Merge tag '5.19-rc-ksmbd-server-fixes' of git://git.samba.org/ksmbd
[linux-2.6-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
66e4c8d9 45#include <net/net_debug.h>
1da177e4 46
9facd941
JK
47/**
48 * DOC: skb checksums
49 *
50 * The interface for checksum offload between the stack and networking drivers
7a6ae71b
TH
51 * is as follows...
52 *
9facd941
JK
53 * IP checksum related features
54 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
7a6ae71b
TH
55 *
56 * Drivers advertise checksum offload capabilities in the features of a device.
db1f00fb
DC
57 * From the stack's point of view these are capabilities offered by the driver.
58 * A driver typically only advertises features that it is capable of offloading
7a6ae71b
TH
59 * to its device.
60 *
9facd941
JK
61 * .. flat-table:: Checksum related device features
62 * :widths: 1 10
63 *
64 * * - %NETIF_F_HW_CSUM
65 * - The driver (or its device) is able to compute one
66 * IP (one's complement) checksum for any combination
67 * of protocols or protocol layering. The checksum is
68 * computed and set in a packet per the CHECKSUM_PARTIAL
69 * interface (see below).
70 *
71 * * - %NETIF_F_IP_CSUM
72 * - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv4. These are specifically
74 * unencapsulated packets of the form IPv4|TCP or
75 * IPv4|UDP where the Protocol field in the IPv4 header
76 * is TCP or UDP. The IPv4 header may contain IP options.
77 * This feature cannot be set in features for a device
78 * with NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * * - %NETIF_F_IPV6_CSUM
82 * - Driver (device) is only able to checksum plain
83 * TCP or UDP packets over IPv6. These are specifically
84 * unencapsulated packets of the form IPv6|TCP or
85 * IPv6|UDP where the Next Header field in the IPv6
86 * header is either TCP or UDP. IPv6 extension headers
87 * are not supported with this feature. This feature
88 * cannot be set in features for a device with
89 * NETIF_F_HW_CSUM also set. This feature is being
90 * DEPRECATED (see below).
91 *
92 * * - %NETIF_F_RXCSUM
93 * - Driver (device) performs receive checksum offload.
94 * This flag is only used to disable the RX checksum
95 * feature for a device. The stack will accept receive
96 * checksum indication in packets received on a device
97 * regardless of whether NETIF_F_RXCSUM is set.
98 *
99 * Checksumming of received packets by device
100 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
101 *
102 * Indication of checksum verification is set in &sk_buff.ip_summed.
103 * Possible values are:
104 *
105 * - %CHECKSUM_NONE
78ea85f1 106 *
7a6ae71b 107 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
108 * The packet contains full (though not verified) checksum in packet but
109 * not in skb->csum. Thus, skb->csum is undefined in this case.
110 *
9facd941 111 * - %CHECKSUM_UNNECESSARY
78ea85f1
DB
112 *
113 * The hardware you're dealing with doesn't calculate the full checksum
9facd941
JK
114 * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
115 * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
116 * if their checksums are okay. &sk_buff.csum is still undefined in this case
7a6ae71b
TH
117 * though. A driver or device must never modify the checksum field in the
118 * packet even if checksum is verified.
77cffe23 119 *
9facd941
JK
120 * %CHECKSUM_UNNECESSARY is applicable to following protocols:
121 *
122 * - TCP: IPv6 and IPv4.
123 * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
77cffe23
TH
124 * zero UDP checksum for either IPv4 or IPv6, the networking stack
125 * may perform further validation in this case.
9facd941
JK
126 * - GRE: only if the checksum is present in the header.
127 * - SCTP: indicates the CRC in SCTP header has been validated.
128 * - FCOE: indicates the CRC in FC frame has been validated.
77cffe23 129 *
9facd941
JK
130 * &sk_buff.csum_level indicates the number of consecutive checksums found in
131 * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
77cffe23
TH
132 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
133 * and a device is able to verify the checksums for UDP (possibly zero),
9facd941 134 * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
77cffe23 135 * two. If the device were only able to verify the UDP checksum and not
db1f00fb 136 * GRE, either because it doesn't support GRE checksum or because GRE
77cffe23
TH
137 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
138 * not considered in this case).
78ea85f1 139 *
9facd941 140 * - %CHECKSUM_COMPLETE
78ea85f1
DB
141 *
142 * This is the most generic way. The device supplied checksum of the _whole_
9facd941 143 * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
78ea85f1
DB
144 * hardware doesn't need to parse L3/L4 headers to implement this.
145 *
b4759dcd 146 * Notes:
9facd941 147 *
b4759dcd
DC
148 * - Even if device supports only some protocols, but is able to produce
149 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
150 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1 151 *
9facd941 152 * - %CHECKSUM_PARTIAL
78ea85f1 153 *
6edec0e6
TH
154 * A checksum is set up to be offloaded to a device as described in the
155 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 156 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
157 * on the same host, or it may be set in the input path in GRO or remote
158 * checksum offload. For the purposes of checksum verification, the checksum
159 * referred to by skb->csum_start + skb->csum_offset and any preceding
160 * checksums in the packet are considered verified. Any checksums in the
161 * packet that are after the checksum being offloaded are not considered to
162 * be verified.
78ea85f1 163 *
9facd941
JK
164 * Checksumming on transmit for non-GSO
165 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
78ea85f1 166 *
9facd941
JK
167 * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
168 * Values are:
169 *
170 * - %CHECKSUM_PARTIAL
78ea85f1 171 *
7a6ae71b 172 * The driver is required to checksum the packet as seen by hard_start_xmit()
9facd941
JK
173 * from &sk_buff.csum_start up to the end, and to record/write the checksum at
174 * offset &sk_buff.csum_start + &sk_buff.csum_offset.
175 * A driver may verify that the
7a6ae71b 176 * csum_start and csum_offset values are valid values given the length and
db1f00fb
DC
177 * offset of the packet, but it should not attempt to validate that the
178 * checksum refers to a legitimate transport layer checksum -- it is the
7a6ae71b
TH
179 * purview of the stack to validate that csum_start and csum_offset are set
180 * correctly.
181 *
182 * When the stack requests checksum offload for a packet, the driver MUST
183 * ensure that the checksum is set correctly. A driver can either offload the
184 * checksum calculation to the device, or call skb_checksum_help (in the case
185 * that the device does not support offload for a particular checksum).
186 *
9facd941
JK
187 * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
188 * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
43c26a1a 189 * checksum offload capability.
9facd941 190 * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
43c26a1a 191 * on network device checksumming capabilities: if a packet does not match
9facd941
JK
192 * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
193 * &sk_buff.csum_not_inet, see :ref:`crc`)
194 * is called to resolve the checksum.
78ea85f1 195 *
9facd941 196 * - %CHECKSUM_NONE
78ea85f1 197 *
7a6ae71b
TH
198 * The skb was already checksummed by the protocol, or a checksum is not
199 * required.
78ea85f1 200 *
9facd941 201 * - %CHECKSUM_UNNECESSARY
78ea85f1 202 *
db1f00fb 203 * This has the same meaning as CHECKSUM_NONE for checksum offload on
7a6ae71b 204 * output.
78ea85f1 205 *
9facd941
JK
206 * - %CHECKSUM_COMPLETE
207 *
7a6ae71b 208 * Not used in checksum output. If a driver observes a packet with this value
9facd941
JK
209 * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
210 *
211 * .. _crc:
212 *
213 * Non-IP checksum (CRC) offloads
214 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
215 *
216 * .. flat-table::
217 * :widths: 1 10
218 *
219 * * - %NETIF_F_SCTP_CRC
220 * - This feature indicates that a device is capable of
221 * offloading the SCTP CRC in a packet. To perform this offload the stack
222 * will set csum_start and csum_offset accordingly, set ip_summed to
223 * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
224 * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
225 * A driver that supports both IP checksum offload and SCTP CRC32c offload
226 * must verify which offload is configured for a packet by testing the
227 * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
228 * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
229 *
230 * * - %NETIF_F_FCOE_CRC
231 * - This feature indicates that a device is capable of offloading the FCOE
232 * CRC in a packet. To perform this offload the stack will set ip_summed
233 * to %CHECKSUM_PARTIAL and set csum_start and csum_offset
234 * accordingly. Note that there is no indication in the skbuff that the
235 * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
236 * both IP checksum offload and FCOE CRC offload must verify which offload
237 * is configured for a packet, presumably by inspecting packet headers.
238 *
239 * Checksumming on output with GSO
240 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
241 *
242 * In the case of a GSO packet (skb_is_gso() is true), checksum offload
7a6ae71b 243 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
9facd941 244 * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
7a6ae71b 245 * part of the GSO operation is implied. If a checksum is being offloaded
9facd941 246 * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
db1f00fb
DC
247 * csum_offset are set to refer to the outermost checksum being offloaded
248 * (two offloaded checksums are possible with UDP encapsulation).
78ea85f1
DB
249 */
250
60476372 251/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
252#define CHECKSUM_NONE 0
253#define CHECKSUM_UNNECESSARY 1
254#define CHECKSUM_COMPLETE 2
255#define CHECKSUM_PARTIAL 3
1da177e4 256
77cffe23
TH
257/* Maximum value in skb->csum_level */
258#define SKB_MAX_CSUM_LEVEL 3
259
0bec8c88 260#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 261#define SKB_WITH_OVERHEAD(X) \
deea84b0 262 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
263#define SKB_MAX_ORDER(X, ORDER) \
264 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
265#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
266#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
267
87fb4b7b
ED
268/* return minimum truesize of one skb containing X bytes of data */
269#define SKB_TRUESIZE(X) ((X) + \
270 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
271 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
272
7999096f 273struct ahash_request;
1da177e4 274struct net_device;
716ea3a7 275struct scatterlist;
9c55e01c 276struct pipe_inode_info;
a8f820aa 277struct iov_iter;
fd11a83d 278struct napi_struct;
d58e468b
PP
279struct bpf_prog;
280union bpf_attr;
df5042f4 281struct skb_ext;
1da177e4 282
34666d46 283#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 284struct nf_bridge_info {
3eaf4025
FW
285 enum {
286 BRNF_PROTO_UNCHANGED,
287 BRNF_PROTO_8021Q,
288 BRNF_PROTO_PPPOE
7fb48c5b 289 } orig_proto:8;
72b1e5e4
FW
290 u8 pkt_otherhost:1;
291 u8 in_prerouting:1;
292 u8 bridged_dnat:1;
411ffb4f 293 __u16 frag_max_size;
bf1ac5ca 294 struct net_device *physindev;
63cdbc06
FW
295
296 /* always valid & non-NULL from FORWARD on, for physdev match */
297 struct net_device *physoutdev;
7fb48c5b 298 union {
72b1e5e4 299 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
300 __be32 ipv4_daddr;
301 struct in6_addr ipv6_daddr;
72b1e5e4
FW
302
303 /* after prerouting + nat detected: store original source
304 * mac since neigh resolution overwrites it, only used while
305 * skb is out in neigh layer.
306 */
307 char neigh_header[8];
72b31f72 308 };
1da177e4
LT
309};
310#endif
311
95a7233c
PB
312#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
313/* Chain in tc_skb_ext will be used to share the tc chain with
314 * ovs recirc_id. It will be set to the current chain by tc
315 * and read by ovs to recirc_id.
316 */
317struct tc_skb_ext {
318 __u32 chain;
038ebb1a 319 __u16 mru;
635d448a 320 __u16 zone;
6f022c2d
PB
321 u8 post_ct:1;
322 u8 post_ct_snat:1;
323 u8 post_ct_dnat:1;
95a7233c
PB
324};
325#endif
326
1da177e4 327struct sk_buff_head {
1a2fb220
KC
328 /* These two members must be first to match sk_buff. */
329 struct_group_tagged(sk_buff_list, list,
330 struct sk_buff *next;
331 struct sk_buff *prev;
332 );
1da177e4
LT
333
334 __u32 qlen;
335 spinlock_t lock;
336};
337
338struct sk_buff;
339
c504e5c2
MD
340/* The reason of skb drop, which is used in kfree_skb_reason().
341 * en...maybe they should be splited by group?
342 *
343 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
344 * used to translate the reason to string.
345 */
346enum skb_drop_reason {
1330b6ef 347 SKB_NOT_DROPPED_YET = 0,
88590b36
MD
348 SKB_DROP_REASON_NOT_SPECIFIED, /* drop reason is not specified */
349 SKB_DROP_REASON_NO_SOCKET, /* socket not found */
350 SKB_DROP_REASON_PKT_TOO_SMALL, /* packet size is too small */
351 SKB_DROP_REASON_TCP_CSUM, /* TCP checksum error */
352 SKB_DROP_REASON_SOCKET_FILTER, /* dropped by socket filter */
353 SKB_DROP_REASON_UDP_CSUM, /* UDP checksum error */
2df3041b 354 SKB_DROP_REASON_NETFILTER_DROP, /* dropped by netfilter */
33cba429
MD
355 SKB_DROP_REASON_OTHERHOST, /* packet don't belong to current
356 * host (interface is in promisc
357 * mode)
358 */
359 SKB_DROP_REASON_IP_CSUM, /* IP checksum error */
360 SKB_DROP_REASON_IP_INHDR, /* there is something wrong with
361 * IP header (see
362 * IPSTATS_MIB_INHDRERRORS)
363 */
c1f166d1
MD
364 SKB_DROP_REASON_IP_RPFILTER, /* IP rpfilter validate failed.
365 * see the document for rp_filter
366 * in ip-sysctl.rst for more
367 * information
368 */
369 SKB_DROP_REASON_UNICAST_IN_L2_MULTICAST, /* destination address of L2
370 * is multicast, but L3 is
371 * unicast.
372 */
10580c47
MD
373 SKB_DROP_REASON_XFRM_POLICY, /* xfrm policy check failed */
374 SKB_DROP_REASON_IP_NOPROTO, /* no support for IP protocol */
08d4c037
MD
375 SKB_DROP_REASON_SOCKET_RCVBUFF, /* socket receive buff is full */
376 SKB_DROP_REASON_PROTO_MEM, /* proto memory limition, such as
377 * udp packet drop out of
378 * udp_memory_allocated.
379 */
643b622b
MD
380 SKB_DROP_REASON_TCP_MD5NOTFOUND, /* no MD5 hash and one
381 * expected, corresponding
382 * to LINUX_MIB_TCPMD5NOTFOUND
383 */
384 SKB_DROP_REASON_TCP_MD5UNEXPECTED, /* MD5 hash and we're not
385 * expecting one, corresponding
386 * to LINUX_MIB_TCPMD5UNEXPECTED
387 */
388 SKB_DROP_REASON_TCP_MD5FAILURE, /* MD5 hash and its wrong,
389 * corresponding to
390 * LINUX_MIB_TCPMD5FAILURE
391 */
7a26dc9e
MD
392 SKB_DROP_REASON_SOCKET_BACKLOG, /* failed to add skb to socket
393 * backlog (see
394 * LINUX_MIB_TCPBACKLOGDROP)
395 */
2a968ef6 396 SKB_DROP_REASON_TCP_FLAGS, /* TCP flags invalid */
a7ec3810
MD
397 SKB_DROP_REASON_TCP_ZEROWINDOW, /* TCP receive window size is zero,
398 * see LINUX_MIB_TCPZEROWINDOWDROP
399 */
400 SKB_DROP_REASON_TCP_OLD_DATA, /* the TCP data reveived is already
401 * received before (spurious retrans
402 * may happened), see
403 * LINUX_MIB_DELAYEDACKLOST
404 */
405 SKB_DROP_REASON_TCP_OVERWINDOW, /* the TCP data is out of window,
406 * the seq of the first byte exceed
407 * the right edges of receive
408 * window
409 */
d25e481b
MD
410 SKB_DROP_REASON_TCP_OFOMERGE, /* the data of skb is already in
411 * the ofo queue, corresponding to
412 * LINUX_MIB_TCPOFOMERGE
413 */
da40b613
ED
414 SKB_DROP_REASON_TCP_RFC7323_PAWS, /* PAWS check, corresponding to
415 * LINUX_MIB_PAWSESTABREJECTED
416 */
417 SKB_DROP_REASON_TCP_INVALID_SEQUENCE, /* Not acceptable SEQ field */
418 SKB_DROP_REASON_TCP_RESET, /* Invalid RST packet */
419 SKB_DROP_REASON_TCP_INVALID_SYN, /* Incoming packet has unexpected SYN flag */
669da7a7
ED
420 SKB_DROP_REASON_TCP_CLOSE, /* TCP socket in CLOSE state */
421 SKB_DROP_REASON_TCP_FASTOPEN, /* dropped by FASTOPEN request socket */
422 SKB_DROP_REASON_TCP_OLD_ACK, /* TCP ACK is old, but in window */
4b506af9
ED
423 SKB_DROP_REASON_TCP_TOO_OLD_ACK, /* TCP ACK is too old */
424 SKB_DROP_REASON_TCP_ACK_UNSENT_DATA, /* TCP ACK for data we haven't sent yet */
e7c89ae4 425 SKB_DROP_REASON_TCP_OFO_QUEUE_PRUNE, /* pruned from TCP OFO queue */
8fbf1957 426 SKB_DROP_REASON_TCP_OFO_DROP, /* data already in receive queue */
5e187189
MD
427 SKB_DROP_REASON_IP_OUTNOROUTES, /* route lookup failed */
428 SKB_DROP_REASON_BPF_CGROUP_EGRESS, /* dropped by
429 * BPF_PROG_TYPE_CGROUP_SKB
430 * eBPF program
431 */
432 SKB_DROP_REASON_IPV6DISABLED, /* IPv6 is disabled on the device */
433 SKB_DROP_REASON_NEIGH_CREATEFAIL, /* failed to create neigh
434 * entry
435 */
a5736edd
MD
436 SKB_DROP_REASON_NEIGH_FAILED, /* neigh entry in failed state */
437 SKB_DROP_REASON_NEIGH_QUEUEFULL, /* arp_queue for neigh
438 * entry is full
439 */
440 SKB_DROP_REASON_NEIGH_DEAD, /* neigh entry is dead */
98b4d7a4 441 SKB_DROP_REASON_TC_EGRESS, /* dropped in TC egress HOOK */
7faef054
MD
442 SKB_DROP_REASON_QDISC_DROP, /* dropped by qdisc when packet
443 * outputting (failed to enqueue to
444 * current qdisc)
445 */
44f0bd40
MD
446 SKB_DROP_REASON_CPU_BACKLOG, /* failed to enqueue the skb to
447 * the per CPU backlog queue. This
448 * can be caused by backlog queue
449 * full (see netdev_max_backlog in
450 * net.rst) or RPS flow limit
451 */
7e726ed8 452 SKB_DROP_REASON_XDP, /* dropped by XDP in input path */
a568aff2 453 SKB_DROP_REASON_TC_INGRESS, /* dropped in TC ingress HOOK */
9f8ed577
MD
454 SKB_DROP_REASON_UNHANDLED_PROTO, /* protocol not implemented
455 * or not supported
456 */
736f16de
DZ
457 SKB_DROP_REASON_SKB_CSUM, /* sk_buff checksum computation
458 * error
459 */
460 SKB_DROP_REASON_SKB_GSO_SEG, /* gso segmentation error */
461 SKB_DROP_REASON_SKB_UCOPY_FAULT, /* failed to copy data from
462 * user space, e.g., via
463 * zerocopy_sg_from_iter()
464 * or skb_orphan_frags_rx()
465 */
466 SKB_DROP_REASON_DEV_HDR, /* device driver specific
467 * header/metadata is invalid
468 */
4b4f052e
DZ
469 /* the device is not ready to xmit/recv due to any of its data
470 * structure that is not up/ready/initialized, e.g., the IFF_UP is
471 * not set, or driver specific tun->tfiles[txq] is not initialized
472 */
473 SKB_DROP_REASON_DEV_READY,
736f16de 474 SKB_DROP_REASON_FULL_RING, /* ring buffer is full */
4b4f052e
DZ
475 SKB_DROP_REASON_NOMEM, /* error due to OOM */
476 SKB_DROP_REASON_HDR_TRUNC, /* failed to trunc/extract the header
477 * from networking data, e.g., failed
478 * to pull the protocol header from
479 * frags via pskb_may_pull()
480 */
481 SKB_DROP_REASON_TAP_FILTER, /* dropped by (ebpf) filter directly
482 * attached to tun/tap, e.g., via
483 * TUNSETFILTEREBPF
484 */
485 SKB_DROP_REASON_TAP_TXFILTER, /* dropped by tx filter implemented
486 * at tun/tap, e.g., check_filter()
487 */
b384c95a
MD
488 SKB_DROP_REASON_ICMP_CSUM, /* ICMP checksum error */
489 SKB_DROP_REASON_INVALID_PROTO, /* the packet doesn't follow RFC
490 * 2211, such as a broadcasts
491 * ICMP_TIMESTAMP
492 */
c4eb6641
MD
493 SKB_DROP_REASON_IP_INADDRERRORS, /* host unreachable, corresponding
494 * to IPSTATS_MIB_INADDRERRORS
495 */
496 SKB_DROP_REASON_IP_INNOROUTES, /* network unreachable, corresponding
497 * to IPSTATS_MIB_INADDRERRORS
498 */
2edc1a38
MD
499 SKB_DROP_REASON_PKT_TOO_BIG, /* packet size is too big (maybe exceed
500 * the MTU)
501 */
c504e5c2
MD
502 SKB_DROP_REASON_MAX,
503};
504
d6d3146c
MD
505#define SKB_DR_INIT(name, reason) \
506 enum skb_drop_reason name = SKB_DROP_REASON_##reason
507#define SKB_DR(name) \
508 SKB_DR_INIT(name, NOT_SPECIFIED)
509#define SKB_DR_SET(name, reason) \
510 (name = SKB_DROP_REASON_##reason)
511#define SKB_DR_OR(name, reason) \
512 do { \
7ebd3f3e
MD
513 if (name == SKB_DROP_REASON_NOT_SPECIFIED || \
514 name == SKB_NOT_DROPPED_YET) \
d6d3146c
MD
515 SKB_DR_SET(name, reason); \
516 } while (0)
517
9d4dde52
IC
518/* To allow 64K frame to be packed as single skb without frag_list we
519 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
520 * buffers which do not start on a page boundary.
