net: move make_writable helper into common code
[linux-2.6-block.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>
5203cd28 37#include <net/flow_keys.h>
1da177e4 38
78ea85f1
DB
39/* A. Checksumming of received packets by device.
40 *
41 * CHECKSUM_NONE:
42 *
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 *
47 * CHECKSUM_UNNECESSARY:
48 *
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
56 *
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * TCP: IPv6 and IPv4.
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
64 *
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
78ea85f1
DB
74 *
75 * CHECKSUM_COMPLETE:
76 *
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
80 *
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
83 *
84 * CHECKSUM_PARTIAL:
85 *
86 * This is identical to the case for output below. This may occur on a packet
87 * received directly from another Linux OS, e.g., a virtualized Linux kernel
88 * on the same host. The packet can be treated in the same way as
89 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
90 * checksum must be filled in by the OS or the hardware.
91 *
92 * B. Checksumming on output.
93 *
94 * CHECKSUM_NONE:
95 *
96 * The skb was already checksummed by the protocol, or a checksum is not
97 * required.
98 *
99 * CHECKSUM_PARTIAL:
100 *
101 * The device is required to checksum the packet as seen by hard_start_xmit()
102 * from skb->csum_start up to the end, and to record/write the checksum at
103 * offset skb->csum_start + skb->csum_offset.
104 *
105 * The device must show its capabilities in dev->features, set up at device
106 * setup time, e.g. netdev_features.h:
107 *
108 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
109 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
110 * IPv4. Sigh. Vendors like this way for an unknown reason.
111 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
112 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
113 * NETIF_F_... - Well, you get the picture.
114 *
115 * CHECKSUM_UNNECESSARY:
116 *
117 * Normally, the device will do per protocol specific checksumming. Protocol
118 * implementations that do not want the NIC to perform the checksum
119 * calculation should use this flag in their outgoing skbs.
120 *
121 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
122 * offload. Correspondingly, the FCoE protocol driver
123 * stack should use CHECKSUM_UNNECESSARY.
124 *
125 * Any questions? No questions, good. --ANK
126 */
127
60476372 128/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
129#define CHECKSUM_NONE 0
130#define CHECKSUM_UNNECESSARY 1
131#define CHECKSUM_COMPLETE 2
132#define CHECKSUM_PARTIAL 3
1da177e4 133
77cffe23
TH
134/* Maximum value in skb->csum_level */
135#define SKB_MAX_CSUM_LEVEL 3
136
0bec8c88 137#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 138#define SKB_WITH_OVERHEAD(X) \
deea84b0 139 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
140#define SKB_MAX_ORDER(X, ORDER) \
141 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
142#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
143#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
144
87fb4b7b
ED
145/* return minimum truesize of one skb containing X bytes of data */
146#define SKB_TRUESIZE(X) ((X) + \
147 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
148 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
149
1da177e4 150struct net_device;
716ea3a7 151struct scatterlist;
9c55e01c 152struct pipe_inode_info;
a8f820aa 153struct iov_iter;
1da177e4 154
5f79e0f9 155#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
156struct nf_conntrack {
157 atomic_t use;
1da177e4 158};
5f79e0f9 159#endif
1da177e4 160
34666d46 161#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 162struct nf_bridge_info {
bf1ac5ca
ED
163 atomic_t use;
164 unsigned int mask;
165 struct net_device *physindev;
166 struct net_device *physoutdev;
167 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
168};
169#endif
170
1da177e4
LT
171struct sk_buff_head {
172 /* These two members must be first. */
173 struct sk_buff *next;
174 struct sk_buff *prev;
175
176 __u32 qlen;
177 spinlock_t lock;
178};
179
180struct sk_buff;
181
9d4dde52
IC
182/* To allow 64K frame to be packed as single skb without frag_list we
183 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
184 * buffers which do not start on a page boundary.
185 *
186 * Since GRO uses frags we allocate at least 16 regardless of page
187 * size.
a715dea3 188 */
9d4dde52 189#if (65536/PAGE_SIZE + 1) < 16
eec00954 190#define MAX_SKB_FRAGS 16UL
a715dea3 191#else
9d4dde52 192#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 193#endif
1da177e4
LT
194
195typedef struct skb_frag_struct skb_frag_t;
196
197struct skb_frag_struct {
a8605c60
IC
198 struct {
199 struct page *p;
200 } page;
cb4dfe56 201#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
202 __u32 page_offset;
203 __u32 size;
cb4dfe56
ED
204#else
205 __u16 page_offset;
206 __u16 size;
207#endif
1da177e4
LT
208};
209
9e903e08
ED
210static inline unsigned int skb_frag_size(const skb_frag_t *frag)
211{
212 return frag->size;
213}
214
215static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
216{
217 frag->size = size;
218}
219
220static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
221{
222 frag->size += delta;
223}
224
225static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
226{
227 frag->size -= delta;
228}
229
ac45f602
PO
230#define HAVE_HW_TIME_STAMP
231
232/**
d3a21be8 233 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
234 * @hwtstamp: hardware time stamp transformed into duration
235 * since arbitrary point in time
ac45f602
PO
236 *
237 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 238 * skb->tstamp.
ac45f602
PO
239 *
240 * hwtstamps can only be compared against other hwtstamps from
241 * the same device.
242 *
243 * This structure is attached to packets as part of the
244 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
245 */
246struct skb_shared_hwtstamps {
247 ktime_t hwtstamp;
ac45f602
PO
248};
249
2244d07b
OH
250/* Definitions for tx_flags in struct skb_shared_info */
251enum {
252 /* generate hardware time stamp */
253 SKBTX_HW_TSTAMP = 1 << 0,
254
e7fd2885 255 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
256 SKBTX_SW_TSTAMP = 1 << 1,
257
258 /* device driver is going to provide hardware time stamp */
259 SKBTX_IN_PROGRESS = 1 << 2,
260
a6686f2f 261 /* device driver supports TX zero-copy buffers */
62b1a8ab 262 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
263
264 /* generate wifi status information (where possible) */
62b1a8ab 265 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
266
267 /* This indicates at least one fragment might be overwritten
268 * (as in vmsplice(), sendfile() ...)
269 * If we need to compute a TX checksum, we'll need to copy
270 * all frags to avoid possible bad checksum
271 */
272 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
273
274 /* generate software time stamp when entering packet scheduling */
275 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
276
277 /* generate software timestamp on peer data acknowledgment */
278 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
279};
280
e1c8a607
WB
281#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
282 SKBTX_SCHED_TSTAMP | \
283 SKBTX_ACK_TSTAMP)
f24b9be5
WB
284#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
285
a6686f2f
SM
286/*
287 * The callback notifies userspace to release buffers when skb DMA is done in
288 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
289 * The zerocopy_success argument is true if zero copy transmit occurred,
290 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
291 * The ctx field is used to track device context.
292 * The desc field is used to track userspace buffer index.
a6686f2f
SM
293 */
294struct ubuf_info {
e19d6763 295 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 296 void *ctx;
a6686f2f 297 unsigned long desc;
ac45f602
PO
298};
299
1da177e4
LT
300/* This data is invariant across clones and lives at
301 * the end of the header data, ie. at skb->end.
302 */
303struct skb_shared_info {
9f42f126
IC
304 unsigned char nr_frags;
305 __u8 tx_flags;
7967168c
HX
306 unsigned short gso_size;
307 /* Warning: this field is not always filled in (UFO)! */
308 unsigned short gso_segs;
309 unsigned short gso_type;
1da177e4 310 struct sk_buff *frag_list;
ac45f602 311 struct skb_shared_hwtstamps hwtstamps;
09c2d251 312 u32 tskey;
9f42f126 313 __be32 ip6_frag_id;
ec7d2f2c
ED
314
315 /*
316 * Warning : all fields before dataref are cleared in __alloc_skb()
317 */
318 atomic_t dataref;
319
69e3c75f
JB
320 /* Intermediate layers must ensure that destructor_arg
321 * remains valid until skb destructor */
322 void * destructor_arg;
a6686f2f 323
fed66381
ED
324 /* must be last field, see pskb_expand_head() */
325 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
326};
327
328/* We divide dataref into two halves. The higher 16 bits hold references
329 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
330 * the entire skb->data. A clone of a headerless skb holds the length of
331 * the header in skb->hdr_len.
1da177e4
LT
332 *
333 * All users must obey the rule that the skb->data reference count must be
334 * greater than or equal to the payload reference count.
335 *
336 * Holding a reference to the payload part means that the user does not
337 * care about modifications to the header part of skb->data.
338 */
339#define SKB_DATAREF_SHIFT 16
340#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
341
d179cd12
DM
342
343enum {
c8753d55
VS
344 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
345 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
346 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
347 SKB_FCLONE_FREE, /* this companion fclone skb is available */
d179cd12
DM
348};
349
7967168c
HX
350enum {
351 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 352 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
353
354 /* This indicates the skb is from an untrusted source. */
355 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
356
357 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
358 SKB_GSO_TCP_ECN = 1 << 3,
359
360 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
361
362 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
363
364 SKB_GSO_GRE = 1 << 6,
73136267 365
4b28252c 366 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 367
4b28252c 368 SKB_GSO_IPIP = 1 << 8,
cb32f511 369
4b28252c 370 SKB_GSO_SIT = 1 << 9,
61c1db7f 371
4b28252c 372 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
373
374 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 375
59b93b41 376 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
377};
378
2e07fa9c
ACM
379#if BITS_PER_LONG > 32
380#define NET_SKBUFF_DATA_USES_OFFSET 1
381#endif
382
383#ifdef NET_SKBUFF_DATA_USES_OFFSET
384typedef unsigned int sk_buff_data_t;
385#else
386typedef unsigned char *sk_buff_data_t;
387#endif
388
363ec392
ED
389/**
390 * struct skb_mstamp - multi resolution time stamps
391 * @stamp_us: timestamp in us resolution
392 * @stamp_jiffies: timestamp in jiffies
393 */
394struct skb_mstamp {
395 union {
396 u64 v64;
397 struct {
398 u32 stamp_us;
399 u32 stamp_jiffies;
400 };
401 };
402};
403
404/**
405 * skb_mstamp_get - get current timestamp
406 * @cl: place to store timestamps
407 */
408static inline void skb_mstamp_get(struct skb_mstamp *cl)
409{
410 u64 val = local_clock();
411
412 do_div(val, NSEC_PER_USEC);
413 cl->stamp_us = (u32)val;
414 cl->stamp_jiffies = (u32)jiffies;
415}
416
417/**
418 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
419 * @t1: pointer to newest sample
420 * @t0: pointer to oldest sample
421 */
422static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
423 const struct skb_mstamp *t0)
424{
425 s32 delta_us = t1->stamp_us - t0->stamp_us;
426 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
427
428 /* If delta_us is negative, this might be because interval is too big,
429 * or local_clock() drift is too big : fallback using jiffies.
