tcp: change TCP_ECN prefixes to lower case
[linux-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;
b1937227 530 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
531#ifdef CONFIG_XFRM
532 struct sec_path *sp;
b1937227
ED
533#endif
534#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
535 struct nf_conntrack *nfct;
536#endif
537#ifdef CONFIG_BRIDGE_NETFILTER
538 struct nf_bridge_info *nf_bridge;
da3f5cf1 539#endif
1da177e4 540 unsigned int len,
334a8132
PM
541 data_len;
542 __u16 mac_len,
543 hdr_len;
b1937227
ED
544
545 /* Following fields are _not_ copied in __copy_skb_header()
546 * Note that queue_mapping is here mostly to fill a hole.
547 */
fe55f6d5 548 kmemcheck_bitfield_begin(flags1);
b1937227
ED
549 __u16 queue_mapping;
550 __u8 cloned:1,
6869c4d8 551 nohdr:1,
b1937227
ED
552 fclone:2,
553 peeked:1,
554 head_frag:1,
555 xmit_more:1;
556 /* one bit hole */
557 kmemcheck_bitfield_end(flags1);
558
559
560
561 /* fields enclosed in headers_start/headers_end are copied
562 * using a single memcpy() in __copy_skb_header()
563 */
564 __u32 headers_start[0];
233577a2
HFS
565
566/* if you move pkt_type around you also must adapt those constants */
567#ifdef __BIG_ENDIAN_BITFIELD
568#define PKT_TYPE_MAX (7 << 5)
569#else
570#define PKT_TYPE_MAX 7
571#endif
572#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
573
574 __u8 __pkt_type_offset[0];
b1937227 575 __u8 pkt_type:3;
c93bdd0e 576 __u8 pfmemalloc:1;
b1937227
ED
577 __u8 ignore_df:1;
578 __u8 nfctinfo:3;
579
580 __u8 nf_trace:1;
581 __u8 ip_summed:2;
3853b584 582 __u8 ooo_okay:1;
61b905da 583 __u8 l4_hash:1;
a3b18ddb 584 __u8 sw_hash:1;
6e3e939f
JB
585 __u8 wifi_acked_valid:1;
586 __u8 wifi_acked:1;
b1937227 587
3bdc0eba 588 __u8 no_fcs:1;
77cffe23 589 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 590 __u8 encapsulation:1;
7e2b10c1 591 __u8 encap_hdr_csum:1;
5d0c2b95 592 __u8 csum_valid:1;
7e3cead5 593 __u8 csum_complete_sw:1;
b1937227
ED
594 __u8 csum_level:2;
595 __u8 csum_bad:1;
fe55f6d5 596
b1937227
ED
597#ifdef CONFIG_IPV6_NDISC_NODETYPE
598 __u8 ndisc_nodetype:2;
599#endif
600 __u8 ipvs_property:1;
601 /* 5 or 7 bit hole */
602
603#ifdef CONFIG_NET_SCHED
604 __u16 tc_index; /* traffic control index */
605#ifdef CONFIG_NET_CLS_ACT
606 __u16 tc_verd; /* traffic control verdict */
607#endif
608#endif
609
610 union {
611 __wsum csum;
612 struct {
613 __u16 csum_start;
614 __u16 csum_offset;
615 };
616 };
617 __u32 priority;
618 int skb_iif;
619 __u32 hash;
620 __be16 vlan_proto;
621 __u16 vlan_tci;
e0d1095a 622#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
623 union {
624 unsigned int napi_id;
625 dma_cookie_t dma_cookie;
626 };
97fc2f08 627#endif
984bc16c
JM
628#ifdef CONFIG_NETWORK_SECMARK
629 __u32 secmark;
630#endif
3b885787
NH
631 union {
632 __u32 mark;
633 __u32 dropcount;
16fad69c 634 __u32 reserved_tailroom;
3b885787 635 };
1da177e4 636
0d89d203 637 __be16 inner_protocol;
1a37e412
SH
638 __u16 inner_transport_header;
639 __u16 inner_network_header;
640 __u16 inner_mac_header;
b1937227
ED
641
642 __be16 protocol;
1a37e412
SH
643 __u16 transport_header;
644 __u16 network_header;
645 __u16 mac_header;
b1937227
ED
646
647 __u32 headers_end[0];
648
1da177e4 649 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 650 sk_buff_data_t tail;
4305b541 651 sk_buff_data_t end;
1da177e4 652 unsigned char *head,
4305b541 653 *data;
27a884dc
ACM
654 unsigned int truesize;
655 atomic_t users;
1da177e4
LT
656};
657
658#ifdef __KERNEL__
659/*
660 * Handling routines are only of interest to the kernel
661 */
662#include <linux/slab.h>
663
1da177e4 664
c93bdd0e
MG
665#define SKB_ALLOC_FCLONE 0x01
666#define SKB_ALLOC_RX 0x02
667
668/* Returns true if the skb was allocated from PFMEMALLOC reserves */
669static inline bool skb_pfmemalloc(const struct sk_buff *skb)
670{
671 return unlikely(skb->pfmemalloc);
672}
673
7fee226a
ED
674/*
675 * skb might have a dst pointer attached, refcounted or not.
676 * _skb_refdst low order bit is set if refcount was _not_ taken
677 */
678#define SKB_DST_NOREF 1UL
679#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
680
681/**
682 * skb_dst - returns skb dst_entry
683 * @skb: buffer
684 *
685 * Returns skb dst_entry, regardless of reference taken or not.
686 */
adf30907
ED
687static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
688{
7fee226a
ED
689 /* If refdst was not refcounted, check we still are in a
690 * rcu_read_lock section
691 */
692 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
693 !rcu_read_lock_held() &&
694 !rcu_read_lock_bh_held());
695 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
696}
697
7fee226a
ED
698/**
699 * skb_dst_set - sets skb dst
700 * @skb: buffer
701 * @dst: dst entry
702 *
703 * Sets skb dst, assuming a reference was taken on dst and should
704 * be released by skb_dst_drop()
705 */
adf30907
ED
706static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
707{
7fee226a
ED
708 skb->_skb_refdst = (unsigned long)dst;
709}
710
7965bd4d
JP
711void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
712 bool force);
932bc4d7
JA
713
714/**
715 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
716 * @skb: buffer
717 * @dst: dst entry
718 *
719 * Sets skb dst, assuming a reference was not taken on dst.
720 * If dst entry is cached, we do not take reference and dst_release
721 * will be avoided by refdst_drop. If dst entry is not cached, we take
722 * reference, so that last dst_release can destroy the dst immediately.
723 */
724static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
725{
726 __skb_dst_set_noref(skb, dst, false);
727}
728
729/**
730 * skb_dst_set_noref_force - sets skb dst, without taking reference
731 * @skb: buffer
732 * @dst: dst entry
733 *
734 * Sets skb dst, assuming a reference was not taken on dst.
735 * No reference is taken and no dst_release will be called. While for
736 * cached dsts deferred reclaim is a basic feature, for entries that are
737 * not cached it is caller's job to guarantee that last dst_release for
738 * provided dst happens when nobody uses it, eg. after a RCU grace period.
739 */
740static inline void skb_dst_set_noref_force(struct sk_buff *skb,
741 struct dst_entry *dst)
742{
743 __skb_dst_set_noref(skb, dst, true);
744}
7fee226a
ED
745
746/**
25985edc 747 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
748 * @skb: buffer
749 */
750static inline bool skb_dst_is_noref(const struct sk_buff *skb)
751{
752 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
753}
754
511c3f92
ED
755static inline struct rtable *skb_rtable(const struct sk_buff *skb)
756{
adf30907 757 return (struct rtable *)skb_dst(skb);
511c3f92
ED
758}
759
7965bd4d
JP
760void kfree_skb(struct sk_buff *skb);
761void kfree_skb_list(struct sk_buff *segs);
762void skb_tx_error(struct sk_buff *skb);
763void consume_skb(struct sk_buff *skb);
764void __kfree_skb(struct sk_buff *skb);
d7e8883c 765extern struct kmem_cache *skbuff_head_cache;
bad43ca8 766
7965bd4d
JP
767void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
768bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
769 bool *fragstolen, int *delta_truesize);
bad43ca8 770
7965bd4d
JP
771struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
772 int node);
773struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 774static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 775 gfp_t priority)
d179cd12 776{
564824b0 777 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
778}
779
2e4e4410
ED
780struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
781 unsigned long data_len,
782 int max_page_order,
783 int *errcode,
784 gfp_t gfp_mask);
785
d179cd12 786static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 787 gfp_t priority)
d179cd12 788{
c93bdd0e 789 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
790}
791
7965bd4d 792struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
793static inline struct sk_buff *alloc_skb_head(gfp_t priority)
794{
795 return __alloc_skb_head(priority, -1);
796}
797
7965bd4d
JP
798struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
799int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
800struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
801struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
802struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
803 gfp_t gfp_mask, bool fclone);
804static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
805 gfp_t gfp_mask)
806{
807 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
808}
7965bd4d
JP
809
810int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
811struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
812 unsigned int headroom);
813struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
814 int newtailroom, gfp_t priority);
25a91d8d
FD
815int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
816 int offset, int len);
7965bd4d
JP
817int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
818 int len);
819int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
820int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 821#define dev_kfree_skb(a) consume_skb(a)
1da177e4 822
7965bd4d
JP
823int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
824 int getfrag(void *from, char *to, int offset,
825 int len, int odd, struct sk_buff *skb),
826 void *from, int length);
e89e9cf5 827
d94d9fee 828struct skb_seq_state {
677e90ed
TG
829 __u32 lower_offset;
830 __u32 upper_offset;
831 __u32 frag_idx;
832 __u32 stepped_offset;
833 struct sk_buff *root_skb;
834 struct sk_buff *cur_skb;
835 __u8 *frag_data;
836};
837
7965bd4d
JP
838void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
839 unsigned int to, struct skb_seq_state *st);
840unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
841 struct skb_seq_state *st);
842void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 843
7965bd4d
JP
844unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
845 unsigned int to, struct ts_config *config,
846 struct ts_state *state);
3fc7e8a6 847
09323cc4
TH
848/*
849 * Packet hash types specify the type of hash in skb_set_hash.
