ipv4: generalize gre_handle_offloads
[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
60476372 37/* Don't change this without changing skb_csum_unnecessary! */
1da177e4 38#define CHECKSUM_NONE 0
60476372
HX
39#define CHECKSUM_UNNECESSARY 1
40#define CHECKSUM_COMPLETE 2
41#define CHECKSUM_PARTIAL 3
1da177e4
LT
42
43#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
44 ~(SMP_CACHE_BYTES - 1))
fc910a27 45#define SKB_WITH_OVERHEAD(X) \
deea84b0 46 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
47#define SKB_MAX_ORDER(X, ORDER) \
48 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
49#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
50#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
51
87fb4b7b
ED
52/* return minimum truesize of one skb containing X bytes of data */
53#define SKB_TRUESIZE(X) ((X) + \
54 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
55 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
56
1da177e4
LT
57/* A. Checksumming of received packets by device.
58 *
59 * NONE: device failed to checksum this packet.
60 * skb->csum is undefined.
61 *
62 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
63 * skb->csum is undefined.
64 * It is bad option, but, unfortunately, many of vendors do this.
65 * Apparently with secret goal to sell you new device, when you
66 * will add new protocol to your host. F.e. IPv6. 8)
67 *
84fa7933 68 * COMPLETE: the most generic way. Device supplied checksum of _all_
1da177e4
LT
69 * the packet as seen by netif_rx in skb->csum.
70 * NOTE: Even if device supports only some protocols, but
84fa7933 71 * is able to produce some skb->csum, it MUST use COMPLETE,
1da177e4
LT
72 * not UNNECESSARY.
73 *
c6c6e3e0
HX
74 * PARTIAL: identical to the case for output below. This may occur
75 * on a packet received directly from another Linux OS, e.g.,
76 * a virtualised Linux kernel on the same host. The packet can
77 * be treated in the same way as UNNECESSARY except that on
78 * output (i.e., forwarding) the checksum must be filled in
79 * by the OS or the hardware.
80 *
1da177e4
LT
81 * B. Checksumming on output.
82 *
83 * NONE: skb is checksummed by protocol or csum is not required.
84 *
84fa7933 85 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
c6c6e3e0
HX
86 * from skb->csum_start to the end and to record the checksum
87 * at skb->csum_start + skb->csum_offset.
1da177e4
LT
88 *
89 * Device must show its capabilities in dev->features, set
90 * at device setup time.
91 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
92 * everything.
1da177e4
LT
93 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
94 * TCP/UDP over IPv4. Sigh. Vendors like this
95 * way by an unknown reason. Though, see comment above
96 * about CHECKSUM_UNNECESSARY. 8)
c6c6e3e0 97 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
1da177e4 98 *
3af79302
YZ
99 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
100 * that do not want net to perform the checksum calculation should use
101 * this flag in their outgoing skbs.
102 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
1da177e4
LT
106 * Any questions? No questions, good. --ANK
107 */
108
1da177e4 109struct net_device;
716ea3a7 110struct scatterlist;
9c55e01c 111struct pipe_inode_info;
1da177e4 112
5f79e0f9 113#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
114struct nf_conntrack {
115 atomic_t use;
1da177e4 116};
5f79e0f9 117#endif
1da177e4
LT
118
119#ifdef CONFIG_BRIDGE_NETFILTER
120struct nf_bridge_info {
bf1ac5ca
ED
121 atomic_t use;
122 unsigned int mask;
123 struct net_device *physindev;
124 struct net_device *physoutdev;
125 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
126};
127#endif
128
1da177e4
LT
129struct sk_buff_head {
130 /* These two members must be first. */
131 struct sk_buff *next;
132 struct sk_buff *prev;
133
134 __u32 qlen;
135 spinlock_t lock;
136};
137
138struct sk_buff;
139
9d4dde52
IC
140/* To allow 64K frame to be packed as single skb without frag_list we
141 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
142 * buffers which do not start on a page boundary.
143 *
144 * Since GRO uses frags we allocate at least 16 regardless of page
145 * size.
a715dea3 146 */
9d4dde52 147#if (65536/PAGE_SIZE + 1) < 16
eec00954 148#define MAX_SKB_FRAGS 16UL
a715dea3 149#else
9d4dde52 150#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 151#endif
1da177e4
LT
152
153typedef struct skb_frag_struct skb_frag_t;
154
155struct skb_frag_struct {
a8605c60
IC
156 struct {
157 struct page *p;
158 } page;
cb4dfe56 159#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
160 __u32 page_offset;
161 __u32 size;
cb4dfe56
ED
162#else
163 __u16 page_offset;
164 __u16 size;
165#endif
1da177e4
LT
166};
167
9e903e08
ED
168static inline unsigned int skb_frag_size(const skb_frag_t *frag)
169{
170 return frag->size;
171}
172
173static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
174{
175 frag->size = size;
176}
177
178static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
179{
180 frag->size += delta;
181}
182
183static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
184{
185 frag->size -= delta;
186}
187
ac45f602
PO
188#define HAVE_HW_TIME_STAMP
189
190/**
d3a21be8 191 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
192 * @hwtstamp: hardware time stamp transformed into duration
193 * since arbitrary point in time
194 * @syststamp: hwtstamp transformed to system time base
195 *
196 * Software time stamps generated by ktime_get_real() are stored in
197 * skb->tstamp. The relation between the different kinds of time
198 * stamps is as follows:
199 *
200 * syststamp and tstamp can be compared against each other in
201 * arbitrary combinations. The accuracy of a
202 * syststamp/tstamp/"syststamp from other device" comparison is
203 * limited by the accuracy of the transformation into system time
204 * base. This depends on the device driver and its underlying
205 * hardware.
206 *
207 * hwtstamps can only be compared against other hwtstamps from
208 * the same device.
209 *
210 * This structure is attached to packets as part of the
211 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
212 */
213struct skb_shared_hwtstamps {
214 ktime_t hwtstamp;
215 ktime_t syststamp;
216};
217
2244d07b
OH
218/* Definitions for tx_flags in struct skb_shared_info */
219enum {
220 /* generate hardware time stamp */
221 SKBTX_HW_TSTAMP = 1 << 0,
222
223 /* generate software time stamp */
224 SKBTX_SW_TSTAMP = 1 << 1,
225
226 /* device driver is going to provide hardware time stamp */
227 SKBTX_IN_PROGRESS = 1 << 2,
228
a6686f2f 229 /* device driver supports TX zero-copy buffers */
62b1a8ab 230 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
231
232 /* generate wifi status information (where possible) */
62b1a8ab 233 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
234
235 /* This indicates at least one fragment might be overwritten
236 * (as in vmsplice(), sendfile() ...)