430 */
431 if (delta_us <= 0 ||
432 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
433
434 delta_us = jiffies_to_usecs(delta_jiffies);
435
436 return delta_us;
437}
438
439
1da177e4
LT
440/**
441 * struct sk_buff - socket buffer
442 * @next: Next buffer in list
443 * @prev: Previous buffer in list
363ec392 444 * @tstamp: Time we arrived/left
56b17425 445 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 446 * @sk: Socket we are owned by
1da177e4 447 * @dev: Device we arrived on/are leaving by
d84e0bd7 448 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 449 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 450 * @sp: the security path, used for xfrm
1da177e4
LT
451 * @len: Length of actual data
452 * @data_len: Data length
453 * @mac_len: Length of link layer header
334a8132 454 * @hdr_len: writable header length of cloned skb
663ead3b
HX
455 * @csum: Checksum (must include start/offset pair)
456 * @csum_start: Offset from skb->head where checksumming should start
457 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 458 * @priority: Packet queueing priority
60ff7467 459 * @ignore_df: allow local fragmentation
1da177e4 460 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 461 * @ip_summed: Driver fed us an IP checksum
1da177e4 462 * @nohdr: Payload reference only, must not modify header
d84e0bd7 463 * @nfctinfo: Relationship of this skb to the connection
1da177e4 464 * @pkt_type: Packet class
c83c2486 465 * @fclone: skbuff clone status
c83c2486 466 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
467 * @peeked: this packet has been seen already, so stats have been
468 * done for it, don't do them again
ba9dda3a 469 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
470 * @protocol: Packet protocol from driver
471 * @destructor: Destruct function
472 * @nfct: Associated connection, if any
1da177e4 473 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 474 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
475 * @tc_index: Traffic control index
476 * @tc_verd: traffic control verdict
61b905da 477 * @hash: the packet hash
d84e0bd7 478 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 479 * @xmit_more: More SKBs are pending for this queue
553a5672 480 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 481 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 482 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 483 * ports.
a3b18ddb 484 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
485 * @wifi_acked_valid: wifi_acked was set
486 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 487 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 488 * @napi_id: id of the NAPI struct this skb came from
984bc16c 489 * @secmark: security marking
d84e0bd7
DB
490 * @mark: Generic packet mark
491 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 492 * @vlan_proto: vlan encapsulation protocol
6aa895b0 493 * @vlan_tci: vlan tag control information
0d89d203 494 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
495 * @inner_transport_header: Inner transport layer header (encapsulation)
496 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 497 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
498 * @transport_header: Transport layer header
499 * @network_header: Network layer header
500 * @mac_header: Link layer header
501 * @tail: Tail pointer
502 * @end: End pointer
503 * @head: Head of buffer
504 * @data: Data head pointer
505 * @truesize: Buffer size
506 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
507 */
508
509struct sk_buff {
363ec392 510 union {
56b17425
ED
511 struct {
512 /* These two members must be first. */
513 struct sk_buff *next;
514 struct sk_buff *prev;
515
516 union {
517 ktime_t tstamp;
518 struct skb_mstamp skb_mstamp;
519 };
520 };
521 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 522 };
da3f5cf1 523 struct sock *sk;
1da177e4 524 struct net_device *dev;
1da177e4 525
1da177e4
LT
526 /*
527 * This is the control buffer. It is free to use for every
528 * layer. Please put your private variables there. If you
529 * want to keep them across layers you have to do a skb_clone()
530 * first. This is owned by whoever has the skb queued ATM.
531 */
da3f5cf1 532 char cb[48] __aligned(8);
1da177e4 533
7fee226a 534 unsigned long _skb_refdst;
b1937227 535 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
536#ifdef CONFIG_XFRM
537 struct sec_path *sp;
b1937227
ED
538#endif
539#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
540 struct nf_conntrack *nfct;
541#endif
85224844 542#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 543 struct nf_bridge_info *nf_bridge;
da3f5cf1 544#endif
1da177e4 545 unsigned int len,
334a8132
PM
546 data_len;
547 __u16 mac_len,
548 hdr_len;
b1937227
ED
549
550 /* Following fields are _not_ copied in __copy_skb_header()
551 * Note that queue_mapping is here mostly to fill a hole.
552 */
fe55f6d5 553 kmemcheck_bitfield_begin(flags1);
b1937227
ED
554 __u16 queue_mapping;
555 __u8 cloned:1,
6869c4d8 556 nohdr:1,
b84f4cc9 557 fclone:2,
a59322be 558 peeked:1,
b1937227
ED
559 head_frag:1,
560 xmit_more:1;
561 /* one bit hole */
fe55f6d5 562 kmemcheck_bitfield_end(flags1);
4031ae6e 563
b1937227
ED
564 /* fields enclosed in headers_start/headers_end are copied
565 * using a single memcpy() in __copy_skb_header()
566 */
ebcf34f3 567 /* private: */
b1937227 568 __u32 headers_start[0];
ebcf34f3 569 /* public: */
4031ae6e 570
233577a2
HFS
571/* if you move pkt_type around you also must adapt those constants */
572#ifdef __BIG_ENDIAN_BITFIELD
573#define PKT_TYPE_MAX (7 << 5)
574#else
575#define PKT_TYPE_MAX 7
1da177e4 576#endif
233577a2 577#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 578
233577a2 579 __u8 __pkt_type_offset[0];
b1937227 580 __u8 pkt_type:3;
c93bdd0e 581 __u8 pfmemalloc:1;
b1937227
ED
582 __u8 ignore_df:1;
583 __u8 nfctinfo:3;
584
585 __u8 nf_trace:1;
586 __u8 ip_summed:2;
3853b584 587 __u8 ooo_okay:1;
61b905da 588 __u8 l4_hash:1;
a3b18ddb 589 __u8 sw_hash:1;
6e3e939f
JB
590 __u8 wifi_acked_valid:1;
591 __u8 wifi_acked:1;
b1937227 592
3bdc0eba 593 __u8 no_fcs:1;
77cffe23 594 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 595 __u8 encapsulation:1;
7e2b10c1 596 __u8 encap_hdr_csum:1;
5d0c2b95 597 __u8 csum_valid:1;
7e3cead5 598 __u8 csum_complete_sw:1;
b1937227
ED
599 __u8 csum_level:2;
600 __u8 csum_bad:1;
fe55f6d5 601
b1937227
ED
602#ifdef CONFIG_IPV6_NDISC_NODETYPE
603 __u8 ndisc_nodetype:2;
604#endif
605 __u8 ipvs_property:1;
8bce6d7d 606 __u8 inner_protocol_type:1;
e585f236
TH
607 __u8 remcsum_offload:1;
608 /* 3 or 5 bit hole */
b1937227
ED
609
610#ifdef CONFIG_NET_SCHED
611 __u16 tc_index; /* traffic control index */
612#ifdef CONFIG_NET_CLS_ACT
613 __u16 tc_verd; /* traffic control verdict */
614#endif
615#endif
fe55f6d5 616
b1937227
ED
617 union {
618 __wsum csum;
619 struct {
620 __u16 csum_start;
621 __u16 csum_offset;
622 };
623 };
624 __u32 priority;
625 int skb_iif;
626 __u32 hash;
627 __be16 vlan_proto;
628 __u16 vlan_tci;
7bced397
DW
629#ifdef CONFIG_NET_RX_BUSY_POLL
630 unsigned int napi_id;
97fc2f08 631#endif
984bc16c
JM
632#ifdef CONFIG_NETWORK_SECMARK
633 __u32 secmark;
634#endif
3b885787
NH
635 union {
636 __u32 mark;
637 __u32 dropcount;
16fad69c 638 __u32 reserved_tailroom;
3b885787 639 };
1da177e4 640
8bce6d7d
TH
641 union {
642 __be16 inner_protocol;
643 __u8 inner_ipproto;
644 };
645
1a37e412
SH
646 __u16 inner_transport_header;
647 __u16 inner_network_header;
648 __u16 inner_mac_header;
b1937227
ED
649
650 __be16 protocol;
1a37e412
SH
651 __u16 transport_header;
652 __u16 network_header;
653 __u16 mac_header;
b1937227 654
ebcf34f3 655 /* private: */
b1937227 656 __u32 headers_end[0];
ebcf34f3 657 /* public: */
b1937227 658
1da177e4 659 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 660 sk_buff_data_t tail;
4305b541 661 sk_buff_data_t end;
1da177e4 662 unsigned char *head,
4305b541 663 *data;
27a884dc
ACM
664 unsigned int truesize;
665 atomic_t users;
1da177e4
LT
666};
667
668#ifdef __KERNEL__
669/*
670 * Handling routines are only of interest to the kernel
671 */
672#include <linux/slab.h>
673
1da177e4 674
c93bdd0e
MG
675#define SKB_ALLOC_FCLONE 0x01
676#define SKB_ALLOC_RX 0x02
677
678/* Returns true if the skb was allocated from PFMEMALLOC reserves */
679static inline bool skb_pfmemalloc(const struct sk_buff *skb)
680{
681 return unlikely(skb->pfmemalloc);
682}
683
7fee226a
ED
684/*
685 * skb might have a dst pointer attached, refcounted or not.
686 * _skb_refdst low order bit is set if refcount was _not_ taken
687 */
688#define SKB_DST_NOREF 1UL
689#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
690
691/**
692 * skb_dst - returns skb dst_entry
693 * @skb: buffer
694 *
695 * Returns skb dst_entry, regardless of reference taken or not.
696 */
adf30907
ED
697static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
698{
7fee226a
ED
699 /* If refdst was not refcounted, check we still are in a
700 * rcu_read_lock section
701 */
702 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
703 !rcu_read_lock_held() &&
704 !rcu_read_lock_bh_held());
705 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
706}
707
7fee226a
ED
708/**
709 * skb_dst_set - sets skb dst
710 * @skb: buffer
711 * @dst: dst entry
712 *
713 * Sets skb dst, assuming a reference was taken on dst and should
714 * be released by skb_dst_drop()
715 */
adf30907
ED
716static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
717{
7fee226a
ED
718 skb->_skb_refdst = (unsigned long)dst;
719}
720
7965bd4d
JP
721void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
722 bool force);
932bc4d7
JA
723
724/**
725 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
726 * @skb: buffer
727 * @dst: dst entry
728 *
729 * Sets skb dst, assuming a reference was not taken on dst.
730 * If dst entry is cached, we do not take reference and dst_release
731 * will be avoided by refdst_drop. If dst entry is not cached, we take
732 * reference, so that last dst_release can destroy the dst immediately.
733 */
734static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
735{
736 __skb_dst_set_noref(skb, dst, false);
737}
738
739/**
740 * skb_dst_set_noref_force - sets skb dst, without taking reference
741 * @skb: buffer
742 * @dst: dst entry
743 *
744 * Sets skb dst, assuming a reference was not taken on dst.
745 * No reference is taken and no dst_release will be called. While for
746 * cached dsts deferred reclaim is a basic feature, for entries that are
747 * not cached it is caller's job to guarantee that last dst_release for
748 * provided dst happens when nobody uses it, eg. after a RCU grace period.
749 */
750static inline void skb_dst_set_noref_force(struct sk_buff *skb,
751 struct dst_entry *dst)
752{
753 __skb_dst_set_noref(skb, dst, true);
754}
7fee226a
ED
755
756/**
25985edc 757 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
758 * @skb: buffer
759 */
760static inline bool skb_dst_is_noref(const struct sk_buff *skb)
761{
762 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
763}
764
511c3f92
ED
765static inline struct rtable *skb_rtable(const struct sk_buff *skb)
766{
adf30907 767 return (struct rtable *)skb_dst(skb);
511c3f92
ED
768}
769
7965bd4d
JP
770void kfree_skb(struct sk_buff *skb);
771void kfree_skb_list(struct sk_buff *segs);
772void skb_tx_error(struct sk_buff *skb);
773void consume_skb(struct sk_buff *skb);
774void __kfree_skb(struct sk_buff *skb);
d7e8883c 775extern struct kmem_cache *skbuff_head_cache;
bad43ca8 776
7965bd4d
JP
777void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
778bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
779 bool *fragstolen, int *delta_truesize);
bad43ca8 780
7965bd4d
JP
781struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
782 int node);
783struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 784static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 785 gfp_t priority)
d179cd12 786{
564824b0 787 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
788}
789
2e4e4410
ED
790struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
791 unsigned long data_len,
792 int max_page_order,
793 int *errcode,
794 gfp_t gfp_mask);
795
d0bf4a9e
ED
796/* Layout of fast clones : [skb1][skb2][fclone_ref] */
797struct sk_buff_fclones {
798 struct sk_buff skb1;
799
800 struct sk_buff skb2;
801
802 atomic_t fclone_ref;
803};
804
805/**
806 * skb_fclone_busy - check if fclone is busy
807 * @skb: buffer
808 *
809 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
810 * Some drivers call skb_orphan() in their ndo_start_xmit(),
811 * so we also check that this didnt happen.
