phy: micrel: add of configuration for LED mode
[linux-2.6-block.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4
LT
22#include <linux/cache.h>
23
60063497 24#include <linux/atomic.h>
1da177e4
LT
25#include <asm/types.h>
26#include <linux/spinlock.h>
1da177e4 27#include <linux/net.h>
3fc7e8a6 28#include <linux/textsearch.h>
1da177e4 29#include <net/checksum.h>
a80958f4 30#include <linux/rcupdate.h>
97fc2f08 31#include <linux/dmaengine.h>
b7aa0bf7 32#include <linux/hrtimer.h>
131ea667 33#include <linux/dma-mapping.h>
c8f44aff 34#include <linux/netdev_features.h>
5203cd28 35#include <net/flow_keys.h>
1da177e4 36
78ea85f1
DB
37/* A. Checksumming of received packets by device.
38 *
39 * CHECKSUM_NONE:
40 *
41 * Device failed to checksum this packet e.g. due to lack of capabilities.
42 * The packet contains full (though not verified) checksum in packet but
43 * not in skb->csum. Thus, skb->csum is undefined in this case.
44 *
45 * CHECKSUM_UNNECESSARY:
46 *
47 * The hardware you're dealing with doesn't calculate the full checksum
48 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
49 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
50 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
51 * undefined in this case though. It is a bad option, but, unfortunately,
52 * nowadays most vendors do this. Apparently with the secret goal to sell
53 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
54 *
55 * CHECKSUM_COMPLETE:
56 *
57 * This is the most generic way. The device supplied checksum of the _whole_
58 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
59 * hardware doesn't need to parse L3/L4 headers to implement this.
60 *
61 * Note: Even if device supports only some protocols, but is able to produce
62 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
63 *
64 * CHECKSUM_PARTIAL:
65 *
66 * This is identical to the case for output below. This may occur on a packet
67 * received directly from another Linux OS, e.g., a virtualized Linux kernel
68 * on the same host. The packet can be treated in the same way as
69 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
70 * checksum must be filled in by the OS or the hardware.
71 *
72 * B. Checksumming on output.
73 *
74 * CHECKSUM_NONE:
75 *
76 * The skb was already checksummed by the protocol, or a checksum is not
77 * required.
78 *
79 * CHECKSUM_PARTIAL:
80 *
81 * The device is required to checksum the packet as seen by hard_start_xmit()
82 * from skb->csum_start up to the end, and to record/write the checksum at
83 * offset skb->csum_start + skb->csum_offset.
84 *
85 * The device must show its capabilities in dev->features, set up at device
86 * setup time, e.g. netdev_features.h:
87 *
88 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
89 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
90 * IPv4. Sigh. Vendors like this way for an unknown reason.
91 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
92 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
93 * NETIF_F_... - Well, you get the picture.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * Normally, the device will do per protocol specific checksumming. Protocol
98 * implementations that do not want the NIC to perform the checksum
99 * calculation should use this flag in their outgoing skbs.
100 *
101 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
105 * Any questions? No questions, good. --ANK
106 */
107
60476372 108/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
109#define CHECKSUM_NONE 0
110#define CHECKSUM_UNNECESSARY 1
111#define CHECKSUM_COMPLETE 2
112#define CHECKSUM_PARTIAL 3
1da177e4
LT
113
114#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
115 ~(SMP_CACHE_BYTES - 1))
fc910a27 116#define SKB_WITH_OVERHEAD(X) \
deea84b0 117 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
118#define SKB_MAX_ORDER(X, ORDER) \
119 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
120#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
121#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
122
87fb4b7b
ED
123/* return minimum truesize of one skb containing X bytes of data */
124#define SKB_TRUESIZE(X) ((X) + \
125 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
126 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
127
1da177e4 128struct net_device;
716ea3a7 129struct scatterlist;
9c55e01c 130struct pipe_inode_info;
1da177e4 131
5f79e0f9 132#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
133struct nf_conntrack {
134 atomic_t use;
1da177e4 135};
5f79e0f9 136#endif
1da177e4
LT
137
138#ifdef CONFIG_BRIDGE_NETFILTER
139struct nf_bridge_info {
bf1ac5ca
ED
140 atomic_t use;
141 unsigned int mask;
142 struct net_device *physindev;
143 struct net_device *physoutdev;
144 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
145};
146#endif
147
1da177e4
LT
148struct sk_buff_head {
149 /* These two members must be first. */
150 struct sk_buff *next;
151 struct sk_buff *prev;
152
153 __u32 qlen;
154 spinlock_t lock;
155};
156
157struct sk_buff;
158
9d4dde52
IC
159/* To allow 64K frame to be packed as single skb without frag_list we
160 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
161 * buffers which do not start on a page boundary.
162 *
163 * Since GRO uses frags we allocate at least 16 regardless of page
164 * size.
a715dea3 165 */
9d4dde52 166#if (65536/PAGE_SIZE + 1) < 16
eec00954 167#define MAX_SKB_FRAGS 16UL
a715dea3 168#else
9d4dde52 169#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 170#endif
1da177e4
LT
171
172typedef struct skb_frag_struct skb_frag_t;
173
174struct skb_frag_struct {
a8605c60
IC
175 struct {
176 struct page *p;
177 } page;
cb4dfe56 178#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
179 __u32 page_offset;
180 __u32 size;
cb4dfe56
ED
181#else
182 __u16 page_offset;
183 __u16 size;
184#endif
1da177e4
LT
185};
186
9e903e08
ED
187static inline unsigned int skb_frag_size(const skb_frag_t *frag)
188{
189 return frag->size;
190}
191
192static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
193{
194 frag->size = size;
195}
196
197static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
198{
199 frag->size += delta;
200}
201
202static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
203{
204 frag->size -= delta;
205}
206
ac45f602
PO
207#define HAVE_HW_TIME_STAMP
208
209/**
d3a21be8 210 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
211 * @hwtstamp: hardware time stamp transformed into duration
212 * since arbitrary point in time
213 * @syststamp: hwtstamp transformed to system time base
214 *
215 * Software time stamps generated by ktime_get_real() are stored in
216 * skb->tstamp. The relation between the different kinds of time
217 * stamps is as follows:
218 *
219 * syststamp and tstamp can be compared against each other in
220 * arbitrary combinations. The accuracy of a
221 * syststamp/tstamp/"syststamp from other device" comparison is
222 * limited by the accuracy of the transformation into system time
223 * base. This depends on the device driver and its underlying
224 * hardware.
225 *
226 * hwtstamps can only be compared against other hwtstamps from
227 * the same device.
228 *
229 * This structure is attached to packets as part of the
230 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
231 */
232struct skb_shared_hwtstamps {
233 ktime_t hwtstamp;
234 ktime_t syststamp;
235};
236
2244d07b
OH
237/* Definitions for tx_flags in struct skb_shared_info */
238enum {
239 /* generate hardware time stamp */
240 SKBTX_HW_TSTAMP = 1 << 0,
241
242 /* generate software time stamp */
243 SKBTX_SW_TSTAMP = 1 << 1,
244
245 /* device driver is going to provide hardware time stamp */
246 SKBTX_IN_PROGRESS = 1 << 2,
247
a6686f2f 248 /* device driver supports TX zero-copy buffers */
62b1a8ab 249 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
250
251 /* generate wifi status information (where possible) */
62b1a8ab 252 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
253
254 /* This indicates at least one fragment might be overwritten
255 * (as in vmsplice(), sendfile() ...)
256 * If we need to compute a TX checksum, we'll need to copy
257 * all frags to avoid possible bad checksum
258 */
259 SKBTX_SHARED_FRAG = 1 << 5,
a6686f2f
SM
260};
261
262/*
263 * The callback notifies userspace to release buffers when skb DMA is done in
264 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
265 * The zerocopy_success argument is true if zero copy transmit occurred,
266 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
267 * The ctx field is used to track device context.
268 * The desc field is used to track userspace buffer index.
a6686f2f
SM
269 */
270struct ubuf_info {
e19d6763 271 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 272 void *ctx;
a6686f2f 273 unsigned long desc;
ac45f602
PO
274};
275
1da177e4
LT
276/* This data is invariant across clones and lives at
277 * the end of the header data, ie. at skb->end.
278 */
279struct skb_shared_info {
9f42f126
IC
280 unsigned char nr_frags;
281 __u8 tx_flags;
7967168c
HX
282 unsigned short gso_size;
283 /* Warning: this field is not always filled in (UFO)! */
284 unsigned short gso_segs;
285 unsigned short gso_type;
1da177e4 286 struct sk_buff *frag_list;
ac45f602 287 struct skb_shared_hwtstamps hwtstamps;
9f42f126 288 __be32 ip6_frag_id;
ec7d2f2c
ED
289
290 /*
291 * Warning : all fields before dataref are cleared in __alloc_skb()
292 */
293 atomic_t dataref;
294
69e3c75f
JB
295 /* Intermediate layers must ensure that destructor_arg
296 * remains valid until skb destructor */
297 void * destructor_arg;
a6686f2f 298
fed66381
ED
299 /* must be last field, see pskb_expand_head() */
300 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
301};
302
303/* We divide dataref into two halves. The higher 16 bits hold references
304 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
305 * the entire skb->data. A clone of a headerless skb holds the length of
306 * the header in skb->hdr_len.
1da177e4
LT
307 *
308 * All users must obey the rule that the skb->data reference count must be
309 * greater than or equal to the payload reference count.
310 *
311 * Holding a reference to the payload part means that the user does not
312 * care about modifications to the header part of skb->data.