237 * If we need to compute a TX checksum, we'll need to copy
238 * all frags to avoid possible bad checksum
239 */
240 SKBTX_SHARED_FRAG = 1 << 5,
a6686f2f
SM
241};
242
243/*
244 * The callback notifies userspace to release buffers when skb DMA is done in
245 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
246 * The zerocopy_success argument is true if zero copy transmit occurred,
247 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
248 * The ctx field is used to track device context.
249 * The desc field is used to track userspace buffer index.
a6686f2f
SM
250 */
251struct ubuf_info {
e19d6763 252 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 253 void *ctx;
a6686f2f 254 unsigned long desc;
ac45f602
PO
255};
256
1da177e4
LT
257/* This data is invariant across clones and lives at
258 * the end of the header data, ie. at skb->end.
259 */
260struct skb_shared_info {
9f42f126
IC
261 unsigned char nr_frags;
262 __u8 tx_flags;
7967168c
HX
263 unsigned short gso_size;
264 /* Warning: this field is not always filled in (UFO)! */
265 unsigned short gso_segs;
266 unsigned short gso_type;
1da177e4 267 struct sk_buff *frag_list;
ac45f602 268 struct skb_shared_hwtstamps hwtstamps;
9f42f126 269 __be32 ip6_frag_id;
ec7d2f2c
ED
270
271 /*
272 * Warning : all fields before dataref are cleared in __alloc_skb()
273 */
274 atomic_t dataref;
275
69e3c75f
JB
276 /* Intermediate layers must ensure that destructor_arg
277 * remains valid until skb destructor */
278 void * destructor_arg;
a6686f2f 279
fed66381
ED
280 /* must be last field, see pskb_expand_head() */
281 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
282};
283
284/* We divide dataref into two halves. The higher 16 bits hold references
285 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
286 * the entire skb->data. A clone of a headerless skb holds the length of
287 * the header in skb->hdr_len.
1da177e4
LT
288 *
289 * All users must obey the rule that the skb->data reference count must be
290 * greater than or equal to the payload reference count.
291 *
292 * Holding a reference to the payload part means that the user does not
293 * care about modifications to the header part of skb->data.
294 */
295#define SKB_DATAREF_SHIFT 16
296#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
297
d179cd12
DM
298
299enum {
300 SKB_FCLONE_UNAVAILABLE,
301 SKB_FCLONE_ORIG,
302 SKB_FCLONE_CLONE,
303};
304
7967168c
HX
305enum {
306 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 307 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
308
309 /* This indicates the skb is from an untrusted source. */
310 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
311
312 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
313 SKB_GSO_TCP_ECN = 1 << 3,
314
315 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
316
317 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
318
319 SKB_GSO_GRE = 1 << 6,
73136267
PS
320
321 SKB_GSO_UDP_TUNNEL = 1 << 7,
0d89d203
SH
322
323 SKB_GSO_MPLS = 1 << 8,
7967168c
HX
324};
325
2e07fa9c
ACM
326#if BITS_PER_LONG > 32
327#define NET_SKBUFF_DATA_USES_OFFSET 1
328#endif
329
330#ifdef NET_SKBUFF_DATA_USES_OFFSET
331typedef unsigned int sk_buff_data_t;
332#else
333typedef unsigned char *sk_buff_data_t;
334#endif
335
2fc72c7b
KK
336#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
337 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
338#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
339#endif
340
1da177e4
LT
341/**
342 * struct sk_buff - socket buffer
343 * @next: Next buffer in list
344 * @prev: Previous buffer in list
325ed823 345 * @tstamp: Time we arrived
d84e0bd7 346 * @sk: Socket we are owned by
1da177e4 347 * @dev: Device we arrived on/are leaving by
d84e0bd7 348 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 349 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 350 * @sp: the security path, used for xfrm
1da177e4
LT
351 * @len: Length of actual data
352 * @data_len: Data length
353 * @mac_len: Length of link layer header
334a8132 354 * @hdr_len: writable header length of cloned skb
663ead3b
HX
355 * @csum: Checksum (must include start/offset pair)
356 * @csum_start: Offset from skb->head where checksumming should start
357 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 358 * @priority: Packet queueing priority
67be2dd1 359 * @local_df: allow local fragmentation
1da177e4 360 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 361 * @ip_summed: Driver fed us an IP checksum
1da177e4 362 * @nohdr: Payload reference only, must not modify header
d84e0bd7 363 * @nfctinfo: Relationship of this skb to the connection
1da177e4 364 * @pkt_type: Packet class
c83c2486 365 * @fclone: skbuff clone status
c83c2486 366 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
367 * @peeked: this packet has been seen already, so stats have been
368 * done for it, don't do them again
ba9dda3a 369 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
370 * @protocol: Packet protocol from driver
371 * @destructor: Destruct function
372 * @nfct: Associated connection, if any
461ddf3b 373 * @nfct_reasm: netfilter conntrack re-assembly pointer
1da177e4 374 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 375 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
376 * @tc_index: Traffic control index
377 * @tc_verd: traffic control verdict
d84e0bd7
DB
378 * @rxhash: the packet hash computed on receive
379 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 380 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 381 * @ooo_okay: allow the mapping of a socket to a queue to be changed
4ca2462e
CG
382 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
383 * ports.
6e3e939f
JB
384 * @wifi_acked_valid: wifi_acked was set
385 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 386 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
387 * @dma_cookie: a cookie to one of several possible DMA operations
388 * done by skb DMA functions
06021292 389 * @napi_id: id of the NAPI struct this skb came from
984bc16c 390 * @secmark: security marking
d84e0bd7
DB
391 * @mark: Generic packet mark
392 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 393 * @vlan_proto: vlan encapsulation protocol
6aa895b0 394 * @vlan_tci: vlan tag control information
0d89d203 395 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
396 * @inner_transport_header: Inner transport layer header (encapsulation)
397 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 398 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
399 * @transport_header: Transport layer header
400 * @network_header: Network layer header
401 * @mac_header: Link layer header
402 * @tail: Tail pointer
403 * @end: End pointer
404 * @head: Head of buffer
405 * @data: Data head pointer
406 * @truesize: Buffer size
407 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
408 */
409
410struct sk_buff {
411 /* These two members must be first. */
412 struct sk_buff *next;
413 struct sk_buff *prev;
414
b7aa0bf7 415 ktime_t tstamp;
da3f5cf1
FF
416
417 struct sock *sk;
1da177e4 418 struct net_device *dev;
1da177e4 419
1da177e4
LT
420 /*
421 * This is the control buffer. It is free to use for every
422 * layer. Please put your private variables there. If you
423 * want to keep them across layers you have to do a skb_clone()
424 * first. This is owned by whoever has the skb queued ATM.