d0bf4a9e 812 */
39bb5e62
ED
813static inline bool skb_fclone_busy(const struct sock *sk,
814 const struct sk_buff *skb)
d0bf4a9e
ED
815{
816 const struct sk_buff_fclones *fclones;
817
818 fclones = container_of(skb, struct sk_buff_fclones, skb1);
819
820 return skb->fclone == SKB_FCLONE_ORIG &&
39bb5e62
ED
821 fclones->skb2.fclone == SKB_FCLONE_CLONE &&
822 fclones->skb2.sk == sk;
d0bf4a9e
ED
823}
824
d179cd12 825static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 826 gfp_t priority)
d179cd12 827{
c93bdd0e 828 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
829}
830
7965bd4d 831struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
832static inline struct sk_buff *alloc_skb_head(gfp_t priority)
833{
834 return __alloc_skb_head(priority, -1);
835}
836
7965bd4d
JP
837struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
838int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
839struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
840struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
841struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
842 gfp_t gfp_mask, bool fclone);
843static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
844 gfp_t gfp_mask)
845{
846 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
847}
7965bd4d
JP
848
849int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
850struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
851 unsigned int headroom);
852struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
853 int newtailroom, gfp_t priority);
25a91d8d
FD
854int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
855 int offset, int len);
7965bd4d
JP
856int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
857 int len);
858int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
859int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 860#define dev_kfree_skb(a) consume_skb(a)
1da177e4 861
7965bd4d
JP
862int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
863 int getfrag(void *from, char *to, int offset,
864 int len, int odd, struct sk_buff *skb),
865 void *from, int length);
e89e9cf5 866
d94d9fee 867struct skb_seq_state {
677e90ed
TG
868 __u32 lower_offset;
869 __u32 upper_offset;
870 __u32 frag_idx;
871 __u32 stepped_offset;
872 struct sk_buff *root_skb;
873 struct sk_buff *cur_skb;
874 __u8 *frag_data;
875};
876
7965bd4d
JP
877void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
878 unsigned int to, struct skb_seq_state *st);
879unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
880 struct skb_seq_state *st);
881void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 882
7965bd4d
JP
883unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
884 unsigned int to, struct ts_config *config,
885 struct ts_state *state);
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
TH
928void __skb_get_hash(struct sk_buff *skb);
929static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 930{
a3b18ddb 931 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 932 __skb_get_hash(skb);
bfb564e7 933
61b905da 934 return skb->hash;
bfb564e7
KK
935}
936
57bdf7f4
TH
937static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
938{
61b905da 939 return skb->hash;
57bdf7f4
TH
940}
941
7539fadc
TH
942static inline void skb_clear_hash(struct sk_buff *skb)
943{
61b905da 944 skb->hash = 0;
a3b18ddb 945 skb->sw_hash = 0;
61b905da 946 skb->l4_hash = 0;
7539fadc
TH
947}
948
949static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
950{
61b905da 951 if (!skb->l4_hash)
7539fadc
TH
952 skb_clear_hash(skb);
953}
954
3df7a74e
TH
955static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
956{
61b905da 957 to->hash = from->hash;
a3b18ddb 958 to->sw_hash = from->sw_hash;
61b905da 959 to->l4_hash = from->l4_hash;
3df7a74e
TH
960};
961
4305b541
ACM
962#ifdef NET_SKBUFF_DATA_USES_OFFSET
963static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
964{
965 return skb->head + skb->end;
966}
ec47ea82
AD
967
968static inline unsigned int skb_end_offset(const struct sk_buff *skb)
969{
970 return skb->end;
971}
4305b541
ACM
972#else
973static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
974{
975 return skb->end;
976}
ec47ea82
AD
977
978static inline unsigned int skb_end_offset(const struct sk_buff *skb)
979{
980 return skb->end - skb->head;
981}
4305b541
ACM
982#endif
983
1da177e4 984/* Internal */
4305b541 985#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 986
ac45f602
PO
987static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
988{
989 return &skb_shinfo(skb)->hwtstamps;
990}
991
1da177e4
LT
992/**
993 * skb_queue_empty - check if a queue is empty
994 * @list: queue head
995 *
996 * Returns true if the queue is empty, false otherwise.
997 */
998static inline int skb_queue_empty(const struct sk_buff_head *list)
999{
fd44b93c 1000 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1001}
1002
fc7ebb21
DM
1003/**
1004 * skb_queue_is_last - check if skb is the last entry in the queue
1005 * @list: queue head
1006 * @skb: buffer
1007 *
1008 * Returns true if @skb is the last buffer on the list.
1009 */
1010static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1011 const struct sk_buff *skb)
1012{
fd44b93c 1013 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1014}
1015
832d11c5
IJ
1016/**
1017 * skb_queue_is_first - check if skb is the first entry in the queue
1018 * @list: queue head
1019 * @skb: buffer
1020 *
1021 * Returns true if @skb is the first buffer on the list.
1022 */
1023static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1024 const struct sk_buff *skb)
1025{
fd44b93c 1026 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1027}
1028
249c8b42
DM
1029/**
1030 * skb_queue_next - return the next packet in the queue
1031 * @list: queue head
1032 * @skb: current buffer
1033 *
1034 * Return the next packet in @list after @skb. It is only valid to
1035 * call this if skb_queue_is_last() evaluates to false.
1036 */
1037static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1038 const struct sk_buff *skb)
1039{
1040 /* This BUG_ON may seem severe, but if we just return then we
1041 * are going to dereference garbage.
1042 */
1043 BUG_ON(skb_queue_is_last(list, skb));
1044 return skb->next;
1045}
1046
832d11c5
IJ
1047/**
1048 * skb_queue_prev - return the prev packet in the queue
1049 * @list: queue head
1050 * @skb: current buffer
1051 *
1052 * Return the prev packet in @list before @skb. It is only valid to
1053 * call this if skb_queue_is_first() evaluates to false.
1054 */
1055static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1056 const struct sk_buff *skb)
1057{
1058 /* This BUG_ON may seem severe, but if we just return then we
1059 * are going to dereference garbage.
1060 */
1061 BUG_ON(skb_queue_is_first(list, skb));
1062 return skb->prev;
1063}
1064
1da177e4
LT
1065/**
1066 * skb_get - reference buffer
1067 * @skb: buffer to reference
1068 *
1069 * Makes another reference to a socket buffer and returns a pointer
1070 * to the buffer.
1071 */
1072static inline struct sk_buff *skb_get(struct sk_buff *skb)
1073{
1074 atomic_inc(&skb->users);
1075 return skb;
1076}
1077
1078/*
1079 * If users == 1, we are the only owner and are can avoid redundant
1080 * atomic change.
1081 */
1082
1da177e4
LT
1083/**
1084 * skb_cloned - is the buffer a clone
1085 * @skb: buffer to check
1086 *
1087 * Returns true if the buffer was generated with skb_clone() and is
1088 * one of multiple shared copies of the buffer. Cloned buffers are
1089 * shared data so must not be written to under normal circumstances.
1090 */
1091static inline int skb_cloned(const struct sk_buff *skb)
1092{
1093 return skb->cloned &&
1094 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1095}
1096
14bbd6a5
PS
1097static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1098{
1099 might_sleep_if(pri & __GFP_WAIT);
1100
1101 if (skb_cloned(skb))
1102 return pskb_expand_head(skb, 0, 0, pri);
1103
1104 return 0;
1105}
1106
1da177e4
LT
1107/**
1108 * skb_header_cloned - is the header a clone
1109 * @skb: buffer to check
1110 *
1111 * Returns true if modifying the header part of the buffer requires
1112 * the data to be copied.
1113 */
1114static inline int skb_header_cloned(const struct sk_buff *skb)
1115{
1116 int dataref;
1117
1118 if (!skb->cloned)
1119 return 0;
1120
1121 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1122 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1123 return dataref != 1;
1124}
1125
1126/**
1127 * skb_header_release - release reference to header
1128 * @skb: buffer to operate on
1129 *
1130 * Drop a reference to the header part of the buffer. This is done
1131 * by acquiring a payload reference. You must not read from the header
1132 * part of skb->data after this.
f4a775d1 1133 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1134 */
1135static inline void skb_header_release(struct sk_buff *skb)
1136{
1137 BUG_ON(skb->nohdr);
1138 skb->nohdr = 1;
1139 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1140}
1141
f4a775d1
ED
1142/**
1143 * __skb_header_release - release reference to header
1144 * @skb: buffer to operate on
1145 *
1146 * Variant of skb_header_release() assuming skb is private to caller.
1147 * We can avoid one atomic operation.
1148 */
1149static inline void __skb_header_release(struct sk_buff *skb)
1150{
1151 skb->nohdr = 1;
1152 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1153}
1154
1155
1da177e4
LT
1156/**
1157 * skb_shared - is the buffer shared
1158 * @skb: buffer to check
1159 *
1160 * Returns true if more than one person has a reference to this
1161 * buffer.
1162 */
1163static inline int skb_shared(const struct sk_buff *skb)
1164{
1165 return atomic_read(&skb->users) != 1;
1166}
1167
1168/**
1169 * skb_share_check - check if buffer is shared and if so clone it
1170 * @skb: buffer to check
1171 * @pri: priority for memory allocation
1172 *
1173 * If the buffer is shared the buffer is cloned and the old copy
1174 * drops a reference. A new clone with a single reference is returned.
1175 * If the buffer is not shared the original buffer is returned. When
1176 * being called from interrupt status or with spinlocks held pri must
1177 * be GFP_ATOMIC.
1178 *
1179 * NULL is returned on a memory allocation failure.
1180 */
47061bc4 1181static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1182{
1183 might_sleep_if(pri & __GFP_WAIT);
1184 if (skb_shared(skb)) {
1185 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1186
1187 if (likely(nskb))
1188 consume_skb(skb);
1189 else
1190 kfree_skb(skb);
1da177e4
LT
1191 skb = nskb;
1192 }
1193 return skb;
1194}
1195
1196/*
1197 * Copy shared buffers into a new sk_buff. We effectively do COW on
1198 * packets to handle cases where we have a local reader and forward
1199 * and a couple of other messy ones. The normal one is tcpdumping
1200 * a packet thats being forwarded.
1201 */
1202
1203/**
1204 * skb_unshare - make a copy of a shared buffer
1205 * @skb: buffer to check
1206 * @pri: priority for memory allocation
1207 *
1208 * If the socket buffer is a clone then this function creates a new
1209 * copy of the data, drops a reference count on the old copy and returns
1210 * the new copy with the reference count at 1. If the buffer is not a clone
1211 * the original buffer is returned. When called with a spinlock held or
1212 * from interrupt state @pri must be %GFP_ATOMIC
1213 *
1214 * %NULL is returned on a memory allocation failure.
1215 */
e2bf521d 1216static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1217 gfp_t pri)
1da177e4
LT
1218{
1219 might_sleep_if(pri & __GFP_WAIT);
1220 if (skb_cloned(skb)) {
1221 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1222
1223 /* Free our shared copy */
1224 if (likely(nskb))
1225 consume_skb(skb);
1226 else
1227 kfree_skb(skb);
1da177e4
LT
1228 skb = nskb;
1229 }
1230 return skb;
1231}
1232
1233/**
1a5778aa 1234 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1235 * @list_: list to peek at
1236 *
1237 * Peek an &sk_buff. Unlike most other operations you _MUST_
1238 * be careful with this one. A peek leaves the buffer on the
1239 * list and someone else may run off with it. You must hold
1240 * the appropriate locks or have a private queue to do this.
1241 *
1242 * Returns %NULL for an empty list or a pointer to the head element.
1243 * The reference count is not incremented and the reference is therefore
1244 * volatile. Use with caution.
1245 */
05bdd2f1 1246static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1247{
18d07000
ED
1248 struct sk_buff *skb = list_->next;
1249
1250 if (skb == (struct sk_buff *)list_)
1251 skb = NULL;
1252 return skb;
1da177e4
LT
1253}
1254
da5ef6e5
PE
1255/**
1256 * skb_peek_next - peek skb following the given one from a queue
1257 * @skb: skb to start from
1258 * @list_: list to peek at
1259 *
1260 * Returns %NULL when the end of the list is met or a pointer to the
1261 * next element. The reference count is not incremented and the
1262 * reference is therefore volatile. Use with caution.
