]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
net: Remove superfluous setting of key_basic
[mirror_ubuntu-bionic-kernel.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
1bd758eb 37#include <net/flow_dissector.h>
a60e3cc7 38#include <linux/splice.h>
1da177e4 39
78ea85f1
DB
40/* A. Checksumming of received packets by device.
41 *
42 * CHECKSUM_NONE:
43 *
44 * Device failed to checksum this packet e.g. due to lack of capabilities.
45 * The packet contains full (though not verified) checksum in packet but
46 * not in skb->csum. Thus, skb->csum is undefined in this case.
47 *
48 * CHECKSUM_UNNECESSARY:
49 *
50 * The hardware you're dealing with doesn't calculate the full checksum
51 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
52 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
53 * if their checksums are okay. skb->csum is still undefined in this case
54 * though. It is a bad option, but, unfortunately, nowadays most vendors do
55 * this. Apparently with the secret goal to sell you new devices, when you
56 * will add new protocol to your host, f.e. IPv6 8)
57 *
58 * CHECKSUM_UNNECESSARY is applicable to following protocols:
59 * TCP: IPv6 and IPv4.
60 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
61 * zero UDP checksum for either IPv4 or IPv6, the networking stack
62 * may perform further validation in this case.
63 * GRE: only if the checksum is present in the header.
64 * SCTP: indicates the CRC in SCTP header has been validated.
65 *
66 * skb->csum_level indicates the number of consecutive checksums found in
67 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
68 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
69 * and a device is able to verify the checksums for UDP (possibly zero),
70 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
71 * two. If the device were only able to verify the UDP checksum and not
72 * GRE, either because it doesn't support GRE checksum of because GRE
73 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
74 * not considered in this case).
78ea85f1
DB
75 *
76 * CHECKSUM_COMPLETE:
77 *
78 * This is the most generic way. The device supplied checksum of the _whole_
79 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
80 * hardware doesn't need to parse L3/L4 headers to implement this.
81 *
82 * Note: Even if device supports only some protocols, but is able to produce
83 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
84 *
85 * CHECKSUM_PARTIAL:
86 *
6edec0e6
TH
87 * A checksum is set up to be offloaded to a device as described in the
88 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 89 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
90 * on the same host, or it may be set in the input path in GRO or remote
91 * checksum offload. For the purposes of checksum verification, the checksum
92 * referred to by skb->csum_start + skb->csum_offset and any preceding
93 * checksums in the packet are considered verified. Any checksums in the
94 * packet that are after the checksum being offloaded are not considered to
95 * be verified.
78ea85f1
DB
96 *
97 * B. Checksumming on output.
98 *
99 * CHECKSUM_NONE:
100 *
101 * The skb was already checksummed by the protocol, or a checksum is not
102 * required.
103 *
104 * CHECKSUM_PARTIAL:
105 *
106 * The device is required to checksum the packet as seen by hard_start_xmit()
107 * from skb->csum_start up to the end, and to record/write the checksum at
108 * offset skb->csum_start + skb->csum_offset.
109 *
110 * The device must show its capabilities in dev->features, set up at device
111 * setup time, e.g. netdev_features.h:
112 *
113 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
114 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
115 * IPv4. Sigh. Vendors like this way for an unknown reason.
116 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
117 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
118 * NETIF_F_... - Well, you get the picture.
119 *
120 * CHECKSUM_UNNECESSARY:
121 *
122 * Normally, the device will do per protocol specific checksumming. Protocol
123 * implementations that do not want the NIC to perform the checksum
124 * calculation should use this flag in their outgoing skbs.
125 *
126 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
127 * offload. Correspondingly, the FCoE protocol driver
128 * stack should use CHECKSUM_UNNECESSARY.
129 *
130 * Any questions? No questions, good. --ANK
131 */
132
60476372 133/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
134#define CHECKSUM_NONE 0
135#define CHECKSUM_UNNECESSARY 1
136#define CHECKSUM_COMPLETE 2
137#define CHECKSUM_PARTIAL 3
1da177e4 138
77cffe23
TH
139/* Maximum value in skb->csum_level */
140#define SKB_MAX_CSUM_LEVEL 3
141
0bec8c88 142#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 143#define SKB_WITH_OVERHEAD(X) \
deea84b0 144 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
145#define SKB_MAX_ORDER(X, ORDER) \
146 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
147#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
148#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
149
87fb4b7b
ED
150/* return minimum truesize of one skb containing X bytes of data */
151#define SKB_TRUESIZE(X) ((X) + \
152 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
153 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
154
1da177e4 155struct net_device;
716ea3a7 156struct scatterlist;
9c55e01c 157struct pipe_inode_info;
a8f820aa 158struct iov_iter;
fd11a83d 159struct napi_struct;
1da177e4 160
5f79e0f9 161#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
162struct nf_conntrack {
163 atomic_t use;
1da177e4 164};
5f79e0f9 165#endif
1da177e4 166
34666d46 167#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 168struct nf_bridge_info {
bf1ac5ca 169 atomic_t use;
3eaf4025
FW
170 enum {
171 BRNF_PROTO_UNCHANGED,
172 BRNF_PROTO_8021Q,
173 BRNF_PROTO_PPPOE
7fb48c5b 174 } orig_proto:8;
a1e67951 175 bool pkt_otherhost;
bf1ac5ca
ED
176 unsigned int mask;
177 struct net_device *physindev;
7fb48c5b
FW
178 union {
179 struct net_device *physoutdev;
180 char neigh_header[8];
181 };
faecbb45 182 __be32 ipv4_daddr;
1da177e4
LT
183};
184#endif
185
1da177e4
LT
186struct sk_buff_head {
187 /* These two members must be first. */
188 struct sk_buff *next;
189 struct sk_buff *prev;
190
191 __u32 qlen;
192 spinlock_t lock;
193};
194
195struct sk_buff;
196
9d4dde52
IC
197/* To allow 64K frame to be packed as single skb without frag_list we
198 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
199 * buffers which do not start on a page boundary.
200 *
201 * Since GRO uses frags we allocate at least 16 regardless of page
202 * size.
a715dea3 203 */
9d4dde52 204#if (65536/PAGE_SIZE + 1) < 16
eec00954 205#define MAX_SKB_FRAGS 16UL
a715dea3 206#else
9d4dde52 207#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 208#endif
1da177e4
LT
209
210typedef struct skb_frag_struct skb_frag_t;
211
212struct skb_frag_struct {
a8605c60
IC
213 struct {
214 struct page *p;
215 } page;
cb4dfe56 216#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
217 __u32 page_offset;
218 __u32 size;
cb4dfe56
ED
219#else
220 __u16 page_offset;
221 __u16 size;
222#endif
1da177e4
LT
223};
224
9e903e08
ED
225static inline unsigned int skb_frag_size(const skb_frag_t *frag)
226{
227 return frag->size;
228}
229
230static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
231{
232 frag->size = size;
233}
234
235static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
236{
237 frag->size += delta;
238}
239
240static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
241{
242 frag->size -= delta;
243}
244
ac45f602
PO
245#define HAVE_HW_TIME_STAMP
246
247/**
d3a21be8 248 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
249 * @hwtstamp: hardware time stamp transformed into duration
250 * since arbitrary point in time
ac45f602
PO
251 *
252 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 253 * skb->tstamp.
ac45f602
PO
254 *
255 * hwtstamps can only be compared against other hwtstamps from
256 * the same device.
257 *
258 * This structure is attached to packets as part of the
259 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
260 */
261struct skb_shared_hwtstamps {
262 ktime_t hwtstamp;
ac45f602
PO
263};
264
2244d07b
OH
265/* Definitions for tx_flags in struct skb_shared_info */
266enum {
267 /* generate hardware time stamp */
268 SKBTX_HW_TSTAMP = 1 << 0,
269
e7fd2885 270 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
271 SKBTX_SW_TSTAMP = 1 << 1,
272
273 /* device driver is going to provide hardware time stamp */
274 SKBTX_IN_PROGRESS = 1 << 2,
275
a6686f2f 276 /* device driver supports TX zero-copy buffers */
62b1a8ab 277 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
278
279 /* generate wifi status information (where possible) */
62b1a8ab 280 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
281
282 /* This indicates at least one fragment might be overwritten
283 * (as in vmsplice(), sendfile() ...)
284 * If we need to compute a TX checksum, we'll need to copy
285 * all frags to avoid possible bad checksum
286 */
287 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
288
289 /* generate software time stamp when entering packet scheduling */
290 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
291
292 /* generate software timestamp on peer data acknowledgment */
293 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
294};
295
e1c8a607
WB
296#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
297 SKBTX_SCHED_TSTAMP | \
298 SKBTX_ACK_TSTAMP)
f24b9be5
WB
299#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
300
a6686f2f
SM
301/*
302 * The callback notifies userspace to release buffers when skb DMA is done in
303 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
304 * The zerocopy_success argument is true if zero copy transmit occurred,
305 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
306 * The ctx field is used to track device context.
307 * The desc field is used to track userspace buffer index.
a6686f2f
SM
308 */
309struct ubuf_info {
e19d6763 310 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 311 void *ctx;
a6686f2f 312 unsigned long desc;
ac45f602
PO
313};
314
1da177e4
LT
315/* This data is invariant across clones and lives at
316 * the end of the header data, ie. at skb->end.
317 */
318struct skb_shared_info {
9f42f126
IC
319 unsigned char nr_frags;
320 __u8 tx_flags;
7967168c
HX
321 unsigned short gso_size;
322 /* Warning: this field is not always filled in (UFO)! */
323 unsigned short gso_segs;
324 unsigned short gso_type;
1da177e4 325 struct sk_buff *frag_list;
ac45f602 326 struct skb_shared_hwtstamps hwtstamps;
09c2d251 327 u32 tskey;
9f42f126 328 __be32 ip6_frag_id;
ec7d2f2c
ED
329
330 /*
331 * Warning : all fields before dataref are cleared in __alloc_skb()
332 */
333 atomic_t dataref;
334
69e3c75f
JB
335 /* Intermediate layers must ensure that destructor_arg
336 * remains valid until skb destructor */
337 void * destructor_arg;
a6686f2f 338
fed66381
ED
339 /* must be last field, see pskb_expand_head() */
340 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
341};
342
343/* We divide dataref into two halves. The higher 16 bits hold references
344 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
345 * the entire skb->data. A clone of a headerless skb holds the length of
346 * the header in skb->hdr_len.
1da177e4
LT
347 *
348 * All users must obey the rule that the skb->data reference count must be
349 * greater than or equal to the payload reference count.
350 *
351 * Holding a reference to the payload part means that the user does not
352 * care about modifications to the header part of skb->data.
353 */
354#define SKB_DATAREF_SHIFT 16
355#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
356
d179cd12
DM
357
358enum {
c8753d55
VS
359 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
360 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
361 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
362};
363
7967168c
HX
364enum {
365 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 366 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
367
368 /* This indicates the skb is from an untrusted source. */
369 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
370
371 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
372 SKB_GSO_TCP_ECN = 1 << 3,
373
374 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
375
376 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
377
378 SKB_GSO_GRE = 1 << 6,
73136267 379
4b28252c 380 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 381
4b28252c 382 SKB_GSO_IPIP = 1 << 8,
cb32f511 383
4b28252c 384 SKB_GSO_SIT = 1 << 9,
61c1db7f 385
4b28252c 386 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
387
388 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 389
59b93b41 390 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
391};
392
2e07fa9c
ACM
393#if BITS_PER_LONG > 32
394#define NET_SKBUFF_DATA_USES_OFFSET 1
395#endif
396
397#ifdef NET_SKBUFF_DATA_USES_OFFSET
398typedef unsigned int sk_buff_data_t;
399#else
400typedef unsigned char *sk_buff_data_t;
401#endif
402
363ec392
ED
403/**
404 * struct skb_mstamp - multi resolution time stamps
405 * @stamp_us: timestamp in us resolution
406 * @stamp_jiffies: timestamp in jiffies
407 */
408struct skb_mstamp {
409 union {
410 u64 v64;
411 struct {
412 u32 stamp_us;
413 u32 stamp_jiffies;
414 };
415 };
416};
417
418/**
419 * skb_mstamp_get - get current timestamp
420 * @cl: place to store timestamps
421 */
422static inline void skb_mstamp_get(struct skb_mstamp *cl)
423{
424 u64 val = local_clock();
425
426 do_div(val, NSEC_PER_USEC);
427 cl->stamp_us = (u32)val;
428 cl->stamp_jiffies = (u32)jiffies;
429}
430
431/**
432 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
433 * @t1: pointer to newest sample
434 * @t0: pointer to oldest sample
435 */
436static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
437 const struct skb_mstamp *t0)
438{
439 s32 delta_us = t1->stamp_us - t0->stamp_us;
440 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
441
442 /* If delta_us is negative, this might be because interval is too big,
443 * or local_clock() drift is too big : fallback using jiffies.
