Merge branch 'timestamping'
[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
LT
22#include <linux/cache.h>
23
60063497 24#include <linux/atomic.h>
1da177e4
LT
25#include <asm/types.h>
26#include <linux/spinlock.h>
1da177e4 27#include <linux/net.h>
3fc7e8a6 28#include <linux/textsearch.h>
1da177e4 29#include <net/checksum.h>
a80958f4 30#include <linux/rcupdate.h>
97fc2f08 31#include <linux/dmaengine.h>
b7aa0bf7 32#include <linux/hrtimer.h>
131ea667 33#include <linux/dma-mapping.h>
c8f44aff 34#include <linux/netdev_features.h>
363ec392 35#include <linux/sched.h>
5203cd28 36#include <net/flow_keys.h>
1da177e4 37
78ea85f1
DB
38/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
50 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
51 * if their checksums are okay. skb->csum is still undefined in this case
52 * though. It is a bad option, but, unfortunately, nowadays most vendors do
53 * this. Apparently with the secret goal to sell you new devices, when you
54 * will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_UNNECESSARY is applicable to following protocols:
57 * TCP: IPv6 and IPv4.
58 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
59 * zero UDP checksum for either IPv4 or IPv6, the networking stack
60 * may perform further validation in this case.
61 * GRE: only if the checksum is present in the header.
62 * SCTP: indicates the CRC in SCTP header has been validated.
63 *
64 * skb->csum_level indicates the number of consecutive checksums found in
65 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
66 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
67 * and a device is able to verify the checksums for UDP (possibly zero),
68 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
69 * two. If the device were only able to verify the UDP checksum and not
70 * GRE, either because it doesn't support GRE checksum of because GRE
71 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
72 * not considered in this case).
78ea85f1
DB
73 *
74 * CHECKSUM_COMPLETE:
75 *
76 * This is the most generic way. The device supplied checksum of the _whole_
77 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
78 * hardware doesn't need to parse L3/L4 headers to implement this.
79 *
80 * Note: Even if device supports only some protocols, but is able to produce
81 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
82 *
83 * CHECKSUM_PARTIAL:
84 *
85 * This is identical to the case for output below. This may occur on a packet
86 * received directly from another Linux OS, e.g., a virtualized Linux kernel
87 * on the same host. The packet can be treated in the same way as
88 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
89 * checksum must be filled in by the OS or the hardware.
90 *
91 * B. Checksumming on output.
92 *
93 * CHECKSUM_NONE:
94 *
95 * The skb was already checksummed by the protocol, or a checksum is not
96 * required.
97 *
98 * CHECKSUM_PARTIAL:
99 *
100 * The device is required to checksum the packet as seen by hard_start_xmit()
101 * from skb->csum_start up to the end, and to record/write the checksum at
102 * offset skb->csum_start + skb->csum_offset.
103 *
104 * The device must show its capabilities in dev->features, set up at device
105 * setup time, e.g. netdev_features.h:
106 *
107 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
108 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
109 * IPv4. Sigh. Vendors like this way for an unknown reason.
110 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
111 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
112 * NETIF_F_... - Well, you get the picture.
113 *
114 * CHECKSUM_UNNECESSARY:
115 *
116 * Normally, the device will do per protocol specific checksumming. Protocol
117 * implementations that do not want the NIC to perform the checksum
118 * calculation should use this flag in their outgoing skbs.
119 *
120 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
121 * offload. Correspondingly, the FCoE protocol driver
122 * stack should use CHECKSUM_UNNECESSARY.
123 *
124 * Any questions? No questions, good. --ANK
125 */
126
60476372 127/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
128#define CHECKSUM_NONE 0
129#define CHECKSUM_UNNECESSARY 1
130#define CHECKSUM_COMPLETE 2
131#define CHECKSUM_PARTIAL 3
1da177e4 132
77cffe23
TH
133/* Maximum value in skb->csum_level */
134#define SKB_MAX_CSUM_LEVEL 3
135
0bec8c88 136#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 137#define SKB_WITH_OVERHEAD(X) \
deea84b0 138 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
139#define SKB_MAX_ORDER(X, ORDER) \
140 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
141#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
142#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
143
87fb4b7b
ED
144/* return minimum truesize of one skb containing X bytes of data */
145#define SKB_TRUESIZE(X) ((X) + \
146 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
147 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
148
1da177e4 149struct net_device;
716ea3a7 150struct scatterlist;
9c55e01c 151struct pipe_inode_info;
1da177e4 152
5f79e0f9 153#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
154struct nf_conntrack {
155 atomic_t use;
1da177e4 156};
5f79e0f9 157#endif
1da177e4
LT
158
159#ifdef CONFIG_BRIDGE_NETFILTER
160struct nf_bridge_info {
bf1ac5ca
ED
161 atomic_t use;
162 unsigned int mask;
163 struct net_device *physindev;
164 struct net_device *physoutdev;
165 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
166};
167#endif
168
1da177e4
LT
169struct sk_buff_head {
170 /* These two members must be first. */
171 struct sk_buff *next;
172 struct sk_buff *prev;
173
174 __u32 qlen;
175 spinlock_t lock;
176};
177
178struct sk_buff;
179
9d4dde52
IC
180/* To allow 64K frame to be packed as single skb without frag_list we
181 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
182 * buffers which do not start on a page boundary.
183 *
184 * Since GRO uses frags we allocate at least 16 regardless of page
185 * size.
a715dea3 186 */
9d4dde52 187#if (65536/PAGE_SIZE + 1) < 16
eec00954 188#define MAX_SKB_FRAGS 16UL
a715dea3 189#else
9d4dde52 190#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 191#endif
1da177e4
LT
192
193typedef struct skb_frag_struct skb_frag_t;
194
195struct skb_frag_struct {
a8605c60
IC
196 struct {
197 struct page *p;
198 } page;
cb4dfe56 199#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
200 __u32 page_offset;
201 __u32 size;
cb4dfe56
ED
202#else
203 __u16 page_offset;
204 __u16 size;
205#endif
1da177e4
LT
206};
207
9e903e08
ED
208static inline unsigned int skb_frag_size(const skb_frag_t *frag)
209{
210 return frag->size;
211}
212
213static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
214{
215 frag->size = size;
216}
217
218static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
219{
220 frag->size += delta;
221}
222
223static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
224{
225 frag->size -= delta;
226}
227
ac45f602
PO
228#define HAVE_HW_TIME_STAMP
229
230/**
d3a21be8 231 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
232 * @hwtstamp: hardware time stamp transformed into duration
233 * since arbitrary point in time
ac45f602
PO
234 *
235 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 236 * skb->tstamp.
ac45f602
PO
237 *
238 * hwtstamps can only be compared against other hwtstamps from
239 * the same device.
240 *
241 * This structure is attached to packets as part of the
242 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
243 */
244struct skb_shared_hwtstamps {
245 ktime_t hwtstamp;
ac45f602
PO
246};
247
2244d07b
OH
248/* Definitions for tx_flags in struct skb_shared_info */
249enum {
250 /* generate hardware time stamp */
251 SKBTX_HW_TSTAMP = 1 << 0,
252
e7fd2885 253 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
254 SKBTX_SW_TSTAMP = 1 << 1,
255
256 /* device driver is going to provide hardware time stamp */
257 SKBTX_IN_PROGRESS = 1 << 2,
258
a6686f2f 259 /* device driver supports TX zero-copy buffers */
62b1a8ab 260 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
261
262 /* generate wifi status information (where possible) */
62b1a8ab 263 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
264
265 /* This indicates at least one fragment might be overwritten
266 * (as in vmsplice(), sendfile() ...)
267 * If we need to compute a TX checksum, we'll need to copy
268 * all frags to avoid possible bad checksum
269 */
270 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
271
272 /* generate software time stamp when entering packet scheduling */
273 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
274
275 /* generate software timestamp on peer data acknowledgment */
276 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
277};
278
e1c8a607
WB
279#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
280 SKBTX_SCHED_TSTAMP | \
281 SKBTX_ACK_TSTAMP)
f24b9be5
WB
282#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
283
a6686f2f
SM
284/*
285 * The callback notifies userspace to release buffers when skb DMA is done in
286 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
287 * The zerocopy_success argument is true if zero copy transmit occurred,
288 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
289 * The ctx field is used to track device context.
290 * The desc field is used to track userspace buffer index.
a6686f2f
SM
291 */
292struct ubuf_info {
e19d6763 293 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 294 void *ctx;
a6686f2f 295 unsigned long desc;
ac45f602
PO
296};
297
1da177e4
LT
298/* This data is invariant across clones and lives at
299 * the end of the header data, ie. at skb->end.
300 */
301struct skb_shared_info {
9f42f126
IC
302 unsigned char nr_frags;
303 __u8 tx_flags;
7967168c
HX
304 unsigned short gso_size;
305 /* Warning: this field is not always filled in (UFO)! */
306 unsigned short gso_segs;
307 unsigned short gso_type;
1da177e4 308 struct sk_buff *frag_list;
ac45f602 309 struct skb_shared_hwtstamps hwtstamps;
09c2d251 310 u32 tskey;
9f42f126 311 __be32 ip6_frag_id;
ec7d2f2c
ED
312
313 /*
314 * Warning : all fields before dataref are cleared in __alloc_skb()
315 */
316 atomic_t dataref;
317
69e3c75f
JB
318 /* Intermediate layers must ensure that destructor_arg
319 * remains valid until skb destructor */
320 void * destructor_arg;
a6686f2f 321
fed66381
ED
322 /* must be last field, see pskb_expand_head() */
323 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
324};
325
326/* We divide dataref into two halves. The higher 16 bits hold references
327 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
328 * the entire skb->data. A clone of a headerless skb holds the length of
329 * the header in skb->hdr_len.
1da177e4
LT
330 *
331 * All users must obey the rule that the skb->data reference count must be
332 * greater than or equal to the payload reference count.
333 *
334 * Holding a reference to the payload part means that the user does not
335 * care about modifications to the header part of skb->data.
336 */
337#define SKB_DATAREF_SHIFT 16
338#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
339
d179cd12
DM
340
341enum {
342 SKB_FCLONE_UNAVAILABLE,
343 SKB_FCLONE_ORIG,
344 SKB_FCLONE_CLONE,
345};
346
7967168c
HX
347enum {
348 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 349 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
350
351 /* This indicates the skb is from an untrusted source. */
352 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
353
354 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
355 SKB_GSO_TCP_ECN = 1 << 3,
356
357 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
358
359 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
360
361 SKB_GSO_GRE = 1 << 6,
73136267 362
4b28252c 363 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 364
4b28252c 365 SKB_GSO_IPIP = 1 << 8,
cb32f511 366
4b28252c 367 SKB_GSO_SIT = 1 << 9,
61c1db7f 368
4b28252c 369 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
370
371 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 372
4b28252c
TH
373 SKB_GSO_MPLS = 1 << 12,
374
7967168c
HX
375};
376
2e07fa9c
ACM
377#if BITS_PER_LONG > 32
378#define NET_SKBUFF_DATA_USES_OFFSET 1
379#endif
380
381#ifdef NET_SKBUFF_DATA_USES_OFFSET
382typedef unsigned int sk_buff_data_t;
383#else
384typedef unsigned char *sk_buff_data_t;
385#endif
386
363ec392
ED
387/**
388 * struct skb_mstamp - multi resolution time stamps
389 * @stamp_us: timestamp in us resolution
390 * @stamp_jiffies: timestamp in jiffies
391 */
392struct skb_mstamp {
393 union {
394 u64 v64;
395 struct {
396 u32 stamp_us;
397 u32 stamp_jiffies;
398 };
399 };
400};
401
402/**
403 * skb_mstamp_get - get current timestamp
404 * @cl: place to store timestamps
405 */
406static inline void skb_mstamp_get(struct skb_mstamp *cl)
407{
408 u64 val = local_clock();
409
410 do_div(val, NSEC_PER_USEC);
411 cl->stamp_us = (u32)val;
412 cl->stamp_jiffies = (u32)jiffies;
413}
414
415/**
416 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
417 * @t1: pointer to newest sample
418 * @t0: pointer to oldest sample
419 */
420static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
421 const struct skb_mstamp *t0)
422{
423 s32 delta_us = t1->stamp_us - t0->stamp_us;
424 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
425
426 /* If delta_us is negative, this might be because interval is too big,
427 * or local_clock() drift is too big : fallback using jiffies.