1263 */
1264static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1265 const struct sk_buff_head *list_)
1266{
1267 struct sk_buff *next = skb->next;
18d07000 1268
da5ef6e5
PE
1269 if (next == (struct sk_buff *)list_)
1270 next = NULL;
1271 return next;
1272}
1273
1da177e4 1274/**
1a5778aa 1275 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1276 * @list_: list to peek at
1277 *
1278 * Peek an &sk_buff. Unlike most other operations you _MUST_
1279 * be careful with this one. A peek leaves the buffer on the
1280 * list and someone else may run off with it. You must hold
1281 * the appropriate locks or have a private queue to do this.
1282 *
1283 * Returns %NULL for an empty list or a pointer to the tail element.
1284 * The reference count is not incremented and the reference is therefore
1285 * volatile. Use with caution.
1286 */
05bdd2f1 1287static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1288{
18d07000
ED
1289 struct sk_buff *skb = list_->prev;
1290
1291 if (skb == (struct sk_buff *)list_)
1292 skb = NULL;
1293 return skb;
1294
1da177e4
LT
1295}
1296
1297/**
1298 * skb_queue_len - get queue length
1299 * @list_: list to measure
1300 *
1301 * Return the length of an &sk_buff queue.
1302 */
1303static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1304{
1305 return list_->qlen;
1306}
1307
67fed459
DM
1308/**
1309 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1310 * @list: queue to initialize
1311 *
1312 * This initializes only the list and queue length aspects of
1313 * an sk_buff_head object. This allows to initialize the list
1314 * aspects of an sk_buff_head without reinitializing things like
1315 * the spinlock. It can also be used for on-stack sk_buff_head
1316 * objects where the spinlock is known to not be used.
1317 */
1318static inline void __skb_queue_head_init(struct sk_buff_head *list)
1319{
1320 list->prev = list->next = (struct sk_buff *)list;
1321 list->qlen = 0;
1322}
1323
76f10ad0
AV
1324/*
1325 * This function creates a split out lock class for each invocation;
1326 * this is needed for now since a whole lot of users of the skb-queue
1327 * infrastructure in drivers have different locking usage (in hardirq)
1328 * than the networking core (in softirq only). In the long run either the
1329 * network layer or drivers should need annotation to consolidate the
1330 * main types of usage into 3 classes.
1331 */
1da177e4
LT
1332static inline void skb_queue_head_init(struct sk_buff_head *list)
1333{
1334 spin_lock_init(&list->lock);
67fed459 1335 __skb_queue_head_init(list);
1da177e4
LT
1336}
1337
c2ecba71
PE
1338static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1339 struct lock_class_key *class)
1340{
1341 skb_queue_head_init(list);
1342 lockdep_set_class(&list->lock, class);
1343}
1344
1da177e4 1345/*
bf299275 1346 * Insert an sk_buff on a list.
1da177e4
LT
1347 *
1348 * The "__skb_xxxx()" functions are the non-atomic ones that
1349 * can only be called with interrupts disabled.
1350 */
7965bd4d
JP
1351void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1352 struct sk_buff_head *list);
bf299275
GR
1353static inline void __skb_insert(struct sk_buff *newsk,
1354 struct sk_buff *prev, struct sk_buff *next,
1355 struct sk_buff_head *list)
1356{
1357 newsk->next = next;
1358 newsk->prev = prev;
1359 next->prev = prev->next = newsk;
1360 list->qlen++;
1361}
1da177e4 1362
67fed459
DM
1363static inline void __skb_queue_splice(const struct sk_buff_head *list,
1364 struct sk_buff *prev,
1365 struct sk_buff *next)
1366{
1367 struct sk_buff *first = list->next;
1368 struct sk_buff *last = list->prev;
1369
1370 first->prev = prev;
1371 prev->next = first;
1372
1373 last->next = next;
1374 next->prev = last;
1375}
1376
1377/**
1378 * skb_queue_splice - join two skb lists, this is designed for stacks
1379 * @list: the new list to add
1380 * @head: the place to add it in the first list
1381 */
1382static inline void skb_queue_splice(const struct sk_buff_head *list,
1383 struct sk_buff_head *head)
1384{
1385 if (!skb_queue_empty(list)) {
1386 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1387 head->qlen += list->qlen;
67fed459
DM
1388 }
1389}
1390
1391/**
d9619496 1392 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1393 * @list: the new list to add
1394 * @head: the place to add it in the first list
1395 *
1396 * The list at @list is reinitialised
1397 */
1398static inline void skb_queue_splice_init(struct sk_buff_head *list,
1399 struct sk_buff_head *head)
1400{
1401 if (!skb_queue_empty(list)) {
1402 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1403 head->qlen += list->qlen;
67fed459
DM
1404 __skb_queue_head_init(list);
1405 }
1406}
1407
1408/**
1409 * skb_queue_splice_tail - join two skb lists, each list being a queue
1410 * @list: the new list to add
1411 * @head: the place to add it in the first list
1412 */
1413static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1414 struct sk_buff_head *head)
1415{
1416 if (!skb_queue_empty(list)) {
1417 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1418 head->qlen += list->qlen;
67fed459
DM
1419 }
1420}
1421
1422/**
d9619496 1423 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1424 * @list: the new list to add
1425 * @head: the place to add it in the first list
1426 *
1427 * Each of the lists is a queue.
1428 * The list at @list is reinitialised
1429 */
1430static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1431 struct sk_buff_head *head)
1432{
1433 if (!skb_queue_empty(list)) {
1434 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1435 head->qlen += list->qlen;
67fed459
DM
1436 __skb_queue_head_init(list);
1437 }
1438}
1439
1da177e4 1440/**
300ce174 1441 * __skb_queue_after - queue a buffer at the list head
1da177e4 1442 * @list: list to use
300ce174 1443 * @prev: place after this buffer
1da177e4
LT
1444 * @newsk: buffer to queue
1445 *
300ce174 1446 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1447 * and you must therefore hold required locks before calling it.
1448 *
1449 * A buffer cannot be placed on two lists at the same time.
1450 */
300ce174
SH
1451static inline void __skb_queue_after(struct sk_buff_head *list,
1452 struct sk_buff *prev,
1453 struct sk_buff *newsk)
1da177e4 1454{
bf299275 1455 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1456}
1457
7965bd4d
JP
1458void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1459 struct sk_buff_head *list);
7de6c033 1460
f5572855
GR
1461static inline void __skb_queue_before(struct sk_buff_head *list,
1462 struct sk_buff *next,
1463 struct sk_buff *newsk)
1464{
1465 __skb_insert(newsk, next->prev, next, list);
1466}
1467
300ce174
SH
1468/**
1469 * __skb_queue_head - queue a buffer at the list head
1470 * @list: list to use
1471 * @newsk: buffer to queue
1472 *
1473 * Queue a buffer at the start of a list. This function takes no locks
1474 * and you must therefore hold required locks before calling it.
1475 *
1476 * A buffer cannot be placed on two lists at the same time.
1477 */
7965bd4d 1478void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1479static inline void __skb_queue_head(struct sk_buff_head *list,
1480 struct sk_buff *newsk)
1481{
1482 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1483}
1484
1da177e4
LT
1485/**
1486 * __skb_queue_tail - queue a buffer at the list tail
1487 * @list: list to use
1488 * @newsk: buffer to queue
1489 *
1490 * Queue a buffer at the end of a list. This function takes no locks
1491 * and you must therefore hold required locks before calling it.
1492 *
1493 * A buffer cannot be placed on two lists at the same time.
1494 */
7965bd4d 1495void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1496static inline void __skb_queue_tail(struct sk_buff_head *list,
1497 struct sk_buff *newsk)
1498{
f5572855 1499 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1500}
1501
1da177e4
LT
1502/*
1503 * remove sk_buff from list. _Must_ be called atomically, and with
1504 * the list known..
1505 */
7965bd4d 1506void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1507static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1508{
1509 struct sk_buff *next, *prev;
1510
1511 list->qlen--;
1512 next = skb->next;
1513 prev = skb->prev;
1514 skb->next = skb->prev = NULL;
1da177e4
LT
1515 next->prev = prev;
1516 prev->next = next;
1517}
1518
f525c06d
GR
1519/**
1520 * __skb_dequeue - remove from the head of the queue
1521 * @list: list to dequeue from
1522 *
1523 * Remove the head of the list. This function does not take any locks
1524 * so must be used with appropriate locks held only. The head item is
1525 * returned or %NULL if the list is empty.
1526 */
7965bd4d 1527struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1528static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1529{
1530 struct sk_buff *skb = skb_peek(list);
1531 if (skb)
1532 __skb_unlink(skb, list);
1533 return skb;
1534}
1da177e4
LT
1535
1536/**
1537 * __skb_dequeue_tail - remove from the tail of the queue
1538 * @list: list to dequeue from
1539 *
1540 * Remove the tail of the list. This function does not take any locks
1541 * so must be used with appropriate locks held only. The tail item is
1542 * returned or %NULL if the list is empty.
1543 */
7965bd4d 1544struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1545static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1546{
1547 struct sk_buff *skb = skb_peek_tail(list);
1548 if (skb)
1549 __skb_unlink(skb, list);
1550 return skb;
1551}
1552
1553
bdcc0924 1554static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1555{
1556 return skb->data_len;
1557}
1558
1559static inline unsigned int skb_headlen(const struct sk_buff *skb)
1560{
1561 return skb->len - skb->data_len;
1562}
1563
1564static inline int skb_pagelen(const struct sk_buff *skb)
1565{
1566 int i, len = 0;
1567
1568 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1569 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1570 return len + skb_headlen(skb);
1571}
1572
131ea667
IC
1573/**
1574 * __skb_fill_page_desc - initialise a paged fragment in an skb
1575 * @skb: buffer containing fragment to be initialised
1576 * @i: paged fragment index to initialise
1577 * @page: the page to use for this fragment
1578 * @off: the offset to the data with @page
1579 * @size: the length of the data
1580 *
1581 * Initialises the @i'th fragment of @skb to point to &size bytes at
1582 * offset @off within @page.
1583 *
1584 * Does not take any additional reference on the fragment.
1585 */
1586static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1587 struct page *page, int off, int size)
1da177e4
LT
1588{
1589 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1590
c48a11c7
MG
1591 /*
1592 * Propagate page->pfmemalloc to the skb if we can. The problem is
1593 * that not all callers have unique ownership of the page. If
1594 * pfmemalloc is set, we check the mapping as a mapping implies
1595 * page->index is set (index and pfmemalloc share space).
1596 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1597 * do not lose pfmemalloc information as the pages would not be
1598 * allocated using __GFP_MEMALLOC.
1599 */
a8605c60 1600 frag->page.p = page;
1da177e4 1601 frag->page_offset = off;
9e903e08 1602 skb_frag_size_set(frag, size);
cca7af38
PE
1603
1604 page = compound_head(page);
1605 if (page->pfmemalloc && !page->mapping)
1606 skb->pfmemalloc = true;
131ea667
IC
1607}
1608
1609/**
1610 * skb_fill_page_desc - initialise a paged fragment in an skb
1611 * @skb: buffer containing fragment to be initialised
1612 * @i: paged fragment index to initialise
1613 * @page: the page to use for this fragment
1614 * @off: the offset to the data with @page
1615 * @size: the length of the data
1616 *
1617 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1618 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1619 * addition updates @skb such that @i is the last fragment.
1620 *
1621 * Does not take any additional reference on the fragment.
