gfp_t gfp_mask)
2954{
2955 return __netdev_alloc_skb(NULL, length, gfp_mask);
2956}
2957
2958/* legacy helper around netdev_alloc_skb() */
2959static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2960{
2961 return netdev_alloc_skb(NULL, length);
2962}
2963
2964
4915a0de
ED
2965static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2966 unsigned int length, gfp_t gfp)
61321bbd 2967{
4915a0de 2968 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2969
2970 if (NET_IP_ALIGN && skb)
2971 skb_reserve(skb, NET_IP_ALIGN);
2972 return skb;
2973}
2974
4915a0de
ED
2975static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2976 unsigned int length)
2977{
2978 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2979}
2980
181edb2b
AD
2981static inline void skb_free_frag(void *addr)
2982{
8c2dd3e4 2983 page_frag_free(addr);
181edb2b
AD
2984}
2985
3f6e687d
KH
2986void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2987
2988static inline void *napi_alloc_frag(unsigned int fragsz)
2989{
2990 return __napi_alloc_frag_align(fragsz, ~0u);
2991}
2992
2993static inline void *napi_alloc_frag_align(unsigned int fragsz,
2994 unsigned int align)
2995{
2996 WARN_ON_ONCE(!is_power_of_2(align));
2997 return __napi_alloc_frag_align(fragsz, -align);
2998}
2999
fd11a83d
AD
3000struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3001 unsigned int length, gfp_t gfp_mask);
3002static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3003 unsigned int length)
3004{
3005 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3006}
795bb1c0
JDB
3007void napi_consume_skb(struct sk_buff *skb, int budget);
3008
9243adfc 3009void napi_skb_free_stolen_head(struct sk_buff *skb);
15fad714 3010void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 3011
71dfda58
AD
3012/**
3013 * __dev_alloc_pages - allocate page for network Rx
3014 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3015 * @order: size of the allocation
3016 *
3017 * Allocate a new page.
3018 *
3019 * %NULL is returned if there is no free memory.
3020*/
3021static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3022 unsigned int order)
3023{
3024 /* This piece of code contains several assumptions.
3025 * 1. This is for device Rx, therefor a cold page is preferred.
3026 * 2. The expectation is the user wants a compound page.
3027 * 3. If requesting a order 0 page it will not be compound
3028 * due to the check to see if order has a value in prep_new_page
3029 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3030 * code in gfp_to_alloc_flags that should be enforcing this.
3031 */
453f85d4 3032 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
3033
3034 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3035}
3036
3037static inline struct page *dev_alloc_pages(unsigned int order)
3038{
95829b3a 3039 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
3040}
3041
3042/**
3043 * __dev_alloc_page - allocate a page for network Rx
3044 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3045 *
3046 * Allocate a new page.
3047 *
3048 * %NULL is returned if there is no free memory.
3049 */
3050static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3051{
3052 return __dev_alloc_pages(gfp_mask, 0);
3053}
3054
3055static inline struct page *dev_alloc_page(void)
3056{
95829b3a 3057 return dev_alloc_pages(0);
71dfda58
AD
3058}
3059
bc38f30f
AL
3060/**
3061 * dev_page_is_reusable - check whether a page can be reused for network Rx
3062 * @page: the page to test
3063 *
3064 * A page shouldn't be considered for reusing/recycling if it was allocated
3065 * under memory pressure or at a distant memory node.
3066 *
3067 * Returns false if this page should be returned to page allocator, true
3068 * otherwise.
3069 */
3070static inline bool dev_page_is_reusable(const struct page *page)
3071{
3072 return likely(page_to_nid(page) == numa_mem_id() &&
3073 !page_is_pfmemalloc(page));
3074}
3075
0614002b
MG
3076/**
3077 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3078 * @page: The page that was allocated from skb_alloc_page
3079 * @skb: The skb that may need pfmemalloc set
3080 */
48f971c9
AL
3081static inline void skb_propagate_pfmemalloc(const struct page *page,
3082 struct sk_buff *skb)
0614002b 3083{
2f064f34 3084 if (page_is_pfmemalloc(page))
0614002b
MG
3085 skb->pfmemalloc = true;
3086}
3087
7240b60c
JL
3088/**
3089 * skb_frag_off() - Returns the offset of a skb fragment
3090 * @frag: the paged fragment
3091 */
3092static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3093{
65c84f14 3094 return frag->bv_offset;
7240b60c
JL
3095}
3096
3097/**
3098 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3099 * @frag: skb fragment
3100 * @delta: value to add
3101 */
3102static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3103{
65c84f14 3104 frag->bv_offset += delta;
7240b60c
JL
3105}
3106
3107/**
3108 * skb_frag_off_set() - Sets the offset of a skb fragment
3109 * @frag: skb fragment
3110 * @offset: offset of fragment
3111 */
3112static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3113{
65c84f14 3114 frag->bv_offset = offset;
7240b60c
JL
3115}
3116
3117/**
3118 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3119 * @fragto: skb fragment where offset is set
3120 * @fragfrom: skb fragment offset is copied from
3121 */
3122static inline void skb_frag_off_copy(skb_frag_t *fragto,
3123 const skb_frag_t *fragfrom)
3124{
65c84f14 3125 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
3126}
3127
131ea667 3128/**
e227867f 3129 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
3130 * @frag: the paged fragment
3131 *
3132 * Returns the &struct page associated with @frag.
3133 */
3134static inline struct page *skb_frag_page(const skb_frag_t *frag)
3135{
1dfa5bd3 3136 return frag->bv_page;
131ea667
IC
3137}
3138
3139/**
3140 * __skb_frag_ref - take an addition reference on a paged fragment.
3141 * @frag: the paged fragment
3142 *
3143 * Takes an additional reference on the paged fragment @frag.
3144 */
3145static inline void __skb_frag_ref(skb_frag_t *frag)
3146{
3147 get_page(skb_frag_page(frag));
3148}
3149
3150/**
3151 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3152 * @skb: the buffer
3153 * @f: the fragment offset.
3154 *
3155 * Takes an additional reference on the @f'th paged fragment of @skb.
3156 */
3157static inline void skb_frag_ref(struct sk_buff *skb, int f)
3158{
3159 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3160}
3161
3162/**
3163 * __skb_frag_unref - release a reference on a paged fragment.
3164 * @frag: the paged fragment
c420c989 3165 * @recycle: recycle the page if allocated via page_pool
131ea667 3166 *
c420c989
MC
3167 * Releases a reference on the paged fragment @frag
3168 * or recycles the page via the page_pool API.
131ea667 3169 */
c420c989 3170static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
131ea667 3171{
6a5bcd84
IA
3172 struct page *page = skb_frag_page(frag);
3173
3174#ifdef CONFIG_PAGE_POOL
3175 if (recycle && page_pool_return_skb_page(page))
3176 return;
3177#endif
3178 put_page(page);
131ea667
IC
3179}
3180
3181/**
3182 * skb_frag_unref - release a reference on a paged fragment of an skb.