521 *
522 * Since GRO uses frags we allocate at least 16 regardless of page
523 * size.
a715dea3 524 */
9d4dde52 525#if (65536/PAGE_SIZE + 1) < 16
eec00954 526#define MAX_SKB_FRAGS 16UL
a715dea3 527#else
9d4dde52 528#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 529#endif
5f74f82e 530extern int sysctl_max_skb_frags;
1da177e4 531
3953c46c
MRL
532/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
533 * segment using its current segmentation instead.
534 */
535#define GSO_BY_FRAGS 0xFFFF
536
8842d285 537typedef struct bio_vec skb_frag_t;
1da177e4 538
161e6137 539/**
7240b60c 540 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
541 * @frag: skb fragment
542 */
9e903e08
ED
543static inline unsigned int skb_frag_size(const skb_frag_t *frag)
544{
b8b576a1 545 return frag->bv_len;
9e903e08
ED
546}
547
161e6137 548/**
7240b60c 549 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
550 * @frag: skb fragment
551 * @size: size of fragment
552 */
9e903e08
ED
553static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
554{
b8b576a1 555 frag->bv_len = size;
9e903e08
ED
556}
557
161e6137 558/**
7240b60c 559 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
560 * @frag: skb fragment
561 * @delta: value to add
562 */
9e903e08
ED
563static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
564{
b8b576a1 565 frag->bv_len += delta;
9e903e08
ED
566}
567
161e6137 568/**
7240b60c 569 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
570 * @frag: skb fragment
571 * @delta: value to subtract
572 */
9e903e08
ED
573static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
574{
b8b576a1 575 frag->bv_len -= delta;
9e903e08
ED
576}
577
161e6137
PT
578/**
579 * skb_frag_must_loop - Test if %p is a high memory page
580 * @p: fragment's page
581 */
c613c209
WB
582static inline bool skb_frag_must_loop(struct page *p)
583{
584#if defined(CONFIG_HIGHMEM)
29766bcf 585 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
586 return true;
587#endif
588 return false;
589}
590
591/**
592 * skb_frag_foreach_page - loop over pages in a fragment
593 *
594 * @f: skb frag to operate on
1dfa5bd3 595 * @f_off: offset from start of f->bv_page
c613c209
WB
596 * @f_len: length from f_off to loop over
597 * @p: (temp var) current page
598 * @p_off: (temp var) offset from start of current page,
599 * non-zero only on first page.
600 * @p_len: (temp var) length in current page,
601 * < PAGE_SIZE only on first and last page.
602 * @copied: (temp var) length so far, excluding current p_len.
603 *
604 * A fragment can hold a compound page, in which case per-page
605 * operations, notably kmap_atomic, must be called for each
606 * regular page.
607 */
608#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
609 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
610 p_off = (f_off) & (PAGE_SIZE - 1), \
611 p_len = skb_frag_must_loop(p) ? \
612 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
613 copied = 0; \
614 copied < f_len; \
615 copied += p_len, p++, p_off = 0, \
616 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
617
ac45f602
PO
618#define HAVE_HW_TIME_STAMP
619
620/**
d3a21be8 621 * struct skb_shared_hwtstamps - hardware time stamps
97dc7cd9
GE
622 * @hwtstamp: hardware time stamp transformed into duration
623 * since arbitrary point in time
624 * @netdev_data: address/cookie of network device driver used as
625 * reference to actual hardware time stamp
ac45f602
PO
626 *
627 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 628 * skb->tstamp.
ac45f602
PO
629 *
630 * hwtstamps can only be compared against other hwtstamps from
631 * the same device.
632 *
633 * This structure is attached to packets as part of the
634 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
635 */
636struct skb_shared_hwtstamps {
97dc7cd9
GE
637 union {
638 ktime_t hwtstamp;
639 void *netdev_data;
640 };
ac45f602
PO
641};
642
2244d07b
OH
643/* Definitions for tx_flags in struct skb_shared_info */
644enum {
645 /* generate hardware time stamp */
646 SKBTX_HW_TSTAMP = 1 << 0,
647
e7fd2885 648 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
649 SKBTX_SW_TSTAMP = 1 << 1,
650
651 /* device driver is going to provide hardware time stamp */
652 SKBTX_IN_PROGRESS = 1 << 2,
653
51eb7492
GE
654 /* generate hardware time stamp based on cycles if supported */
655 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3,
656
6e3e939f 657 /* generate wifi status information (where possible) */
62b1a8ab 658 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 659
97dc7cd9
GE
660 /* determine hardware time stamp based on time or cycles */
661 SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
662
e7fd2885
WB
663 /* generate software time stamp when entering packet scheduling */
664 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
665};
666
e1c8a607 667#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 668 SKBTX_SCHED_TSTAMP)
51eb7492
GE
669#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \
670 SKBTX_HW_TSTAMP_USE_CYCLES | \
671 SKBTX_ANY_SW_TSTAMP)
f24b9be5 672
06b4feb3
JL
673/* Definitions for flags in struct skb_shared_info */
674enum {
675 /* use zcopy routines */
676 SKBFL_ZEROCOPY_ENABLE = BIT(0),
677
678 /* This indicates at least one fragment might be overwritten
679 * (as in vmsplice(), sendfile() ...)
680 * If we need to compute a TX checksum, we'll need to copy
681 * all frags to avoid possible bad checksum
682 */
683 SKBFL_SHARED_FRAG = BIT(1),
9b65b17d
TA
684
685 /* segment contains only zerocopy data and should not be
686 * charged to the kernel memory.
687 */
688 SKBFL_PURE_ZEROCOPY = BIT(2),
06b4feb3
JL
689};
690
691#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
9b65b17d 692#define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY)
06b4feb3 693
a6686f2f
SM
694/*
695 * The callback notifies userspace to release buffers when skb DMA is done in
696 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
697 * The zerocopy_success argument is true if zero copy transmit occurred,
698 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
699 * The ctx field is used to track device context.
700 * The desc field is used to track userspace buffer index.
a6686f2f
SM
701 */
702struct ubuf_info {
36177832
JL
703 void (*callback)(struct sk_buff *, struct ubuf_info *,
704 bool zerocopy_success);
4ab6c99d
WB
705 union {
706 struct {
707 unsigned long desc;
708 void *ctx;
709 };
710 struct {
711 u32 id;
712 u16 len;
713 u16 zerocopy:1;
714 u32 bytelen;
715 };
716 };
c1d1b437 717 refcount_t refcnt;
04c2d33e 718 u8 flags;
a91dbff5
WB
719
720 struct mmpin {
721 struct user_struct *user;
722 unsigned int num_pg;
723 } mmp;
ac45f602
PO
724};
725
52267790
WB
726#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
727
6f89dbce
SV
728int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
729void mm_unaccount_pinned_pages(struct mmpin *mmp);
730
1da177e4
LT
731/* This data is invariant across clones and lives at
732 * the end of the header data, ie. at skb->end.
733 */
734struct skb_shared_info {
06b4feb3 735 __u8 flags;
de8f3a83
DB
736 __u8 meta_len;
737 __u8 nr_frags;
9f42f126 738 __u8 tx_flags;
7967168c
HX
739 unsigned short gso_size;
740 /* Warning: this field is not always filled in (UFO)! */
741 unsigned short gso_segs;
1da177e4 742 struct sk_buff *frag_list;
ac45f602 743 struct skb_shared_hwtstamps hwtstamps;
7f564528 744 unsigned int gso_type;
09c2d251 745 u32 tskey;
ec7d2f2c
ED
746
747 /*
748 * Warning : all fields before dataref are cleared in __alloc_skb()
749 */
750 atomic_t dataref;
d16697cb 751 unsigned int xdp_frags_size;
ec7d2f2c 752
69e3c75f
JB
753 /* Intermediate layers must ensure that destructor_arg
754 * remains valid until skb destructor */
755 void * destructor_arg;
a6686f2f 756
fed66381
ED
757 /* must be last field, see pskb_expand_head() */
758 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
759};
760
9ec7ea14
JK
761/**
762 * DOC: dataref and headerless skbs
763 *
764 * Transport layers send out clones of payload skbs they hold for
765 * retransmissions. To allow lower layers of the stack to prepend their headers
766 * we split &skb_shared_info.dataref into two halves.
767 * The lower 16 bits count the overall number of references.
768 * The higher 16 bits indicate how many of the references are payload-only.
769 * skb_header_cloned() checks if skb is allowed to add / write the headers.
770 *
771 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
772 * (via __skb_header_release()). Any clone created from marked skb will get
773 * &sk_buff.hdr_len populated with the available headroom.
774 * If there's the only clone in existence it's able to modify the headroom
775 * at will. The sequence of calls inside the transport layer is::
776 *
777 * <alloc skb>
778 * skb_reserve()
779 * __skb_header_release()
780 * skb_clone()
781 * // send the clone down the stack
782 *
783 * This is not a very generic construct and it depends on the transport layers
784 * doing the right thing. In practice there's usually only one payload-only skb.
785 * Having multiple payload-only skbs with different lengths of hdr_len is not
786 * possible. The payload-only skbs should never leave their owner.
1da177e4
LT
787 */
788#define SKB_DATAREF_SHIFT 16
789#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
790
d179cd12
DM
791
792enum {
c8753d55
VS
793 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
794 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
795 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
796};
797
7967168c
HX
798enum {
799 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
800
801 /* This indicates the skb is from an untrusted source. */
d9d30adf 802 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
803
804 /* This indicates the tcp segment has CWR set. */
d9d30adf 805 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 806
d9d30adf 807 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 808
d9d30adf 809 SKB_GSO_TCPV6 = 1 << 4,
68c33163 810
d9d30adf 811 SKB_GSO_FCOE = 1 << 5,
73136267 812
d9d30adf 813 SKB_GSO_GRE = 1 << 6,
0d89d203 814
d9d30adf 815 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 816
d9d30adf 817 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 818
d9d30adf 819 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 820
d9d30adf 821 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 822
d9d30adf 823 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 824
d9d30adf 825 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 826
d9d30adf 827 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 828
d9d30adf 829 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 830
d9d30adf 831 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
832
833 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
834
835 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
836
837 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
838};
839
2e07fa9c
ACM
840#if BITS_PER_LONG > 32
841#define NET_SKBUFF_DATA_USES_OFFSET 1
842#endif
843
844#ifdef NET_SKBUFF_DATA_USES_OFFSET
845typedef unsigned int sk_buff_data_t;
846#else
847typedef unsigned char *sk_buff_data_t;
848#endif
849
ddccc9ef
JK
850/**
851 * DOC: Basic sk_buff geometry
852 *
853 * struct sk_buff itself is a metadata structure and does not hold any packet
854 * data. All the data is held in associated buffers.
855 *
856 * &sk_buff.head points to the main "head" buffer. The head buffer is divided
857 * into two parts:
858 *
859 * - data buffer, containing headers and sometimes payload;
860 * this is the part of the skb operated on by the common helpers
861 * such as skb_put() or skb_pull();
862 * - shared info (struct skb_shared_info) which holds an array of pointers
863 * to read-only data in the (page, offset, length) format.
864 *
865 * Optionally &skb_shared_info.frag_list may point to another skb.
866 *
867 * Basic diagram may look like this::
868 *
869 * ---------------
870 * | sk_buff |
871 * ---------------
872 * ,--------------------------- + head
873 * / ,----------------- + data
874 * / / ,----------- + tail
875 * | | | , + end
876 * | | | |
877 * v v v v
878 * -----------------------------------------------
879 * | headroom | data | tailroom | skb_shared_info |
880 * -----------------------------------------------
881 * + [page frag]
882 * + [page frag]
883 * + [page frag]
884 * + [page frag] ---------
885 * + frag_list --> | sk_buff |
886 * ---------
887 *
888 */
889
161e6137 890/**
1da177e4
LT
891 * struct sk_buff - socket buffer
892 * @next: Next buffer in list
893 * @prev: Previous buffer in list
363ec392 894 * @tstamp: Time we arrived/left
d2f273f0
RD
895 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
896 * for retransmit timer
56b17425 897 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 898 * @list: queue head
df6160de 899 * @ll_node: anchor in an llist (eg socket defer_list)
d84e0bd7 900 * @sk: Socket we are owned by
d2f273f0
RD
901 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
902 * fragmentation management
1da177e4 903 * @dev: Device we arrived on/are leaving by
d2f273f0 904 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 905 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 906 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 907 * @sp: the security path, used for xfrm
1da177e4
LT
908 * @len: Length of actual data
909 * @data_len: Data length
910 * @mac_len: Length of link layer header
334a8132 911 * @hdr_len: writable header length of cloned skb
663ead3b
HX
912 * @csum: Checksum (must include start/offset pair)
913 * @csum_start: Offset from skb->head where checksumming should start
914 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 915 * @priority: Packet queueing priority
60ff7467 916 * @ignore_df: allow local fragmentation
1da177e4 917 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 918 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
919 * @nohdr: Payload reference only, must not modify header
920 * @pkt_type: Packet class
c83c2486 921 * @fclone: skbuff clone status
c83c2486 922 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
923 * @inner_protocol_type: whether the inner protocol is
924 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
925 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
926 * @offload_fwd_mark: Packet was L2-forwarded in hardware
927 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 928 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 929 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
930 * @redirected: packet was redirected by packet classifier
931 * @from_ingress: packet was redirected from the ingress path
42df6e1d 932 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
31729363
RD
933 * @peeked: this packet has been seen already, so stats have been
934 * done for it, don't do them again
ba9dda3a 935 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
936 * @protocol: Packet protocol from driver
937 * @destructor: Destruct function
e2080072 938 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
6ed6e1c7 939 * @_sk_redir: socket redirection information for skmsg
a9e419dc 940 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 941 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 942 * @skb_iif: ifindex of device we arrived on
1da177e4 943 * @tc_index: Traffic control index
61b905da 944 * @hash: the packet hash
d84e0bd7 945 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
946 * @head_frag: skb was allocated from page fragments,
947 * not allocated by kmalloc() or vmalloc().
8b700862 948 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
6a5bcd84
IA
949 * @pp_recycle: mark the packet for recycling instead of freeing (implies
950 * page_pool support on driver)
df5042f4 951 * @active_extensions: active extensions (skb_ext_id types)
553a5672 952 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 953 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 954 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 955 * ports.
a3b18ddb 956 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
957 * @wifi_acked_valid: wifi_acked was set
958 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 959 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
960 * @encapsulation: indicates the inner headers in the skbuff are valid
961 * @encap_hdr_csum: software checksum is needed
962 * @csum_valid: checksum is already valid
dba00306 963 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
964 * @csum_complete_sw: checksum was completed by software
965 * @csum_level: indicates the number of consecutive checksums found in
966 * the packet minus one that have been verified as
967 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 968 * @dst_pending_confirm: need to confirm neighbour
a48d189e 969 * @decrypted: Decrypted SKB
5fc88f93 970 * @slow_gro: state present at GRO time, slower prepare step required
a1ac9c8a
MKL
971 * @mono_delivery_time: When set, skb->tstamp has the
972 * delivery_time in mono clock base (i.e. EDT). Otherwise, the
973 * skb->tstamp has the (rcv) timestamp at ingress and
974 * delivery_time at egress.
161e6137 975 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 976 * @sender_cpu: (aka @napi_id) source CPU in XPS
68822bdf 977 * @alloc_cpu: CPU which did the skb allocation.
984bc16c 978 * @secmark: security marking
d84e0bd7 979 * @mark: Generic packet mark
d2f273f0
RD
980 * @reserved_tailroom: (aka @mark) number of bytes of free space available
981 * at the tail of an sk_buff
982 * @vlan_present: VLAN tag is present
86a9bad3 983 * @vlan_proto: vlan encapsulation protocol
6aa895b0 984 * @vlan_tci: vlan tag control information
0d89d203 985 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
986 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
987 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
988 * @inner_transport_header: Inner transport layer header (encapsulation)
989 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 990 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
991 * @transport_header: Transport layer header
992 * @network_header: Network layer header
993 * @mac_header: Link layer header
fa69ee5a 994 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
995 * @tail: Tail pointer
996 * @end: End pointer
997 * @head: Head of buffer
998 * @data: Data head pointer
999 * @truesize: Buffer size
1000 * @users: User count - see {datagram,tcp}.c
df5042f4 1001 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
1002 */
1003
1004struct sk_buff {
363ec392 1005 union {
56b17425 1006 struct {
1a2fb220 1007 /* These two members must be first to match sk_buff_head. */
56b17425
ED
1008 struct sk_buff *next;
1009 struct sk_buff *prev;
1010
1011 union {
bffa72cf
ED
1012 struct net_device *dev;
1013 /* Some protocols might use this space to store information,
1014 * while device pointer would be NULL.
1015 * UDP receive path is one user.
1016 */
1017 unsigned long dev_scratch;
56b17425
ED
1018 };
1019 };
fa0f5273 1020 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 1021 struct list_head list;
f35f8219 1022 struct llist_node ll_node;
363ec392 1023 };
fa0f5273
PO
1024
1025 union {
1026 struct sock *sk;
1027 int ip_defrag_offset;
1028 };
1da177e4 1029
c84d9490 1030 union {
bffa72cf 1031 ktime_t tstamp;
d3edd06e 1032 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 1033 };
1da177e4
LT
1034 /*
1035 * This is the control buffer. It is free to use for every
1036 * layer. Please put your private variables there. If you
1037 * want to keep them across layers you have to do a skb_clone()
1038 * first. This is owned by whoever has the skb queued ATM.
1039 */
da3f5cf1 1040 char cb[48] __aligned(8);
1da177e4 1041
e2080072
ED
1042 union {
1043 struct {
1044 unsigned long _skb_refdst;
1045 void (*destructor)(struct sk_buff *skb);
1046 };
1047 struct list_head tcp_tsorted_anchor;
e3526bb9
CW
1048#ifdef CONFIG_NET_SOCK_MSG
1049 unsigned long _sk_redir;
1050#endif
e2080072
ED
1051 };
1052
b1937227 1053#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 1054 unsigned long _nfct;
da3f5cf1 1055#endif
1da177e4 1056 unsigned int len,
334a8132
PM
1057 data_len;
1058 __u16 mac_len,
1059 hdr_len;
b1937227
ED
1060
1061 /* Following fields are _not_ copied in __copy_skb_header()
1062 * Note that queue_mapping is here mostly to fill a hole.