428 */
429 if (delta_us <= 0 ||
430 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
431
432 delta_us = jiffies_to_usecs(delta_jiffies);
433
434 return delta_us;
435}
436
437
1da177e4
LT
438/**
439 * struct sk_buff - socket buffer
440 * @next: Next buffer in list
441 * @prev: Previous buffer in list
363ec392 442 * @tstamp: Time we arrived/left
d84e0bd7 443 * @sk: Socket we are owned by
1da177e4 444 * @dev: Device we arrived on/are leaving by
d84e0bd7 445 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 446 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 447 * @sp: the security path, used for xfrm
1da177e4
LT
448 * @len: Length of actual data
449 * @data_len: Data length
450 * @mac_len: Length of link layer header
334a8132 451 * @hdr_len: writable header length of cloned skb
663ead3b
HX
452 * @csum: Checksum (must include start/offset pair)
453 * @csum_start: Offset from skb->head where checksumming should start
454 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 455 * @priority: Packet queueing priority
60ff7467 456 * @ignore_df: allow local fragmentation
1da177e4 457 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 458 * @ip_summed: Driver fed us an IP checksum
1da177e4 459 * @nohdr: Payload reference only, must not modify header
d84e0bd7 460 * @nfctinfo: Relationship of this skb to the connection
1da177e4 461 * @pkt_type: Packet class
c83c2486 462 * @fclone: skbuff clone status
c83c2486 463 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
464 * @peeked: this packet has been seen already, so stats have been
465 * done for it, don't do them again
ba9dda3a 466 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
467 * @protocol: Packet protocol from driver
468 * @destructor: Destruct function
469 * @nfct: Associated connection, if any
1da177e4 470 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 471 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
472 * @tc_index: Traffic control index
473 * @tc_verd: traffic control verdict
61b905da 474 * @hash: the packet hash
d84e0bd7 475 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 476 * @xmit_more: More SKBs are pending for this queue
553a5672 477 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 478 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 479 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 480 * ports.
a3b18ddb 481 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
482 * @wifi_acked_valid: wifi_acked was set
483 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 484 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
485 * @dma_cookie: a cookie to one of several possible DMA operations
486 * done by skb DMA functions
06021292 487 * @napi_id: id of the NAPI struct this skb came from
984bc16c 488 * @secmark: security marking
d84e0bd7
DB
489 * @mark: Generic packet mark
490 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 491 * @vlan_proto: vlan encapsulation protocol
6aa895b0 492 * @vlan_tci: vlan tag control information
0d89d203 493 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
494 * @inner_transport_header: Inner transport layer header (encapsulation)
495 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 496 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
497 * @transport_header: Transport layer header
498 * @network_header: Network layer header
499 * @mac_header: Link layer header
500 * @tail: Tail pointer
501 * @end: End pointer
502 * @head: Head of buffer
503 * @data: Data head pointer
504 * @truesize: Buffer size
505 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
506 */
507
508struct sk_buff {
509 /* These two members must be first. */
510 struct sk_buff *next;
511 struct sk_buff *prev;
512
363ec392
ED
513 union {
514 ktime_t tstamp;
515 struct skb_mstamp skb_mstamp;
516 };
da3f5cf1
FF
517
518 struct sock *sk;
1da177e4 519 struct net_device *dev;
1da177e4 520
1da177e4
LT
521 /*
522 * This is the control buffer. It is free to use for every
523 * layer. Please put your private variables there. If you
524 * want to keep them across layers you have to do a skb_clone()
525 * first. This is owned by whoever has the skb queued ATM.
526 */
da3f5cf1 527 char cb[48] __aligned(8);
1da177e4 528
7fee226a 529 unsigned long _skb_refdst;
da3f5cf1
FF
530#ifdef CONFIG_XFRM
531 struct sec_path *sp;
532#endif
1da177e4 533 unsigned int len,
334a8132
PM
534 data_len;
535 __u16 mac_len,
536 hdr_len;
ff1dcadb
AV
537 union {
538 __wsum csum;
663ead3b
HX
539 struct {
540 __u16 csum_start;
541 __u16 csum_offset;
542 };
ff1dcadb 543 };
1da177e4 544 __u32 priority;
fe55f6d5 545 kmemcheck_bitfield_begin(flags1);
60ff7467 546 __u8 ignore_df:1,
1cbb3380
TG
547 cloned:1,
548 ip_summed:2,
6869c4d8
HW
549 nohdr:1,
550 nfctinfo:3;
d179cd12 551 __u8 pkt_type:3,
b84f4cc9 552 fclone:2,
ba9dda3a 553 ipvs_property:1,
a59322be 554 peeked:1,
ba9dda3a 555 nf_trace:1;
fe55f6d5 556 kmemcheck_bitfield_end(flags1);
4ab408de 557 __be16 protocol;
1da177e4
LT
558
559 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 560#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 561 struct nf_conntrack *nfct;
2fc72c7b 562#endif
1da177e4
LT
563#ifdef CONFIG_BRIDGE_NETFILTER
564 struct nf_bridge_info *nf_bridge;
565#endif
f25f4e44 566
8964be4a 567 int skb_iif;
4031ae6e 568
61b905da 569 __u32 hash;
4031ae6e 570
86a9bad3 571 __be16 vlan_proto;
4031ae6e
AD
572 __u16 vlan_tci;
573
1da177e4 574#ifdef CONFIG_NET_SCHED
b6b99eb5 575 __u16 tc_index; /* traffic control index */
1da177e4 576#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 577 __u16 tc_verd; /* traffic control verdict */
1da177e4 578#endif
1da177e4 579#endif
fe55f6d5 580
0a14842f 581 __u16 queue_mapping;
fe55f6d5 582 kmemcheck_bitfield_begin(flags2);
0b725a2c 583 __u8 xmit_more:1;
de357cc0 584#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 585 __u8 ndisc_nodetype:2;
d0f09804 586#endif
c93bdd0e 587 __u8 pfmemalloc:1;
3853b584 588 __u8 ooo_okay:1;
61b905da 589 __u8 l4_hash:1;
a3b18ddb 590 __u8 sw_hash:1;
6e3e939f
JB
591 __u8 wifi_acked_valid:1;
592 __u8 wifi_acked:1;
3bdc0eba 593 __u8 no_fcs:1;
d3836f21 594 __u8 head_frag:1;
77cffe23 595 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 596 __u8 encapsulation:1;
7e2b10c1 597 __u8 encap_hdr_csum:1;
5d0c2b95 598 __u8 csum_valid:1;
7e3cead5 599 __u8 csum_complete_sw:1;
5c21403d 600 /* 1/3 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
601 kmemcheck_bitfield_end(flags2);
602
e0d1095a 603#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
604 union {
605 unsigned int napi_id;
606 dma_cookie_t dma_cookie;
607 };
97fc2f08 608#endif
984bc16c
JM
609#ifdef CONFIG_NETWORK_SECMARK
610 __u32 secmark;
611#endif
3b885787
NH
612 union {
613 __u32 mark;
614 __u32 dropcount;
16fad69c 615 __u32 reserved_tailroom;
3b885787 616 };
1da177e4 617
de20fe8e 618 kmemcheck_bitfield_begin(flags3);
77cffe23 619 __u8 csum_level:2;
5a212329
TH
620 __u8 csum_bad:1;
621 /* 13 bit hole */
de20fe8e
TH
622 kmemcheck_bitfield_end(flags3);
623
0d89d203 624 __be16 inner_protocol;
1a37e412
SH
625 __u16 inner_transport_header;
626 __u16 inner_network_header;
627 __u16 inner_mac_header;
628 __u16 transport_header;
629 __u16 network_header;
630 __u16 mac_header;
1da177e4 631 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 632 sk_buff_data_t tail;
4305b541 633 sk_buff_data_t end;
1da177e4 634 unsigned char *head,
4305b541 635 *data;
27a884dc
ACM
636 unsigned int truesize;
637 atomic_t users;
1da177e4
LT
638};
639
640#ifdef __KERNEL__
641/*
642 * Handling routines are only of interest to the kernel
643 */
644#include <linux/slab.h>
645
1da177e4 646
c93bdd0e
MG
647#define SKB_ALLOC_FCLONE 0x01
648#define SKB_ALLOC_RX 0x02
649
650/* Returns true if the skb was allocated from PFMEMALLOC reserves */
651static inline bool skb_pfmemalloc(const struct sk_buff *skb)
652{
653 return unlikely(skb->pfmemalloc);
654}
655
7fee226a
ED
656/*
657 * skb might have a dst pointer attached, refcounted or not.
658 * _skb_refdst low order bit is set if refcount was _not_ taken
659 */
660#define SKB_DST_NOREF 1UL
661#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
662
663/**
664 * skb_dst - returns skb dst_entry
665 * @skb: buffer
666 *
667 * Returns skb dst_entry, regardless of reference taken or not.
668 */
adf30907
ED
669static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
670{
7fee226a
ED
671 /* If refdst was not refcounted, check we still are in a
672 * rcu_read_lock section
673 */
674 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
675 !rcu_read_lock_held() &&
676 !rcu_read_lock_bh_held());
677 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
678}
679
7fee226a
ED
680/**
681 * skb_dst_set - sets skb dst
682 * @skb: buffer
683 * @dst: dst entry
684 *
685 * Sets skb dst, assuming a reference was taken on dst and should
686 * be released by skb_dst_drop()
687 */
adf30907
ED
688static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
689{
7fee226a
ED
690 skb->_skb_refdst = (unsigned long)dst;
691}
692
7965bd4d
JP
693void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
694 bool force);
932bc4d7
JA
695
696/**
697 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
698 * @skb: buffer
699 * @dst: dst entry
700 *
701 * Sets skb dst, assuming a reference was not taken on dst.
702 * If dst entry is cached, we do not take reference and dst_release
703 * will be avoided by refdst_drop. If dst entry is not cached, we take
704 * reference, so that last dst_release can destroy the dst immediately.
705 */
706static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
707{
708 __skb_dst_set_noref(skb, dst, false);
709}
710
711/**
712 * skb_dst_set_noref_force - sets skb dst, without taking reference
713 * @skb: buffer
714 * @dst: dst entry
715 *
716 * Sets skb dst, assuming a reference was not taken on dst.
717 * No reference is taken and no dst_release will be called. While for
718 * cached dsts deferred reclaim is a basic feature, for entries that are
719 * not cached it is caller's job to guarantee that last dst_release for
720 * provided dst happens when nobody uses it, eg. after a RCU grace period.