425 */
da3f5cf1 426 char cb[48] __aligned(8);
1da177e4 427
7fee226a 428 unsigned long _skb_refdst;
da3f5cf1
FF
429#ifdef CONFIG_XFRM
430 struct sec_path *sp;
431#endif
1da177e4 432 unsigned int len,
334a8132
PM
433 data_len;
434 __u16 mac_len,
435 hdr_len;
ff1dcadb
AV
436 union {
437 __wsum csum;
663ead3b
HX
438 struct {
439 __u16 csum_start;
440 __u16 csum_offset;
441 };
ff1dcadb 442 };
1da177e4 443 __u32 priority;
fe55f6d5 444 kmemcheck_bitfield_begin(flags1);
1cbb3380
TG
445 __u8 local_df:1,
446 cloned:1,
447 ip_summed:2,
6869c4d8
HW
448 nohdr:1,
449 nfctinfo:3;
d179cd12 450 __u8 pkt_type:3,
b84f4cc9 451 fclone:2,
ba9dda3a 452 ipvs_property:1,
a59322be 453 peeked:1,
ba9dda3a 454 nf_trace:1;
fe55f6d5 455 kmemcheck_bitfield_end(flags1);
4ab408de 456 __be16 protocol;
1da177e4
LT
457
458 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 459#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 460 struct nf_conntrack *nfct;
2fc72c7b
KK
461#endif
462#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
9fb9cbb1
YK
463 struct sk_buff *nfct_reasm;
464#endif
1da177e4
LT
465#ifdef CONFIG_BRIDGE_NETFILTER
466 struct nf_bridge_info *nf_bridge;
467#endif
f25f4e44 468
8964be4a 469 int skb_iif;
4031ae6e
AD
470
471 __u32 rxhash;
472
86a9bad3 473 __be16 vlan_proto;
4031ae6e
AD
474 __u16 vlan_tci;
475
1da177e4 476#ifdef CONFIG_NET_SCHED
b6b99eb5 477 __u16 tc_index; /* traffic control index */
1da177e4 478#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 479 __u16 tc_verd; /* traffic control verdict */
1da177e4 480#endif
1da177e4 481#endif
fe55f6d5 482
0a14842f 483 __u16 queue_mapping;
fe55f6d5 484 kmemcheck_bitfield_begin(flags2);
de357cc0 485#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 486 __u8 ndisc_nodetype:2;
d0f09804 487#endif
c93bdd0e 488 __u8 pfmemalloc:1;
3853b584 489 __u8 ooo_okay:1;
bdeab991 490 __u8 l4_rxhash:1;
6e3e939f
JB
491 __u8 wifi_acked_valid:1;
492 __u8 wifi_acked:1;
3bdc0eba 493 __u8 no_fcs:1;
d3836f21 494 __u8 head_frag:1;
6a674e9c
JG
495 /* Encapsulation protocol and NIC drivers should use
496 * this flag to indicate to each other if the skb contains
497 * encapsulated packet or not and maybe use the inner packet
498 * headers if needed
499 */
500 __u8 encapsulation:1;
45906723 501 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
502 kmemcheck_bitfield_end(flags2);
503
e0d1095a 504#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
505 union {
506 unsigned int napi_id;
507 dma_cookie_t dma_cookie;
508 };
97fc2f08 509#endif
984bc16c
JM
510#ifdef CONFIG_NETWORK_SECMARK
511 __u32 secmark;
512#endif
3b885787
NH
513 union {
514 __u32 mark;
515 __u32 dropcount;
16fad69c 516 __u32 reserved_tailroom;
3b885787 517 };
1da177e4 518
0d89d203 519 __be16 inner_protocol;
1a37e412
SH
520 __u16 inner_transport_header;
521 __u16 inner_network_header;
522 __u16 inner_mac_header;
523 __u16 transport_header;
524 __u16 network_header;
525 __u16 mac_header;
1da177e4 526 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 527 sk_buff_data_t tail;
4305b541 528 sk_buff_data_t end;
1da177e4 529 unsigned char *head,
4305b541 530 *data;
27a884dc
ACM
531 unsigned int truesize;
532 atomic_t users;
1da177e4
LT
533};
534
535#ifdef __KERNEL__
536/*
537 * Handling routines are only of interest to the kernel
538 */
539#include <linux/slab.h>
540
1da177e4 541
c93bdd0e
MG
542#define SKB_ALLOC_FCLONE 0x01
543#define SKB_ALLOC_RX 0x02
544
545/* Returns true if the skb was allocated from PFMEMALLOC reserves */
546static inline bool skb_pfmemalloc(const struct sk_buff *skb)
547{
548 return unlikely(skb->pfmemalloc);
549}
550
7fee226a
ED
551/*
552 * skb might have a dst pointer attached, refcounted or not.
553 * _skb_refdst low order bit is set if refcount was _not_ taken
554 */
555#define SKB_DST_NOREF 1UL
556#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
557
558/**
559 * skb_dst - returns skb dst_entry
560 * @skb: buffer
561 *
562 * Returns skb dst_entry, regardless of reference taken or not.
563 */
adf30907
ED
564static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
565{
7fee226a
ED
566 /* If refdst was not refcounted, check we still are in a
567 * rcu_read_lock section
568 */
569 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
570 !rcu_read_lock_held() &&
571 !rcu_read_lock_bh_held());
572 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
573}
574
7fee226a
ED
575/**
576 * skb_dst_set - sets skb dst
577 * @skb: buffer
578 * @dst: dst entry
579 *
580 * Sets skb dst, assuming a reference was taken on dst and should
581 * be released by skb_dst_drop()
582 */
adf30907
ED
583static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
584{
7fee226a
ED
585 skb->_skb_refdst = (unsigned long)dst;
586}
587
7965bd4d
JP
588void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
589 bool force);
932bc4d7
JA
590
591/**
592 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
593 * @skb: buffer
594 * @dst: dst entry
595 *
596 * Sets skb dst, assuming a reference was not taken on dst.
597 * If dst entry is cached, we do not take reference and dst_release
598 * will be avoided by refdst_drop. If dst entry is not cached, we take
599 * reference, so that last dst_release can destroy the dst immediately.
600 */
601static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
602{
603 __skb_dst_set_noref(skb, dst, false);
604}
605
606/**
607 * skb_dst_set_noref_force - sets skb dst, without taking reference
608 * @skb: buffer
609 * @dst: dst entry
610 *
611 * Sets skb dst, assuming a reference was not taken on dst.
612 * No reference is taken and no dst_release will be called. While for
613 * cached dsts deferred reclaim is a basic feature, for entries that are
614 * not cached it is caller's job to guarantee that last dst_release for
615 * provided dst happens when nobody uses it, eg. after a RCU grace period.