850 *
851 * Hash types refer to the protocol layer addresses which are used to
852 * construct a packet's hash. The hashes are used to differentiate or identify
853 * flows of the protocol layer for the hash type. Hash types are either
854 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
855 *
856 * Properties of hashes:
857 *
858 * 1) Two packets in different flows have different hash values
859 * 2) Two packets in the same flow should have the same hash value
860 *
861 * A hash at a higher layer is considered to be more specific. A driver should
862 * set the most specific hash possible.
863 *
864 * A driver cannot indicate a more specific hash than the layer at which a hash
865 * was computed. For instance an L3 hash cannot be set as an L4 hash.
866 *
867 * A driver may indicate a hash level which is less specific than the
868 * actual layer the hash was computed on. For instance, a hash computed
869 * at L4 may be considered an L3 hash. This should only be done if the
870 * driver can't unambiguously determine that the HW computed the hash at
871 * the higher layer. Note that the "should" in the second property above
872 * permits this.
873 */
874enum pkt_hash_types {
875 PKT_HASH_TYPE_NONE, /* Undefined type */
876 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
877 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
878 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
879};
880
881static inline void
882skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
883{
61b905da 884 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 885 skb->sw_hash = 0;
61b905da 886 skb->hash = hash;
09323cc4
TH
887}
888
3958afa1
TH
889void __skb_get_hash(struct sk_buff *skb);
890static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 891{
a3b18ddb 892 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 893 __skb_get_hash(skb);
bfb564e7 894
61b905da 895 return skb->hash;
bfb564e7
KK
896}
897
57bdf7f4
TH
898static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
899{
61b905da 900 return skb->hash;
57bdf7f4
TH
901}
902
7539fadc
TH
903static inline void skb_clear_hash(struct sk_buff *skb)
904{
61b905da 905 skb->hash = 0;
a3b18ddb 906 skb->sw_hash = 0;
61b905da 907 skb->l4_hash = 0;
7539fadc
TH
908}
909
910static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
911{
61b905da 912 if (!skb->l4_hash)
7539fadc
TH
913 skb_clear_hash(skb);
914}
915
3df7a74e
TH
916static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
917{
61b905da 918 to->hash = from->hash;
a3b18ddb 919 to->sw_hash = from->sw_hash;
61b905da 920 to->l4_hash = from->l4_hash;
3df7a74e
TH
921};
922
4305b541
ACM
923#ifdef NET_SKBUFF_DATA_USES_OFFSET
924static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
925{
926 return skb->head + skb->end;
927}
ec47ea82
AD
928
929static inline unsigned int skb_end_offset(const struct sk_buff *skb)
930{
931 return skb->end;
932}
4305b541
ACM
933#else
934static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
935{
936 return skb->end;
937}
ec47ea82
AD
938
939static inline unsigned int skb_end_offset(const struct sk_buff *skb)
940{
941 return skb->end - skb->head;
942}
4305b541
ACM
943#endif
944
1da177e4 945/* Internal */
4305b541 946#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 947
ac45f602
PO
948static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
949{
950 return &skb_shinfo(skb)->hwtstamps;
951}
952
1da177e4
LT
953/**
954 * skb_queue_empty - check if a queue is empty
955 * @list: queue head
956 *
957 * Returns true if the queue is empty, false otherwise.
958 */
959static inline int skb_queue_empty(const struct sk_buff_head *list)
960{
fd44b93c 961 return list->next == (const struct sk_buff *) list;
1da177e4
LT
962}
963
fc7ebb21
DM
964/**
965 * skb_queue_is_last - check if skb is the last entry in the queue
966 * @list: queue head
967 * @skb: buffer
968 *
969 * Returns true if @skb is the last buffer on the list.
970 */
971static inline bool skb_queue_is_last(const struct sk_buff_head *list,
972 const struct sk_buff *skb)
973{
fd44b93c 974 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
975}
976
832d11c5
IJ
977/**
978 * skb_queue_is_first - check if skb is the first entry in the queue
979 * @list: queue head
980 * @skb: buffer
981 *
982 * Returns true if @skb is the first buffer on the list.
983 */
984static inline bool skb_queue_is_first(const struct sk_buff_head *list,
985 const struct sk_buff *skb)
986{
fd44b93c 987 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
988}
989
249c8b42
DM
990/**
991 * skb_queue_next - return the next packet in the queue
992 * @list: queue head
993 * @skb: current buffer
994 *
995 * Return the next packet in @list after @skb. It is only valid to
996 * call this if skb_queue_is_last() evaluates to false.
997 */
998static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
999 const struct sk_buff *skb)
1000{
1001 /* This BUG_ON may seem severe, but if we just return then we
1002 * are going to dereference garbage.
1003 */
1004 BUG_ON(skb_queue_is_last(list, skb));
1005 return skb->next;
1006}
1007
832d11c5
IJ
1008/**
1009 * skb_queue_prev - return the prev packet in the queue
1010 * @list: queue head
1011 * @skb: current buffer
1012 *
1013 * Return the prev packet in @list before @skb. It is only valid to
1014 * call this if skb_queue_is_first() evaluates to false.
1015 */
1016static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1017 const struct sk_buff *skb)
1018{
1019 /* This BUG_ON may seem severe, but if we just return then we
1020 * are going to dereference garbage.
1021 */
1022 BUG_ON(skb_queue_is_first(list, skb));
1023 return skb->prev;
1024}
1025
1da177e4
LT
1026/**
1027 * skb_get - reference buffer
1028 * @skb: buffer to reference
1029 *
1030 * Makes another reference to a socket buffer and returns a pointer
1031 * to the buffer.
1032 */
1033static inline struct sk_buff *skb_get(struct sk_buff *skb)
1034{
1035 atomic_inc(&skb->users);
1036 return skb;
1037}
1038
1039/*
1040 * If users == 1, we are the only owner and are can avoid redundant
1041 * atomic change.
1042 */
1043
1da177e4
LT
1044/**
1045 * skb_cloned - is the buffer a clone
1046 * @skb: buffer to check
1047 *
1048 * Returns true if the buffer was generated with skb_clone() and is
1049 * one of multiple shared copies of the buffer. Cloned buffers are
1050 * shared data so must not be written to under normal circumstances.
1051 */
1052static inline int skb_cloned(const struct sk_buff *skb)
1053{
1054 return skb->cloned &&
1055 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1056}
1057
14bbd6a5
PS
1058static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1059{
1060 might_sleep_if(pri & __GFP_WAIT);
1061
1062 if (skb_cloned(skb))
1063 return pskb_expand_head(skb, 0, 0, pri);
1064
1065 return 0;
1066}
1067
1da177e4
LT
1068/**
1069 * skb_header_cloned - is the header a clone
1070 * @skb: buffer to check
1071 *
1072 * Returns true if modifying the header part of the buffer requires
1073 * the data to be copied.
1074 */
1075static inline int skb_header_cloned(const struct sk_buff *skb)
1076{
1077 int dataref;
1078
1079 if (!skb->cloned)
1080 return 0;
1081
1082 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1083 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1084 return dataref != 1;
1085}
1086
1087/**
1088 * skb_header_release - release reference to header
1089 * @skb: buffer to operate on
1090 *
1091 * Drop a reference to the header part of the buffer. This is done
1092 * by acquiring a payload reference. You must not read from the header
1093 * part of skb->data after this.
f4a775d1 1094 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1095 */
1096static inline void skb_header_release(struct sk_buff *skb)
1097{
1098 BUG_ON(skb->nohdr);
1099 skb->nohdr = 1;
1100 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1101}
1102
f4a775d1
ED
1103/**
1104 * __skb_header_release - release reference to header
1105 * @skb: buffer to operate on
1106 *
1107 * Variant of skb_header_release() assuming skb is private to caller.
1108 * We can avoid one atomic operation.
1109 */
1110static inline void __skb_header_release(struct sk_buff *skb)
1111{
1112 skb->nohdr = 1;
1113 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1114}
1115
1116
1da177e4
LT
1117/**
1118 * skb_shared - is the buffer shared
1119 * @skb: buffer to check
1120 *
1121 * Returns true if more than one person has a reference to this
1122 * buffer.
1123 */
1124static inline int skb_shared(const struct sk_buff *skb)
1125{
1126 return atomic_read(&skb->users) != 1;
1127}
1128
1129/**
1130 * skb_share_check - check if buffer is shared and if so clone it
1131 * @skb: buffer to check
1132 * @pri: priority for memory allocation
1133 *
1134 * If the buffer is shared the buffer is cloned and the old copy
1135 * drops a reference. A new clone with a single reference is returned.
1136 * If the buffer is not shared the original buffer is returned. When
1137 * being called from interrupt status or with spinlocks held pri must
1138 * be GFP_ATOMIC.
1139 *
1140 * NULL is returned on a memory allocation failure.
1141 */
47061bc4 1142static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1143{
1144 might_sleep_if(pri & __GFP_WAIT);
1145 if (skb_shared(skb)) {
1146 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1147
1148 if (likely(nskb))
1149 consume_skb(skb);
1150 else
1151 kfree_skb(skb);
1da177e4
LT
1152 skb = nskb;
1153 }
1154 return skb;
1155}
1156
1157/*
1158 * Copy shared buffers into a new sk_buff. We effectively do COW on
1159 * packets to handle cases where we have a local reader and forward
1160 * and a couple of other messy ones. The normal one is tcpdumping
1161 * a packet thats being forwarded.
1162 */
1163
1164/**
1165 * skb_unshare - make a copy of a shared buffer
1166 * @skb: buffer to check
1167 * @pri: priority for memory allocation
1168 *
1169 * If the socket buffer is a clone then this function creates a new
1170 * copy of the data, drops a reference count on the old copy and returns
1171 * the new copy with the reference count at 1. If the buffer is not a clone
1172 * the original buffer is returned. When called with a spinlock held or
1173 * from interrupt state @pri must be %GFP_ATOMIC
1174 *
1175 * %NULL is returned on a memory allocation failure.