3183 * @skb: the buffer
3184 * @f: the fragment offset
3185 *
3186 * Releases a reference on the @f'th paged fragment of @skb.
3187 */
3188static inline void skb_frag_unref(struct sk_buff *skb, int f)
3189{
6a5bcd84 3190 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
131ea667
IC
3191}
3192
3193/**
3194 * skb_frag_address - gets the address of the data contained in a paged fragment
3195 * @frag: the paged fragment buffer
3196 *
3197 * Returns the address of the data within @frag. The page must already
3198 * be mapped.
3199 */
3200static inline void *skb_frag_address(const skb_frag_t *frag)
3201{
7240b60c 3202 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3203}
3204
3205/**
3206 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3207 * @frag: the paged fragment buffer
3208 *
3209 * Returns the address of the data within @frag. Checks that the page
3210 * is mapped and returns %NULL otherwise.
3211 */
3212static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3213{
3214 void *ptr = page_address(skb_frag_page(frag));
3215 if (unlikely(!ptr))
3216 return NULL;
3217
7240b60c
JL
3218 return ptr + skb_frag_off(frag);
3219}
3220
3221/**
3222 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3223 * @fragto: skb fragment where page is set
3224 * @fragfrom: skb fragment page is copied from
3225 */
3226static inline void skb_frag_page_copy(skb_frag_t *fragto,
3227 const skb_frag_t *fragfrom)
3228{
3229 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3230}
3231
3232/**
3233 * __skb_frag_set_page - sets the page contained in a paged fragment
3234 * @frag: the paged fragment
3235 * @page: the page to set
3236 *
3237 * Sets the fragment @frag to contain @page.
3238 */
3239static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3240{
1dfa5bd3 3241 frag->bv_page = page;
131ea667
IC
3242}
3243
3244/**
3245 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3246 * @skb: the buffer
3247 * @f: the fragment offset
3248 * @page: the page to set
3249 *
3250 * Sets the @f'th fragment of @skb to contain @page.
3251 */
3252static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3253 struct page *page)
3254{
3255 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3256}
3257
400dfd3a
ED
3258bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3259
131ea667
IC
3260/**
3261 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3262 * @dev: the device to map the fragment to
131ea667
IC
3263 * @frag: the paged fragment to map
3264 * @offset: the offset within the fragment (starting at the
3265 * fragment's own offset)
3266 * @size: the number of bytes to map
771b00a8 3267 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3268 *
3269 * Maps the page associated with @frag to @device.
3270 */
3271static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3272 const skb_frag_t *frag,
3273 size_t offset, size_t size,
3274 enum dma_data_direction dir)
3275{
3276 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3277 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3278}
3279
117632e6
ED
3280static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3281 gfp_t gfp_mask)
3282{
3283 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3284}
3285
bad93e9d
OP
3286
3287static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3288 gfp_t gfp_mask)
3289{
3290 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3291}
3292
3293
334a8132
PM
3294/**
3295 * skb_clone_writable - is the header of a clone writable
3296 * @skb: buffer to check
3297 * @len: length up to which to write
3298 *
3299 * Returns true if modifying the header part of the cloned buffer
3300 * does not requires the data to be copied.
3301 */
05bdd2f1 3302static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3303{
3304 return !skb_header_cloned(skb) &&
3305 skb_headroom(skb) + len <= skb->hdr_len;
3306}
3307
3697649f
DB
3308static inline int skb_try_make_writable(struct sk_buff *skb,
3309 unsigned int write_len)
3310{
3311 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3312 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3313}
3314
d9cc2048
HX
3315static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3316 int cloned)
3317{
3318 int delta = 0;
3319
d9cc2048
HX
3320 if (headroom > skb_headroom(skb))
3321 delta = headroom - skb_headroom(skb);
3322
3323 if (delta || cloned)
3324 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3325 GFP_ATOMIC);
3326 return 0;
3327}
3328
1da177e4
LT
3329/**
3330 * skb_cow - copy header of skb when it is required
3331 * @skb: buffer to cow
3332 * @headroom: needed headroom
3333 *
3334 * If the skb passed lacks sufficient headroom or its data part
3335 * is shared, data is reallocated. If reallocation fails, an error
3336 * is returned and original skb is not changed.
3337 *
3338 * The result is skb with writable area skb->head...skb->tail
3339 * and at least @headroom of space at head.
3340 */
3341static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3342{
d9cc2048
HX
3343 return __skb_cow(skb, headroom, skb_cloned(skb));
3344}
1da177e4 3345
d9cc2048
HX
3346/**
3347 * skb_cow_head - skb_cow but only making the head writable
3348 * @skb: buffer to cow
3349 * @headroom: needed headroom
3350 *
3351 * This function is identical to skb_cow except that we replace the
3352 * skb_cloned check by skb_header_cloned. It should be used when
3353 * you only need to push on some header and do not need to modify
3354 * the data.
3355 */
3356static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3357{
3358 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3359}
3360
3361/**
3362 * skb_padto - pad an skbuff up to a minimal size
3363 * @skb: buffer to pad
3364 * @len: minimal length
3365 *
3366 * Pads up a buffer to ensure the trailing bytes exist and are
3367 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3368 * is untouched. Otherwise it is extended. Returns zero on
3369 * success. The skb is freed on error.
1da177e4 3370 */
5b057c6b 3371static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3372{
3373 unsigned int size = skb->len;
3374 if (likely(size >= len))
5b057c6b 3375 return 0;
987c402a 3376 return skb_pad(skb, len - size);
1da177e4
LT
3377}
3378
9c0c1124 3379/**
4ea7b0cf 3380 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3381 * @skb: buffer to pad
3382 * @len: minimal length
cd0a137a 3383 * @free_on_error: free buffer on error
9c0c1124
AD
3384 *
3385 * Pads up a buffer to ensure the trailing bytes exist and are
3386 * blanked. If the buffer already contains sufficient data it
3387 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3388 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3389 */
4a009cb0
ED
3390static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3391 unsigned int len,
3392 bool free_on_error)
9c0c1124
AD
3393{
3394 unsigned int size = skb->len;
3395
3396 if (unlikely(size < len)) {
3397 len -= size;
cd0a137a 3398 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3399 return -ENOMEM;
3400 __skb_put(skb, len);
3401 }
3402 return 0;
3403}
3404
cd0a137a
FF
3405/**
3406 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3407 * @skb: buffer to pad
3408 * @len: minimal length
3409 *
3410 * Pads up a buffer to ensure the trailing bytes exist and are
3411 * blanked. If the buffer already contains sufficient data it
3412 * is untouched. Otherwise it is extended. Returns zero on
3413 * success. The skb is freed on error.