721 */
722static inline void skb_dst_set_noref_force(struct sk_buff *skb,
723 struct dst_entry *dst)
724{
725 __skb_dst_set_noref(skb, dst, true);
726}
7fee226a
ED
727
728/**
25985edc 729 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
730 * @skb: buffer
731 */
732static inline bool skb_dst_is_noref(const struct sk_buff *skb)
733{
734 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
735}
736
511c3f92
ED
737static inline struct rtable *skb_rtable(const struct sk_buff *skb)
738{
adf30907 739 return (struct rtable *)skb_dst(skb);
511c3f92
ED
740}
741
7965bd4d
JP
742void kfree_skb(struct sk_buff *skb);
743void kfree_skb_list(struct sk_buff *segs);
744void skb_tx_error(struct sk_buff *skb);
745void consume_skb(struct sk_buff *skb);
746void __kfree_skb(struct sk_buff *skb);
d7e8883c 747extern struct kmem_cache *skbuff_head_cache;
bad43ca8 748
7965bd4d
JP
749void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
750bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
751 bool *fragstolen, int *delta_truesize);
bad43ca8 752
7965bd4d
JP
753struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
754 int node);
755struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 756static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 757 gfp_t priority)
d179cd12 758{
564824b0 759 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
760}
761
762static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 763 gfp_t priority)
d179cd12 764{
c93bdd0e 765 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
766}
767
7965bd4d 768struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
769static inline struct sk_buff *alloc_skb_head(gfp_t priority)
770{
771 return __alloc_skb_head(priority, -1);
772}
773
7965bd4d
JP
774struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
775int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
776struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
777struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
778struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
779 gfp_t gfp_mask, bool fclone);
780static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
781 gfp_t gfp_mask)
782{
783 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
784}
7965bd4d
JP
785
786int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
787struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
788 unsigned int headroom);
789struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
790 int newtailroom, gfp_t priority);
25a91d8d
FD
791int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
792 int offset, int len);
7965bd4d
JP
793int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
794 int len);
795int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
796int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 797#define dev_kfree_skb(a) consume_skb(a)
1da177e4 798
7965bd4d
JP
799int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
800 int getfrag(void *from, char *to, int offset,
801 int len, int odd, struct sk_buff *skb),
802 void *from, int length);
e89e9cf5 803
d94d9fee 804struct skb_seq_state {
677e90ed
TG
805 __u32 lower_offset;
806 __u32 upper_offset;
807 __u32 frag_idx;
808 __u32 stepped_offset;
809 struct sk_buff *root_skb;
810 struct sk_buff *cur_skb;
811 __u8 *frag_data;
812};
813
7965bd4d
JP
814void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
815 unsigned int to, struct skb_seq_state *st);
816unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
817 struct skb_seq_state *st);
818void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 819
7965bd4d
JP
820unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
821 unsigned int to, struct ts_config *config,
822 struct ts_state *state);
3fc7e8a6 823
09323cc4
TH
824/*
825 * Packet hash types specify the type of hash in skb_set_hash.
826 *
827 * Hash types refer to the protocol layer addresses which are used to
828 * construct a packet's hash. The hashes are used to differentiate or identify
829 * flows of the protocol layer for the hash type. Hash types are either
830 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
831 *
832 * Properties of hashes:
833 *
834 * 1) Two packets in different flows have different hash values
835 * 2) Two packets in the same flow should have the same hash value
836 *
837 * A hash at a higher layer is considered to be more specific. A driver should
838 * set the most specific hash possible.
839 *
840 * A driver cannot indicate a more specific hash than the layer at which a hash
841 * was computed. For instance an L3 hash cannot be set as an L4 hash.
842 *
843 * A driver may indicate a hash level which is less specific than the
844 * actual layer the hash was computed on. For instance, a hash computed
845 * at L4 may be considered an L3 hash. This should only be done if the
846 * driver can't unambiguously determine that the HW computed the hash at
847 * the higher layer. Note that the "should" in the second property above
848 * permits this.
849 */
850enum pkt_hash_types {
851 PKT_HASH_TYPE_NONE, /* Undefined type */
852 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
853 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
854 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
855};
856
857static inline void
858skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
859{
61b905da 860 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 861 skb->sw_hash = 0;
61b905da 862 skb->hash = hash;
09323cc4
TH
863}
864
3958afa1
TH
865void __skb_get_hash(struct sk_buff *skb);
866static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 867{
a3b18ddb 868 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 869 __skb_get_hash(skb);
bfb564e7 870
61b905da 871 return skb->hash;
bfb564e7
KK
872}
873
57bdf7f4
TH
874static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
875{
61b905da 876 return skb->hash;
57bdf7f4
TH
877}
878
7539fadc
TH
879static inline void skb_clear_hash(struct sk_buff *skb)
880{
61b905da 881 skb->hash = 0;
a3b18ddb 882 skb->sw_hash = 0;
61b905da 883 skb->l4_hash = 0;
7539fadc
TH
884}
885
886static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
887{
61b905da 888 if (!skb->l4_hash)
7539fadc
TH
889 skb_clear_hash(skb);
890}
891
3df7a74e
TH
892static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
893{
61b905da 894 to->hash = from->hash;
a3b18ddb 895 to->sw_hash = from->sw_hash;
61b905da 896 to->l4_hash = from->l4_hash;
3df7a74e
TH
897};
898
4305b541
ACM
899#ifdef NET_SKBUFF_DATA_USES_OFFSET
900static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
901{
902 return skb->head + skb->end;
903}
ec47ea82
AD
904
905static inline unsigned int skb_end_offset(const struct sk_buff *skb)
906{
907 return skb->end;
908}
4305b541
ACM
909#else
910static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
911{
912 return skb->end;
913}
ec47ea82
AD
914
915static inline unsigned int skb_end_offset(const struct sk_buff *skb)
916{
917 return skb->end - skb->head;
918}
4305b541
ACM
919#endif
920
1da177e4 921/* Internal */
4305b541 922#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 923
ac45f602
PO
924static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
925{
926 return &skb_shinfo(skb)->hwtstamps;
927}
928
1da177e4
LT
929/**
930 * skb_queue_empty - check if a queue is empty
931 * @list: queue head
932 *
933 * Returns true if the queue is empty, false otherwise.
934 */
935static inline int skb_queue_empty(const struct sk_buff_head *list)
936{
fd44b93c 937 return list->next == (const struct sk_buff *) list;
1da177e4
LT
938}
939
fc7ebb21
DM
940/**
941 * skb_queue_is_last - check if skb is the last entry in the queue
942 * @list: queue head
943 * @skb: buffer
944 *
945 * Returns true if @skb is the last buffer on the list.
946 */
947static inline bool skb_queue_is_last(const struct sk_buff_head *list,
948 const struct sk_buff *skb)
949{
fd44b93c 950 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
951}
952
832d11c5
IJ
953/**
954 * skb_queue_is_first - check if skb is the first entry in the queue
955 * @list: queue head
956 * @skb: buffer
957 *
958 * Returns true if @skb is the first buffer on the list.
959 */
960static inline bool skb_queue_is_first(const struct sk_buff_head *list,
961 const struct sk_buff *skb)
962{
fd44b93c 963 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
964}
965
249c8b42
DM
966/**
967 * skb_queue_next - return the next packet in the queue
968 * @list: queue head
969 * @skb: current buffer
970 *
971 * Return the next packet in @list after @skb. It is only valid to
972 * call this if skb_queue_is_last() evaluates to false.
973 */
974static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
975 const struct sk_buff *skb)
976{
977 /* This BUG_ON may seem severe, but if we just return then we
978 * are going to dereference garbage.
979 */
980 BUG_ON(skb_queue_is_last(list, skb));
981 return skb->next;
982}
983
832d11c5
IJ
984/**
985 * skb_queue_prev - return the prev packet in the queue
986 * @list: queue head
987 * @skb: current buffer
988 *
989 * Return the prev packet in @list before @skb. It is only valid to
990 * call this if skb_queue_is_first() evaluates to false.
991 */
992static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
993 const struct sk_buff *skb)
994{
995 /* This BUG_ON may seem severe, but if we just return then we
996 * are going to dereference garbage.
997 */
998 BUG_ON(skb_queue_is_first(list, skb));
999 return skb->prev;
1000}
1001
1da177e4
LT
1002/**
1003 * skb_get - reference buffer
1004 * @skb: buffer to reference
1005 *
1006 * Makes another reference to a socket buffer and returns a pointer
1007 * to the buffer.
1008 */
1009static inline struct sk_buff *skb_get(struct sk_buff *skb)
1010{
1011 atomic_inc(&skb->users);
1012 return skb;
1013}
1014
1015/*
1016 * If users == 1, we are the only owner and are can avoid redundant
1017 * atomic change.
1018 */
1019
1da177e4
LT
1020/**
1021 * skb_cloned - is the buffer a clone
1022 * @skb: buffer to check
1023 *
1024 * Returns true if the buffer was generated with skb_clone() and is
1025 * one of multiple shared copies of the buffer. Cloned buffers are
1026 * shared data so must not be written to under normal circumstances.
1027 */
1028static inline int skb_cloned(const struct sk_buff *skb)
1029{
1030 return skb->cloned &&
1031 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1032}
1033
14bbd6a5
PS
1034static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1035{
1036 might_sleep_if(pri & __GFP_WAIT);
1037
1038 if (skb_cloned(skb))
1039 return pskb_expand_head(skb, 0, 0, pri);
1040
1041 return 0;
1042}
1043
1da177e4
LT
1044/**
1045 * skb_header_cloned - is the header a clone
1046 * @skb: buffer to check
1047 *
1048 * Returns true if modifying the header part of the buffer requires
1049 * the data to be copied.
1050 */
1051static inline int skb_header_cloned(const struct sk_buff *skb)
1052{
1053 int dataref;
1054
1055 if (!skb->cloned)
1056 return 0;
1057
1058 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1059 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1060 return dataref != 1;
1061}
1062
1063/**
1064 * skb_header_release - release reference to header
1065 * @skb: buffer to operate on
1066 *
1067 * Drop a reference to the header part of the buffer. This is done
1068 * by acquiring a payload reference. You must not read from the header
1069 * part of skb->data after this.
1070 */
1071static inline void skb_header_release(struct sk_buff *skb)
1072{
1073 BUG_ON(skb->nohdr);
1074 skb->nohdr = 1;
1075 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1076}
1077
1078/**
1079 * skb_shared - is the buffer shared
1080 * @skb: buffer to check
1081 *
1082 * Returns true if more than one person has a reference to this
1083 * buffer.
1084 */
1085static inline int skb_shared(const struct sk_buff *skb)
1086{
1087 return atomic_read(&skb->users) != 1;
1088}
1089
1090/**
1091 * skb_share_check - check if buffer is shared and if so clone it
1092 * @skb: buffer to check
1093 * @pri: priority for memory allocation
1094 *
1095 * If the buffer is shared the buffer is cloned and the old copy
1096 * drops a reference. A new clone with a single reference is returned.
1097 * If the buffer is not shared the original buffer is returned. When
1098 * being called from interrupt status or with spinlocks held pri must
1099 * be GFP_ATOMIC.
1100 *
1101 * NULL is returned on a memory allocation failure.
1102 */
47061bc4 1103static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1104{
1105 might_sleep_if(pri & __GFP_WAIT);
1106 if (skb_shared(skb)) {
1107 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1108
1109 if (likely(nskb))
1110 consume_skb(skb);
1111 else
1112 kfree_skb(skb);
1da177e4
LT
1113 skb = nskb;
1114 }
1115 return skb;
1116}
1117
1118/*
1119 * Copy shared buffers into a new sk_buff. We effectively do COW on
1120 * packets to handle cases where we have a local reader and forward
1121 * and a couple of other messy ones. The normal one is tcpdumping
1122 * a packet thats being forwarded.
1123 */
1124
1125/**
1126 * skb_unshare - make a copy of a shared buffer
1127 * @skb: buffer to check
1128 * @pri: priority for memory allocation
1129 *
1130 * If the socket buffer is a clone then this function creates a new
1131 * copy of the data, drops a reference count on the old copy and returns
1132 * the new copy with the reference count at 1. If the buffer is not a clone
1133 * the original buffer is returned. When called with a spinlock held or
1134 * from interrupt state @pri must be %GFP_ATOMIC
1135 *
1136 * %NULL is returned on a memory allocation failure.
1137 */
e2bf521d 1138static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1139 gfp_t pri)
1da177e4
LT
1140{
1141 might_sleep_if(pri & __GFP_WAIT);
1142 if (skb_cloned(skb)) {
1143 struct sk_buff *nskb = skb_copy(skb, pri);
1144 kfree_skb(skb); /* Free our shared copy */
1145 skb = nskb;
1146 }
1147 return skb;
1148}
1149
1150/**
1a5778aa 1151 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1152 * @list_: list to peek at
1153 *
1154 * Peek an &sk_buff. Unlike most other operations you _MUST_
1155 * be careful with this one. A peek leaves the buffer on the
1156 * list and someone else may run off with it. You must hold
1157 * the appropriate locks or have a private queue to do this.
1158 *
1159 * Returns %NULL for an empty list or a pointer to the head element.
1160 * The reference count is not incremented and the reference is therefore
1161 * volatile. Use with caution.