616 */
617static inline void skb_dst_set_noref_force(struct sk_buff *skb,
618 struct dst_entry *dst)
619{
620 __skb_dst_set_noref(skb, dst, true);
621}
7fee226a
ED
622
623/**
25985edc 624 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
625 * @skb: buffer
626 */
627static inline bool skb_dst_is_noref(const struct sk_buff *skb)
628{
629 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
630}
631
511c3f92
ED
632static inline struct rtable *skb_rtable(const struct sk_buff *skb)
633{
adf30907 634 return (struct rtable *)skb_dst(skb);
511c3f92
ED
635}
636
7965bd4d
JP
637void kfree_skb(struct sk_buff *skb);
638void kfree_skb_list(struct sk_buff *segs);
639void skb_tx_error(struct sk_buff *skb);
640void consume_skb(struct sk_buff *skb);
641void __kfree_skb(struct sk_buff *skb);
d7e8883c 642extern struct kmem_cache *skbuff_head_cache;
bad43ca8 643
7965bd4d
JP
644void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
645bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
646 bool *fragstolen, int *delta_truesize);
bad43ca8 647
7965bd4d
JP
648struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
649 int node);
650struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 651static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 652 gfp_t priority)
d179cd12 653{
564824b0 654 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
655}
656
657static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 658 gfp_t priority)
d179cd12 659{
c93bdd0e 660 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
661}
662
7965bd4d 663struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
664static inline struct sk_buff *alloc_skb_head(gfp_t priority)
665{
666 return __alloc_skb_head(priority, -1);
667}
668
7965bd4d
JP
669struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
670int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
671struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
672struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
673struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
674
675int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
676struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
677 unsigned int headroom);
678struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
679 int newtailroom, gfp_t priority);
680int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
681 int len);
682int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
683int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 684#define dev_kfree_skb(a) consume_skb(a)
1da177e4 685
7965bd4d
JP
686int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
687 int getfrag(void *from, char *to, int offset,
688 int len, int odd, struct sk_buff *skb),
689 void *from, int length);
e89e9cf5 690
d94d9fee 691struct skb_seq_state {
677e90ed
TG
692 __u32 lower_offset;
693 __u32 upper_offset;
694 __u32 frag_idx;
695 __u32 stepped_offset;
696 struct sk_buff *root_skb;
697 struct sk_buff *cur_skb;
698 __u8 *frag_data;
699};
700
7965bd4d
JP
701void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
702 unsigned int to, struct skb_seq_state *st);
703unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
704 struct skb_seq_state *st);
705void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 706
7965bd4d
JP
707unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
708 unsigned int to, struct ts_config *config,
709 struct ts_state *state);
3fc7e8a6 710
7965bd4d 711void __skb_get_rxhash(struct sk_buff *skb);
bfb564e7
KK
712static inline __u32 skb_get_rxhash(struct sk_buff *skb)
713{
ecd5cf5d 714 if (!skb->l4_rxhash)
bdeab991 715 __skb_get_rxhash(skb);
bfb564e7
KK
716
717 return skb->rxhash;
718}
719
4305b541
ACM
720#ifdef NET_SKBUFF_DATA_USES_OFFSET
721static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
722{
723 return skb->head + skb->end;
724}
ec47ea82
AD
725
726static inline unsigned int skb_end_offset(const struct sk_buff *skb)
727{
728 return skb->end;
729}
4305b541
ACM
730#else
731static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
732{
733 return skb->end;
734}
ec47ea82
AD
735
736static inline unsigned int skb_end_offset(const struct sk_buff *skb)
737{
738 return skb->end - skb->head;
739}
4305b541
ACM
740#endif
741
1da177e4 742/* Internal */
4305b541 743#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 744
ac45f602
PO
745static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
746{
747 return &skb_shinfo(skb)->hwtstamps;
748}
749
1da177e4
LT
750/**
751 * skb_queue_empty - check if a queue is empty
752 * @list: queue head
753 *
754 * Returns true if the queue is empty, false otherwise.
755 */
756static inline int skb_queue_empty(const struct sk_buff_head *list)
757{
758 return list->next == (struct sk_buff *)list;
759}
760
fc7ebb21
DM
761/**
762 * skb_queue_is_last - check if skb is the last entry in the queue
763 * @list: queue head
764 * @skb: buffer
765 *
766 * Returns true if @skb is the last buffer on the list.
767 */
768static inline bool skb_queue_is_last(const struct sk_buff_head *list,
769 const struct sk_buff *skb)
770{
a02cec21 771 return skb->next == (struct sk_buff *)list;
fc7ebb21
DM
772}
773
832d11c5
IJ
774/**
775 * skb_queue_is_first - check if skb is the first entry in the queue
776 * @list: queue head
777 * @skb: buffer
778 *
779 * Returns true if @skb is the first buffer on the list.
780 */
781static inline bool skb_queue_is_first(const struct sk_buff_head *list,
782 const struct sk_buff *skb)
783{
a02cec21 784 return skb->prev == (struct sk_buff *)list;
832d11c5
IJ
785}
786
249c8b42
DM
787/**
788 * skb_queue_next - return the next packet in the queue
789 * @list: queue head
790 * @skb: current buffer
791 *
792 * Return the next packet in @list after @skb. It is only valid to
793 * call this if skb_queue_is_last() evaluates to false.
794 */
795static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
796 const struct sk_buff *skb)
797{
798 /* This BUG_ON may seem severe, but if we just return then we
799 * are going to dereference garbage.
800 */
801 BUG_ON(skb_queue_is_last(list, skb));
802 return skb->next;
803}
804
832d11c5
IJ
805/**
806 * skb_queue_prev - return the prev packet in the queue
807 * @list: queue head
808 * @skb: current buffer
809 *
810 * Return the prev packet in @list before @skb. It is only valid to
811 * call this if skb_queue_is_first() evaluates to false.
812 */
813static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
814 const struct sk_buff *skb)
815{
816 /* This BUG_ON may seem severe, but if we just return then we
817 * are going to dereference garbage.
818 */
819 BUG_ON(skb_queue_is_first(list, skb));
820 return skb->prev;
821}
822
1da177e4
LT
823/**
824 * skb_get - reference buffer
825 * @skb: buffer to reference
826 *
827 * Makes another reference to a socket buffer and returns a pointer
828 * to the buffer.
829 */
830static inline struct sk_buff *skb_get(struct sk_buff *skb)
831{
832 atomic_inc(&skb->users);
833 return skb;
834}
835
836/*
837 * If users == 1, we are the only owner and are can avoid redundant
838 * atomic change.
839 */
840
1da177e4
LT
841/**
842 * skb_cloned - is the buffer a clone
843 * @skb: buffer to check
844 *
845 * Returns true if the buffer was generated with skb_clone() and is
846 * one of multiple shared copies of the buffer. Cloned buffers are
847 * shared data so must not be written to under normal circumstances.