1176 */
e2bf521d 1177static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1178 gfp_t pri)
1da177e4
LT
1179{
1180 might_sleep_if(pri & __GFP_WAIT);
1181 if (skb_cloned(skb)) {
1182 struct sk_buff *nskb = skb_copy(skb, pri);
1183 kfree_skb(skb); /* Free our shared copy */
1184 skb = nskb;
1185 }
1186 return skb;
1187}
1188
1189/**
1a5778aa 1190 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1191 * @list_: list to peek at
1192 *
1193 * Peek an &sk_buff. Unlike most other operations you _MUST_
1194 * be careful with this one. A peek leaves the buffer on the
1195 * list and someone else may run off with it. You must hold
1196 * the appropriate locks or have a private queue to do this.
1197 *
1198 * Returns %NULL for an empty list or a pointer to the head element.
1199 * The reference count is not incremented and the reference is therefore
1200 * volatile. Use with caution.
1201 */
05bdd2f1 1202static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1203{
18d07000
ED
1204 struct sk_buff *skb = list_->next;
1205
1206 if (skb == (struct sk_buff *)list_)
1207 skb = NULL;
1208 return skb;
1da177e4
LT
1209}
1210
da5ef6e5
PE
1211/**
1212 * skb_peek_next - peek skb following the given one from a queue
1213 * @skb: skb to start from
1214 * @list_: list to peek at
1215 *
1216 * Returns %NULL when the end of the list is met or a pointer to the
1217 * next element. The reference count is not incremented and the
1218 * reference is therefore volatile. Use with caution.
1219 */
1220static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1221 const struct sk_buff_head *list_)
1222{
1223 struct sk_buff *next = skb->next;
18d07000 1224
da5ef6e5
PE
1225 if (next == (struct sk_buff *)list_)
1226 next = NULL;
1227 return next;
1228}
1229
1da177e4 1230/**
1a5778aa 1231 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1232 * @list_: list to peek at
1233 *
1234 * Peek an &sk_buff. Unlike most other operations you _MUST_
1235 * be careful with this one. A peek leaves the buffer on the
1236 * list and someone else may run off with it. You must hold
1237 * the appropriate locks or have a private queue to do this.
1238 *
1239 * Returns %NULL for an empty list or a pointer to the tail element.
1240 * The reference count is not incremented and the reference is therefore
1241 * volatile. Use with caution.
1242 */
05bdd2f1 1243static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1244{
18d07000
ED
1245 struct sk_buff *skb = list_->prev;
1246
1247 if (skb == (struct sk_buff *)list_)
1248 skb = NULL;
1249 return skb;
1250
1da177e4
LT
1251}
1252
1253/**
1254 * skb_queue_len - get queue length
1255 * @list_: list to measure
1256 *
1257 * Return the length of an &sk_buff queue.
1258 */
1259static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1260{
1261 return list_->qlen;
1262}
1263
67fed459
DM
1264/**
1265 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1266 * @list: queue to initialize
1267 *
1268 * This initializes only the list and queue length aspects of
1269 * an sk_buff_head object. This allows to initialize the list
1270 * aspects of an sk_buff_head without reinitializing things like
1271 * the spinlock. It can also be used for on-stack sk_buff_head
1272 * objects where the spinlock is known to not be used.
1273 */
1274static inline void __skb_queue_head_init(struct sk_buff_head *list)
1275{
1276 list->prev = list->next = (struct sk_buff *)list;
1277 list->qlen = 0;
1278}
1279
76f10ad0
AV
1280/*
1281 * This function creates a split out lock class for each invocation;
1282 * this is needed for now since a whole lot of users of the skb-queue
1283 * infrastructure in drivers have different locking usage (in hardirq)
1284 * than the networking core (in softirq only). In the long run either the
1285 * network layer or drivers should need annotation to consolidate the
1286 * main types of usage into 3 classes.
1287 */
1da177e4
LT
1288static inline void skb_queue_head_init(struct sk_buff_head *list)
1289{
1290 spin_lock_init(&list->lock);
67fed459 1291 __skb_queue_head_init(list);
1da177e4
LT
1292}
1293
c2ecba71
PE
1294static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1295 struct lock_class_key *class)
1296{
1297 skb_queue_head_init(list);
1298 lockdep_set_class(&list->lock, class);
1299}
1300
1da177e4 1301/*
bf299275 1302 * Insert an sk_buff on a list.
1da177e4
LT
1303 *
1304 * The "__skb_xxxx()" functions are the non-atomic ones that
1305 * can only be called with interrupts disabled.
1306 */
7965bd4d
JP
1307void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1308 struct sk_buff_head *list);
bf299275
GR
1309static inline void __skb_insert(struct sk_buff *newsk,
1310 struct sk_buff *prev, struct sk_buff *next,
1311 struct sk_buff_head *list)
1312{
1313 newsk->next = next;
1314 newsk->prev = prev;
1315 next->prev = prev->next = newsk;
1316 list->qlen++;
1317}
1da177e4 1318
67fed459
DM
1319static inline void __skb_queue_splice(const struct sk_buff_head *list,
1320 struct sk_buff *prev,
1321 struct sk_buff *next)
1322{
1323 struct sk_buff *first = list->next;
1324 struct sk_buff *last = list->prev;
1325
1326 first->prev = prev;
1327 prev->next = first;
1328
1329 last->next = next;
1330 next->prev = last;
1331}
1332
1333/**
1334 * skb_queue_splice - join two skb lists, this is designed for stacks
1335 * @list: the new list to add
1336 * @head: the place to add it in the first list
1337 */
1338static inline void skb_queue_splice(const struct sk_buff_head *list,
1339 struct sk_buff_head *head)
1340{
1341 if (!skb_queue_empty(list)) {
1342 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1343 head->qlen += list->qlen;
67fed459
DM
1344 }
1345}
1346
1347/**
d9619496 1348 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1349 * @list: the new list to add
1350 * @head: the place to add it in the first list
1351 *
1352 * The list at @list is reinitialised
1353 */
1354static inline void skb_queue_splice_init(struct sk_buff_head *list,
1355 struct sk_buff_head *head)
1356{
1357 if (!skb_queue_empty(list)) {
1358 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1359 head->qlen += list->qlen;
67fed459
DM
1360 __skb_queue_head_init(list);
1361 }
1362}
1363
1364/**
1365 * skb_queue_splice_tail - join two skb lists, each list being a queue
1366 * @list: the new list to add
1367 * @head: the place to add it in the first list
1368 */
1369static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1370 struct sk_buff_head *head)
1371{
1372 if (!skb_queue_empty(list)) {
1373 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1374 head->qlen += list->qlen;
67fed459
DM
1375 }
1376}
1377
1378/**
d9619496 1379 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1380 * @list: the new list to add
1381 * @head: the place to add it in the first list
1382 *
1383 * Each of the lists is a queue.
1384 * The list at @list is reinitialised
1385 */
1386static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1387 struct sk_buff_head *head)
1388{
1389 if (!skb_queue_empty(list)) {
1390 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1391 head->qlen += list->qlen;
67fed459
DM
1392 __skb_queue_head_init(list);
1393 }
1394}
1395
1da177e4 1396/**
300ce174 1397 * __skb_queue_after - queue a buffer at the list head
1da177e4 1398 * @list: list to use
300ce174 1399 * @prev: place after this buffer
1da177e4
LT
1400 * @newsk: buffer to queue
1401 *
300ce174 1402 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1403 * and you must therefore hold required locks before calling it.
1404 *
1405 * A buffer cannot be placed on two lists at the same time.
1406 */
300ce174
SH
1407static inline void __skb_queue_after(struct sk_buff_head *list,
1408 struct sk_buff *prev,
1409 struct sk_buff *newsk)
1da177e4 1410{
bf299275 1411 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1412}
1413
7965bd4d
JP
1414void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1415 struct sk_buff_head *list);
7de6c033 1416
f5572855
GR
1417static inline void __skb_queue_before(struct sk_buff_head *list,
1418 struct sk_buff *next,
1419 struct sk_buff *newsk)
1420{
1421 __skb_insert(newsk, next->prev, next, list);
1422}
1423
300ce174
SH
1424/**
1425 * __skb_queue_head - queue a buffer at the list head
1426 * @list: list to use
1427 * @newsk: buffer to queue
1428 *
1429 * Queue a buffer at the start of a list. This function takes no locks
1430 * and you must therefore hold required locks before calling it.
1431 *
1432 * A buffer cannot be placed on two lists at the same time.
1433 */
7965bd4d 1434void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1435static inline void __skb_queue_head(struct sk_buff_head *list,
1436 struct sk_buff *newsk)
1437{
1438 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1439}
1440
1da177e4
LT
1441/**
1442 * __skb_queue_tail - queue a buffer at the list tail
1443 * @list: list to use
1444 * @newsk: buffer to queue
1445 *
1446 * Queue a buffer at the end of a list. This function takes no locks
1447 * and you must therefore hold required locks before calling it.
1448 *
1449 * A buffer cannot be placed on two lists at the same time.
1450 */
7965bd4d 1451void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1452static inline void __skb_queue_tail(struct sk_buff_head *list,
1453 struct sk_buff *newsk)
1454{
f5572855 1455 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1456}
1457
1da177e4
LT
1458/*
1459 * remove sk_buff from list. _Must_ be called atomically, and with
1460 * the list known..
1461 */
7965bd4d 1462void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1463static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1464{
1465 struct sk_buff *next, *prev;
1466
1467 list->qlen--;
1468 next = skb->next;
1469 prev = skb->prev;
1470 skb->next = skb->prev = NULL;
1da177e4
LT
1471 next->prev = prev;
1472 prev->next = next;
1473}
1474
f525c06d
GR
1475/**
1476 * __skb_dequeue - remove from the head of the queue
1477 * @list: list to dequeue from
1478 *
1479 * Remove the head of the list. This function does not take any locks
1480 * so must be used with appropriate locks held only. The head item is
1481 * returned or %NULL if the list is empty.
1482 */
7965bd4d 1483struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1484static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1485{
1486 struct sk_buff *skb = skb_peek(list);
1487 if (skb)
1488 __skb_unlink(skb, list);
1489 return skb;
1490}
1da177e4
LT
1491
1492/**
1493 * __skb_dequeue_tail - remove from the tail of the queue
1494 * @list: list to dequeue from
1495 *
1496 * Remove the tail of the list. This function does not take any locks
1497 * so must be used with appropriate locks held only. The tail item is
1498 * returned or %NULL if the list is empty.