1622 */
1623static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1624 struct page *page, int off, int size)
1625{
1626 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1627 skb_shinfo(skb)->nr_frags = i + 1;
1628}
1629
7965bd4d
JP
1630void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1631 int size, unsigned int truesize);
654bed16 1632
f8e617e1
JW
1633void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1634 unsigned int truesize);
1635
1da177e4 1636#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1637#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1638#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1639
27a884dc
ACM
1640#ifdef NET_SKBUFF_DATA_USES_OFFSET
1641static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1642{
1643 return skb->head + skb->tail;
1644}
1645
1646static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1647{
1648 skb->tail = skb->data - skb->head;
1649}
1650
1651static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1652{
1653 skb_reset_tail_pointer(skb);
1654 skb->tail += offset;
1655}
7cc46190 1656
27a884dc
ACM
1657#else /* NET_SKBUFF_DATA_USES_OFFSET */
1658static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1659{
1660 return skb->tail;
1661}
1662
1663static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1664{
1665 skb->tail = skb->data;
1666}
1667
1668static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1669{
1670 skb->tail = skb->data + offset;
1671}
4305b541 1672
27a884dc
ACM
1673#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1674
1da177e4
LT
1675/*
1676 * Add data to an sk_buff
1677 */
0c7ddf36 1678unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1679unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1680static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1681{
27a884dc 1682 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1683 SKB_LINEAR_ASSERT(skb);
1684 skb->tail += len;
1685 skb->len += len;
1686 return tmp;
1687}
1688
7965bd4d 1689unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1690static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1691{
1692 skb->data -= len;
1693 skb->len += len;
1694 return skb->data;
1695}
1696
7965bd4d 1697unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1698static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1699{
1700 skb->len -= len;
1701 BUG_ON(skb->len < skb->data_len);
1702 return skb->data += len;
1703}
1704
47d29646
DM
1705static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1706{
1707 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1708}
1709
7965bd4d 1710unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1711
1712static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1713{
1714 if (len > skb_headlen(skb) &&
987c402a 1715 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1716 return NULL;
1717 skb->len -= len;
1718 return skb->data += len;
1719}
1720
1721static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1722{
1723 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1724}
1725
1726static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1727{
1728 if (likely(len <= skb_headlen(skb)))
1729 return 1;
1730 if (unlikely(len > skb->len))
1731 return 0;
987c402a 1732 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1733}
1734
1735/**
1736 * skb_headroom - bytes at buffer head
1737 * @skb: buffer to check
1738 *
1739 * Return the number of bytes of free space at the head of an &sk_buff.
1740 */
c2636b4d 1741static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1742{
1743 return skb->data - skb->head;
1744}
1745
1746/**
1747 * skb_tailroom - bytes at buffer end
1748 * @skb: buffer to check
1749 *
1750 * Return the number of bytes of free space at the tail of an sk_buff
1751 */
1752static inline int skb_tailroom(const struct sk_buff *skb)
1753{
4305b541 1754 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1755}
1756
a21d4572
ED
1757/**
1758 * skb_availroom - bytes at buffer end
1759 * @skb: buffer to check
1760 *
1761 * Return the number of bytes of free space at the tail of an sk_buff
1762 * allocated by sk_stream_alloc()
1763 */
1764static inline int skb_availroom(const struct sk_buff *skb)
1765{
16fad69c
ED
1766 if (skb_is_nonlinear(skb))
1767 return 0;
1768
1769 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1770}
1771
1da177e4
LT
1772/**
1773 * skb_reserve - adjust headroom
1774 * @skb: buffer to alter
1775 * @len: bytes to move
1776 *
1777 * Increase the headroom of an empty &sk_buff by reducing the tail
1778 * room. This is only allowed for an empty buffer.
1779 */
8243126c 1780static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1781{
1782 skb->data += len;
1783 skb->tail += len;
1784}
1785
8bce6d7d
TH
1786#define ENCAP_TYPE_ETHER 0
1787#define ENCAP_TYPE_IPPROTO 1
1788
1789static inline void skb_set_inner_protocol(struct sk_buff *skb,
1790 __be16 protocol)
1791{
1792 skb->inner_protocol = protocol;
1793 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1794}
1795
1796static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1797 __u8 ipproto)
1798{
1799 skb->inner_ipproto = ipproto;
1800 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1801}
1802
6a674e9c
JG
1803static inline void skb_reset_inner_headers(struct sk_buff *skb)
1804{
aefbd2b3 1805 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1806 skb->inner_network_header = skb->network_header;
1807 skb->inner_transport_header = skb->transport_header;
1808}
1809
0b5c9db1
JP
1810static inline void skb_reset_mac_len(struct sk_buff *skb)
1811{
1812 skb->mac_len = skb->network_header - skb->mac_header;
1813}
1814
6a674e9c
JG
1815static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1816 *skb)
1817{
1818 return skb->head + skb->inner_transport_header;
1819}
1820
1821static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1822{
1823 skb->inner_transport_header = skb->data - skb->head;
1824}
1825
1826static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1827 const int offset)
1828{
1829 skb_reset_inner_transport_header(skb);
1830 skb->inner_transport_header += offset;
1831}
1832
1833static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1834{
1835 return skb->head + skb->inner_network_header;
1836}
1837
1838static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1839{
1840 skb->inner_network_header = skb->data - skb->head;
1841}
1842
1843static inline void skb_set_inner_network_header(struct sk_buff *skb,
1844 const int offset)
1845{
1846 skb_reset_inner_network_header(skb);
1847 skb->inner_network_header += offset;
1848}
1849
aefbd2b3
PS
1850static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1851{
1852 return skb->head + skb->inner_mac_header;
1853}
1854
1855static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1856{
1857 skb->inner_mac_header = skb->data - skb->head;
1858}
1859
1860static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1861 const int offset)
1862{
1863 skb_reset_inner_mac_header(skb);
1864 skb->inner_mac_header += offset;
1865}
fda55eca
ED
1866static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1867{
35d04610 1868 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1869}
1870
9c70220b
ACM
1871static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1872{
2e07fa9c 1873 return skb->head + skb->transport_header;
9c70220b
ACM
1874}
1875
badff6d0
ACM
1876static inline void skb_reset_transport_header(struct sk_buff *skb)
1877{
2e07fa9c 1878 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1879}
1880
967b05f6
ACM
1881static inline void skb_set_transport_header(struct sk_buff *skb,
1882 const int offset)
1883{
2e07fa9c
ACM
1884 skb_reset_transport_header(skb);
1885 skb->transport_header += offset;
ea2ae17d
ACM
1886}
1887
d56f90a7
ACM
1888static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1889{
2e07fa9c 1890 return skb->head + skb->network_header;
d56f90a7
ACM
1891}
1892
c1d2bbe1
ACM
1893static inline void skb_reset_network_header(struct sk_buff *skb)
1894{
2e07fa9c 1895 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1896}
1897
c14d2450
ACM
1898static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1899{
2e07fa9c
ACM
1900 skb_reset_network_header(skb);
1901 skb->network_header += offset;
c14d2450
ACM
1902}
1903
2e07fa9c 1904static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1905{
2e07fa9c 1906 return skb->head + skb->mac_header;
bbe735e4
ACM
1907}
1908
2e07fa9c 1909static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1910{
35d04610 1911 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1912}
1913
1914static inline void skb_reset_mac_header(struct sk_buff *skb)
1915{
1916 skb->mac_header = skb->data - skb->head;
1917}
1918
1919static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1920{
1921 skb_reset_mac_header(skb);
1922 skb->mac_header += offset;
1923}
1924
0e3da5bb
TT
1925static inline void skb_pop_mac_header(struct sk_buff *skb)
1926{
1927 skb->mac_header = skb->network_header;
1928}
1929
fbbdb8f0
YX
1930static inline void skb_probe_transport_header(struct sk_buff *skb,
1931 const int offset_hint)
1932{
1933 struct flow_keys keys;
1934
1935 if (skb_transport_header_was_set(skb))
1936 return;
1937 else if (skb_flow_dissect(skb, &keys))
1938 skb_set_transport_header(skb, keys.thoff);
1939 else
1940 skb_set_transport_header(skb, offset_hint);
1941}
1942
03606895
ED
1943static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1944{
1945 if (skb_mac_header_was_set(skb)) {
1946 const unsigned char *old_mac = skb_mac_header(skb);
1947
1948 skb_set_mac_header(skb, -skb->mac_len);
1949 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1950 }
1951}
1952
04fb451e
MM
1953static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1954{
1955 return skb->csum_start - skb_headroom(skb);
1956}
1957
2e07fa9c
ACM
1958static inline int skb_transport_offset(const struct sk_buff *skb)
1959{
1960 return skb_transport_header(skb) - skb->data;
1961}
1962
1963static inline u32 skb_network_header_len(const struct sk_buff *skb)
1964{
1965 return skb->transport_header - skb->network_header;
1966}
1967
6a674e9c
JG
1968static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1969{
1970 return skb->inner_transport_header - skb->inner_network_header;
1971}
1972
2e07fa9c
ACM
1973static inline int skb_network_offset(const struct sk_buff *skb)
1974{
1975 return skb_network_header(skb) - skb->data;
1976}
48d49d0c 1977
6a674e9c
JG
1978static inline int skb_inner_network_offset(const struct sk_buff *skb)
1979{
1980 return skb_inner_network_header(skb) - skb->data;
1981}
1982
f9599ce1
CG
1983static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1984{
1985 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1986}
1987
1da177e4
LT
1988/*
1989 * CPUs often take a performance hit when accessing unaligned memory
1990 * locations. The actual performance hit varies, it can be small if the
1991 * hardware handles it or large if we have to take an exception and fix it
1992 * in software.
1993 *
1994 * Since an ethernet header is 14 bytes network drivers often end up with
1995 * the IP header at an unaligned offset. The IP header can be aligned by
1996 * shifting the start of the packet by 2 bytes. Drivers should do this
1997 * with:
1998 *
8660c124 1999 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2000 *
2001 * The downside to this alignment of the IP header is that the DMA is now
2002 * unaligned. On some architectures the cost of an unaligned DMA is high
2003 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2004 *
1da177e4
LT
2005 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2006 * to be overridden.
2007 */
2008#ifndef NET_IP_ALIGN
2009#define NET_IP_ALIGN 2
2010#endif
2011
025be81e
AB
2012/*
2013 * The networking layer reserves some headroom in skb data (via
2014 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2015 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2016 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2017 *
2018 * Unfortunately this headroom changes the DMA alignment of the resulting
2019 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2020 * on some architectures. An architecture can override this value,
2021 * perhaps setting it to a cacheline in size (since that will maintain
2022 * cacheline alignment of the DMA). It must be a power of 2.
2023 *
d6301d3d 2024 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2025 * headroom, you should not reduce this.
5933dd2f
ED
2026 *
2027 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2028 * to reduce average number of cache lines per packet.
2029 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2030 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2031 */
2032#ifndef NET_SKB_PAD
5933dd2f 2033#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2034#endif
2035
7965bd4d 2036int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2037
2038static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2039{
c4264f27 2040 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2041 WARN_ON(1);
2042 return;
2043 }
27a884dc
ACM
2044 skb->len = len;
2045 skb_set_tail_pointer(skb, len);
1da177e4
LT
2046}
2047
7965bd4d 2048void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2049
2050static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2051{
3cc0e873
HX
2052 if (skb->data_len)
2053 return ___pskb_trim(skb, len);
2054 __skb_trim(skb, len);
2055 return 0;
1da177e4
LT
2056}
2057
2058static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2059{
2060 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2061}
2062
e9fa4f7b
HX
2063/**
2064 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2065 * @skb: buffer to alter
2066 * @len: new length
2067 *
2068 * This is identical to pskb_trim except that the caller knows that
2069 * the skb is not cloned so we should never get an error due to out-
2070 * of-memory.
2071 */
2072static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2073{
2074 int err = pskb_trim(skb, len);
2075 BUG_ON(err);
2076}
2077
1da177e4
LT
2078/**
2079 * skb_orphan - orphan a buffer
2080 * @skb: buffer to orphan
2081 *
2082 * If a buffer currently has an owner then we call the owner's
2083 * destructor function and make the @skb unowned. The buffer continues
2084 * to exist but is no longer charged to its former owner.
2085 */
2086static inline void skb_orphan(struct sk_buff *skb)
2087{
c34a7612 2088 if (skb->destructor) {
1da177e4 2089 skb->destructor(skb);
c34a7612
ED
2090 skb->destructor = NULL;
2091 skb->sk = NULL;
376c7311
ED
2092 } else {
2093 BUG_ON(skb->sk);
c34a7612 2094 }
1da177e4
LT
2095}
2096
a353e0ce
MT
2097/**
2098 * skb_orphan_frags - orphan the frags contained in a buffer
2099 * @skb: buffer to orphan frags from
2100 * @gfp_mask: allocation mask for replacement pages
2101 *
2102 * For each frag in the SKB which needs a destructor (i.e. has an
2103 * owner) create a copy of that frag and release the original
2104 * page by calling the destructor.