1162 */
05bdd2f1 1163static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1164{
18d07000
ED
1165 struct sk_buff *skb = list_->next;
1166
1167 if (skb == (struct sk_buff *)list_)
1168 skb = NULL;
1169 return skb;
1da177e4
LT
1170}
1171
da5ef6e5
PE
1172/**
1173 * skb_peek_next - peek skb following the given one from a queue
1174 * @skb: skb to start from
1175 * @list_: list to peek at
1176 *
1177 * Returns %NULL when the end of the list is met or a pointer to the
1178 * next element. The reference count is not incremented and the
1179 * reference is therefore volatile. Use with caution.
1180 */
1181static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1182 const struct sk_buff_head *list_)
1183{
1184 struct sk_buff *next = skb->next;
18d07000 1185
da5ef6e5
PE
1186 if (next == (struct sk_buff *)list_)
1187 next = NULL;
1188 return next;
1189}
1190
1da177e4 1191/**
1a5778aa 1192 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1193 * @list_: list to peek at
1194 *
1195 * Peek an &sk_buff. Unlike most other operations you _MUST_
1196 * be careful with this one. A peek leaves the buffer on the
1197 * list and someone else may run off with it. You must hold
1198 * the appropriate locks or have a private queue to do this.
1199 *
1200 * Returns %NULL for an empty list or a pointer to the tail element.
1201 * The reference count is not incremented and the reference is therefore
1202 * volatile. Use with caution.
1203 */
05bdd2f1 1204static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1205{
18d07000
ED
1206 struct sk_buff *skb = list_->prev;
1207
1208 if (skb == (struct sk_buff *)list_)
1209 skb = NULL;
1210 return skb;
1211
1da177e4
LT
1212}
1213
1214/**
1215 * skb_queue_len - get queue length
1216 * @list_: list to measure
1217 *
1218 * Return the length of an &sk_buff queue.
1219 */
1220static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1221{
1222 return list_->qlen;
1223}
1224
67fed459
DM
1225/**
1226 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1227 * @list: queue to initialize
1228 *
1229 * This initializes only the list and queue length aspects of
1230 * an sk_buff_head object. This allows to initialize the list
1231 * aspects of an sk_buff_head without reinitializing things like
1232 * the spinlock. It can also be used for on-stack sk_buff_head
1233 * objects where the spinlock is known to not be used.
1234 */
1235static inline void __skb_queue_head_init(struct sk_buff_head *list)
1236{
1237 list->prev = list->next = (struct sk_buff *)list;
1238 list->qlen = 0;
1239}
1240
76f10ad0
AV
1241/*
1242 * This function creates a split out lock class for each invocation;
1243 * this is needed for now since a whole lot of users of the skb-queue
1244 * infrastructure in drivers have different locking usage (in hardirq)
1245 * than the networking core (in softirq only). In the long run either the
1246 * network layer or drivers should need annotation to consolidate the
1247 * main types of usage into 3 classes.
1248 */
1da177e4
LT
1249static inline void skb_queue_head_init(struct sk_buff_head *list)
1250{
1251 spin_lock_init(&list->lock);
67fed459 1252 __skb_queue_head_init(list);
1da177e4
LT
1253}
1254
c2ecba71
PE
1255static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1256 struct lock_class_key *class)
1257{
1258 skb_queue_head_init(list);
1259 lockdep_set_class(&list->lock, class);
1260}
1261
1da177e4 1262/*
bf299275 1263 * Insert an sk_buff on a list.
1da177e4
LT
1264 *
1265 * The "__skb_xxxx()" functions are the non-atomic ones that
1266 * can only be called with interrupts disabled.
1267 */
7965bd4d
JP
1268void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1269 struct sk_buff_head *list);
bf299275
GR
1270static inline void __skb_insert(struct sk_buff *newsk,
1271 struct sk_buff *prev, struct sk_buff *next,
1272 struct sk_buff_head *list)
1273{
1274 newsk->next = next;
1275 newsk->prev = prev;
1276 next->prev = prev->next = newsk;
1277 list->qlen++;
1278}
1da177e4 1279
67fed459
DM
1280static inline void __skb_queue_splice(const struct sk_buff_head *list,
1281 struct sk_buff *prev,
1282 struct sk_buff *next)
1283{
1284 struct sk_buff *first = list->next;
1285 struct sk_buff *last = list->prev;
1286
1287 first->prev = prev;
1288 prev->next = first;
1289
1290 last->next = next;
1291 next->prev = last;
1292}
1293
1294/**
1295 * skb_queue_splice - join two skb lists, this is designed for stacks
1296 * @list: the new list to add
1297 * @head: the place to add it in the first list
1298 */
1299static inline void skb_queue_splice(const struct sk_buff_head *list,
1300 struct sk_buff_head *head)
1301{
1302 if (!skb_queue_empty(list)) {
1303 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1304 head->qlen += list->qlen;
67fed459
DM
1305 }
1306}
1307
1308/**
d9619496 1309 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1310 * @list: the new list to add
1311 * @head: the place to add it in the first list
1312 *
1313 * The list at @list is reinitialised
1314 */
1315static inline void skb_queue_splice_init(struct sk_buff_head *list,
1316 struct sk_buff_head *head)
1317{
1318 if (!skb_queue_empty(list)) {
1319 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1320 head->qlen += list->qlen;
67fed459
DM
1321 __skb_queue_head_init(list);
1322 }
1323}
1324
1325/**
1326 * skb_queue_splice_tail - join two skb lists, each list being a queue
1327 * @list: the new list to add
1328 * @head: the place to add it in the first list
1329 */
1330static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1331 struct sk_buff_head *head)
1332{
1333 if (!skb_queue_empty(list)) {
1334 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1335 head->qlen += list->qlen;
67fed459
DM
1336 }
1337}
1338
1339/**
d9619496 1340 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1341 * @list: the new list to add
1342 * @head: the place to add it in the first list
1343 *
1344 * Each of the lists is a queue.
1345 * The list at @list is reinitialised
1346 */
1347static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1348 struct sk_buff_head *head)
1349{
1350 if (!skb_queue_empty(list)) {
1351 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1352 head->qlen += list->qlen;
67fed459
DM
1353 __skb_queue_head_init(list);
1354 }
1355}
1356
1da177e4 1357/**
300ce174 1358 * __skb_queue_after - queue a buffer at the list head
1da177e4 1359 * @list: list to use
300ce174 1360 * @prev: place after this buffer
1da177e4
LT
1361 * @newsk: buffer to queue
1362 *
300ce174 1363 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1364 * and you must therefore hold required locks before calling it.
1365 *
1366 * A buffer cannot be placed on two lists at the same time.
1367 */
300ce174
SH
1368static inline void __skb_queue_after(struct sk_buff_head *list,
1369 struct sk_buff *prev,
1370 struct sk_buff *newsk)
1da177e4 1371{
bf299275 1372 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1373}
1374
7965bd4d
JP
1375void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1376 struct sk_buff_head *list);
7de6c033 1377
f5572855
GR
1378static inline void __skb_queue_before(struct sk_buff_head *list,
1379 struct sk_buff *next,
1380 struct sk_buff *newsk)
1381{
1382 __skb_insert(newsk, next->prev, next, list);
1383}
1384
300ce174
SH
1385/**
1386 * __skb_queue_head - queue a buffer at the list head
1387 * @list: list to use
1388 * @newsk: buffer to queue
1389 *
1390 * Queue a buffer at the start of a list. This function takes no locks
1391 * and you must therefore hold required locks before calling it.
1392 *
1393 * A buffer cannot be placed on two lists at the same time.
1394 */
7965bd4d 1395void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1396static inline void __skb_queue_head(struct sk_buff_head *list,
1397 struct sk_buff *newsk)
1398{
1399 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1400}
1401
1da177e4
LT
1402/**
1403 * __skb_queue_tail - queue a buffer at the list tail
1404 * @list: list to use
1405 * @newsk: buffer to queue
1406 *
1407 * Queue a buffer at the end of a list. This function takes no locks
1408 * and you must therefore hold required locks before calling it.
1409 *
1410 * A buffer cannot be placed on two lists at the same time.
1411 */
7965bd4d 1412void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1413static inline void __skb_queue_tail(struct sk_buff_head *list,
1414 struct sk_buff *newsk)
1415{
f5572855 1416 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1417}
1418
1da177e4
LT
1419/*
1420 * remove sk_buff from list. _Must_ be called atomically, and with
1421 * the list known..
1422 */
7965bd4d 1423void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1424static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1425{
1426 struct sk_buff *next, *prev;
1427
1428 list->qlen--;
1429 next = skb->next;
1430 prev = skb->prev;
1431 skb->next = skb->prev = NULL;
1da177e4
LT
1432 next->prev = prev;
1433 prev->next = next;
1434}
1435
f525c06d
GR
1436/**
1437 * __skb_dequeue - remove from the head of the queue
1438 * @list: list to dequeue from
1439 *
1440 * Remove the head of the list. This function does not take any locks
1441 * so must be used with appropriate locks held only. The head item is
1442 * returned or %NULL if the list is empty.
1443 */
7965bd4d 1444struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1445static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1446{
1447 struct sk_buff *skb = skb_peek(list);
1448 if (skb)
1449 __skb_unlink(skb, list);
1450 return skb;
1451}
1da177e4
LT
1452
1453/**
1454 * __skb_dequeue_tail - remove from the tail of the queue
1455 * @list: list to dequeue from
1456 *
1457 * Remove the tail of the list. This function does not take any locks
1458 * so must be used with appropriate locks held only. The tail item is
1459 * returned or %NULL if the list is empty.
1460 */
7965bd4d 1461struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1462static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1463{
1464 struct sk_buff *skb = skb_peek_tail(list);
1465 if (skb)
1466 __skb_unlink(skb, list);
1467 return skb;
1468}
1469
1470
bdcc0924 1471static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1472{
1473 return skb->data_len;
1474}
1475
1476static inline unsigned int skb_headlen(const struct sk_buff *skb)
1477{
1478 return skb->len - skb->data_len;
1479}
1480
1481static inline int skb_pagelen(const struct sk_buff *skb)
1482{
1483 int i, len = 0;
1484
1485 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1486 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1487 return len + skb_headlen(skb);
1488}
1489
131ea667
IC
1490/**
1491 * __skb_fill_page_desc - initialise a paged fragment in an skb
1492 * @skb: buffer containing fragment to be initialised
1493 * @i: paged fragment index to initialise
1494 * @page: the page to use for this fragment
1495 * @off: the offset to the data with @page
1496 * @size: the length of the data
1497 *
1498 * Initialises the @i'th fragment of @skb to point to &size bytes at
1499 * offset @off within @page.
1500 *
1501 * Does not take any additional reference on the fragment.
1502 */
1503static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1504 struct page *page, int off, int size)
1da177e4
LT
1505{
1506 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1507
c48a11c7
MG
1508 /*
1509 * Propagate page->pfmemalloc to the skb if we can. The problem is
1510 * that not all callers have unique ownership of the page. If
1511 * pfmemalloc is set, we check the mapping as a mapping implies
1512 * page->index is set (index and pfmemalloc share space).
1513 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1514 * do not lose pfmemalloc information as the pages would not be
1515 * allocated using __GFP_MEMALLOC.
1516 */
a8605c60 1517 frag->page.p = page;
1da177e4 1518 frag->page_offset = off;
9e903e08 1519 skb_frag_size_set(frag, size);
cca7af38
PE
1520
1521 page = compound_head(page);
1522 if (page->pfmemalloc && !page->mapping)
1523 skb->pfmemalloc = true;
131ea667
IC
1524}
1525
1526/**
1527 * skb_fill_page_desc - initialise a paged fragment in an skb
1528 * @skb: buffer containing fragment to be initialised
1529 * @i: paged fragment index to initialise
1530 * @page: the page to use for this fragment
1531 * @off: the offset to the data with @page
1532 * @size: the length of the data
1533 *
1534 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1535 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1536 * addition updates @skb such that @i is the last fragment.
1537 *
1538 * Does not take any additional reference on the fragment.