1499 */
7965bd4d 1500struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1501static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1502{
1503 struct sk_buff *skb = skb_peek_tail(list);
1504 if (skb)
1505 __skb_unlink(skb, list);
1506 return skb;
1507}
1508
1509
bdcc0924 1510static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1511{
1512 return skb->data_len;
1513}
1514
1515static inline unsigned int skb_headlen(const struct sk_buff *skb)
1516{
1517 return skb->len - skb->data_len;
1518}
1519
1520static inline int skb_pagelen(const struct sk_buff *skb)
1521{
1522 int i, len = 0;
1523
1524 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1525 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1526 return len + skb_headlen(skb);
1527}
1528
131ea667
IC
1529/**
1530 * __skb_fill_page_desc - initialise a paged fragment in an skb
1531 * @skb: buffer containing fragment to be initialised
1532 * @i: paged fragment index to initialise
1533 * @page: the page to use for this fragment
1534 * @off: the offset to the data with @page
1535 * @size: the length of the data
1536 *
1537 * Initialises the @i'th fragment of @skb to point to &size bytes at
1538 * offset @off within @page.
1539 *
1540 * Does not take any additional reference on the fragment.
1541 */
1542static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1543 struct page *page, int off, int size)
1da177e4
LT
1544{
1545 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1546
c48a11c7
MG
1547 /*
1548 * Propagate page->pfmemalloc to the skb if we can. The problem is
1549 * that not all callers have unique ownership of the page. If
1550 * pfmemalloc is set, we check the mapping as a mapping implies
1551 * page->index is set (index and pfmemalloc share space).
1552 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1553 * do not lose pfmemalloc information as the pages would not be
1554 * allocated using __GFP_MEMALLOC.
1555 */
a8605c60 1556 frag->page.p = page;
1da177e4 1557 frag->page_offset = off;
9e903e08 1558 skb_frag_size_set(frag, size);
cca7af38
PE
1559
1560 page = compound_head(page);
1561 if (page->pfmemalloc && !page->mapping)
1562 skb->pfmemalloc = true;
131ea667
IC
1563}
1564
1565/**
1566 * skb_fill_page_desc - initialise a paged fragment in an skb
1567 * @skb: buffer containing fragment to be initialised
1568 * @i: paged fragment index to initialise
1569 * @page: the page to use for this fragment
1570 * @off: the offset to the data with @page
1571 * @size: the length of the data
1572 *
1573 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1574 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1575 * addition updates @skb such that @i is the last fragment.
1576 *
1577 * Does not take any additional reference on the fragment.
1578 */
1579static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1580 struct page *page, int off, int size)
1581{
1582 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1583 skb_shinfo(skb)->nr_frags = i + 1;
1584}
1585
7965bd4d
JP
1586void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1587 int size, unsigned int truesize);
654bed16 1588
f8e617e1
JW
1589void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1590 unsigned int truesize);
1591
1da177e4 1592#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1593#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1594#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1595
27a884dc
ACM
1596#ifdef NET_SKBUFF_DATA_USES_OFFSET
1597static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1598{
1599 return skb->head + skb->tail;
1600}
1601
1602static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1603{
1604 skb->tail = skb->data - skb->head;
1605}
1606
1607static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1608{
1609 skb_reset_tail_pointer(skb);
1610 skb->tail += offset;
1611}
7cc46190 1612
27a884dc
ACM
1613#else /* NET_SKBUFF_DATA_USES_OFFSET */
1614static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1615{
1616 return skb->tail;
1617}
1618
1619static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1620{
1621 skb->tail = skb->data;
1622}
1623
1624static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1625{
1626 skb->tail = skb->data + offset;
1627}
4305b541 1628
27a884dc
ACM
1629#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1630
1da177e4
LT
1631/*
1632 * Add data to an sk_buff
1633 */
0c7ddf36 1634unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1635unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1636static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1637{
27a884dc 1638 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1639 SKB_LINEAR_ASSERT(skb);
1640 skb->tail += len;
1641 skb->len += len;
1642 return tmp;
1643}
1644
7965bd4d 1645unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1646static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1647{
1648 skb->data -= len;
1649 skb->len += len;
1650 return skb->data;
1651}
1652
7965bd4d 1653unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1654static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1655{
1656 skb->len -= len;
1657 BUG_ON(skb->len < skb->data_len);
1658 return skb->data += len;
1659}
1660
47d29646
DM
1661static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1662{
1663 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1664}
1665
7965bd4d 1666unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1667
1668static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1669{
1670 if (len > skb_headlen(skb) &&
987c402a 1671 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1672 return NULL;
1673 skb->len -= len;
1674 return skb->data += len;
1675}
1676
1677static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1678{
1679 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1680}
1681
1682static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1683{
1684 if (likely(len <= skb_headlen(skb)))
1685 return 1;
1686 if (unlikely(len > skb->len))
1687 return 0;
987c402a 1688 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1689}
1690
1691/**
1692 * skb_headroom - bytes at buffer head
1693 * @skb: buffer to check
1694 *
1695 * Return the number of bytes of free space at the head of an &sk_buff.
1696 */
c2636b4d 1697static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1698{
1699 return skb->data - skb->head;
1700}
1701
1702/**
1703 * skb_tailroom - bytes at buffer end
1704 * @skb: buffer to check
1705 *
1706 * Return the number of bytes of free space at the tail of an sk_buff
1707 */
1708static inline int skb_tailroom(const struct sk_buff *skb)
1709{
4305b541 1710 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1711}
1712
a21d4572
ED
1713/**
1714 * skb_availroom - bytes at buffer end
1715 * @skb: buffer to check
1716 *
1717 * Return the number of bytes of free space at the tail of an sk_buff
1718 * allocated by sk_stream_alloc()
1719 */
1720static inline int skb_availroom(const struct sk_buff *skb)
1721{
16fad69c
ED
1722 if (skb_is_nonlinear(skb))
1723 return 0;
1724
1725 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1726}
1727
1da177e4
LT
1728/**
1729 * skb_reserve - adjust headroom
1730 * @skb: buffer to alter
1731 * @len: bytes to move
1732 *
1733 * Increase the headroom of an empty &sk_buff by reducing the tail
1734 * room. This is only allowed for an empty buffer.
1735 */
8243126c 1736static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1737{
1738 skb->data += len;
1739 skb->tail += len;
1740}
1741
6a674e9c
JG
1742static inline void skb_reset_inner_headers(struct sk_buff *skb)
1743{
aefbd2b3 1744 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1745 skb->inner_network_header = skb->network_header;
1746 skb->inner_transport_header = skb->transport_header;
1747}
1748
0b5c9db1
JP
1749static inline void skb_reset_mac_len(struct sk_buff *skb)
1750{
1751 skb->mac_len = skb->network_header - skb->mac_header;
1752}
1753
6a674e9c
JG
1754static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1755 *skb)
1756{
1757 return skb->head + skb->inner_transport_header;
1758}
1759
1760static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1761{
1762 skb->inner_transport_header = skb->data - skb->head;
1763}
1764
1765static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1766 const int offset)
1767{
1768 skb_reset_inner_transport_header(skb);
1769 skb->inner_transport_header += offset;
1770}
1771
1772static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1773{
1774 return skb->head + skb->inner_network_header;
1775}
1776
1777static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1778{
1779 skb->inner_network_header = skb->data - skb->head;
1780}
1781
1782static inline void skb_set_inner_network_header(struct sk_buff *skb,
1783 const int offset)
1784{
1785 skb_reset_inner_network_header(skb);
1786 skb->inner_network_header += offset;
1787}
1788
aefbd2b3
PS
1789static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1790{
1791 return skb->head + skb->inner_mac_header;
1792}
1793
1794static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1795{
1796 skb->inner_mac_header = skb->data - skb->head;
1797}
1798
1799static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1800 const int offset)
1801{
1802 skb_reset_inner_mac_header(skb);
1803 skb->inner_mac_header += offset;
1804}
fda55eca
ED
1805static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1806{
35d04610 1807 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1808}
1809
9c70220b
ACM
1810static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1811{
2e07fa9c 1812 return skb->head + skb->transport_header;
9c70220b
ACM
1813}
1814
badff6d0
ACM
1815static inline void skb_reset_transport_header(struct sk_buff *skb)
1816{
2e07fa9c 1817 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1818}
1819
967b05f6
ACM
1820static inline void skb_set_transport_header(struct sk_buff *skb,
1821 const int offset)
1822{
2e07fa9c
ACM
1823 skb_reset_transport_header(skb);
1824 skb->transport_header += offset;
ea2ae17d
ACM
1825}
1826
d56f90a7
ACM
1827static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1828{
2e07fa9c 1829 return skb->head + skb->network_header;
d56f90a7
ACM
1830}
1831
c1d2bbe1
ACM
1832static inline void skb_reset_network_header(struct sk_buff *skb)
1833{
2e07fa9c 1834 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1835}
1836
c14d2450
ACM
1837static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1838{
2e07fa9c
ACM
1839 skb_reset_network_header(skb);
1840 skb->network_header += offset;
c14d2450
ACM
1841}
1842
2e07fa9c 1843static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1844{
2e07fa9c 1845 return skb->head + skb->mac_header;
bbe735e4
ACM
1846}
1847
2e07fa9c 1848static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1849{
35d04610 1850 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1851}
1852
1853static inline void skb_reset_mac_header(struct sk_buff *skb)
1854{
1855 skb->mac_header = skb->data - skb->head;
1856}
1857
1858static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1859{
1860 skb_reset_mac_header(skb);
1861 skb->mac_header += offset;
1862}
1863
0e3da5bb
TT
1864static inline void skb_pop_mac_header(struct sk_buff *skb)
1865{
1866 skb->mac_header = skb->network_header;
1867}
1868
fbbdb8f0
YX
1869static inline void skb_probe_transport_header(struct sk_buff *skb,
1870 const int offset_hint)
1871{
1872 struct flow_keys keys;
1873
1874 if (skb_transport_header_was_set(skb))
1875 return;
1876 else if (skb_flow_dissect(skb, &keys))
1877 skb_set_transport_header(skb, keys.thoff);
1878 else
1879 skb_set_transport_header(skb, offset_hint);
1880}
1881
03606895
ED
1882static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1883{
1884 if (skb_mac_header_was_set(skb)) {
1885 const unsigned char *old_mac = skb_mac_header(skb);
1886
1887 skb_set_mac_header(skb, -skb->mac_len);
1888 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1889 }
1890}
1891
04fb451e
MM
1892static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1893{
1894 return skb->csum_start - skb_headroom(skb);
1895}
1896
2e07fa9c
ACM
1897static inline int skb_transport_offset(const struct sk_buff *skb)
1898{
1899 return skb_transport_header(skb) - skb->data;
1900}
1901
1902static inline u32 skb_network_header_len(const struct sk_buff *skb)
1903{
1904 return skb->transport_header - skb->network_header;
1905}
1906
6a674e9c
JG
1907static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1908{
1909 return skb->inner_transport_header - skb->inner_network_header;
1910}
1911
2e07fa9c
ACM
1912static inline int skb_network_offset(const struct sk_buff *skb)
1913{
1914 return skb_network_header(skb) - skb->data;
1915}
48d49d0c 1916
6a674e9c
JG
1917static inline int skb_inner_network_offset(const struct sk_buff *skb)
1918{
1919 return skb_inner_network_header(skb) - skb->data;
1920}
1921
f9599ce1
CG
1922static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1923{
1924 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1925}
1926
1da177e4
LT
1927/*
1928 * CPUs often take a performance hit when accessing unaligned memory
1929 * locations. The actual performance hit varies, it can be small if the
1930 * hardware handles it or large if we have to take an exception and fix it
1931 * in software.