313 */
314#define SKB_DATAREF_SHIFT 16
315#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
316
d179cd12
DM
317
318enum {
319 SKB_FCLONE_UNAVAILABLE,
320 SKB_FCLONE_ORIG,
321 SKB_FCLONE_CLONE,
322};
323
7967168c
HX
324enum {
325 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 326 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
327
328 /* This indicates the skb is from an untrusted source. */
329 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
330
331 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
332 SKB_GSO_TCP_ECN = 1 << 3,
333
334 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
335
336 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
337
338 SKB_GSO_GRE = 1 << 6,
73136267 339
cb32f511 340 SKB_GSO_IPIP = 1 << 7,
0d89d203 341
61c1db7f 342 SKB_GSO_SIT = 1 << 8,
cb32f511 343
61c1db7f
ED
344 SKB_GSO_UDP_TUNNEL = 1 << 9,
345
346 SKB_GSO_MPLS = 1 << 10,
7967168c
HX
347};
348
2e07fa9c
ACM
349#if BITS_PER_LONG > 32
350#define NET_SKBUFF_DATA_USES_OFFSET 1
351#endif
352
353#ifdef NET_SKBUFF_DATA_USES_OFFSET
354typedef unsigned int sk_buff_data_t;
355#else
356typedef unsigned char *sk_buff_data_t;
357#endif
358
1da177e4
LT
359/**
360 * struct sk_buff - socket buffer
361 * @next: Next buffer in list
362 * @prev: Previous buffer in list
325ed823 363 * @tstamp: Time we arrived
d84e0bd7 364 * @sk: Socket we are owned by
1da177e4 365 * @dev: Device we arrived on/are leaving by
d84e0bd7 366 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 367 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 368 * @sp: the security path, used for xfrm
1da177e4
LT
369 * @len: Length of actual data
370 * @data_len: Data length
371 * @mac_len: Length of link layer header
334a8132 372 * @hdr_len: writable header length of cloned skb
663ead3b
HX
373 * @csum: Checksum (must include start/offset pair)
374 * @csum_start: Offset from skb->head where checksumming should start
375 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 376 * @priority: Packet queueing priority
67be2dd1 377 * @local_df: allow local fragmentation
1da177e4 378 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 379 * @ip_summed: Driver fed us an IP checksum
1da177e4 380 * @nohdr: Payload reference only, must not modify header
d84e0bd7 381 * @nfctinfo: Relationship of this skb to the connection
1da177e4 382 * @pkt_type: Packet class
c83c2486 383 * @fclone: skbuff clone status
c83c2486 384 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
385 * @peeked: this packet has been seen already, so stats have been
386 * done for it, don't do them again
ba9dda3a 387 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
388 * @protocol: Packet protocol from driver
389 * @destructor: Destruct function
390 * @nfct: Associated connection, if any
1da177e4 391 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 392 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
393 * @tc_index: Traffic control index
394 * @tc_verd: traffic control verdict
d84e0bd7
DB
395 * @rxhash: the packet hash computed on receive
396 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 397 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 398 * @ooo_okay: allow the mapping of a socket to a queue to be changed
4ca2462e
CG
399 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
400 * ports.
6e3e939f
JB
401 * @wifi_acked_valid: wifi_acked was set
402 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 403 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
404 * @dma_cookie: a cookie to one of several possible DMA operations
405 * done by skb DMA functions
06021292 406 * @napi_id: id of the NAPI struct this skb came from
984bc16c 407 * @secmark: security marking
d84e0bd7
DB
408 * @mark: Generic packet mark
409 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 410 * @vlan_proto: vlan encapsulation protocol
6aa895b0 411 * @vlan_tci: vlan tag control information
0d89d203 412 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
413 * @inner_transport_header: Inner transport layer header (encapsulation)
414 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 415 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
416 * @transport_header: Transport layer header
417 * @network_header: Network layer header
418 * @mac_header: Link layer header
419 * @tail: Tail pointer
420 * @end: End pointer
421 * @head: Head of buffer
422 * @data: Data head pointer
423 * @truesize: Buffer size
424 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
425 */
426
427struct sk_buff {
428 /* These two members must be first. */
429 struct sk_buff *next;
430 struct sk_buff *prev;
431
b7aa0bf7 432 ktime_t tstamp;
da3f5cf1
FF
433
434 struct sock *sk;
1da177e4 435 struct net_device *dev;
1da177e4 436
1da177e4
LT
437 /*
438 * This is the control buffer. It is free to use for every
439 * layer. Please put your private variables there. If you
440 * want to keep them across layers you have to do a skb_clone()
441 * first. This is owned by whoever has the skb queued ATM.
442 */
da3f5cf1 443 char cb[48] __aligned(8);
1da177e4 444
7fee226a 445 unsigned long _skb_refdst;
da3f5cf1
FF
446#ifdef CONFIG_XFRM
447 struct sec_path *sp;
448#endif
1da177e4 449 unsigned int len,
334a8132
PM
450 data_len;
451 __u16 mac_len,
452 hdr_len;
ff1dcadb
AV
453 union {
454 __wsum csum;
663ead3b
HX
455 struct {
456 __u16 csum_start;
457 __u16 csum_offset;
458 };
ff1dcadb 459 };
1da177e4 460 __u32 priority;
fe55f6d5 461 kmemcheck_bitfield_begin(flags1);
1cbb3380
TG
462 __u8 local_df:1,
463 cloned:1,
464 ip_summed:2,
6869c4d8
HW
465 nohdr:1,
466 nfctinfo:3;
d179cd12 467 __u8 pkt_type:3,
b84f4cc9 468 fclone:2,
ba9dda3a 469 ipvs_property:1,
a59322be 470 peeked:1,
ba9dda3a 471 nf_trace:1;
fe55f6d5 472 kmemcheck_bitfield_end(flags1);
4ab408de 473 __be16 protocol;
1da177e4
LT
474
475 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 476#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 477 struct nf_conntrack *nfct;
2fc72c7b 478#endif
1da177e4
LT
479#ifdef CONFIG_BRIDGE_NETFILTER
480 struct nf_bridge_info *nf_bridge;
481#endif
f25f4e44 482
8964be4a 483 int skb_iif;
4031ae6e
AD
484
485 __u32 rxhash;
486
86a9bad3 487 __be16 vlan_proto;
4031ae6e
AD
488 __u16 vlan_tci;
489
1da177e4 490#ifdef CONFIG_NET_SCHED
b6b99eb5 491 __u16 tc_index; /* traffic control index */
1da177e4 492#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 493 __u16 tc_verd; /* traffic control verdict */
1da177e4 494#endif
1da177e4 495#endif
fe55f6d5 496
0a14842f 497 __u16 queue_mapping;
fe55f6d5 498 kmemcheck_bitfield_begin(flags2);
de357cc0 499#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 500 __u8 ndisc_nodetype:2;
d0f09804 501#endif
c93bdd0e 502 __u8 pfmemalloc:1;
3853b584 503 __u8 ooo_okay:1;
bdeab991 504 __u8 l4_rxhash:1;
6e3e939f
JB
505 __u8 wifi_acked_valid:1;
506 __u8 wifi_acked:1;
3bdc0eba 507 __u8 no_fcs:1;
d3836f21 508 __u8 head_frag:1;
6a674e9c
JG
509 /* Encapsulation protocol and NIC drivers should use
510 * this flag to indicate to each other if the skb contains
511 * encapsulated packet or not and maybe use the inner packet
512 * headers if needed
513 */
514 __u8 encapsulation:1;
45906723 515 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
516 kmemcheck_bitfield_end(flags2);
517
e0d1095a 518#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
519 union {
520 unsigned int napi_id;
521 dma_cookie_t dma_cookie;
522 };
97fc2f08 523#endif
984bc16c
JM
524#ifdef CONFIG_NETWORK_SECMARK
525 __u32 secmark;
526#endif
3b885787
NH
527 union {
528 __u32 mark;
529 __u32 dropcount;
16fad69c 530 __u32 reserved_tailroom;
3b885787 531 };
1da177e4 532
0d89d203 533 __be16 inner_protocol;
1a37e412
SH
534 __u16 inner_transport_header;
535 __u16 inner_network_header;
536 __u16 inner_mac_header;
537 __u16 transport_header;
538 __u16 network_header;
539 __u16 mac_header;
1da177e4 540 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 541 sk_buff_data_t tail;
4305b541 542 sk_buff_data_t end;
1da177e4 543 unsigned char *head,
4305b541 544 *data;
27a884dc
ACM
545 unsigned int truesize;
546 atomic_t users;
1da177e4
LT
547};
548
549#ifdef __KERNEL__
550/*
551 * Handling routines are only of interest to the kernel
552 */
553#include <linux/slab.h>
554
1da177e4 555
c93bdd0e
MG
556#define SKB_ALLOC_FCLONE 0x01
557#define SKB_ALLOC_RX 0x02
558
559/* Returns true if the skb was allocated from PFMEMALLOC reserves */
560static inline bool skb_pfmemalloc(const struct sk_buff *skb)
561{
562 return unlikely(skb->pfmemalloc);
563}
564
7fee226a
ED
565/*
566 * skb might have a dst pointer attached, refcounted or not.
567 * _skb_refdst low order bit is set if refcount was _not_ taken
568 */
569#define SKB_DST_NOREF 1UL
570#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
571
572/**
573 * skb_dst - returns skb dst_entry
574 * @skb: buffer
575 *
576 * Returns skb dst_entry, regardless of reference taken or not.
577 */
adf30907
ED
578static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
579{
7fee226a
ED
580 /* If refdst was not refcounted, check we still are in a
581 * rcu_read_lock section
582 */
583 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
584 !rcu_read_lock_held() &&
585 !rcu_read_lock_bh_held());
586 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
587}
588
7fee226a
ED
589/**
590 * skb_dst_set - sets skb dst
591 * @skb: buffer
592 * @dst: dst entry
593 *
594 * Sets skb dst, assuming a reference was taken on dst and should
595 * be released by skb_dst_drop()
596 */
adf30907
ED
597static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
598{
7fee226a
ED
599 skb->_skb_refdst = (unsigned long)dst;
600}
601
7965bd4d
JP
602void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
603 bool force);
932bc4d7
JA
604
605/**
606 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
607 * @skb: buffer
608 * @dst: dst entry
609 *
610 * Sets skb dst, assuming a reference was not taken on dst.
611 * If dst entry is cached, we do not take reference and dst_release
612 * will be avoided by refdst_drop. If dst entry is not cached, we take
613 * reference, so that last dst_release can destroy the dst immediately.
614 */
615static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
616{
617 __skb_dst_set_noref(skb, dst, false);
618}
619
620/**
621 * skb_dst_set_noref_force - sets skb dst, without taking reference
622 * @skb: buffer
623 * @dst: dst entry
624 *
625 * Sets skb dst, assuming a reference was not taken on dst.
626 * No reference is taken and no dst_release will be called. While for
627 * cached dsts deferred reclaim is a basic feature, for entries that are
628 * not cached it is caller's job to guarantee that last dst_release for
629 * provided dst happens when nobody uses it, eg. after a RCU grace period.