1539 */
1540static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1541 struct page *page, int off, int size)
1542{
1543 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1544 skb_shinfo(skb)->nr_frags = i + 1;
1545}
1546
7965bd4d
JP
1547void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1548 int size, unsigned int truesize);
654bed16 1549
f8e617e1
JW
1550void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1551 unsigned int truesize);
1552
1da177e4 1553#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1554#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1555#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1556
27a884dc
ACM
1557#ifdef NET_SKBUFF_DATA_USES_OFFSET
1558static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1559{
1560 return skb->head + skb->tail;
1561}
1562
1563static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1564{
1565 skb->tail = skb->data - skb->head;
1566}
1567
1568static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1569{
1570 skb_reset_tail_pointer(skb);
1571 skb->tail += offset;
1572}
7cc46190 1573
27a884dc
ACM
1574#else /* NET_SKBUFF_DATA_USES_OFFSET */
1575static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1576{
1577 return skb->tail;
1578}
1579
1580static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1581{
1582 skb->tail = skb->data;
1583}
1584
1585static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1586{
1587 skb->tail = skb->data + offset;
1588}
4305b541 1589
27a884dc
ACM
1590#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1591
1da177e4
LT
1592/*
1593 * Add data to an sk_buff
1594 */
0c7ddf36 1595unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1596unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1597static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1598{
27a884dc 1599 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1600 SKB_LINEAR_ASSERT(skb);
1601 skb->tail += len;
1602 skb->len += len;
1603 return tmp;
1604}
1605
7965bd4d 1606unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1607static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1608{
1609 skb->data -= len;
1610 skb->len += len;
1611 return skb->data;
1612}
1613
7965bd4d 1614unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1615static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1616{
1617 skb->len -= len;
1618 BUG_ON(skb->len < skb->data_len);
1619 return skb->data += len;
1620}
1621
47d29646
DM
1622static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1623{
1624 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1625}
1626
7965bd4d 1627unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1628
1629static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1630{
1631 if (len > skb_headlen(skb) &&
987c402a 1632 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1633 return NULL;
1634 skb->len -= len;
1635 return skb->data += len;
1636}
1637
1638static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1639{
1640 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1641}
1642
1643static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1644{
1645 if (likely(len <= skb_headlen(skb)))
1646 return 1;
1647 if (unlikely(len > skb->len))
1648 return 0;
987c402a 1649 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1650}
1651
1652/**
1653 * skb_headroom - bytes at buffer head
1654 * @skb: buffer to check
1655 *
1656 * Return the number of bytes of free space at the head of an &sk_buff.
1657 */
c2636b4d 1658static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1659{
1660 return skb->data - skb->head;
1661}
1662
1663/**
1664 * skb_tailroom - bytes at buffer end
1665 * @skb: buffer to check
1666 *
1667 * Return the number of bytes of free space at the tail of an sk_buff
1668 */
1669static inline int skb_tailroom(const struct sk_buff *skb)
1670{
4305b541 1671 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1672}
1673
a21d4572
ED
1674/**
1675 * skb_availroom - bytes at buffer end
1676 * @skb: buffer to check
1677 *
1678 * Return the number of bytes of free space at the tail of an sk_buff
1679 * allocated by sk_stream_alloc()
1680 */
1681static inline int skb_availroom(const struct sk_buff *skb)
1682{
16fad69c
ED
1683 if (skb_is_nonlinear(skb))
1684 return 0;
1685
1686 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1687}
1688
1da177e4
LT
1689/**
1690 * skb_reserve - adjust headroom
1691 * @skb: buffer to alter
1692 * @len: bytes to move
1693 *
1694 * Increase the headroom of an empty &sk_buff by reducing the tail
1695 * room. This is only allowed for an empty buffer.
1696 */
8243126c 1697static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1698{
1699 skb->data += len;
1700 skb->tail += len;
1701}
1702
6a674e9c
JG
1703static inline void skb_reset_inner_headers(struct sk_buff *skb)
1704{
aefbd2b3 1705 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1706 skb->inner_network_header = skb->network_header;
1707 skb->inner_transport_header = skb->transport_header;
1708}
1709
0b5c9db1
JP
1710static inline void skb_reset_mac_len(struct sk_buff *skb)
1711{
1712 skb->mac_len = skb->network_header - skb->mac_header;
1713}
1714
6a674e9c
JG
1715static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1716 *skb)
1717{
1718 return skb->head + skb->inner_transport_header;
1719}
1720
1721static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1722{
1723 skb->inner_transport_header = skb->data - skb->head;
1724}
1725
1726static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1727 const int offset)
1728{
1729 skb_reset_inner_transport_header(skb);
1730 skb->inner_transport_header += offset;
1731}
1732
1733static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1734{
1735 return skb->head + skb->inner_network_header;
1736}
1737
1738static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1739{
1740 skb->inner_network_header = skb->data - skb->head;
1741}
1742
1743static inline void skb_set_inner_network_header(struct sk_buff *skb,
1744 const int offset)
1745{
1746 skb_reset_inner_network_header(skb);
1747 skb->inner_network_header += offset;
1748}
1749
aefbd2b3
PS
1750static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1751{
1752 return skb->head + skb->inner_mac_header;
1753}
1754
1755static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1756{
1757 skb->inner_mac_header = skb->data - skb->head;
1758}
1759
1760static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1761 const int offset)
1762{
1763 skb_reset_inner_mac_header(skb);
1764 skb->inner_mac_header += offset;
1765}
fda55eca
ED
1766static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1767{
35d04610 1768 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1769}
1770
9c70220b
ACM
1771static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1772{
2e07fa9c 1773 return skb->head + skb->transport_header;
9c70220b
ACM
1774}
1775
badff6d0
ACM
1776static inline void skb_reset_transport_header(struct sk_buff *skb)
1777{
2e07fa9c 1778 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1779}
1780
967b05f6
ACM
1781static inline void skb_set_transport_header(struct sk_buff *skb,
1782 const int offset)
1783{
2e07fa9c
ACM
1784 skb_reset_transport_header(skb);
1785 skb->transport_header += offset;
ea2ae17d
ACM
1786}
1787
d56f90a7
ACM
1788static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1789{
2e07fa9c 1790 return skb->head + skb->network_header;
d56f90a7
ACM
1791}
1792
c1d2bbe1
ACM
1793static inline void skb_reset_network_header(struct sk_buff *skb)
1794{
2e07fa9c 1795 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1796}
1797
c14d2450
ACM
1798static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1799{
2e07fa9c
ACM
1800 skb_reset_network_header(skb);
1801 skb->network_header += offset;
c14d2450
ACM
1802}
1803
2e07fa9c 1804static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1805{
2e07fa9c 1806 return skb->head + skb->mac_header;
bbe735e4
ACM
1807}
1808
2e07fa9c 1809static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1810{
35d04610 1811 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1812}
1813
1814static inline void skb_reset_mac_header(struct sk_buff *skb)
1815{
1816 skb->mac_header = skb->data - skb->head;
1817}
1818
1819static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1820{
1821 skb_reset_mac_header(skb);
1822 skb->mac_header += offset;
1823}
1824
0e3da5bb
TT
1825static inline void skb_pop_mac_header(struct sk_buff *skb)
1826{
1827 skb->mac_header = skb->network_header;
1828}
1829
fbbdb8f0
YX
1830static inline void skb_probe_transport_header(struct sk_buff *skb,
1831 const int offset_hint)
1832{
1833 struct flow_keys keys;
1834
1835 if (skb_transport_header_was_set(skb))
1836 return;
1837 else if (skb_flow_dissect(skb, &keys))
1838 skb_set_transport_header(skb, keys.thoff);
1839 else
1840 skb_set_transport_header(skb, offset_hint);
1841}
1842
03606895
ED
1843static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1844{
1845 if (skb_mac_header_was_set(skb)) {
1846 const unsigned char *old_mac = skb_mac_header(skb);
1847
1848 skb_set_mac_header(skb, -skb->mac_len);
1849 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1850 }
1851}
1852
04fb451e
MM
1853static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1854{
1855 return skb->csum_start - skb_headroom(skb);
1856}
1857
2e07fa9c
ACM
1858static inline int skb_transport_offset(const struct sk_buff *skb)
1859{
1860 return skb_transport_header(skb) - skb->data;
1861}
1862
1863static inline u32 skb_network_header_len(const struct sk_buff *skb)
1864{
1865 return skb->transport_header - skb->network_header;
1866}
1867
6a674e9c
JG
1868static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1869{
1870 return skb->inner_transport_header - skb->inner_network_header;
1871}
1872
2e07fa9c
ACM
1873static inline int skb_network_offset(const struct sk_buff *skb)
1874{
1875 return skb_network_header(skb) - skb->data;
1876}
48d49d0c 1877
6a674e9c
JG
1878static inline int skb_inner_network_offset(const struct sk_buff *skb)
1879{
1880 return skb_inner_network_header(skb) - skb->data;
1881}
1882
f9599ce1
CG
1883static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1884{
1885 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1886}
1887
1da177e4
LT
1888/*
1889 * CPUs often take a performance hit when accessing unaligned memory
1890 * locations. The actual performance hit varies, it can be small if the
1891 * hardware handles it or large if we have to take an exception and fix it
1892 * in software.
1893 *
1894 * Since an ethernet header is 14 bytes network drivers often end up with
1895 * the IP header at an unaligned offset. The IP header can be aligned by
1896 * shifting the start of the packet by 2 bytes. Drivers should do this
1897 * with:
1898 *
8660c124 1899 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1900 *
1901 * The downside to this alignment of the IP header is that the DMA is now
1902 * unaligned. On some architectures the cost of an unaligned DMA is high
1903 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1904 *
1da177e4
LT
1905 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1906 * to be overridden.
1907 */
1908#ifndef NET_IP_ALIGN
1909#define NET_IP_ALIGN 2
1910#endif
1911
025be81e
AB
1912/*
1913 * The networking layer reserves some headroom in skb data (via
1914 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1915 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1916 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1917 *
1918 * Unfortunately this headroom changes the DMA alignment of the resulting
1919 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1920 * on some architectures. An architecture can override this value,
1921 * perhaps setting it to a cacheline in size (since that will maintain
1922 * cacheline alignment of the DMA). It must be a power of 2.
1923 *
d6301d3d 1924 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1925 * headroom, you should not reduce this.
5933dd2f
ED
1926 *
1927 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1928 * to reduce average number of cache lines per packet.
1929 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1930 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1931 */
1932#ifndef NET_SKB_PAD
5933dd2f 1933#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1934#endif
1935
7965bd4d 1936int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1937
1938static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1939{
c4264f27 1940 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1941 WARN_ON(1);
1942 return;
1943 }
27a884dc
ACM
1944 skb->len = len;
1945 skb_set_tail_pointer(skb, len);
1da177e4
LT
1946}
1947
7965bd4d 1948void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1949
1950static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1951{
3cc0e873
HX
1952 if (skb->data_len)
1953 return ___pskb_trim(skb, len);
1954 __skb_trim(skb, len);
1955 return 0;
1da177e4
LT
1956}
1957
1958static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1959{
1960 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1961}
1962
e9fa4f7b
HX
1963/**
1964 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1965 * @skb: buffer to alter
1966 * @len: new length
1967 *
1968 * This is identical to pskb_trim except that the caller knows that
1969 * the skb is not cloned so we should never get an error due to out-
1970 * of-memory.
1971 */
1972static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1973{
1974 int err = pskb_trim(skb, len);
1975 BUG_ON(err);
1976}
1977
1da177e4
LT
1978/**
1979 * skb_orphan - orphan a buffer
1980 * @skb: buffer to orphan
1981 *
1982 * If a buffer currently has an owner then we call the owner's
1983 * destructor function and make the @skb unowned. The buffer continues
1984 * to exist but is no longer charged to its former owner.
1985 */
1986static inline void skb_orphan(struct sk_buff *skb)
1987{
c34a7612 1988 if (skb->destructor) {
1da177e4 1989 skb->destructor(skb);
c34a7612
ED
1990 skb->destructor = NULL;
1991 skb->sk = NULL;
376c7311
ED
1992 } else {
1993 BUG_ON(skb->sk);
c34a7612 1994 }
1da177e4
LT
1995}
1996
a353e0ce
MT
1997/**
1998 * skb_orphan_frags - orphan the frags contained in a buffer
1999 * @skb: buffer to orphan frags from
2000 * @gfp_mask: allocation mask for replacement pages
2001 *
2002 * For each frag in the SKB which needs a destructor (i.e. has an
2003 * owner) create a copy of that frag and release the original
2004 * page by calling the destructor.