848 */
849static inline int skb_cloned(const struct sk_buff *skb)
850{
851 return skb->cloned &&
852 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
853}
854
14bbd6a5
PS
855static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
856{
857 might_sleep_if(pri & __GFP_WAIT);
858
859 if (skb_cloned(skb))
860 return pskb_expand_head(skb, 0, 0, pri);
861
862 return 0;
863}
864
1da177e4
LT
865/**
866 * skb_header_cloned - is the header a clone
867 * @skb: buffer to check
868 *
869 * Returns true if modifying the header part of the buffer requires
870 * the data to be copied.
871 */
872static inline int skb_header_cloned(const struct sk_buff *skb)
873{
874 int dataref;
875
876 if (!skb->cloned)
877 return 0;
878
879 dataref = atomic_read(&skb_shinfo(skb)->dataref);
880 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
881 return dataref != 1;
882}
883
884/**
885 * skb_header_release - release reference to header
886 * @skb: buffer to operate on
887 *
888 * Drop a reference to the header part of the buffer. This is done
889 * by acquiring a payload reference. You must not read from the header
890 * part of skb->data after this.
891 */
892static inline void skb_header_release(struct sk_buff *skb)
893{
894 BUG_ON(skb->nohdr);
895 skb->nohdr = 1;
896 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
897}
898
899/**
900 * skb_shared - is the buffer shared
901 * @skb: buffer to check
902 *
903 * Returns true if more than one person has a reference to this
904 * buffer.
905 */
906static inline int skb_shared(const struct sk_buff *skb)
907{
908 return atomic_read(&skb->users) != 1;
909}
910
911/**
912 * skb_share_check - check if buffer is shared and if so clone it
913 * @skb: buffer to check
914 * @pri: priority for memory allocation
915 *
916 * If the buffer is shared the buffer is cloned and the old copy
917 * drops a reference. A new clone with a single reference is returned.
918 * If the buffer is not shared the original buffer is returned. When
919 * being called from interrupt status or with spinlocks held pri must
920 * be GFP_ATOMIC.
921 *
922 * NULL is returned on a memory allocation failure.
923 */
47061bc4 924static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
925{
926 might_sleep_if(pri & __GFP_WAIT);
927 if (skb_shared(skb)) {
928 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
929
930 if (likely(nskb))
931 consume_skb(skb);
932 else
933 kfree_skb(skb);
1da177e4
LT
934 skb = nskb;
935 }
936 return skb;
937}
938
939/*
940 * Copy shared buffers into a new sk_buff. We effectively do COW on
941 * packets to handle cases where we have a local reader and forward
942 * and a couple of other messy ones. The normal one is tcpdumping
943 * a packet thats being forwarded.
944 */
945
946/**
947 * skb_unshare - make a copy of a shared buffer
948 * @skb: buffer to check
949 * @pri: priority for memory allocation
950 *
951 * If the socket buffer is a clone then this function creates a new
952 * copy of the data, drops a reference count on the old copy and returns
953 * the new copy with the reference count at 1. If the buffer is not a clone
954 * the original buffer is returned. When called with a spinlock held or
955 * from interrupt state @pri must be %GFP_ATOMIC
956 *
957 * %NULL is returned on a memory allocation failure.
958 */
e2bf521d 959static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 960 gfp_t pri)
1da177e4
LT
961{
962 might_sleep_if(pri & __GFP_WAIT);
963 if (skb_cloned(skb)) {
964 struct sk_buff *nskb = skb_copy(skb, pri);
965 kfree_skb(skb); /* Free our shared copy */
966 skb = nskb;
967 }
968 return skb;
969}
970
971/**
1a5778aa 972 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
973 * @list_: list to peek at
974 *
975 * Peek an &sk_buff. Unlike most other operations you _MUST_
976 * be careful with this one. A peek leaves the buffer on the
977 * list and someone else may run off with it. You must hold
978 * the appropriate locks or have a private queue to do this.
979 *
980 * Returns %NULL for an empty list or a pointer to the head element.
981 * The reference count is not incremented and the reference is therefore
982 * volatile. Use with caution.
983 */
05bdd2f1 984static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 985{
18d07000
ED
986 struct sk_buff *skb = list_->next;
987
988 if (skb == (struct sk_buff *)list_)
989 skb = NULL;
990 return skb;
1da177e4
LT
991}
992
da5ef6e5
PE
993/**
994 * skb_peek_next - peek skb following the given one from a queue
995 * @skb: skb to start from
996 * @list_: list to peek at
997 *
998 * Returns %NULL when the end of the list is met or a pointer to the
999 * next element. The reference count is not incremented and the
1000 * reference is therefore volatile. Use with caution.
1001 */
1002static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1003 const struct sk_buff_head *list_)
1004{
1005 struct sk_buff *next = skb->next;
18d07000 1006
da5ef6e5
PE
1007 if (next == (struct sk_buff *)list_)
1008 next = NULL;
1009 return next;
1010}
1011
1da177e4 1012/**
1a5778aa 1013 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1014 * @list_: list to peek at
1015 *
1016 * Peek an &sk_buff. Unlike most other operations you _MUST_
1017 * be careful with this one. A peek leaves the buffer on the
1018 * list and someone else may run off with it. You must hold
1019 * the appropriate locks or have a private queue to do this.
1020 *
1021 * Returns %NULL for an empty list or a pointer to the tail element.
1022 * The reference count is not incremented and the reference is therefore
1023 * volatile. Use with caution.
1024 */
05bdd2f1 1025static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1026{
18d07000
ED
1027 struct sk_buff *skb = list_->prev;
1028
1029 if (skb == (struct sk_buff *)list_)
1030 skb = NULL;
1031 return skb;
1032
1da177e4
LT
1033}
1034
1035/**
1036 * skb_queue_len - get queue length
1037 * @list_: list to measure
1038 *
1039 * Return the length of an &sk_buff queue.
1040 */
1041static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1042{
1043 return list_->qlen;
1044}
1045
67fed459
DM
1046/**
1047 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1048 * @list: queue to initialize
1049 *
1050 * This initializes only the list and queue length aspects of
1051 * an sk_buff_head object. This allows to initialize the list
1052 * aspects of an sk_buff_head without reinitializing things like
1053 * the spinlock. It can also be used for on-stack sk_buff_head
1054 * objects where the spinlock is known to not be used.
1055 */
1056static inline void __skb_queue_head_init(struct sk_buff_head *list)
1057{
1058 list->prev = list->next = (struct sk_buff *)list;
1059 list->qlen = 0;
1060}
1061
76f10ad0
AV
1062/*
1063 * This function creates a split out lock class for each invocation;
1064 * this is needed for now since a whole lot of users of the skb-queue
1065 * infrastructure in drivers have different locking usage (in hardirq)
1066 * than the networking core (in softirq only). In the long run either the
1067 * network layer or drivers should need annotation to consolidate the
1068 * main types of usage into 3 classes.