630 */
631static inline void skb_dst_set_noref_force(struct sk_buff *skb,
632 struct dst_entry *dst)
633{
634 __skb_dst_set_noref(skb, dst, true);
635}
7fee226a
ED
636
637/**
25985edc 638 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
639 * @skb: buffer
640 */
641static inline bool skb_dst_is_noref(const struct sk_buff *skb)
642{
643 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
644}
645
511c3f92
ED
646static inline struct rtable *skb_rtable(const struct sk_buff *skb)
647{
adf30907 648 return (struct rtable *)skb_dst(skb);
511c3f92
ED
649}
650
7965bd4d
JP
651void kfree_skb(struct sk_buff *skb);
652void kfree_skb_list(struct sk_buff *segs);
653void skb_tx_error(struct sk_buff *skb);
654void consume_skb(struct sk_buff *skb);
655void __kfree_skb(struct sk_buff *skb);
d7e8883c 656extern struct kmem_cache *skbuff_head_cache;
bad43ca8 657
7965bd4d
JP
658void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
659bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
660 bool *fragstolen, int *delta_truesize);
bad43ca8 661
7965bd4d
JP
662struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
663 int node);
664struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 665static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 666 gfp_t priority)
d179cd12 667{
564824b0 668 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
669}
670
671static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 672 gfp_t priority)
d179cd12 673{
c93bdd0e 674 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
675}
676
7965bd4d 677struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
678static inline struct sk_buff *alloc_skb_head(gfp_t priority)
679{
680 return __alloc_skb_head(priority, -1);
681}
682
7965bd4d
JP
683struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
684int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
685struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
686struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
687struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
688
689int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
690struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
691 unsigned int headroom);
692struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
693 int newtailroom, gfp_t priority);
25a91d8d
FD
694int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
695 int offset, int len);
7965bd4d
JP
696int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
697 int len);
698int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
699int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 700#define dev_kfree_skb(a) consume_skb(a)
1da177e4 701
7965bd4d
JP
702int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
703 int getfrag(void *from, char *to, int offset,
704 int len, int odd, struct sk_buff *skb),
705 void *from, int length);
e89e9cf5 706
d94d9fee 707struct skb_seq_state {
677e90ed
TG
708 __u32 lower_offset;
709 __u32 upper_offset;
710 __u32 frag_idx;
711 __u32 stepped_offset;
712 struct sk_buff *root_skb;
713 struct sk_buff *cur_skb;
714 __u8 *frag_data;
715};
716
7965bd4d
JP
717void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
718 unsigned int to, struct skb_seq_state *st);
719unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
720 struct skb_seq_state *st);
721void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 722
7965bd4d
JP
723unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
724 unsigned int to, struct ts_config *config,
725 struct ts_state *state);
3fc7e8a6 726
09323cc4
TH
727/*
728 * Packet hash types specify the type of hash in skb_set_hash.
729 *
730 * Hash types refer to the protocol layer addresses which are used to
731 * construct a packet's hash. The hashes are used to differentiate or identify
732 * flows of the protocol layer for the hash type. Hash types are either
733 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
734 *
735 * Properties of hashes:
736 *
737 * 1) Two packets in different flows have different hash values
738 * 2) Two packets in the same flow should have the same hash value
739 *
740 * A hash at a higher layer is considered to be more specific. A driver should
741 * set the most specific hash possible.
742 *
743 * A driver cannot indicate a more specific hash than the layer at which a hash
744 * was computed. For instance an L3 hash cannot be set as an L4 hash.
745 *
746 * A driver may indicate a hash level which is less specific than the
747 * actual layer the hash was computed on. For instance, a hash computed
748 * at L4 may be considered an L3 hash. This should only be done if the
749 * driver can't unambiguously determine that the HW computed the hash at
750 * the higher layer. Note that the "should" in the second property above
751 * permits this.
752 */
753enum pkt_hash_types {
754 PKT_HASH_TYPE_NONE, /* Undefined type */
755 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
756 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
757 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
758};
759
760static inline void
761skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
762{
763 skb->l4_rxhash = (type == PKT_HASH_TYPE_L4);
764 skb->rxhash = hash;
765}
766
3958afa1
TH
767void __skb_get_hash(struct sk_buff *skb);
768static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 769{
ecd5cf5d 770 if (!skb->l4_rxhash)
3958afa1 771 __skb_get_hash(skb);
bfb564e7
KK
772
773 return skb->rxhash;
774}
775
57bdf7f4
TH
776static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
777{
778 return skb->rxhash;
779}
780
7539fadc
TH
781static inline void skb_clear_hash(struct sk_buff *skb)
782{
783 skb->rxhash = 0;
784 skb->l4_rxhash = 0;
785}
786
787static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
788{
789 if (!skb->l4_rxhash)
790 skb_clear_hash(skb);
791}
792
3df7a74e
TH
793static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
794{
795 to->rxhash = from->rxhash;
796 to->l4_rxhash = from->l4_rxhash;
797};
798
4305b541
ACM
799#ifdef NET_SKBUFF_DATA_USES_OFFSET
800static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
801{
802 return skb->head + skb->end;
803}
ec47ea82
AD
804
805static inline unsigned int skb_end_offset(const struct sk_buff *skb)
806{
807 return skb->end;
808}
4305b541
ACM
809#else
810static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
811{
812 return skb->end;
813}
ec47ea82
AD
814
815static inline unsigned int skb_end_offset(const struct sk_buff *skb)
816{
817 return skb->end - skb->head;
818}
4305b541
ACM
819#endif
820
1da177e4 821/* Internal */
4305b541 822#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 823
ac45f602
PO
824static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
825{
826 return &skb_shinfo(skb)->hwtstamps;
827}
828
1da177e4
LT
829/**
830 * skb_queue_empty - check if a queue is empty
831 * @list: queue head
832 *
833 * Returns true if the queue is empty, false otherwise.
834 */
835static inline int skb_queue_empty(const struct sk_buff_head *list)
836{
fd44b93c 837 return list->next == (const struct sk_buff *) list;
1da177e4
LT
838}
839
fc7ebb21
DM
840/**
841 * skb_queue_is_last - check if skb is the last entry in the queue
842 * @list: queue head
843 * @skb: buffer
844 *
845 * Returns true if @skb is the last buffer on the list.
846 */
847static inline bool skb_queue_is_last(const struct sk_buff_head *list,
848 const struct sk_buff *skb)
849{
fd44b93c 850 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
851}
852
832d11c5
IJ
853/**
854 * skb_queue_is_first - check if skb is the first entry in the queue
855 * @list: queue head
856 * @skb: buffer
857 *
858 * Returns true if @skb is the first buffer on the list.
859 */
860static inline bool skb_queue_is_first(const struct sk_buff_head *list,
861 const struct sk_buff *skb)
862{
fd44b93c 863 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
864}
865
249c8b42
DM
866/**
867 * skb_queue_next - return the next packet in the queue
868 * @list: queue head
869 * @skb: current buffer
870 *
871 * Return the next packet in @list after @skb. It is only valid to
872 * call this if skb_queue_is_last() evaluates to false.
873 */
874static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
875 const struct sk_buff *skb)
876{
877 /* This BUG_ON may seem severe, but if we just return then we
878 * are going to dereference garbage.
879 */
880 BUG_ON(skb_queue_is_last(list, skb));
881 return skb->next;
882}
883
832d11c5
IJ
884/**
885 * skb_queue_prev - return the prev packet in the queue
886 * @list: queue head
887 * @skb: current buffer
888 *
889 * Return the prev packet in @list before @skb. It is only valid to
890 * call this if skb_queue_is_first() evaluates to false.
891 */
892static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
893 const struct sk_buff *skb)
894{
895 /* This BUG_ON may seem severe, but if we just return then we
896 * are going to dereference garbage.
897 */
898 BUG_ON(skb_queue_is_first(list, skb));
899 return skb->prev;
900}
901
1da177e4
LT
902/**
903 * skb_get - reference buffer
904 * @skb: buffer to reference
905 *
906 * Makes another reference to a socket buffer and returns a pointer
907 * to the buffer.
908 */
909static inline struct sk_buff *skb_get(struct sk_buff *skb)
910{
911 atomic_inc(&skb->users);
912 return skb;
913}
914
915/*
916 * If users == 1, we are the only owner and are can avoid redundant
917 * atomic change.
918 */
919
1da177e4
LT
920/**
921 * skb_cloned - is the buffer a clone
922 * @skb: buffer to check
923 *
924 * Returns true if the buffer was generated with skb_clone() and is
925 * one of multiple shared copies of the buffer. Cloned buffers are
926 * shared data so must not be written to under normal circumstances.
927 */
928static inline int skb_cloned(const struct sk_buff *skb)
929{
930 return skb->cloned &&
931 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
932}
933
14bbd6a5
PS
934static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
935{
936 might_sleep_if(pri & __GFP_WAIT);
937
938 if (skb_cloned(skb))
939 return pskb_expand_head(skb, 0, 0, pri);
940
941 return 0;
942}
943
1da177e4
LT
944/**
945 * skb_header_cloned - is the header a clone
946 * @skb: buffer to check
947 *
948 * Returns true if modifying the header part of the buffer requires
949 * the data to be copied.
950 */
951static inline int skb_header_cloned(const struct sk_buff *skb)
952{
953 int dataref;
954
955 if (!skb->cloned)
956 return 0;
957
958 dataref = atomic_read(&skb_shinfo(skb)->dataref);
959 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
960 return dataref != 1;
961}
962
963/**
964 * skb_header_release - release reference to header
965 * @skb: buffer to operate on
966 *
967 * Drop a reference to the header part of the buffer. This is done
968 * by acquiring a payload reference. You must not read from the header
969 * part of skb->data after this.
970 */
971static inline void skb_header_release(struct sk_buff *skb)
972{
973 BUG_ON(skb->nohdr);
974 skb->nohdr = 1;
975 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
976}
977
978/**
979 * skb_shared - is the buffer shared
980 * @skb: buffer to check
981 *
982 * Returns true if more than one person has a reference to this
983 * buffer.
984 */
985static inline int skb_shared(const struct sk_buff *skb)
986{
987 return atomic_read(&skb->users) != 1;
988}
989
990/**
991 * skb_share_check - check if buffer is shared and if so clone it
992 * @skb: buffer to check
993 * @pri: priority for memory allocation
994 *
995 * If the buffer is shared the buffer is cloned and the old copy
996 * drops a reference. A new clone with a single reference is returned.
997 * If the buffer is not shared the original buffer is returned. When
998 * being called from interrupt status or with spinlocks held pri must
999 * be GFP_ATOMIC.
1000 *
1001 * NULL is returned on a memory allocation failure.
1002 */
47061bc4 1003static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1004{
1005 might_sleep_if(pri & __GFP_WAIT);
1006 if (skb_shared(skb)) {
1007 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1008
1009 if (likely(nskb))
1010 consume_skb(skb);
1011 else
1012 kfree_skb(skb);
1da177e4
LT
1013 skb = nskb;
1014 }
1015 return skb;
1016}
1017
1018/*
1019 * Copy shared buffers into a new sk_buff. We effectively do COW on
1020 * packets to handle cases where we have a local reader and forward
1021 * and a couple of other messy ones. The normal one is tcpdumping
1022 * a packet thats being forwarded.
1023 */
1024
1025/**
1026 * skb_unshare - make a copy of a shared buffer
1027 * @skb: buffer to check
1028 * @pri: priority for memory allocation
1029 *
1030 * If the socket buffer is a clone then this function creates a new
1031 * copy of the data, drops a reference count on the old copy and returns
1032 * the new copy with the reference count at 1. If the buffer is not a clone
1033 * the original buffer is returned. When called with a spinlock held or
1034 * from interrupt state @pri must be %GFP_ATOMIC
1035 *
1036 * %NULL is returned on a memory allocation failure.
1037 */
e2bf521d 1038static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1039 gfp_t pri)
1da177e4
LT
1040{
1041 might_sleep_if(pri & __GFP_WAIT);
1042 if (skb_cloned(skb)) {
1043 struct sk_buff *nskb = skb_copy(skb, pri);
1044 kfree_skb(skb); /* Free our shared copy */
1045 skb = nskb;
1046 }
1047 return skb;
1048}
1049
1050/**
1a5778aa 1051 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1052 * @list_: list to peek at
1053 *
1054 * Peek an &sk_buff. Unlike most other operations you _MUST_
1055 * be careful with this one. A peek leaves the buffer on the
1056 * list and someone else may run off with it. You must hold
1057 * the appropriate locks or have a private queue to do this.