444 */
445 if (delta_us <= 0 ||
446 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
447
448 delta_us = jiffies_to_usecs(delta_jiffies);
449
450 return delta_us;
451}
452
453
1da177e4
LT
454/**
455 * struct sk_buff - socket buffer
456 * @next: Next buffer in list
457 * @prev: Previous buffer in list
363ec392 458 * @tstamp: Time we arrived/left
56b17425 459 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 460 * @sk: Socket we are owned by
1da177e4 461 * @dev: Device we arrived on/are leaving by
d84e0bd7 462 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 463 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 464 * @sp: the security path, used for xfrm
1da177e4
LT
465 * @len: Length of actual data
466 * @data_len: Data length
467 * @mac_len: Length of link layer header
334a8132 468 * @hdr_len: writable header length of cloned skb
663ead3b
HX
469 * @csum: Checksum (must include start/offset pair)
470 * @csum_start: Offset from skb->head where checksumming should start
471 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 472 * @priority: Packet queueing priority
60ff7467 473 * @ignore_df: allow local fragmentation
1da177e4 474 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 475 * @ip_summed: Driver fed us an IP checksum
1da177e4 476 * @nohdr: Payload reference only, must not modify header
d84e0bd7 477 * @nfctinfo: Relationship of this skb to the connection
1da177e4 478 * @pkt_type: Packet class
c83c2486 479 * @fclone: skbuff clone status
c83c2486 480 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
481 * @peeked: this packet has been seen already, so stats have been
482 * done for it, don't do them again
ba9dda3a 483 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
484 * @protocol: Packet protocol from driver
485 * @destructor: Destruct function
486 * @nfct: Associated connection, if any
1da177e4 487 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 488 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
489 * @tc_index: Traffic control index
490 * @tc_verd: traffic control verdict
61b905da 491 * @hash: the packet hash
d84e0bd7 492 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 493 * @xmit_more: More SKBs are pending for this queue
553a5672 494 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 495 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 496 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 497 * ports.
a3b18ddb 498 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
499 * @wifi_acked_valid: wifi_acked was set
500 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 501 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 502 * @napi_id: id of the NAPI struct this skb came from
984bc16c 503 * @secmark: security marking
d84e0bd7 504 * @mark: Generic packet mark
86a9bad3 505 * @vlan_proto: vlan encapsulation protocol
6aa895b0 506 * @vlan_tci: vlan tag control information
0d89d203 507 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
508 * @inner_transport_header: Inner transport layer header (encapsulation)
509 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 510 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
511 * @transport_header: Transport layer header
512 * @network_header: Network layer header
513 * @mac_header: Link layer header
514 * @tail: Tail pointer
515 * @end: End pointer
516 * @head: Head of buffer
517 * @data: Data head pointer
518 * @truesize: Buffer size
519 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
520 */
521
522struct sk_buff {
363ec392 523 union {
56b17425
ED
524 struct {
525 /* These two members must be first. */
526 struct sk_buff *next;
527 struct sk_buff *prev;
528
529 union {
530 ktime_t tstamp;
531 struct skb_mstamp skb_mstamp;
532 };
533 };
534 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 535 };
da3f5cf1 536 struct sock *sk;
1da177e4 537 struct net_device *dev;
1da177e4 538
1da177e4
LT
539 /*
540 * This is the control buffer. It is free to use for every
541 * layer. Please put your private variables there. If you
542 * want to keep them across layers you have to do a skb_clone()
543 * first. This is owned by whoever has the skb queued ATM.
544 */
da3f5cf1 545 char cb[48] __aligned(8);
1da177e4 546
7fee226a 547 unsigned long _skb_refdst;
b1937227 548 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
549#ifdef CONFIG_XFRM
550 struct sec_path *sp;
b1937227
ED
551#endif
552#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
553 struct nf_conntrack *nfct;
554#endif
85224844 555#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 556 struct nf_bridge_info *nf_bridge;
da3f5cf1 557#endif
1da177e4 558 unsigned int len,
334a8132
PM
559 data_len;
560 __u16 mac_len,
561 hdr_len;
b1937227
ED
562
563 /* Following fields are _not_ copied in __copy_skb_header()
564 * Note that queue_mapping is here mostly to fill a hole.
565 */
fe55f6d5 566 kmemcheck_bitfield_begin(flags1);
b1937227
ED
567 __u16 queue_mapping;
568 __u8 cloned:1,
6869c4d8 569 nohdr:1,
b84f4cc9 570 fclone:2,
a59322be 571 peeked:1,
b1937227
ED
572 head_frag:1,
573 xmit_more:1;
574 /* one bit hole */
fe55f6d5 575 kmemcheck_bitfield_end(flags1);
4031ae6e 576
b1937227
ED
577 /* fields enclosed in headers_start/headers_end are copied
578 * using a single memcpy() in __copy_skb_header()
579 */
ebcf34f3 580 /* private: */
b1937227 581 __u32 headers_start[0];
ebcf34f3 582 /* public: */
4031ae6e 583
233577a2
HFS
584/* if you move pkt_type around you also must adapt those constants */
585#ifdef __BIG_ENDIAN_BITFIELD
586#define PKT_TYPE_MAX (7 << 5)
587#else
588#define PKT_TYPE_MAX 7
1da177e4 589#endif
233577a2 590#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 591
233577a2 592 __u8 __pkt_type_offset[0];
b1937227 593 __u8 pkt_type:3;
c93bdd0e 594 __u8 pfmemalloc:1;
b1937227
ED
595 __u8 ignore_df:1;
596 __u8 nfctinfo:3;
597
598 __u8 nf_trace:1;
599 __u8 ip_summed:2;
3853b584 600 __u8 ooo_okay:1;
61b905da 601 __u8 l4_hash:1;
a3b18ddb 602 __u8 sw_hash:1;
6e3e939f
JB
603 __u8 wifi_acked_valid:1;
604 __u8 wifi_acked:1;
b1937227 605
3bdc0eba 606 __u8 no_fcs:1;
77cffe23 607 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 608 __u8 encapsulation:1;
7e2b10c1 609 __u8 encap_hdr_csum:1;
5d0c2b95 610 __u8 csum_valid:1;
7e3cead5 611 __u8 csum_complete_sw:1;
b1937227
ED
612 __u8 csum_level:2;
613 __u8 csum_bad:1;
fe55f6d5 614
b1937227
ED
615#ifdef CONFIG_IPV6_NDISC_NODETYPE
616 __u8 ndisc_nodetype:2;
617#endif
618 __u8 ipvs_property:1;
8bce6d7d 619 __u8 inner_protocol_type:1;
e585f236
TH
620 __u8 remcsum_offload:1;
621 /* 3 or 5 bit hole */
b1937227
ED
622
623#ifdef CONFIG_NET_SCHED
624 __u16 tc_index; /* traffic control index */
625#ifdef CONFIG_NET_CLS_ACT
626 __u16 tc_verd; /* traffic control verdict */
627#endif
628#endif
fe55f6d5 629
b1937227
ED
630 union {
631 __wsum csum;
632 struct {
633 __u16 csum_start;
634 __u16 csum_offset;
635 };
636 };
637 __u32 priority;
638 int skb_iif;
639 __u32 hash;
640 __be16 vlan_proto;
641 __u16 vlan_tci;
2bd82484
ED
642#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
643 union {
644 unsigned int napi_id;
645 unsigned int sender_cpu;
646 };
97fc2f08 647#endif
984bc16c
JM
648#ifdef CONFIG_NETWORK_SECMARK
649 __u32 secmark;
650#endif
3b885787
NH
651 union {
652 __u32 mark;
16fad69c 653 __u32 reserved_tailroom;
3b885787 654 };
1da177e4 655
8bce6d7d
TH
656 union {
657 __be16 inner_protocol;
658 __u8 inner_ipproto;
659 };
660
1a37e412
SH
661 __u16 inner_transport_header;
662 __u16 inner_network_header;
663 __u16 inner_mac_header;
b1937227
ED
664
665 __be16 protocol;
1a37e412
SH
666 __u16 transport_header;
667 __u16 network_header;
668 __u16 mac_header;
b1937227 669
ebcf34f3 670 /* private: */
b1937227 671 __u32 headers_end[0];
ebcf34f3 672 /* public: */
b1937227 673
1da177e4 674 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 675 sk_buff_data_t tail;
4305b541 676 sk_buff_data_t end;
1da177e4 677 unsigned char *head,
4305b541 678 *data;
27a884dc
ACM
679 unsigned int truesize;
680 atomic_t users;
1da177e4
LT
681};
682
683#ifdef __KERNEL__
684/*
685 * Handling routines are only of interest to the kernel
686 */
687#include <linux/slab.h>
688
1da177e4 689
c93bdd0e
MG
690#define SKB_ALLOC_FCLONE 0x01
691#define SKB_ALLOC_RX 0x02
fd11a83d 692#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
693
694/* Returns true if the skb was allocated from PFMEMALLOC reserves */
695static inline bool skb_pfmemalloc(const struct sk_buff *skb)
696{
697 return unlikely(skb->pfmemalloc);
698}
699
7fee226a
ED
700/*
701 * skb might have a dst pointer attached, refcounted or not.
702 * _skb_refdst low order bit is set if refcount was _not_ taken
703 */
704#define SKB_DST_NOREF 1UL
705#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
706
707/**
708 * skb_dst - returns skb dst_entry
709 * @skb: buffer
710 *
711 * Returns skb dst_entry, regardless of reference taken or not.
712 */
adf30907
ED
713static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
714{
7fee226a
ED
715 /* If refdst was not refcounted, check we still are in a
716 * rcu_read_lock section
717 */
718 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
719 !rcu_read_lock_held() &&
720 !rcu_read_lock_bh_held());
721 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
722}
723
7fee226a
ED
724/**
725 * skb_dst_set - sets skb dst
726 * @skb: buffer
727 * @dst: dst entry
728 *
729 * Sets skb dst, assuming a reference was taken on dst and should
730 * be released by skb_dst_drop()
731 */
adf30907
ED
732static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
733{
7fee226a
ED
734 skb->_skb_refdst = (unsigned long)dst;
735}
736
932bc4d7
JA
737/**
738 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
739 * @skb: buffer
740 * @dst: dst entry
741 *
742 * Sets skb dst, assuming a reference was not taken on dst.
743 * If dst entry is cached, we do not take reference and dst_release
744 * will be avoided by refdst_drop. If dst entry is not cached, we take
745 * reference, so that last dst_release can destroy the dst immediately.
746 */
747static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
748{
dbfc4fb7
HFS
749 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
750 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 751}
7fee226a
ED
752
753/**
25985edc 754 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
755 * @skb: buffer
756 */
757static inline bool skb_dst_is_noref(const struct sk_buff *skb)
758{
759 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
760}
761
511c3f92
ED
762static inline struct rtable *skb_rtable(const struct sk_buff *skb)
763{
adf30907 764 return (struct rtable *)skb_dst(skb);
511c3f92
ED
765}
766
7965bd4d
JP
767void kfree_skb(struct sk_buff *skb);
768void kfree_skb_list(struct sk_buff *segs);
769void skb_tx_error(struct sk_buff *skb);
770void consume_skb(struct sk_buff *skb);
771void __kfree_skb(struct sk_buff *skb);
d7e8883c 772extern struct kmem_cache *skbuff_head_cache;
bad43ca8 773
7965bd4d
JP
774void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
775bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
776 bool *fragstolen, int *delta_truesize);
bad43ca8 777
7965bd4d
JP
778struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
779 int node);
2ea2f62c 780struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 781struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 782static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 783 gfp_t priority)
d179cd12 784{
564824b0 785 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
786}
787
2e4e4410
ED
788struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
789 unsigned long data_len,
790 int max_page_order,
791 int *errcode,
792 gfp_t gfp_mask);
793
d0bf4a9e
ED
794/* Layout of fast clones : [skb1][skb2][fclone_ref] */
795struct sk_buff_fclones {
796 struct sk_buff skb1;
797
798 struct sk_buff skb2;
799
800 atomic_t fclone_ref;
801};
802
803/**
804 * skb_fclone_busy - check if fclone is busy
805 * @skb: buffer
806 *
807 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
808 * Some drivers call skb_orphan() in their ndo_start_xmit(),
809 * so we also check that this didnt happen.
d0bf4a9e 810 */
39bb5e62
ED
811static inline bool skb_fclone_busy(const struct sock *sk,
812 const struct sk_buff *skb)
d0bf4a9e
ED
813{
814 const struct sk_buff_fclones *fclones;
815
816 fclones = container_of(skb, struct sk_buff_fclones, skb1);
817
818 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 819 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 820 fclones->skb2.sk == sk;
d0bf4a9e
ED
821}
822
d179cd12 823static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 824 gfp_t priority)
d179cd12 825{
c93bdd0e 826 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
827}
828
7965bd4d 829struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
830static inline struct sk_buff *alloc_skb_head(gfp_t priority)
831{
832 return __alloc_skb_head(priority, -1);
833}
834
7965bd4d
JP
835struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
836int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
837struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
838struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
839struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
840 gfp_t gfp_mask, bool fclone);
841static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
842 gfp_t gfp_mask)
843{
844 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
845}
7965bd4d
JP
846
847int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
848struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
849 unsigned int headroom);
850struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
851 int newtailroom, gfp_t priority);
25a91d8d
FD
852int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
853 int offset, int len);
7965bd4d
JP
854int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
855 int len);
856int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
857int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 858#define dev_kfree_skb(a) consume_skb(a)
1da177e4 859
7965bd4d
JP
860int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
861 int getfrag(void *from, char *to, int offset,
862 int len, int odd, struct sk_buff *skb),
863 void *from, int length);
e89e9cf5 864
be12a1fe
HFS
865int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
866 int offset, size_t size);
867
d94d9fee 868struct skb_seq_state {
677e90ed
TG
869 __u32 lower_offset;
870 __u32 upper_offset;
871 __u32 frag_idx;
872 __u32 stepped_offset;
873 struct sk_buff *root_skb;
874 struct sk_buff *cur_skb;
875 __u8 *frag_data;
876};
877
7965bd4d
JP
878void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
879 unsigned int to, struct skb_seq_state *st);
880unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
881 struct skb_seq_state *st);
882void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 883
7965bd4d 884unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 885 unsigned int to, struct ts_config *config);
3fc7e8a6 886
09323cc4
TH
887/*
888 * Packet hash types specify the type of hash in skb_set_hash.