1063 */
b1937227 1064 __u16 queue_mapping;
36bbef52
DB
1065
1066/* if you move cloned around you also must adapt those constants */
1067#ifdef __BIG_ENDIAN_BITFIELD
1068#define CLONED_MASK (1 << 7)
1069#else
1070#define CLONED_MASK 1
1071#endif
fba84957 1072#define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
36bbef52 1073
d2f273f0 1074 /* private: */
36bbef52 1075 __u8 __cloned_offset[0];
d2f273f0 1076 /* public: */
b1937227 1077 __u8 cloned:1,
6869c4d8 1078 nohdr:1,
b84f4cc9 1079 fclone:2,
a59322be 1080 peeked:1,
b1937227 1081 head_frag:1,
6a5bcd84
IA
1082 pfmemalloc:1,
1083 pp_recycle:1; /* page_pool recycle indicator */
df5042f4
FW
1084#ifdef CONFIG_SKB_EXTENSIONS
1085 __u8 active_extensions;
1086#endif
6a5bcd84 1087
03f61041 1088 /* Fields enclosed in headers group are copied
b1937227
ED
1089 * using a single memcpy() in __copy_skb_header()
1090 */
03f61041 1091 struct_group(headers,
4031ae6e 1092
d2f273f0 1093 /* private: */
233577a2 1094 __u8 __pkt_type_offset[0];
d2f273f0 1095 /* public: */
fba84957 1096 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
b1937227 1097 __u8 ignore_df:1;
b1937227
ED
1098 __u8 nf_trace:1;
1099 __u8 ip_summed:2;
3853b584 1100 __u8 ooo_okay:1;
8b700862 1101
61b905da 1102 __u8 l4_hash:1;
a3b18ddb 1103 __u8 sw_hash:1;
6e3e939f
JB
1104 __u8 wifi_acked_valid:1;
1105 __u8 wifi_acked:1;
3bdc0eba 1106 __u8 no_fcs:1;
77cffe23 1107 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 1108 __u8 encapsulation:1;
7e2b10c1 1109 __u8 encap_hdr_csum:1;
5d0c2b95 1110 __u8 csum_valid:1;
8b700862 1111
d2f273f0 1112 /* private: */
0c4b2d37 1113 __u8 __pkt_vlan_present_offset[0];
d2f273f0 1114 /* public: */
fba84957 1115 __u8 vlan_present:1; /* See PKT_VLAN_PRESENT_BIT */
7e3cead5 1116 __u8 csum_complete_sw:1;
b1937227 1117 __u8 csum_level:2;
4ff06203 1118 __u8 dst_pending_confirm:1;
3b5d4ddf 1119 __u8 mono_delivery_time:1; /* See SKB_MONO_DELIVERY_TIME_MASK */
7449197d
MKL
1120#ifdef CONFIG_NET_CLS_ACT
1121 __u8 tc_skip_classify:1;
3b5d4ddf 1122 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */
7449197d 1123#endif
b1937227
ED
1124#ifdef CONFIG_IPV6_NDISC_NODETYPE
1125 __u8 ndisc_nodetype:2;
1126#endif
8b700862 1127
0c4b2d37 1128 __u8 ipvs_property:1;
8bce6d7d 1129 __u8 inner_protocol_type:1;
e585f236 1130 __u8 remcsum_offload:1;
6bc506b4
IS
1131#ifdef CONFIG_NET_SWITCHDEV
1132 __u8 offload_fwd_mark:1;
875e8939 1133 __u8 offload_l3_fwd_mark:1;
2c64605b 1134#endif
2c64605b 1135 __u8 redirected:1;
11941f8a 1136#ifdef CONFIG_NET_REDIRECT
2c64605b 1137 __u8 from_ingress:1;
e7246e12 1138#endif
42df6e1d
LW
1139#ifdef CONFIG_NETFILTER_SKIP_EGRESS
1140 __u8 nf_skip_egress:1;
1141#endif
a48d189e
SB
1142#ifdef CONFIG_TLS_DEVICE
1143 __u8 decrypted:1;
1144#endif
5fc88f93 1145 __u8 slow_gro:1;
7449197d 1146 __u8 csum_not_inet:1;
b1937227
ED
1147
1148#ifdef CONFIG_NET_SCHED
1149 __u16 tc_index; /* traffic control index */
b1937227 1150#endif
fe55f6d5 1151
b1937227
ED
1152 union {
1153 __wsum csum;
1154 struct {
1155 __u16 csum_start;
1156 __u16 csum_offset;
1157 };
1158 };
1159 __u32 priority;
1160 int skb_iif;
1161 __u32 hash;
1162 __be16 vlan_proto;
1163 __u16 vlan_tci;
2bd82484
ED
1164#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1165 union {
1166 unsigned int napi_id;
1167 unsigned int sender_cpu;
1168 };
97fc2f08 1169#endif
68822bdf 1170 u16 alloc_cpu;
984bc16c 1171#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 1172 __u32 secmark;
0c4f691f 1173#endif
0c4f691f 1174
3b885787
NH
1175 union {
1176 __u32 mark;
16fad69c 1177 __u32 reserved_tailroom;
3b885787 1178 };
1da177e4 1179
8bce6d7d
TH
1180 union {
1181 __be16 inner_protocol;
1182 __u8 inner_ipproto;
1183 };
1184
1a37e412
SH
1185 __u16 inner_transport_header;
1186 __u16 inner_network_header;
1187 __u16 inner_mac_header;
b1937227
ED
1188
1189 __be16 protocol;
1a37e412
SH
1190 __u16 transport_header;
1191 __u16 network_header;
1192 __u16 mac_header;
b1937227 1193
fa69ee5a
ME
1194#ifdef CONFIG_KCOV
1195 u64 kcov_handle;
1196#endif
1197
03f61041 1198 ); /* end headers group */
b1937227 1199
1da177e4 1200 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 1201 sk_buff_data_t tail;
4305b541 1202 sk_buff_data_t end;
1da177e4 1203 unsigned char *head,
4305b541 1204 *data;
27a884dc 1205 unsigned int truesize;
63354797 1206 refcount_t users;
df5042f4
FW
1207
1208#ifdef CONFIG_SKB_EXTENSIONS
1209 /* only useable after checking ->active_extensions != 0 */
1210 struct skb_ext *extensions;
1211#endif
1da177e4
LT
1212};
1213
fba84957
KC
1214/* if you move pkt_type around you also must adapt those constants */
1215#ifdef __BIG_ENDIAN_BITFIELD
1216#define PKT_TYPE_MAX (7 << 5)
1217#else
1218#define PKT_TYPE_MAX 7
1219#endif
1220#define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset)
1221
3b5d4ddf
MKL
1222/* if you move pkt_vlan_present, tc_at_ingress, or mono_delivery_time
1223 * around, you also must adapt these constants.
1224 */
fba84957
KC
1225#ifdef __BIG_ENDIAN_BITFIELD
1226#define PKT_VLAN_PRESENT_BIT 7
7449197d
MKL
1227#define TC_AT_INGRESS_MASK (1 << 0)
1228#define SKB_MONO_DELIVERY_TIME_MASK (1 << 2)
fba84957
KC
1229#else
1230#define PKT_VLAN_PRESENT_BIT 0
7449197d
MKL
1231#define TC_AT_INGRESS_MASK (1 << 7)
1232#define SKB_MONO_DELIVERY_TIME_MASK (1 << 5)
fba84957
KC
1233#endif
1234#define PKT_VLAN_PRESENT_OFFSET offsetof(struct sk_buff, __pkt_vlan_present_offset)
1235
1da177e4
LT
1236#ifdef __KERNEL__
1237/*
1238 * Handling routines are only of interest to the kernel
1239 */
1da177e4 1240
c93bdd0e
MG
1241#define SKB_ALLOC_FCLONE 0x01
1242#define SKB_ALLOC_RX 0x02
fd11a83d 1243#define SKB_ALLOC_NAPI 0x04
c93bdd0e 1244
161e6137
PT
1245/**
1246 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1247 * @skb: buffer
1248 */
c93bdd0e
MG
1249static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1250{
1251 return unlikely(skb->pfmemalloc);
1252}
1253
7fee226a
ED
1254/*
1255 * skb might have a dst pointer attached, refcounted or not.
1256 * _skb_refdst low order bit is set if refcount was _not_ taken
1257 */
1258#define SKB_DST_NOREF 1UL
1259#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1260
1261/**
1262 * skb_dst - returns skb dst_entry
1263 * @skb: buffer
1264 *
1265 * Returns skb dst_entry, regardless of reference taken or not.
1266 */
adf30907
ED
1267static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1268{
161e6137 1269 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
1270 * rcu_read_lock section
1271 */
1272 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1273 !rcu_read_lock_held() &&
1274 !rcu_read_lock_bh_held());
1275 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
1276}
1277
7fee226a
ED
1278/**
1279 * skb_dst_set - sets skb dst
1280 * @skb: buffer
1281 * @dst: dst entry
1282 *
1283 * Sets skb dst, assuming a reference was taken on dst and should
1284 * be released by skb_dst_drop()
1285 */
adf30907
ED
1286static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1287{
8a886b14 1288 skb->slow_gro |= !!dst;
7fee226a
ED
1289 skb->_skb_refdst = (unsigned long)dst;
1290}
1291
932bc4d7
JA
1292/**
1293 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1294 * @skb: buffer
1295 * @dst: dst entry
1296 *
1297 * Sets skb dst, assuming a reference was not taken on dst.
1298 * If dst entry is cached, we do not take reference and dst_release
1299 * will be avoided by refdst_drop. If dst entry is not cached, we take
1300 * reference, so that last dst_release can destroy the dst immediately.
1301 */
1302static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1303{
dbfc4fb7 1304 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
a432934a 1305 skb->slow_gro |= !!dst;
dbfc4fb7 1306 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1307}
7fee226a
ED
1308
1309/**
25985edc 1310 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1311 * @skb: buffer
1312 */
1313static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1314{
1315 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1316}
1317
161e6137
PT
1318/**
1319 * skb_rtable - Returns the skb &rtable
1320 * @skb: buffer
1321 */
511c3f92
ED
1322static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1323{
adf30907 1324 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1325}
1326
8b10cab6
JHS
1327/* For mangling skb->pkt_type from user space side from applications
1328 * such as nft, tc, etc, we only allow a conservative subset of
1329 * possible pkt_types to be set.
1330*/
1331static inline bool skb_pkt_type_ok(u32 ptype)
1332{
1333 return ptype <= PACKET_OTHERHOST;
1334}
1335
161e6137
PT
1336/**
1337 * skb_napi_id - Returns the skb's NAPI id
1338 * @skb: buffer
1339 */
90b602f8
ML
1340static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1341{
1342#ifdef CONFIG_NET_RX_BUSY_POLL
1343 return skb->napi_id;
1344#else
1345 return 0;
1346#endif
1347}
1348
161e6137
PT
1349/**
1350 * skb_unref - decrement the skb's reference count
1351 * @skb: buffer
1352 *
1353 * Returns true if we can free the skb.
1354 */
3889a803
PA
1355static inline bool skb_unref(struct sk_buff *skb)
1356{
1357 if (unlikely(!skb))
1358 return false;
63354797 1359 if (likely(refcount_read(&skb->users) == 1))
3889a803 1360 smp_rmb();
63354797 1361 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1362 return false;
1363
1364 return true;
1365}
1366
c504e5c2
MD
1367void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1368
1369/**
1370 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1371 * @skb: buffer to free
1372 */
1373static inline void kfree_skb(struct sk_buff *skb)
1374{
1375 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1376}
1377
0a463c78 1378void skb_release_head_state(struct sk_buff *skb);
215b0f19
MD
1379void kfree_skb_list_reason(struct sk_buff *segs,
1380 enum skb_drop_reason reason);
6413139d 1381void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1382void skb_tx_error(struct sk_buff *skb);
be769db2 1383
215b0f19
MD
1384static inline void kfree_skb_list(struct sk_buff *segs)
1385{
1386 kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED);
1387}
1388
be769db2 1389#ifdef CONFIG_TRACEPOINTS
7965bd4d 1390void consume_skb(struct sk_buff *skb);
be769db2
HX
1391#else
1392static inline void consume_skb(struct sk_buff *skb)
1393{
1394 return kfree_skb(skb);
1395}
1396#endif
1397
ca2c1418 1398void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1399void __kfree_skb(struct sk_buff *skb);
d7e8883c 1400extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1401
7965bd4d
JP
1402void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1403bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1404 bool *fragstolen, int *delta_truesize);
bad43ca8 1405
7965bd4d
JP
1406struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1407 int node);
2ea2f62c 1408struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1409struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1410struct sk_buff *build_skb_around(struct sk_buff *skb,
1411 void *data, unsigned int frag_size);
68822bdf 1412void skb_attempt_defer_free(struct sk_buff *skb);
161e6137 1413
f450d539
AL
1414struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1415
161e6137
PT
1416/**
1417 * alloc_skb - allocate a network buffer
1418 * @size: size to allocate
1419 * @priority: allocation mask
1420 *
1421 * This function is a convenient wrapper around __alloc_skb().
1422 */
d179cd12 1423static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1424 gfp_t priority)
d179cd12 1425{
564824b0 1426 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1427}
1428
2e4e4410
ED
1429struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1430 unsigned long data_len,
1431 int max_page_order,
1432 int *errcode,
1433 gfp_t gfp_mask);
da29e4b4 1434struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1435
d0bf4a9e
ED
1436/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1437struct sk_buff_fclones {
1438 struct sk_buff skb1;
1439
1440 struct sk_buff skb2;
1441
2638595a 1442 refcount_t fclone_ref;
d0bf4a9e
ED
1443};
1444
1445/**
1446 * skb_fclone_busy - check if fclone is busy
293de7de 1447 * @sk: socket
d0bf4a9e
ED
1448 * @skb: buffer
1449 *
bda13fed 1450 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1451 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1452 * so we also check that this didnt happen.
d0bf4a9e 1453 */
39bb5e62
ED
1454static inline bool skb_fclone_busy(const struct sock *sk,
1455 const struct sk_buff *skb)
d0bf4a9e
ED
1456{
1457 const struct sk_buff_fclones *fclones;
1458
1459 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1460
1461 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1462 refcount_read(&fclones->fclone_ref) > 1 &&
f4dae54e 1463 READ_ONCE(fclones->skb2.sk) == sk;
d0bf4a9e
ED
1464}
1465
161e6137
PT
1466/**
1467 * alloc_skb_fclone - allocate a network buffer from fclone cache
1468 * @size: size to allocate
1469 * @priority: allocation mask
1470 *
1471 * This function is a convenient wrapper around __alloc_skb().
1472 */
d179cd12 1473static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1474 gfp_t priority)
d179cd12 1475{
c93bdd0e 1476 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1477}
1478
7965bd4d 1479struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1480void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1481int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1482struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1483void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1484struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1485struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1486 gfp_t gfp_mask, bool fclone);
1487static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1488 gfp_t gfp_mask)
1489{
1490 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1491}
7965bd4d
JP
1492
1493int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1494struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1495 unsigned int headroom);
f1260ff1 1496struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
7965bd4d
JP
1497struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1498 int newtailroom, gfp_t priority);
48a1df65
JD
1499int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1500 int offset, int len);
1501int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1502 int offset, int len);
7965bd4d 1503int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1504int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1505
1506/**
1507 * skb_pad - zero pad the tail of an skb
1508 * @skb: buffer to pad
1509 * @pad: space to pad
1510 *
1511 * Ensure that a buffer is followed by a padding area that is zero
1512 * filled. Used by network drivers which may DMA or transfer data
1513 * beyond the buffer end onto the wire.
1514 *
1515 * May return error in out of memory cases. The skb is freed on error.
1516 */
1517static inline int skb_pad(struct sk_buff *skb, int pad)
1518{
1519 return __skb_pad(skb, pad, true);
1520}
ead2ceb0 1521#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1522
be12a1fe
HFS
1523int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1524 int offset, size_t size);
1525
d94d9fee 1526struct skb_seq_state {
677e90ed
TG
1527 __u32 lower_offset;
1528 __u32 upper_offset;
1529 __u32 frag_idx;
1530 __u32 stepped_offset;
1531 struct sk_buff *root_skb;
1532 struct sk_buff *cur_skb;
1533 __u8 *frag_data;
97550f6f 1534 __u32 frag_off;
677e90ed
TG
1535};
1536
7965bd4d
JP
1537void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1538 unsigned int to, struct skb_seq_state *st);
1539unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1540 struct skb_seq_state *st);
1541void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1542
7965bd4d 1543unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1544 unsigned int to, struct ts_config *config);
3fc7e8a6 1545
09323cc4
TH
1546/*
1547 * Packet hash types specify the type of hash in skb_set_hash.
1548 *
1549 * Hash types refer to the protocol layer addresses which are used to
1550 * construct a packet's hash. The hashes are used to differentiate or identify
1551 * flows of the protocol layer for the hash type. Hash types are either
1552 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1553 *
1554 * Properties of hashes:
1555 *
1556 * 1) Two packets in different flows have different hash values
1557 * 2) Two packets in the same flow should have the same hash value
1558 *
1559 * A hash at a higher layer is considered to be more specific. A driver should
1560 * set the most specific hash possible.
1561 *
1562 * A driver cannot indicate a more specific hash than the layer at which a hash
1563 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1564 *
1565 * A driver may indicate a hash level which is less specific than the
1566 * actual layer the hash was computed on. For instance, a hash computed
1567 * at L4 may be considered an L3 hash. This should only be done if the
1568 * driver can't unambiguously determine that the HW computed the hash at
1569 * the higher layer. Note that the "should" in the second property above
1570 * permits this.
1571 */
1572enum pkt_hash_types {
1573 PKT_HASH_TYPE_NONE, /* Undefined type */
1574 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1575 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1576 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1577};
1578
bcc83839 1579static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1580{
bcc83839 1581 skb->hash = 0;
a3b18ddb 1582 skb->sw_hash = 0;
bcc83839
TH
1583 skb->l4_hash = 0;
1584}
1585
1586static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1587{
1588 if (!skb->l4_hash)
1589 skb_clear_hash(skb);
1590}
1591
1592static inline void
1593__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1594{
1595 skb->l4_hash = is_l4;
1596 skb->sw_hash = is_sw;
61b905da 1597 skb->hash = hash;
09323cc4
TH
1598}
1599
bcc83839
TH
1600static inline void
1601skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1602{
1603 /* Used by drivers to set hash from HW */
1604 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1605}
1606
1607static inline void
1608__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1609{
1610 __skb_set_hash(skb, hash, true, is_l4);
1611}
1612
e5276937 1613void __skb_get_hash(struct sk_buff *skb);
b917783c 1614u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937 1615u32 skb_get_poff(const struct sk_buff *skb);
f96533cd 1616u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
72a338bc 1617 const struct flow_keys_basic *keys, int hlen);
e5276937 1618__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
f96533cd 1619 const void *data, int hlen_proto);
e5276937
TH
1620
1621static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1622 int thoff, u8 ip_proto)
1623{
1624 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1625}
1626
1627void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1628 const struct flow_dissector_key *key,
1629 unsigned int key_count);
1630
089b19a9
SF
1631struct bpf_flow_dissector;
1632bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1633 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1634
3cbf4ffb
SF
1635bool __skb_flow_dissect(const struct net *net,
1636 const struct sk_buff *skb,
e5276937 1637 struct flow_dissector *flow_dissector,
f96533cd
AL
1638 void *target_container, const void *data,
1639 __be16 proto, int nhoff, int hlen, unsigned int flags);
e5276937
TH
1640
1641static inline bool skb_flow_dissect(const struct sk_buff *skb,
1642 struct flow_dissector *flow_dissector,
cd79a238 1643 void *target_container, unsigned int flags)
e5276937 1644{
3cbf4ffb
SF
1645 return __skb_flow_dissect(NULL, skb, flow_dissector,
1646 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1647}
1648
1649static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1650 struct flow_keys *flow,
1651 unsigned int flags)
e5276937
TH
1652{
1653 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1654 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1655 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1656}
1657
72a338bc 1658static inline bool
3cbf4ffb
SF
1659skb_flow_dissect_flow_keys_basic(const struct net *net,
1660 const struct sk_buff *skb,
f96533cd
AL
1661 struct flow_keys_basic *flow,
1662 const void *data, __be16 proto,
1663 int nhoff, int hlen, unsigned int flags)
e5276937
TH
1664{
1665 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1666 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1667 data, proto, nhoff, hlen, flags);
e5276937
TH
1668}
1669
82828b88
JP
1670void skb_flow_dissect_meta(const struct sk_buff *skb,
1671 struct flow_dissector *flow_dissector,
1672 void *target_container);
1673
75a56758 1674/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1675 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1676 * to user states.