3414 */
4a009cb0 3415static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3416{
3417 return __skb_put_padto(skb, len, true);
3418}
3419
1da177e4 3420static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3421 struct iov_iter *from, int copy)
1da177e4
LT
3422{
3423 const int off = skb->len;
3424
3425 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3426 __wsum csum = 0;
15e6cb46
AV
3427 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3428 &csum, from)) {
1da177e4
LT
3429 skb->csum = csum_block_add(skb->csum, csum, off);
3430 return 0;
3431 }
15e6cb46 3432 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3433 return 0;
3434
3435 __skb_trim(skb, off);
3436 return -EFAULT;
3437}
3438
38ba0a65
ED
3439static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3440 const struct page *page, int off)
1da177e4 3441{
1f8b977a
WB
3442 if (skb_zcopy(skb))
3443 return false;
1da177e4 3444 if (i) {
d8e18a51 3445 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3446
ea2ab693 3447 return page == skb_frag_page(frag) &&
7240b60c 3448 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3449 }
38ba0a65 3450 return false;
1da177e4
LT
3451}
3452
364c6bad
HX
3453static inline int __skb_linearize(struct sk_buff *skb)
3454{
3455 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3456}
3457
1da177e4
LT
3458/**
3459 * skb_linearize - convert paged skb to linear one
3460 * @skb: buffer to linarize
1da177e4
LT
3461 *
3462 * If there is no free memory -ENOMEM is returned, otherwise zero
3463 * is returned and the old skb data released.
3464 */
364c6bad
HX
3465static inline int skb_linearize(struct sk_buff *skb)
3466{
3467 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3468}
3469
cef401de
ED
3470/**
3471 * skb_has_shared_frag - can any frag be overwritten
3472 * @skb: buffer to test
3473 *
3474 * Return true if the skb has at least one frag that might be modified
3475 * by an external entity (as in vmsplice()/sendfile())
3476 */
3477static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3478{
c9af6db4 3479 return skb_is_nonlinear(skb) &&
06b4feb3 3480 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3481}
3482
364c6bad
HX
3483/**
3484 * skb_linearize_cow - make sure skb is linear and writable
3485 * @skb: buffer to process
3486 *
3487 * If there is no free memory -ENOMEM is returned, otherwise zero
3488 * is returned and the old skb data released.
3489 */
3490static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3491{
364c6bad
HX
3492 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3493 __skb_linearize(skb) : 0;
1da177e4
LT
3494}
3495
479ffccc
DB
3496static __always_inline void
3497__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3498 unsigned int off)
3499{
3500 if (skb->ip_summed == CHECKSUM_COMPLETE)
3501 skb->csum = csum_block_sub(skb->csum,
3502 csum_partial(start, len, 0), off);
3503 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3504 skb_checksum_start_offset(skb) < 0)
3505 skb->ip_summed = CHECKSUM_NONE;
3506}
3507
1da177e4
LT
3508/**
3509 * skb_postpull_rcsum - update checksum for received skb after pull
3510 * @skb: buffer to update
3511 * @start: start of data before pull
3512 * @len: length of data pulled
3513 *
3514 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3515 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3516 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3517 */
1da177e4 3518static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3519 const void *start, unsigned int len)
1da177e4 3520{
29c30026 3521 if (skb->ip_summed == CHECKSUM_COMPLETE)
45cac675
ED
3522 skb->csum = wsum_negate(csum_partial(start, len,
3523 wsum_negate(skb->csum)));
29c30026
ED
3524 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3525 skb_checksum_start_offset(skb) < 0)
3526 skb->ip_summed = CHECKSUM_NONE;
1da177e4
LT
3527}
3528
479ffccc
DB
3529static __always_inline void
3530__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3531 unsigned int off)
3532{
3533 if (skb->ip_summed == CHECKSUM_COMPLETE)
3534 skb->csum = csum_block_add(skb->csum,
3535 csum_partial(start, len, 0), off);
3536}
cbb042f9 3537
479ffccc
DB
3538/**
3539 * skb_postpush_rcsum - update checksum for received skb after push
3540 * @skb: buffer to update
3541 * @start: start of data after push
3542 * @len: length of data pushed
3543 *
3544 * After doing a push on a received packet, you need to call this to
3545 * update the CHECKSUM_COMPLETE checksum.
3546 */
f8ffad69
DB
3547static inline void skb_postpush_rcsum(struct sk_buff *skb,
3548 const void *start, unsigned int len)
3549{
479ffccc 3550 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3551}
3552
af72868b 3553void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3554
82a31b92
WC
3555/**
3556 * skb_push_rcsum - push skb and update receive checksum
3557 * @skb: buffer to update
3558 * @len: length of data pulled
3559 *
3560 * This function performs an skb_push on the packet and updates
3561 * the CHECKSUM_COMPLETE checksum. It should be used on
3562 * receive path processing instead of skb_push unless you know
3563 * that the checksum difference is zero (e.g., a valid IP header)
3564 * or you are setting ip_summed to CHECKSUM_NONE.
3565 */
d58ff351 3566static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3567{
3568 skb_push(skb, len);
3569 skb_postpush_rcsum(skb, skb->data, len);
3570 return skb->data;
3571}
3572
88078d98 3573int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3574/**
3575 * pskb_trim_rcsum - trim received skb and update checksum
3576 * @skb: buffer to trim
3577 * @len: new length
3578 *
3579 * This is exactly the same as pskb_trim except that it ensures the
3580 * checksum of received packets are still valid after the operation.
6c57f045 3581 * It can change skb pointers.