889 *
890 * Hash types refer to the protocol layer addresses which are used to
891 * construct a packet's hash. The hashes are used to differentiate or identify
892 * flows of the protocol layer for the hash type. Hash types are either
893 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
894 *
895 * Properties of hashes:
896 *
897 * 1) Two packets in different flows have different hash values
898 * 2) Two packets in the same flow should have the same hash value
899 *
900 * A hash at a higher layer is considered to be more specific. A driver should
901 * set the most specific hash possible.
902 *
903 * A driver cannot indicate a more specific hash than the layer at which a hash
904 * was computed. For instance an L3 hash cannot be set as an L4 hash.
905 *
906 * A driver may indicate a hash level which is less specific than the
907 * actual layer the hash was computed on. For instance, a hash computed
908 * at L4 may be considered an L3 hash. This should only be done if the
909 * driver can't unambiguously determine that the HW computed the hash at
910 * the higher layer. Note that the "should" in the second property above
911 * permits this.
912 */
913enum pkt_hash_types {
914 PKT_HASH_TYPE_NONE, /* Undefined type */
915 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
916 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
917 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
918};
919
920static inline void
921skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
922{
61b905da 923 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 924 skb->sw_hash = 0;
61b905da 925 skb->hash = hash;
09323cc4
TH
926}
927
3958afa1 928static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 929{
a3b18ddb 930 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 931 __skb_get_hash(skb);
bfb564e7 932
61b905da 933 return skb->hash;
bfb564e7
KK
934}
935
50fb7992
TH
936__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
937
57bdf7f4
TH
938static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
939{
61b905da 940 return skb->hash;
57bdf7f4
TH
941}
942
7539fadc
TH
943static inline void skb_clear_hash(struct sk_buff *skb)
944{
61b905da 945 skb->hash = 0;
a3b18ddb 946 skb->sw_hash = 0;
61b905da 947 skb->l4_hash = 0;
7539fadc
TH
948}
949
950static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
951{
61b905da 952 if (!skb->l4_hash)
7539fadc
TH
953 skb_clear_hash(skb);
954}
955
3df7a74e
TH
956static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
957{
61b905da 958 to->hash = from->hash;
a3b18ddb 959 to->sw_hash = from->sw_hash;
61b905da 960 to->l4_hash = from->l4_hash;
3df7a74e
TH
961};
962
c29390c6
ED
963static inline void skb_sender_cpu_clear(struct sk_buff *skb)
964{
965#ifdef CONFIG_XPS
966 skb->sender_cpu = 0;
967#endif
968}
969
4305b541
ACM
970#ifdef NET_SKBUFF_DATA_USES_OFFSET
971static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
972{
973 return skb->head + skb->end;
974}
ec47ea82
AD
975
976static inline unsigned int skb_end_offset(const struct sk_buff *skb)
977{
978 return skb->end;
979}
4305b541
ACM
980#else
981static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
982{
983 return skb->end;
984}
ec47ea82
AD
985
986static inline unsigned int skb_end_offset(const struct sk_buff *skb)
987{
988 return skb->end - skb->head;
989}
4305b541
ACM
990#endif
991
1da177e4 992/* Internal */
4305b541 993#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 994
ac45f602
PO
995static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
996{
997 return &skb_shinfo(skb)->hwtstamps;
998}
999
1da177e4
LT
1000/**
1001 * skb_queue_empty - check if a queue is empty
1002 * @list: queue head
1003 *
1004 * Returns true if the queue is empty, false otherwise.
1005 */
1006static inline int skb_queue_empty(const struct sk_buff_head *list)
1007{
fd44b93c 1008 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1009}
1010
fc7ebb21
DM
1011/**
1012 * skb_queue_is_last - check if skb is the last entry in the queue
1013 * @list: queue head
1014 * @skb: buffer
1015 *
1016 * Returns true if @skb is the last buffer on the list.
1017 */
1018static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1019 const struct sk_buff *skb)
1020{
fd44b93c 1021 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1022}
1023
832d11c5
IJ
1024/**
1025 * skb_queue_is_first - check if skb is the first entry in the queue
1026 * @list: queue head
1027 * @skb: buffer
1028 *
1029 * Returns true if @skb is the first buffer on the list.
1030 */
1031static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1032 const struct sk_buff *skb)
1033{
fd44b93c 1034 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1035}
1036
249c8b42
DM
1037/**
1038 * skb_queue_next - return the next packet in the queue
1039 * @list: queue head
1040 * @skb: current buffer
1041 *
1042 * Return the next packet in @list after @skb. It is only valid to
1043 * call this if skb_queue_is_last() evaluates to false.
1044 */
1045static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1046 const struct sk_buff *skb)
1047{
1048 /* This BUG_ON may seem severe, but if we just return then we
1049 * are going to dereference garbage.
1050 */
1051 BUG_ON(skb_queue_is_last(list, skb));
1052 return skb->next;
1053}
1054
832d11c5
IJ
1055/**
1056 * skb_queue_prev - return the prev packet in the queue
1057 * @list: queue head
1058 * @skb: current buffer
1059 *
1060 * Return the prev packet in @list before @skb. It is only valid to
1061 * call this if skb_queue_is_first() evaluates to false.
1062 */
1063static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1064 const struct sk_buff *skb)
1065{
1066 /* This BUG_ON may seem severe, but if we just return then we
1067 * are going to dereference garbage.
1068 */
1069 BUG_ON(skb_queue_is_first(list, skb));
1070 return skb->prev;
1071}
1072
1da177e4
LT
1073/**
1074 * skb_get - reference buffer
1075 * @skb: buffer to reference
1076 *
1077 * Makes another reference to a socket buffer and returns a pointer
1078 * to the buffer.
1079 */
1080static inline struct sk_buff *skb_get(struct sk_buff *skb)
1081{
1082 atomic_inc(&skb->users);
1083 return skb;
1084}
1085
1086/*
1087 * If users == 1, we are the only owner and are can avoid redundant
1088 * atomic change.
1089 */
1090
1da177e4
LT
1091/**
1092 * skb_cloned - is the buffer a clone
1093 * @skb: buffer to check
1094 *
1095 * Returns true if the buffer was generated with skb_clone() and is
1096 * one of multiple shared copies of the buffer. Cloned buffers are
1097 * shared data so must not be written to under normal circumstances.
1098 */
1099static inline int skb_cloned(const struct sk_buff *skb)
1100{
1101 return skb->cloned &&
1102 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1103}
1104
14bbd6a5
PS
1105static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1106{
1107 might_sleep_if(pri & __GFP_WAIT);
1108
1109 if (skb_cloned(skb))
1110 return pskb_expand_head(skb, 0, 0, pri);
1111
1112 return 0;
1113}
1114
1da177e4
LT
1115/**
1116 * skb_header_cloned - is the header a clone
1117 * @skb: buffer to check
1118 *
1119 * Returns true if modifying the header part of the buffer requires
1120 * the data to be copied.
1121 */
1122static inline int skb_header_cloned(const struct sk_buff *skb)
1123{
1124 int dataref;
1125
1126 if (!skb->cloned)
1127 return 0;
1128
1129 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1130 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1131 return dataref != 1;
1132}
1133
1134/**
1135 * skb_header_release - release reference to header
1136 * @skb: buffer to operate on
1137 *
1138 * Drop a reference to the header part of the buffer. This is done
1139 * by acquiring a payload reference. You must not read from the header
1140 * part of skb->data after this.
f4a775d1 1141 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1142 */
1143static inline void skb_header_release(struct sk_buff *skb)
1144{
1145 BUG_ON(skb->nohdr);
1146 skb->nohdr = 1;
1147 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1148}
1149
f4a775d1
ED
1150/**
1151 * __skb_header_release - release reference to header
1152 * @skb: buffer to operate on
1153 *
1154 * Variant of skb_header_release() assuming skb is private to caller.
1155 * We can avoid one atomic operation.
1156 */
1157static inline void __skb_header_release(struct sk_buff *skb)
1158{
1159 skb->nohdr = 1;
1160 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1161}
1162
1163
1da177e4
LT
1164/**
1165 * skb_shared - is the buffer shared
1166 * @skb: buffer to check
1167 *
1168 * Returns true if more than one person has a reference to this
1169 * buffer.
1170 */
1171static inline int skb_shared(const struct sk_buff *skb)
1172{
1173 return atomic_read(&skb->users) != 1;
1174}
1175
1176/**
1177 * skb_share_check - check if buffer is shared and if so clone it
1178 * @skb: buffer to check
1179 * @pri: priority for memory allocation
1180 *
1181 * If the buffer is shared the buffer is cloned and the old copy
1182 * drops a reference. A new clone with a single reference is returned.
1183 * If the buffer is not shared the original buffer is returned. When
1184 * being called from interrupt status or with spinlocks held pri must
1185 * be GFP_ATOMIC.
1186 *
1187 * NULL is returned on a memory allocation failure.
1188 */
47061bc4 1189static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1190{
1191 might_sleep_if(pri & __GFP_WAIT);
1192 if (skb_shared(skb)) {
1193 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1194
1195 if (likely(nskb))
1196 consume_skb(skb);
1197 else
1198 kfree_skb(skb);
1da177e4
LT
1199 skb = nskb;
1200 }
1201 return skb;
1202}
1203
1204/*
1205 * Copy shared buffers into a new sk_buff. We effectively do COW on
1206 * packets to handle cases where we have a local reader and forward
1207 * and a couple of other messy ones. The normal one is tcpdumping
1208 * a packet thats being forwarded.
1209 */
1210
1211/**
1212 * skb_unshare - make a copy of a shared buffer
1213 * @skb: buffer to check
1214 * @pri: priority for memory allocation
1215 *
1216 * If the socket buffer is a clone then this function creates a new
1217 * copy of the data, drops a reference count on the old copy and returns
1218 * the new copy with the reference count at 1. If the buffer is not a clone
1219 * the original buffer is returned. When called with a spinlock held or
1220 * from interrupt state @pri must be %GFP_ATOMIC
1221 *
1222 * %NULL is returned on a memory allocation failure.
1223 */
e2bf521d 1224static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1225 gfp_t pri)
1da177e4
LT
1226{
1227 might_sleep_if(pri & __GFP_WAIT);
1228 if (skb_cloned(skb)) {
1229 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1230
1231 /* Free our shared copy */
1232 if (likely(nskb))
1233 consume_skb(skb);
1234 else
1235 kfree_skb(skb);
1da177e4
LT
1236 skb = nskb;
1237 }
1238 return skb;
1239}
1240
1241/**
1a5778aa 1242 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1243 * @list_: list to peek at
1244 *
1245 * Peek an &sk_buff. Unlike most other operations you _MUST_
1246 * be careful with this one. A peek leaves the buffer on the
1247 * list and someone else may run off with it. You must hold
1248 * the appropriate locks or have a private queue to do this.
1249 *
1250 * Returns %NULL for an empty list or a pointer to the head element.
1251 * The reference count is not incremented and the reference is therefore
1252 * volatile. Use with caution.