1058 *
1059 * Returns %NULL for an empty list or a pointer to the head element.
1060 * The reference count is not incremented and the reference is therefore
1061 * volatile. Use with caution.
1062 */
05bdd2f1 1063static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1064{
18d07000
ED
1065 struct sk_buff *skb = list_->next;
1066
1067 if (skb == (struct sk_buff *)list_)
1068 skb = NULL;
1069 return skb;
1da177e4
LT
1070}
1071
da5ef6e5
PE
1072/**
1073 * skb_peek_next - peek skb following the given one from a queue
1074 * @skb: skb to start from
1075 * @list_: list to peek at
1076 *
1077 * Returns %NULL when the end of the list is met or a pointer to the
1078 * next element. The reference count is not incremented and the
1079 * reference is therefore volatile. Use with caution.
1080 */
1081static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1082 const struct sk_buff_head *list_)
1083{
1084 struct sk_buff *next = skb->next;
18d07000 1085
da5ef6e5
PE
1086 if (next == (struct sk_buff *)list_)
1087 next = NULL;
1088 return next;
1089}
1090
1da177e4 1091/**
1a5778aa 1092 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1093 * @list_: list to peek at
1094 *
1095 * Peek an &sk_buff. Unlike most other operations you _MUST_
1096 * be careful with this one. A peek leaves the buffer on the
1097 * list and someone else may run off with it. You must hold
1098 * the appropriate locks or have a private queue to do this.
1099 *
1100 * Returns %NULL for an empty list or a pointer to the tail element.
1101 * The reference count is not incremented and the reference is therefore
1102 * volatile. Use with caution.
1103 */
05bdd2f1 1104static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1105{
18d07000
ED
1106 struct sk_buff *skb = list_->prev;
1107
1108 if (skb == (struct sk_buff *)list_)
1109 skb = NULL;
1110 return skb;
1111
1da177e4
LT
1112}
1113
1114/**
1115 * skb_queue_len - get queue length
1116 * @list_: list to measure
1117 *
1118 * Return the length of an &sk_buff queue.
1119 */
1120static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1121{
1122 return list_->qlen;
1123}
1124
67fed459
DM
1125/**
1126 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1127 * @list: queue to initialize
1128 *
1129 * This initializes only the list and queue length aspects of
1130 * an sk_buff_head object. This allows to initialize the list
1131 * aspects of an sk_buff_head without reinitializing things like
1132 * the spinlock. It can also be used for on-stack sk_buff_head
1133 * objects where the spinlock is known to not be used.
1134 */
1135static inline void __skb_queue_head_init(struct sk_buff_head *list)
1136{
1137 list->prev = list->next = (struct sk_buff *)list;
1138 list->qlen = 0;
1139}
1140
76f10ad0
AV
1141/*
1142 * This function creates a split out lock class for each invocation;
1143 * this is needed for now since a whole lot of users of the skb-queue
1144 * infrastructure in drivers have different locking usage (in hardirq)
1145 * than the networking core (in softirq only). In the long run either the
1146 * network layer or drivers should need annotation to consolidate the
1147 * main types of usage into 3 classes.
1148 */
1da177e4
LT
1149static inline void skb_queue_head_init(struct sk_buff_head *list)
1150{
1151 spin_lock_init(&list->lock);
67fed459 1152 __skb_queue_head_init(list);
1da177e4
LT
1153}
1154
c2ecba71
PE
1155static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1156 struct lock_class_key *class)
1157{
1158 skb_queue_head_init(list);
1159 lockdep_set_class(&list->lock, class);
1160}
1161
1da177e4 1162/*
bf299275 1163 * Insert an sk_buff on a list.
1da177e4
LT
1164 *
1165 * The "__skb_xxxx()" functions are the non-atomic ones that
1166 * can only be called with interrupts disabled.
1167 */
7965bd4d
JP
1168void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1169 struct sk_buff_head *list);
bf299275
GR
1170static inline void __skb_insert(struct sk_buff *newsk,
1171 struct sk_buff *prev, struct sk_buff *next,
1172 struct sk_buff_head *list)
1173{
1174 newsk->next = next;
1175 newsk->prev = prev;
1176 next->prev = prev->next = newsk;
1177 list->qlen++;
1178}
1da177e4 1179
67fed459
DM
1180static inline void __skb_queue_splice(const struct sk_buff_head *list,
1181 struct sk_buff *prev,
1182 struct sk_buff *next)
1183{
1184 struct sk_buff *first = list->next;
1185 struct sk_buff *last = list->prev;
1186
1187 first->prev = prev;
1188 prev->next = first;
1189
1190 last->next = next;
1191 next->prev = last;
1192}
1193
1194/**
1195 * skb_queue_splice - join two skb lists, this is designed for stacks
1196 * @list: the new list to add
1197 * @head: the place to add it in the first list
1198 */
1199static inline void skb_queue_splice(const struct sk_buff_head *list,
1200 struct sk_buff_head *head)
1201{
1202 if (!skb_queue_empty(list)) {
1203 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1204 head->qlen += list->qlen;
67fed459
DM
1205 }
1206}
1207
1208/**
d9619496 1209 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1210 * @list: the new list to add
1211 * @head: the place to add it in the first list
1212 *
1213 * The list at @list is reinitialised
1214 */
1215static inline void skb_queue_splice_init(struct sk_buff_head *list,
1216 struct sk_buff_head *head)
1217{
1218 if (!skb_queue_empty(list)) {
1219 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1220 head->qlen += list->qlen;
67fed459
DM
1221 __skb_queue_head_init(list);
1222 }
1223}
1224
1225/**
1226 * skb_queue_splice_tail - join two skb lists, each list being a queue
1227 * @list: the new list to add
1228 * @head: the place to add it in the first list
1229 */
1230static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1231 struct sk_buff_head *head)
1232{
1233 if (!skb_queue_empty(list)) {
1234 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1235 head->qlen += list->qlen;
67fed459
DM
1236 }
1237}
1238
1239/**
d9619496 1240 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1241 * @list: the new list to add
1242 * @head: the place to add it in the first list
1243 *
1244 * Each of the lists is a queue.
1245 * The list at @list is reinitialised
1246 */
1247static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1248 struct sk_buff_head *head)
1249{
1250 if (!skb_queue_empty(list)) {
1251 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1252 head->qlen += list->qlen;
67fed459
DM
1253 __skb_queue_head_init(list);
1254 }
1255}
1256
1da177e4 1257/**
300ce174 1258 * __skb_queue_after - queue a buffer at the list head
1da177e4 1259 * @list: list to use
300ce174 1260 * @prev: place after this buffer
1da177e4
LT
1261 * @newsk: buffer to queue
1262 *
300ce174 1263 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1264 * and you must therefore hold required locks before calling it.
1265 *
1266 * A buffer cannot be placed on two lists at the same time.
1267 */
300ce174
SH
1268static inline void __skb_queue_after(struct sk_buff_head *list,
1269 struct sk_buff *prev,
1270 struct sk_buff *newsk)
1da177e4 1271{
bf299275 1272 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1273}
1274
7965bd4d
JP
1275void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1276 struct sk_buff_head *list);
7de6c033 1277
f5572855
GR
1278static inline void __skb_queue_before(struct sk_buff_head *list,
1279 struct sk_buff *next,
1280 struct sk_buff *newsk)
1281{
1282 __skb_insert(newsk, next->prev, next, list);
1283}
1284
300ce174
SH
1285/**
1286 * __skb_queue_head - queue a buffer at the list head
1287 * @list: list to use
1288 * @newsk: buffer to queue
1289 *
1290 * Queue a buffer at the start of a list. This function takes no locks
1291 * and you must therefore hold required locks before calling it.
1292 *
1293 * A buffer cannot be placed on two lists at the same time.
1294 */
7965bd4d 1295void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1296static inline void __skb_queue_head(struct sk_buff_head *list,
1297 struct sk_buff *newsk)
1298{
1299 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1300}
1301
1da177e4
LT
1302/**
1303 * __skb_queue_tail - queue a buffer at the list tail
1304 * @list: list to use
1305 * @newsk: buffer to queue
1306 *
1307 * Queue a buffer at the end of a list. This function takes no locks
1308 * and you must therefore hold required locks before calling it.
1309 *
1310 * A buffer cannot be placed on two lists at the same time.
1311 */
7965bd4d 1312void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1313static inline void __skb_queue_tail(struct sk_buff_head *list,
1314 struct sk_buff *newsk)
1315{
f5572855 1316 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1317}
1318
1da177e4
LT
1319/*
1320 * remove sk_buff from list. _Must_ be called atomically, and with
1321 * the list known..
1322 */
7965bd4d 1323void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1324static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1325{
1326 struct sk_buff *next, *prev;
1327
1328 list->qlen--;
1329 next = skb->next;
1330 prev = skb->prev;
1331 skb->next = skb->prev = NULL;
1da177e4
LT
1332 next->prev = prev;
1333 prev->next = next;
1334}
1335
f525c06d
GR
1336/**
1337 * __skb_dequeue - remove from the head of the queue
1338 * @list: list to dequeue from
1339 *
1340 * Remove the head of the list. This function does not take any locks
1341 * so must be used with appropriate locks held only. The head item is
1342 * returned or %NULL if the list is empty.
1343 */
7965bd4d 1344struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1345static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1346{
1347 struct sk_buff *skb = skb_peek(list);
1348 if (skb)
1349 __skb_unlink(skb, list);
1350 return skb;
1351}
1da177e4
LT
1352
1353/**
1354 * __skb_dequeue_tail - remove from the tail of the queue
1355 * @list: list to dequeue from
1356 *
1357 * Remove the tail of the list. This function does not take any locks
1358 * so must be used with appropriate locks held only. The tail item is
1359 * returned or %NULL if the list is empty.
1360 */
7965bd4d 1361struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1362static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1363{
1364 struct sk_buff *skb = skb_peek_tail(list);
1365 if (skb)
1366 __skb_unlink(skb, list);
1367 return skb;
1368}
1369
1370
bdcc0924 1371static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1372{
1373 return skb->data_len;
1374}
1375
1376static inline unsigned int skb_headlen(const struct sk_buff *skb)
1377{
1378 return skb->len - skb->data_len;
1379}
1380
1381static inline int skb_pagelen(const struct sk_buff *skb)
1382{
1383 int i, len = 0;
1384
1385 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1386 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1387 return len + skb_headlen(skb);
1388}
1389
131ea667
IC
1390/**
1391 * __skb_fill_page_desc - initialise a paged fragment in an skb
1392 * @skb: buffer containing fragment to be initialised
1393 * @i: paged fragment index to initialise
1394 * @page: the page to use for this fragment
1395 * @off: the offset to the data with @page
1396 * @size: the length of the data
1397 *
1398 * Initialises the @i'th fragment of @skb to point to &size bytes at
1399 * offset @off within @page.