7ce5a27f
DM
3582 */
3583
3584static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3585{
3586 if (likely(len >= skb->len))
3587 return 0;
88078d98 3588 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3589}
3590
5293efe6
DB
3591static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3592{
3593 if (skb->ip_summed == CHECKSUM_COMPLETE)
3594 skb->ip_summed = CHECKSUM_NONE;
3595 __skb_trim(skb, len);
3596 return 0;
3597}
3598
3599static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3600{
3601 if (skb->ip_summed == CHECKSUM_COMPLETE)
3602 skb->ip_summed = CHECKSUM_NONE;
3603 return __skb_grow(skb, len);
3604}
3605
18a4c0ea
ED
3606#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3607#define skb_rb_first(root) rb_to_skb(rb_first(root))
3608#define skb_rb_last(root) rb_to_skb(rb_last(root))
3609#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3610#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3611
1da177e4
LT
3612#define skb_queue_walk(queue, skb) \
3613 for (skb = (queue)->next; \
a1e4891f 3614 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3615 skb = skb->next)
3616
46f8914e
JC
3617#define skb_queue_walk_safe(queue, skb, tmp) \
3618 for (skb = (queue)->next, tmp = skb->next; \
3619 skb != (struct sk_buff *)(queue); \
3620 skb = tmp, tmp = skb->next)
3621
1164f52a 3622#define skb_queue_walk_from(queue, skb) \
a1e4891f 3623 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3624 skb = skb->next)
3625
18a4c0ea
ED
3626#define skb_rbtree_walk(skb, root) \
3627 for (skb = skb_rb_first(root); skb != NULL; \
3628 skb = skb_rb_next(skb))
3629
3630#define skb_rbtree_walk_from(skb) \
3631 for (; skb != NULL; \
3632 skb = skb_rb_next(skb))
3633
3634#define skb_rbtree_walk_from_safe(skb, tmp) \
3635 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3636 skb = tmp)
3637
1164f52a
DM
3638#define skb_queue_walk_from_safe(queue, skb, tmp) \
3639 for (tmp = skb->next; \
3640 skb != (struct sk_buff *)(queue); \
3641 skb = tmp, tmp = skb->next)
3642
300ce174
SH
3643#define skb_queue_reverse_walk(queue, skb) \
3644 for (skb = (queue)->prev; \
a1e4891f 3645 skb != (struct sk_buff *)(queue); \
300ce174
SH
3646 skb = skb->prev)
3647
686a2955
DM
3648#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3649 for (skb = (queue)->prev, tmp = skb->prev; \
3650 skb != (struct sk_buff *)(queue); \
3651 skb = tmp, tmp = skb->prev)
3652
3653#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3654 for (tmp = skb->prev; \
3655 skb != (struct sk_buff *)(queue); \
3656 skb = tmp, tmp = skb->prev)
1da177e4 3657
21dc3301 3658static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3659{
3660 return skb_shinfo(skb)->frag_list != NULL;
3661}
3662
3663static inline void skb_frag_list_init(struct sk_buff *skb)
3664{
3665 skb_shinfo(skb)->frag_list = NULL;
3666}
3667
ee039871
DM
3668#define skb_walk_frags(skb, iter) \
3669 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3670
ea3793ee 3671
b50b0580
SD
3672int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3673 int *err, long *timeo_p,
ea3793ee 3674 const struct sk_buff *skb);
65101aec
PA
3675struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3676 struct sk_buff_head *queue,
3677 unsigned int flags,
fd69c399 3678 int *off, int *err,
65101aec 3679 struct sk_buff **last);
b50b0580
SD
3680struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3681 struct sk_buff_head *queue,
e427cad6 3682 unsigned int flags, int *off, int *err,
ea3793ee 3683 struct sk_buff **last);
b50b0580
SD
3684struct sk_buff *__skb_recv_datagram(struct sock *sk,
3685 struct sk_buff_head *sk_queue,
e427cad6 3686 unsigned int flags, int *off, int *err);
7965bd4d
JP
3687struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3688 int *err);
a11e1d43
LT
3689__poll_t datagram_poll(struct file *file, struct socket *sock,
3690 struct poll_table_struct *wait);
c0371da6
AV
3691int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3692 struct iov_iter *to, int size);
51f3d02b
DM
3693static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3694 struct msghdr *msg, int size)
3695{
e5a4b0bb 3696 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3697}
e5a4b0bb
AV
3698int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3699 struct msghdr *msg);
65d69e25
SG
3700int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3701 struct iov_iter *to, int len,
3702 struct ahash_request *hash);
3a654f97
AV
3703int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3704 struct iov_iter *from, int len);
3a654f97 3705int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3706void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3707void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3708static inline void skb_free_datagram_locked(struct sock *sk,
3709 struct sk_buff *skb)
3710{
3711 __skb_free_datagram_locked(sk, skb, 0);
3712}
7965bd4d 3713int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3714int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3715int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3716__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 3717 int len);
a60e3cc7 3718int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3719 struct pipe_inode_info *pipe, unsigned int len,
25869262 3720 unsigned int flags);
20bf50de
TH
3721int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3722 int len);
0739cd28 3723int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
7965bd4d 3724void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3725unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3726int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3727 int len, int hlen);
7965bd4d
JP
3728void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3729int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3730void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3731bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3732bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3733struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3734struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3735 unsigned int offset);
0d5501c1 3736struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3737int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3738int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3739int skb_vlan_pop(struct sk_buff *skb);
3740int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
3741int skb_eth_pop(struct sk_buff *skb);
3742int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3743 const unsigned char *src);
fa4e0f88 3744int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3745 int mac_len, bool ethernet);
040b5cfb
MV
3746int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3747 bool ethernet);
d27cf5c5 3748int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3749int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3750struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3751 gfp_t gfp);
20380731 3752
6ce8e9ce
AV
3753static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3754{
3073f070 3755 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3756}
3757
7eab8d9e
AV
3758static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3759{
e5a4b0bb 3760 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3761}
3762
2817a336
DB
3763struct skb_checksum_ops {
3764 __wsum (*update)(const void *mem, int len, __wsum wsum);
3765 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3766};
3767
9617813d
DC
3768extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3769
2817a336
DB
3770__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3771 __wsum csum, const struct skb_checksum_ops *ops);
3772__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3773 __wsum csum);
3774
1e98a0f0 3775static inline void * __must_check
e3305138
AL
3776__skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3777 const void *data, int hlen, void *buffer)
1da177e4 3778{
d206121f 3779 if (likely(hlen - offset >= len))
e3305138 3780 return (void *)data + offset;
1da177e4 3781
d206121f 3782 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
1da177e4
LT
3783 return NULL;
3784
3785 return buffer;
3786}
3787
1e98a0f0
ED
3788static inline void * __must_check
3789skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3790{
3791 return __skb_header_pointer(skb, offset, len, skb->data,
3792 skb_headlen(skb), buffer);
3793}
3794
4262e5cc
DB
3795/**
3796 * skb_needs_linearize - check if we need to linearize a given skb
3797 * depending on the given device features.
3798 * @skb: socket buffer to check
3799 * @features: net device features
3800 *
3801 * Returns true if either:
3802 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3803 * 2. skb is fragmented and the device does not support SG.
3804 */
3805static inline bool skb_needs_linearize(struct sk_buff *skb,
3806 netdev_features_t features)
3807{
3808 return skb_is_nonlinear(skb) &&
3809 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3810 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3811}
3812
d626f62b
ACM
3813static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3814 void *to,
3815 const unsigned int len)
3816{
3817 memcpy(to, skb->data, len);
3818}
3819
3820static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3821 const int offset, void *to,
3822 const unsigned int len)
3823{
3824 memcpy(to, skb->data + offset, len);
3825}
3826
27d7ff46
ACM
3827static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3828 const void *from,
3829 const unsigned int len)
3830{
3831 memcpy(skb->data, from, len);
3832}
3833
3834static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3835 const int offset,
3836 const void *from,
3837 const unsigned int len)
3838{
3839 memcpy(skb->data + offset, from, len);
3840}
3841
7965bd4d 3842void skb_init(void);
1da177e4 3843
ac45f602
PO
3844static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3845{
3846 return skb->tstamp;
3847}
3848
a61bbcf2
PM
3849/**
3850 * skb_get_timestamp - get timestamp from a skb
3851 * @skb: skb to get stamp from
13c6ee2a 3852 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3853 *
3854 * Timestamps are stored in the skb as offsets to a base timestamp.