1253 */
05bdd2f1 1254static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1255{
18d07000
ED
1256 struct sk_buff *skb = list_->next;
1257
1258 if (skb == (struct sk_buff *)list_)
1259 skb = NULL;
1260 return skb;
1da177e4
LT
1261}
1262
da5ef6e5
PE
1263/**
1264 * skb_peek_next - peek skb following the given one from a queue
1265 * @skb: skb to start from
1266 * @list_: list to peek at
1267 *
1268 * Returns %NULL when the end of the list is met or a pointer to the
1269 * next element. The reference count is not incremented and the
1270 * reference is therefore volatile. Use with caution.
1271 */
1272static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1273 const struct sk_buff_head *list_)
1274{
1275 struct sk_buff *next = skb->next;
18d07000 1276
da5ef6e5
PE
1277 if (next == (struct sk_buff *)list_)
1278 next = NULL;
1279 return next;
1280}
1281
1da177e4 1282/**
1a5778aa 1283 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1284 * @list_: list to peek at
1285 *
1286 * Peek an &sk_buff. Unlike most other operations you _MUST_
1287 * be careful with this one. A peek leaves the buffer on the
1288 * list and someone else may run off with it. You must hold
1289 * the appropriate locks or have a private queue to do this.
1290 *
1291 * Returns %NULL for an empty list or a pointer to the tail element.
1292 * The reference count is not incremented and the reference is therefore
1293 * volatile. Use with caution.
1294 */
05bdd2f1 1295static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1296{
18d07000
ED
1297 struct sk_buff *skb = list_->prev;
1298
1299 if (skb == (struct sk_buff *)list_)
1300 skb = NULL;
1301 return skb;
1302
1da177e4
LT
1303}
1304
1305/**
1306 * skb_queue_len - get queue length
1307 * @list_: list to measure
1308 *
1309 * Return the length of an &sk_buff queue.
1310 */
1311static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1312{
1313 return list_->qlen;
1314}
1315
67fed459
DM
1316/**
1317 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1318 * @list: queue to initialize
1319 *
1320 * This initializes only the list and queue length aspects of
1321 * an sk_buff_head object. This allows to initialize the list
1322 * aspects of an sk_buff_head without reinitializing things like
1323 * the spinlock. It can also be used for on-stack sk_buff_head
1324 * objects where the spinlock is known to not be used.
1325 */
1326static inline void __skb_queue_head_init(struct sk_buff_head *list)
1327{
1328 list->prev = list->next = (struct sk_buff *)list;
1329 list->qlen = 0;
1330}
1331
76f10ad0
AV
1332/*
1333 * This function creates a split out lock class for each invocation;
1334 * this is needed for now since a whole lot of users of the skb-queue
1335 * infrastructure in drivers have different locking usage (in hardirq)
1336 * than the networking core (in softirq only). In the long run either the
1337 * network layer or drivers should need annotation to consolidate the
1338 * main types of usage into 3 classes.
1339 */
1da177e4
LT
1340static inline void skb_queue_head_init(struct sk_buff_head *list)
1341{
1342 spin_lock_init(&list->lock);
67fed459 1343 __skb_queue_head_init(list);
1da177e4
LT
1344}
1345
c2ecba71
PE
1346static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1347 struct lock_class_key *class)
1348{
1349 skb_queue_head_init(list);
1350 lockdep_set_class(&list->lock, class);
1351}
1352
1da177e4 1353/*
bf299275 1354 * Insert an sk_buff on a list.
1da177e4
LT
1355 *
1356 * The "__skb_xxxx()" functions are the non-atomic ones that
1357 * can only be called with interrupts disabled.
1358 */
7965bd4d
JP
1359void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1360 struct sk_buff_head *list);
bf299275
GR
1361static inline void __skb_insert(struct sk_buff *newsk,
1362 struct sk_buff *prev, struct sk_buff *next,
1363 struct sk_buff_head *list)
1364{
1365 newsk->next = next;
1366 newsk->prev = prev;
1367 next->prev = prev->next = newsk;
1368 list->qlen++;
1369}
1da177e4 1370
67fed459
DM
1371static inline void __skb_queue_splice(const struct sk_buff_head *list,
1372 struct sk_buff *prev,
1373 struct sk_buff *next)
1374{
1375 struct sk_buff *first = list->next;
1376 struct sk_buff *last = list->prev;
1377
1378 first->prev = prev;
1379 prev->next = first;
1380
1381 last->next = next;
1382 next->prev = last;
1383}
1384
1385/**
1386 * skb_queue_splice - join two skb lists, this is designed for stacks
1387 * @list: the new list to add
1388 * @head: the place to add it in the first list
1389 */
1390static inline void skb_queue_splice(const struct sk_buff_head *list,
1391 struct sk_buff_head *head)
1392{
1393 if (!skb_queue_empty(list)) {
1394 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1395 head->qlen += list->qlen;
67fed459
DM
1396 }
1397}
1398
1399/**
d9619496 1400 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1401 * @list: the new list to add
1402 * @head: the place to add it in the first list
1403 *
1404 * The list at @list is reinitialised
1405 */
1406static inline void skb_queue_splice_init(struct sk_buff_head *list,
1407 struct sk_buff_head *head)
1408{
1409 if (!skb_queue_empty(list)) {
1410 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1411 head->qlen += list->qlen;
67fed459
DM
1412 __skb_queue_head_init(list);
1413 }
1414}
1415
1416/**
1417 * skb_queue_splice_tail - join two skb lists, each list being a queue
1418 * @list: the new list to add
1419 * @head: the place to add it in the first list
1420 */
1421static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1422 struct sk_buff_head *head)
1423{
1424 if (!skb_queue_empty(list)) {
1425 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1426 head->qlen += list->qlen;
67fed459
DM
1427 }
1428}
1429
1430/**
d9619496 1431 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1432 * @list: the new list to add
1433 * @head: the place to add it in the first list
1434 *
1435 * Each of the lists is a queue.
1436 * The list at @list is reinitialised
1437 */
1438static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1439 struct sk_buff_head *head)
1440{
1441 if (!skb_queue_empty(list)) {
1442 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1443 head->qlen += list->qlen;
67fed459
DM
1444 __skb_queue_head_init(list);
1445 }
1446}
1447
1da177e4 1448/**
300ce174 1449 * __skb_queue_after - queue a buffer at the list head
1da177e4 1450 * @list: list to use
300ce174 1451 * @prev: place after this buffer
1da177e4
LT
1452 * @newsk: buffer to queue
1453 *
300ce174 1454 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1455 * and you must therefore hold required locks before calling it.
1456 *
1457 * A buffer cannot be placed on two lists at the same time.
1458 */
300ce174
SH
1459static inline void __skb_queue_after(struct sk_buff_head *list,
1460 struct sk_buff *prev,
1461 struct sk_buff *newsk)
1da177e4 1462{
bf299275 1463 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1464}
1465
7965bd4d
JP
1466void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1467 struct sk_buff_head *list);
7de6c033 1468
f5572855
GR
1469static inline void __skb_queue_before(struct sk_buff_head *list,
1470 struct sk_buff *next,
1471 struct sk_buff *newsk)
1472{
1473 __skb_insert(newsk, next->prev, next, list);
1474}
1475
300ce174
SH
1476/**
1477 * __skb_queue_head - queue a buffer at the list head
1478 * @list: list to use
1479 * @newsk: buffer to queue
1480 *
1481 * Queue a buffer at the start of a list. This function takes no locks
1482 * and you must therefore hold required locks before calling it.
1483 *
1484 * A buffer cannot be placed on two lists at the same time.
1485 */
7965bd4d 1486void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1487static inline void __skb_queue_head(struct sk_buff_head *list,
1488 struct sk_buff *newsk)
1489{
1490 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1491}
1492
1da177e4
LT
1493/**
1494 * __skb_queue_tail - queue a buffer at the list tail
1495 * @list: list to use
1496 * @newsk: buffer to queue
1497 *
1498 * Queue a buffer at the end of a list. This function takes no locks
1499 * and you must therefore hold required locks before calling it.
1500 *
1501 * A buffer cannot be placed on two lists at the same time.
1502 */
7965bd4d 1503void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1504static inline void __skb_queue_tail(struct sk_buff_head *list,
1505 struct sk_buff *newsk)
1506{
f5572855 1507 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1508}
1509
1da177e4
LT
1510/*
1511 * remove sk_buff from list. _Must_ be called atomically, and with
1512 * the list known..
1513 */
7965bd4d 1514void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1515static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1516{
1517 struct sk_buff *next, *prev;
1518
1519 list->qlen--;
1520 next = skb->next;
1521 prev = skb->prev;
1522 skb->next = skb->prev = NULL;
1da177e4
LT
1523 next->prev = prev;
1524 prev->next = next;
1525}
1526
f525c06d
GR
1527/**
1528 * __skb_dequeue - remove from the head of the queue
1529 * @list: list to dequeue from
1530 *
1531 * Remove the head of the list. This function does not take any locks
1532 * so must be used with appropriate locks held only. The head item is
1533 * returned or %NULL if the list is empty.
1534 */
7965bd4d 1535struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1536static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1537{
1538 struct sk_buff *skb = skb_peek(list);
1539 if (skb)
1540 __skb_unlink(skb, list);
1541 return skb;
1542}
1da177e4
LT
1543
1544/**
1545 * __skb_dequeue_tail - remove from the tail of the queue
1546 * @list: list to dequeue from
1547 *
1548 * Remove the tail of the list. This function does not take any locks
1549 * so must be used with appropriate locks held only. The tail item is
1550 * returned or %NULL if the list is empty.
1551 */
7965bd4d 1552struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1553static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1554{
1555 struct sk_buff *skb = skb_peek_tail(list);
1556 if (skb)
1557 __skb_unlink(skb, list);
1558 return skb;
1559}
1560
1561
bdcc0924 1562static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1563{
1564 return skb->data_len;
1565}
1566
1567static inline unsigned int skb_headlen(const struct sk_buff *skb)
1568{
1569 return skb->len - skb->data_len;
1570}
1571
1572static inline int skb_pagelen(const struct sk_buff *skb)
1573{
1574 int i, len = 0;
1575
1576 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1577 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1578 return len + skb_headlen(skb);
1579}
1580
131ea667
IC
1581/**
1582 * __skb_fill_page_desc - initialise a paged fragment in an skb
1583 * @skb: buffer containing fragment to be initialised
1584 * @i: paged fragment index to initialise
1585 * @page: the page to use for this fragment
1586 * @off: the offset to the data with @page
1587 * @size: the length of the data
1588 *
1589 * Initialises the @i'th fragment of @skb to point to &size bytes at
1590 * offset @off within @page.
1591 *
1592 * Does not take any additional reference on the fragment.
1593 */
1594static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1595 struct page *page, int off, int size)
1da177e4
LT
1596{
1597 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1598
c48a11c7
MG
1599 /*
1600 * Propagate page->pfmemalloc to the skb if we can. The problem is
1601 * that not all callers have unique ownership of the page. If
1602 * pfmemalloc is set, we check the mapping as a mapping implies
1603 * page->index is set (index and pfmemalloc share space).
1604 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1605 * do not lose pfmemalloc information as the pages would not be
1606 * allocated using __GFP_MEMALLOC.
1607 */
a8605c60 1608 frag->page.p = page;
1da177e4 1609 frag->page_offset = off;
9e903e08 1610 skb_frag_size_set(frag, size);
cca7af38
PE
1611
1612 page = compound_head(page);
1613 if (page->pfmemalloc && !page->mapping)
1614 skb->pfmemalloc = true;
131ea667
IC
1615}
1616
1617/**
1618 * skb_fill_page_desc - initialise a paged fragment in an skb
1619 * @skb: buffer containing fragment to be initialised
1620 * @i: paged fragment index to initialise
1621 * @page: the page to use for this fragment
1622 * @off: the offset to the data with @page
1623 * @size: the length of the data
1624 *
1625 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1626 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1627 * addition updates @skb such that @i is the last fragment.
1628 *
1629 * Does not take any additional reference on the fragment.