1932 *
1933 * Since an ethernet header is 14 bytes network drivers often end up with
1934 * the IP header at an unaligned offset. The IP header can be aligned by
1935 * shifting the start of the packet by 2 bytes. Drivers should do this
1936 * with:
1937 *
8660c124 1938 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1939 *
1940 * The downside to this alignment of the IP header is that the DMA is now
1941 * unaligned. On some architectures the cost of an unaligned DMA is high
1942 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1943 *
1da177e4
LT
1944 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1945 * to be overridden.
1946 */
1947#ifndef NET_IP_ALIGN
1948#define NET_IP_ALIGN 2
1949#endif
1950
025be81e
AB
1951/*
1952 * The networking layer reserves some headroom in skb data (via
1953 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1954 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1955 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1956 *
1957 * Unfortunately this headroom changes the DMA alignment of the resulting
1958 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1959 * on some architectures. An architecture can override this value,
1960 * perhaps setting it to a cacheline in size (since that will maintain
1961 * cacheline alignment of the DMA). It must be a power of 2.
1962 *
d6301d3d 1963 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1964 * headroom, you should not reduce this.
5933dd2f
ED
1965 *
1966 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1967 * to reduce average number of cache lines per packet.
1968 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1969 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1970 */
1971#ifndef NET_SKB_PAD
5933dd2f 1972#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1973#endif
1974
7965bd4d 1975int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1976
1977static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1978{
c4264f27 1979 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1980 WARN_ON(1);
1981 return;
1982 }
27a884dc
ACM
1983 skb->len = len;
1984 skb_set_tail_pointer(skb, len);
1da177e4
LT
1985}
1986
7965bd4d 1987void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1988
1989static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1990{
3cc0e873
HX
1991 if (skb->data_len)
1992 return ___pskb_trim(skb, len);
1993 __skb_trim(skb, len);
1994 return 0;
1da177e4
LT
1995}
1996
1997static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1998{
1999 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2000}
2001
e9fa4f7b
HX
2002/**
2003 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2004 * @skb: buffer to alter
2005 * @len: new length
2006 *
2007 * This is identical to pskb_trim except that the caller knows that
2008 * the skb is not cloned so we should never get an error due to out-
2009 * of-memory.
2010 */
2011static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2012{
2013 int err = pskb_trim(skb, len);
2014 BUG_ON(err);
2015}
2016
1da177e4
LT
2017/**
2018 * skb_orphan - orphan a buffer
2019 * @skb: buffer to orphan
2020 *
2021 * If a buffer currently has an owner then we call the owner's
2022 * destructor function and make the @skb unowned. The buffer continues
2023 * to exist but is no longer charged to its former owner.
2024 */
2025static inline void skb_orphan(struct sk_buff *skb)
2026{
c34a7612 2027 if (skb->destructor) {
1da177e4 2028 skb->destructor(skb);
c34a7612
ED
2029 skb->destructor = NULL;
2030 skb->sk = NULL;
376c7311
ED
2031 } else {
2032 BUG_ON(skb->sk);
c34a7612 2033 }
1da177e4
LT
2034}
2035
a353e0ce
MT
2036/**
2037 * skb_orphan_frags - orphan the frags contained in a buffer
2038 * @skb: buffer to orphan frags from
2039 * @gfp_mask: allocation mask for replacement pages
2040 *
2041 * For each frag in the SKB which needs a destructor (i.e. has an
2042 * owner) create a copy of that frag and release the original
2043 * page by calling the destructor.
2044 */
2045static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2046{
2047 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2048 return 0;
2049 return skb_copy_ubufs(skb, gfp_mask);
2050}
2051
1da177e4
LT
2052/**
2053 * __skb_queue_purge - empty a list
2054 * @list: list to empty
2055 *
2056 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2057 * the list and one reference dropped. This function does not take the
2058 * list lock and the caller must hold the relevant locks to use it.
2059 */
7965bd4d 2060void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2061static inline void __skb_queue_purge(struct sk_buff_head *list)
2062{
2063 struct sk_buff *skb;
2064 while ((skb = __skb_dequeue(list)) != NULL)
2065 kfree_skb(skb);
2066}
2067
e5e67305
AD
2068#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2069#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2070#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2071
7965bd4d 2072void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2073
7965bd4d
JP
2074struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2075 gfp_t gfp_mask);
8af27456
CH
2076
2077/**
2078 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2079 * @dev: network device to receive on
2080 * @length: length to allocate
2081 *
2082 * Allocate a new &sk_buff and assign it a usage count of one. The
2083 * buffer has unspecified headroom built in. Users should allocate
2084 * the headroom they think they need without accounting for the
2085 * built in space. The built in space is used for optimisations.
2086 *
2087 * %NULL is returned if there is no free memory. Although this function
2088 * allocates memory it can be called from an interrupt.
2089 */
2090static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2091 unsigned int length)
8af27456
CH
2092{
2093 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2094}
2095
6f532612
ED
2096/* legacy helper around __netdev_alloc_skb() */
2097static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2098 gfp_t gfp_mask)
2099{
2100 return __netdev_alloc_skb(NULL, length, gfp_mask);
2101}
2102
2103/* legacy helper around netdev_alloc_skb() */
2104static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2105{
2106 return netdev_alloc_skb(NULL, length);
2107}
2108
2109
4915a0de
ED
2110static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2111 unsigned int length, gfp_t gfp)
61321bbd 2112{
4915a0de 2113 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2114
2115 if (NET_IP_ALIGN && skb)
2116 skb_reserve(skb, NET_IP_ALIGN);
2117 return skb;
2118}
2119
4915a0de
ED
2120static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2121 unsigned int length)
2122{
2123 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2124}
2125
bc6fc9fa
FF
2126/**
2127 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2128 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2129 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2130 * @order: size of the allocation
2131 *
2132 * Allocate a new page.
2133 *
2134 * %NULL is returned if there is no free memory.
2135*/
2136static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2137 struct sk_buff *skb,
2138 unsigned int order)
2139{
2140 struct page *page;
2141
2142 gfp_mask |= __GFP_COLD;
2143
2144 if (!(gfp_mask & __GFP_NOMEMALLOC))
2145 gfp_mask |= __GFP_MEMALLOC;
2146
2147 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2148 if (skb && page && page->pfmemalloc)
2149 skb->pfmemalloc = true;
2150
2151 return page;
2152}
2153
2154/**
2155 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2156 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2157 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2158 *
2159 * Allocate a new page.
2160 *
2161 * %NULL is returned if there is no free memory.
2162 */
2163static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2164 struct sk_buff *skb)
2165{
2166 return __skb_alloc_pages(gfp_mask, skb, 0);
2167}
2168
2169/**
2170 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2171 * @page: The page that was allocated from skb_alloc_page
2172 * @skb: The skb that may need pfmemalloc set
2173 */
2174static inline void skb_propagate_pfmemalloc(struct page *page,
2175 struct sk_buff *skb)
2176{
2177 if (page && page->pfmemalloc)
2178 skb->pfmemalloc = true;
2179}
2180
131ea667 2181/**
e227867f 2182 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2183 * @frag: the paged fragment
2184 *
2185 * Returns the &struct page associated with @frag.
2186 */
2187static inline struct page *skb_frag_page(const skb_frag_t *frag)
2188{
a8605c60 2189 return frag->page.p;
131ea667
IC
2190}
2191
2192/**
2193 * __skb_frag_ref - take an addition reference on a paged fragment.
2194 * @frag: the paged fragment
2195 *
2196 * Takes an additional reference on the paged fragment @frag.
2197 */
2198static inline void __skb_frag_ref(skb_frag_t *frag)
2199{
2200 get_page(skb_frag_page(frag));
2201}
2202
2203/**
2204 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2205 * @skb: the buffer
2206 * @f: the fragment offset.
2207 *
2208 * Takes an additional reference on the @f'th paged fragment of @skb.
2209 */
2210static inline void skb_frag_ref(struct sk_buff *skb, int f)
2211{
2212 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2213}
2214
2215/**
2216 * __skb_frag_unref - release a reference on a paged fragment.
2217 * @frag: the paged fragment
2218 *
2219 * Releases a reference on the paged fragment @frag.
2220 */
2221static inline void __skb_frag_unref(skb_frag_t *frag)
2222{
2223 put_page(skb_frag_page(frag));
2224}
2225
2226/**
2227 * skb_frag_unref - release a reference on a paged fragment of an skb.
2228 * @skb: the buffer
2229 * @f: the fragment offset
2230 *
2231 * Releases a reference on the @f'th paged fragment of @skb.
2232 */
2233static inline void skb_frag_unref(struct sk_buff *skb, int f)
2234{
2235 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2236}
2237
2238/**
2239 * skb_frag_address - gets the address of the data contained in a paged fragment
2240 * @frag: the paged fragment buffer
2241 *
2242 * Returns the address of the data within @frag. The page must already
2243 * be mapped.