1069 */
1da177e4
LT
1070static inline void skb_queue_head_init(struct sk_buff_head *list)
1071{
1072 spin_lock_init(&list->lock);
67fed459 1073 __skb_queue_head_init(list);
1da177e4
LT
1074}
1075
c2ecba71
PE
1076static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1077 struct lock_class_key *class)
1078{
1079 skb_queue_head_init(list);
1080 lockdep_set_class(&list->lock, class);
1081}
1082
1da177e4 1083/*
bf299275 1084 * Insert an sk_buff on a list.
1da177e4
LT
1085 *
1086 * The "__skb_xxxx()" functions are the non-atomic ones that
1087 * can only be called with interrupts disabled.
1088 */
7965bd4d
JP
1089void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1090 struct sk_buff_head *list);
bf299275
GR
1091static inline void __skb_insert(struct sk_buff *newsk,
1092 struct sk_buff *prev, struct sk_buff *next,
1093 struct sk_buff_head *list)
1094{
1095 newsk->next = next;
1096 newsk->prev = prev;
1097 next->prev = prev->next = newsk;
1098 list->qlen++;
1099}
1da177e4 1100
67fed459
DM
1101static inline void __skb_queue_splice(const struct sk_buff_head *list,
1102 struct sk_buff *prev,
1103 struct sk_buff *next)
1104{
1105 struct sk_buff *first = list->next;
1106 struct sk_buff *last = list->prev;
1107
1108 first->prev = prev;
1109 prev->next = first;
1110
1111 last->next = next;
1112 next->prev = last;
1113}
1114
1115/**
1116 * skb_queue_splice - join two skb lists, this is designed for stacks
1117 * @list: the new list to add
1118 * @head: the place to add it in the first list
1119 */
1120static inline void skb_queue_splice(const struct sk_buff_head *list,
1121 struct sk_buff_head *head)
1122{
1123 if (!skb_queue_empty(list)) {
1124 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1125 head->qlen += list->qlen;
67fed459
DM
1126 }
1127}
1128
1129/**
d9619496 1130 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1131 * @list: the new list to add
1132 * @head: the place to add it in the first list
1133 *
1134 * The list at @list is reinitialised
1135 */
1136static inline void skb_queue_splice_init(struct sk_buff_head *list,
1137 struct sk_buff_head *head)
1138{
1139 if (!skb_queue_empty(list)) {
1140 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1141 head->qlen += list->qlen;
67fed459
DM
1142 __skb_queue_head_init(list);
1143 }
1144}
1145
1146/**
1147 * skb_queue_splice_tail - join two skb lists, each list being a queue
1148 * @list: the new list to add
1149 * @head: the place to add it in the first list
1150 */
1151static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1152 struct sk_buff_head *head)
1153{
1154 if (!skb_queue_empty(list)) {
1155 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1156 head->qlen += list->qlen;
67fed459
DM
1157 }
1158}
1159
1160/**
d9619496 1161 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1162 * @list: the new list to add
1163 * @head: the place to add it in the first list
1164 *
1165 * Each of the lists is a queue.
1166 * The list at @list is reinitialised
1167 */
1168static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1169 struct sk_buff_head *head)
1170{
1171 if (!skb_queue_empty(list)) {
1172 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1173 head->qlen += list->qlen;
67fed459
DM
1174 __skb_queue_head_init(list);
1175 }
1176}
1177
1da177e4 1178/**
300ce174 1179 * __skb_queue_after - queue a buffer at the list head
1da177e4 1180 * @list: list to use
300ce174 1181 * @prev: place after this buffer
1da177e4
LT
1182 * @newsk: buffer to queue
1183 *
300ce174 1184 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1185 * and you must therefore hold required locks before calling it.
1186 *
1187 * A buffer cannot be placed on two lists at the same time.
1188 */
300ce174
SH
1189static inline void __skb_queue_after(struct sk_buff_head *list,
1190 struct sk_buff *prev,
1191 struct sk_buff *newsk)
1da177e4 1192{
bf299275 1193 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1194}
1195
7965bd4d
JP
1196void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1197 struct sk_buff_head *list);
7de6c033 1198
f5572855
GR
1199static inline void __skb_queue_before(struct sk_buff_head *list,
1200 struct sk_buff *next,
1201 struct sk_buff *newsk)
1202{
1203 __skb_insert(newsk, next->prev, next, list);
1204}
1205
300ce174
SH
1206/**
1207 * __skb_queue_head - queue a buffer at the list head
1208 * @list: list to use
1209 * @newsk: buffer to queue
1210 *
1211 * Queue a buffer at the start of a list. This function takes no locks
1212 * and you must therefore hold required locks before calling it.
1213 *
1214 * A buffer cannot be placed on two lists at the same time.
1215 */
7965bd4d 1216void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1217static inline void __skb_queue_head(struct sk_buff_head *list,
1218 struct sk_buff *newsk)
1219{
1220 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1221}
1222
1da177e4
LT
1223/**
1224 * __skb_queue_tail - queue a buffer at the list tail
1225 * @list: list to use
1226 * @newsk: buffer to queue
1227 *
1228 * Queue a buffer at the end of a list. This function takes no locks
1229 * and you must therefore hold required locks before calling it.
1230 *
1231 * A buffer cannot be placed on two lists at the same time.
1232 */
7965bd4d 1233void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1234static inline void __skb_queue_tail(struct sk_buff_head *list,
1235 struct sk_buff *newsk)
1236{
f5572855 1237 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1238}
1239
1da177e4
LT
1240/*
1241 * remove sk_buff from list. _Must_ be called atomically, and with
1242 * the list known..
1243 */
7965bd4d 1244void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1245static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1246{
1247 struct sk_buff *next, *prev;
1248
1249 list->qlen--;
1250 next = skb->next;
1251 prev = skb->prev;
1252 skb->next = skb->prev = NULL;
1da177e4
LT
1253 next->prev = prev;
1254 prev->next = next;
1255}
1256
f525c06d
GR
1257/**
1258 * __skb_dequeue - remove from the head of the queue
1259 * @list: list to dequeue from
1260 *
1261 * Remove the head of the list. This function does not take any locks
1262 * so must be used with appropriate locks held only. The head item is
1263 * returned or %NULL if the list is empty.