1677 */
1678void
1679skb_flow_dissect_ct(const struct sk_buff *skb,
1680 struct flow_dissector *flow_dissector,
1681 void *target_container,
7baf2429 1682 u16 *ctinfo_map, size_t mapsize,
38495958 1683 bool post_ct, u16 zone);
62b32379
SH
1684void
1685skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1686 struct flow_dissector *flow_dissector,
1687 void *target_container);
1688
0cb09aff
AL
1689void skb_flow_dissect_hash(const struct sk_buff *skb,
1690 struct flow_dissector *flow_dissector,
1691 void *target_container);
1692
3958afa1 1693static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1694{
a3b18ddb 1695 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1696 __skb_get_hash(skb);
bfb564e7 1697
61b905da 1698 return skb->hash;
bfb564e7
KK
1699}
1700
20a17bf6 1701static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1702{
c6cc1ca7
TH
1703 if (!skb->l4_hash && !skb->sw_hash) {
1704 struct flow_keys keys;
de4c1f8b 1705 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1706
de4c1f8b 1707 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1708 }
f70ea018
TH
1709
1710 return skb->hash;
1711}
1712
55667441
ED
1713__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1714 const siphash_key_t *perturb);
50fb7992 1715
57bdf7f4
TH
1716static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1717{
61b905da 1718 return skb->hash;
57bdf7f4
TH
1719}
1720
3df7a74e
TH
1721static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1722{
61b905da 1723 to->hash = from->hash;
a3b18ddb 1724 to->sw_hash = from->sw_hash;
61b905da 1725 to->l4_hash = from->l4_hash;
3df7a74e
TH
1726};
1727
41477662
JK
1728static inline void skb_copy_decrypted(struct sk_buff *to,
1729 const struct sk_buff *from)
1730{
1731#ifdef CONFIG_TLS_DEVICE
1732 to->decrypted = from->decrypted;
1733#endif
1734}
1735
4305b541
ACM
1736#ifdef NET_SKBUFF_DATA_USES_OFFSET
1737static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1738{
1739 return skb->head + skb->end;
1740}
ec47ea82
AD
1741
1742static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1743{
1744 return skb->end;
1745}
763087da
ED
1746
1747static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1748{
1749 skb->end = offset;
1750}
4305b541
ACM
1751#else
1752static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1753{
1754 return skb->end;
1755}
ec47ea82
AD
1756
1757static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1758{
1759 return skb->end - skb->head;
1760}
763087da
ED
1761
1762static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1763{
1764 skb->end = skb->head + offset;
1765}
4305b541
ACM
1766#endif
1767
657dd5f9
PB
1768struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
1769 struct ubuf_info *uarg);
1770
1771void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
1772
1773void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
1774 bool success);
1775
1776int __zerocopy_sg_from_iter(struct sock *sk, struct sk_buff *skb,
1777 struct iov_iter *from, size_t length);
1778
1779static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb,
1780 struct msghdr *msg, int len)
1781{
1782 return __zerocopy_sg_from_iter(skb->sk, skb, &msg->msg_iter, len);
1783}
1784
1785int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
1786 struct msghdr *msg, int len,
1787 struct ubuf_info *uarg);
1788
1da177e4 1789/* Internal */
4305b541 1790#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1791
ac45f602
PO
1792static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1793{
1794 return &skb_shinfo(skb)->hwtstamps;
1795}
1796
52267790
WB
1797static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1798{
06b4feb3 1799 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1800
1801 return is_zcopy ? skb_uarg(skb) : NULL;
1802}
1803
9b65b17d
TA
1804static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1805{
1806 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1807}
1808
1809static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1810 const struct sk_buff *skb2)
1811{
1812 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1813}
1814
8e044917 1815static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1816{
1817 refcount_inc(&uarg->refcnt);
1818}
1819
9ee5e5ad
JL
1820static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1821{
1822 skb_shinfo(skb)->destructor_arg = uarg;
1823 skb_shinfo(skb)->flags |= uarg->flags;
1824}
1825
52900d22
WB
1826static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1827 bool *have_ref)
52267790
WB
1828{
1829 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1830 if (unlikely(have_ref && *have_ref))
1831 *have_ref = false;
1832 else
8e044917 1833 net_zcopy_get(uarg);
9ee5e5ad 1834 skb_zcopy_init(skb, uarg);
52267790
WB
1835 }
1836}
1837
5cd8d46e
WB
1838static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1839{
1840 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1841 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1842}
1843
1844static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1845{
1846 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1847}
1848
1849static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1850{
1851 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1852}
1853
8e044917 1854static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1855{
1856 if (uarg)
36177832 1857 uarg->callback(NULL, uarg, true);
59776362
JL
1858}
1859
8e044917 1860static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1861{
1862 if (uarg) {
8c793822
JL
1863 if (uarg->callback == msg_zerocopy_callback)
1864 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1865 else if (have_uref)
8e044917 1866 net_zcopy_put(uarg);
236a6b1c
JL
1867 }
1868}
1869
52267790 1870/* Release a reference on a zerocopy structure */
36177832 1871static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1872{
1873 struct ubuf_info *uarg = skb_zcopy(skb);
1874
1875 if (uarg) {
36177832
JL
1876 if (!skb_zcopy_is_nouarg(skb))
1877 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1878
9b65b17d 1879 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
52267790
WB
1880 }
1881}
1882
a8305bff
DM
1883static inline void skb_mark_not_on_list(struct sk_buff *skb)
1884{
1885 skb->next = NULL;
1886}
1887
dcfea72e 1888/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1889#define skb_list_walk_safe(first, skb, next_skb) \
1890 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1891 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1892
992cba7e
DM
1893static inline void skb_list_del_init(struct sk_buff *skb)
1894{
1895 __list_del_entry(&skb->list);
1896 skb_mark_not_on_list(skb);
1897}
1898
1da177e4
LT
1899/**
1900 * skb_queue_empty - check if a queue is empty
1901 * @list: queue head
1902 *
1903 * Returns true if the queue is empty, false otherwise.
1904 */
1905static inline int skb_queue_empty(const struct sk_buff_head *list)
1906{
fd44b93c 1907 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1908}
1909
d7d16a89
ED
1910/**
1911 * skb_queue_empty_lockless - check if a queue is empty
1912 * @list: queue head
1913 *
1914 * Returns true if the queue is empty, false otherwise.
1915 * This variant can be used in lockless contexts.
1916 */
1917static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1918{
1919 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1920}
1921
1922
fc7ebb21
DM
1923/**
1924 * skb_queue_is_last - check if skb is the last entry in the queue
1925 * @list: queue head
1926 * @skb: buffer
1927 *
1928 * Returns true if @skb is the last buffer on the list.
1929 */
1930static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1931 const struct sk_buff *skb)
1932{
fd44b93c 1933 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1934}
1935
832d11c5
IJ
1936/**
1937 * skb_queue_is_first - check if skb is the first entry in the queue
1938 * @list: queue head
1939 * @skb: buffer
1940 *
1941 * Returns true if @skb is the first buffer on the list.
1942 */
1943static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1944 const struct sk_buff *skb)
1945{
fd44b93c 1946 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1947}
1948
249c8b42
DM
1949/**
1950 * skb_queue_next - return the next packet in the queue
1951 * @list: queue head
1952 * @skb: current buffer
1953 *
1954 * Return the next packet in @list after @skb. It is only valid to
1955 * call this if skb_queue_is_last() evaluates to false.
1956 */
1957static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1958 const struct sk_buff *skb)
1959{
1960 /* This BUG_ON may seem severe, but if we just return then we
1961 * are going to dereference garbage.
1962 */
1963 BUG_ON(skb_queue_is_last(list, skb));
1964 return skb->next;
1965}
1966
832d11c5
IJ
1967/**
1968 * skb_queue_prev - return the prev packet in the queue
1969 * @list: queue head
1970 * @skb: current buffer
1971 *
1972 * Return the prev packet in @list before @skb. It is only valid to
1973 * call this if skb_queue_is_first() evaluates to false.
1974 */
1975static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1976 const struct sk_buff *skb)
1977{
1978 /* This BUG_ON may seem severe, but if we just return then we
1979 * are going to dereference garbage.
1980 */
1981 BUG_ON(skb_queue_is_first(list, skb));
1982 return skb->prev;
1983}
1984
1da177e4
LT
1985/**
1986 * skb_get - reference buffer
1987 * @skb: buffer to reference
1988 *
1989 * Makes another reference to a socket buffer and returns a pointer
1990 * to the buffer.
1991 */
1992static inline struct sk_buff *skb_get(struct sk_buff *skb)
1993{
63354797 1994 refcount_inc(&skb->users);
1da177e4
LT
1995 return skb;
1996}
1997
1998/*
f8821f96 1999 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
2000 */
2001
1da177e4
LT
2002/**
2003 * skb_cloned - is the buffer a clone
2004 * @skb: buffer to check
2005 *
2006 * Returns true if the buffer was generated with skb_clone() and is
2007 * one of multiple shared copies of the buffer. Cloned buffers are
2008 * shared data so must not be written to under normal circumstances.
2009 */
2010static inline int skb_cloned(const struct sk_buff *skb)
2011{
2012 return skb->cloned &&
2013 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
2014}
2015
14bbd6a5
PS
2016static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
2017{
d0164adc 2018 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
2019
2020 if (skb_cloned(skb))
2021 return pskb_expand_head(skb, 0, 0, pri);
2022
2023 return 0;
2024}
2025
2b88cba5
ED
2026/* This variant of skb_unclone() makes sure skb->truesize
2027 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
2028 *
2029 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
2030 * when various debugging features are in place.
2031 */
2032int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
c4777efa
ED
2033static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
2034{
2035 might_sleep_if(gfpflags_allow_blocking(pri));
2036
2b88cba5
ED
2037 if (skb_cloned(skb))
2038 return __skb_unclone_keeptruesize(skb, pri);
c4777efa
ED
2039 return 0;
2040}
2041
1da177e4
LT
2042/**
2043 * skb_header_cloned - is the header a clone
2044 * @skb: buffer to check
2045 *
2046 * Returns true if modifying the header part of the buffer requires
2047 * the data to be copied.
2048 */
2049static inline int skb_header_cloned(const struct sk_buff *skb)
2050{
2051 int dataref;
2052
2053 if (!skb->cloned)
2054 return 0;
2055
2056 dataref = atomic_read(&skb_shinfo(skb)->dataref);
2057 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
2058 return dataref != 1;
2059}
2060
9580bf2e
ED
2061static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
2062{
2063 might_sleep_if(gfpflags_allow_blocking(pri));
2064
2065 if (skb_header_cloned(skb))
2066 return pskb_expand_head(skb, 0, 0, pri);
2067
2068 return 0;
2069}
2070
f4a775d1 2071/**
9ec7ea14
JK
2072 * __skb_header_release() - allow clones to use the headroom
2073 * @skb: buffer to operate on
2074 *
2075 * See "DOC: dataref and headerless skbs".
f4a775d1
ED
2076 */
2077static inline void __skb_header_release(struct sk_buff *skb)
2078{
2079 skb->nohdr = 1;
2080 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
2081}
2082
2083
1da177e4
LT
2084/**
2085 * skb_shared - is the buffer shared
2086 * @skb: buffer to check
2087 *
2088 * Returns true if more than one person has a reference to this
2089 * buffer.
2090 */
2091static inline int skb_shared(const struct sk_buff *skb)
2092{
63354797 2093 return refcount_read(&skb->users) != 1;
1da177e4
LT
2094}
2095
2096/**
2097 * skb_share_check - check if buffer is shared and if so clone it
2098 * @skb: buffer to check
2099 * @pri: priority for memory allocation
2100 *
2101 * If the buffer is shared the buffer is cloned and the old copy
2102 * drops a reference. A new clone with a single reference is returned.
2103 * If the buffer is not shared the original buffer is returned. When
2104 * being called from interrupt status or with spinlocks held pri must
2105 * be GFP_ATOMIC.
2106 *
2107 * NULL is returned on a memory allocation failure.
2108 */
47061bc4 2109static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 2110{
d0164adc 2111 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
2112 if (skb_shared(skb)) {
2113 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
2114
2115 if (likely(nskb))
2116 consume_skb(skb);
2117 else
2118 kfree_skb(skb);
1da177e4
LT
2119 skb = nskb;
2120 }
2121 return skb;
2122}
2123
2124/*
2125 * Copy shared buffers into a new sk_buff. We effectively do COW on
2126 * packets to handle cases where we have a local reader and forward
2127 * and a couple of other messy ones. The normal one is tcpdumping
2128 * a packet thats being forwarded.
2129 */
2130
2131/**
2132 * skb_unshare - make a copy of a shared buffer
2133 * @skb: buffer to check
2134 * @pri: priority for memory allocation
2135 *
2136 * If the socket buffer is a clone then this function creates a new
2137 * copy of the data, drops a reference count on the old copy and returns
2138 * the new copy with the reference count at 1. If the buffer is not a clone
2139 * the original buffer is returned. When called with a spinlock held or
2140 * from interrupt state @pri must be %GFP_ATOMIC
2141 *
2142 * %NULL is returned on a memory allocation failure.
2143 */
e2bf521d 2144static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 2145 gfp_t pri)
1da177e4 2146{
d0164adc 2147 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
2148 if (skb_cloned(skb)) {
2149 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
2150
2151 /* Free our shared copy */
2152 if (likely(nskb))
2153 consume_skb(skb);
2154 else
2155 kfree_skb(skb);
1da177e4
LT
2156 skb = nskb;
2157 }
2158 return skb;
2159}
2160
2161/**
1a5778aa 2162 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
2163 * @list_: list to peek at
2164 *
2165 * Peek an &sk_buff. Unlike most other operations you _MUST_
2166 * be careful with this one. A peek leaves the buffer on the
2167 * list and someone else may run off with it. You must hold
2168 * the appropriate locks or have a private queue to do this.
2169 *
2170 * Returns %NULL for an empty list or a pointer to the head element.
2171 * The reference count is not incremented and the reference is therefore
2172 * volatile. Use with caution.
2173 */
05bdd2f1 2174static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 2175{
18d07000
ED
2176 struct sk_buff *skb = list_->next;
2177
2178 if (skb == (struct sk_buff *)list_)
2179 skb = NULL;
2180 return skb;
1da177e4
LT
2181}
2182
8b69bd7d
DM
2183/**
2184 * __skb_peek - peek at the head of a non-empty &sk_buff_head
2185 * @list_: list to peek at
2186 *
2187 * Like skb_peek(), but the caller knows that the list is not empty.
2188 */
2189static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
2190{
2191 return list_->next;
2192}
2193
da5ef6e5
PE
2194/**
2195 * skb_peek_next - peek skb following the given one from a queue
2196 * @skb: skb to start from
2197 * @list_: list to peek at
2198 *
2199 * Returns %NULL when the end of the list is met or a pointer to the
2200 * next element. The reference count is not incremented and the
2201 * reference is therefore volatile. Use with caution.
2202 */
2203static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
2204 const struct sk_buff_head *list_)
2205{
2206 struct sk_buff *next = skb->next;
18d07000 2207
da5ef6e5
PE
2208 if (next == (struct sk_buff *)list_)
2209 next = NULL;
2210 return next;
2211}
2212
1da177e4 2213/**
1a5778aa 2214 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
2215 * @list_: list to peek at
2216 *
2217 * Peek an &sk_buff. Unlike most other operations you _MUST_
2218 * be careful with this one. A peek leaves the buffer on the
2219 * list and someone else may run off with it. You must hold
2220 * the appropriate locks or have a private queue to do this.
2221 *
2222 * Returns %NULL for an empty list or a pointer to the tail element.
2223 * The reference count is not incremented and the reference is therefore
2224 * volatile. Use with caution.
2225 */
05bdd2f1 2226static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 2227{
f8cc62ca 2228 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
2229
2230 if (skb == (struct sk_buff *)list_)
2231 skb = NULL;
2232 return skb;
2233
1da177e4
LT
2234}
2235
2236/**
2237 * skb_queue_len - get queue length
2238 * @list_: list to measure
2239 *
2240 * Return the length of an &sk_buff queue.
2241 */
2242static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
2243{
2244 return list_->qlen;
2245}
2246
86b18aaa
QC
2247/**
2248 * skb_queue_len_lockless - get queue length
2249 * @list_: list to measure
2250 *
2251 * Return the length of an &sk_buff queue.
2252 * This variant can be used in lockless contexts.
2253 */
2254static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
2255{
2256 return READ_ONCE(list_->qlen);
2257}
2258
67fed459
DM
2259/**
2260 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
2261 * @list: queue to initialize
2262 *
2263 * This initializes only the list and queue length aspects of
2264 * an sk_buff_head object. This allows to initialize the list
2265 * aspects of an sk_buff_head without reinitializing things like
2266 * the spinlock. It can also be used for on-stack sk_buff_head
2267 * objects where the spinlock is known to not be used.
2268 */
2269static inline void __skb_queue_head_init(struct sk_buff_head *list)
2270{
2271 list->prev = list->next = (struct sk_buff *)list;
2272 list->qlen = 0;
2273}
2274
76f10ad0
AV
2275/*
2276 * This function creates a split out lock class for each invocation;
2277 * this is needed for now since a whole lot of users of the skb-queue
2278 * infrastructure in drivers have different locking usage (in hardirq)
2279 * than the networking core (in softirq only). In the long run either the
2280 * network layer or drivers should need annotation to consolidate the
2281 * main types of usage into 3 classes.
2282 */
1da177e4
LT
2283static inline void skb_queue_head_init(struct sk_buff_head *list)
2284{
2285 spin_lock_init(&list->lock);
67fed459 2286 __skb_queue_head_init(list);
1da177e4
LT
2287}
2288
c2ecba71
PE
2289static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2290 struct lock_class_key *class)
2291{
2292 skb_queue_head_init(list);
2293 lockdep_set_class(&list->lock, class);
2294}
2295
1da177e4 2296/*
bf299275 2297 * Insert an sk_buff on a list.
1da177e4
LT
2298 *
2299 * The "__skb_xxxx()" functions are the non-atomic ones that
2300 * can only be called with interrupts disabled.
2301 */
bf299275
GR
2302static inline void __skb_insert(struct sk_buff *newsk,
2303 struct sk_buff *prev, struct sk_buff *next,
2304 struct sk_buff_head *list)
2305{
f8cc62ca
ED
2306 /* See skb_queue_empty_lockless() and skb_peek_tail()
2307 * for the opposite READ_ONCE()
2308 */
d7d16a89
ED
2309 WRITE_ONCE(newsk->next, next);
2310 WRITE_ONCE(newsk->prev, prev);
1a2fb220
KC
2311 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2312 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
04f08eb4 2313 WRITE_ONCE(list->qlen, list->qlen + 1);
bf299275 2314}
1da177e4 2315
67fed459
DM
2316static inline void __skb_queue_splice(const struct sk_buff_head *list,
2317 struct sk_buff *prev,
2318 struct sk_buff *next)
2319{
2320 struct sk_buff *first = list->next;
2321 struct sk_buff *last = list->prev;
2322
d7d16a89
ED
2323 WRITE_ONCE(first->prev, prev);
2324 WRITE_ONCE(prev->next, first);
67fed459 2325
d7d16a89
ED
2326 WRITE_ONCE(last->next, next);
2327 WRITE_ONCE(next->prev, last);
67fed459
DM
2328}
2329
2330/**
2331 * skb_queue_splice - join two skb lists, this is designed for stacks
2332 * @list: the new list to add
2333 * @head: the place to add it in the first list
2334 */
2335static inline void skb_queue_splice(const struct sk_buff_head *list,
2336 struct sk_buff_head *head)
2337{
2338 if (!skb_queue_empty(list)) {
2339 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 2340 head->qlen += list->qlen;
67fed459
DM
2341 }
2342}
2343
2344/**
d9619496 2345 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2346 * @list: the new list to add
2347 * @head: the place to add it in the first list
2348 *
2349 * The list at @list is reinitialised
2350 */
2351static inline void skb_queue_splice_init(struct sk_buff_head *list,
2352 struct sk_buff_head *head)
2353{
2354 if (!skb_queue_empty(list)) {
2355 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 2356 head->qlen += list->qlen;
67fed459
DM
2357 __skb_queue_head_init(list);
2358 }
2359}
2360
2361/**
2362 * skb_queue_splice_tail - join two skb lists, each list being a queue
2363 * @list: the new list to add
2364 * @head: the place to add it in the first list
2365 */
2366static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2367 struct sk_buff_head *head)
2368{
2369 if (!skb_queue_empty(list)) {
2370 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2371 head->qlen += list->qlen;
67fed459
DM
2372 }
2373}
2374
2375/**
d9619496 2376 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2377 * @list: the new list to add
2378 * @head: the place to add it in the first list
2379 *
2380 * Each of the lists is a queue.