1400 *
1401 * Does not take any additional reference on the fragment.
1402 */
1403static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1404 struct page *page, int off, int size)
1da177e4
LT
1405{
1406 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1407
c48a11c7
MG
1408 /*
1409 * Propagate page->pfmemalloc to the skb if we can. The problem is
1410 * that not all callers have unique ownership of the page. If
1411 * pfmemalloc is set, we check the mapping as a mapping implies
1412 * page->index is set (index and pfmemalloc share space).
1413 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1414 * do not lose pfmemalloc information as the pages would not be
1415 * allocated using __GFP_MEMALLOC.
1416 */
a8605c60 1417 frag->page.p = page;
1da177e4 1418 frag->page_offset = off;
9e903e08 1419 skb_frag_size_set(frag, size);
cca7af38
PE
1420
1421 page = compound_head(page);
1422 if (page->pfmemalloc && !page->mapping)
1423 skb->pfmemalloc = true;
131ea667
IC
1424}
1425
1426/**
1427 * skb_fill_page_desc - initialise a paged fragment in an skb
1428 * @skb: buffer containing fragment to be initialised
1429 * @i: paged fragment index to initialise
1430 * @page: the page to use for this fragment
1431 * @off: the offset to the data with @page
1432 * @size: the length of the data
1433 *
1434 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1435 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1436 * addition updates @skb such that @i is the last fragment.
1437 *
1438 * Does not take any additional reference on the fragment.
1439 */
1440static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1441 struct page *page, int off, int size)
1442{
1443 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1444 skb_shinfo(skb)->nr_frags = i + 1;
1445}
1446
7965bd4d
JP
1447void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1448 int size, unsigned int truesize);
654bed16 1449
f8e617e1
JW
1450void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1451 unsigned int truesize);
1452
1da177e4 1453#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1454#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1455#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1456
27a884dc
ACM
1457#ifdef NET_SKBUFF_DATA_USES_OFFSET
1458static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1459{
1460 return skb->head + skb->tail;
1461}
1462
1463static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1464{
1465 skb->tail = skb->data - skb->head;
1466}
1467
1468static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1469{
1470 skb_reset_tail_pointer(skb);
1471 skb->tail += offset;
1472}
7cc46190 1473
27a884dc
ACM
1474#else /* NET_SKBUFF_DATA_USES_OFFSET */
1475static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1476{
1477 return skb->tail;
1478}
1479
1480static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1481{
1482 skb->tail = skb->data;
1483}
1484
1485static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1486{
1487 skb->tail = skb->data + offset;
1488}
4305b541 1489
27a884dc
ACM
1490#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1491
1da177e4
LT
1492/*
1493 * Add data to an sk_buff
1494 */
0c7ddf36 1495unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1496unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1497static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1498{
27a884dc 1499 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1500 SKB_LINEAR_ASSERT(skb);
1501 skb->tail += len;
1502 skb->len += len;
1503 return tmp;
1504}
1505
7965bd4d 1506unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1507static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1508{
1509 skb->data -= len;
1510 skb->len += len;
1511 return skb->data;
1512}
1513
7965bd4d 1514unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1515static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1516{
1517 skb->len -= len;
1518 BUG_ON(skb->len < skb->data_len);
1519 return skb->data += len;
1520}
1521
47d29646
DM
1522static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1523{
1524 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1525}
1526
7965bd4d 1527unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1528
1529static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1530{
1531 if (len > skb_headlen(skb) &&
987c402a 1532 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1533 return NULL;
1534 skb->len -= len;
1535 return skb->data += len;
1536}
1537
1538static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1539{
1540 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1541}
1542
1543static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1544{
1545 if (likely(len <= skb_headlen(skb)))
1546 return 1;
1547 if (unlikely(len > skb->len))
1548 return 0;
987c402a 1549 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1550}
1551
1552/**
1553 * skb_headroom - bytes at buffer head
1554 * @skb: buffer to check
1555 *
1556 * Return the number of bytes of free space at the head of an &sk_buff.
1557 */
c2636b4d 1558static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1559{
1560 return skb->data - skb->head;
1561}
1562
1563/**
1564 * skb_tailroom - bytes at buffer end
1565 * @skb: buffer to check
1566 *
1567 * Return the number of bytes of free space at the tail of an sk_buff
1568 */
1569static inline int skb_tailroom(const struct sk_buff *skb)
1570{
4305b541 1571 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1572}
1573
a21d4572
ED
1574/**
1575 * skb_availroom - bytes at buffer end
1576 * @skb: buffer to check
1577 *
1578 * Return the number of bytes of free space at the tail of an sk_buff
1579 * allocated by sk_stream_alloc()
1580 */
1581static inline int skb_availroom(const struct sk_buff *skb)
1582{
16fad69c
ED
1583 if (skb_is_nonlinear(skb))
1584 return 0;
1585
1586 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1587}
1588
1da177e4
LT
1589/**
1590 * skb_reserve - adjust headroom
1591 * @skb: buffer to alter
1592 * @len: bytes to move
1593 *
1594 * Increase the headroom of an empty &sk_buff by reducing the tail
1595 * room. This is only allowed for an empty buffer.
1596 */
8243126c 1597static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1598{
1599 skb->data += len;
1600 skb->tail += len;
1601}
1602
6a674e9c
JG
1603static inline void skb_reset_inner_headers(struct sk_buff *skb)
1604{
aefbd2b3 1605 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1606 skb->inner_network_header = skb->network_header;
1607 skb->inner_transport_header = skb->transport_header;
1608}
1609
0b5c9db1
JP
1610static inline void skb_reset_mac_len(struct sk_buff *skb)
1611{
1612 skb->mac_len = skb->network_header - skb->mac_header;
1613}
1614
6a674e9c
JG
1615static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1616 *skb)
1617{
1618 return skb->head + skb->inner_transport_header;
1619}
1620
1621static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1622{
1623 skb->inner_transport_header = skb->data - skb->head;
1624}
1625
1626static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1627 const int offset)
1628{
1629 skb_reset_inner_transport_header(skb);
1630 skb->inner_transport_header += offset;
1631}
1632
1633static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1634{
1635 return skb->head + skb->inner_network_header;
1636}
1637
1638static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1639{
1640 skb->inner_network_header = skb->data - skb->head;
1641}
1642
1643static inline void skb_set_inner_network_header(struct sk_buff *skb,
1644 const int offset)
1645{
1646 skb_reset_inner_network_header(skb);
1647 skb->inner_network_header += offset;
1648}
1649
aefbd2b3
PS
1650static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1651{
1652 return skb->head + skb->inner_mac_header;
1653}
1654
1655static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1656{
1657 skb->inner_mac_header = skb->data - skb->head;
1658}
1659
1660static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1661 const int offset)
1662{
1663 skb_reset_inner_mac_header(skb);
1664 skb->inner_mac_header += offset;
1665}
fda55eca
ED
1666static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1667{
35d04610 1668 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1669}
1670
9c70220b
ACM
1671static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1672{
2e07fa9c 1673 return skb->head + skb->transport_header;
9c70220b
ACM
1674}
1675
badff6d0
ACM
1676static inline void skb_reset_transport_header(struct sk_buff *skb)
1677{
2e07fa9c 1678 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1679}
1680
967b05f6
ACM
1681static inline void skb_set_transport_header(struct sk_buff *skb,
1682 const int offset)
1683{
2e07fa9c
ACM
1684 skb_reset_transport_header(skb);
1685 skb->transport_header += offset;
ea2ae17d
ACM
1686}
1687
d56f90a7
ACM
1688static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1689{
2e07fa9c 1690 return skb->head + skb->network_header;
d56f90a7
ACM
1691}
1692
c1d2bbe1
ACM
1693static inline void skb_reset_network_header(struct sk_buff *skb)
1694{
2e07fa9c 1695 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1696}
1697
c14d2450
ACM
1698static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1699{
2e07fa9c
ACM
1700 skb_reset_network_header(skb);
1701 skb->network_header += offset;
c14d2450
ACM
1702}
1703
2e07fa9c 1704static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1705{
2e07fa9c 1706 return skb->head + skb->mac_header;
bbe735e4
ACM
1707}
1708
2e07fa9c 1709static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1710{
35d04610 1711 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1712}
1713
1714static inline void skb_reset_mac_header(struct sk_buff *skb)
1715{
1716 skb->mac_header = skb->data - skb->head;
1717}
1718
1719static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1720{
1721 skb_reset_mac_header(skb);
1722 skb->mac_header += offset;
1723}
1724
0e3da5bb
TT
1725static inline void skb_pop_mac_header(struct sk_buff *skb)
1726{
1727 skb->mac_header = skb->network_header;
1728}
1729
fbbdb8f0
YX
1730static inline void skb_probe_transport_header(struct sk_buff *skb,
1731 const int offset_hint)
1732{
1733 struct flow_keys keys;
1734
1735 if (skb_transport_header_was_set(skb))
1736 return;
1737 else if (skb_flow_dissect(skb, &keys))
1738 skb_set_transport_header(skb, keys.thoff);
1739 else
1740 skb_set_transport_header(skb, offset_hint);
1741}
1742
03606895
ED
1743static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1744{
1745 if (skb_mac_header_was_set(skb)) {
1746 const unsigned char *old_mac = skb_mac_header(skb);
1747
1748 skb_set_mac_header(skb, -skb->mac_len);
1749 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1750 }
1751}
1752
04fb451e
MM
1753static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1754{
1755 return skb->csum_start - skb_headroom(skb);
1756}
1757
2e07fa9c
ACM
1758static inline int skb_transport_offset(const struct sk_buff *skb)
1759{
1760 return skb_transport_header(skb) - skb->data;
1761}
1762
1763static inline u32 skb_network_header_len(const struct sk_buff *skb)
1764{
1765 return skb->transport_header - skb->network_header;
1766}
1767
6a674e9c
JG
1768static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1769{
1770 return skb->inner_transport_header - skb->inner_network_header;
1771}
1772
2e07fa9c
ACM
1773static inline int skb_network_offset(const struct sk_buff *skb)
1774{
1775 return skb_network_header(skb) - skb->data;
1776}
48d49d0c 1777
6a674e9c
JG
1778static inline int skb_inner_network_offset(const struct sk_buff *skb)
1779{
1780 return skb_inner_network_header(skb) - skb->data;
1781}
1782
f9599ce1
CG
1783static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1784{
1785 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1786}
1787
1da177e4
LT
1788/*
1789 * CPUs often take a performance hit when accessing unaligned memory
1790 * locations. The actual performance hit varies, it can be small if the
1791 * hardware handles it or large if we have to take an exception and fix it
1792 * in software.