2005 */
2006static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2007{
2008 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2009 return 0;
2010 return skb_copy_ubufs(skb, gfp_mask);
2011}
2012
1da177e4
LT
2013/**
2014 * __skb_queue_purge - empty a list
2015 * @list: list to empty
2016 *
2017 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2018 * the list and one reference dropped. This function does not take the
2019 * list lock and the caller must hold the relevant locks to use it.
2020 */
7965bd4d 2021void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2022static inline void __skb_queue_purge(struct sk_buff_head *list)
2023{
2024 struct sk_buff *skb;
2025 while ((skb = __skb_dequeue(list)) != NULL)
2026 kfree_skb(skb);
2027}
2028
e5e67305
AD
2029#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2030#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2031#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2032
7965bd4d 2033void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2034
7965bd4d
JP
2035struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2036 gfp_t gfp_mask);
8af27456
CH
2037
2038/**
2039 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2040 * @dev: network device to receive on
2041 * @length: length to allocate
2042 *
2043 * Allocate a new &sk_buff and assign it a usage count of one. The
2044 * buffer has unspecified headroom built in. Users should allocate
2045 * the headroom they think they need without accounting for the
2046 * built in space. The built in space is used for optimisations.
2047 *
2048 * %NULL is returned if there is no free memory. Although this function
2049 * allocates memory it can be called from an interrupt.
2050 */
2051static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2052 unsigned int length)
8af27456
CH
2053{
2054 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2055}
2056
6f532612
ED
2057/* legacy helper around __netdev_alloc_skb() */
2058static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2059 gfp_t gfp_mask)
2060{
2061 return __netdev_alloc_skb(NULL, length, gfp_mask);
2062}
2063
2064/* legacy helper around netdev_alloc_skb() */
2065static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2066{
2067 return netdev_alloc_skb(NULL, length);
2068}
2069
2070
4915a0de
ED
2071static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2072 unsigned int length, gfp_t gfp)
61321bbd 2073{
4915a0de 2074 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2075
2076 if (NET_IP_ALIGN && skb)
2077 skb_reserve(skb, NET_IP_ALIGN);
2078 return skb;
2079}
2080
4915a0de
ED
2081static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2082 unsigned int length)
2083{
2084 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2085}
2086
bc6fc9fa
FF
2087/**
2088 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2089 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2090 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2091 * @order: size of the allocation
2092 *
2093 * Allocate a new page.
2094 *
2095 * %NULL is returned if there is no free memory.
2096*/
2097static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2098 struct sk_buff *skb,
2099 unsigned int order)
2100{
2101 struct page *page;
2102
2103 gfp_mask |= __GFP_COLD;
2104
2105 if (!(gfp_mask & __GFP_NOMEMALLOC))
2106 gfp_mask |= __GFP_MEMALLOC;
2107
2108 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2109 if (skb && page && page->pfmemalloc)
2110 skb->pfmemalloc = true;
2111
2112 return page;
2113}
2114
2115/**
2116 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2117 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2118 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2119 *
2120 * Allocate a new page.
2121 *
2122 * %NULL is returned if there is no free memory.
2123 */
2124static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2125 struct sk_buff *skb)
2126{
2127 return __skb_alloc_pages(gfp_mask, skb, 0);
2128}
2129
2130/**
2131 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2132 * @page: The page that was allocated from skb_alloc_page
2133 * @skb: The skb that may need pfmemalloc set
2134 */
2135static inline void skb_propagate_pfmemalloc(struct page *page,
2136 struct sk_buff *skb)
2137{
2138 if (page && page->pfmemalloc)
2139 skb->pfmemalloc = true;
2140}
2141
131ea667 2142/**
e227867f 2143 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2144 * @frag: the paged fragment
2145 *
2146 * Returns the &struct page associated with @frag.
2147 */
2148static inline struct page *skb_frag_page(const skb_frag_t *frag)
2149{
a8605c60 2150 return frag->page.p;
131ea667
IC
2151}
2152
2153/**
2154 * __skb_frag_ref - take an addition reference on a paged fragment.
2155 * @frag: the paged fragment
2156 *
2157 * Takes an additional reference on the paged fragment @frag.
2158 */
2159static inline void __skb_frag_ref(skb_frag_t *frag)
2160{
2161 get_page(skb_frag_page(frag));
2162}
2163
2164/**
2165 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2166 * @skb: the buffer
2167 * @f: the fragment offset.
2168 *
2169 * Takes an additional reference on the @f'th paged fragment of @skb.
2170 */
2171static inline void skb_frag_ref(struct sk_buff *skb, int f)
2172{
2173 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2174}
2175
2176/**
2177 * __skb_frag_unref - release a reference on a paged fragment.
2178 * @frag: the paged fragment
2179 *
2180 * Releases a reference on the paged fragment @frag.
2181 */
2182static inline void __skb_frag_unref(skb_frag_t *frag)
2183{
2184 put_page(skb_frag_page(frag));
2185}
2186
2187/**
2188 * skb_frag_unref - release a reference on a paged fragment of an skb.
2189 * @skb: the buffer
2190 * @f: the fragment offset
2191 *
2192 * Releases a reference on the @f'th paged fragment of @skb.
2193 */
2194static inline void skb_frag_unref(struct sk_buff *skb, int f)
2195{
2196 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2197}
2198
2199/**
2200 * skb_frag_address - gets the address of the data contained in a paged fragment
2201 * @frag: the paged fragment buffer
2202 *
2203 * Returns the address of the data within @frag. The page must already
2204 * be mapped.
2205 */
2206static inline void *skb_frag_address(const skb_frag_t *frag)
2207{
2208 return page_address(skb_frag_page(frag)) + frag->page_offset;
2209}
2210
2211/**
2212 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2213 * @frag: the paged fragment buffer
2214 *
2215 * Returns the address of the data within @frag. Checks that the page
2216 * is mapped and returns %NULL otherwise.
2217 */
2218static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2219{
2220 void *ptr = page_address(skb_frag_page(frag));
2221 if (unlikely(!ptr))
2222 return NULL;
2223
2224 return ptr + frag->page_offset;
2225}
2226
2227/**
2228 * __skb_frag_set_page - sets the page contained in a paged fragment
2229 * @frag: the paged fragment
2230 * @page: the page to set
2231 *
2232 * Sets the fragment @frag to contain @page.
2233 */
2234static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2235{
a8605c60 2236 frag->page.p = page;
131ea667
IC
2237}
2238
2239/**
2240 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2241 * @skb: the buffer
2242 * @f: the fragment offset
2243 * @page: the page to set
2244 *
2245 * Sets the @f'th fragment of @skb to contain @page.
2246 */
2247static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2248 struct page *page)
2249{
2250 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2251}
2252
400dfd3a
ED
2253bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2254
131ea667
IC
2255/**
2256 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2257 * @dev: the device to map the fragment to
131ea667
IC
2258 * @frag: the paged fragment to map
2259 * @offset: the offset within the fragment (starting at the
2260 * fragment's own offset)
2261 * @size: the number of bytes to map
f83347df 2262 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2263 *
2264 * Maps the page associated with @frag to @device.
2265 */
2266static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2267 const skb_frag_t *frag,
2268 size_t offset, size_t size,
2269 enum dma_data_direction dir)
2270{
2271 return dma_map_page(dev, skb_frag_page(frag),
2272 frag->page_offset + offset, size, dir);
2273}
2274
117632e6
ED
2275static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2276 gfp_t gfp_mask)
2277{
2278 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2279}
2280
bad93e9d
OP
2281
2282static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2283 gfp_t gfp_mask)
2284{
2285 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2286}
2287
2288
334a8132
PM
2289/**
2290 * skb_clone_writable - is the header of a clone writable
2291 * @skb: buffer to check
2292 * @len: length up to which to write
2293 *
2294 * Returns true if modifying the header part of the cloned buffer
2295 * does not requires the data to be copied.
2296 */
05bdd2f1 2297static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2298{
2299 return !skb_header_cloned(skb) &&
2300 skb_headroom(skb) + len <= skb->hdr_len;
2301}
2302
d9cc2048
HX
2303static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2304 int cloned)
2305{
2306 int delta = 0;
2307
d9cc2048
HX
2308 if (headroom > skb_headroom(skb))
2309 delta = headroom - skb_headroom(skb);
2310
2311 if (delta || cloned)
2312 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2313 GFP_ATOMIC);
2314 return 0;
2315}
2316
1da177e4
LT
2317/**
2318 * skb_cow - copy header of skb when it is required
2319 * @skb: buffer to cow
2320 * @headroom: needed headroom
2321 *
2322 * If the skb passed lacks sufficient headroom or its data part
2323 * is shared, data is reallocated. If reallocation fails, an error
2324 * is returned and original skb is not changed.
2325 *
2326 * The result is skb with writable area skb->head...skb->tail
2327 * and at least @headroom of space at head.
2328 */
2329static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2330{
d9cc2048
HX
2331 return __skb_cow(skb, headroom, skb_cloned(skb));
2332}
1da177e4 2333
d9cc2048
HX
2334/**
2335 * skb_cow_head - skb_cow but only making the head writable
2336 * @skb: buffer to cow
2337 * @headroom: needed headroom
2338 *
2339 * This function is identical to skb_cow except that we replace the
2340 * skb_cloned check by skb_header_cloned. It should be used when
2341 * you only need to push on some header and do not need to modify
2342 * the data.
2343 */
2344static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2345{
2346 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2347}
2348
2349/**
2350 * skb_padto - pad an skbuff up to a minimal size
2351 * @skb: buffer to pad
2352 * @len: minimal length
2353 *
2354 * Pads up a buffer to ensure the trailing bytes exist and are
2355 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2356 * is untouched. Otherwise it is extended. Returns zero on
2357 * success. The skb is freed on error.
1da177e4
LT
2358 */
2359
5b057c6b 2360static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2361{
2362 unsigned int size = skb->len;
2363 if (likely(size >= len))
5b057c6b 2364 return 0;
987c402a 2365 return skb_pad(skb, len - size);
1da177e4
LT
2366}
2367
2368static inline int skb_add_data(struct sk_buff *skb,
2369 char __user *from, int copy)
2370{
2371 const int off = skb->len;
2372
2373 if (skb->ip_summed == CHECKSUM_NONE) {
2374 int err = 0;
5084205f 2375 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2376 copy, 0, &err);
2377 if (!err) {
2378 skb->csum = csum_block_add(skb->csum, csum, off);
2379 return 0;
2380 }
2381 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2382 return 0;
2383
2384 __skb_trim(skb, off);
2385 return -EFAULT;
2386}
2387
38ba0a65
ED
2388static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2389 const struct page *page, int off)
1da177e4
LT
2390{
2391 if (i) {
9e903e08 2392 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2393
ea2ab693 2394 return page == skb_frag_page(frag) &&
9e903e08 2395 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2396 }
38ba0a65 2397 return false;
1da177e4
LT
2398}
2399
364c6bad
HX
2400static inline int __skb_linearize(struct sk_buff *skb)
2401{
2402 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2403}
2404
1da177e4
LT
2405/**
2406 * skb_linearize - convert paged skb to linear one
2407 * @skb: buffer to linarize
1da177e4
LT
2408 *
2409 * If there is no free memory -ENOMEM is returned, otherwise zero
2410 * is returned and the old skb data released.
2411 */
364c6bad
HX
2412static inline int skb_linearize(struct sk_buff *skb)
2413{
2414 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2415}
2416
cef401de
ED
2417/**
2418 * skb_has_shared_frag - can any frag be overwritten
2419 * @skb: buffer to test
2420 *
2421 * Return true if the skb has at least one frag that might be modified
2422 * by an external entity (as in vmsplice()/sendfile())
2423 */
2424static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2425{
c9af6db4
PS
2426 return skb_is_nonlinear(skb) &&
2427 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2428}
2429
364c6bad
HX
2430/**
2431 * skb_linearize_cow - make sure skb is linear and writable
2432 * @skb: buffer to process
2433 *
2434 * If there is no free memory -ENOMEM is returned, otherwise zero
2435 * is returned and the old skb data released.