3855 * This function converts the offset back to a struct timeval and stores
3856 * it in stamp.
3857 */
ac45f602 3858static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3859 struct __kernel_old_timeval *stamp)
a61bbcf2 3860{
13c6ee2a 3861 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3862}
3863
887feae3
DD
3864static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3865 struct __kernel_sock_timeval *stamp)
3866{
3867 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3868
3869 stamp->tv_sec = ts.tv_sec;
3870 stamp->tv_usec = ts.tv_nsec / 1000;
3871}
3872
ac45f602 3873static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3874 struct __kernel_old_timespec *stamp)
ac45f602 3875{
df1b4ba9
AB
3876 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3877
3878 stamp->tv_sec = ts.tv_sec;
3879 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3880}
3881
887feae3
DD
3882static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3883 struct __kernel_timespec *stamp)
3884{
3885 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3886
3887 stamp->tv_sec = ts.tv_sec;
3888 stamp->tv_nsec = ts.tv_nsec;
3889}
3890
b7aa0bf7 3891static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3892{
b7aa0bf7 3893 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3894}
3895
164891aa
SH
3896static inline ktime_t net_timedelta(ktime_t t)
3897{
3898 return ktime_sub(ktime_get_real(), t);
3899}
3900
b9ce204f
IJ
3901static inline ktime_t net_invalid_timestamp(void)
3902{
8b0e1953 3903 return 0;
b9ce204f 3904}
a61bbcf2 3905
de8f3a83
DB
3906static inline u8 skb_metadata_len(const struct sk_buff *skb)
3907{
3908 return skb_shinfo(skb)->meta_len;
3909}
3910
3911static inline void *skb_metadata_end(const struct sk_buff *skb)
3912{
3913 return skb_mac_header(skb);
3914}
3915
3916static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3917 const struct sk_buff *skb_b,
3918 u8 meta_len)
3919{
3920 const void *a = skb_metadata_end(skb_a);
3921 const void *b = skb_metadata_end(skb_b);
3922 /* Using more efficient varaiant than plain call to memcmp(). */
3923#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3924 u64 diffs = 0;
3925
3926 switch (meta_len) {
3927#define __it(x, op) (x -= sizeof(u##op))
3928#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3929 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3930 fallthrough;
de8f3a83 3931 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3932 fallthrough;
de8f3a83 3933 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3934 fallthrough;
de8f3a83
DB
3935 case 8: diffs |= __it_diff(a, b, 64);
3936 break;
3937 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3938 fallthrough;
de8f3a83 3939 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3940 fallthrough;
de8f3a83 3941 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3942 fallthrough;
de8f3a83
DB
3943 case 4: diffs |= __it_diff(a, b, 32);
3944 break;
3945 }
3946 return diffs;
3947#else
3948 return memcmp(a - meta_len, b - meta_len, meta_len);
3949#endif
3950}
3951
3952static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3953 const struct sk_buff *skb_b)
3954{
3955 u8 len_a = skb_metadata_len(skb_a);
3956 u8 len_b = skb_metadata_len(skb_b);
3957
3958 if (!(len_a | len_b))
3959 return false;
3960
3961 return len_a != len_b ?
3962 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3963}
3964
3965static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3966{
3967 skb_shinfo(skb)->meta_len = meta_len;
3968}
3969
3970static inline void skb_metadata_clear(struct sk_buff *skb)
3971{
3972 skb_metadata_set(skb, 0);
3973}
3974
62bccb8c
AD
3975struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3976
c1f19b51
RC
3977#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3978
7965bd4d
JP
3979void skb_clone_tx_timestamp(struct sk_buff *skb);
3980bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3981
3982#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3983
3984static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3985{
3986}
3987
3988static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3989{
3990 return false;
3991}
3992
3993#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3994
3995/**
3996 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3997 *
da92b194
RC
3998 * PHY drivers may accept clones of transmitted packets for
3999 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
4000 * must call this function to return the skb back to the stack with a
4001 * timestamp.
da92b194 4002 *
2ff17117 4003 * @skb: clone of the original outgoing packet
7a76a021 4004 * @hwtstamps: hardware time stamps
c1f19b51
RC
4005 *
4006 */
4007void skb_complete_tx_timestamp(struct sk_buff *skb,
4008 struct skb_shared_hwtstamps *hwtstamps);
4009
e7ed11ee 4010void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
4011 struct skb_shared_hwtstamps *hwtstamps,
4012 struct sock *sk, int tstype);
4013
ac45f602
PO
4014/**
4015 * skb_tstamp_tx - queue clone of skb with send time stamps
4016 * @orig_skb: the original outgoing packet
4017 * @hwtstamps: hardware time stamps, may be NULL if not available
4018 *
4019 * If the skb has a socket associated, then this function clones the
4020 * skb (thus sharing the actual data and optional structures), stores
4021 * the optional hardware time stamping information (if non NULL) or
4022 * generates a software time stamp (otherwise), then queues the clone
4023 * to the error queue of the socket. Errors are silently ignored.
4024 */
7965bd4d
JP
4025void skb_tstamp_tx(struct sk_buff *orig_skb,
4026 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 4027
4507a715
RC
4028/**
4029 * skb_tx_timestamp() - Driver hook for transmit timestamping
4030 *
4031 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 4032 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 4033 *
73409f3b
DM
4034 * Specifically, one should make absolutely sure that this function is
4035 * called before TX completion of this packet can trigger. Otherwise
4036 * the packet could potentially already be freed.
4037 *
4507a715
RC
4038 * @skb: A socket buffer.
4039 */
4040static inline void skb_tx_timestamp(struct sk_buff *skb)
4041{
c1f19b51 4042 skb_clone_tx_timestamp(skb);
b50a5c70
ML
4043 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4044 skb_tstamp_tx(skb, NULL);
4507a715
RC
4045}
4046
6e3e939f
JB
4047/**
4048 * skb_complete_wifi_ack - deliver skb with wifi status
4049 *
4050 * @skb: the original outgoing packet
4051 * @acked: ack status
4052 *
4053 */
4054void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4055
7965bd4d
JP
4056__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4057__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 4058
60476372
HX
4059static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4060{
6edec0e6
TH
4061 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4062 skb->csum_valid ||
4063 (skb->ip_summed == CHECKSUM_PARTIAL &&
4064 skb_checksum_start_offset(skb) >= 0));
60476372
HX
4065}
4066
fb286bb2
HX
4067/**
4068 * skb_checksum_complete - Calculate checksum of an entire packet
4069 * @skb: packet to process
4070 *
4071 * This function calculates the checksum over the entire packet plus
4072 * the value of skb->csum. The latter can be used to supply the
4073 * checksum of a pseudo header as used by TCP/UDP. It returns the
4074 * checksum.