1264 */
7965bd4d 1265struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1266static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1267{
1268 struct sk_buff *skb = skb_peek(list);
1269 if (skb)
1270 __skb_unlink(skb, list);
1271 return skb;
1272}
1da177e4
LT
1273
1274/**
1275 * __skb_dequeue_tail - remove from the tail of the queue
1276 * @list: list to dequeue from
1277 *
1278 * Remove the tail of the list. This function does not take any locks
1279 * so must be used with appropriate locks held only. The tail item is
1280 * returned or %NULL if the list is empty.
1281 */
7965bd4d 1282struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1283static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1284{
1285 struct sk_buff *skb = skb_peek_tail(list);
1286 if (skb)
1287 __skb_unlink(skb, list);
1288 return skb;
1289}
1290
1291
bdcc0924 1292static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1293{
1294 return skb->data_len;
1295}
1296
1297static inline unsigned int skb_headlen(const struct sk_buff *skb)
1298{
1299 return skb->len - skb->data_len;
1300}
1301
1302static inline int skb_pagelen(const struct sk_buff *skb)
1303{
1304 int i, len = 0;
1305
1306 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1307 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1308 return len + skb_headlen(skb);
1309}
1310
131ea667
IC
1311/**
1312 * __skb_fill_page_desc - initialise a paged fragment in an skb
1313 * @skb: buffer containing fragment to be initialised
1314 * @i: paged fragment index to initialise
1315 * @page: the page to use for this fragment
1316 * @off: the offset to the data with @page
1317 * @size: the length of the data
1318 *
1319 * Initialises the @i'th fragment of @skb to point to &size bytes at
1320 * offset @off within @page.
1321 *
1322 * Does not take any additional reference on the fragment.
1323 */
1324static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1325 struct page *page, int off, int size)
1da177e4
LT
1326{
1327 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1328
c48a11c7
MG
1329 /*
1330 * Propagate page->pfmemalloc to the skb if we can. The problem is
1331 * that not all callers have unique ownership of the page. If
1332 * pfmemalloc is set, we check the mapping as a mapping implies
1333 * page->index is set (index and pfmemalloc share space).
1334 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1335 * do not lose pfmemalloc information as the pages would not be
1336 * allocated using __GFP_MEMALLOC.
1337 */
a8605c60 1338 frag->page.p = page;
1da177e4 1339 frag->page_offset = off;
9e903e08 1340 skb_frag_size_set(frag, size);
cca7af38
PE
1341
1342 page = compound_head(page);
1343 if (page->pfmemalloc && !page->mapping)
1344 skb->pfmemalloc = true;
131ea667
IC
1345}
1346
1347/**
1348 * skb_fill_page_desc - initialise a paged fragment in an skb
1349 * @skb: buffer containing fragment to be initialised
1350 * @i: paged fragment index to initialise
1351 * @page: the page to use for this fragment
1352 * @off: the offset to the data with @page
1353 * @size: the length of the data
1354 *
1355 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1356 * @skb to point to &size bytes at offset @off within @page. In
1357 * addition updates @skb such that @i is the last fragment.
1358 *
1359 * Does not take any additional reference on the fragment.
1360 */
1361static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1362 struct page *page, int off, int size)
1363{
1364 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1365 skb_shinfo(skb)->nr_frags = i + 1;
1366}
1367
7965bd4d
JP
1368void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1369 int size, unsigned int truesize);
654bed16 1370
1da177e4 1371#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1372#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1373#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1374
27a884dc
ACM
1375#ifdef NET_SKBUFF_DATA_USES_OFFSET
1376static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1377{
1378 return skb->head + skb->tail;
1379}
1380
1381static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1382{
1383 skb->tail = skb->data - skb->head;
1384}
1385
1386static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1387{
1388 skb_reset_tail_pointer(skb);
1389 skb->tail += offset;
1390}
7cc46190 1391
27a884dc
ACM
1392#else /* NET_SKBUFF_DATA_USES_OFFSET */
1393static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1394{
1395 return skb->tail;
1396}
1397
1398static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1399{
1400 skb->tail = skb->data;
1401}
1402
1403static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1404{
1405 skb->tail = skb->data + offset;
1406}
4305b541 1407
27a884dc
ACM
1408#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1409
1da177e4
LT
1410/*
1411 * Add data to an sk_buff
1412 */
7965bd4d 1413unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1414static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1415{
27a884dc 1416 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1417 SKB_LINEAR_ASSERT(skb);
1418 skb->tail += len;
1419 skb->len += len;
1420 return tmp;
1421}
1422
7965bd4d 1423unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1424static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1425{
1426 skb->data -= len;
1427 skb->len += len;
1428 return skb->data;
1429}
1430
7965bd4d 1431unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1432static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1433{
1434 skb->len -= len;
1435 BUG_ON(skb->len < skb->data_len);
1436 return skb->data += len;
1437}
1438
47d29646
DM
1439static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1440{
1441 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1442}
1443
7965bd4d 1444unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1445
1446static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1447{
1448 if (len > skb_headlen(skb) &&
987c402a 1449 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1450 return NULL;
1451 skb->len -= len;
1452 return skb->data += len;
1453}
1454
1455static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1456{
1457 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1458}
1459
1460static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1461{
1462 if (likely(len <= skb_headlen(skb)))
1463 return 1;
1464 if (unlikely(len > skb->len))
1465 return 0;
987c402a 1466 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1467}
1468
1469/**
1470 * skb_headroom - bytes at buffer head
1471 * @skb: buffer to check
1472 *
1473 * Return the number of bytes of free space at the head of an &sk_buff.
1474 */
c2636b4d 1475static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1476{
1477 return skb->data - skb->head;
1478}
1479
1480/**
1481 * skb_tailroom - bytes at buffer end
1482 * @skb: buffer to check
1483 *
1484 * Return the number of bytes of free space at the tail of an sk_buff
1485 */
1486static inline int skb_tailroom(const struct sk_buff *skb)
1487{
4305b541 1488 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1489}
1490
a21d4572
ED
1491/**
1492 * skb_availroom - bytes at buffer end
1493 * @skb: buffer to check
1494 *
1495 * Return the number of bytes of free space at the tail of an sk_buff
1496 * allocated by sk_stream_alloc()
1497 */
1498static inline int skb_availroom(const struct sk_buff *skb)
1499{
16fad69c
ED
1500 if (skb_is_nonlinear(skb))
1501 return 0;
1502
1503 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1504}
1505
1da177e4
LT
1506/**
1507 * skb_reserve - adjust headroom
1508 * @skb: buffer to alter
1509 * @len: bytes to move
1510 *
1511 * Increase the headroom of an empty &sk_buff by reducing the tail
1512 * room. This is only allowed for an empty buffer.