2436 */
2437static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2438{
364c6bad
HX
2439 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2440 __skb_linearize(skb) : 0;
1da177e4
LT
2441}
2442
2443/**
2444 * skb_postpull_rcsum - update checksum for received skb after pull
2445 * @skb: buffer to update
2446 * @start: start of data before pull
2447 * @len: length of data pulled
2448 *
2449 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2450 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2451 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2452 */
2453
2454static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2455 const void *start, unsigned int len)
1da177e4 2456{
84fa7933 2457 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2458 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2459}
2460
cbb042f9
HX
2461unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2462
7ce5a27f
DM
2463/**
2464 * pskb_trim_rcsum - trim received skb and update checksum
2465 * @skb: buffer to trim
2466 * @len: new length
2467 *
2468 * This is exactly the same as pskb_trim except that it ensures the
2469 * checksum of received packets are still valid after the operation.
2470 */
2471
2472static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2473{
2474 if (likely(len >= skb->len))
2475 return 0;
2476 if (skb->ip_summed == CHECKSUM_COMPLETE)
2477 skb->ip_summed = CHECKSUM_NONE;
2478 return __pskb_trim(skb, len);
2479}
2480
1da177e4
LT
2481#define skb_queue_walk(queue, skb) \
2482 for (skb = (queue)->next; \
a1e4891f 2483 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2484 skb = skb->next)
2485
46f8914e
JC
2486#define skb_queue_walk_safe(queue, skb, tmp) \
2487 for (skb = (queue)->next, tmp = skb->next; \
2488 skb != (struct sk_buff *)(queue); \
2489 skb = tmp, tmp = skb->next)
2490
1164f52a 2491#define skb_queue_walk_from(queue, skb) \
a1e4891f 2492 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2493 skb = skb->next)
2494
2495#define skb_queue_walk_from_safe(queue, skb, tmp) \
2496 for (tmp = skb->next; \
2497 skb != (struct sk_buff *)(queue); \
2498 skb = tmp, tmp = skb->next)
2499
300ce174
SH
2500#define skb_queue_reverse_walk(queue, skb) \
2501 for (skb = (queue)->prev; \
a1e4891f 2502 skb != (struct sk_buff *)(queue); \
300ce174
SH
2503 skb = skb->prev)
2504
686a2955
DM
2505#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2506 for (skb = (queue)->prev, tmp = skb->prev; \
2507 skb != (struct sk_buff *)(queue); \
2508 skb = tmp, tmp = skb->prev)
2509
2510#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2511 for (tmp = skb->prev; \
2512 skb != (struct sk_buff *)(queue); \
2513 skb = tmp, tmp = skb->prev)
1da177e4 2514
21dc3301 2515static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2516{
2517 return skb_shinfo(skb)->frag_list != NULL;
2518}
2519
2520static inline void skb_frag_list_init(struct sk_buff *skb)
2521{
2522 skb_shinfo(skb)->frag_list = NULL;
2523}
2524
2525static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2526{
2527 frag->next = skb_shinfo(skb)->frag_list;
2528 skb_shinfo(skb)->frag_list = frag;
2529}
2530
2531#define skb_walk_frags(skb, iter) \
2532 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2533
7965bd4d
JP
2534struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2535 int *peeked, int *off, int *err);
2536struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2537 int *err);
2538unsigned int datagram_poll(struct file *file, struct socket *sock,
2539 struct poll_table_struct *wait);
2540int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2541 struct iovec *to, int size);
2542int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2543 struct iovec *iov);
2544int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2545 const struct iovec *from, int from_offset,
2546 int len);
2547int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2548 int offset, size_t count);
2549int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2550 const struct iovec *to, int to_offset,
2551 int size);
2552void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2553void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2554int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2555int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2556int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2557__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2558 int len, __wsum csum);
2559int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2560 struct pipe_inode_info *pipe, unsigned int len,
2561 unsigned int flags);
2562void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2563unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2564int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2565 int len, int hlen);
7965bd4d
JP
2566void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2567int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2568void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2569unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2570struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2571struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2572
2817a336
DB
2573struct skb_checksum_ops {
2574 __wsum (*update)(const void *mem, int len, __wsum wsum);
2575 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2576};
2577
2578__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2579 __wsum csum, const struct skb_checksum_ops *ops);
2580__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2581 __wsum csum);
2582
690e36e7
DM
2583static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2584 int len, void *data, int hlen, void *buffer)
1da177e4 2585{
55820ee2 2586 if (hlen - offset >= len)
690e36e7 2587 return data + offset;
1da177e4 2588
690e36e7
DM
2589 if (!skb ||
2590 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2591 return NULL;
2592
2593 return buffer;
2594}
2595
690e36e7
DM
2596static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2597 int len, void *buffer)
2598{
2599 return __skb_header_pointer(skb, offset, len, skb->data,
2600 skb_headlen(skb), buffer);
2601}
2602
4262e5cc
DB
2603/**
2604 * skb_needs_linearize - check if we need to linearize a given skb
2605 * depending on the given device features.
2606 * @skb: socket buffer to check
2607 * @features: net device features
2608 *
2609 * Returns true if either:
2610 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2611 * 2. skb is fragmented and the device does not support SG.
2612 */
2613static inline bool skb_needs_linearize(struct sk_buff *skb,
2614 netdev_features_t features)
2615{
2616 return skb_is_nonlinear(skb) &&
2617 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2618 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2619}
2620
d626f62b
ACM
2621static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2622 void *to,
2623 const unsigned int len)
2624{
2625 memcpy(to, skb->data, len);
2626}
2627
2628static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2629 const int offset, void *to,
2630 const unsigned int len)
2631{
2632 memcpy(to, skb->data + offset, len);
2633}
2634
27d7ff46
ACM
2635static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2636 const void *from,
2637 const unsigned int len)
2638{
2639 memcpy(skb->data, from, len);
2640}
2641
2642static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2643 const int offset,
2644 const void *from,
2645 const unsigned int len)
2646{
2647 memcpy(skb->data + offset, from, len);
2648}
2649
7965bd4d 2650void skb_init(void);
1da177e4 2651
ac45f602
PO
2652static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2653{
2654 return skb->tstamp;
2655}
2656
a61bbcf2
PM
2657/**
2658 * skb_get_timestamp - get timestamp from a skb
2659 * @skb: skb to get stamp from
2660 * @stamp: pointer to struct timeval to store stamp in
2661 *
2662 * Timestamps are stored in the skb as offsets to a base timestamp.
2663 * This function converts the offset back to a struct timeval and stores
2664 * it in stamp.
2665 */
ac45f602
PO
2666static inline void skb_get_timestamp(const struct sk_buff *skb,
2667 struct timeval *stamp)
a61bbcf2 2668{
b7aa0bf7 2669 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2670}
2671
ac45f602
PO
2672static inline void skb_get_timestampns(const struct sk_buff *skb,
2673 struct timespec *stamp)
2674{
2675 *stamp = ktime_to_timespec(skb->tstamp);
2676}
2677
b7aa0bf7 2678static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2679{
b7aa0bf7 2680 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2681}
2682
164891aa
SH
2683static inline ktime_t net_timedelta(ktime_t t)
2684{
2685 return ktime_sub(ktime_get_real(), t);
2686}
2687
b9ce204f
IJ
2688static inline ktime_t net_invalid_timestamp(void)
2689{
2690 return ktime_set(0, 0);
2691}
a61bbcf2 2692
62bccb8c
AD
2693struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2694
c1f19b51
RC
2695#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2696
7965bd4d
JP
2697void skb_clone_tx_timestamp(struct sk_buff *skb);
2698bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2699
2700#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2701
2702static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2703{
2704}
2705
2706static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2707{
2708 return false;
2709}
2710
2711#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2712
2713/**
2714 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2715 *
da92b194
RC
2716 * PHY drivers may accept clones of transmitted packets for
2717 * timestamping via their phy_driver.txtstamp method. These drivers
2718 * must call this function to return the skb back to the stack, with
2719 * or without a timestamp.
2720 *
c1f19b51 2721 * @skb: clone of the the original outgoing packet
da92b194 2722 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2723 *
2724 */
2725void skb_complete_tx_timestamp(struct sk_buff *skb,
2726 struct skb_shared_hwtstamps *hwtstamps);
2727
e7fd2885
WB
2728void __skb_tstamp_tx(struct sk_buff *orig_skb,
2729 struct skb_shared_hwtstamps *hwtstamps,
2730 struct sock *sk, int tstype);
2731
ac45f602
PO
2732/**
2733 * skb_tstamp_tx - queue clone of skb with send time stamps
2734 * @orig_skb: the original outgoing packet
2735 * @hwtstamps: hardware time stamps, may be NULL if not available
2736 *
2737 * If the skb has a socket associated, then this function clones the
2738 * skb (thus sharing the actual data and optional structures), stores
2739 * the optional hardware time stamping information (if non NULL) or
2740 * generates a software time stamp (otherwise), then queues the clone
2741 * to the error queue of the socket. Errors are silently ignored.
2742 */
7965bd4d
JP
2743void skb_tstamp_tx(struct sk_buff *orig_skb,
2744 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2745
4507a715
RC
2746static inline void sw_tx_timestamp(struct sk_buff *skb)
2747{
2244d07b
OH
2748 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2749 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2750 skb_tstamp_tx(skb, NULL);
2751}
2752
2753/**
2754 * skb_tx_timestamp() - Driver hook for transmit timestamping
2755 *
2756 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2757 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2758 *
73409f3b
DM
2759 * Specifically, one should make absolutely sure that this function is
2760 * called before TX completion of this packet can trigger. Otherwise
2761 * the packet could potentially already be freed.
2762 *
4507a715
RC
2763 * @skb: A socket buffer.
2764 */
2765static inline void skb_tx_timestamp(struct sk_buff *skb)
2766{
c1f19b51 2767 skb_clone_tx_timestamp(skb);
4507a715
RC
2768 sw_tx_timestamp(skb);
2769}
2770
6e3e939f
JB
2771/**
2772 * skb_complete_wifi_ack - deliver skb with wifi status
2773 *
2774 * @skb: the original outgoing packet
2775 * @acked: ack status
2776 *
2777 */
2778void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2779
7965bd4d
JP
2780__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2781__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2782
60476372
HX
2783static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2784{
5d0c2b95 2785 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2786}
2787
fb286bb2
HX
2788/**
2789 * skb_checksum_complete - Calculate checksum of an entire packet
2790 * @skb: packet to process
2791 *
2792 * This function calculates the checksum over the entire packet plus
2793 * the value of skb->csum. The latter can be used to supply the
2794 * checksum of a pseudo header as used by TCP/UDP. It returns the
2795 * checksum.
2796 *
2797 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2798 * this function can be used to verify that checksum on received
2799 * packets. In that case the function should return zero if the
2800 * checksum is correct. In particular, this function will return zero
2801 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2802 * hardware has already verified the correctness of the checksum.
2803 */
4381ca3c 2804static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2805{
60476372
HX
2806 return skb_csum_unnecessary(skb) ?
2807 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2808}
2809
77cffe23
TH
2810static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2811{
2812 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2813 if (skb->csum_level == 0)
2814 skb->ip_summed = CHECKSUM_NONE;
2815 else
2816 skb->csum_level--;
2817 }
2818}
2819
2820static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2821{
2822 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2823 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2824 skb->csum_level++;
2825 } else if (skb->ip_summed == CHECKSUM_NONE) {
2826 skb->ip_summed = CHECKSUM_UNNECESSARY;
2827 skb->csum_level = 0;
2828 }
2829}
2830
5a212329
TH
2831static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2832{
2833 /* Mark current checksum as bad (typically called from GRO
2834 * path). In the case that ip_summed is CHECKSUM_NONE
2835 * this must be the first checksum encountered in the packet.