4075 *
4076 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4077 * this function can be used to verify that checksum on received
4078 * packets. In that case the function should return zero if the
4079 * checksum is correct. In particular, this function will return zero
4080 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4081 * hardware has already verified the correctness of the checksum.
4082 */
4381ca3c 4083static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 4084{
60476372
HX
4085 return skb_csum_unnecessary(skb) ?
4086 0 : __skb_checksum_complete(skb);
fb286bb2
HX
4087}
4088
77cffe23
TH
4089static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4090{
4091 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4092 if (skb->csum_level == 0)
4093 skb->ip_summed = CHECKSUM_NONE;
4094 else
4095 skb->csum_level--;
4096 }
4097}
4098
4099static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4100{
4101 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4102 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4103 skb->csum_level++;
4104 } else if (skb->ip_summed == CHECKSUM_NONE) {
4105 skb->ip_summed = CHECKSUM_UNNECESSARY;
4106 skb->csum_level = 0;
4107 }
4108}
4109
836e66c2
DB
4110static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4111{
4112 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4113 skb->ip_summed = CHECKSUM_NONE;
4114 skb->csum_level = 0;
4115 }
4116}
4117
76ba0aae
TH
4118/* Check if we need to perform checksum complete validation.
4119 *
4120 * Returns true if checksum complete is needed, false otherwise
4121 * (either checksum is unnecessary or zero checksum is allowed).
4122 */
4123static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4124 bool zero_okay,
4125 __sum16 check)
4126{
5d0c2b95
TH
4127 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4128 skb->csum_valid = 1;
77cffe23 4129 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
4130 return false;
4131 }
4132
4133 return true;
4134}
4135
da279887 4136/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
4137 * in checksum_init.
4138 */
4139#define CHECKSUM_BREAK 76
4140
4e18b9ad
TH
4141/* Unset checksum-complete
4142 *
4143 * Unset checksum complete can be done when packet is being modified
4144 * (uncompressed for instance) and checksum-complete value is
4145 * invalidated.
4146 */
4147static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4148{
4149 if (skb->ip_summed == CHECKSUM_COMPLETE)
4150 skb->ip_summed = CHECKSUM_NONE;
4151}
4152
76ba0aae
TH
4153/* Validate (init) checksum based on checksum complete.
4154 *
4155 * Return values:
4156 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4157 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4158 * checksum is stored in skb->csum for use in __skb_checksum_complete
4159 * non-zero: value of invalid checksum
4160 *
4161 */
4162static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4163 bool complete,
4164 __wsum psum)
4165{
4166 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4167 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4168 skb->csum_valid = 1;
76ba0aae
TH
4169 return 0;
4170 }
4171 }
4172
4173 skb->csum = psum;
4174
5d0c2b95
TH
4175 if (complete || skb->len <= CHECKSUM_BREAK) {
4176 __sum16 csum;
4177
4178 csum = __skb_checksum_complete(skb);
4179 skb->csum_valid = !csum;
4180 return csum;
4181 }
76ba0aae
TH
4182
4183 return 0;
4184}
4185
4186static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4187{
4188 return 0;
4189}
4190
4191/* Perform checksum validate (init). Note that this is a macro since we only
4192 * want to calculate the pseudo header which is an input function if necessary.
4193 * First we try to validate without any computation (checksum unnecessary) and
4194 * then calculate based on checksum complete calling the function to compute
4195 * pseudo header.
4196 *
4197 * Return values:
4198 * 0: checksum is validated or try to in skb_checksum_complete
4199 * non-zero: value of invalid checksum
4200 */
4201#define __skb_checksum_validate(skb, proto, complete, \
4202 zero_okay, check, compute_pseudo) \
4203({ \
4204 __sum16 __ret = 0; \
5d0c2b95 4205 skb->csum_valid = 0; \
76ba0aae
TH
4206 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4207 __ret = __skb_checksum_validate_complete(skb, \
4208 complete, compute_pseudo(skb, proto)); \
4209 __ret; \
4210})
4211
4212#define skb_checksum_init(skb, proto, compute_pseudo) \
4213 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4214
4215#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4216 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4217
4218#define skb_checksum_validate(skb, proto, compute_pseudo) \
4219 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4220
4221#define skb_checksum_validate_zero_check(skb, proto, check, \
4222 compute_pseudo) \
096a4cfa 4223 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4224
4225#define skb_checksum_simple_validate(skb) \
4226 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4227
d96535a1
TH
4228static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4229{
219f1d79 4230 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4231}
4232
e4aa33ad 4233static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4234{
4235 skb->csum = ~pseudo;
4236 skb->ip_summed = CHECKSUM_COMPLETE;
4237}
4238
e4aa33ad 4239#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4240do { \
4241 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4242 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4243} while (0)
4244
15e2396d
TH
4245static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4246 u16 start, u16 offset)
4247{
4248 skb->ip_summed = CHECKSUM_PARTIAL;
4249 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4250 skb->csum_offset = offset - start;
4251}
4252
dcdc8994
TH
4253/* Update skbuf and packet to reflect the remote checksum offload operation.
4254 * When called, ptr indicates the starting point for skb->csum when
4255 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4256 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4257 */
4258static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4259 int start, int offset, bool nopartial)
dcdc8994
TH
4260{
4261 __wsum delta;
4262
15e2396d
TH
4263 if (!nopartial) {
4264 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4265 return;
4266 }
4267
10a2308f 4268 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
dcdc8994
TH
4269 __skb_checksum_complete(skb);
4270 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4271 }
4272
4273 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4274
4275 /* Adjust skb->csum since we changed the packet */
4276 skb->csum = csum_add(skb->csum, delta);
4277}
4278
cb9c6836
FW
4279static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4280{
4281#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4282 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4283#else
4284 return NULL;
4285#endif
4286}
4287
261db6c2 4288static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4289{
261db6c2
JS
4290#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4291 return skb->_nfct;
4292#else
4293 return 0UL;
4294#endif
1da177e4 4295}
261db6c2
JS
4296
4297static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4298{
261db6c2 4299#if IS_ENABLED(CONFIG_NF_CONNTRACK)
5fc88f93 4300 skb->slow_gro |= !!nfct;
261db6c2 4301 skb->_nfct = nfct;
2fc72c7b 4302#endif
261db6c2 4303}
df5042f4
FW
4304
4305#ifdef CONFIG_SKB_EXTENSIONS
4306enum skb_ext_id {
4307#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4308 SKB_EXT_BRIDGE_NF,
4165079b
FW
4309#endif
4310#ifdef CONFIG_XFRM
4311 SKB_EXT_SEC_PATH,
95a7233c
PB
4312#endif
4313#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4314 TC_SKB_EXT,
3ee17bc7
MM
4315#endif
4316#if IS_ENABLED(CONFIG_MPTCP)
4317 SKB_EXT_MPTCP,
78476d31
JK
4318#endif
4319#if IS_ENABLED(CONFIG_MCTP_FLOWS)
4320 SKB_EXT_MCTP,
df5042f4
FW
4321#endif
4322 SKB_EXT_NUM, /* must be last */
4323};
4324
4325/**
4326 * struct skb_ext - sk_buff extensions
4327 * @refcnt: 1 on allocation, deallocated on 0
4328 * @offset: offset to add to @data to obtain extension address
4329 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4330 * @data: start of extension data, variable sized
4331 *
4332 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4333 * to use 'u8' types while allowing up to 2kb worth of extension data.