1630 */
1631static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1632 struct page *page, int off, int size)
1633{
1634 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1635 skb_shinfo(skb)->nr_frags = i + 1;
1636}
1637
7965bd4d
JP
1638void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1639 int size, unsigned int truesize);
654bed16 1640
f8e617e1
JW
1641void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1642 unsigned int truesize);
1643
1da177e4 1644#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1645#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1646#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1647
27a884dc
ACM
1648#ifdef NET_SKBUFF_DATA_USES_OFFSET
1649static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1650{
1651 return skb->head + skb->tail;
1652}
1653
1654static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1655{
1656 skb->tail = skb->data - skb->head;
1657}
1658
1659static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1660{
1661 skb_reset_tail_pointer(skb);
1662 skb->tail += offset;
1663}
7cc46190 1664
27a884dc
ACM
1665#else /* NET_SKBUFF_DATA_USES_OFFSET */
1666static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1667{
1668 return skb->tail;
1669}
1670
1671static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1672{
1673 skb->tail = skb->data;
1674}
1675
1676static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1677{
1678 skb->tail = skb->data + offset;
1679}
4305b541 1680
27a884dc
ACM
1681#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1682
1da177e4
LT
1683/*
1684 * Add data to an sk_buff
1685 */
0c7ddf36 1686unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1687unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1688static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1689{
27a884dc 1690 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1691 SKB_LINEAR_ASSERT(skb);
1692 skb->tail += len;
1693 skb->len += len;
1694 return tmp;
1695}
1696
7965bd4d 1697unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1698static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1699{
1700 skb->data -= len;
1701 skb->len += len;
1702 return skb->data;
1703}
1704
7965bd4d 1705unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1706static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1707{
1708 skb->len -= len;
1709 BUG_ON(skb->len < skb->data_len);
1710 return skb->data += len;
1711}
1712
47d29646
DM
1713static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1714{
1715 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1716}
1717
7965bd4d 1718unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1719
1720static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1721{
1722 if (len > skb_headlen(skb) &&
987c402a 1723 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1724 return NULL;
1725 skb->len -= len;
1726 return skb->data += len;
1727}
1728
1729static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1730{
1731 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1732}
1733
1734static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1735{
1736 if (likely(len <= skb_headlen(skb)))
1737 return 1;
1738 if (unlikely(len > skb->len))
1739 return 0;
987c402a 1740 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1741}
1742
1743/**
1744 * skb_headroom - bytes at buffer head
1745 * @skb: buffer to check
1746 *
1747 * Return the number of bytes of free space at the head of an &sk_buff.
1748 */
c2636b4d 1749static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1750{
1751 return skb->data - skb->head;
1752}
1753
1754/**
1755 * skb_tailroom - bytes at buffer end
1756 * @skb: buffer to check
1757 *
1758 * Return the number of bytes of free space at the tail of an sk_buff
1759 */
1760static inline int skb_tailroom(const struct sk_buff *skb)
1761{
4305b541 1762 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1763}
1764
a21d4572
ED
1765/**
1766 * skb_availroom - bytes at buffer end
1767 * @skb: buffer to check
1768 *
1769 * Return the number of bytes of free space at the tail of an sk_buff
1770 * allocated by sk_stream_alloc()
1771 */
1772static inline int skb_availroom(const struct sk_buff *skb)
1773{
16fad69c
ED
1774 if (skb_is_nonlinear(skb))
1775 return 0;
1776
1777 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1778}
1779
1da177e4
LT
1780/**
1781 * skb_reserve - adjust headroom
1782 * @skb: buffer to alter
1783 * @len: bytes to move
1784 *
1785 * Increase the headroom of an empty &sk_buff by reducing the tail
1786 * room. This is only allowed for an empty buffer.
1787 */
8243126c 1788static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1789{
1790 skb->data += len;
1791 skb->tail += len;
1792}
1793
8bce6d7d
TH
1794#define ENCAP_TYPE_ETHER 0
1795#define ENCAP_TYPE_IPPROTO 1
1796
1797static inline void skb_set_inner_protocol(struct sk_buff *skb,
1798 __be16 protocol)
1799{
1800 skb->inner_protocol = protocol;
1801 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1802}
1803
1804static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1805 __u8 ipproto)
1806{
1807 skb->inner_ipproto = ipproto;
1808 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1809}
1810
6a674e9c
JG
1811static inline void skb_reset_inner_headers(struct sk_buff *skb)
1812{
aefbd2b3 1813 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1814 skb->inner_network_header = skb->network_header;
1815 skb->inner_transport_header = skb->transport_header;
1816}
1817
0b5c9db1
JP
1818static inline void skb_reset_mac_len(struct sk_buff *skb)
1819{
1820 skb->mac_len = skb->network_header - skb->mac_header;
1821}
1822
6a674e9c
JG
1823static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1824 *skb)
1825{
1826 return skb->head + skb->inner_transport_header;
1827}
1828
1829static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1830{
1831 skb->inner_transport_header = skb->data - skb->head;
1832}
1833
1834static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1835 const int offset)
1836{
1837 skb_reset_inner_transport_header(skb);
1838 skb->inner_transport_header += offset;
1839}
1840
1841static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1842{
1843 return skb->head + skb->inner_network_header;
1844}
1845
1846static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1847{
1848 skb->inner_network_header = skb->data - skb->head;
1849}
1850
1851static inline void skb_set_inner_network_header(struct sk_buff *skb,
1852 const int offset)
1853{
1854 skb_reset_inner_network_header(skb);
1855 skb->inner_network_header += offset;
1856}
1857
aefbd2b3
PS
1858static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1859{
1860 return skb->head + skb->inner_mac_header;
1861}
1862
1863static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1864{
1865 skb->inner_mac_header = skb->data - skb->head;
1866}
1867
1868static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1869 const int offset)
1870{
1871 skb_reset_inner_mac_header(skb);
1872 skb->inner_mac_header += offset;
1873}
fda55eca
ED
1874static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1875{
35d04610 1876 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1877}
1878
9c70220b
ACM
1879static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1880{
2e07fa9c 1881 return skb->head + skb->transport_header;
9c70220b
ACM
1882}
1883
badff6d0
ACM
1884static inline void skb_reset_transport_header(struct sk_buff *skb)
1885{
2e07fa9c 1886 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1887}
1888
967b05f6
ACM
1889static inline void skb_set_transport_header(struct sk_buff *skb,
1890 const int offset)
1891{
2e07fa9c
ACM
1892 skb_reset_transport_header(skb);
1893 skb->transport_header += offset;
ea2ae17d
ACM
1894}
1895
d56f90a7
ACM
1896static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1897{
2e07fa9c 1898 return skb->head + skb->network_header;
d56f90a7
ACM
1899}
1900
c1d2bbe1
ACM
1901static inline void skb_reset_network_header(struct sk_buff *skb)
1902{
2e07fa9c 1903 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1904}
1905
c14d2450
ACM
1906static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1907{
2e07fa9c
ACM
1908 skb_reset_network_header(skb);
1909 skb->network_header += offset;
c14d2450
ACM
1910}
1911
2e07fa9c 1912static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1913{
2e07fa9c 1914 return skb->head + skb->mac_header;
bbe735e4
ACM
1915}
1916
2e07fa9c 1917static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1918{
35d04610 1919 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1920}
1921
1922static inline void skb_reset_mac_header(struct sk_buff *skb)
1923{
1924 skb->mac_header = skb->data - skb->head;
1925}
1926
1927static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1928{
1929 skb_reset_mac_header(skb);
1930 skb->mac_header += offset;
1931}
1932
0e3da5bb
TT
1933static inline void skb_pop_mac_header(struct sk_buff *skb)
1934{
1935 skb->mac_header = skb->network_header;
1936}
1937
fbbdb8f0
YX
1938static inline void skb_probe_transport_header(struct sk_buff *skb,
1939 const int offset_hint)
1940{
1941 struct flow_keys keys;
1942
1943 if (skb_transport_header_was_set(skb))
1944 return;
06635a35
JP
1945 else if (skb_flow_dissect_flow_keys(skb, &keys))
1946 skb_set_transport_header(skb, keys.basic.thoff);
fbbdb8f0
YX
1947 else
1948 skb_set_transport_header(skb, offset_hint);
1949}
1950
03606895
ED
1951static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1952{
1953 if (skb_mac_header_was_set(skb)) {
1954 const unsigned char *old_mac = skb_mac_header(skb);
1955
1956 skb_set_mac_header(skb, -skb->mac_len);
1957 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1958 }
1959}
1960
04fb451e
MM
1961static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1962{
1963 return skb->csum_start - skb_headroom(skb);
1964}
1965
2e07fa9c
ACM
1966static inline int skb_transport_offset(const struct sk_buff *skb)
1967{
1968 return skb_transport_header(skb) - skb->data;
1969}
1970
1971static inline u32 skb_network_header_len(const struct sk_buff *skb)
1972{
1973 return skb->transport_header - skb->network_header;
1974}
1975
6a674e9c
JG
1976static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1977{
1978 return skb->inner_transport_header - skb->inner_network_header;
1979}
1980
2e07fa9c
ACM
1981static inline int skb_network_offset(const struct sk_buff *skb)
1982{
1983 return skb_network_header(skb) - skb->data;
1984}
48d49d0c 1985
6a674e9c
JG
1986static inline int skb_inner_network_offset(const struct sk_buff *skb)
1987{
1988 return skb_inner_network_header(skb) - skb->data;
1989}
1990
f9599ce1
CG
1991static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1992{
1993 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1994}
1995
1da177e4
LT
1996/*
1997 * CPUs often take a performance hit when accessing unaligned memory
1998 * locations. The actual performance hit varies, it can be small if the
1999 * hardware handles it or large if we have to take an exception and fix it
2000 * in software.
2001 *
2002 * Since an ethernet header is 14 bytes network drivers often end up with
2003 * the IP header at an unaligned offset. The IP header can be aligned by
2004 * shifting the start of the packet by 2 bytes. Drivers should do this
2005 * with:
2006 *
8660c124 2007 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2008 *
2009 * The downside to this alignment of the IP header is that the DMA is now
2010 * unaligned. On some architectures the cost of an unaligned DMA is high
2011 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2012 *
1da177e4
LT
2013 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2014 * to be overridden.
2015 */
2016#ifndef NET_IP_ALIGN
2017#define NET_IP_ALIGN 2
2018#endif
2019
025be81e
AB
2020/*
2021 * The networking layer reserves some headroom in skb data (via
2022 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2023 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2024 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2025 *
2026 * Unfortunately this headroom changes the DMA alignment of the resulting
2027 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2028 * on some architectures. An architecture can override this value,
2029 * perhaps setting it to a cacheline in size (since that will maintain
2030 * cacheline alignment of the DMA). It must be a power of 2.
2031 *
d6301d3d 2032 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2033 * headroom, you should not reduce this.
5933dd2f
ED
2034 *
2035 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2036 * to reduce average number of cache lines per packet.
2037 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2038 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2039 */
2040#ifndef NET_SKB_PAD
5933dd2f 2041#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2042#endif
2043
7965bd4d 2044int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2045
2046static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2047{
c4264f27 2048 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2049 WARN_ON(1);
2050 return;
2051 }
27a884dc
ACM
2052 skb->len = len;
2053 skb_set_tail_pointer(skb, len);
1da177e4
LT
2054}
2055
7965bd4d 2056void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2057
2058static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2059{
3cc0e873
HX
2060 if (skb->data_len)
2061 return ___pskb_trim(skb, len);
2062 __skb_trim(skb, len);
2063 return 0;
1da177e4
LT
2064}
2065
2066static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2067{
2068 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2069}
2070
e9fa4f7b
HX
2071/**
2072 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2073 * @skb: buffer to alter
2074 * @len: new length
2075 *
2076 * This is identical to pskb_trim except that the caller knows that
2077 * the skb is not cloned so we should never get an error due to out-
2078 * of-memory.
2079 */
2080static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2081{
2082 int err = pskb_trim(skb, len);
2083 BUG_ON(err);
2084}
2085
1da177e4
LT
2086/**
2087 * skb_orphan - orphan a buffer
2088 * @skb: buffer to orphan
2089 *
2090 * If a buffer currently has an owner then we call the owner's
2091 * destructor function and make the @skb unowned. The buffer continues
2092 * to exist but is no longer charged to its former owner.
2093 */
2094static inline void skb_orphan(struct sk_buff *skb)
2095{
c34a7612 2096 if (skb->destructor) {
1da177e4 2097 skb->destructor(skb);
c34a7612
ED
2098 skb->destructor = NULL;
2099 skb->sk = NULL;
376c7311
ED
2100 } else {
2101 BUG_ON(skb->sk);
c34a7612 2102 }
1da177e4
LT
2103}
2104
a353e0ce
MT
2105/**
2106 * skb_orphan_frags - orphan the frags contained in a buffer
2107 * @skb: buffer to orphan frags from
2108 * @gfp_mask: allocation mask for replacement pages
2109 *
2110 * For each frag in the SKB which needs a destructor (i.e. has an
2111 * owner) create a copy of that frag and release the original
2112 * page by calling the destructor.
2113 */
2114static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2115{
2116 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2117 return 0;
2118 return skb_copy_ubufs(skb, gfp_mask);
2119}
2120
1da177e4
LT
2121/**
2122 * __skb_queue_purge - empty a list
2123 * @list: list to empty
2124 *
2125 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2126 * the list and one reference dropped. This function does not take the
2127 * list lock and the caller must hold the relevant locks to use it.