2105 */
2106static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2107{
2108 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2109 return 0;
2110 return skb_copy_ubufs(skb, gfp_mask);
2111}
2112
1da177e4
LT
2113/**
2114 * __skb_queue_purge - empty a list
2115 * @list: list to empty
2116 *
2117 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2118 * the list and one reference dropped. This function does not take the
2119 * list lock and the caller must hold the relevant locks to use it.
2120 */
7965bd4d 2121void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2122static inline void __skb_queue_purge(struct sk_buff_head *list)
2123{
2124 struct sk_buff *skb;
2125 while ((skb = __skb_dequeue(list)) != NULL)
2126 kfree_skb(skb);
2127}
2128
e5e67305
AD
2129#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2130#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2131#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2132
7965bd4d 2133void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2134
7965bd4d
JP
2135struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2136 gfp_t gfp_mask);
8af27456
CH
2137
2138/**
2139 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2140 * @dev: network device to receive on
2141 * @length: length to allocate
2142 *
2143 * Allocate a new &sk_buff and assign it a usage count of one. The
2144 * buffer has unspecified headroom built in. Users should allocate
2145 * the headroom they think they need without accounting for the
2146 * built in space. The built in space is used for optimisations.
2147 *
2148 * %NULL is returned if there is no free memory. Although this function
2149 * allocates memory it can be called from an interrupt.
2150 */
2151static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2152 unsigned int length)
8af27456
CH
2153{
2154 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2155}
2156
6f532612
ED
2157/* legacy helper around __netdev_alloc_skb() */
2158static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2159 gfp_t gfp_mask)
2160{
2161 return __netdev_alloc_skb(NULL, length, gfp_mask);
2162}
2163
2164/* legacy helper around netdev_alloc_skb() */
2165static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2166{
2167 return netdev_alloc_skb(NULL, length);
2168}
2169
2170
4915a0de
ED
2171static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2172 unsigned int length, gfp_t gfp)
61321bbd 2173{
4915a0de 2174 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2175
2176 if (NET_IP_ALIGN && skb)
2177 skb_reserve(skb, NET_IP_ALIGN);
2178 return skb;
2179}
2180
4915a0de
ED
2181static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2182 unsigned int length)
2183{
2184 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2185}
2186
71dfda58
AD
2187/**
2188 * __dev_alloc_pages - allocate page for network Rx
2189 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2190 * @order: size of the allocation
2191 *
2192 * Allocate a new page.
2193 *
2194 * %NULL is returned if there is no free memory.
2195*/
2196static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2197 unsigned int order)
2198{
2199 /* This piece of code contains several assumptions.
2200 * 1. This is for device Rx, therefor a cold page is preferred.
2201 * 2. The expectation is the user wants a compound page.
2202 * 3. If requesting a order 0 page it will not be compound
2203 * due to the check to see if order has a value in prep_new_page
2204 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2205 * code in gfp_to_alloc_flags that should be enforcing this.
2206 */
2207 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2208
2209 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2210}
2211
2212static inline struct page *dev_alloc_pages(unsigned int order)
2213{
2214 return __dev_alloc_pages(GFP_ATOMIC, order);
2215}
2216
2217/**
2218 * __dev_alloc_page - allocate a page for network Rx
2219 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2220 *
2221 * Allocate a new page.
2222 *
2223 * %NULL is returned if there is no free memory.
2224 */
2225static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2226{
2227 return __dev_alloc_pages(gfp_mask, 0);
2228}
2229
2230static inline struct page *dev_alloc_page(void)
2231{
2232 return __dev_alloc_page(GFP_ATOMIC);
2233}
2234
0614002b
MG
2235/**
2236 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2237 * @page: The page that was allocated from skb_alloc_page
2238 * @skb: The skb that may need pfmemalloc set
2239 */
2240static inline void skb_propagate_pfmemalloc(struct page *page,
2241 struct sk_buff *skb)
2242{
2243 if (page && page->pfmemalloc)
2244 skb->pfmemalloc = true;
2245}
2246
131ea667 2247/**
e227867f 2248 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2249 * @frag: the paged fragment
2250 *
2251 * Returns the &struct page associated with @frag.
2252 */
2253static inline struct page *skb_frag_page(const skb_frag_t *frag)
2254{
a8605c60 2255 return frag->page.p;
131ea667
IC
2256}
2257
2258/**
2259 * __skb_frag_ref - take an addition reference on a paged fragment.
2260 * @frag: the paged fragment
2261 *
2262 * Takes an additional reference on the paged fragment @frag.
2263 */
2264static inline void __skb_frag_ref(skb_frag_t *frag)
2265{
2266 get_page(skb_frag_page(frag));
2267}
2268
2269/**
2270 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2271 * @skb: the buffer
2272 * @f: the fragment offset.
2273 *
2274 * Takes an additional reference on the @f'th paged fragment of @skb.
2275 */
2276static inline void skb_frag_ref(struct sk_buff *skb, int f)
2277{
2278 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2279}
2280
2281/**
2282 * __skb_frag_unref - release a reference on a paged fragment.
2283 * @frag: the paged fragment
2284 *
2285 * Releases a reference on the paged fragment @frag.
2286 */
2287static inline void __skb_frag_unref(skb_frag_t *frag)
2288{
2289 put_page(skb_frag_page(frag));
2290}
2291
2292/**
2293 * skb_frag_unref - release a reference on a paged fragment of an skb.
2294 * @skb: the buffer
2295 * @f: the fragment offset
2296 *
2297 * Releases a reference on the @f'th paged fragment of @skb.
2298 */
2299static inline void skb_frag_unref(struct sk_buff *skb, int f)
2300{
2301 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2302}
2303
2304/**
2305 * skb_frag_address - gets the address of the data contained in a paged fragment
2306 * @frag: the paged fragment buffer
2307 *
2308 * Returns the address of the data within @frag. The page must already
2309 * be mapped.
2310 */
2311static inline void *skb_frag_address(const skb_frag_t *frag)
2312{
2313 return page_address(skb_frag_page(frag)) + frag->page_offset;
2314}
2315
2316/**
2317 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2318 * @frag: the paged fragment buffer
2319 *
2320 * Returns the address of the data within @frag. Checks that the page
2321 * is mapped and returns %NULL otherwise.
2322 */
2323static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2324{
2325 void *ptr = page_address(skb_frag_page(frag));
2326 if (unlikely(!ptr))
2327 return NULL;
2328
2329 return ptr + frag->page_offset;
2330}
2331
2332/**
2333 * __skb_frag_set_page - sets the page contained in a paged fragment
2334 * @frag: the paged fragment
2335 * @page: the page to set
2336 *
2337 * Sets the fragment @frag to contain @page.
2338 */
2339static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2340{
a8605c60 2341 frag->page.p = page;
131ea667
IC
2342}
2343
2344/**
2345 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2346 * @skb: the buffer
2347 * @f: the fragment offset
2348 * @page: the page to set
2349 *
2350 * Sets the @f'th fragment of @skb to contain @page.
2351 */
2352static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2353 struct page *page)
2354{
2355 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2356}
2357
400dfd3a
ED
2358bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2359
131ea667
IC
2360/**
2361 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2362 * @dev: the device to map the fragment to
131ea667
IC
2363 * @frag: the paged fragment to map
2364 * @offset: the offset within the fragment (starting at the
2365 * fragment's own offset)
2366 * @size: the number of bytes to map
f83347df 2367 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2368 *
2369 * Maps the page associated with @frag to @device.
2370 */
2371static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2372 const skb_frag_t *frag,
2373 size_t offset, size_t size,
2374 enum dma_data_direction dir)
2375{
2376 return dma_map_page(dev, skb_frag_page(frag),
2377 frag->page_offset + offset, size, dir);
2378}
2379
117632e6
ED
2380static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2381 gfp_t gfp_mask)
2382{
2383 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2384}
2385
bad93e9d
OP
2386
2387static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2388 gfp_t gfp_mask)
2389{
2390 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2391}
2392
2393
334a8132
PM
2394/**
2395 * skb_clone_writable - is the header of a clone writable
2396 * @skb: buffer to check
2397 * @len: length up to which to write
2398 *
2399 * Returns true if modifying the header part of the cloned buffer
2400 * does not requires the data to be copied.
2401 */
05bdd2f1 2402static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2403{
2404 return !skb_header_cloned(skb) &&
2405 skb_headroom(skb) + len <= skb->hdr_len;
2406}
2407
d9cc2048
HX
2408static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2409 int cloned)
2410{
2411 int delta = 0;
2412
d9cc2048
HX
2413 if (headroom > skb_headroom(skb))
2414 delta = headroom - skb_headroom(skb);
2415
2416 if (delta || cloned)
2417 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2418 GFP_ATOMIC);
2419 return 0;
2420}
2421
1da177e4
LT
2422/**
2423 * skb_cow - copy header of skb when it is required
2424 * @skb: buffer to cow
2425 * @headroom: needed headroom
2426 *
2427 * If the skb passed lacks sufficient headroom or its data part
2428 * is shared, data is reallocated. If reallocation fails, an error
2429 * is returned and original skb is not changed.
2430 *
2431 * The result is skb with writable area skb->head...skb->tail
2432 * and at least @headroom of space at head.
2433 */
2434static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2435{
d9cc2048
HX
2436 return __skb_cow(skb, headroom, skb_cloned(skb));
2437}
1da177e4 2438
d9cc2048
HX
2439/**
2440 * skb_cow_head - skb_cow but only making the head writable
2441 * @skb: buffer to cow
2442 * @headroom: needed headroom
2443 *
2444 * This function is identical to skb_cow except that we replace the
2445 * skb_cloned check by skb_header_cloned. It should be used when
2446 * you only need to push on some header and do not need to modify
2447 * the data.
2448 */
2449static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2450{
2451 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2452}
2453
2454/**
2455 * skb_padto - pad an skbuff up to a minimal size
2456 * @skb: buffer to pad
2457 * @len: minimal length
2458 *
2459 * Pads up a buffer to ensure the trailing bytes exist and are
2460 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2461 * is untouched. Otherwise it is extended. Returns zero on
2462 * success. The skb is freed on error.
1da177e4
LT
2463 */
2464
5b057c6b 2465static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2466{
2467 unsigned int size = skb->len;
2468 if (likely(size >= len))
5b057c6b 2469 return 0;
987c402a 2470 return skb_pad(skb, len - size);
1da177e4
LT
2471}
2472
2473static inline int skb_add_data(struct sk_buff *skb,
2474 char __user *from, int copy)
2475{
2476 const int off = skb->len;
2477
2478 if (skb->ip_summed == CHECKSUM_NONE) {
2479 int err = 0;
5084205f 2480 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2481 copy, 0, &err);
2482 if (!err) {
2483 skb->csum = csum_block_add(skb->csum, csum, off);
2484 return 0;
2485 }
2486 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2487 return 0;
2488
2489 __skb_trim(skb, off);
2490 return -EFAULT;
2491}
2492
38ba0a65
ED
2493static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2494 const struct page *page, int off)
1da177e4
LT
2495{
2496 if (i) {
9e903e08 2497 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2498
ea2ab693 2499 return page == skb_frag_page(frag) &&
9e903e08 2500 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2501 }
38ba0a65 2502 return false;
1da177e4
LT
2503}
2504
364c6bad
HX
2505static inline int __skb_linearize(struct sk_buff *skb)
2506{
2507 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2508}
2509
1da177e4
LT
2510/**
2511 * skb_linearize - convert paged skb to linear one
2512 * @skb: buffer to linarize
1da177e4
LT
2513 *
2514 * If there is no free memory -ENOMEM is returned, otherwise zero
2515 * is returned and the old skb data released.
2516 */
364c6bad
HX
2517static inline int skb_linearize(struct sk_buff *skb)
2518{
2519 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2520}
2521
cef401de
ED
2522/**
2523 * skb_has_shared_frag - can any frag be overwritten
2524 * @skb: buffer to test
2525 *
2526 * Return true if the skb has at least one frag that might be modified
2527 * by an external entity (as in vmsplice()/sendfile())
2528 */
2529static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2530{
c9af6db4
PS
2531 return skb_is_nonlinear(skb) &&
2532 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2533}
2534
364c6bad
HX
2535/**
2536 * skb_linearize_cow - make sure skb is linear and writable
2537 * @skb: buffer to process
2538 *
2539 * If there is no free memory -ENOMEM is returned, otherwise zero
2540 * is returned and the old skb data released.