2381 * The list at @list is reinitialised
2382 */
2383static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2384 struct sk_buff_head *head)
2385{
2386 if (!skb_queue_empty(list)) {
2387 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2388 head->qlen += list->qlen;
67fed459
DM
2389 __skb_queue_head_init(list);
2390 }
2391}
2392
1da177e4 2393/**
300ce174 2394 * __skb_queue_after - queue a buffer at the list head
1da177e4 2395 * @list: list to use
300ce174 2396 * @prev: place after this buffer
1da177e4
LT
2397 * @newsk: buffer to queue
2398 *
300ce174 2399 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2400 * and you must therefore hold required locks before calling it.
2401 *
2402 * A buffer cannot be placed on two lists at the same time.
2403 */
300ce174
SH
2404static inline void __skb_queue_after(struct sk_buff_head *list,
2405 struct sk_buff *prev,
2406 struct sk_buff *newsk)
1da177e4 2407{
1a2fb220 2408 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
1da177e4
LT
2409}
2410
7965bd4d
JP
2411void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2412 struct sk_buff_head *list);
7de6c033 2413
f5572855
GR
2414static inline void __skb_queue_before(struct sk_buff_head *list,
2415 struct sk_buff *next,
2416 struct sk_buff *newsk)
2417{
1a2fb220 2418 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
f5572855
GR
2419}
2420
300ce174
SH
2421/**
2422 * __skb_queue_head - queue a buffer at the list head
2423 * @list: list to use
2424 * @newsk: buffer to queue
2425 *
2426 * Queue a buffer at the start of a list. This function takes no locks
2427 * and you must therefore hold required locks before calling it.
2428 *
2429 * A buffer cannot be placed on two lists at the same time.
2430 */
300ce174
SH
2431static inline void __skb_queue_head(struct sk_buff_head *list,
2432 struct sk_buff *newsk)
2433{
2434 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2435}
4ea7b0cf 2436void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2437
1da177e4
LT
2438/**
2439 * __skb_queue_tail - queue a buffer at the list tail
2440 * @list: list to use
2441 * @newsk: buffer to queue
2442 *
2443 * Queue a buffer at the end of a list. This function takes no locks
2444 * and you must therefore hold required locks before calling it.
2445 *
2446 * A buffer cannot be placed on two lists at the same time.
2447 */
1da177e4
LT
2448static inline void __skb_queue_tail(struct sk_buff_head *list,
2449 struct sk_buff *newsk)
2450{
f5572855 2451 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2452}
4ea7b0cf 2453void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2454
1da177e4
LT
2455/*
2456 * remove sk_buff from list. _Must_ be called atomically, and with
2457 * the list known..
2458 */
7965bd4d 2459void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2460static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2461{
2462 struct sk_buff *next, *prev;
2463
86b18aaa 2464 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2465 next = skb->next;
2466 prev = skb->prev;
2467 skb->next = skb->prev = NULL;
d7d16a89
ED
2468 WRITE_ONCE(next->prev, prev);
2469 WRITE_ONCE(prev->next, next);
1da177e4
LT
2470}
2471
f525c06d
GR
2472/**
2473 * __skb_dequeue - remove from the head of the queue
2474 * @list: list to dequeue from
2475 *
2476 * Remove the head of the list. This function does not take any locks
2477 * so must be used with appropriate locks held only. The head item is
2478 * returned or %NULL if the list is empty.
2479 */
f525c06d
GR
2480static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2481{
2482 struct sk_buff *skb = skb_peek(list);
2483 if (skb)
2484 __skb_unlink(skb, list);
2485 return skb;
2486}
4ea7b0cf 2487struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2488
2489/**
2490 * __skb_dequeue_tail - remove from the tail of the queue
2491 * @list: list to dequeue from
2492 *
2493 * Remove the tail of the list. This function does not take any locks
2494 * so must be used with appropriate locks held only. The tail item is
2495 * returned or %NULL if the list is empty.
2496 */
1da177e4
LT
2497static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2498{
2499 struct sk_buff *skb = skb_peek_tail(list);
2500 if (skb)
2501 __skb_unlink(skb, list);
2502 return skb;
2503}
4ea7b0cf 2504struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2505
2506
bdcc0924 2507static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2508{
2509 return skb->data_len;
2510}
2511
2512static inline unsigned int skb_headlen(const struct sk_buff *skb)
2513{
2514 return skb->len - skb->data_len;
2515}
2516
3ece7826 2517static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2518{
c72d8cda 2519 unsigned int i, len = 0;
1da177e4 2520
c72d8cda 2521 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2522 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2523 return len;
2524}
2525
2526static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2527{
2528 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2529}
2530
131ea667
IC
2531/**
2532 * __skb_fill_page_desc - initialise a paged fragment in an skb
2533 * @skb: buffer containing fragment to be initialised
2534 * @i: paged fragment index to initialise
2535 * @page: the page to use for this fragment
2536 * @off: the offset to the data with @page
2537 * @size: the length of the data
2538 *
2539 * Initialises the @i'th fragment of @skb to point to &size bytes at
2540 * offset @off within @page.
2541 *
2542 * Does not take any additional reference on the fragment.
2543 */
2544static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2545 struct page *page, int off, int size)
1da177e4
LT
2546{
2547 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2548
c48a11c7 2549 /*
2f064f34
MH
2550 * Propagate page pfmemalloc to the skb if we can. The problem is
2551 * that not all callers have unique ownership of the page but rely
2552 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2553 */
1dfa5bd3 2554 frag->bv_page = page;
65c84f14 2555 frag->bv_offset = off;
9e903e08 2556 skb_frag_size_set(frag, size);
cca7af38
PE
2557
2558 page = compound_head(page);
2f064f34 2559 if (page_is_pfmemalloc(page))
cca7af38 2560 skb->pfmemalloc = true;
131ea667
IC
2561}
2562
2563/**
2564 * skb_fill_page_desc - initialise a paged fragment in an skb
2565 * @skb: buffer containing fragment to be initialised
2566 * @i: paged fragment index to initialise
2567 * @page: the page to use for this fragment
2568 * @off: the offset to the data with @page
2569 * @size: the length of the data
2570 *
2571 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2572 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2573 * addition updates @skb such that @i is the last fragment.
2574 *
2575 * Does not take any additional reference on the fragment.
2576 */
2577static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2578 struct page *page, int off, int size)
2579{
2580 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2581 skb_shinfo(skb)->nr_frags = i + 1;
2582}
2583
7965bd4d
JP
2584void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2585 int size, unsigned int truesize);
654bed16 2586
f8e617e1
JW
2587void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2588 unsigned int truesize);
2589
1da177e4
LT
2590#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2591
27a884dc
ACM
2592#ifdef NET_SKBUFF_DATA_USES_OFFSET
2593static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2594{
2595 return skb->head + skb->tail;
2596}
2597
2598static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2599{
2600 skb->tail = skb->data - skb->head;
2601}
2602
2603static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2604{
2605 skb_reset_tail_pointer(skb);
2606 skb->tail += offset;
2607}
7cc46190 2608
27a884dc
ACM
2609#else /* NET_SKBUFF_DATA_USES_OFFSET */
2610static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2611{
2612 return skb->tail;
2613}
2614
2615static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2616{
2617 skb->tail = skb->data;
2618}
2619
2620static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2621{
2622 skb->tail = skb->data + offset;
2623}
4305b541 2624
27a884dc
ACM
2625#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2626
1da177e4
LT
2627/*
2628 * Add data to an sk_buff
2629 */
4df864c1
JB
2630void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2631void *skb_put(struct sk_buff *skb, unsigned int len);
2632static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2633{
4df864c1 2634 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2635 SKB_LINEAR_ASSERT(skb);
2636 skb->tail += len;
2637 skb->len += len;
2638 return tmp;
2639}
2640
de77b966 2641static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2642{
2643 void *tmp = __skb_put(skb, len);
2644
2645 memset(tmp, 0, len);
2646 return tmp;
2647}
2648
2649static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2650 unsigned int len)
2651{
2652 void *tmp = __skb_put(skb, len);
2653
2654 memcpy(tmp, data, len);
2655 return tmp;
2656}
2657
2658static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2659{
2660 *(u8 *)__skb_put(skb, 1) = val;
2661}
2662
83ad357d 2663static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2664{
83ad357d 2665 void *tmp = skb_put(skb, len);
e45a79da
JB
2666
2667 memset(tmp, 0, len);
2668
2669 return tmp;
2670}
2671
59ae1d12
JB
2672static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2673 unsigned int len)
2674{
2675 void *tmp = skb_put(skb, len);
2676
2677 memcpy(tmp, data, len);
2678
2679 return tmp;
2680}
2681
634fef61
JB
2682static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2683{
2684 *(u8 *)skb_put(skb, 1) = val;
2685}
2686
d58ff351
JB
2687void *skb_push(struct sk_buff *skb, unsigned int len);
2688static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2689{
2690 skb->data -= len;
2691 skb->len += len;
2692 return skb->data;
2693}
2694
af72868b
JB
2695void *skb_pull(struct sk_buff *skb, unsigned int len);
2696static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2697{
2698 skb->len -= len;
2699 BUG_ON(skb->len < skb->data_len);
2700 return skb->data += len;
2701}
2702
af72868b 2703static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2704{
2705 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2706}
2707
13244ccc
LAD
2708void *skb_pull_data(struct sk_buff *skb, size_t len);
2709
af72868b 2710void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2711
af72868b 2712static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2713{
2714 if (len > skb_headlen(skb) &&
987c402a 2715 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2716 return NULL;
2717 skb->len -= len;
2718 return skb->data += len;
2719}
2720
af72868b 2721static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2722{
2723 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2724}
2725
b9df4fd7 2726static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2727{
2728 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2729 return true;
1da177e4 2730 if (unlikely(len > skb->len))
b9df4fd7 2731 return false;
987c402a 2732 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2733}
2734
c8c8b127
ED
2735void skb_condense(struct sk_buff *skb);
2736
1da177e4
LT
2737/**
2738 * skb_headroom - bytes at buffer head
2739 * @skb: buffer to check
2740 *
2741 * Return the number of bytes of free space at the head of an &sk_buff.
2742 */
c2636b4d 2743static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2744{
2745 return skb->data - skb->head;
2746}
2747
2748/**
2749 * skb_tailroom - bytes at buffer end
2750 * @skb: buffer to check
2751 *
2752 * Return the number of bytes of free space at the tail of an sk_buff
2753 */
2754static inline int skb_tailroom(const struct sk_buff *skb)
2755{
4305b541 2756 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2757}
2758
a21d4572
ED
2759/**
2760 * skb_availroom - bytes at buffer end
2761 * @skb: buffer to check
2762 *
2763 * Return the number of bytes of free space at the tail of an sk_buff
2764 * allocated by sk_stream_alloc()
2765 */
2766static inline int skb_availroom(const struct sk_buff *skb)
2767{
16fad69c
ED
2768 if (skb_is_nonlinear(skb))
2769 return 0;
2770
2771 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2772}
2773
1da177e4
LT
2774/**
2775 * skb_reserve - adjust headroom
2776 * @skb: buffer to alter
2777 * @len: bytes to move
2778 *
2779 * Increase the headroom of an empty &sk_buff by reducing the tail
2780 * room. This is only allowed for an empty buffer.
2781 */
8243126c 2782static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2783{
2784 skb->data += len;
2785 skb->tail += len;
2786}
2787
1837b2e2
BP
2788/**
2789 * skb_tailroom_reserve - adjust reserved_tailroom
2790 * @skb: buffer to alter
2791 * @mtu: maximum amount of headlen permitted
2792 * @needed_tailroom: minimum amount of reserved_tailroom
2793 *
2794 * Set reserved_tailroom so that headlen can be as large as possible but
2795 * not larger than mtu and tailroom cannot be smaller than
2796 * needed_tailroom.
2797 * The required headroom should already have been reserved before using
2798 * this function.
2799 */
2800static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2801 unsigned int needed_tailroom)
2802{
2803 SKB_LINEAR_ASSERT(skb);
2804 if (mtu < skb_tailroom(skb) - needed_tailroom)
2805 /* use at most mtu */
2806 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2807 else
2808 /* use up to all available space */
2809 skb->reserved_tailroom = needed_tailroom;
2810}
2811
8bce6d7d
TH
2812#define ENCAP_TYPE_ETHER 0
2813#define ENCAP_TYPE_IPPROTO 1
2814
2815static inline void skb_set_inner_protocol(struct sk_buff *skb,
2816 __be16 protocol)
2817{
2818 skb->inner_protocol = protocol;
2819 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2820}
2821
2822static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2823 __u8 ipproto)
2824{
2825 skb->inner_ipproto = ipproto;
2826 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2827}
2828
6a674e9c
JG
2829static inline void skb_reset_inner_headers(struct sk_buff *skb)
2830{
aefbd2b3 2831 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2832 skb->inner_network_header = skb->network_header;
2833 skb->inner_transport_header = skb->transport_header;
2834}
2835
0b5c9db1
JP
2836static inline void skb_reset_mac_len(struct sk_buff *skb)
2837{
2838 skb->mac_len = skb->network_header - skb->mac_header;
2839}
2840
6a674e9c
JG
2841static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2842 *skb)
2843{
2844 return skb->head + skb->inner_transport_header;
2845}
2846
55dc5a9f
TH
2847static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2848{
2849 return skb_inner_transport_header(skb) - skb->data;
2850}
2851
6a674e9c
JG
2852static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2853{
2854 skb->inner_transport_header = skb->data - skb->head;
2855}
2856
2857static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2858 const int offset)
2859{
2860 skb_reset_inner_transport_header(skb);
2861 skb->inner_transport_header += offset;
2862}
2863
2864static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2865{
2866 return skb->head + skb->inner_network_header;
2867}
2868
2869static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2870{
2871 skb->inner_network_header = skb->data - skb->head;
2872}
2873
2874static inline void skb_set_inner_network_header(struct sk_buff *skb,
2875 const int offset)
2876{
2877 skb_reset_inner_network_header(skb);
2878 skb->inner_network_header += offset;
2879}
2880
aefbd2b3
PS
2881static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2882{
2883 return skb->head + skb->inner_mac_header;
2884}
2885
2886static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2887{
2888 skb->inner_mac_header = skb->data - skb->head;
2889}
2890
2891static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2892 const int offset)
2893{
2894 skb_reset_inner_mac_header(skb);
2895 skb->inner_mac_header += offset;
2896}
fda55eca
ED
2897static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2898{
35d04610 2899 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2900}
2901
9c70220b
ACM
2902static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2903{
66e4c8d9 2904 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
2e07fa9c 2905 return skb->head + skb->transport_header;
9c70220b
ACM
2906}
2907
badff6d0
ACM
2908static inline void skb_reset_transport_header(struct sk_buff *skb)
2909{
2e07fa9c 2910 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2911}
2912
967b05f6
ACM
2913static inline void skb_set_transport_header(struct sk_buff *skb,
2914 const int offset)
2915{
2e07fa9c
ACM
2916 skb_reset_transport_header(skb);
2917 skb->transport_header += offset;
ea2ae17d
ACM
2918}
2919
d56f90a7
ACM
2920static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2921{
2e07fa9c 2922 return skb->head + skb->network_header;
d56f90a7
ACM
2923}
2924
c1d2bbe1
ACM
2925static inline void skb_reset_network_header(struct sk_buff *skb)
2926{
2e07fa9c 2927 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2928}
2929
c14d2450
ACM
2930static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2931{
2e07fa9c
ACM
2932 skb_reset_network_header(skb);
2933 skb->network_header += offset;
c14d2450
ACM
2934}
2935
2e07fa9c 2936static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2937{
2e07fa9c 2938 return skb->head + skb->mac_header;
bbe735e4
ACM
2939}
2940
ea6da4fd
AV
2941static inline int skb_mac_offset(const struct sk_buff *skb)
2942{
2943 return skb_mac_header(skb) - skb->data;
2944}
2945
0daf4349
DB
2946static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2947{
2948 return skb->network_header - skb->mac_header;
2949}
2950
2e07fa9c 2951static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2952{
35d04610 2953 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2954}
2955
b4ab3141
DB
2956static inline void skb_unset_mac_header(struct sk_buff *skb)
2957{
2958 skb->mac_header = (typeof(skb->mac_header))~0U;
2959}
2960
2e07fa9c
ACM
2961static inline void skb_reset_mac_header(struct sk_buff *skb)
2962{
2963 skb->mac_header = skb->data - skb->head;
2964}
2965
2966static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2967{
2968 skb_reset_mac_header(skb);
2969 skb->mac_header += offset;
2970}
2971
0e3da5bb
TT
2972static inline void skb_pop_mac_header(struct sk_buff *skb)
2973{
2974 skb->mac_header = skb->network_header;
2975}
2976
d2aa125d 2977static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2978{
72a338bc 2979 struct flow_keys_basic keys;
fbbdb8f0
YX
2980
2981 if (skb_transport_header_was_set(skb))
2982 return;
72a338bc 2983
3cbf4ffb
SF
2984 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2985 NULL, 0, 0, 0, 0))
42aecaa9 2986 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2987}
2988
03606895
ED
2989static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2990{
2991 if (skb_mac_header_was_set(skb)) {
2992 const unsigned char *old_mac = skb_mac_header(skb);
2993
2994 skb_set_mac_header(skb, -skb->mac_len);
2995 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2996 }
2997}
2998
04fb451e
MM
2999static inline int skb_checksum_start_offset(const struct sk_buff *skb)
3000{
3001 return skb->csum_start - skb_headroom(skb);
3002}
3003
08b64fcc
AD
3004static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
3005{
3006 return skb->head + skb->csum_start;
3007}
3008
2e07fa9c
ACM
3009static inline int skb_transport_offset(const struct sk_buff *skb)
3010{
3011 return skb_transport_header(skb) - skb->data;
3012}
3013
3014static inline u32 skb_network_header_len(const struct sk_buff *skb)
3015{
3016 return skb->transport_header - skb->network_header;
3017}
3018
6a674e9c
JG
3019static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
3020{
3021 return skb->inner_transport_header - skb->inner_network_header;
3022}
3023
2e07fa9c
ACM
3024static inline int skb_network_offset(const struct sk_buff *skb)
3025{
3026 return skb_network_header(skb) - skb->data;
3027}
48d49d0c 3028
6a674e9c
JG
3029static inline int skb_inner_network_offset(const struct sk_buff *skb)
3030{
3031 return skb_inner_network_header(skb) - skb->data;
3032}
3033
f9599ce1
CG
3034static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
3035{
3036 return pskb_may_pull(skb, skb_network_offset(skb) + len);
3037}
3038
1da177e4
LT
3039/*
3040 * CPUs often take a performance hit when accessing unaligned memory
3041 * locations. The actual performance hit varies, it can be small if the
3042 * hardware handles it or large if we have to take an exception and fix it
3043 * in software.
3044 *
3045 * Since an ethernet header is 14 bytes network drivers often end up with
3046 * the IP header at an unaligned offset. The IP header can be aligned by
3047 * shifting the start of the packet by 2 bytes. Drivers should do this
3048 * with:
3049 *
8660c124 3050 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
3051 *
3052 * The downside to this alignment of the IP header is that the DMA is now
3053 * unaligned. On some architectures the cost of an unaligned DMA is high
3054 * and this cost outweighs the gains made by aligning the IP header.
8660c124 3055 *
1da177e4
LT
3056 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
3057 * to be overridden.
3058 */
3059#ifndef NET_IP_ALIGN
3060#define NET_IP_ALIGN 2
3061#endif
3062
025be81e
AB
3063/*
3064 * The networking layer reserves some headroom in skb data (via
3065 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
3066 * the header has to grow. In the default case, if the header has to grow
d6301d3d 3067 * 32 bytes or less we avoid the reallocation.
025be81e
AB
3068 *
3069 * Unfortunately this headroom changes the DMA alignment of the resulting
3070 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
3071 * on some architectures. An architecture can override this value,
3072 * perhaps setting it to a cacheline in size (since that will maintain
3073 * cacheline alignment of the DMA). It must be a power of 2.
3074 *
d6301d3d 3075 * Various parts of the networking layer expect at least 32 bytes of
025be81e 3076 * headroom, you should not reduce this.