4334 */
4335struct skb_ext {
4336 refcount_t refcnt;
4337 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4338 u8 chunks; /* same */
5c91aa1d 4339 char data[] __aligned(8);
df5042f4
FW
4340};
4341
4930f483 4342struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4343void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4344 struct skb_ext *ext);
df5042f4
FW
4345void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4346void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4347void __skb_ext_put(struct skb_ext *ext);
4348
4349static inline void skb_ext_put(struct sk_buff *skb)
4350{
4351 if (skb->active_extensions)
4352 __skb_ext_put(skb->extensions);
4353}
4354
df5042f4
FW
4355static inline void __skb_ext_copy(struct sk_buff *dst,
4356 const struct sk_buff *src)
4357{
4358 dst->active_extensions = src->active_extensions;
4359
4360 if (src->active_extensions) {
4361 struct skb_ext *ext = src->extensions;
4362
4363 refcount_inc(&ext->refcnt);
4364 dst->extensions = ext;
4365 }
4366}
4367
4368static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4369{
4370 skb_ext_put(dst);
4371 __skb_ext_copy(dst, src);
4372}
4373
4374static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4375{
4376 return !!ext->offset[i];
4377}
4378
4379static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4380{
4381 return skb->active_extensions & (1 << id);
4382}
4383
4384static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4385{
4386 if (skb_ext_exist(skb, id))
4387 __skb_ext_del(skb, id);
4388}
4389
4390static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4391{
4392 if (skb_ext_exist(skb, id)) {
4393 struct skb_ext *ext = skb->extensions;
4394
4395 return (void *)ext + (ext->offset[id] << 3);
4396 }
4397
4398 return NULL;
4399}
174e2381
FW
4400
4401static inline void skb_ext_reset(struct sk_buff *skb)
4402{
4403 if (unlikely(skb->active_extensions)) {
4404 __skb_ext_put(skb->extensions);
4405 skb->active_extensions = 0;
4406 }
4407}
677bf08c
FW
4408
4409static inline bool skb_has_extensions(struct sk_buff *skb)
4410{
4411 return unlikely(skb->active_extensions);
4412}
df5042f4
FW
4413#else
4414static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4415static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4416static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4417static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4418static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4419static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4420#endif /* CONFIG_SKB_EXTENSIONS */
4421
895b5c9f 4422static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4423{
5f79e0f9 4424#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4425 nf_conntrack_put(skb_nfct(skb));
4426 skb->_nfct = 0;
2fc72c7b 4427#endif
a193a4ab
PM
4428}
4429
124dff01
PM
4430static inline void nf_reset_trace(struct sk_buff *skb)
4431{
478b360a 4432#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4433 skb->nf_trace = 0;
4434#endif
a193a4ab
PM
4435}
4436
2b5ec1a5
YY
4437static inline void ipvs_reset(struct sk_buff *skb)
4438{
4439#if IS_ENABLED(CONFIG_IP_VS)
4440 skb->ipvs_property = 0;
4441#endif
4442}
4443
de8bda1d 4444/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4445static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4446 bool copy)
edda553c 4447{
5f79e0f9 4448#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4449 dst->_nfct = src->_nfct;
4450 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4451#endif
478b360a 4452#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4453 if (copy)
4454 dst->nf_trace = src->nf_trace;
478b360a 4455#endif
edda553c
YK
4456}
4457
e7ac05f3
YK
4458static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4459{
e7ac05f3 4460#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4461 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4462#endif
5fc88f93 4463 dst->slow_gro = src->slow_gro;
b1937227 4464 __nf_copy(dst, src, true);
e7ac05f3
YK
4465}
4466
984bc16c
JM
4467#ifdef CONFIG_NETWORK_SECMARK
4468static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4469{
4470 to->secmark = from->secmark;
4471}
4472
4473static inline void skb_init_secmark(struct sk_buff *skb)
4474{
4475 skb->secmark = 0;
4476}
4477#else
4478static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4479{ }
4480
4481static inline void skb_init_secmark(struct sk_buff *skb)
4482{ }
4483#endif
4484
7af8f4ca
FW
4485static inline int secpath_exists(const struct sk_buff *skb)
4486{
4487#ifdef CONFIG_XFRM
4165079b 4488 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4489#else
4490 return 0;
4491#endif
4492}
4493
574f7194
EB
4494static inline bool skb_irq_freeable(const struct sk_buff *skb)
4495{
4496 return !skb->destructor &&
7af8f4ca 4497 !secpath_exists(skb) &&
cb9c6836 4498 !skb_nfct(skb) &&
574f7194
EB
4499 !skb->_skb_refdst &&
4500 !skb_has_frag_list(skb);
4501}
4502
f25f4e44
PWJ
4503static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4504{
f25f4e44 4505 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4506}
4507
9247744e 4508static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4509{
4e3ab47a 4510 return skb->queue_mapping;
4e3ab47a
PE
4511}
4512
f25f4e44
PWJ
4513static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4514{
f25f4e44 4515 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4516}
4517
d5a9e24a
DM
4518static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4519{
4520 skb->queue_mapping = rx_queue + 1;
4521}
4522
9247744e 4523static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4524{
4525 return skb->queue_mapping - 1;
4526}
4527
9247744e 4528static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4529{
a02cec21 4530 return skb->queue_mapping != 0;
d5a9e24a
DM
4531}
4532
4ff06203
JA
4533static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4534{
4535 skb->dst_pending_confirm = val;
4536}
4537
4538static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4539{
4540 return skb->dst_pending_confirm != 0;
4541}
4542
2294be0f 4543static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4544{
0b3d8e08 4545#ifdef CONFIG_XFRM
4165079b 4546 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4547#else
def8b4fa 4548 return NULL;
def8b4fa 4549#endif
0b3d8e08 4550}
def8b4fa 4551
68c33163
PS
4552/* Keeps track of mac header offset relative to skb->head.
4553 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4554 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4555 * tunnel skb it points to outer mac header.
4556 * Keeps track of level of encapsulation of network headers.