2541 */
2542static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2543{
364c6bad
HX
2544 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2545 __skb_linearize(skb) : 0;
1da177e4
LT
2546}
2547
2548/**
2549 * skb_postpull_rcsum - update checksum for received skb after pull
2550 * @skb: buffer to update
2551 * @start: start of data before pull
2552 * @len: length of data pulled
2553 *
2554 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2555 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2556 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2557 */
2558
2559static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2560 const void *start, unsigned int len)
1da177e4 2561{
84fa7933 2562 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2563 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2564}
2565
cbb042f9
HX
2566unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2567
7ce5a27f
DM
2568/**
2569 * pskb_trim_rcsum - trim received skb and update checksum
2570 * @skb: buffer to trim
2571 * @len: new length
2572 *
2573 * This is exactly the same as pskb_trim except that it ensures the
2574 * checksum of received packets are still valid after the operation.
2575 */
2576
2577static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2578{
2579 if (likely(len >= skb->len))
2580 return 0;
2581 if (skb->ip_summed == CHECKSUM_COMPLETE)
2582 skb->ip_summed = CHECKSUM_NONE;
2583 return __pskb_trim(skb, len);
2584}
2585
1da177e4
LT
2586#define skb_queue_walk(queue, skb) \
2587 for (skb = (queue)->next; \
a1e4891f 2588 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2589 skb = skb->next)
2590
46f8914e
JC
2591#define skb_queue_walk_safe(queue, skb, tmp) \
2592 for (skb = (queue)->next, tmp = skb->next; \
2593 skb != (struct sk_buff *)(queue); \
2594 skb = tmp, tmp = skb->next)
2595
1164f52a 2596#define skb_queue_walk_from(queue, skb) \
a1e4891f 2597 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2598 skb = skb->next)
2599
2600#define skb_queue_walk_from_safe(queue, skb, tmp) \
2601 for (tmp = skb->next; \
2602 skb != (struct sk_buff *)(queue); \
2603 skb = tmp, tmp = skb->next)
2604
300ce174
SH
2605#define skb_queue_reverse_walk(queue, skb) \
2606 for (skb = (queue)->prev; \
a1e4891f 2607 skb != (struct sk_buff *)(queue); \
300ce174
SH
2608 skb = skb->prev)
2609
686a2955
DM
2610#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2611 for (skb = (queue)->prev, tmp = skb->prev; \
2612 skb != (struct sk_buff *)(queue); \
2613 skb = tmp, tmp = skb->prev)
2614
2615#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2616 for (tmp = skb->prev; \
2617 skb != (struct sk_buff *)(queue); \
2618 skb = tmp, tmp = skb->prev)
1da177e4 2619
21dc3301 2620static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2621{
2622 return skb_shinfo(skb)->frag_list != NULL;
2623}
2624
2625static inline void skb_frag_list_init(struct sk_buff *skb)
2626{
2627 skb_shinfo(skb)->frag_list = NULL;
2628}
2629
2630static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2631{
2632 frag->next = skb_shinfo(skb)->frag_list;
2633 skb_shinfo(skb)->frag_list = frag;
2634}
2635
2636#define skb_walk_frags(skb, iter) \
2637 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2638
7965bd4d
JP
2639struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2640 int *peeked, int *off, int *err);
2641struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2642 int *err);
2643unsigned int datagram_poll(struct file *file, struct socket *sock,
2644 struct poll_table_struct *wait);
2645int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2646 struct iovec *to, int size);
51f3d02b
DM
2647static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2648 struct msghdr *msg, int size)
2649{
2650 return skb_copy_datagram_iovec(from, offset, msg->msg_iov, size);
2651}
7965bd4d
JP
2652int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2653 struct iovec *iov);
2654int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2655 const struct iovec *from, int from_offset,
2656 int len);
2657int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2658 int offset, size_t count);
a8f820aa
HX
2659int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2660 struct iov_iter *to, int size);
7965bd4d
JP
2661void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2662void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2663int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2664int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2665int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2666__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2667 int len, __wsum csum);
2668int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2669 struct pipe_inode_info *pipe, unsigned int len,
2670 unsigned int flags);
2671void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2672unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2673int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2674 int len, int hlen);
7965bd4d
JP
2675void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2676int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2677void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2678unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2679struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2680struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2681int skb_ensure_writable(struct sk_buff *skb, int write_len);
20380731 2682
2817a336
DB
2683struct skb_checksum_ops {
2684 __wsum (*update)(const void *mem, int len, __wsum wsum);
2685 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2686};
2687
2688__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2689 __wsum csum, const struct skb_checksum_ops *ops);
2690__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2691 __wsum csum);
2692
690e36e7
DM
2693static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2694 int len, void *data, int hlen, void *buffer)
1da177e4 2695{
55820ee2 2696 if (hlen - offset >= len)
690e36e7 2697 return data + offset;
1da177e4 2698
690e36e7
DM
2699 if (!skb ||
2700 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2701 return NULL;
2702
2703 return buffer;
2704}
2705
690e36e7
DM
2706static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2707 int len, void *buffer)
2708{
2709 return __skb_header_pointer(skb, offset, len, skb->data,
2710 skb_headlen(skb), buffer);
2711}
2712
4262e5cc
DB
2713/**
2714 * skb_needs_linearize - check if we need to linearize a given skb
2715 * depending on the given device features.
2716 * @skb: socket buffer to check
2717 * @features: net device features
2718 *
2719 * Returns true if either:
2720 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2721 * 2. skb is fragmented and the device does not support SG.
2722 */
2723static inline bool skb_needs_linearize(struct sk_buff *skb,
2724 netdev_features_t features)
2725{
2726 return skb_is_nonlinear(skb) &&
2727 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2728 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2729}
2730
d626f62b
ACM
2731static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2732 void *to,
2733 const unsigned int len)
2734{
2735 memcpy(to, skb->data, len);
2736}
2737
2738static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2739 const int offset, void *to,
2740 const unsigned int len)
2741{
2742 memcpy(to, skb->data + offset, len);
2743}
2744
27d7ff46
ACM
2745static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2746 const void *from,
2747 const unsigned int len)
2748{
2749 memcpy(skb->data, from, len);
2750}
2751
2752static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2753 const int offset,
2754 const void *from,
2755 const unsigned int len)
2756{
2757 memcpy(skb->data + offset, from, len);
2758}
2759
7965bd4d 2760void skb_init(void);
1da177e4 2761
ac45f602
PO
2762static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2763{
2764 return skb->tstamp;
2765}
2766
a61bbcf2
PM
2767/**
2768 * skb_get_timestamp - get timestamp from a skb
2769 * @skb: skb to get stamp from
2770 * @stamp: pointer to struct timeval to store stamp in
2771 *
2772 * Timestamps are stored in the skb as offsets to a base timestamp.
2773 * This function converts the offset back to a struct timeval and stores
2774 * it in stamp.
2775 */
ac45f602
PO
2776static inline void skb_get_timestamp(const struct sk_buff *skb,
2777 struct timeval *stamp)
a61bbcf2 2778{
b7aa0bf7 2779 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2780}
2781
ac45f602
PO
2782static inline void skb_get_timestampns(const struct sk_buff *skb,
2783 struct timespec *stamp)
2784{
2785 *stamp = ktime_to_timespec(skb->tstamp);
2786}
2787
b7aa0bf7 2788static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2789{
b7aa0bf7 2790 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2791}
2792
164891aa
SH
2793static inline ktime_t net_timedelta(ktime_t t)
2794{
2795 return ktime_sub(ktime_get_real(), t);
2796}
2797
b9ce204f
IJ
2798static inline ktime_t net_invalid_timestamp(void)
2799{
2800 return ktime_set(0, 0);
2801}
a61bbcf2 2802
62bccb8c
AD
2803struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2804
c1f19b51
RC
2805#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2806
7965bd4d
JP
2807void skb_clone_tx_timestamp(struct sk_buff *skb);
2808bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2809
2810#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2811
2812static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2813{
2814}
2815
2816static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2817{
2818 return false;
2819}
2820
2821#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2822
2823/**
2824 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2825 *
da92b194
RC
2826 * PHY drivers may accept clones of transmitted packets for
2827 * timestamping via their phy_driver.txtstamp method. These drivers
2828 * must call this function to return the skb back to the stack, with
2829 * or without a timestamp.
2830 *
c1f19b51 2831 * @skb: clone of the the original outgoing packet
da92b194 2832 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2833 *
2834 */
2835void skb_complete_tx_timestamp(struct sk_buff *skb,
2836 struct skb_shared_hwtstamps *hwtstamps);
2837
e7fd2885
WB
2838void __skb_tstamp_tx(struct sk_buff *orig_skb,
2839 struct skb_shared_hwtstamps *hwtstamps,
2840 struct sock *sk, int tstype);
2841
ac45f602
PO
2842/**
2843 * skb_tstamp_tx - queue clone of skb with send time stamps
2844 * @orig_skb: the original outgoing packet
2845 * @hwtstamps: hardware time stamps, may be NULL if not available
2846 *
2847 * If the skb has a socket associated, then this function clones the
2848 * skb (thus sharing the actual data and optional structures), stores
2849 * the optional hardware time stamping information (if non NULL) or
2850 * generates a software time stamp (otherwise), then queues the clone
2851 * to the error queue of the socket. Errors are silently ignored.
2852 */
7965bd4d
JP
2853void skb_tstamp_tx(struct sk_buff *orig_skb,
2854 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2855
4507a715
RC
2856static inline void sw_tx_timestamp(struct sk_buff *skb)
2857{
2244d07b
OH
2858 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2859 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2860 skb_tstamp_tx(skb, NULL);
2861}
2862
2863/**
2864 * skb_tx_timestamp() - Driver hook for transmit timestamping
2865 *
2866 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2867 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2868 *
73409f3b
DM
2869 * Specifically, one should make absolutely sure that this function is
2870 * called before TX completion of this packet can trigger. Otherwise
2871 * the packet could potentially already be freed.
2872 *
4507a715
RC
2873 * @skb: A socket buffer.
2874 */
2875static inline void skb_tx_timestamp(struct sk_buff *skb)
2876{
c1f19b51 2877 skb_clone_tx_timestamp(skb);
4507a715
RC
2878 sw_tx_timestamp(skb);
2879}
2880
6e3e939f
JB
2881/**
2882 * skb_complete_wifi_ack - deliver skb with wifi status
2883 *
2884 * @skb: the original outgoing packet
2885 * @acked: ack status
2886 *
2887 */
2888void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2889
7965bd4d
JP
2890__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2891__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2892
60476372
HX
2893static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2894{
5d0c2b95 2895 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2896}
2897
fb286bb2
HX
2898/**
2899 * skb_checksum_complete - Calculate checksum of an entire packet
2900 * @skb: packet to process
2901 *
2902 * This function calculates the checksum over the entire packet plus
2903 * the value of skb->csum. The latter can be used to supply the
2904 * checksum of a pseudo header as used by TCP/UDP. It returns the
2905 * checksum.
2906 *
2907 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2908 * this function can be used to verify that checksum on received
2909 * packets. In that case the function should return zero if the
2910 * checksum is correct. In particular, this function will return zero
2911 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2912 * hardware has already verified the correctness of the checksum.
2913 */
4381ca3c 2914static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2915{
60476372
HX
2916 return skb_csum_unnecessary(skb) ?
2917 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2918}
2919
77cffe23
TH
2920static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2921{
2922 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2923 if (skb->csum_level == 0)
2924 skb->ip_summed = CHECKSUM_NONE;
2925 else
2926 skb->csum_level--;
2927 }
2928}
2929
2930static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2931{
2932 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2933 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2934 skb->csum_level++;
2935 } else if (skb->ip_summed == CHECKSUM_NONE) {
2936 skb->ip_summed = CHECKSUM_UNNECESSARY;
2937 skb->csum_level = 0;
2938 }
2939}
2940
5a212329
TH
2941static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2942{
2943 /* Mark current checksum as bad (typically called from GRO
2944 * path). In the case that ip_summed is CHECKSUM_NONE
2945 * this must be the first checksum encountered in the packet.