5933dd2f
ED
3077 *
3078 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
3079 * to reduce average number of cache lines per packet.
645f0897 3080 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 3081 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
3082 */
3083#ifndef NET_SKB_PAD
5933dd2f 3084#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
3085#endif
3086
7965bd4d 3087int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 3088
5293efe6 3089static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 3090{
5e1abdc3 3091 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 3092 return;
27a884dc
ACM
3093 skb->len = len;
3094 skb_set_tail_pointer(skb, len);
1da177e4
LT
3095}
3096
5293efe6
DB
3097static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3098{
3099 __skb_set_length(skb, len);
3100}
3101
7965bd4d 3102void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
3103
3104static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3105{
3cc0e873
HX
3106 if (skb->data_len)
3107 return ___pskb_trim(skb, len);
3108 __skb_trim(skb, len);
3109 return 0;
1da177e4
LT
3110}
3111
3112static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3113{
3114 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3115}
3116
e9fa4f7b
HX
3117/**
3118 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3119 * @skb: buffer to alter
3120 * @len: new length
3121 *
3122 * This is identical to pskb_trim except that the caller knows that
3123 * the skb is not cloned so we should never get an error due to out-
3124 * of-memory.
3125 */
3126static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3127{
3128 int err = pskb_trim(skb, len);
3129 BUG_ON(err);
3130}
3131
5293efe6
DB
3132static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3133{
3134 unsigned int diff = len - skb->len;
3135
3136 if (skb_tailroom(skb) < diff) {
3137 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3138 GFP_ATOMIC);
3139 if (ret)
3140 return ret;
3141 }
3142 __skb_set_length(skb, len);
3143 return 0;
3144}
3145
1da177e4
LT
3146/**
3147 * skb_orphan - orphan a buffer
3148 * @skb: buffer to orphan
3149 *
3150 * If a buffer currently has an owner then we call the owner's
3151 * destructor function and make the @skb unowned. The buffer continues
3152 * to exist but is no longer charged to its former owner.
3153 */
3154static inline void skb_orphan(struct sk_buff *skb)
3155{
c34a7612 3156 if (skb->destructor) {
1da177e4 3157 skb->destructor(skb);
c34a7612
ED
3158 skb->destructor = NULL;
3159 skb->sk = NULL;
376c7311
ED
3160 } else {
3161 BUG_ON(skb->sk);
c34a7612 3162 }
1da177e4
LT
3163}
3164
a353e0ce
MT
3165/**
3166 * skb_orphan_frags - orphan the frags contained in a buffer
3167 * @skb: buffer to orphan frags from
3168 * @gfp_mask: allocation mask for replacement pages
3169 *
3170 * For each frag in the SKB which needs a destructor (i.e. has an
3171 * owner) create a copy of that frag and release the original
3172 * page by calling the destructor.
3173 */
3174static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3175{
1f8b977a
WB
3176 if (likely(!skb_zcopy(skb)))
3177 return 0;
185ce5c3 3178 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 3179 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
3180 return 0;
3181 return skb_copy_ubufs(skb, gfp_mask);
3182}
3183
3184/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
3185static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3186{
3187 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
3188 return 0;
3189 return skb_copy_ubufs(skb, gfp_mask);
3190}
3191
1da177e4
LT
3192/**
3193 * __skb_queue_purge - empty a list
3194 * @list: list to empty
3195 *
3196 * Delete all buffers on an &sk_buff list. Each buffer is removed from
3197 * the list and one reference dropped. This function does not take the
3198 * list lock and the caller must hold the relevant locks to use it.
3199 */
1da177e4
LT
3200static inline void __skb_queue_purge(struct sk_buff_head *list)
3201{
3202 struct sk_buff *skb;
3203 while ((skb = __skb_dequeue(list)) != NULL)
3204 kfree_skb(skb);
3205}
4ea7b0cf 3206void skb_queue_purge(struct sk_buff_head *list);
1da177e4 3207
385114de 3208unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 3209
3f6e687d
KH
3210void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3211
3212/**
3213 * netdev_alloc_frag - allocate a page fragment
3214 * @fragsz: fragment size
3215 *
3216 * Allocates a frag from a page for receive buffer.
3217 * Uses GFP_ATOMIC allocations.
3218 */
3219static inline void *netdev_alloc_frag(unsigned int fragsz)
3220{
3221 return __netdev_alloc_frag_align(fragsz, ~0u);
3222}
3223
3224static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3225 unsigned int align)
3226{
3227 WARN_ON_ONCE(!is_power_of_2(align));
3228 return __netdev_alloc_frag_align(fragsz, -align);
3229}
1da177e4 3230
7965bd4d
JP
3231struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3232 gfp_t gfp_mask);
8af27456
CH
3233
3234/**
3235 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
3236 * @dev: network device to receive on
3237 * @length: length to allocate
3238 *
3239 * Allocate a new &sk_buff and assign it a usage count of one. The
3240 * buffer has unspecified headroom built in. Users should allocate
3241 * the headroom they think they need without accounting for the
3242 * built in space. The built in space is used for optimisations.
3243 *
3244 * %NULL is returned if there is no free memory. Although this function
3245 * allocates memory it can be called from an interrupt.
3246 */
3247static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 3248 unsigned int length)
8af27456
CH
3249{
3250 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3251}
3252
6f532612
ED
3253/* legacy helper around __netdev_alloc_skb() */
3254static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3255 gfp_t gfp_mask)
3256{
3257 return __netdev_alloc_skb(NULL, length, gfp_mask);
3258}
3259
3260/* legacy helper around netdev_alloc_skb() */
3261static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3262{
3263 return netdev_alloc_skb(NULL, length);
3264}
3265
3266
4915a0de
ED
3267static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3268 unsigned int length, gfp_t gfp)
61321bbd 3269{
4915a0de 3270 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
3271
3272 if (NET_IP_ALIGN && skb)
3273 skb_reserve(skb, NET_IP_ALIGN);
3274 return skb;
3275}
3276
4915a0de
ED
3277static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3278 unsigned int length)
3279{
3280 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3281}
3282
181edb2b
AD
3283static inline void skb_free_frag(void *addr)
3284{
8c2dd3e4 3285 page_frag_free(addr);
181edb2b
AD
3286}
3287
3f6e687d
KH
3288void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3289
3290static inline void *napi_alloc_frag(unsigned int fragsz)
3291{
3292 return __napi_alloc_frag_align(fragsz, ~0u);
3293}
3294
3295static inline void *napi_alloc_frag_align(unsigned int fragsz,
3296 unsigned int align)
3297{
3298 WARN_ON_ONCE(!is_power_of_2(align));
3299 return __napi_alloc_frag_align(fragsz, -align);
3300}
3301
fd11a83d
AD
3302struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3303 unsigned int length, gfp_t gfp_mask);
3304static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3305 unsigned int length)
3306{
3307 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3308}
795bb1c0
JDB
3309void napi_consume_skb(struct sk_buff *skb, int budget);
3310
9243adfc 3311void napi_skb_free_stolen_head(struct sk_buff *skb);
15fad714 3312void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 3313
71dfda58
AD
3314/**
3315 * __dev_alloc_pages - allocate page for network Rx
3316 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3317 * @order: size of the allocation
3318 *
3319 * Allocate a new page.
3320 *
3321 * %NULL is returned if there is no free memory.
3322*/
3323static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3324 unsigned int order)
3325{
3326 /* This piece of code contains several assumptions.
3327 * 1. This is for device Rx, therefor a cold page is preferred.
3328 * 2. The expectation is the user wants a compound page.
3329 * 3. If requesting a order 0 page it will not be compound
3330 * due to the check to see if order has a value in prep_new_page
3331 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3332 * code in gfp_to_alloc_flags that should be enforcing this.
3333 */
453f85d4 3334 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
3335
3336 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3337}
3338
3339static inline struct page *dev_alloc_pages(unsigned int order)
3340{
95829b3a 3341 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
3342}
3343
3344/**
3345 * __dev_alloc_page - allocate a page for network Rx
3346 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3347 *
3348 * Allocate a new page.
3349 *
3350 * %NULL is returned if there is no free memory.
3351 */
3352static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3353{
3354 return __dev_alloc_pages(gfp_mask, 0);
3355}
3356
3357static inline struct page *dev_alloc_page(void)
3358{
95829b3a 3359 return dev_alloc_pages(0);
71dfda58
AD
3360}
3361
bc38f30f
AL
3362/**
3363 * dev_page_is_reusable - check whether a page can be reused for network Rx
3364 * @page: the page to test
3365 *
3366 * A page shouldn't be considered for reusing/recycling if it was allocated
3367 * under memory pressure or at a distant memory node.
3368 *
3369 * Returns false if this page should be returned to page allocator, true
3370 * otherwise.
3371 */
3372static inline bool dev_page_is_reusable(const struct page *page)
3373{
3374 return likely(page_to_nid(page) == numa_mem_id() &&
3375 !page_is_pfmemalloc(page));
3376}
3377
0614002b
MG
3378/**
3379 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3380 * @page: The page that was allocated from skb_alloc_page
3381 * @skb: The skb that may need pfmemalloc set
3382 */
48f971c9
AL
3383static inline void skb_propagate_pfmemalloc(const struct page *page,
3384 struct sk_buff *skb)
0614002b 3385{
2f064f34 3386 if (page_is_pfmemalloc(page))
0614002b
MG
3387 skb->pfmemalloc = true;
3388}
3389
7240b60c
JL
3390/**
3391 * skb_frag_off() - Returns the offset of a skb fragment
3392 * @frag: the paged fragment
3393 */
3394static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3395{
65c84f14 3396 return frag->bv_offset;
7240b60c
JL
3397}
3398
3399/**
3400 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3401 * @frag: skb fragment
3402 * @delta: value to add
3403 */
3404static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3405{
65c84f14 3406 frag->bv_offset += delta;
7240b60c
JL
3407}
3408
3409/**
3410 * skb_frag_off_set() - Sets the offset of a skb fragment
3411 * @frag: skb fragment
3412 * @offset: offset of fragment
3413 */
3414static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3415{
65c84f14 3416 frag->bv_offset = offset;
7240b60c
JL
3417}
3418
3419/**
3420 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3421 * @fragto: skb fragment where offset is set
3422 * @fragfrom: skb fragment offset is copied from
3423 */
3424static inline void skb_frag_off_copy(skb_frag_t *fragto,
3425 const skb_frag_t *fragfrom)
3426{
65c84f14 3427 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
3428}
3429
131ea667 3430/**
e227867f 3431 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
3432 * @frag: the paged fragment
3433 *
3434 * Returns the &struct page associated with @frag.
3435 */
3436static inline struct page *skb_frag_page(const skb_frag_t *frag)
3437{
1dfa5bd3 3438 return frag->bv_page;
131ea667
IC
3439}
3440
3441/**
3442 * __skb_frag_ref - take an addition reference on a paged fragment.
3443 * @frag: the paged fragment
3444 *
3445 * Takes an additional reference on the paged fragment @frag.
3446 */
3447static inline void __skb_frag_ref(skb_frag_t *frag)
3448{
3449 get_page(skb_frag_page(frag));
3450}
3451
3452/**
3453 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3454 * @skb: the buffer
3455 * @f: the fragment offset.
3456 *
3457 * Takes an additional reference on the @f'th paged fragment of @skb.
3458 */
3459static inline void skb_frag_ref(struct sk_buff *skb, int f)
3460{
3461 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3462}
3463
3464/**
3465 * __skb_frag_unref - release a reference on a paged fragment.
3466 * @frag: the paged fragment
c420c989 3467 * @recycle: recycle the page if allocated via page_pool
131ea667 3468 *
c420c989
MC
3469 * Releases a reference on the paged fragment @frag
3470 * or recycles the page via the page_pool API.
131ea667 3471 */
c420c989 3472static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
131ea667 3473{
6a5bcd84
IA
3474 struct page *page = skb_frag_page(frag);
3475
3476#ifdef CONFIG_PAGE_POOL
3477 if (recycle && page_pool_return_skb_page(page))
3478 return;
3479#endif
3480 put_page(page);
131ea667
IC
3481}
3482
3483/**
3484 * skb_frag_unref - release a reference on a paged fragment of an skb.
3485 * @skb: the buffer
3486 * @f: the fragment offset
3487 *
3488 * Releases a reference on the @f'th paged fragment of @skb.
3489 */
3490static inline void skb_frag_unref(struct sk_buff *skb, int f)
3491{
6a5bcd84 3492 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
131ea667
IC
3493}
3494
3495/**
3496 * skb_frag_address - gets the address of the data contained in a paged fragment
3497 * @frag: the paged fragment buffer
3498 *
3499 * Returns the address of the data within @frag. The page must already
3500 * be mapped.
3501 */
3502static inline void *skb_frag_address(const skb_frag_t *frag)
3503{
7240b60c 3504 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3505}
3506
3507/**
3508 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3509 * @frag: the paged fragment buffer
3510 *
3511 * Returns the address of the data within @frag. Checks that the page
3512 * is mapped and returns %NULL otherwise.
3513 */
3514static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3515{
3516 void *ptr = page_address(skb_frag_page(frag));
3517 if (unlikely(!ptr))
3518 return NULL;
3519
7240b60c
JL
3520 return ptr + skb_frag_off(frag);
3521}
3522
3523/**
3524 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3525 * @fragto: skb fragment where page is set
3526 * @fragfrom: skb fragment page is copied from
3527 */
3528static inline void skb_frag_page_copy(skb_frag_t *fragto,
3529 const skb_frag_t *fragfrom)
3530{
3531 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3532}
3533
3534/**
3535 * __skb_frag_set_page - sets the page contained in a paged fragment
3536 * @frag: the paged fragment
3537 * @page: the page to set
3538 *
3539 * Sets the fragment @frag to contain @page.
3540 */
3541static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3542{
1dfa5bd3 3543 frag->bv_page = page;
131ea667
IC
3544}
3545
3546/**
3547 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3548 * @skb: the buffer
3549 * @f: the fragment offset
3550 * @page: the page to set
3551 *
3552 * Sets the @f'th fragment of @skb to contain @page.
3553 */
3554static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3555 struct page *page)
3556{
3557 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3558}
3559
400dfd3a
ED
3560bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3561
131ea667
IC
3562/**
3563 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3564 * @dev: the device to map the fragment to
131ea667
IC
3565 * @frag: the paged fragment to map
3566 * @offset: the offset within the fragment (starting at the
3567 * fragment's own offset)
3568 * @size: the number of bytes to map
771b00a8 3569 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3570 *
3571 * Maps the page associated with @frag to @device.
3572 */
3573static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3574 const skb_frag_t *frag,
3575 size_t offset, size_t size,
3576 enum dma_data_direction dir)
3577{
3578 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3579 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3580}
3581
117632e6
ED
3582static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3583 gfp_t gfp_mask)
3584{
3585 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3586}
3587
bad93e9d
OP
3588
3589static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3590 gfp_t gfp_mask)
3591{
3592 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3593}
3594
3595
334a8132
PM
3596/**
3597 * skb_clone_writable - is the header of a clone writable
3598 * @skb: buffer to check
3599 * @len: length up to which to write
3600 *
3601 * Returns true if modifying the header part of the cloned buffer
3602 * does not requires the data to be copied.
3603 */
05bdd2f1 3604static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3605{
3606 return !skb_header_cloned(skb) &&
3607 skb_headroom(skb) + len <= skb->hdr_len;
3608}
3609
3697649f
DB
3610static inline int skb_try_make_writable(struct sk_buff *skb,
3611 unsigned int write_len)
3612{
3613 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3614 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3615}
3616
d9cc2048
HX
3617static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3618 int cloned)
3619{
3620 int delta = 0;
3621
d9cc2048
HX
3622 if (headroom > skb_headroom(skb))
3623 delta = headroom - skb_headroom(skb);
3624
3625 if (delta || cloned)
3626 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3627 GFP_ATOMIC);
3628 return 0;
3629}
3630
1da177e4
LT
3631/**
3632 * skb_cow - copy header of skb when it is required
3633 * @skb: buffer to cow
3634 * @headroom: needed headroom
3635 *
3636 * If the skb passed lacks sufficient headroom or its data part
3637 * is shared, data is reallocated. If reallocation fails, an error
3638 * is returned and original skb is not changed.
3639 *
3640 * The result is skb with writable area skb->head...skb->tail
3641 * and at least @headroom of space at head.
3642 */
3643static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3644{
d9cc2048
HX
3645 return __skb_cow(skb, headroom, skb_cloned(skb));
3646}
1da177e4 3647
d9cc2048
HX
3648/**
3649 * skb_cow_head - skb_cow but only making the head writable
3650 * @skb: buffer to cow
3651 * @headroom: needed headroom
3652 *
3653 * This function is identical to skb_cow except that we replace the
3654 * skb_cloned check by skb_header_cloned. It should be used when
3655 * you only need to push on some header and do not need to modify
3656 * the data.
3657 */
3658static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3659{
3660 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3661}
3662
3663/**
3664 * skb_padto - pad an skbuff up to a minimal size
3665 * @skb: buffer to pad
3666 * @len: minimal length
3667 *
3668 * Pads up a buffer to ensure the trailing bytes exist and are
3669 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3670 * is untouched. Otherwise it is extended. Returns zero on
3671 * success. The skb is freed on error.
1da177e4 3672 */
5b057c6b 3673static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3674{
3675 unsigned int size = skb->len;
3676 if (likely(size >= len))
5b057c6b 3677 return 0;
987c402a 3678 return skb_pad(skb, len - size);
1da177e4
LT
3679}
3680
9c0c1124 3681/**
4ea7b0cf 3682 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3683 * @skb: buffer to pad
3684 * @len: minimal length
cd0a137a 3685 * @free_on_error: free buffer on error
9c0c1124
AD
3686 *
3687 * Pads up a buffer to ensure the trailing bytes exist and are
3688 * blanked. If the buffer already contains sufficient data it
3689 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3690 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3691 */
4a009cb0
ED
3692static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3693 unsigned int len,
3694 bool free_on_error)
9c0c1124
AD
3695{
3696 unsigned int size = skb->len;
3697
3698 if (unlikely(size < len)) {
3699 len -= size;
cd0a137a 3700 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3701 return -ENOMEM;
3702 __skb_put(skb, len);
3703 }
3704 return 0;
3705}
3706
cd0a137a
FF
3707/**
3708 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3709 * @skb: buffer to pad
3710 * @len: minimal length
3711 *
3712 * Pads up a buffer to ensure the trailing bytes exist and are
3713 * blanked. If the buffer already contains sufficient data it
3714 * is untouched. Otherwise it is extended. Returns zero on
3715 * success. The skb is freed on error.
3716 */
4a009cb0 3717static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3718{
3719 return __skb_put_padto(skb, len, true);
3720}
3721
1da177e4 3722static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3723 struct iov_iter *from, int copy)
1da177e4
LT
3724{
3725 const int off = skb->len;
3726
3727 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3728 __wsum csum = 0;
15e6cb46
AV
3729 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3730 &csum, from)) {
1da177e4
LT
3731 skb->csum = csum_block_add(skb->csum, csum, off);
3732 return 0;
3733 }
15e6cb46 3734 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3735 return 0;
3736
3737 __skb_trim(skb, off);
3738 return -EFAULT;
3739}
3740
38ba0a65
ED
3741static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3742 const struct page *page, int off)
1da177e4 3743{
1f8b977a
WB
3744 if (skb_zcopy(skb))
3745 return false;
1da177e4 3746 if (i) {
d8e18a51 3747 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3748
ea2ab693 3749 return page == skb_frag_page(frag) &&
7240b60c 3750 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3751 }
38ba0a65 3752 return false;
1da177e4
LT
3753}
3754
364c6bad
HX
3755static inline int __skb_linearize(struct sk_buff *skb)
3756{
3757 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3758}
3759
1da177e4
LT
3760/**
3761 * skb_linearize - convert paged skb to linear one
3762 * @skb: buffer to linarize
1da177e4
LT
3763 *
3764 * If there is no free memory -ENOMEM is returned, otherwise zero
3765 * is returned and the old skb data released.
3766 */
364c6bad
HX
3767static inline int skb_linearize(struct sk_buff *skb)
3768{
3769 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3770}
3771
cef401de
ED
3772/**
3773 * skb_has_shared_frag - can any frag be overwritten
3774 * @skb: buffer to test
3775 *
3776 * Return true if the skb has at least one frag that might be modified
3777 * by an external entity (as in vmsplice()/sendfile())
3778 */
3779static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3780{
c9af6db4 3781 return skb_is_nonlinear(skb) &&
06b4feb3 3782 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3783}
3784
364c6bad
HX
3785/**
3786 * skb_linearize_cow - make sure skb is linear and writable
3787 * @skb: buffer to process
3788 *
3789 * If there is no free memory -ENOMEM is returned, otherwise zero
3790 * is returned and the old skb data released.