2244 */
2245static inline void *skb_frag_address(const skb_frag_t *frag)
2246{
2247 return page_address(skb_frag_page(frag)) + frag->page_offset;
2248}
2249
2250/**
2251 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2252 * @frag: the paged fragment buffer
2253 *
2254 * Returns the address of the data within @frag. Checks that the page
2255 * is mapped and returns %NULL otherwise.
2256 */
2257static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2258{
2259 void *ptr = page_address(skb_frag_page(frag));
2260 if (unlikely(!ptr))
2261 return NULL;
2262
2263 return ptr + frag->page_offset;
2264}
2265
2266/**
2267 * __skb_frag_set_page - sets the page contained in a paged fragment
2268 * @frag: the paged fragment
2269 * @page: the page to set
2270 *
2271 * Sets the fragment @frag to contain @page.
2272 */
2273static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2274{
a8605c60 2275 frag->page.p = page;
131ea667
IC
2276}
2277
2278/**
2279 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2280 * @skb: the buffer
2281 * @f: the fragment offset
2282 * @page: the page to set
2283 *
2284 * Sets the @f'th fragment of @skb to contain @page.
2285 */
2286static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2287 struct page *page)
2288{
2289 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2290}
2291
400dfd3a
ED
2292bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2293
131ea667
IC
2294/**
2295 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2296 * @dev: the device to map the fragment to
131ea667
IC
2297 * @frag: the paged fragment to map
2298 * @offset: the offset within the fragment (starting at the
2299 * fragment's own offset)
2300 * @size: the number of bytes to map
f83347df 2301 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2302 *
2303 * Maps the page associated with @frag to @device.
2304 */
2305static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2306 const skb_frag_t *frag,
2307 size_t offset, size_t size,
2308 enum dma_data_direction dir)
2309{
2310 return dma_map_page(dev, skb_frag_page(frag),
2311 frag->page_offset + offset, size, dir);
2312}
2313
117632e6
ED
2314static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2315 gfp_t gfp_mask)
2316{
2317 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2318}
2319
bad93e9d
OP
2320
2321static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2322 gfp_t gfp_mask)
2323{
2324 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2325}
2326
2327
334a8132
PM
2328/**
2329 * skb_clone_writable - is the header of a clone writable
2330 * @skb: buffer to check
2331 * @len: length up to which to write
2332 *
2333 * Returns true if modifying the header part of the cloned buffer
2334 * does not requires the data to be copied.
2335 */
05bdd2f1 2336static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2337{
2338 return !skb_header_cloned(skb) &&
2339 skb_headroom(skb) + len <= skb->hdr_len;
2340}
2341
d9cc2048
HX
2342static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2343 int cloned)
2344{
2345 int delta = 0;
2346
d9cc2048
HX
2347 if (headroom > skb_headroom(skb))
2348 delta = headroom - skb_headroom(skb);
2349
2350 if (delta || cloned)
2351 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2352 GFP_ATOMIC);
2353 return 0;
2354}
2355
1da177e4
LT
2356/**
2357 * skb_cow - copy header of skb when it is required
2358 * @skb: buffer to cow
2359 * @headroom: needed headroom
2360 *
2361 * If the skb passed lacks sufficient headroom or its data part
2362 * is shared, data is reallocated. If reallocation fails, an error
2363 * is returned and original skb is not changed.
2364 *
2365 * The result is skb with writable area skb->head...skb->tail
2366 * and at least @headroom of space at head.
2367 */
2368static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2369{
d9cc2048
HX
2370 return __skb_cow(skb, headroom, skb_cloned(skb));
2371}
1da177e4 2372
d9cc2048
HX
2373/**
2374 * skb_cow_head - skb_cow but only making the head writable
2375 * @skb: buffer to cow
2376 * @headroom: needed headroom
2377 *
2378 * This function is identical to skb_cow except that we replace the
2379 * skb_cloned check by skb_header_cloned. It should be used when
2380 * you only need to push on some header and do not need to modify
2381 * the data.
2382 */
2383static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2384{
2385 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2386}
2387
2388/**
2389 * skb_padto - pad an skbuff up to a minimal size
2390 * @skb: buffer to pad
2391 * @len: minimal length
2392 *
2393 * Pads up a buffer to ensure the trailing bytes exist and are
2394 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2395 * is untouched. Otherwise it is extended. Returns zero on
2396 * success. The skb is freed on error.
1da177e4
LT
2397 */
2398
5b057c6b 2399static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2400{
2401 unsigned int size = skb->len;
2402 if (likely(size >= len))
5b057c6b 2403 return 0;
987c402a 2404 return skb_pad(skb, len - size);
1da177e4
LT
2405}
2406
2407static inline int skb_add_data(struct sk_buff *skb,
2408 char __user *from, int copy)
2409{
2410 const int off = skb->len;
2411
2412 if (skb->ip_summed == CHECKSUM_NONE) {
2413 int err = 0;
5084205f 2414 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2415 copy, 0, &err);
2416 if (!err) {
2417 skb->csum = csum_block_add(skb->csum, csum, off);
2418 return 0;
2419 }
2420 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2421 return 0;
2422
2423 __skb_trim(skb, off);
2424 return -EFAULT;
2425}
2426
38ba0a65
ED
2427static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2428 const struct page *page, int off)
1da177e4
LT
2429{
2430 if (i) {
9e903e08 2431 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2432
ea2ab693 2433 return page == skb_frag_page(frag) &&
9e903e08 2434 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2435 }
38ba0a65 2436 return false;
1da177e4
LT
2437}
2438
364c6bad
HX
2439static inline int __skb_linearize(struct sk_buff *skb)
2440{
2441 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2442}
2443
1da177e4
LT
2444/**
2445 * skb_linearize - convert paged skb to linear one
2446 * @skb: buffer to linarize
1da177e4
LT
2447 *
2448 * If there is no free memory -ENOMEM is returned, otherwise zero
2449 * is returned and the old skb data released.
2450 */
364c6bad
HX
2451static inline int skb_linearize(struct sk_buff *skb)
2452{
2453 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2454}
2455
cef401de
ED
2456/**
2457 * skb_has_shared_frag - can any frag be overwritten
2458 * @skb: buffer to test
2459 *
2460 * Return true if the skb has at least one frag that might be modified
2461 * by an external entity (as in vmsplice()/sendfile())
2462 */
2463static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2464{
c9af6db4
PS
2465 return skb_is_nonlinear(skb) &&
2466 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2467}
2468
364c6bad
HX
2469/**
2470 * skb_linearize_cow - make sure skb is linear and writable
2471 * @skb: buffer to process
2472 *
2473 * If there is no free memory -ENOMEM is returned, otherwise zero
2474 * is returned and the old skb data released.
2475 */
2476static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2477{
364c6bad
HX
2478 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2479 __skb_linearize(skb) : 0;
1da177e4
LT
2480}
2481
2482/**
2483 * skb_postpull_rcsum - update checksum for received skb after pull
2484 * @skb: buffer to update
2485 * @start: start of data before pull
2486 * @len: length of data pulled
2487 *
2488 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2489 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2490 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2491 */
2492
2493static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2494 const void *start, unsigned int len)
1da177e4 2495{
84fa7933 2496 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2497 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2498}
2499
cbb042f9
HX
2500unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2501
7ce5a27f
DM
2502/**
2503 * pskb_trim_rcsum - trim received skb and update checksum
2504 * @skb: buffer to trim
2505 * @len: new length
2506 *
2507 * This is exactly the same as pskb_trim except that it ensures the
2508 * checksum of received packets are still valid after the operation.
2509 */
2510
2511static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2512{
2513 if (likely(len >= skb->len))
2514 return 0;
2515 if (skb->ip_summed == CHECKSUM_COMPLETE)
2516 skb->ip_summed = CHECKSUM_NONE;
2517 return __pskb_trim(skb, len);
2518}
2519
1da177e4
LT
2520#define skb_queue_walk(queue, skb) \
2521 for (skb = (queue)->next; \
a1e4891f 2522 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2523 skb = skb->next)
2524
46f8914e
JC
2525#define skb_queue_walk_safe(queue, skb, tmp) \
2526 for (skb = (queue)->next, tmp = skb->next; \
2527 skb != (struct sk_buff *)(queue); \
2528 skb = tmp, tmp = skb->next)
2529
1164f52a 2530#define skb_queue_walk_from(queue, skb) \
a1e4891f 2531 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2532 skb = skb->next)
2533
2534#define skb_queue_walk_from_safe(queue, skb, tmp) \
2535 for (tmp = skb->next; \
2536 skb != (struct sk_buff *)(queue); \
2537 skb = tmp, tmp = skb->next)
2538
300ce174
SH
2539#define skb_queue_reverse_walk(queue, skb) \
2540 for (skb = (queue)->prev; \
a1e4891f 2541 skb != (struct sk_buff *)(queue); \
300ce174
SH
2542 skb = skb->prev)
2543
686a2955
DM
2544#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2545 for (skb = (queue)->prev, tmp = skb->prev; \
2546 skb != (struct sk_buff *)(queue); \
2547 skb = tmp, tmp = skb->prev)
2548
2549#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2550 for (tmp = skb->prev; \
2551 skb != (struct sk_buff *)(queue); \
2552 skb = tmp, tmp = skb->prev)
1da177e4 2553
21dc3301 2554static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2555{
2556 return skb_shinfo(skb)->frag_list != NULL;
2557}
2558
2559static inline void skb_frag_list_init(struct sk_buff *skb)
2560{
2561 skb_shinfo(skb)->frag_list = NULL;
2562}
2563
2564static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2565{
2566 frag->next = skb_shinfo(skb)->frag_list;
2567 skb_shinfo(skb)->frag_list = frag;
2568}
2569
2570#define skb_walk_frags(skb, iter) \
2571 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2572
7965bd4d
JP
2573struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2574 int *peeked, int *off, int *err);
2575struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2576 int *err);
2577unsigned int datagram_poll(struct file *file, struct socket *sock,
2578 struct poll_table_struct *wait);
2579int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2580 struct iovec *to, int size);
2581int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2582 struct iovec *iov);
2583int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2584 const struct iovec *from, int from_offset,
2585 int len);
2586int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2587 int offset, size_t count);
2588int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2589 const struct iovec *to, int to_offset,
2590 int size);
2591void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2592void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2593int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2594int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2595int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2596__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2597 int len, __wsum csum);
2598int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2599 struct pipe_inode_info *pipe, unsigned int len,
2600 unsigned int flags);
2601void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2602unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2603int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2604 int len, int hlen);
7965bd4d
JP
2605void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2606int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2607void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2608unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2609struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2610struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2611
2817a336
DB
2612struct skb_checksum_ops {
2613 __wsum (*update)(const void *mem, int len, __wsum wsum);
2614 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2615};
2616
2617__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2618 __wsum csum, const struct skb_checksum_ops *ops);
2619__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2620 __wsum csum);
2621
690e36e7
DM
2622static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2623 int len, void *data, int hlen, void *buffer)
1da177e4 2624{
55820ee2 2625 if (hlen - offset >= len)
690e36e7 2626 return data + offset;
1da177e4 2627
690e36e7
DM
2628 if (!skb ||
2629 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2630 return NULL;
2631
2632 return buffer;
2633}
2634
690e36e7
DM
2635static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2636 int len, void *buffer)
2637{
2638 return __skb_header_pointer(skb, offset, len, skb->data,
2639 skb_headlen(skb), buffer);
2640}
2641
4262e5cc
DB
2642/**
2643 * skb_needs_linearize - check if we need to linearize a given skb
2644 * depending on the given device features.