2946 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2947 * checksum after the last one validated. For UDP, a zero
2948 * checksum can not be marked as bad.
2949 */
2950
2951 if (skb->ip_summed == CHECKSUM_NONE ||
2952 skb->ip_summed == CHECKSUM_UNNECESSARY)
2953 skb->csum_bad = 1;
2954}
2955
76ba0aae
TH
2956/* Check if we need to perform checksum complete validation.
2957 *
2958 * Returns true if checksum complete is needed, false otherwise
2959 * (either checksum is unnecessary or zero checksum is allowed).
2960 */
2961static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2962 bool zero_okay,
2963 __sum16 check)
2964{
5d0c2b95
TH
2965 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2966 skb->csum_valid = 1;
77cffe23 2967 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2968 return false;
2969 }
2970
2971 return true;
2972}
2973
2974/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2975 * in checksum_init.
2976 */
2977#define CHECKSUM_BREAK 76
2978
2979/* Validate (init) checksum based on checksum complete.
2980 *
2981 * Return values:
2982 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2983 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2984 * checksum is stored in skb->csum for use in __skb_checksum_complete
2985 * non-zero: value of invalid checksum
2986 *
2987 */
2988static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2989 bool complete,
2990 __wsum psum)
2991{
2992 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2993 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2994 skb->csum_valid = 1;
76ba0aae
TH
2995 return 0;
2996 }
5a212329
TH
2997 } else if (skb->csum_bad) {
2998 /* ip_summed == CHECKSUM_NONE in this case */
2999 return 1;
76ba0aae
TH
3000 }
3001
3002 skb->csum = psum;
3003
5d0c2b95
TH
3004 if (complete || skb->len <= CHECKSUM_BREAK) {
3005 __sum16 csum;
3006
3007 csum = __skb_checksum_complete(skb);
3008 skb->csum_valid = !csum;
3009 return csum;
3010 }
76ba0aae
TH
3011
3012 return 0;
3013}
3014
3015static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3016{
3017 return 0;
3018}
3019
3020/* Perform checksum validate (init). Note that this is a macro since we only
3021 * want to calculate the pseudo header which is an input function if necessary.
3022 * First we try to validate without any computation (checksum unnecessary) and
3023 * then calculate based on checksum complete calling the function to compute
3024 * pseudo header.
3025 *
3026 * Return values:
3027 * 0: checksum is validated or try to in skb_checksum_complete
3028 * non-zero: value of invalid checksum
3029 */
3030#define __skb_checksum_validate(skb, proto, complete, \
3031 zero_okay, check, compute_pseudo) \
3032({ \
3033 __sum16 __ret = 0; \
5d0c2b95 3034 skb->csum_valid = 0; \
76ba0aae
TH
3035 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3036 __ret = __skb_checksum_validate_complete(skb, \
3037 complete, compute_pseudo(skb, proto)); \
3038 __ret; \
3039})
3040
3041#define skb_checksum_init(skb, proto, compute_pseudo) \
3042 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3043
3044#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3045 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3046
3047#define skb_checksum_validate(skb, proto, compute_pseudo) \
3048 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3049
3050#define skb_checksum_validate_zero_check(skb, proto, check, \
3051 compute_pseudo) \
3052 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3053
3054#define skb_checksum_simple_validate(skb) \
3055 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3056
d96535a1
TH
3057static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3058{
3059 return (skb->ip_summed == CHECKSUM_NONE &&
3060 skb->csum_valid && !skb->csum_bad);
3061}
3062
3063static inline void __skb_checksum_convert(struct sk_buff *skb,
3064 __sum16 check, __wsum pseudo)
3065{
3066 skb->csum = ~pseudo;
3067 skb->ip_summed = CHECKSUM_COMPLETE;
3068}
3069
3070#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3071do { \
3072 if (__skb_checksum_convert_check(skb)) \
3073 __skb_checksum_convert(skb, check, \
3074 compute_pseudo(skb, proto)); \
3075} while (0)
3076
5f79e0f9 3077#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3078void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3079static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3080{
3081 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3082 nf_conntrack_destroy(nfct);
1da177e4
LT
3083}
3084static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3085{
3086 if (nfct)
3087 atomic_inc(&nfct->use);
3088}
2fc72c7b 3089#endif
34666d46 3090#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3091static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3092{
3093 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3094 kfree(nf_bridge);
3095}
3096static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3097{
3098 if (nf_bridge)
3099 atomic_inc(&nf_bridge->use);
3100}
3101#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3102static inline void nf_reset(struct sk_buff *skb)
3103{
5f79e0f9 3104#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3105 nf_conntrack_put(skb->nfct);
3106 skb->nfct = NULL;
2fc72c7b 3107#endif
34666d46 3108#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3109 nf_bridge_put(skb->nf_bridge);
3110 skb->nf_bridge = NULL;
3111#endif
3112}
3113
124dff01
PM
3114static inline void nf_reset_trace(struct sk_buff *skb)
3115{
478b360a 3116#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3117 skb->nf_trace = 0;
3118#endif
a193a4ab
PM
3119}
3120
edda553c 3121/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3122static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3123 bool copy)
edda553c 3124{
5f79e0f9 3125#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3126 dst->nfct = src->nfct;
3127 nf_conntrack_get(src->nfct);
b1937227
ED
3128 if (copy)
3129 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3130#endif
34666d46 3131#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3132 dst->nf_bridge = src->nf_bridge;
3133 nf_bridge_get(src->nf_bridge);
3134#endif
478b360a 3135#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3136 if (copy)
3137 dst->nf_trace = src->nf_trace;
478b360a 3138#endif
edda553c
YK
3139}
3140
e7ac05f3
YK
3141static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3142{
e7ac05f3 3143#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3144 nf_conntrack_put(dst->nfct);
2fc72c7b 3145#endif
34666d46 3146#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3147 nf_bridge_put(dst->nf_bridge);
3148#endif
b1937227 3149 __nf_copy(dst, src, true);
e7ac05f3
YK
3150}
3151
984bc16c
JM
3152#ifdef CONFIG_NETWORK_SECMARK
3153static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3154{
3155 to->secmark = from->secmark;
3156}
3157
3158static inline void skb_init_secmark(struct sk_buff *skb)
3159{
3160 skb->secmark = 0;
3161}
3162#else
3163static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3164{ }
3165
3166static inline void skb_init_secmark(struct sk_buff *skb)
3167{ }
3168#endif
3169
574f7194
EB
3170static inline bool skb_irq_freeable(const struct sk_buff *skb)
3171{
3172 return !skb->destructor &&
3173#if IS_ENABLED(CONFIG_XFRM)
3174 !skb->sp &&
3175#endif
3176#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3177 !skb->nfct &&
3178#endif
3179 !skb->_skb_refdst &&
3180 !skb_has_frag_list(skb);
3181}
3182
f25f4e44
PWJ
3183static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3184{
f25f4e44 3185 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3186}
3187
9247744e 3188static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3189{
4e3ab47a 3190 return skb->queue_mapping;
4e3ab47a
PE
3191}
3192
f25f4e44
PWJ
3193static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3194{
f25f4e44 3195 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3196}
3197
d5a9e24a
DM
3198static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3199{
3200 skb->queue_mapping = rx_queue + 1;
3201}
3202
9247744e 3203static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3204{
3205 return skb->queue_mapping - 1;
3206}
3207
9247744e 3208static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3209{
a02cec21 3210 return skb->queue_mapping != 0;
d5a9e24a
DM
3211}
3212
0e001614 3213u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3214 unsigned int num_tx_queues);
9247744e 3215
def8b4fa
AD
3216static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3217{
0b3d8e08 3218#ifdef CONFIG_XFRM
def8b4fa 3219 return skb->sp;
def8b4fa 3220#else
def8b4fa 3221 return NULL;
def8b4fa 3222#endif
0b3d8e08 3223}
def8b4fa 3224
68c33163
PS
3225/* Keeps track of mac header offset relative to skb->head.
3226 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3227 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3228 * tunnel skb it points to outer mac header.
3229 * Keeps track of level of encapsulation of network headers.
3230 */
68c33163 3231struct skb_gso_cb {
3347c960
ED
3232 int mac_offset;
3233 int encap_level;
7e2b10c1 3234 __u16 csum_start;
68c33163
PS
3235};
3236#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3237
3238static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3239{
3240 return (skb_mac_header(inner_skb) - inner_skb->head) -
3241 SKB_GSO_CB(inner_skb)->mac_offset;
3242}
3243
1e2bd517
PS
3244static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3245{
3246 int new_headroom, headroom;
3247 int ret;
3248
3249 headroom = skb_headroom(skb);
3250 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3251 if (ret)
3252 return ret;
3253
3254 new_headroom = skb_headroom(skb);
3255 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3256 return 0;
3257}
3258
7e2b10c1
TH
3259/* Compute the checksum for a gso segment. First compute the checksum value
3260 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3261 * then add in skb->csum (checksum from csum_start to end of packet).
3262 * skb->csum and csum_start are then updated to reflect the checksum of the
3263 * resultant packet starting from the transport header-- the resultant checksum
3264 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3265 * header.
3266 */
3267static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3268{
3269 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3270 skb_transport_offset(skb);
3271 __u16 csum;
3272
3273 csum = csum_fold(csum_partial(skb_transport_header(skb),
3274 plen, skb->csum));
3275 skb->csum = res;
3276 SKB_GSO_CB(skb)->csum_start -= plen;
3277
3278 return csum;
3279}
3280
bdcc0924 3281static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3282{
3283 return skb_shinfo(skb)->gso_size;
3284}
3285
36a8f39e 3286/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3287static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3288{
3289 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3290}
3291
7965bd4d 3292void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3293
3294static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3295{
3296 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3297 * wanted then gso_type will be set. */
05bdd2f1
ED
3298 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3299
b78462eb
AD
3300 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3301 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3302 __skb_warn_lro_forwarding(skb);
3303 return true;
3304 }
3305 return false;
3306}
3307
35fc92a9
HX
3308static inline void skb_forward_csum(struct sk_buff *skb)
3309{
3310 /* Unfortunately we don't support this one. Any brave souls? */
3311 if (skb->ip_summed == CHECKSUM_COMPLETE)
3312 skb->ip_summed = CHECKSUM_NONE;
3313}
3314
bc8acf2c
ED
3315/**
3316 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3317 * @skb: skb to check
3318 *
3319 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3320 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3321 * use this helper, to document places where we make this assertion.
3322 */
05bdd2f1 3323static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3324{
3325#ifdef DEBUG
3326 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3327#endif
3328}
3329
f35d9d8a 3330bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3331
ed1f50c3
PD
3332int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3333
56193d1b
AD
3334u32 skb_get_poff(const struct sk_buff *skb);
3335u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3336 const struct flow_keys *keys, int hlen);
f77668dc 3337
3a7c1ee4
AD
3338/**
3339 * skb_head_is_locked - Determine if the skb->head is locked down
3340 * @skb: skb to check
3341 *
3342 * The head on skbs build around a head frag can be removed if they are
3343 * not cloned. This function returns true if the skb head is locked down
3344 * due to either being allocated via kmalloc, or by being a clone with
3345 * multiple references to the head.
3346 */
3347static inline bool skb_head_is_locked(const struct sk_buff *skb)
3348{
3349 return !skb->head_frag || skb_cloned(skb);
3350}
fe6cc55f
FW
3351
3352/**
3353 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3354 *
3355 * @skb: GSO skb
3356 *
3357 * skb_gso_network_seglen is used to determine the real size of the
3358 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3359 *
3360 * The MAC/L2 header is not accounted for.
3361 */
3362static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3363{
3364 unsigned int hdr_len = skb_transport_header(skb) -
3365 skb_network_header(skb);
3366 return hdr_len + skb_gso_transport_seglen(skb);
3367}
1da177e4
LT
3368#endif /* __KERNEL__ */
3369#endif /* _LINUX_SKBUFF_H */