3791 */
3792static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3793{
364c6bad
HX
3794 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3795 __skb_linearize(skb) : 0;
1da177e4
LT
3796}
3797
479ffccc
DB
3798static __always_inline void
3799__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3800 unsigned int off)
3801{
3802 if (skb->ip_summed == CHECKSUM_COMPLETE)
3803 skb->csum = csum_block_sub(skb->csum,
3804 csum_partial(start, len, 0), off);
3805 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3806 skb_checksum_start_offset(skb) < 0)
3807 skb->ip_summed = CHECKSUM_NONE;
3808}
3809
1da177e4
LT
3810/**
3811 * skb_postpull_rcsum - update checksum for received skb after pull
3812 * @skb: buffer to update
3813 * @start: start of data before pull
3814 * @len: length of data pulled
3815 *
3816 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3817 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3818 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3819 */
1da177e4 3820static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3821 const void *start, unsigned int len)
1da177e4 3822{
29c30026 3823 if (skb->ip_summed == CHECKSUM_COMPLETE)
45cac675
ED
3824 skb->csum = wsum_negate(csum_partial(start, len,
3825 wsum_negate(skb->csum)));
29c30026
ED
3826 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3827 skb_checksum_start_offset(skb) < 0)
3828 skb->ip_summed = CHECKSUM_NONE;
1da177e4
LT
3829}
3830
479ffccc
DB
3831static __always_inline void
3832__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3833 unsigned int off)
3834{
3835 if (skb->ip_summed == CHECKSUM_COMPLETE)
3836 skb->csum = csum_block_add(skb->csum,
3837 csum_partial(start, len, 0), off);
3838}
cbb042f9 3839
479ffccc
DB
3840/**
3841 * skb_postpush_rcsum - update checksum for received skb after push
3842 * @skb: buffer to update
3843 * @start: start of data after push
3844 * @len: length of data pushed
3845 *
3846 * After doing a push on a received packet, you need to call this to
3847 * update the CHECKSUM_COMPLETE checksum.
3848 */
f8ffad69
DB
3849static inline void skb_postpush_rcsum(struct sk_buff *skb,
3850 const void *start, unsigned int len)
3851{
479ffccc 3852 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3853}
3854
af72868b 3855void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3856
82a31b92
WC
3857/**
3858 * skb_push_rcsum - push skb and update receive checksum
3859 * @skb: buffer to update
3860 * @len: length of data pulled
3861 *
3862 * This function performs an skb_push on the packet and updates
3863 * the CHECKSUM_COMPLETE checksum. It should be used on
3864 * receive path processing instead of skb_push unless you know
3865 * that the checksum difference is zero (e.g., a valid IP header)
3866 * or you are setting ip_summed to CHECKSUM_NONE.
3867 */
d58ff351 3868static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3869{
3870 skb_push(skb, len);
3871 skb_postpush_rcsum(skb, skb->data, len);
3872 return skb->data;
3873}
3874
88078d98 3875int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3876/**
3877 * pskb_trim_rcsum - trim received skb and update checksum
3878 * @skb: buffer to trim
3879 * @len: new length
3880 *
3881 * This is exactly the same as pskb_trim except that it ensures the
3882 * checksum of received packets are still valid after the operation.
6c57f045 3883 * It can change skb pointers.
7ce5a27f
DM
3884 */
3885
3886static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3887{
3888 if (likely(len >= skb->len))
3889 return 0;
88078d98 3890 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3891}
3892
5293efe6
DB
3893static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3894{
3895 if (skb->ip_summed == CHECKSUM_COMPLETE)
3896 skb->ip_summed = CHECKSUM_NONE;
3897 __skb_trim(skb, len);
3898 return 0;
3899}
3900
3901static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3902{
3903 if (skb->ip_summed == CHECKSUM_COMPLETE)
3904 skb->ip_summed = CHECKSUM_NONE;
3905 return __skb_grow(skb, len);
3906}
3907
18a4c0ea
ED
3908#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3909#define skb_rb_first(root) rb_to_skb(rb_first(root))
3910#define skb_rb_last(root) rb_to_skb(rb_last(root))
3911#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3912#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3913
1da177e4
LT
3914#define skb_queue_walk(queue, skb) \
3915 for (skb = (queue)->next; \
a1e4891f 3916 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3917 skb = skb->next)
3918
46f8914e
JC
3919#define skb_queue_walk_safe(queue, skb, tmp) \
3920 for (skb = (queue)->next, tmp = skb->next; \
3921 skb != (struct sk_buff *)(queue); \
3922 skb = tmp, tmp = skb->next)
3923
1164f52a 3924#define skb_queue_walk_from(queue, skb) \
a1e4891f 3925 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3926 skb = skb->next)
3927
18a4c0ea
ED
3928#define skb_rbtree_walk(skb, root) \
3929 for (skb = skb_rb_first(root); skb != NULL; \
3930 skb = skb_rb_next(skb))
3931
3932#define skb_rbtree_walk_from(skb) \
3933 for (; skb != NULL; \
3934 skb = skb_rb_next(skb))
3935
3936#define skb_rbtree_walk_from_safe(skb, tmp) \
3937 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3938 skb = tmp)
3939
1164f52a
DM
3940#define skb_queue_walk_from_safe(queue, skb, tmp) \
3941 for (tmp = skb->next; \
3942 skb != (struct sk_buff *)(queue); \
3943 skb = tmp, tmp = skb->next)
3944
300ce174
SH
3945#define skb_queue_reverse_walk(queue, skb) \
3946 for (skb = (queue)->prev; \
a1e4891f 3947 skb != (struct sk_buff *)(queue); \
300ce174
SH
3948 skb = skb->prev)
3949
686a2955
DM
3950#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3951 for (skb = (queue)->prev, tmp = skb->prev; \
3952 skb != (struct sk_buff *)(queue); \
3953 skb = tmp, tmp = skb->prev)
3954
3955#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3956 for (tmp = skb->prev; \
3957 skb != (struct sk_buff *)(queue); \
3958 skb = tmp, tmp = skb->prev)
1da177e4 3959
21dc3301 3960static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3961{
3962 return skb_shinfo(skb)->frag_list != NULL;
3963}
3964
3965static inline void skb_frag_list_init(struct sk_buff *skb)
3966{
3967 skb_shinfo(skb)->frag_list = NULL;
3968}
3969
ee039871
DM
3970#define skb_walk_frags(skb, iter) \
3971 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3972
ea3793ee 3973
b50b0580
SD
3974int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3975 int *err, long *timeo_p,
ea3793ee 3976 const struct sk_buff *skb);
65101aec
PA
3977struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3978 struct sk_buff_head *queue,
3979 unsigned int flags,
fd69c399 3980 int *off, int *err,
65101aec 3981 struct sk_buff **last);
b50b0580
SD
3982struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3983 struct sk_buff_head *queue,
e427cad6 3984 unsigned int flags, int *off, int *err,
ea3793ee 3985 struct sk_buff **last);
b50b0580
SD
3986struct sk_buff *__skb_recv_datagram(struct sock *sk,
3987 struct sk_buff_head *sk_queue,
e427cad6 3988 unsigned int flags, int *off, int *err);
f4b41f06 3989struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
a11e1d43
LT
3990__poll_t datagram_poll(struct file *file, struct socket *sock,
3991 struct poll_table_struct *wait);
c0371da6
AV
3992int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3993 struct iov_iter *to, int size);
51f3d02b
DM
3994static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3995 struct msghdr *msg, int size)
3996{
e5a4b0bb 3997 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3998}
e5a4b0bb
AV
3999int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
4000 struct msghdr *msg);
65d69e25
SG
4001int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
4002 struct iov_iter *to, int len,
4003 struct ahash_request *hash);
3a654f97
AV
4004int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
4005 struct iov_iter *from, int len);
3a654f97 4006int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 4007void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 4008void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
4009static inline void skb_free_datagram_locked(struct sock *sk,
4010 struct sk_buff *skb)
4011{
4012 __skb_free_datagram_locked(sk, skb, 0);
4013}
7965bd4d 4014int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
4015int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
4016int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
4017__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 4018 int len);
a60e3cc7 4019int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 4020 struct pipe_inode_info *pipe, unsigned int len,
25869262 4021 unsigned int flags);
20bf50de
TH
4022int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
4023 int len);
0739cd28 4024int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
7965bd4d 4025void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 4026unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
4027int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
4028 int len, int hlen);
7965bd4d
JP
4029void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
4030int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
4031void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 4032bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 4033bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 4034struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
4035struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
4036 unsigned int offset);
0d5501c1 4037struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
92ece280 4038int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
bfca4c52 4039int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
4040int skb_vlan_pop(struct sk_buff *skb);
4041int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
4042int skb_eth_pop(struct sk_buff *skb);
4043int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
4044 const unsigned char *src);
fa4e0f88 4045int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 4046 int mac_len, bool ethernet);
040b5cfb
MV
4047int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
4048 bool ethernet);
d27cf5c5 4049int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 4050int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
4051struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
4052 gfp_t gfp);
20380731 4053
6ce8e9ce
AV
4054static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
4055{
3073f070 4056 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
4057}
4058
7eab8d9e
AV
4059static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
4060{
e5a4b0bb 4061 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
4062}
4063
2817a336
DB
4064struct skb_checksum_ops {
4065 __wsum (*update)(const void *mem, int len, __wsum wsum);
4066 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
4067};
4068
9617813d
DC
4069extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
4070
2817a336
DB
4071__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
4072 __wsum csum, const struct skb_checksum_ops *ops);
4073__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
4074 __wsum csum);
4075
1e98a0f0 4076static inline void * __must_check
e3305138
AL
4077__skb_header_pointer(const struct sk_buff *skb, int offset, int len,
4078 const void *data, int hlen, void *buffer)
1da177e4 4079{
d206121f 4080 if (likely(hlen - offset >= len))
e3305138 4081 return (void *)data + offset;
1da177e4 4082
d206121f 4083 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
1da177e4
LT
4084 return NULL;
4085
4086 return buffer;
4087}
4088
1e98a0f0
ED
4089static inline void * __must_check
4090skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
4091{
4092 return __skb_header_pointer(skb, offset, len, skb->data,
4093 skb_headlen(skb), buffer);
4094}
4095
4262e5cc
DB
4096/**
4097 * skb_needs_linearize - check if we need to linearize a given skb
4098 * depending on the given device features.
4099 * @skb: socket buffer to check
4100 * @features: net device features
4101 *
4102 * Returns true if either:
4103 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
4104 * 2. skb is fragmented and the device does not support SG.
4105 */
4106static inline bool skb_needs_linearize(struct sk_buff *skb,
4107 netdev_features_t features)
4108{
4109 return skb_is_nonlinear(skb) &&
4110 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4111 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4112}
4113
d626f62b
ACM
4114static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4115 void *to,
4116 const unsigned int len)
4117{
4118 memcpy(to, skb->data, len);
4119}
4120
4121static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4122 const int offset, void *to,
4123 const unsigned int len)
4124{
4125 memcpy(to, skb->data + offset, len);
4126}
4127
27d7ff46
ACM
4128static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4129 const void *from,
4130 const unsigned int len)
4131{
4132 memcpy(skb->data, from, len);
4133}
4134
4135static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4136 const int offset,
4137 const void *from,
4138 const unsigned int len)
4139{
4140 memcpy(skb->data + offset, from, len);
4141}
4142
7965bd4d 4143void skb_init(void);
1da177e4 4144
ac45f602
PO
4145static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4146{
4147 return skb->tstamp;
4148}
4149
a61bbcf2
PM
4150/**
4151 * skb_get_timestamp - get timestamp from a skb
4152 * @skb: skb to get stamp from
13c6ee2a 4153 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
4154 *
4155 * Timestamps are stored in the skb as offsets to a base timestamp.
4156 * This function converts the offset back to a struct timeval and stores
4157 * it in stamp.
4158 */
ac45f602 4159static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 4160 struct __kernel_old_timeval *stamp)
a61bbcf2 4161{
13c6ee2a 4162 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
4163}
4164
887feae3
DD
4165static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4166 struct __kernel_sock_timeval *stamp)
4167{
4168 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4169
4170 stamp->tv_sec = ts.tv_sec;
4171 stamp->tv_usec = ts.tv_nsec / 1000;
4172}
4173
ac45f602 4174static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 4175 struct __kernel_old_timespec *stamp)
ac45f602 4176{
df1b4ba9
AB
4177 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4178
4179 stamp->tv_sec = ts.tv_sec;
4180 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
4181}
4182
887feae3
DD
4183static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4184 struct __kernel_timespec *stamp)
4185{
4186 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4187
4188 stamp->tv_sec = ts.tv_sec;
4189 stamp->tv_nsec = ts.tv_nsec;
4190}
4191
b7aa0bf7 4192static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 4193{
b7aa0bf7 4194 skb->tstamp = ktime_get_real();
d93376f5 4195 skb->mono_delivery_time = 0;
a61bbcf2
PM
4196}
4197
164891aa
SH
4198static inline ktime_t net_timedelta(ktime_t t)
4199{
4200 return ktime_sub(ktime_get_real(), t);
4201}
4202
a1ac9c8a
MKL
4203static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4204 bool mono)
4205{
4206 skb->tstamp = kt;
d98d58a0 4207 skb->mono_delivery_time = kt && mono;
a1ac9c8a
MKL
4208}
4209
27942a15
MKL
4210DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4211
4212/* It is used in the ingress path to clear the delivery_time.
4213 * If needed, set the skb->tstamp to the (rcv) timestamp.
4214 */
4215static inline void skb_clear_delivery_time(struct sk_buff *skb)
4216{
4217 if (skb->mono_delivery_time) {
4218 skb->mono_delivery_time = 0;
4219 if (static_branch_unlikely(&netstamp_needed_key))
4220 skb->tstamp = ktime_get_real();
4221 else
4222 skb->tstamp = 0;
4223 }
4224}
4225
de799101
MKL
4226static inline void skb_clear_tstamp(struct sk_buff *skb)
4227{
4228 if (skb->mono_delivery_time)
4229 return;
4230
4231 skb->tstamp = 0;
4232}
4233
27942a15
MKL
4234static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4235{
4236 if (skb->mono_delivery_time)
4237 return 0;
4238
4239 return skb->tstamp;
4240}
4241
b6561f84
MKL
4242static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4243{
4244 if (!skb->mono_delivery_time && skb->tstamp)
4245 return skb->tstamp;
4246
4247 if (static_branch_unlikely(&netstamp_needed_key) || cond)
4248 return ktime_get_real();
4249
4250 return 0;
4251}
4252
de8f3a83
DB
4253static inline u8 skb_metadata_len(const struct sk_buff *skb)
4254{
4255 return skb_shinfo(skb)->meta_len;
4256}
4257
4258static inline void *skb_metadata_end(const struct sk_buff *skb)
4259{
4260 return skb_mac_header(skb);
4261}
4262
4263static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4264 const struct sk_buff *skb_b,
4265 u8 meta_len)
4266{
4267 const void *a = skb_metadata_end(skb_a);
4268 const void *b = skb_metadata_end(skb_b);
4269 /* Using more efficient varaiant than plain call to memcmp(). */
4270#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
4271 u64 diffs = 0;
4272
4273 switch (meta_len) {
4274#define __it(x, op) (x -= sizeof(u##op))
4275#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4276 case 32: diffs |= __it_diff(a, b, 64);
df561f66 4277 fallthrough;
de8f3a83 4278 case 24: diffs |= __it_diff(a, b, 64);
df561f66 4279 fallthrough;
de8f3a83 4280 case 16: diffs |= __it_diff(a, b, 64);
df561f66 4281 fallthrough;
de8f3a83
DB
4282 case 8: diffs |= __it_diff(a, b, 64);
4283 break;
4284 case 28: diffs |= __it_diff(a, b, 64);
df561f66 4285 fallthrough;
de8f3a83 4286 case 20: diffs |= __it_diff(a, b, 64);
df561f66 4287 fallthrough;
de8f3a83 4288 case 12: diffs |= __it_diff(a, b, 64);
df561f66 4289 fallthrough;
de8f3a83
DB
4290 case 4: diffs |= __it_diff(a, b, 32);
4291 break;
4292 }
4293 return diffs;
4294#else
4295 return memcmp(a - meta_len, b - meta_len, meta_len);
4296#endif
4297}
4298
4299static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4300 const struct sk_buff *skb_b)
4301{
4302 u8 len_a = skb_metadata_len(skb_a);
4303 u8 len_b = skb_metadata_len(skb_b);
4304
4305 if (!(len_a | len_b))
4306 return false;
4307
4308 return len_a != len_b ?
4309 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4310}
4311
4312static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4313{
4314 skb_shinfo(skb)->meta_len = meta_len;
4315}
4316
4317static inline void skb_metadata_clear(struct sk_buff *skb)
4318{
4319 skb_metadata_set(skb, 0);
4320}
4321
62bccb8c
AD
4322struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4323
c1f19b51
RC
4324#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4325
7965bd4d
JP
4326void skb_clone_tx_timestamp(struct sk_buff *skb);
4327bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
4328
4329#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4330
4331static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4332{
4333}
4334
4335static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4336{
4337 return false;
4338}
4339
4340#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4341
4342/**
4343 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4344 *
da92b194
RC
4345 * PHY drivers may accept clones of transmitted packets for
4346 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
4347 * must call this function to return the skb back to the stack with a
4348 * timestamp.
da92b194 4349 *
2ff17117 4350 * @skb: clone of the original outgoing packet
7a76a021 4351 * @hwtstamps: hardware time stamps
c1f19b51
RC
4352 *
4353 */
4354void skb_complete_tx_timestamp(struct sk_buff *skb,
4355 struct skb_shared_hwtstamps *hwtstamps);
4356
e7ed11ee 4357void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
4358 struct skb_shared_hwtstamps *hwtstamps,
4359 struct sock *sk, int tstype);
4360
ac45f602
PO
4361/**
4362 * skb_tstamp_tx - queue clone of skb with send time stamps
4363 * @orig_skb: the original outgoing packet
4364 * @hwtstamps: hardware time stamps, may be NULL if not available
4365 *
4366 * If the skb has a socket associated, then this function clones the
4367 * skb (thus sharing the actual data and optional structures), stores
4368 * the optional hardware time stamping information (if non NULL) or
4369 * generates a software time stamp (otherwise), then queues the clone
4370 * to the error queue of the socket. Errors are silently ignored.
4371 */
7965bd4d
JP
4372void skb_tstamp_tx(struct sk_buff *orig_skb,
4373 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 4374
4507a715
RC
4375/**
4376 * skb_tx_timestamp() - Driver hook for transmit timestamping
4377 *
4378 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 4379 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 4380 *
73409f3b
DM
4381 * Specifically, one should make absolutely sure that this function is
4382 * called before TX completion of this packet can trigger. Otherwise
4383 * the packet could potentially already be freed.
4384 *
4507a715
RC
4385 * @skb: A socket buffer.
4386 */
4387static inline void skb_tx_timestamp(struct sk_buff *skb)
4388{
c1f19b51 4389 skb_clone_tx_timestamp(skb);
b50a5c70
ML
4390 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4391 skb_tstamp_tx(skb, NULL);
4507a715
RC
4392}
4393
6e3e939f
JB
4394/**
4395 * skb_complete_wifi_ack - deliver skb with wifi status
4396 *
4397 * @skb: the original outgoing packet
4398 * @acked: ack status
4399 *
4400 */
4401void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4402
7965bd4d
JP
4403__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4404__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 4405
60476372
HX
4406static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4407{
6edec0e6
TH
4408 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4409 skb->csum_valid ||
4410 (skb->ip_summed == CHECKSUM_PARTIAL &&
4411 skb_checksum_start_offset(skb) >= 0));
60476372
HX
4412}
4413
fb286bb2
HX
4414/**
4415 * skb_checksum_complete - Calculate checksum of an entire packet
4416 * @skb: packet to process
4417 *
4418 * This function calculates the checksum over the entire packet plus
4419 * the value of skb->csum. The latter can be used to supply the
4420 * checksum of a pseudo header as used by TCP/UDP. It returns the
4421 * checksum.