1793 *
1794 * Since an ethernet header is 14 bytes network drivers often end up with
1795 * the IP header at an unaligned offset. The IP header can be aligned by
1796 * shifting the start of the packet by 2 bytes. Drivers should do this
1797 * with:
1798 *
8660c124 1799 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1800 *
1801 * The downside to this alignment of the IP header is that the DMA is now
1802 * unaligned. On some architectures the cost of an unaligned DMA is high
1803 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1804 *
1da177e4
LT
1805 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1806 * to be overridden.
1807 */
1808#ifndef NET_IP_ALIGN
1809#define NET_IP_ALIGN 2
1810#endif
1811
025be81e
AB
1812/*
1813 * The networking layer reserves some headroom in skb data (via
1814 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1815 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1816 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1817 *
1818 * Unfortunately this headroom changes the DMA alignment of the resulting
1819 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1820 * on some architectures. An architecture can override this value,
1821 * perhaps setting it to a cacheline in size (since that will maintain
1822 * cacheline alignment of the DMA). It must be a power of 2.
1823 *
d6301d3d 1824 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1825 * headroom, you should not reduce this.
5933dd2f
ED
1826 *
1827 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1828 * to reduce average number of cache lines per packet.
1829 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1830 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1831 */
1832#ifndef NET_SKB_PAD
5933dd2f 1833#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1834#endif
1835
7965bd4d 1836int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1837
1838static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1839{
c4264f27 1840 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1841 WARN_ON(1);
1842 return;
1843 }
27a884dc
ACM
1844 skb->len = len;
1845 skb_set_tail_pointer(skb, len);
1da177e4
LT
1846}
1847
7965bd4d 1848void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1849
1850static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1851{
3cc0e873
HX
1852 if (skb->data_len)
1853 return ___pskb_trim(skb, len);
1854 __skb_trim(skb, len);
1855 return 0;
1da177e4
LT
1856}
1857
1858static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1859{
1860 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1861}
1862
e9fa4f7b
HX
1863/**
1864 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1865 * @skb: buffer to alter
1866 * @len: new length
1867 *
1868 * This is identical to pskb_trim except that the caller knows that
1869 * the skb is not cloned so we should never get an error due to out-
1870 * of-memory.
1871 */
1872static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1873{
1874 int err = pskb_trim(skb, len);
1875 BUG_ON(err);
1876}
1877
1da177e4
LT
1878/**
1879 * skb_orphan - orphan a buffer
1880 * @skb: buffer to orphan
1881 *
1882 * If a buffer currently has an owner then we call the owner's
1883 * destructor function and make the @skb unowned. The buffer continues
1884 * to exist but is no longer charged to its former owner.
1885 */
1886static inline void skb_orphan(struct sk_buff *skb)
1887{
c34a7612 1888 if (skb->destructor) {
1da177e4 1889 skb->destructor(skb);
c34a7612
ED
1890 skb->destructor = NULL;
1891 skb->sk = NULL;
376c7311
ED
1892 } else {
1893 BUG_ON(skb->sk);
c34a7612 1894 }
1da177e4
LT
1895}
1896
a353e0ce
MT
1897/**
1898 * skb_orphan_frags - orphan the frags contained in a buffer
1899 * @skb: buffer to orphan frags from
1900 * @gfp_mask: allocation mask for replacement pages
1901 *
1902 * For each frag in the SKB which needs a destructor (i.e. has an
1903 * owner) create a copy of that frag and release the original
1904 * page by calling the destructor.
1905 */
1906static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1907{
1908 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1909 return 0;
1910 return skb_copy_ubufs(skb, gfp_mask);
1911}
1912
1da177e4
LT
1913/**
1914 * __skb_queue_purge - empty a list
1915 * @list: list to empty
1916 *
1917 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1918 * the list and one reference dropped. This function does not take the
1919 * list lock and the caller must hold the relevant locks to use it.
1920 */
7965bd4d 1921void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
1922static inline void __skb_queue_purge(struct sk_buff_head *list)
1923{
1924 struct sk_buff *skb;
1925 while ((skb = __skb_dequeue(list)) != NULL)
1926 kfree_skb(skb);
1927}
1928
e5e67305
AD
1929#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1930#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1931#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1932
7965bd4d 1933void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 1934
7965bd4d
JP
1935struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1936 gfp_t gfp_mask);
8af27456
CH
1937
1938/**
1939 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1940 * @dev: network device to receive on
1941 * @length: length to allocate
1942 *
1943 * Allocate a new &sk_buff and assign it a usage count of one. The
1944 * buffer has unspecified headroom built in. Users should allocate
1945 * the headroom they think they need without accounting for the
1946 * built in space. The built in space is used for optimisations.
1947 *
1948 * %NULL is returned if there is no free memory. Although this function
1949 * allocates memory it can be called from an interrupt.
1950 */
1951static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 1952 unsigned int length)
8af27456
CH
1953{
1954 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1955}
1956
6f532612
ED
1957/* legacy helper around __netdev_alloc_skb() */
1958static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1959 gfp_t gfp_mask)
1960{
1961 return __netdev_alloc_skb(NULL, length, gfp_mask);
1962}
1963
1964/* legacy helper around netdev_alloc_skb() */
1965static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1966{
1967 return netdev_alloc_skb(NULL, length);
1968}
1969
1970
4915a0de
ED
1971static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1972 unsigned int length, gfp_t gfp)
61321bbd 1973{
4915a0de 1974 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
1975
1976 if (NET_IP_ALIGN && skb)
1977 skb_reserve(skb, NET_IP_ALIGN);
1978 return skb;
1979}
1980
4915a0de
ED
1981static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1982 unsigned int length)
1983{
1984 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1985}
1986
bc6fc9fa
FF
1987/**
1988 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
1989 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1990 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1991 * @order: size of the allocation
1992 *
1993 * Allocate a new page.
1994 *
1995 * %NULL is returned if there is no free memory.
1996*/
1997static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1998 struct sk_buff *skb,
1999 unsigned int order)
2000{
2001 struct page *page;
2002
2003 gfp_mask |= __GFP_COLD;
2004
2005 if (!(gfp_mask & __GFP_NOMEMALLOC))
2006 gfp_mask |= __GFP_MEMALLOC;
2007
2008 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2009 if (skb && page && page->pfmemalloc)
2010 skb->pfmemalloc = true;
2011
2012 return page;
2013}
2014
2015/**
2016 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2017 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2018 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2019 *
2020 * Allocate a new page.
2021 *
2022 * %NULL is returned if there is no free memory.
2023 */
2024static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2025 struct sk_buff *skb)
2026{
2027 return __skb_alloc_pages(gfp_mask, skb, 0);
2028}
2029
2030/**
2031 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2032 * @page: The page that was allocated from skb_alloc_page
2033 * @skb: The skb that may need pfmemalloc set
2034 */
2035static inline void skb_propagate_pfmemalloc(struct page *page,
2036 struct sk_buff *skb)
2037{
2038 if (page && page->pfmemalloc)
2039 skb->pfmemalloc = true;
2040}
2041
131ea667
IC
2042/**
2043 * skb_frag_page - retrieve the page refered to by a paged fragment
2044 * @frag: the paged fragment
2045 *
2046 * Returns the &struct page associated with @frag.
2047 */
2048static inline struct page *skb_frag_page(const skb_frag_t *frag)
2049{
a8605c60 2050 return frag->page.p;
131ea667
IC
2051}
2052
2053/**
2054 * __skb_frag_ref - take an addition reference on a paged fragment.
2055 * @frag: the paged fragment
2056 *
2057 * Takes an additional reference on the paged fragment @frag.
2058 */
2059static inline void __skb_frag_ref(skb_frag_t *frag)
2060{
2061 get_page(skb_frag_page(frag));
2062}
2063
2064/**
2065 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2066 * @skb: the buffer
2067 * @f: the fragment offset.
2068 *
2069 * Takes an additional reference on the @f'th paged fragment of @skb.
2070 */
2071static inline void skb_frag_ref(struct sk_buff *skb, int f)
2072{
2073 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2074}
2075
2076/**
2077 * __skb_frag_unref - release a reference on a paged fragment.
2078 * @frag: the paged fragment
2079 *
2080 * Releases a reference on the paged fragment @frag.
2081 */
2082static inline void __skb_frag_unref(skb_frag_t *frag)
2083{
2084 put_page(skb_frag_page(frag));
2085}
2086
2087/**
2088 * skb_frag_unref - release a reference on a paged fragment of an skb.
2089 * @skb: the buffer
2090 * @f: the fragment offset
2091 *
2092 * Releases a reference on the @f'th paged fragment of @skb.
2093 */
2094static inline void skb_frag_unref(struct sk_buff *skb, int f)
2095{
2096 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2097}
2098
2099/**
2100 * skb_frag_address - gets the address of the data contained in a paged fragment
2101 * @frag: the paged fragment buffer
2102 *
2103 * Returns the address of the data within @frag. The page must already
2104 * be mapped.
2105 */
2106static inline void *skb_frag_address(const skb_frag_t *frag)
2107{
2108 return page_address(skb_frag_page(frag)) + frag->page_offset;
2109}
2110
2111/**
2112 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2113 * @frag: the paged fragment buffer
2114 *
2115 * Returns the address of the data within @frag. Checks that the page
2116 * is mapped and returns %NULL otherwise.
2117 */
2118static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2119{
2120 void *ptr = page_address(skb_frag_page(frag));
2121 if (unlikely(!ptr))
2122 return NULL;
2123
2124 return ptr + frag->page_offset;
2125}
2126
2127/**
2128 * __skb_frag_set_page - sets the page contained in a paged fragment
2129 * @frag: the paged fragment
2130 * @page: the page to set
2131 *
2132 * Sets the fragment @frag to contain @page.
2133 */
2134static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2135{
a8605c60 2136 frag->page.p = page;
131ea667
IC
2137}
2138
2139/**
2140 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2141 * @skb: the buffer
2142 * @f: the fragment offset
2143 * @page: the page to set
2144 *
2145 * Sets the @f'th fragment of @skb to contain @page.
2146 */
2147static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2148 struct page *page)
2149{
2150 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2151}
2152
400dfd3a
ED
2153bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2154
131ea667
IC
2155/**
2156 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2157 * @dev: the device to map the fragment to
131ea667
IC
2158 * @frag: the paged fragment to map
2159 * @offset: the offset within the fragment (starting at the
2160 * fragment's own offset)
2161 * @size: the number of bytes to map
f83347df 2162 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2163 *
2164 * Maps the page associated with @frag to @device.
2165 */
2166static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2167 const skb_frag_t *frag,
2168 size_t offset, size_t size,
2169 enum dma_data_direction dir)
2170{
2171 return dma_map_page(dev, skb_frag_page(frag),
2172 frag->page_offset + offset, size, dir);
2173}
2174
117632e6
ED
2175static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2176 gfp_t gfp_mask)
2177{
2178 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2179}
2180
334a8132
PM
2181/**
2182 * skb_clone_writable - is the header of a clone writable
2183 * @skb: buffer to check
2184 * @len: length up to which to write
2185 *
2186 * Returns true if modifying the header part of the cloned buffer
2187 * does not requires the data to be copied.