2645 * @skb: socket buffer to check
2646 * @features: net device features
2647 *
2648 * Returns true if either:
2649 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2650 * 2. skb is fragmented and the device does not support SG.
2651 */
2652static inline bool skb_needs_linearize(struct sk_buff *skb,
2653 netdev_features_t features)
2654{
2655 return skb_is_nonlinear(skb) &&
2656 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2657 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2658}
2659
d626f62b
ACM
2660static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2661 void *to,
2662 const unsigned int len)
2663{
2664 memcpy(to, skb->data, len);
2665}
2666
2667static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2668 const int offset, void *to,
2669 const unsigned int len)
2670{
2671 memcpy(to, skb->data + offset, len);
2672}
2673
27d7ff46
ACM
2674static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2675 const void *from,
2676 const unsigned int len)
2677{
2678 memcpy(skb->data, from, len);
2679}
2680
2681static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2682 const int offset,
2683 const void *from,
2684 const unsigned int len)
2685{
2686 memcpy(skb->data + offset, from, len);
2687}
2688
7965bd4d 2689void skb_init(void);
1da177e4 2690
ac45f602
PO
2691static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2692{
2693 return skb->tstamp;
2694}
2695
a61bbcf2
PM
2696/**
2697 * skb_get_timestamp - get timestamp from a skb
2698 * @skb: skb to get stamp from
2699 * @stamp: pointer to struct timeval to store stamp in
2700 *
2701 * Timestamps are stored in the skb as offsets to a base timestamp.
2702 * This function converts the offset back to a struct timeval and stores
2703 * it in stamp.
2704 */
ac45f602
PO
2705static inline void skb_get_timestamp(const struct sk_buff *skb,
2706 struct timeval *stamp)
a61bbcf2 2707{
b7aa0bf7 2708 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2709}
2710
ac45f602
PO
2711static inline void skb_get_timestampns(const struct sk_buff *skb,
2712 struct timespec *stamp)
2713{
2714 *stamp = ktime_to_timespec(skb->tstamp);
2715}
2716
b7aa0bf7 2717static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2718{
b7aa0bf7 2719 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2720}
2721
164891aa
SH
2722static inline ktime_t net_timedelta(ktime_t t)
2723{
2724 return ktime_sub(ktime_get_real(), t);
2725}
2726
b9ce204f
IJ
2727static inline ktime_t net_invalid_timestamp(void)
2728{
2729 return ktime_set(0, 0);
2730}
a61bbcf2 2731
62bccb8c
AD
2732struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2733
c1f19b51
RC
2734#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2735
7965bd4d
JP
2736void skb_clone_tx_timestamp(struct sk_buff *skb);
2737bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2738
2739#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2740
2741static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2742{
2743}
2744
2745static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2746{
2747 return false;
2748}
2749
2750#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2751
2752/**
2753 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2754 *
da92b194
RC
2755 * PHY drivers may accept clones of transmitted packets for
2756 * timestamping via their phy_driver.txtstamp method. These drivers
2757 * must call this function to return the skb back to the stack, with
2758 * or without a timestamp.
2759 *
c1f19b51 2760 * @skb: clone of the the original outgoing packet
da92b194 2761 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2762 *
2763 */
2764void skb_complete_tx_timestamp(struct sk_buff *skb,
2765 struct skb_shared_hwtstamps *hwtstamps);
2766
e7fd2885
WB
2767void __skb_tstamp_tx(struct sk_buff *orig_skb,
2768 struct skb_shared_hwtstamps *hwtstamps,
2769 struct sock *sk, int tstype);
2770
ac45f602
PO
2771/**
2772 * skb_tstamp_tx - queue clone of skb with send time stamps
2773 * @orig_skb: the original outgoing packet
2774 * @hwtstamps: hardware time stamps, may be NULL if not available
2775 *
2776 * If the skb has a socket associated, then this function clones the
2777 * skb (thus sharing the actual data and optional structures), stores
2778 * the optional hardware time stamping information (if non NULL) or
2779 * generates a software time stamp (otherwise), then queues the clone
2780 * to the error queue of the socket. Errors are silently ignored.
2781 */
7965bd4d
JP
2782void skb_tstamp_tx(struct sk_buff *orig_skb,
2783 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2784
4507a715
RC
2785static inline void sw_tx_timestamp(struct sk_buff *skb)
2786{
2244d07b
OH
2787 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2788 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2789 skb_tstamp_tx(skb, NULL);
2790}
2791
2792/**
2793 * skb_tx_timestamp() - Driver hook for transmit timestamping
2794 *
2795 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2796 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2797 *
73409f3b
DM
2798 * Specifically, one should make absolutely sure that this function is
2799 * called before TX completion of this packet can trigger. Otherwise
2800 * the packet could potentially already be freed.
2801 *
4507a715
RC
2802 * @skb: A socket buffer.
2803 */
2804static inline void skb_tx_timestamp(struct sk_buff *skb)
2805{
c1f19b51 2806 skb_clone_tx_timestamp(skb);
4507a715
RC
2807 sw_tx_timestamp(skb);
2808}
2809
6e3e939f
JB
2810/**
2811 * skb_complete_wifi_ack - deliver skb with wifi status
2812 *
2813 * @skb: the original outgoing packet
2814 * @acked: ack status
2815 *
2816 */
2817void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2818
7965bd4d
JP
2819__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2820__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2821
60476372
HX
2822static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2823{
5d0c2b95 2824 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2825}
2826
fb286bb2
HX
2827/**
2828 * skb_checksum_complete - Calculate checksum of an entire packet
2829 * @skb: packet to process
2830 *
2831 * This function calculates the checksum over the entire packet plus
2832 * the value of skb->csum. The latter can be used to supply the
2833 * checksum of a pseudo header as used by TCP/UDP. It returns the
2834 * checksum.
2835 *
2836 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2837 * this function can be used to verify that checksum on received
2838 * packets. In that case the function should return zero if the
2839 * checksum is correct. In particular, this function will return zero
2840 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2841 * hardware has already verified the correctness of the checksum.
2842 */
4381ca3c 2843static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2844{
60476372
HX
2845 return skb_csum_unnecessary(skb) ?
2846 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2847}
2848
77cffe23
TH
2849static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2850{
2851 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2852 if (skb->csum_level == 0)
2853 skb->ip_summed = CHECKSUM_NONE;
2854 else
2855 skb->csum_level--;
2856 }
2857}
2858
2859static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2860{
2861 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2862 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2863 skb->csum_level++;
2864 } else if (skb->ip_summed == CHECKSUM_NONE) {
2865 skb->ip_summed = CHECKSUM_UNNECESSARY;
2866 skb->csum_level = 0;
2867 }
2868}
2869
5a212329
TH
2870static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2871{
2872 /* Mark current checksum as bad (typically called from GRO
2873 * path). In the case that ip_summed is CHECKSUM_NONE
2874 * this must be the first checksum encountered in the packet.
2875 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2876 * checksum after the last one validated. For UDP, a zero
2877 * checksum can not be marked as bad.
2878 */
2879
2880 if (skb->ip_summed == CHECKSUM_NONE ||
2881 skb->ip_summed == CHECKSUM_UNNECESSARY)
2882 skb->csum_bad = 1;
2883}
2884
76ba0aae
TH
2885/* Check if we need to perform checksum complete validation.
2886 *
2887 * Returns true if checksum complete is needed, false otherwise
2888 * (either checksum is unnecessary or zero checksum is allowed).
2889 */
2890static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2891 bool zero_okay,
2892 __sum16 check)
2893{
5d0c2b95
TH
2894 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2895 skb->csum_valid = 1;
77cffe23 2896 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2897 return false;
2898 }
2899
2900 return true;
2901}
2902
2903/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2904 * in checksum_init.
2905 */
2906#define CHECKSUM_BREAK 76
2907
2908/* Validate (init) checksum based on checksum complete.
2909 *
2910 * Return values:
2911 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2912 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2913 * checksum is stored in skb->csum for use in __skb_checksum_complete
2914 * non-zero: value of invalid checksum
2915 *
2916 */
2917static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2918 bool complete,
2919 __wsum psum)
2920{
2921 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2922 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2923 skb->csum_valid = 1;
76ba0aae
TH
2924 return 0;
2925 }
5a212329
TH
2926 } else if (skb->csum_bad) {
2927 /* ip_summed == CHECKSUM_NONE in this case */
2928 return 1;
76ba0aae
TH
2929 }
2930
2931 skb->csum = psum;
2932
5d0c2b95
TH
2933 if (complete || skb->len <= CHECKSUM_BREAK) {
2934 __sum16 csum;
2935
2936 csum = __skb_checksum_complete(skb);
2937 skb->csum_valid = !csum;
2938 return csum;
2939 }
76ba0aae
TH
2940
2941 return 0;
2942}
2943
2944static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2945{
2946 return 0;
2947}
2948
2949/* Perform checksum validate (init). Note that this is a macro since we only
2950 * want to calculate the pseudo header which is an input function if necessary.