4422 *
4423 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4424 * this function can be used to verify that checksum on received
4425 * packets. In that case the function should return zero if the
4426 * checksum is correct. In particular, this function will return zero
4427 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4428 * hardware has already verified the correctness of the checksum.
4429 */
4381ca3c 4430static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 4431{
60476372
HX
4432 return skb_csum_unnecessary(skb) ?
4433 0 : __skb_checksum_complete(skb);
fb286bb2
HX
4434}
4435
77cffe23
TH
4436static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4437{
4438 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4439 if (skb->csum_level == 0)
4440 skb->ip_summed = CHECKSUM_NONE;
4441 else
4442 skb->csum_level--;
4443 }
4444}
4445
4446static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4447{
4448 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4449 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4450 skb->csum_level++;
4451 } else if (skb->ip_summed == CHECKSUM_NONE) {
4452 skb->ip_summed = CHECKSUM_UNNECESSARY;
4453 skb->csum_level = 0;
4454 }
4455}
4456
836e66c2
DB
4457static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4458{
4459 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4460 skb->ip_summed = CHECKSUM_NONE;
4461 skb->csum_level = 0;
4462 }
4463}
4464
76ba0aae
TH
4465/* Check if we need to perform checksum complete validation.
4466 *
4467 * Returns true if checksum complete is needed, false otherwise
4468 * (either checksum is unnecessary or zero checksum is allowed).
4469 */
4470static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4471 bool zero_okay,
4472 __sum16 check)
4473{
5d0c2b95
TH
4474 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4475 skb->csum_valid = 1;
77cffe23 4476 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
4477 return false;
4478 }
4479
4480 return true;
4481}
4482
da279887 4483/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
4484 * in checksum_init.
4485 */
4486#define CHECKSUM_BREAK 76
4487
4e18b9ad
TH
4488/* Unset checksum-complete
4489 *
4490 * Unset checksum complete can be done when packet is being modified
4491 * (uncompressed for instance) and checksum-complete value is
4492 * invalidated.
4493 */
4494static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4495{
4496 if (skb->ip_summed == CHECKSUM_COMPLETE)
4497 skb->ip_summed = CHECKSUM_NONE;
4498}
4499
76ba0aae
TH
4500/* Validate (init) checksum based on checksum complete.
4501 *
4502 * Return values:
4503 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4504 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4505 * checksum is stored in skb->csum for use in __skb_checksum_complete
4506 * non-zero: value of invalid checksum
4507 *
4508 */
4509static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4510 bool complete,
4511 __wsum psum)
4512{
4513 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4514 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4515 skb->csum_valid = 1;
76ba0aae
TH
4516 return 0;
4517 }
4518 }
4519
4520 skb->csum = psum;
4521
5d0c2b95
TH
4522 if (complete || skb->len <= CHECKSUM_BREAK) {
4523 __sum16 csum;
4524
4525 csum = __skb_checksum_complete(skb);
4526 skb->csum_valid = !csum;
4527 return csum;
4528 }
76ba0aae
TH
4529
4530 return 0;
4531}
4532
4533static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4534{
4535 return 0;
4536}
4537
4538/* Perform checksum validate (init). Note that this is a macro since we only
4539 * want to calculate the pseudo header which is an input function if necessary.
4540 * First we try to validate without any computation (checksum unnecessary) and
4541 * then calculate based on checksum complete calling the function to compute
4542 * pseudo header.
4543 *
4544 * Return values:
4545 * 0: checksum is validated or try to in skb_checksum_complete
4546 * non-zero: value of invalid checksum
4547 */
4548#define __skb_checksum_validate(skb, proto, complete, \
4549 zero_okay, check, compute_pseudo) \
4550({ \
4551 __sum16 __ret = 0; \
5d0c2b95 4552 skb->csum_valid = 0; \
76ba0aae
TH
4553 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4554 __ret = __skb_checksum_validate_complete(skb, \
4555 complete, compute_pseudo(skb, proto)); \
4556 __ret; \
4557})
4558
4559#define skb_checksum_init(skb, proto, compute_pseudo) \
4560 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4561
4562#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4563 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4564
4565#define skb_checksum_validate(skb, proto, compute_pseudo) \
4566 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4567
4568#define skb_checksum_validate_zero_check(skb, proto, check, \
4569 compute_pseudo) \
096a4cfa 4570 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4571
4572#define skb_checksum_simple_validate(skb) \
4573 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4574
d96535a1
TH
4575static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4576{
219f1d79 4577 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4578}
4579
e4aa33ad 4580static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4581{
4582 skb->csum = ~pseudo;
4583 skb->ip_summed = CHECKSUM_COMPLETE;
4584}
4585
e4aa33ad 4586#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4587do { \
4588 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4589 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4590} while (0)
4591
15e2396d
TH
4592static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4593 u16 start, u16 offset)
4594{
4595 skb->ip_summed = CHECKSUM_PARTIAL;
4596 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4597 skb->csum_offset = offset - start;
4598}
4599
dcdc8994
TH
4600/* Update skbuf and packet to reflect the remote checksum offload operation.
4601 * When called, ptr indicates the starting point for skb->csum when
4602 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4603 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4604 */
4605static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4606 int start, int offset, bool nopartial)
dcdc8994
TH
4607{
4608 __wsum delta;
4609
15e2396d
TH
4610 if (!nopartial) {
4611 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4612 return;
4613 }
4614
10a2308f 4615 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
dcdc8994
TH
4616 __skb_checksum_complete(skb);
4617 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4618 }
4619
4620 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4621
4622 /* Adjust skb->csum since we changed the packet */
4623 skb->csum = csum_add(skb->csum, delta);
4624}
4625
cb9c6836
FW
4626static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4627{
4628#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4629 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4630#else
4631 return NULL;
4632#endif
4633}
4634
261db6c2 4635static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4636{
261db6c2
JS
4637#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4638 return skb->_nfct;
4639#else
4640 return 0UL;
4641#endif
1da177e4 4642}
261db6c2
JS
4643
4644static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4645{
261db6c2 4646#if IS_ENABLED(CONFIG_NF_CONNTRACK)
5fc88f93 4647 skb->slow_gro |= !!nfct;
261db6c2 4648 skb->_nfct = nfct;
2fc72c7b 4649#endif
261db6c2 4650}
df5042f4
FW
4651
4652#ifdef CONFIG_SKB_EXTENSIONS
4653enum skb_ext_id {
4654#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4655 SKB_EXT_BRIDGE_NF,
4165079b
FW
4656#endif
4657#ifdef CONFIG_XFRM
4658 SKB_EXT_SEC_PATH,
95a7233c
PB
4659#endif
4660#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4661 TC_SKB_EXT,
3ee17bc7
MM
4662#endif
4663#if IS_ENABLED(CONFIG_MPTCP)
4664 SKB_EXT_MPTCP,
78476d31
JK
4665#endif
4666#if IS_ENABLED(CONFIG_MCTP_FLOWS)
4667 SKB_EXT_MCTP,
df5042f4
FW
4668#endif
4669 SKB_EXT_NUM, /* must be last */
4670};
4671
4672/**
4673 * struct skb_ext - sk_buff extensions
4674 * @refcnt: 1 on allocation, deallocated on 0
4675 * @offset: offset to add to @data to obtain extension address
4676 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4677 * @data: start of extension data, variable sized
4678 *
4679 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4680 * to use 'u8' types while allowing up to 2kb worth of extension data.
4681 */
4682struct skb_ext {
4683 refcount_t refcnt;
4684 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4685 u8 chunks; /* same */
5c91aa1d 4686 char data[] __aligned(8);
df5042f4
FW
4687};
4688
4930f483 4689struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4690void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4691 struct skb_ext *ext);
df5042f4
FW
4692void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4693void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4694void __skb_ext_put(struct skb_ext *ext);
4695
4696static inline void skb_ext_put(struct sk_buff *skb)
4697{
4698 if (skb->active_extensions)
4699 __skb_ext_put(skb->extensions);
4700}
4701
df5042f4
FW
4702static inline void __skb_ext_copy(struct sk_buff *dst,
4703 const struct sk_buff *src)
4704{
4705 dst->active_extensions = src->active_extensions;
4706
4707 if (src->active_extensions) {
4708 struct skb_ext *ext = src->extensions;
4709
4710 refcount_inc(&ext->refcnt);
4711 dst->extensions = ext;
4712 }
4713}
4714
4715static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4716{
4717 skb_ext_put(dst);
4718 __skb_ext_copy(dst, src);
4719}
4720
4721static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4722{
4723 return !!ext->offset[i];
4724}
4725
4726static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4727{
4728 return skb->active_extensions & (1 << id);
4729}
4730
4731static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4732{
4733 if (skb_ext_exist(skb, id))
4734 __skb_ext_del(skb, id);
4735}
4736
4737static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4738{
4739 if (skb_ext_exist(skb, id)) {
4740 struct skb_ext *ext = skb->extensions;
4741
4742 return (void *)ext + (ext->offset[id] << 3);
4743 }
4744
4745 return NULL;
4746}
174e2381
FW
4747
4748static inline void skb_ext_reset(struct sk_buff *skb)
4749{
4750 if (unlikely(skb->active_extensions)) {
4751 __skb_ext_put(skb->extensions);
4752 skb->active_extensions = 0;
4753 }
4754}
677bf08c
FW
4755
4756static inline bool skb_has_extensions(struct sk_buff *skb)
4757{
4758 return unlikely(skb->active_extensions);
4759}
df5042f4
FW
4760#else
4761static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4762static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4763static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4764static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4765static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4766static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4767#endif /* CONFIG_SKB_EXTENSIONS */
4768
895b5c9f 4769static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4770{
5f79e0f9 4771#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4772 nf_conntrack_put(skb_nfct(skb));
4773 skb->_nfct = 0;
2fc72c7b 4774#endif
a193a4ab
PM
4775}
4776
124dff01
PM
4777static inline void nf_reset_trace(struct sk_buff *skb)
4778{
478b360a 4779#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4780 skb->nf_trace = 0;
4781#endif
a193a4ab
PM
4782}
4783
2b5ec1a5
YY
4784static inline void ipvs_reset(struct sk_buff *skb)
4785{
4786#if IS_ENABLED(CONFIG_IP_VS)
4787 skb->ipvs_property = 0;
4788#endif
4789}
4790
de8bda1d 4791/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4792static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4793 bool copy)
edda553c 4794{
5f79e0f9 4795#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4796 dst->_nfct = src->_nfct;
4797 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4798#endif
478b360a 4799#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4800 if (copy)
4801 dst->nf_trace = src->nf_trace;
478b360a 4802#endif
edda553c
YK
4803}
4804
e7ac05f3
YK
4805static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4806{
e7ac05f3 4807#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4808 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4809#endif
5fc88f93 4810 dst->slow_gro = src->slow_gro;
b1937227 4811 __nf_copy(dst, src, true);
e7ac05f3
YK
4812}
4813
984bc16c
JM
4814#ifdef CONFIG_NETWORK_SECMARK
4815static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4816{
4817 to->secmark = from->secmark;
4818}
4819
4820static inline void skb_init_secmark(struct sk_buff *skb)
4821{
4822 skb->secmark = 0;
4823}
4824#else
4825static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4826{ }
4827
4828static inline void skb_init_secmark(struct sk_buff *skb)
4829{ }
4830#endif
4831
7af8f4ca
FW
4832static inline int secpath_exists(const struct sk_buff *skb)
4833{
4834#ifdef CONFIG_XFRM
4165079b 4835 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4836#else
4837 return 0;
4838#endif
4839}
4840
574f7194
EB
4841static inline bool skb_irq_freeable(const struct sk_buff *skb)
4842{
4843 return !skb->destructor &&
7af8f4ca 4844 !secpath_exists(skb) &&
cb9c6836 4845 !skb_nfct(skb) &&
574f7194
EB
4846 !skb->_skb_refdst &&
4847 !skb_has_frag_list(skb);
4848}
4849
f25f4e44
PWJ
4850static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4851{
f25f4e44 4852 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4853}
4854
9247744e 4855static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4856{
4e3ab47a 4857 return skb->queue_mapping;
4e3ab47a
PE
4858}
4859
f25f4e44
PWJ
4860static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4861{
f25f4e44 4862 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4863}
4864
d5a9e24a
DM
4865static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4866{
4867 skb->queue_mapping = rx_queue + 1;
4868}
4869
9247744e 4870static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4871{
4872 return skb->queue_mapping - 1;
4873}
4874
9247744e 4875static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4876{
a02cec21 4877 return skb->queue_mapping != 0;
d5a9e24a
DM
4878}
4879
4ff06203
JA
4880static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4881{
4882 skb->dst_pending_confirm = val;
4883}
4884
4885static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4886{
4887 return skb->dst_pending_confirm != 0;
4888}
4889
2294be0f 4890static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4891{
0b3d8e08 4892#ifdef CONFIG_XFRM
4165079b 4893 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4894#else
def8b4fa 4895 return NULL;
def8b4fa 4896#endif
0b3d8e08 4897}
def8b4fa 4898
68c33163
PS
4899/* Keeps track of mac header offset relative to skb->head.
4900 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4901 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4902 * tunnel skb it points to outer mac header.
4903 * Keeps track of level of encapsulation of network headers.
4904 */
68c33163 4905struct skb_gso_cb {
802ab55a
AD
4906 union {
4907 int mac_offset;
4908 int data_offset;
4909 };
3347c960 4910 int encap_level;
76443456 4911 __wsum csum;
7e2b10c1 4912 __u16 csum_start;
68c33163 4913};
a08e7fd9
CZ
4914#define SKB_GSO_CB_OFFSET 32
4915#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4916
4917static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4918{
4919 return (skb_mac_header(inner_skb) - inner_skb->head) -
4920 SKB_GSO_CB(inner_skb)->mac_offset;
4921}
4922
1e2bd517
PS
4923static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4924{
4925 int new_headroom, headroom;
4926 int ret;
4927
4928 headroom = skb_headroom(skb);
4929 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4930 if (ret)
4931 return ret;
4932
4933 new_headroom = skb_headroom(skb);
4934 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4935 return 0;
4936}
4937
08b64fcc
AD
4938static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4939{
4940 /* Do not update partial checksums if remote checksum is enabled. */
4941 if (skb->remcsum_offload)
4942 return;
4943
4944 SKB_GSO_CB(skb)->csum = res;
4945 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4946}
4947
7e2b10c1
TH
4948/* Compute the checksum for a gso segment. First compute the checksum value
4949 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4950 * then add in skb->csum (checksum from csum_start to end of packet).
4951 * skb->csum and csum_start are then updated to reflect the checksum of the
4952 * resultant packet starting from the transport header-- the resultant checksum
4953 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4954 * header.
4955 */
4956static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4957{
76443456
AD
4958 unsigned char *csum_start = skb_transport_header(skb);
4959 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4960 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4961
76443456
AD
4962 SKB_GSO_CB(skb)->csum = res;
4963 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4964
76443456 4965 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4966}
4967
bdcc0924 4968static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4969{
4970 return skb_shinfo(skb)->gso_size;
4971}
4972
36a8f39e 4973/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4974static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4975{
4976 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4977}
4978
d02f51cb
DA
4979/* Note: Should be called only if skb_is_gso(skb) is true */
4980static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4981{
4982 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4983}
4984
4c3024de 4985/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4986static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4987{
4c3024de 4988 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4989}
4990
5293efe6
DB
4991static inline void skb_gso_reset(struct sk_buff *skb)
4992{
4993 skb_shinfo(skb)->gso_size = 0;
4994 skb_shinfo(skb)->gso_segs = 0;
4995 skb_shinfo(skb)->gso_type = 0;
4996}
4997
d02f51cb
DA
4998static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4999 u16 increment)
5000{
5001 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5002 return;
5003 shinfo->gso_size += increment;
5004}
5005
5006static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
5007 u16 decrement)
5008{
5009 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5010 return;
5011 shinfo->gso_size -= decrement;
5012}
5013
7965bd4d 5014void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
5015
5016static inline bool skb_warn_if_lro(const struct sk_buff *skb)
5017{
5018 /* LRO sets gso_size but not gso_type, whereas if GSO is really
5019 * wanted then gso_type will be set. */
05bdd2f1
ED
5020 const struct skb_shared_info *shinfo = skb_shinfo(skb);
5021
b78462eb
AD
5022 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
5023 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
5024 __skb_warn_lro_forwarding(skb);
5025 return true;
5026 }
5027 return false;
5028}
5029
35fc92a9
HX
5030static inline void skb_forward_csum(struct sk_buff *skb)
5031{
5032 /* Unfortunately we don't support this one. Any brave souls? */
5033 if (skb->ip_summed == CHECKSUM_COMPLETE)
5034 skb->ip_summed = CHECKSUM_NONE;
5035}
5036
bc8acf2c
ED
5037/**
5038 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
5039 * @skb: skb to check
5040 *
5041 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
5042 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
5043 * use this helper, to document places where we make this assertion.
5044 */
05bdd2f1 5045static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c 5046{
0df65743 5047 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
bc8acf2c
ED
5048}
5049
f35d9d8a 5050bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 5051
ed1f50c3 5052int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
5053struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
5054 unsigned int transport_len,
5055 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 5056
3a7c1ee4
AD
5057/**
5058 * skb_head_is_locked - Determine if the skb->head is locked down
5059 * @skb: skb to check
5060 *
5061 * The head on skbs build around a head frag can be removed if they are
5062 * not cloned. This function returns true if the skb head is locked down
5063 * due to either being allocated via kmalloc, or by being a clone with
5064 * multiple references to the head.
5065 */
5066static inline bool skb_head_is_locked(const struct sk_buff *skb)
5067{
5068 return !skb->head_frag || skb_cloned(skb);
5069}
fe6cc55f 5070
179bc67f
EC
5071/* Local Checksum Offload.
5072 * Compute outer checksum based on the assumption that the
5073 * inner checksum will be offloaded later.
d0dcde64 5074 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 5075 * explanation of how this works.
179bc67f
EC
5076 * Fill in outer checksum adjustment (e.g. with sum of outer
5077 * pseudo-header) before calling.
5078 * Also ensure that inner checksum is in linear data area.
5079 */
5080static inline __wsum lco_csum(struct sk_buff *skb)
5081{
9e74a6da
AD
5082 unsigned char *csum_start = skb_checksum_start(skb);
5083 unsigned char *l4_hdr = skb_transport_header(skb);
5084 __wsum partial;
179bc67f
EC
5085
5086 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
5087 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
5088 skb->csum_offset));
5089
179bc67f 5090 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 5091 * adjustment filled in by caller) and return result.
179bc67f 5092 */
9e74a6da 5093 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
5094}
5095
2c64605b
PNA
5096static inline bool skb_is_redirected(const struct sk_buff *skb)
5097{
2c64605b 5098 return skb->redirected;
2c64605b
PNA
5099}
5100
5101static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5102{
2c64605b 5103 skb->redirected = 1;
11941f8a 5104#ifdef CONFIG_NET_REDIRECT
2c64605b
PNA
5105 skb->from_ingress = from_ingress;
5106 if (skb->from_ingress)
de799101 5107 skb_clear_tstamp(skb);
2c64605b
PNA
5108#endif
5109}
5110
5111static inline void skb_reset_redirect(struct sk_buff *skb)
5112{
2c64605b 5113 skb->redirected = 0;
2c64605b
PNA
5114}
5115
fa821170
XL
5116static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5117{
5118 return skb->csum_not_inet;
5119}
5120
6370cc3b
AN
5121static inline void skb_set_kcov_handle(struct sk_buff *skb,
5122 const u64 kcov_handle)
5123{
fa69ee5a
ME
5124#ifdef CONFIG_KCOV
5125 skb->kcov_handle = kcov_handle;
5126#endif
6370cc3b
AN
5127}
5128
5129static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5130{
fa69ee5a
ME
5131#ifdef CONFIG_KCOV
5132 return skb->kcov_handle;
6370cc3b 5133#else
fa69ee5a
ME
5134 return 0;
5135#endif
5136}
6370cc3b 5137
6a5bcd84 5138#ifdef CONFIG_PAGE_POOL
57f05bc2 5139static inline void skb_mark_for_recycle(struct sk_buff *skb)
6a5bcd84
IA
5140{
5141 skb->pp_recycle = 1;
6a5bcd84
IA
5142}
5143#endif
5144
5145static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
5146{
5147 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
5148 return false;
5149 return page_pool_return_skb_page(virt_to_page(data));
5150}
5151
1da177e4
LT
5152#endif /* __KERNEL__ */
5153#endif /* _LINUX_SKBUFF_H */