2188 */
05bdd2f1 2189static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2190{
2191 return !skb_header_cloned(skb) &&
2192 skb_headroom(skb) + len <= skb->hdr_len;
2193}
2194
d9cc2048
HX
2195static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2196 int cloned)
2197{
2198 int delta = 0;
2199
d9cc2048
HX
2200 if (headroom > skb_headroom(skb))
2201 delta = headroom - skb_headroom(skb);
2202
2203 if (delta || cloned)
2204 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2205 GFP_ATOMIC);
2206 return 0;
2207}
2208
1da177e4
LT
2209/**
2210 * skb_cow - copy header of skb when it is required
2211 * @skb: buffer to cow
2212 * @headroom: needed headroom
2213 *
2214 * If the skb passed lacks sufficient headroom or its data part
2215 * is shared, data is reallocated. If reallocation fails, an error
2216 * is returned and original skb is not changed.
2217 *
2218 * The result is skb with writable area skb->head...skb->tail
2219 * and at least @headroom of space at head.
2220 */
2221static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2222{
d9cc2048
HX
2223 return __skb_cow(skb, headroom, skb_cloned(skb));
2224}
1da177e4 2225
d9cc2048
HX
2226/**
2227 * skb_cow_head - skb_cow but only making the head writable
2228 * @skb: buffer to cow
2229 * @headroom: needed headroom
2230 *
2231 * This function is identical to skb_cow except that we replace the
2232 * skb_cloned check by skb_header_cloned. It should be used when
2233 * you only need to push on some header and do not need to modify
2234 * the data.
2235 */
2236static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2237{
2238 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2239}
2240
2241/**
2242 * skb_padto - pad an skbuff up to a minimal size
2243 * @skb: buffer to pad
2244 * @len: minimal length
2245 *
2246 * Pads up a buffer to ensure the trailing bytes exist and are
2247 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2248 * is untouched. Otherwise it is extended. Returns zero on
2249 * success. The skb is freed on error.
1da177e4
LT
2250 */
2251
5b057c6b 2252static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2253{
2254 unsigned int size = skb->len;
2255 if (likely(size >= len))
5b057c6b 2256 return 0;
987c402a 2257 return skb_pad(skb, len - size);
1da177e4
LT
2258}
2259
2260static inline int skb_add_data(struct sk_buff *skb,
2261 char __user *from, int copy)
2262{
2263 const int off = skb->len;
2264
2265 if (skb->ip_summed == CHECKSUM_NONE) {
2266 int err = 0;
5084205f 2267 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2268 copy, 0, &err);
2269 if (!err) {
2270 skb->csum = csum_block_add(skb->csum, csum, off);
2271 return 0;
2272 }
2273 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2274 return 0;
2275
2276 __skb_trim(skb, off);
2277 return -EFAULT;
2278}
2279
38ba0a65
ED
2280static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2281 const struct page *page, int off)
1da177e4
LT
2282{
2283 if (i) {
9e903e08 2284 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2285
ea2ab693 2286 return page == skb_frag_page(frag) &&
9e903e08 2287 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2288 }
38ba0a65 2289 return false;
1da177e4
LT
2290}
2291
364c6bad
HX
2292static inline int __skb_linearize(struct sk_buff *skb)
2293{
2294 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2295}
2296
1da177e4
LT
2297/**
2298 * skb_linearize - convert paged skb to linear one
2299 * @skb: buffer to linarize
1da177e4
LT
2300 *
2301 * If there is no free memory -ENOMEM is returned, otherwise zero
2302 * is returned and the old skb data released.
2303 */
364c6bad
HX
2304static inline int skb_linearize(struct sk_buff *skb)
2305{
2306 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2307}
2308
cef401de
ED
2309/**
2310 * skb_has_shared_frag - can any frag be overwritten
2311 * @skb: buffer to test
2312 *
2313 * Return true if the skb has at least one frag that might be modified
2314 * by an external entity (as in vmsplice()/sendfile())
2315 */
2316static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2317{
c9af6db4
PS
2318 return skb_is_nonlinear(skb) &&
2319 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2320}
2321
364c6bad
HX
2322/**
2323 * skb_linearize_cow - make sure skb is linear and writable
2324 * @skb: buffer to process
2325 *
2326 * If there is no free memory -ENOMEM is returned, otherwise zero
2327 * is returned and the old skb data released.
2328 */
2329static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2330{
364c6bad
HX
2331 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2332 __skb_linearize(skb) : 0;
1da177e4
LT
2333}
2334
2335/**
2336 * skb_postpull_rcsum - update checksum for received skb after pull
2337 * @skb: buffer to update
2338 * @start: start of data before pull
2339 * @len: length of data pulled
2340 *
2341 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2342 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2343 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2344 */
2345
2346static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2347 const void *start, unsigned int len)
1da177e4 2348{
84fa7933 2349 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2350 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2351}
2352
cbb042f9
HX
2353unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2354
7ce5a27f
DM
2355/**
2356 * pskb_trim_rcsum - trim received skb and update checksum
2357 * @skb: buffer to trim
2358 * @len: new length
2359 *
2360 * This is exactly the same as pskb_trim except that it ensures the
2361 * checksum of received packets are still valid after the operation.
2362 */
2363
2364static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2365{
2366 if (likely(len >= skb->len))
2367 return 0;
2368 if (skb->ip_summed == CHECKSUM_COMPLETE)
2369 skb->ip_summed = CHECKSUM_NONE;
2370 return __pskb_trim(skb, len);
2371}
2372
1da177e4
LT
2373#define skb_queue_walk(queue, skb) \
2374 for (skb = (queue)->next; \
a1e4891f 2375 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2376 skb = skb->next)
2377
46f8914e
JC
2378#define skb_queue_walk_safe(queue, skb, tmp) \
2379 for (skb = (queue)->next, tmp = skb->next; \
2380 skb != (struct sk_buff *)(queue); \
2381 skb = tmp, tmp = skb->next)
2382
1164f52a 2383#define skb_queue_walk_from(queue, skb) \
a1e4891f 2384 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2385 skb = skb->next)
2386
2387#define skb_queue_walk_from_safe(queue, skb, tmp) \
2388 for (tmp = skb->next; \
2389 skb != (struct sk_buff *)(queue); \
2390 skb = tmp, tmp = skb->next)
2391
300ce174
SH
2392#define skb_queue_reverse_walk(queue, skb) \
2393 for (skb = (queue)->prev; \
a1e4891f 2394 skb != (struct sk_buff *)(queue); \
300ce174
SH
2395 skb = skb->prev)
2396
686a2955
DM
2397#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2398 for (skb = (queue)->prev, tmp = skb->prev; \
2399 skb != (struct sk_buff *)(queue); \
2400 skb = tmp, tmp = skb->prev)
2401
2402#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2403 for (tmp = skb->prev; \
2404 skb != (struct sk_buff *)(queue); \
2405 skb = tmp, tmp = skb->prev)
1da177e4 2406
21dc3301 2407static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2408{
2409 return skb_shinfo(skb)->frag_list != NULL;
2410}
2411
2412static inline void skb_frag_list_init(struct sk_buff *skb)
2413{
2414 skb_shinfo(skb)->frag_list = NULL;
2415}
2416
2417static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2418{
2419 frag->next = skb_shinfo(skb)->frag_list;
2420 skb_shinfo(skb)->frag_list = frag;
2421}
2422
2423#define skb_walk_frags(skb, iter) \
2424 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2425
7965bd4d
JP
2426struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2427 int *peeked, int *off, int *err);
2428struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2429 int *err);
2430unsigned int datagram_poll(struct file *file, struct socket *sock,
2431 struct poll_table_struct *wait);
2432int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2433 struct iovec *to, int size);
2434int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2435 struct iovec *iov);
2436int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2437 const struct iovec *from, int from_offset,
2438 int len);
2439int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2440 int offset, size_t count);
2441int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2442 const struct iovec *to, int to_offset,
2443 int size);
2444void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2445void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2446int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2447int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2448int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2449__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2450 int len, __wsum csum);
2451int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2452 struct pipe_inode_info *pipe, unsigned int len,
2453 unsigned int flags);
2454void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8
TG
2455unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2456void skb_zerocopy(struct sk_buff *to, const struct sk_buff *from,
2457 int len, int hlen);
7965bd4d
JP
2458void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2459int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2460void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2461unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2462struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2463
2817a336
DB
2464struct skb_checksum_ops {
2465 __wsum (*update)(const void *mem, int len, __wsum wsum);
2466 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2467};
2468
2469__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2470 __wsum csum, const struct skb_checksum_ops *ops);
2471__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2472 __wsum csum);
2473
1da177e4
LT
2474static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2475 int len, void *buffer)
2476{
2477 int hlen = skb_headlen(skb);
2478
55820ee2 2479 if (hlen - offset >= len)
1da177e4
LT
2480 return skb->data + offset;
2481
2482 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2483 return NULL;
2484
2485 return buffer;
2486}
2487
4262e5cc
DB
2488/**
2489 * skb_needs_linearize - check if we need to linearize a given skb
2490 * depending on the given device features.
2491 * @skb: socket buffer to check
2492 * @features: net device features
2493 *
2494 * Returns true if either:
2495 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2496 * 2. skb is fragmented and the device does not support SG.
2497 */
2498static inline bool skb_needs_linearize(struct sk_buff *skb,
2499 netdev_features_t features)
2500{
2501 return skb_is_nonlinear(skb) &&
2502 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2503 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2504}
2505
d626f62b
ACM
2506static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2507 void *to,
2508 const unsigned int len)
2509{
2510 memcpy(to, skb->data, len);
2511}
2512
2513static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2514 const int offset, void *to,
2515 const unsigned int len)
2516{
2517 memcpy(to, skb->data + offset, len);
2518}
2519
27d7ff46
ACM
2520static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2521 const void *from,
2522 const unsigned int len)
2523{
2524 memcpy(skb->data, from, len);
2525}
2526
2527static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2528 const int offset,
2529 const void *from,
2530 const unsigned int len)
2531{
2532 memcpy(skb->data + offset, from, len);
2533}
2534
7965bd4d 2535void skb_init(void);
1da177e4 2536
ac45f602
PO
2537static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2538{
2539 return skb->tstamp;
2540}
2541
a61bbcf2
PM
2542/**
2543 * skb_get_timestamp - get timestamp from a skb
2544 * @skb: skb to get stamp from
2545 * @stamp: pointer to struct timeval to store stamp in
2546 *
2547 * Timestamps are stored in the skb as offsets to a base timestamp.
2548 * This function converts the offset back to a struct timeval and stores
2549 * it in stamp.