4557 */
68c33163 4558struct skb_gso_cb {
802ab55a
AD
4559 union {
4560 int mac_offset;
4561 int data_offset;
4562 };
3347c960 4563 int encap_level;
76443456 4564 __wsum csum;
7e2b10c1 4565 __u16 csum_start;
68c33163 4566};
a08e7fd9
CZ
4567#define SKB_GSO_CB_OFFSET 32
4568#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4569
4570static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4571{
4572 return (skb_mac_header(inner_skb) - inner_skb->head) -
4573 SKB_GSO_CB(inner_skb)->mac_offset;
4574}
4575
1e2bd517
PS
4576static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4577{
4578 int new_headroom, headroom;
4579 int ret;
4580
4581 headroom = skb_headroom(skb);
4582 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4583 if (ret)
4584 return ret;
4585
4586 new_headroom = skb_headroom(skb);
4587 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4588 return 0;
4589}
4590
08b64fcc
AD
4591static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4592{
4593 /* Do not update partial checksums if remote checksum is enabled. */
4594 if (skb->remcsum_offload)
4595 return;
4596
4597 SKB_GSO_CB(skb)->csum = res;
4598 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4599}
4600
7e2b10c1
TH
4601/* Compute the checksum for a gso segment. First compute the checksum value
4602 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4603 * then add in skb->csum (checksum from csum_start to end of packet).
4604 * skb->csum and csum_start are then updated to reflect the checksum of the
4605 * resultant packet starting from the transport header-- the resultant checksum
4606 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4607 * header.
4608 */
4609static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4610{
76443456
AD
4611 unsigned char *csum_start = skb_transport_header(skb);
4612 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4613 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4614
76443456
AD
4615 SKB_GSO_CB(skb)->csum = res;
4616 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4617
76443456 4618 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4619}
4620
bdcc0924 4621static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4622{
4623 return skb_shinfo(skb)->gso_size;
4624}
4625
36a8f39e 4626/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4627static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4628{
4629 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4630}
4631
d02f51cb
DA
4632/* Note: Should be called only if skb_is_gso(skb) is true */
4633static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4634{
4635 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4636}
4637
4c3024de 4638/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4639static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4640{
4c3024de 4641 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4642}
4643
5293efe6
DB
4644static inline void skb_gso_reset(struct sk_buff *skb)
4645{
4646 skb_shinfo(skb)->gso_size = 0;
4647 skb_shinfo(skb)->gso_segs = 0;
4648 skb_shinfo(skb)->gso_type = 0;
4649}
4650
d02f51cb
DA
4651static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4652 u16 increment)
4653{
4654 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4655 return;
4656 shinfo->gso_size += increment;
4657}
4658
4659static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4660 u16 decrement)
4661{
4662 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4663 return;
4664 shinfo->gso_size -= decrement;
4665}
4666
7965bd4d 4667void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4668
4669static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4670{
4671 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4672 * wanted then gso_type will be set. */
05bdd2f1
ED
4673 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4674
b78462eb
AD
4675 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4676 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4677 __skb_warn_lro_forwarding(skb);
4678 return true;
4679 }
4680 return false;
4681}
4682
35fc92a9
HX
4683static inline void skb_forward_csum(struct sk_buff *skb)
4684{
4685 /* Unfortunately we don't support this one. Any brave souls? */
4686 if (skb->ip_summed == CHECKSUM_COMPLETE)
4687 skb->ip_summed = CHECKSUM_NONE;
4688}
4689
bc8acf2c
ED
4690/**
4691 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4692 * @skb: skb to check
4693 *
4694 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4695 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4696 * use this helper, to document places where we make this assertion.
4697 */
05bdd2f1 4698static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4699{
4700#ifdef DEBUG
4701 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4702#endif
4703}
4704
f35d9d8a 4705bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4706
ed1f50c3 4707int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4708struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4709 unsigned int transport_len,
4710 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4711
3a7c1ee4
AD
4712/**
4713 * skb_head_is_locked - Determine if the skb->head is locked down
4714 * @skb: skb to check
4715 *
4716 * The head on skbs build around a head frag can be removed if they are
4717 * not cloned. This function returns true if the skb head is locked down
4718 * due to either being allocated via kmalloc, or by being a clone with
4719 * multiple references to the head.
4720 */
4721static inline bool skb_head_is_locked(const struct sk_buff *skb)
4722{
4723 return !skb->head_frag || skb_cloned(skb);
4724}
fe6cc55f 4725
179bc67f
EC
4726/* Local Checksum Offload.
4727 * Compute outer checksum based on the assumption that the
4728 * inner checksum will be offloaded later.
d0dcde64 4729 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4730 * explanation of how this works.
179bc67f
EC
4731 * Fill in outer checksum adjustment (e.g. with sum of outer
4732 * pseudo-header) before calling.
4733 * Also ensure that inner checksum is in linear data area.
4734 */
4735static inline __wsum lco_csum(struct sk_buff *skb)
4736{
9e74a6da
AD
4737 unsigned char *csum_start = skb_checksum_start(skb);
4738 unsigned char *l4_hdr = skb_transport_header(skb);
4739 __wsum partial;
179bc67f
EC
4740
4741 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4742 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4743 skb->csum_offset));
4744
179bc67f 4745 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4746 * adjustment filled in by caller) and return result.
179bc67f 4747 */
9e74a6da 4748 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4749}
4750
2c64605b
PNA
4751static inline bool skb_is_redirected(const struct sk_buff *skb)
4752{
2c64605b 4753 return skb->redirected;
2c64605b
PNA
4754}
4755
4756static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4757{
2c64605b 4758 skb->redirected = 1;
11941f8a 4759#ifdef CONFIG_NET_REDIRECT
2c64605b
PNA
4760 skb->from_ingress = from_ingress;
4761 if (skb->from_ingress)
4762 skb->tstamp = 0;
4763#endif
4764}
4765
4766static inline void skb_reset_redirect(struct sk_buff *skb)
4767{
2c64605b 4768 skb->redirected = 0;
2c64605b
PNA
4769}
4770
fa821170
XL
4771static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4772{
4773 return skb->csum_not_inet;
4774}
4775
6370cc3b
AN
4776static inline void skb_set_kcov_handle(struct sk_buff *skb,
4777 const u64 kcov_handle)
4778{
fa69ee5a
ME
4779#ifdef CONFIG_KCOV
4780 skb->kcov_handle = kcov_handle;
4781#endif
6370cc3b
AN
4782}
4783
4784static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4785{
fa69ee5a
ME
4786#ifdef CONFIG_KCOV
4787 return skb->kcov_handle;
6370cc3b 4788#else
fa69ee5a
ME
4789 return 0;
4790#endif
4791}
6370cc3b 4792
6a5bcd84 4793#ifdef CONFIG_PAGE_POOL
57f05bc2 4794static inline void skb_mark_for_recycle(struct sk_buff *skb)
6a5bcd84
IA
4795{
4796 skb->pp_recycle = 1;
6a5bcd84
IA
4797}
4798#endif
4799
4800static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4801{
4802 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4803 return false;
4804 return page_pool_return_skb_page(virt_to_page(data));
4805}
4806
1da177e4
LT
4807#endif /* __KERNEL__ */
4808#endif /* _LINUX_SKBUFF_H */