2951 * First we try to validate without any computation (checksum unnecessary) and
2952 * then calculate based on checksum complete calling the function to compute
2953 * pseudo header.
2954 *
2955 * Return values:
2956 * 0: checksum is validated or try to in skb_checksum_complete
2957 * non-zero: value of invalid checksum
2958 */
2959#define __skb_checksum_validate(skb, proto, complete, \
2960 zero_okay, check, compute_pseudo) \
2961({ \
2962 __sum16 __ret = 0; \
5d0c2b95 2963 skb->csum_valid = 0; \
76ba0aae
TH
2964 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2965 __ret = __skb_checksum_validate_complete(skb, \
2966 complete, compute_pseudo(skb, proto)); \
2967 __ret; \
2968})
2969
2970#define skb_checksum_init(skb, proto, compute_pseudo) \
2971 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2972
2973#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2974 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2975
2976#define skb_checksum_validate(skb, proto, compute_pseudo) \
2977 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2978
2979#define skb_checksum_validate_zero_check(skb, proto, check, \
2980 compute_pseudo) \
2981 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2982
2983#define skb_checksum_simple_validate(skb) \
2984 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2985
d96535a1
TH
2986static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
2987{
2988 return (skb->ip_summed == CHECKSUM_NONE &&
2989 skb->csum_valid && !skb->csum_bad);
2990}
2991
2992static inline void __skb_checksum_convert(struct sk_buff *skb,
2993 __sum16 check, __wsum pseudo)
2994{
2995 skb->csum = ~pseudo;
2996 skb->ip_summed = CHECKSUM_COMPLETE;
2997}
2998
2999#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3000do { \
3001 if (__skb_checksum_convert_check(skb)) \
3002 __skb_checksum_convert(skb, check, \
3003 compute_pseudo(skb, proto)); \
3004} while (0)
3005
5f79e0f9 3006#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3007void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3008static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3009{
3010 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3011 nf_conntrack_destroy(nfct);
1da177e4
LT
3012}
3013static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3014{
3015 if (nfct)
3016 atomic_inc(&nfct->use);
3017}
2fc72c7b 3018#endif
1da177e4
LT
3019#ifdef CONFIG_BRIDGE_NETFILTER
3020static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3021{
3022 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3023 kfree(nf_bridge);
3024}
3025static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3026{
3027 if (nf_bridge)
3028 atomic_inc(&nf_bridge->use);
3029}
3030#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3031static inline void nf_reset(struct sk_buff *skb)
3032{
5f79e0f9 3033#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3034 nf_conntrack_put(skb->nfct);
3035 skb->nfct = NULL;
2fc72c7b 3036#endif
a193a4ab
PM
3037#ifdef CONFIG_BRIDGE_NETFILTER
3038 nf_bridge_put(skb->nf_bridge);
3039 skb->nf_bridge = NULL;
3040#endif
3041}
3042
124dff01
PM
3043static inline void nf_reset_trace(struct sk_buff *skb)
3044{
478b360a 3045#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3046 skb->nf_trace = 0;
3047#endif
a193a4ab
PM
3048}
3049
edda553c 3050/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3051static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3052 bool copy)
edda553c 3053{
5f79e0f9 3054#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3055 dst->nfct = src->nfct;
3056 nf_conntrack_get(src->nfct);
b1937227
ED
3057 if (copy)
3058 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3059#endif
edda553c
YK
3060#ifdef CONFIG_BRIDGE_NETFILTER
3061 dst->nf_bridge = src->nf_bridge;
3062 nf_bridge_get(src->nf_bridge);
3063#endif
478b360a 3064#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3065 if (copy)
3066 dst->nf_trace = src->nf_trace;
478b360a 3067#endif
edda553c
YK
3068}
3069
e7ac05f3
YK
3070static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3071{
e7ac05f3 3072#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3073 nf_conntrack_put(dst->nfct);
2fc72c7b 3074#endif
e7ac05f3
YK
3075#ifdef CONFIG_BRIDGE_NETFILTER
3076 nf_bridge_put(dst->nf_bridge);
3077#endif
b1937227 3078 __nf_copy(dst, src, true);
e7ac05f3
YK
3079}
3080
984bc16c
JM
3081#ifdef CONFIG_NETWORK_SECMARK
3082static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3083{
3084 to->secmark = from->secmark;
3085}
3086
3087static inline void skb_init_secmark(struct sk_buff *skb)
3088{
3089 skb->secmark = 0;
3090}
3091#else
3092static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3093{ }
3094
3095static inline void skb_init_secmark(struct sk_buff *skb)
3096{ }
3097#endif
3098
574f7194
EB
3099static inline bool skb_irq_freeable(const struct sk_buff *skb)
3100{
3101 return !skb->destructor &&
3102#if IS_ENABLED(CONFIG_XFRM)
3103 !skb->sp &&
3104#endif
3105#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3106 !skb->nfct &&
3107#endif
3108 !skb->_skb_refdst &&
3109 !skb_has_frag_list(skb);
3110}
3111
f25f4e44
PWJ
3112static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3113{
f25f4e44 3114 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3115}
3116
9247744e 3117static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3118{
4e3ab47a 3119 return skb->queue_mapping;
4e3ab47a
PE
3120}
3121
f25f4e44
PWJ
3122static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3123{
f25f4e44 3124 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3125}
3126
d5a9e24a
DM
3127static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3128{
3129 skb->queue_mapping = rx_queue + 1;
3130}
3131
9247744e 3132static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3133{
3134 return skb->queue_mapping - 1;
3135}
3136
9247744e 3137static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3138{
a02cec21 3139 return skb->queue_mapping != 0;
d5a9e24a
DM
3140}
3141
0e001614 3142u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3143 unsigned int num_tx_queues);
9247744e 3144
def8b4fa
AD
3145static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3146{
0b3d8e08 3147#ifdef CONFIG_XFRM
def8b4fa 3148 return skb->sp;
def8b4fa 3149#else
def8b4fa 3150 return NULL;
def8b4fa 3151#endif
0b3d8e08 3152}
def8b4fa 3153
68c33163
PS
3154/* Keeps track of mac header offset relative to skb->head.
3155 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3156 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3157 * tunnel skb it points to outer mac header.
3158 * Keeps track of level of encapsulation of network headers.
3159 */
68c33163 3160struct skb_gso_cb {
3347c960
ED
3161 int mac_offset;
3162 int encap_level;
7e2b10c1 3163 __u16 csum_start;
68c33163
PS
3164};
3165#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3166
3167static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3168{
3169 return (skb_mac_header(inner_skb) - inner_skb->head) -
3170 SKB_GSO_CB(inner_skb)->mac_offset;
3171}
3172
1e2bd517
PS
3173static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3174{
3175 int new_headroom, headroom;
3176 int ret;
3177
3178 headroom = skb_headroom(skb);
3179 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3180 if (ret)
3181 return ret;
3182
3183 new_headroom = skb_headroom(skb);
3184 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3185 return 0;
3186}
3187
7e2b10c1
TH
3188/* Compute the checksum for a gso segment. First compute the checksum value
3189 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3190 * then add in skb->csum (checksum from csum_start to end of packet).
3191 * skb->csum and csum_start are then updated to reflect the checksum of the
3192 * resultant packet starting from the transport header-- the resultant checksum
3193 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3194 * header.
3195 */
3196static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3197{
3198 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3199 skb_transport_offset(skb);
3200 __u16 csum;
3201
3202 csum = csum_fold(csum_partial(skb_transport_header(skb),
3203 plen, skb->csum));
3204 skb->csum = res;
3205 SKB_GSO_CB(skb)->csum_start -= plen;
3206
3207 return csum;
3208}
3209
bdcc0924 3210static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3211{
3212 return skb_shinfo(skb)->gso_size;
3213}
3214
36a8f39e 3215/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3216static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3217{
3218 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3219}
3220
7965bd4d 3221void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3222
3223static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3224{
3225 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3226 * wanted then gso_type will be set. */
05bdd2f1
ED
3227 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3228
b78462eb
AD
3229 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3230 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3231 __skb_warn_lro_forwarding(skb);
3232 return true;
3233 }
3234 return false;
3235}
3236
35fc92a9
HX
3237static inline void skb_forward_csum(struct sk_buff *skb)
3238{
3239 /* Unfortunately we don't support this one. Any brave souls? */
3240 if (skb->ip_summed == CHECKSUM_COMPLETE)
3241 skb->ip_summed = CHECKSUM_NONE;
3242}
3243
bc8acf2c
ED
3244/**
3245 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3246 * @skb: skb to check
3247 *
3248 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3249 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3250 * use this helper, to document places where we make this assertion.
3251 */
05bdd2f1 3252static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3253{
3254#ifdef DEBUG
3255 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3256#endif
3257}
3258
f35d9d8a 3259bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3260
ed1f50c3
PD
3261int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3262
56193d1b
AD
3263u32 skb_get_poff(const struct sk_buff *skb);
3264u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3265 const struct flow_keys *keys, int hlen);
f77668dc 3266
3a7c1ee4
AD
3267/**
3268 * skb_head_is_locked - Determine if the skb->head is locked down
3269 * @skb: skb to check
3270 *
3271 * The head on skbs build around a head frag can be removed if they are
3272 * not cloned. This function returns true if the skb head is locked down
3273 * due to either being allocated via kmalloc, or by being a clone with
3274 * multiple references to the head.
3275 */
3276static inline bool skb_head_is_locked(const struct sk_buff *skb)
3277{
3278 return !skb->head_frag || skb_cloned(skb);
3279}
fe6cc55f
FW
3280
3281/**
3282 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3283 *
3284 * @skb: GSO skb
3285 *
3286 * skb_gso_network_seglen is used to determine the real size of the
3287 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3288 *
3289 * The MAC/L2 header is not accounted for.
3290 */
3291static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3292{
3293 unsigned int hdr_len = skb_transport_header(skb) -
3294 skb_network_header(skb);
3295 return hdr_len + skb_gso_transport_seglen(skb);
3296}
1da177e4
LT
3297#endif /* __KERNEL__ */
3298#endif /* _LINUX_SKBUFF_H */