2550 */
ac45f602
PO
2551static inline void skb_get_timestamp(const struct sk_buff *skb,
2552 struct timeval *stamp)
a61bbcf2 2553{
b7aa0bf7 2554 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2555}
2556
ac45f602
PO
2557static inline void skb_get_timestampns(const struct sk_buff *skb,
2558 struct timespec *stamp)
2559{
2560 *stamp = ktime_to_timespec(skb->tstamp);
2561}
2562
b7aa0bf7 2563static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2564{
b7aa0bf7 2565 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2566}
2567
164891aa
SH
2568static inline ktime_t net_timedelta(ktime_t t)
2569{
2570 return ktime_sub(ktime_get_real(), t);
2571}
2572
b9ce204f
IJ
2573static inline ktime_t net_invalid_timestamp(void)
2574{
2575 return ktime_set(0, 0);
2576}
a61bbcf2 2577
7965bd4d 2578void skb_timestamping_init(void);
c1f19b51
RC
2579
2580#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2581
7965bd4d
JP
2582void skb_clone_tx_timestamp(struct sk_buff *skb);
2583bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2584
2585#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2586
2587static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2588{
2589}
2590
2591static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2592{
2593 return false;
2594}
2595
2596#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2597
2598/**
2599 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2600 *
da92b194
RC
2601 * PHY drivers may accept clones of transmitted packets for
2602 * timestamping via their phy_driver.txtstamp method. These drivers
2603 * must call this function to return the skb back to the stack, with
2604 * or without a timestamp.
2605 *
c1f19b51 2606 * @skb: clone of the the original outgoing packet
da92b194 2607 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2608 *
2609 */
2610void skb_complete_tx_timestamp(struct sk_buff *skb,
2611 struct skb_shared_hwtstamps *hwtstamps);
2612
ac45f602
PO
2613/**
2614 * skb_tstamp_tx - queue clone of skb with send time stamps
2615 * @orig_skb: the original outgoing packet
2616 * @hwtstamps: hardware time stamps, may be NULL if not available
2617 *
2618 * If the skb has a socket associated, then this function clones the
2619 * skb (thus sharing the actual data and optional structures), stores
2620 * the optional hardware time stamping information (if non NULL) or
2621 * generates a software time stamp (otherwise), then queues the clone
2622 * to the error queue of the socket. Errors are silently ignored.
2623 */
7965bd4d
JP
2624void skb_tstamp_tx(struct sk_buff *orig_skb,
2625 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2626
4507a715
RC
2627static inline void sw_tx_timestamp(struct sk_buff *skb)
2628{
2244d07b
OH
2629 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2630 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2631 skb_tstamp_tx(skb, NULL);
2632}
2633
2634/**
2635 * skb_tx_timestamp() - Driver hook for transmit timestamping
2636 *
2637 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2638 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2639 *
73409f3b
DM
2640 * Specifically, one should make absolutely sure that this function is
2641 * called before TX completion of this packet can trigger. Otherwise
2642 * the packet could potentially already be freed.
2643 *
4507a715
RC
2644 * @skb: A socket buffer.
2645 */
2646static inline void skb_tx_timestamp(struct sk_buff *skb)
2647{
c1f19b51 2648 skb_clone_tx_timestamp(skb);
4507a715
RC
2649 sw_tx_timestamp(skb);
2650}
2651
6e3e939f
JB
2652/**
2653 * skb_complete_wifi_ack - deliver skb with wifi status
2654 *
2655 * @skb: the original outgoing packet
2656 * @acked: ack status
2657 *
2658 */
2659void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2660
7965bd4d
JP
2661__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2662__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2663
60476372
HX
2664static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2665{
2666 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2667}
2668
fb286bb2
HX
2669/**
2670 * skb_checksum_complete - Calculate checksum of an entire packet
2671 * @skb: packet to process
2672 *
2673 * This function calculates the checksum over the entire packet plus
2674 * the value of skb->csum. The latter can be used to supply the
2675 * checksum of a pseudo header as used by TCP/UDP. It returns the
2676 * checksum.
2677 *
2678 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2679 * this function can be used to verify that checksum on received
2680 * packets. In that case the function should return zero if the
2681 * checksum is correct. In particular, this function will return zero
2682 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2683 * hardware has already verified the correctness of the checksum.
2684 */
4381ca3c 2685static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2686{
60476372
HX
2687 return skb_csum_unnecessary(skb) ?
2688 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2689}
2690
5f79e0f9 2691#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2692void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2693static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2694{
2695 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2696 nf_conntrack_destroy(nfct);
1da177e4
LT
2697}
2698static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2699{
2700 if (nfct)
2701 atomic_inc(&nfct->use);
2702}
2fc72c7b 2703#endif
1da177e4
LT
2704#ifdef CONFIG_BRIDGE_NETFILTER
2705static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2706{
2707 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2708 kfree(nf_bridge);
2709}
2710static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2711{
2712 if (nf_bridge)
2713 atomic_inc(&nf_bridge->use);
2714}
2715#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2716static inline void nf_reset(struct sk_buff *skb)
2717{
5f79e0f9 2718#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2719 nf_conntrack_put(skb->nfct);
2720 skb->nfct = NULL;
2fc72c7b 2721#endif
a193a4ab
PM
2722#ifdef CONFIG_BRIDGE_NETFILTER
2723 nf_bridge_put(skb->nf_bridge);
2724 skb->nf_bridge = NULL;
2725#endif
2726}
2727
124dff01
PM
2728static inline void nf_reset_trace(struct sk_buff *skb)
2729{
130549fe
G
2730#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2731 skb->nf_trace = 0;
2732#endif
a193a4ab
PM
2733}
2734
edda553c
YK
2735/* Note: This doesn't put any conntrack and bridge info in dst. */
2736static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2737{
5f79e0f9 2738#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2739 dst->nfct = src->nfct;
2740 nf_conntrack_get(src->nfct);
2741 dst->nfctinfo = src->nfctinfo;
2fc72c7b 2742#endif
edda553c
YK
2743#ifdef CONFIG_BRIDGE_NETFILTER
2744 dst->nf_bridge = src->nf_bridge;
2745 nf_bridge_get(src->nf_bridge);
2746#endif
2747}
2748
e7ac05f3
YK
2749static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2750{
e7ac05f3 2751#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2752 nf_conntrack_put(dst->nfct);
2fc72c7b 2753#endif
e7ac05f3
YK
2754#ifdef CONFIG_BRIDGE_NETFILTER
2755 nf_bridge_put(dst->nf_bridge);
2756#endif
2757 __nf_copy(dst, src);
2758}
2759
984bc16c
JM
2760#ifdef CONFIG_NETWORK_SECMARK
2761static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2762{
2763 to->secmark = from->secmark;
2764}
2765
2766static inline void skb_init_secmark(struct sk_buff *skb)
2767{
2768 skb->secmark = 0;
2769}
2770#else
2771static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2772{ }
2773
2774static inline void skb_init_secmark(struct sk_buff *skb)
2775{ }
2776#endif
2777
f25f4e44
PWJ
2778static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2779{
f25f4e44 2780 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2781}
2782
9247744e 2783static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2784{
4e3ab47a 2785 return skb->queue_mapping;
4e3ab47a
PE
2786}
2787
f25f4e44
PWJ
2788static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2789{
f25f4e44 2790 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2791}
2792
d5a9e24a
DM
2793static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2794{
2795 skb->queue_mapping = rx_queue + 1;
2796}
2797
9247744e 2798static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
2799{
2800 return skb->queue_mapping - 1;
2801}
2802
9247744e 2803static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 2804{
a02cec21 2805 return skb->queue_mapping != 0;
d5a9e24a
DM
2806}
2807
7965bd4d
JP
2808u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2809 unsigned int num_tx_queues);
9247744e 2810
def8b4fa
AD
2811static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2812{
0b3d8e08 2813#ifdef CONFIG_XFRM
def8b4fa 2814 return skb->sp;
def8b4fa 2815#else
def8b4fa 2816 return NULL;
def8b4fa 2817#endif
0b3d8e08 2818}
def8b4fa 2819
68c33163
PS
2820/* Keeps track of mac header offset relative to skb->head.
2821 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2822 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
2823 * tunnel skb it points to outer mac header.
2824 * Keeps track of level of encapsulation of network headers.
2825 */
68c33163 2826struct skb_gso_cb {
3347c960
ED
2827 int mac_offset;
2828 int encap_level;
68c33163
PS
2829};
2830#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2831
2832static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2833{
2834 return (skb_mac_header(inner_skb) - inner_skb->head) -
2835 SKB_GSO_CB(inner_skb)->mac_offset;
2836}
2837
1e2bd517
PS
2838static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2839{
2840 int new_headroom, headroom;
2841 int ret;
2842
2843 headroom = skb_headroom(skb);
2844 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2845 if (ret)
2846 return ret;
2847
2848 new_headroom = skb_headroom(skb);
2849 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2850 return 0;
2851}
2852
bdcc0924 2853static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
2854{
2855 return skb_shinfo(skb)->gso_size;
2856}
2857
36a8f39e 2858/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 2859static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
2860{
2861 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2862}
2863
7965bd4d 2864void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
2865
2866static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2867{
2868 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2869 * wanted then gso_type will be set. */
05bdd2f1
ED
2870 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2871
b78462eb
AD
2872 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2873 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
2874 __skb_warn_lro_forwarding(skb);
2875 return true;
2876 }
2877 return false;
2878}
2879
35fc92a9
HX
2880static inline void skb_forward_csum(struct sk_buff *skb)
2881{
2882 /* Unfortunately we don't support this one. Any brave souls? */
2883 if (skb->ip_summed == CHECKSUM_COMPLETE)
2884 skb->ip_summed = CHECKSUM_NONE;
2885}
2886
bc8acf2c
ED
2887/**
2888 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2889 * @skb: skb to check
2890 *
2891 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2892 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2893 * use this helper, to document places where we make this assertion.
2894 */
05bdd2f1 2895static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
2896{
2897#ifdef DEBUG
2898 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2899#endif
2900}
2901
f35d9d8a 2902bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 2903
ed1f50c3
PD
2904int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
2905
f77668dc
DB
2906u32 __skb_get_poff(const struct sk_buff *skb);
2907
3a7c1ee4
AD
2908/**
2909 * skb_head_is_locked - Determine if the skb->head is locked down
2910 * @skb: skb to check
2911 *
2912 * The head on skbs build around a head frag can be removed if they are
2913 * not cloned. This function returns true if the skb head is locked down
2914 * due to either being allocated via kmalloc, or by being a clone with
2915 * multiple references to the head.
2916 */
2917static inline bool skb_head_is_locked(const struct sk_buff *skb)
2918{
2919 return !skb->head_frag || skb_cloned(skb);
2920}
fe6cc55f
FW
2921
2922/**
2923 * skb_gso_network_seglen - Return length of individual segments of a gso packet
2924 *
2925 * @skb: GSO skb
2926 *
2927 * skb_gso_network_seglen is used to determine the real size of the
2928 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
2929 *
2930 * The MAC/L2 header is not accounted for.
2931 */
2932static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
2933{
2934 unsigned int hdr_len = skb_transport_header(skb) -
2935 skb_network_header(skb);
2936 return hdr_len + skb_gso_transport_seglen(skb);
2937}
1da177e4
LT
2938#endif /* __KERNEL__ */
2939#endif /* _LINUX_SKBUFF_H */