2128 */
7965bd4d 2129void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2130static inline void __skb_queue_purge(struct sk_buff_head *list)
2131{
2132 struct sk_buff *skb;
2133 while ((skb = __skb_dequeue(list)) != NULL)
2134 kfree_skb(skb);
2135}
2136
7965bd4d 2137void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2138
7965bd4d
JP
2139struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2140 gfp_t gfp_mask);
8af27456
CH
2141
2142/**
2143 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2144 * @dev: network device to receive on
2145 * @length: length to allocate
2146 *
2147 * Allocate a new &sk_buff and assign it a usage count of one. The
2148 * buffer has unspecified headroom built in. Users should allocate
2149 * the headroom they think they need without accounting for the
2150 * built in space. The built in space is used for optimisations.
2151 *
2152 * %NULL is returned if there is no free memory. Although this function
2153 * allocates memory it can be called from an interrupt.
2154 */
2155static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2156 unsigned int length)
8af27456
CH
2157{
2158 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2159}
2160
6f532612
ED
2161/* legacy helper around __netdev_alloc_skb() */
2162static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2163 gfp_t gfp_mask)
2164{
2165 return __netdev_alloc_skb(NULL, length, gfp_mask);
2166}
2167
2168/* legacy helper around netdev_alloc_skb() */
2169static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2170{
2171 return netdev_alloc_skb(NULL, length);
2172}
2173
2174
4915a0de
ED
2175static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2176 unsigned int length, gfp_t gfp)
61321bbd 2177{
4915a0de 2178 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2179
2180 if (NET_IP_ALIGN && skb)
2181 skb_reserve(skb, NET_IP_ALIGN);
2182 return skb;
2183}
2184
4915a0de
ED
2185static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2186 unsigned int length)
2187{
2188 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2189}
2190
181edb2b
AD
2191static inline void skb_free_frag(void *addr)
2192{
2193 __free_page_frag(addr);
2194}
2195
ffde7328 2196void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2197struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2198 unsigned int length, gfp_t gfp_mask);
2199static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2200 unsigned int length)
2201{
2202 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2203}
ffde7328 2204
71dfda58
AD
2205/**
2206 * __dev_alloc_pages - allocate page for network Rx
2207 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2208 * @order: size of the allocation
2209 *
2210 * Allocate a new page.
2211 *
2212 * %NULL is returned if there is no free memory.
2213*/
2214static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2215 unsigned int order)
2216{
2217 /* This piece of code contains several assumptions.
2218 * 1. This is for device Rx, therefor a cold page is preferred.
2219 * 2. The expectation is the user wants a compound page.
2220 * 3. If requesting a order 0 page it will not be compound
2221 * due to the check to see if order has a value in prep_new_page
2222 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2223 * code in gfp_to_alloc_flags that should be enforcing this.
2224 */
2225 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2226
2227 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2228}
2229
2230static inline struct page *dev_alloc_pages(unsigned int order)
2231{
2232 return __dev_alloc_pages(GFP_ATOMIC, order);
2233}
2234
2235/**
2236 * __dev_alloc_page - allocate a page for network Rx
2237 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2238 *
2239 * Allocate a new page.
2240 *
2241 * %NULL is returned if there is no free memory.
2242 */
2243static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2244{
2245 return __dev_alloc_pages(gfp_mask, 0);
2246}
2247
2248static inline struct page *dev_alloc_page(void)
2249{
2250 return __dev_alloc_page(GFP_ATOMIC);
2251}
2252
0614002b
MG
2253/**
2254 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2255 * @page: The page that was allocated from skb_alloc_page
2256 * @skb: The skb that may need pfmemalloc set
2257 */
2258static inline void skb_propagate_pfmemalloc(struct page *page,
2259 struct sk_buff *skb)
2260{
2261 if (page && page->pfmemalloc)
2262 skb->pfmemalloc = true;
2263}
2264
131ea667 2265/**
e227867f 2266 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2267 * @frag: the paged fragment
2268 *
2269 * Returns the &struct page associated with @frag.
2270 */
2271static inline struct page *skb_frag_page(const skb_frag_t *frag)
2272{
a8605c60 2273 return frag->page.p;
131ea667
IC
2274}
2275
2276/**
2277 * __skb_frag_ref - take an addition reference on a paged fragment.
2278 * @frag: the paged fragment
2279 *
2280 * Takes an additional reference on the paged fragment @frag.
2281 */
2282static inline void __skb_frag_ref(skb_frag_t *frag)
2283{
2284 get_page(skb_frag_page(frag));
2285}
2286
2287/**
2288 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2289 * @skb: the buffer
2290 * @f: the fragment offset.
2291 *
2292 * Takes an additional reference on the @f'th paged fragment of @skb.
2293 */
2294static inline void skb_frag_ref(struct sk_buff *skb, int f)
2295{
2296 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2297}
2298
2299/**
2300 * __skb_frag_unref - release a reference on a paged fragment.
2301 * @frag: the paged fragment
2302 *
2303 * Releases a reference on the paged fragment @frag.
2304 */
2305static inline void __skb_frag_unref(skb_frag_t *frag)
2306{
2307 put_page(skb_frag_page(frag));
2308}
2309
2310/**
2311 * skb_frag_unref - release a reference on a paged fragment of an skb.
2312 * @skb: the buffer
2313 * @f: the fragment offset
2314 *
2315 * Releases a reference on the @f'th paged fragment of @skb.
2316 */
2317static inline void skb_frag_unref(struct sk_buff *skb, int f)
2318{
2319 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2320}
2321
2322/**
2323 * skb_frag_address - gets the address of the data contained in a paged fragment
2324 * @frag: the paged fragment buffer
2325 *
2326 * Returns the address of the data within @frag. The page must already
2327 * be mapped.
2328 */
2329static inline void *skb_frag_address(const skb_frag_t *frag)
2330{
2331 return page_address(skb_frag_page(frag)) + frag->page_offset;
2332}
2333
2334/**
2335 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2336 * @frag: the paged fragment buffer
2337 *
2338 * Returns the address of the data within @frag. Checks that the page
2339 * is mapped and returns %NULL otherwise.
2340 */
2341static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2342{
2343 void *ptr = page_address(skb_frag_page(frag));
2344 if (unlikely(!ptr))
2345 return NULL;
2346
2347 return ptr + frag->page_offset;
2348}
2349
2350/**
2351 * __skb_frag_set_page - sets the page contained in a paged fragment
2352 * @frag: the paged fragment
2353 * @page: the page to set
2354 *
2355 * Sets the fragment @frag to contain @page.
2356 */
2357static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2358{
a8605c60 2359 frag->page.p = page;
131ea667
IC
2360}
2361
2362/**
2363 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2364 * @skb: the buffer
2365 * @f: the fragment offset
2366 * @page: the page to set
2367 *
2368 * Sets the @f'th fragment of @skb to contain @page.
2369 */
2370static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2371 struct page *page)
2372{
2373 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2374}
2375
400dfd3a
ED
2376bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2377
131ea667
IC
2378/**
2379 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2380 * @dev: the device to map the fragment to
131ea667
IC
2381 * @frag: the paged fragment to map
2382 * @offset: the offset within the fragment (starting at the
2383 * fragment's own offset)
2384 * @size: the number of bytes to map
f83347df 2385 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2386 *
2387 * Maps the page associated with @frag to @device.
2388 */
2389static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2390 const skb_frag_t *frag,
2391 size_t offset, size_t size,
2392 enum dma_data_direction dir)
2393{
2394 return dma_map_page(dev, skb_frag_page(frag),
2395 frag->page_offset + offset, size, dir);
2396}
2397
117632e6
ED
2398static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2399 gfp_t gfp_mask)
2400{
2401 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2402}
2403
bad93e9d
OP
2404
2405static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2406 gfp_t gfp_mask)
2407{
2408 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2409}
2410
2411
334a8132
PM
2412/**
2413 * skb_clone_writable - is the header of a clone writable
2414 * @skb: buffer to check
2415 * @len: length up to which to write
2416 *
2417 * Returns true if modifying the header part of the cloned buffer
2418 * does not requires the data to be copied.
2419 */
05bdd2f1 2420static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2421{
2422 return !skb_header_cloned(skb) &&
2423 skb_headroom(skb) + len <= skb->hdr_len;
2424}
2425
d9cc2048
HX
2426static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2427 int cloned)
2428{
2429 int delta = 0;
2430
d9cc2048
HX
2431 if (headroom > skb_headroom(skb))
2432 delta = headroom - skb_headroom(skb);
2433
2434 if (delta || cloned)
2435 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2436 GFP_ATOMIC);
2437 return 0;
2438}
2439
1da177e4
LT
2440/**
2441 * skb_cow - copy header of skb when it is required
2442 * @skb: buffer to cow
2443 * @headroom: needed headroom
2444 *
2445 * If the skb passed lacks sufficient headroom or its data part
2446 * is shared, data is reallocated. If reallocation fails, an error
2447 * is returned and original skb is not changed.
2448 *
2449 * The result is skb with writable area skb->head...skb->tail
2450 * and at least @headroom of space at head.
2451 */
2452static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2453{
d9cc2048
HX
2454 return __skb_cow(skb, headroom, skb_cloned(skb));
2455}
1da177e4 2456
d9cc2048
HX
2457/**
2458 * skb_cow_head - skb_cow but only making the head writable
2459 * @skb: buffer to cow
2460 * @headroom: needed headroom
2461 *
2462 * This function is identical to skb_cow except that we replace the
2463 * skb_cloned check by skb_header_cloned. It should be used when
2464 * you only need to push on some header and do not need to modify
2465 * the data.
2466 */
2467static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2468{
2469 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2470}
2471
2472/**
2473 * skb_padto - pad an skbuff up to a minimal size
2474 * @skb: buffer to pad
2475 * @len: minimal length
2476 *
2477 * Pads up a buffer to ensure the trailing bytes exist and are
2478 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2479 * is untouched. Otherwise it is extended. Returns zero on
2480 * success. The skb is freed on error.
1da177e4 2481 */
5b057c6b 2482static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2483{
2484 unsigned int size = skb->len;
2485 if (likely(size >= len))
5b057c6b 2486 return 0;
987c402a 2487 return skb_pad(skb, len - size);
1da177e4
LT
2488}
2489
9c0c1124
AD
2490/**
2491 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2492 * @skb: buffer to pad
2493 * @len: minimal length
2494 *
2495 * Pads up a buffer to ensure the trailing bytes exist and are
2496 * blanked. If the buffer already contains sufficient data it
2497 * is untouched. Otherwise it is extended. Returns zero on
2498 * success. The skb is freed on error.
2499 */
2500static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2501{
2502 unsigned int size = skb->len;
2503
2504 if (unlikely(size < len)) {
2505 len -= size;
2506 if (skb_pad(skb, len))
2507 return -ENOMEM;
2508 __skb_put(skb, len);
2509 }
2510 return 0;
2511}
2512
1da177e4 2513static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2514 struct iov_iter *from, int copy)
1da177e4
LT
2515{
2516 const int off = skb->len;
2517
2518 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2519 __wsum csum = 0;
2520 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2521 &csum, from) == copy) {
1da177e4
LT
2522 skb->csum = csum_block_add(skb->csum, csum, off);
2523 return 0;
2524 }
af2b040e 2525 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2526 return 0;
2527
2528 __skb_trim(skb, off);
2529 return -EFAULT;
2530}
2531
38ba0a65
ED
2532static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2533 const struct page *page, int off)
1da177e4
LT
2534{
2535 if (i) {
9e903e08 2536 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2537
ea2ab693 2538 return page == skb_frag_page(frag) &&
9e903e08 2539 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2540 }
38ba0a65 2541 return false;
1da177e4
LT
2542}
2543
364c6bad
HX
2544static inline int __skb_linearize(struct sk_buff *skb)
2545{
2546 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2547}
2548
1da177e4
LT
2549/**
2550 * skb_linearize - convert paged skb to linear one
2551 * @skb: buffer to linarize
1da177e4
LT
2552 *
2553 * If there is no free memory -ENOMEM is returned, otherwise zero
2554 * is returned and the old skb data released.
2555 */
364c6bad
HX
2556static inline int skb_linearize(struct sk_buff *skb)
2557{
2558 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2559}
2560
cef401de
ED
2561/**
2562 * skb_has_shared_frag - can any frag be overwritten
2563 * @skb: buffer to test
2564 *
2565 * Return true if the skb has at least one frag that might be modified
2566 * by an external entity (as in vmsplice()/sendfile())
2567 */
2568static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2569{
c9af6db4
PS
2570 return skb_is_nonlinear(skb) &&
2571 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2572}
2573
364c6bad
HX
2574/**
2575 * skb_linearize_cow - make sure skb is linear and writable
2576 * @skb: buffer to process
2577 *
2578 * If there is no free memory -ENOMEM is returned, otherwise zero
2579 * is returned and the old skb data released.