2836 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2837 * checksum after the last one validated. For UDP, a zero
2838 * checksum can not be marked as bad.
2839 */
2840
2841 if (skb->ip_summed == CHECKSUM_NONE ||
2842 skb->ip_summed == CHECKSUM_UNNECESSARY)
2843 skb->csum_bad = 1;
2844}
2845
76ba0aae
TH
2846/* Check if we need to perform checksum complete validation.
2847 *
2848 * Returns true if checksum complete is needed, false otherwise
2849 * (either checksum is unnecessary or zero checksum is allowed).
2850 */
2851static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2852 bool zero_okay,
2853 __sum16 check)
2854{
5d0c2b95
TH
2855 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2856 skb->csum_valid = 1;
77cffe23 2857 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2858 return false;
2859 }
2860
2861 return true;
2862}
2863
2864/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2865 * in checksum_init.
2866 */
2867#define CHECKSUM_BREAK 76
2868
2869/* Validate (init) checksum based on checksum complete.
2870 *
2871 * Return values:
2872 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2873 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2874 * checksum is stored in skb->csum for use in __skb_checksum_complete
2875 * non-zero: value of invalid checksum
2876 *
2877 */
2878static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2879 bool complete,
2880 __wsum psum)
2881{
2882 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2883 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2884 skb->csum_valid = 1;
76ba0aae
TH
2885 return 0;
2886 }
5a212329
TH
2887 } else if (skb->csum_bad) {
2888 /* ip_summed == CHECKSUM_NONE in this case */
2889 return 1;
76ba0aae
TH
2890 }
2891
2892 skb->csum = psum;
2893
5d0c2b95
TH
2894 if (complete || skb->len <= CHECKSUM_BREAK) {
2895 __sum16 csum;
2896
2897 csum = __skb_checksum_complete(skb);
2898 skb->csum_valid = !csum;
2899 return csum;
2900 }
76ba0aae
TH
2901
2902 return 0;
2903}
2904
2905static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2906{
2907 return 0;
2908}
2909
2910/* Perform checksum validate (init). Note that this is a macro since we only
2911 * want to calculate the pseudo header which is an input function if necessary.
2912 * First we try to validate without any computation (checksum unnecessary) and
2913 * then calculate based on checksum complete calling the function to compute
2914 * pseudo header.
2915 *
2916 * Return values:
2917 * 0: checksum is validated or try to in skb_checksum_complete
2918 * non-zero: value of invalid checksum
2919 */
2920#define __skb_checksum_validate(skb, proto, complete, \
2921 zero_okay, check, compute_pseudo) \
2922({ \
2923 __sum16 __ret = 0; \
5d0c2b95 2924 skb->csum_valid = 0; \
76ba0aae
TH
2925 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2926 __ret = __skb_checksum_validate_complete(skb, \
2927 complete, compute_pseudo(skb, proto)); \
2928 __ret; \
2929})
2930
2931#define skb_checksum_init(skb, proto, compute_pseudo) \
2932 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2933
2934#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2935 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2936
2937#define skb_checksum_validate(skb, proto, compute_pseudo) \
2938 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2939
2940#define skb_checksum_validate_zero_check(skb, proto, check, \
2941 compute_pseudo) \
2942 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2943
2944#define skb_checksum_simple_validate(skb) \
2945 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2946
d96535a1
TH
2947static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
2948{
2949 return (skb->ip_summed == CHECKSUM_NONE &&
2950 skb->csum_valid && !skb->csum_bad);
2951}
2952
2953static inline void __skb_checksum_convert(struct sk_buff *skb,
2954 __sum16 check, __wsum pseudo)
2955{
2956 skb->csum = ~pseudo;
2957 skb->ip_summed = CHECKSUM_COMPLETE;
2958}
2959
2960#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
2961do { \
2962 if (__skb_checksum_convert_check(skb)) \
2963 __skb_checksum_convert(skb, check, \
2964 compute_pseudo(skb, proto)); \
2965} while (0)
2966
5f79e0f9 2967#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2968void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2969static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2970{
2971 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2972 nf_conntrack_destroy(nfct);
1da177e4
LT
2973}
2974static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2975{
2976 if (nfct)
2977 atomic_inc(&nfct->use);
2978}
2fc72c7b 2979#endif
1da177e4
LT
2980#ifdef CONFIG_BRIDGE_NETFILTER
2981static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2982{
2983 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2984 kfree(nf_bridge);
2985}
2986static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2987{
2988 if (nf_bridge)
2989 atomic_inc(&nf_bridge->use);
2990}
2991#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2992static inline void nf_reset(struct sk_buff *skb)
2993{
5f79e0f9 2994#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2995 nf_conntrack_put(skb->nfct);
2996 skb->nfct = NULL;
2fc72c7b 2997#endif
a193a4ab
PM
2998#ifdef CONFIG_BRIDGE_NETFILTER
2999 nf_bridge_put(skb->nf_bridge);
3000 skb->nf_bridge = NULL;
3001#endif
3002}
3003
124dff01
PM
3004static inline void nf_reset_trace(struct sk_buff *skb)
3005{
478b360a 3006#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3007 skb->nf_trace = 0;
3008#endif
a193a4ab
PM
3009}
3010
edda553c
YK
3011/* Note: This doesn't put any conntrack and bridge info in dst. */
3012static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3013{
5f79e0f9 3014#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3015 dst->nfct = src->nfct;
3016 nf_conntrack_get(src->nfct);
3017 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3018#endif
edda553c
YK
3019#ifdef CONFIG_BRIDGE_NETFILTER
3020 dst->nf_bridge = src->nf_bridge;
3021 nf_bridge_get(src->nf_bridge);
3022#endif
478b360a
FW
3023#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3024 dst->nf_trace = src->nf_trace;
3025#endif
edda553c
YK
3026}
3027
e7ac05f3
YK
3028static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3029{
e7ac05f3 3030#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3031 nf_conntrack_put(dst->nfct);
2fc72c7b 3032#endif
e7ac05f3
YK
3033#ifdef CONFIG_BRIDGE_NETFILTER
3034 nf_bridge_put(dst->nf_bridge);
3035#endif
3036 __nf_copy(dst, src);
3037}
3038
984bc16c
JM
3039#ifdef CONFIG_NETWORK_SECMARK
3040static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3041{
3042 to->secmark = from->secmark;
3043}
3044
3045static inline void skb_init_secmark(struct sk_buff *skb)
3046{
3047 skb->secmark = 0;
3048}
3049#else
3050static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3051{ }
3052
3053static inline void skb_init_secmark(struct sk_buff *skb)
3054{ }
3055#endif
3056
574f7194
EB
3057static inline bool skb_irq_freeable(const struct sk_buff *skb)
3058{
3059 return !skb->destructor &&
3060#if IS_ENABLED(CONFIG_XFRM)
3061 !skb->sp &&
3062#endif
3063#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3064 !skb->nfct &&
3065#endif
3066 !skb->_skb_refdst &&
3067 !skb_has_frag_list(skb);
3068}
3069
f25f4e44
PWJ
3070static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3071{
f25f4e44 3072 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3073}
3074
9247744e 3075static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3076{
4e3ab47a 3077 return skb->queue_mapping;
4e3ab47a
PE
3078}
3079
f25f4e44
PWJ
3080static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3081{
f25f4e44 3082 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3083}
3084
d5a9e24a
DM
3085static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3086{
3087 skb->queue_mapping = rx_queue + 1;
3088}
3089
9247744e 3090static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3091{
3092 return skb->queue_mapping - 1;
3093}
3094
9247744e 3095static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3096{
a02cec21 3097 return skb->queue_mapping != 0;
d5a9e24a
DM
3098}
3099
0e001614 3100u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3101 unsigned int num_tx_queues);
9247744e 3102
def8b4fa
AD
3103static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3104{
0b3d8e08 3105#ifdef CONFIG_XFRM
def8b4fa 3106 return skb->sp;
def8b4fa 3107#else
def8b4fa 3108 return NULL;
def8b4fa 3109#endif
0b3d8e08 3110}
def8b4fa 3111
68c33163
PS
3112/* Keeps track of mac header offset relative to skb->head.
3113 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3114 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3115 * tunnel skb it points to outer mac header.
3116 * Keeps track of level of encapsulation of network headers.
3117 */
68c33163 3118struct skb_gso_cb {
3347c960
ED
3119 int mac_offset;
3120 int encap_level;
7e2b10c1 3121 __u16 csum_start;
68c33163
PS
3122};
3123#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3124
3125static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3126{
3127 return (skb_mac_header(inner_skb) - inner_skb->head) -
3128 SKB_GSO_CB(inner_skb)->mac_offset;
3129}
3130
1e2bd517
PS
3131static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3132{
3133 int new_headroom, headroom;
3134 int ret;
3135
3136 headroom = skb_headroom(skb);
3137 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3138 if (ret)
3139 return ret;
3140
3141 new_headroom = skb_headroom(skb);
3142 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3143 return 0;
3144}
3145
7e2b10c1
TH
3146/* Compute the checksum for a gso segment. First compute the checksum value
3147 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3148 * then add in skb->csum (checksum from csum_start to end of packet).
3149 * skb->csum and csum_start are then updated to reflect the checksum of the
3150 * resultant packet starting from the transport header-- the resultant checksum
3151 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3152 * header.
3153 */
3154static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3155{
3156 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3157 skb_transport_offset(skb);
3158 __u16 csum;
3159
3160 csum = csum_fold(csum_partial(skb_transport_header(skb),
3161 plen, skb->csum));
3162 skb->csum = res;
3163 SKB_GSO_CB(skb)->csum_start -= plen;
3164
3165 return csum;
3166}
3167
bdcc0924 3168static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3169{
3170 return skb_shinfo(skb)->gso_size;
3171}
3172
36a8f39e 3173/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3174static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3175{
3176 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3177}
3178
7965bd4d 3179void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3180
3181static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3182{
3183 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3184 * wanted then gso_type will be set. */
05bdd2f1
ED
3185 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3186
b78462eb
AD
3187 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3188 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3189 __skb_warn_lro_forwarding(skb);
3190 return true;
3191 }
3192 return false;
3193}
3194
35fc92a9
HX
3195static inline void skb_forward_csum(struct sk_buff *skb)
3196{
3197 /* Unfortunately we don't support this one. Any brave souls? */
3198 if (skb->ip_summed == CHECKSUM_COMPLETE)
3199 skb->ip_summed = CHECKSUM_NONE;
3200}
3201
bc8acf2c
ED
3202/**
3203 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3204 * @skb: skb to check
3205 *
3206 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3207 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3208 * use this helper, to document places where we make this assertion.
3209 */
05bdd2f1 3210static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3211{
3212#ifdef DEBUG
3213 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3214#endif
3215}
3216
f35d9d8a 3217bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3218
ed1f50c3
PD
3219int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3220
f77668dc
DB
3221u32 __skb_get_poff(const struct sk_buff *skb);
3222
3a7c1ee4
AD
3223/**
3224 * skb_head_is_locked - Determine if the skb->head is locked down
3225 * @skb: skb to check
3226 *
3227 * The head on skbs build around a head frag can be removed if they are
3228 * not cloned. This function returns true if the skb head is locked down
3229 * due to either being allocated via kmalloc, or by being a clone with
3230 * multiple references to the head.
3231 */
3232static inline bool skb_head_is_locked(const struct sk_buff *skb)
3233{
3234 return !skb->head_frag || skb_cloned(skb);
3235}
fe6cc55f
FW
3236
3237/**
3238 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3239 *
3240 * @skb: GSO skb
3241 *
3242 * skb_gso_network_seglen is used to determine the real size of the
3243 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3244 *
3245 * The MAC/L2 header is not accounted for.
3246 */
3247static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3248{
3249 unsigned int hdr_len = skb_transport_header(skb) -
3250 skb_network_header(skb);
3251 return hdr_len + skb_gso_transport_seglen(skb);
3252}
1da177e4
LT
3253#endif /* __KERNEL__ */
3254#endif /* _LINUX_SKBUFF_H */