2580 */
2581static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2582{
364c6bad
HX
2583 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2584 __skb_linearize(skb) : 0;
1da177e4
LT
2585}
2586
2587/**
2588 * skb_postpull_rcsum - update checksum for received skb after pull
2589 * @skb: buffer to update
2590 * @start: start of data before pull
2591 * @len: length of data pulled
2592 *
2593 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2594 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2595 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2596 */
2597
2598static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2599 const void *start, unsigned int len)
1da177e4 2600{
84fa7933 2601 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2602 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2603}
2604
cbb042f9
HX
2605unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2606
7ce5a27f
DM
2607/**
2608 * pskb_trim_rcsum - trim received skb and update checksum
2609 * @skb: buffer to trim
2610 * @len: new length
2611 *
2612 * This is exactly the same as pskb_trim except that it ensures the
2613 * checksum of received packets are still valid after the operation.
2614 */
2615
2616static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2617{
2618 if (likely(len >= skb->len))
2619 return 0;
2620 if (skb->ip_summed == CHECKSUM_COMPLETE)
2621 skb->ip_summed = CHECKSUM_NONE;
2622 return __pskb_trim(skb, len);
2623}
2624
1da177e4
LT
2625#define skb_queue_walk(queue, skb) \
2626 for (skb = (queue)->next; \
a1e4891f 2627 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2628 skb = skb->next)
2629
46f8914e
JC
2630#define skb_queue_walk_safe(queue, skb, tmp) \
2631 for (skb = (queue)->next, tmp = skb->next; \
2632 skb != (struct sk_buff *)(queue); \
2633 skb = tmp, tmp = skb->next)
2634
1164f52a 2635#define skb_queue_walk_from(queue, skb) \
a1e4891f 2636 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2637 skb = skb->next)
2638
2639#define skb_queue_walk_from_safe(queue, skb, tmp) \
2640 for (tmp = skb->next; \
2641 skb != (struct sk_buff *)(queue); \
2642 skb = tmp, tmp = skb->next)
2643
300ce174
SH
2644#define skb_queue_reverse_walk(queue, skb) \
2645 for (skb = (queue)->prev; \
a1e4891f 2646 skb != (struct sk_buff *)(queue); \
300ce174
SH
2647 skb = skb->prev)
2648
686a2955
DM
2649#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2650 for (skb = (queue)->prev, tmp = skb->prev; \
2651 skb != (struct sk_buff *)(queue); \
2652 skb = tmp, tmp = skb->prev)
2653
2654#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2655 for (tmp = skb->prev; \
2656 skb != (struct sk_buff *)(queue); \
2657 skb = tmp, tmp = skb->prev)
1da177e4 2658
21dc3301 2659static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2660{
2661 return skb_shinfo(skb)->frag_list != NULL;
2662}
2663
2664static inline void skb_frag_list_init(struct sk_buff *skb)
2665{
2666 skb_shinfo(skb)->frag_list = NULL;
2667}
2668
2669static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2670{
2671 frag->next = skb_shinfo(skb)->frag_list;
2672 skb_shinfo(skb)->frag_list = frag;
2673}
2674
2675#define skb_walk_frags(skb, iter) \
2676 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2677
7965bd4d
JP
2678struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2679 int *peeked, int *off, int *err);
2680struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2681 int *err);
2682unsigned int datagram_poll(struct file *file, struct socket *sock,
2683 struct poll_table_struct *wait);
c0371da6
AV
2684int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2685 struct iov_iter *to, int size);
51f3d02b
DM
2686static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2687 struct msghdr *msg, int size)
2688{
e5a4b0bb 2689 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2690}
e5a4b0bb
AV
2691int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2692 struct msghdr *msg);
3a654f97
AV
2693int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2694 struct iov_iter *from, int len);
3a654f97 2695int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2696void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2697void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2698int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2699int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2700int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2701__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2702 int len, __wsum csum);
a60e3cc7
HFS
2703ssize_t skb_socket_splice(struct sock *sk,
2704 struct pipe_inode_info *pipe,
2705 struct splice_pipe_desc *spd);
2706int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 2707 struct pipe_inode_info *pipe, unsigned int len,
a60e3cc7
HFS
2708 unsigned int flags,
2709 ssize_t (*splice_cb)(struct sock *,
2710 struct pipe_inode_info *,
2711 struct splice_pipe_desc *));
7965bd4d 2712void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2713unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2714int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2715 int len, int hlen);
7965bd4d
JP
2716void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2717int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2718void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2719unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2720struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2721struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2722int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2723int skb_vlan_pop(struct sk_buff *skb);
2724int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2725
6ce8e9ce
AV
2726static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2727{
21226abb 2728 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
2729}
2730
7eab8d9e
AV
2731static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2732{
e5a4b0bb 2733 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2734}
2735
2817a336
DB
2736struct skb_checksum_ops {
2737 __wsum (*update)(const void *mem, int len, __wsum wsum);
2738 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2739};
2740
2741__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2742 __wsum csum, const struct skb_checksum_ops *ops);
2743__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2744 __wsum csum);
2745
690e36e7
DM
2746static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2747 int len, void *data, int hlen, void *buffer)
1da177e4 2748{
55820ee2 2749 if (hlen - offset >= len)
690e36e7 2750 return data + offset;
1da177e4 2751
690e36e7
DM
2752 if (!skb ||
2753 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2754 return NULL;
2755
2756 return buffer;
2757}
2758
690e36e7
DM
2759static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2760 int len, void *buffer)
2761{
2762 return __skb_header_pointer(skb, offset, len, skb->data,
2763 skb_headlen(skb), buffer);
2764}
2765
4262e5cc
DB
2766/**
2767 * skb_needs_linearize - check if we need to linearize a given skb
2768 * depending on the given device features.
2769 * @skb: socket buffer to check
2770 * @features: net device features
2771 *
2772 * Returns true if either:
2773 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2774 * 2. skb is fragmented and the device does not support SG.
2775 */
2776static inline bool skb_needs_linearize(struct sk_buff *skb,
2777 netdev_features_t features)
2778{
2779 return skb_is_nonlinear(skb) &&
2780 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2781 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2782}
2783
d626f62b
ACM
2784static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2785 void *to,
2786 const unsigned int len)
2787{
2788 memcpy(to, skb->data, len);
2789}
2790
2791static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2792 const int offset, void *to,
2793 const unsigned int len)
2794{
2795 memcpy(to, skb->data + offset, len);
2796}
2797
27d7ff46
ACM
2798static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2799 const void *from,
2800 const unsigned int len)
2801{
2802 memcpy(skb->data, from, len);
2803}
2804
2805static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2806 const int offset,
2807 const void *from,
2808 const unsigned int len)
2809{
2810 memcpy(skb->data + offset, from, len);
2811}
2812
7965bd4d 2813void skb_init(void);
1da177e4 2814
ac45f602
PO
2815static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2816{
2817 return skb->tstamp;
2818}
2819
a61bbcf2
PM
2820/**
2821 * skb_get_timestamp - get timestamp from a skb
2822 * @skb: skb to get stamp from
2823 * @stamp: pointer to struct timeval to store stamp in
2824 *
2825 * Timestamps are stored in the skb as offsets to a base timestamp.
2826 * This function converts the offset back to a struct timeval and stores
2827 * it in stamp.
2828 */
ac45f602
PO
2829static inline void skb_get_timestamp(const struct sk_buff *skb,
2830 struct timeval *stamp)
a61bbcf2 2831{
b7aa0bf7 2832 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2833}
2834
ac45f602
PO
2835static inline void skb_get_timestampns(const struct sk_buff *skb,
2836 struct timespec *stamp)
2837{
2838 *stamp = ktime_to_timespec(skb->tstamp);
2839}
2840
b7aa0bf7 2841static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2842{
b7aa0bf7 2843 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2844}
2845
164891aa
SH
2846static inline ktime_t net_timedelta(ktime_t t)
2847{
2848 return ktime_sub(ktime_get_real(), t);
2849}
2850
b9ce204f
IJ
2851static inline ktime_t net_invalid_timestamp(void)
2852{
2853 return ktime_set(0, 0);
2854}
a61bbcf2 2855
62bccb8c
AD
2856struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2857
c1f19b51
RC
2858#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2859
7965bd4d
JP
2860void skb_clone_tx_timestamp(struct sk_buff *skb);
2861bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2862
2863#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2864
2865static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2866{
2867}
2868
2869static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2870{
2871 return false;
2872}
2873
2874#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2875
2876/**
2877 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2878 *
da92b194
RC
2879 * PHY drivers may accept clones of transmitted packets for
2880 * timestamping via their phy_driver.txtstamp method. These drivers
2881 * must call this function to return the skb back to the stack, with
2882 * or without a timestamp.
2883 *
c1f19b51 2884 * @skb: clone of the the original outgoing packet
da92b194 2885 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2886 *
2887 */
2888void skb_complete_tx_timestamp(struct sk_buff *skb,
2889 struct skb_shared_hwtstamps *hwtstamps);
2890
e7fd2885
WB
2891void __skb_tstamp_tx(struct sk_buff *orig_skb,
2892 struct skb_shared_hwtstamps *hwtstamps,
2893 struct sock *sk, int tstype);
2894
ac45f602
PO
2895/**
2896 * skb_tstamp_tx - queue clone of skb with send time stamps
2897 * @orig_skb: the original outgoing packet
2898 * @hwtstamps: hardware time stamps, may be NULL if not available
2899 *
2900 * If the skb has a socket associated, then this function clones the
2901 * skb (thus sharing the actual data and optional structures), stores
2902 * the optional hardware time stamping information (if non NULL) or
2903 * generates a software time stamp (otherwise), then queues the clone
2904 * to the error queue of the socket. Errors are silently ignored.
2905 */
7965bd4d
JP
2906void skb_tstamp_tx(struct sk_buff *orig_skb,
2907 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2908
4507a715
RC
2909static inline void sw_tx_timestamp(struct sk_buff *skb)
2910{
2244d07b
OH
2911 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2912 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2913 skb_tstamp_tx(skb, NULL);
2914}
2915
2916/**
2917 * skb_tx_timestamp() - Driver hook for transmit timestamping
2918 *
2919 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2920 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2921 *
73409f3b
DM
2922 * Specifically, one should make absolutely sure that this function is
2923 * called before TX completion of this packet can trigger. Otherwise
2924 * the packet could potentially already be freed.
2925 *
4507a715
RC
2926 * @skb: A socket buffer.
2927 */
2928static inline void skb_tx_timestamp(struct sk_buff *skb)
2929{
c1f19b51 2930 skb_clone_tx_timestamp(skb);
4507a715
RC
2931 sw_tx_timestamp(skb);
2932}
2933
6e3e939f
JB
2934/**
2935 * skb_complete_wifi_ack - deliver skb with wifi status
2936 *
2937 * @skb: the original outgoing packet
2938 * @acked: ack status
2939 *
2940 */
2941void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2942
7965bd4d
JP
2943__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2944__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2945
60476372
HX
2946static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2947{
6edec0e6
TH
2948 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2949 skb->csum_valid ||
2950 (skb->ip_summed == CHECKSUM_PARTIAL &&
2951 skb_checksum_start_offset(skb) >= 0));
60476372
HX
2952}
2953
fb286bb2
HX
2954/**
2955 * skb_checksum_complete - Calculate checksum of an entire packet
2956 * @skb: packet to process
2957 *
2958 * This function calculates the checksum over the entire packet plus
2959 * the value of skb->csum. The latter can be used to supply the
2960 * checksum of a pseudo header as used by TCP/UDP. It returns the
2961 * checksum.
2962 *
2963 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2964 * this function can be used to verify that checksum on received
2965 * packets. In that case the function should return zero if the
2966 * checksum is correct. In particular, this function will return zero
2967 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2968 * hardware has already verified the correctness of the checksum.
2969 */
4381ca3c 2970static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2971{
60476372
HX
2972 return skb_csum_unnecessary(skb) ?
2973 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2974}
2975
77cffe23
TH
2976static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2977{
2978 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2979 if (skb->csum_level == 0)
2980 skb->ip_summed = CHECKSUM_NONE;
2981 else
2982 skb->csum_level--;
2983 }
2984}
2985
2986static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2987{
2988 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2989 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2990 skb->csum_level++;
2991 } else if (skb->ip_summed == CHECKSUM_NONE) {
2992 skb->ip_summed = CHECKSUM_UNNECESSARY;
2993 skb->csum_level = 0;
2994 }
2995}
2996
5a212329
TH
2997static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2998{
2999 /* Mark current checksum as bad (typically called from GRO
3000 * path). In the case that ip_summed is CHECKSUM_NONE
3001 * this must be the first checksum encountered in the packet.
3002 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3003 * checksum after the last one validated. For UDP, a zero
3004 * checksum can not be marked as bad.
3005 */
3006
3007 if (skb->ip_summed == CHECKSUM_NONE ||
3008 skb->ip_summed == CHECKSUM_UNNECESSARY)
3009 skb->csum_bad = 1;
3010}
3011
76ba0aae
TH
3012/* Check if we need to perform checksum complete validation.
3013 *
3014 * Returns true if checksum complete is needed, false otherwise
3015 * (either checksum is unnecessary or zero checksum is allowed).
3016 */
3017static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3018 bool zero_okay,
3019 __sum16 check)
3020{
5d0c2b95
TH
3021 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3022 skb->csum_valid = 1;
77cffe23 3023 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3024 return false;
3025 }
3026
3027 return true;
3028}
3029
3030/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3031 * in checksum_init.
3032 */
3033#define CHECKSUM_BREAK 76
3034
4e18b9ad
TH
3035/* Unset checksum-complete
3036 *
3037 * Unset checksum complete can be done when packet is being modified
3038 * (uncompressed for instance) and checksum-complete value is
3039 * invalidated.
3040 */
3041static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3042{
3043 if (skb->ip_summed == CHECKSUM_COMPLETE)
3044 skb->ip_summed = CHECKSUM_NONE;
3045}
3046
76ba0aae
TH
3047/* Validate (init) checksum based on checksum complete.
3048 *
3049 * Return values:
3050 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3051 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3052 * checksum is stored in skb->csum for use in __skb_checksum_complete
3053 * non-zero: value of invalid checksum
3054 *
3055 */
3056static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3057 bool complete,
3058 __wsum psum)
3059{
3060 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3061 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3062 skb->csum_valid = 1;
76ba0aae
TH
3063 return 0;
3064 }
5a212329
TH
3065 } else if (skb->csum_bad) {
3066 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3067 return (__force __sum16)1;
76ba0aae
TH
3068 }
3069
3070 skb->csum = psum;
3071
5d0c2b95
TH
3072 if (complete || skb->len <= CHECKSUM_BREAK) {
3073 __sum16 csum;
3074
3075 csum = __skb_checksum_complete(skb);
3076 skb->csum_valid = !csum;
3077 return csum;
3078 }
76ba0aae
TH
3079
3080 return 0;
3081}
3082
3083static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3084{
3085 return 0;
3086}
3087
3088/* Perform checksum validate (init). Note that this is a macro since we only
3089 * want to calculate the pseudo header which is an input function if necessary.
3090 * First we try to validate without any computation (checksum unnecessary) and
3091 * then calculate based on checksum complete calling the function to compute
3092 * pseudo header.
3093 *
3094 * Return values:
3095 * 0: checksum is validated or try to in skb_checksum_complete
3096 * non-zero: value of invalid checksum
3097 */
3098#define __skb_checksum_validate(skb, proto, complete, \
3099 zero_okay, check, compute_pseudo) \
3100({ \
3101 __sum16 __ret = 0; \
5d0c2b95 3102 skb->csum_valid = 0; \
76ba0aae
TH
3103 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3104 __ret = __skb_checksum_validate_complete(skb, \
3105 complete, compute_pseudo(skb, proto)); \
3106 __ret; \
3107})
3108
3109#define skb_checksum_init(skb, proto, compute_pseudo) \
3110 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3111
3112#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3113 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3114
3115#define skb_checksum_validate(skb, proto, compute_pseudo) \
3116 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3117
3118#define skb_checksum_validate_zero_check(skb, proto, check, \
3119 compute_pseudo) \
096a4cfa 3120 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3121
3122#define skb_checksum_simple_validate(skb) \
3123 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3124
d96535a1
TH
3125static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3126{
3127 return (skb->ip_summed == CHECKSUM_NONE &&
3128 skb->csum_valid && !skb->csum_bad);
3129}
3130
3131static inline void __skb_checksum_convert(struct sk_buff *skb,
3132 __sum16 check, __wsum pseudo)
3133{
3134 skb->csum = ~pseudo;
3135 skb->ip_summed = CHECKSUM_COMPLETE;
3136}
3137
3138#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3139do { \
3140 if (__skb_checksum_convert_check(skb)) \
3141 __skb_checksum_convert(skb, check, \
3142 compute_pseudo(skb, proto)); \
3143} while (0)
3144
15e2396d
TH
3145static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3146 u16 start, u16 offset)
3147{
3148 skb->ip_summed = CHECKSUM_PARTIAL;
3149 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3150 skb->csum_offset = offset - start;
3151}
3152
dcdc8994
TH
3153/* Update skbuf and packet to reflect the remote checksum offload operation.
3154 * When called, ptr indicates the starting point for skb->csum when
3155 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3156 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3157 */
3158static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3159 int start, int offset, bool nopartial)
dcdc8994
TH
3160{
3161 __wsum delta;
3162
15e2396d
TH
3163 if (!nopartial) {
3164 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3165 return;
3166 }
3167
dcdc8994
TH
3168 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3169 __skb_checksum_complete(skb);
3170 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3171 }
3172
3173 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3174
3175 /* Adjust skb->csum since we changed the packet */
3176 skb->csum = csum_add(skb->csum, delta);
3177}
3178
5f79e0f9 3179#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3180void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3181static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3182{
3183 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3184 nf_conntrack_destroy(nfct);
1da177e4
LT
3185}
3186static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3187{
3188 if (nfct)
3189 atomic_inc(&nfct->use);
3190}
2fc72c7b 3191#endif
34666d46 3192#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3193static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3194{
3195 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3196 kfree(nf_bridge);
3197}
3198static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3199{
3200 if (nf_bridge)
3201 atomic_inc(&nf_bridge->use);
3202}
3203#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3204static inline void nf_reset(struct sk_buff *skb)
3205{
5f79e0f9 3206#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3207 nf_conntrack_put(skb->nfct);
3208 skb->nfct = NULL;
2fc72c7b 3209#endif
34666d46 3210#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3211 nf_bridge_put(skb->nf_bridge);
3212 skb->nf_bridge = NULL;
3213#endif
3214}
3215
124dff01
PM
3216static inline void nf_reset_trace(struct sk_buff *skb)
3217{
478b360a 3218#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3219 skb->nf_trace = 0;
3220#endif
a193a4ab
PM
3221}
3222
edda553c 3223/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3224static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3225 bool copy)
edda553c 3226{
5f79e0f9 3227#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3228 dst->nfct = src->nfct;
3229 nf_conntrack_get(src->nfct);
b1937227
ED
3230 if (copy)
3231 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3232#endif
34666d46 3233#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3234 dst->nf_bridge = src->nf_bridge;
3235 nf_bridge_get(src->nf_bridge);
3236#endif
478b360a 3237#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3238 if (copy)
3239 dst->nf_trace = src->nf_trace;
478b360a 3240#endif
edda553c
YK
3241}
3242
e7ac05f3
YK
3243static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3244{
e7ac05f3 3245#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3246 nf_conntrack_put(dst->nfct);
2fc72c7b 3247#endif
34666d46 3248#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3249 nf_bridge_put(dst->nf_bridge);
3250#endif
b1937227 3251 __nf_copy(dst, src, true);
e7ac05f3
YK
3252}
3253
984bc16c
JM
3254#ifdef CONFIG_NETWORK_SECMARK
3255static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3256{
3257 to->secmark = from->secmark;
3258}
3259
3260static inline void skb_init_secmark(struct sk_buff *skb)
3261{
3262 skb->secmark = 0;
3263}
3264#else
3265static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3266{ }
3267
3268static inline void skb_init_secmark(struct sk_buff *skb)
3269{ }
3270#endif
3271
574f7194
EB
3272static inline bool skb_irq_freeable(const struct sk_buff *skb)
3273{
3274 return !skb->destructor &&
3275#if IS_ENABLED(CONFIG_XFRM)
3276 !skb->sp &&
3277#endif
3278#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3279 !skb->nfct &&
3280#endif
3281 !skb->_skb_refdst &&
3282 !skb_has_frag_list(skb);
3283}
3284
f25f4e44
PWJ
3285static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3286{
f25f4e44 3287 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3288}
3289
9247744e 3290static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3291{
4e3ab47a 3292 return skb->queue_mapping;
4e3ab47a
PE
3293}
3294
f25f4e44
PWJ
3295static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3296{
f25f4e44 3297 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3298}
3299
d5a9e24a
DM
3300static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3301{
3302 skb->queue_mapping = rx_queue + 1;
3303}
3304
9247744e 3305static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3306{
3307 return skb->queue_mapping - 1;
3308}
3309
9247744e 3310static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3311{
a02cec21 3312 return skb->queue_mapping != 0;
d5a9e24a
DM
3313}
3314
def8b4fa
AD
3315static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3316{
0b3d8e08 3317#ifdef CONFIG_XFRM
def8b4fa 3318 return skb->sp;
def8b4fa 3319#else
def8b4fa 3320 return NULL;
def8b4fa 3321#endif
0b3d8e08 3322}
def8b4fa 3323
68c33163
PS
3324/* Keeps track of mac header offset relative to skb->head.
3325 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3326 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3327 * tunnel skb it points to outer mac header.
3328 * Keeps track of level of encapsulation of network headers.
3329 */
68c33163 3330struct skb_gso_cb {
3347c960
ED
3331 int mac_offset;
3332 int encap_level;
7e2b10c1 3333 __u16 csum_start;
68c33163
PS
3334};
3335#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3336
3337static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3338{
3339 return (skb_mac_header(inner_skb) - inner_skb->head) -
3340 SKB_GSO_CB(inner_skb)->mac_offset;
3341}
3342
1e2bd517
PS
3343static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3344{
3345 int new_headroom, headroom;
3346 int ret;
3347
3348 headroom = skb_headroom(skb);
3349 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3350 if (ret)
3351 return ret;
3352
3353 new_headroom = skb_headroom(skb);
3354 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3355 return 0;
3356}
3357
7e2b10c1
TH
3358/* Compute the checksum for a gso segment. First compute the checksum value
3359 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3360 * then add in skb->csum (checksum from csum_start to end of packet).
3361 * skb->csum and csum_start are then updated to reflect the checksum of the
3362 * resultant packet starting from the transport header-- the resultant checksum
3363 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3364 * header.
3365 */
3366static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3367{
3368 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
c91d4606
ED
3369 skb_transport_offset(skb);
3370 __wsum partial;
7e2b10c1 3371
c91d4606 3372 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
7e2b10c1
TH
3373 skb->csum = res;
3374 SKB_GSO_CB(skb)->csum_start -= plen;
3375
c91d4606 3376 return csum_fold(partial);
7e2b10c1
TH
3377}
3378
bdcc0924 3379static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3380{
3381 return skb_shinfo(skb)->gso_size;
3382}
3383
36a8f39e 3384/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3385static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3386{
3387 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3388}
3389
7965bd4d 3390void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3391
3392static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3393{
3394 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3395 * wanted then gso_type will be set. */
05bdd2f1
ED
3396 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3397
b78462eb
AD
3398 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3399 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3400 __skb_warn_lro_forwarding(skb);
3401 return true;
3402 }
3403 return false;
3404}
3405
35fc92a9
HX
3406static inline void skb_forward_csum(struct sk_buff *skb)
3407{
3408 /* Unfortunately we don't support this one. Any brave souls? */
3409 if (skb->ip_summed == CHECKSUM_COMPLETE)
3410 skb->ip_summed = CHECKSUM_NONE;
3411}
3412
bc8acf2c
ED
3413/**
3414 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3415 * @skb: skb to check
3416 *
3417 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3418 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3419 * use this helper, to document places where we make this assertion.
3420 */
05bdd2f1 3421static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3422{
3423#ifdef DEBUG
3424 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3425#endif
3426}
3427
f35d9d8a 3428bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3429
ed1f50c3 3430int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3431struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3432 unsigned int transport_len,
3433 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3434
3a7c1ee4
AD
3435/**
3436 * skb_head_is_locked - Determine if the skb->head is locked down
3437 * @skb: skb to check
3438 *
3439 * The head on skbs build around a head frag can be removed if they are
3440 * not cloned. This function returns true if the skb head is locked down
3441 * due to either being allocated via kmalloc, or by being a clone with
3442 * multiple references to the head.
3443 */
3444static inline bool skb_head_is_locked(const struct sk_buff *skb)
3445{
3446 return !skb->head_frag || skb_cloned(skb);
3447}
fe6cc55f
FW
3448
3449/**
3450 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3451 *
3452 * @skb: GSO skb
3453 *
3454 * skb_gso_network_seglen is used to determine the real size of the
3455 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3456 *
3457 * The MAC/L2 header is not accounted for.
3458 */
3459static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3460{
3461 unsigned int hdr_len = skb_transport_header(skb) -
3462 skb_network_header(skb);
3463 return hdr_len + skb_gso_transport_seglen(skb);
3464}
1da177e4
LT
3465#endif /* __KERNEL__ */
3466#endif /* _LINUX_SKBUFF_H */