1513 */
8243126c 1514static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1515{
1516 skb->data += len;
1517 skb->tail += len;
1518}
1519
6a674e9c
JG
1520static inline void skb_reset_inner_headers(struct sk_buff *skb)
1521{
aefbd2b3 1522 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1523 skb->inner_network_header = skb->network_header;
1524 skb->inner_transport_header = skb->transport_header;
1525}
1526
0b5c9db1
JP
1527static inline void skb_reset_mac_len(struct sk_buff *skb)
1528{
1529 skb->mac_len = skb->network_header - skb->mac_header;
1530}
1531
6a674e9c
JG
1532static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1533 *skb)
1534{
1535 return skb->head + skb->inner_transport_header;
1536}
1537
1538static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1539{
1540 skb->inner_transport_header = skb->data - skb->head;
1541}
1542
1543static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1544 const int offset)
1545{
1546 skb_reset_inner_transport_header(skb);
1547 skb->inner_transport_header += offset;
1548}
1549
1550static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1551{
1552 return skb->head + skb->inner_network_header;
1553}
1554
1555static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1556{
1557 skb->inner_network_header = skb->data - skb->head;
1558}
1559
1560static inline void skb_set_inner_network_header(struct sk_buff *skb,
1561 const int offset)
1562{
1563 skb_reset_inner_network_header(skb);
1564 skb->inner_network_header += offset;
1565}
1566
aefbd2b3
PS
1567static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1568{
1569 return skb->head + skb->inner_mac_header;
1570}
1571
1572static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1573{
1574 skb->inner_mac_header = skb->data - skb->head;
1575}
1576
1577static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1578 const int offset)
1579{
1580 skb_reset_inner_mac_header(skb);
1581 skb->inner_mac_header += offset;
1582}
fda55eca
ED
1583static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1584{
35d04610 1585 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1586}
1587
9c70220b
ACM
1588static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1589{
2e07fa9c 1590 return skb->head + skb->transport_header;
9c70220b
ACM
1591}
1592
badff6d0
ACM
1593static inline void skb_reset_transport_header(struct sk_buff *skb)
1594{
2e07fa9c 1595 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1596}
1597
967b05f6
ACM
1598static inline void skb_set_transport_header(struct sk_buff *skb,
1599 const int offset)
1600{
2e07fa9c
ACM
1601 skb_reset_transport_header(skb);
1602 skb->transport_header += offset;
ea2ae17d
ACM
1603}
1604
d56f90a7
ACM
1605static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1606{
2e07fa9c 1607 return skb->head + skb->network_header;
d56f90a7
ACM
1608}
1609
c1d2bbe1
ACM
1610static inline void skb_reset_network_header(struct sk_buff *skb)
1611{
2e07fa9c 1612 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1613}
1614
c14d2450
ACM
1615static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1616{
2e07fa9c
ACM
1617 skb_reset_network_header(skb);
1618 skb->network_header += offset;
c14d2450
ACM
1619}
1620
2e07fa9c 1621static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1622{
2e07fa9c 1623 return skb->head + skb->mac_header;
bbe735e4
ACM
1624}
1625
2e07fa9c 1626static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1627{
35d04610 1628 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1629}
1630
1631static inline void skb_reset_mac_header(struct sk_buff *skb)
1632{
1633 skb->mac_header = skb->data - skb->head;
1634}
1635
1636static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1637{
1638 skb_reset_mac_header(skb);
1639 skb->mac_header += offset;
1640}
1641
fbbdb8f0
YX
1642static inline void skb_probe_transport_header(struct sk_buff *skb,
1643 const int offset_hint)
1644{
1645 struct flow_keys keys;
1646
1647 if (skb_transport_header_was_set(skb))
1648 return;
1649 else if (skb_flow_dissect(skb, &keys))
1650 skb_set_transport_header(skb, keys.thoff);
1651 else
1652 skb_set_transport_header(skb, offset_hint);
1653}
1654
03606895
ED
1655static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1656{
1657 if (skb_mac_header_was_set(skb)) {
1658 const unsigned char *old_mac = skb_mac_header(skb);
1659
1660 skb_set_mac_header(skb, -skb->mac_len);
1661 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1662 }
1663}
1664
04fb451e
MM
1665static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1666{
1667 return skb->csum_start - skb_headroom(skb);
1668}
1669
2e07fa9c
ACM
1670static inline int skb_transport_offset(const struct sk_buff *skb)
1671{
1672 return skb_transport_header(skb) - skb->data;
1673}
1674
1675static inline u32 skb_network_header_len(const struct sk_buff *skb)
1676{
1677 return skb->transport_header - skb->network_header;
1678}
1679
6a674e9c
JG
1680static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1681{
1682 return skb->inner_transport_header - skb->inner_network_header;
1683}
1684
2e07fa9c
ACM
1685static inline int skb_network_offset(const struct sk_buff *skb)
1686{
1687 return skb_network_header(skb) - skb->data;
1688}
48d49d0c 1689
6a674e9c
JG
1690static inline int skb_inner_network_offset(const struct sk_buff *skb)
1691{
1692 return skb_inner_network_header(skb) - skb->data;
1693}
1694
f9599ce1
CG
1695static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1696{
1697 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1698}
1699
1da177e4
LT
1700/*
1701 * CPUs often take a performance hit when accessing unaligned memory
1702 * locations. The actual performance hit varies, it can be small if the
1703 * hardware handles it or large if we have to take an exception and fix it
1704 * in software.
1705 *
1706 * Since an ethernet header is 14 bytes network drivers often end up with
1707 * the IP header at an unaligned offset. The IP header can be aligned by
1708 * shifting the start of the packet by 2 bytes. Drivers should do this
1709 * with:
1710 *
8660c124 1711 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1712 *
1713 * The downside to this alignment of the IP header is that the DMA is now
1714 * unaligned. On some architectures the cost of an unaligned DMA is high
1715 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1716 *
1da177e4
LT
1717 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1718 * to be overridden.
1719 */
1720#ifndef NET_IP_ALIGN
1721#define NET_IP_ALIGN 2
1722#endif
1723
025be81e
AB
1724/*
1725 * The networking layer reserves some headroom in skb data (via
1726 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1727 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1728 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1729 *
1730 * Unfortunately this headroom changes the DMA alignment of the resulting
1731 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1732 * on some architectures. An architecture can override this value,
1733 * perhaps setting it to a cacheline in size (since that will maintain
1734 * cacheline alignment of the DMA). It must be a power of 2.
1735 *
d6301d3d 1736 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1737 * headroom, you should not reduce this.
5933dd2f
ED
1738 *
1739 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1740 * to reduce average number of cache lines per packet.
1741 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1742 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1743 */
1744#ifndef NET_SKB_PAD
5933dd2f 1745#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1746#endif
1747
7965bd4d 1748int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1749
1750static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1751{
c4264f27 1752 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1753 WARN_ON(1);
1754 return;
1755 }
27a884dc
ACM
1756 skb->len = len;
1757 skb_set_tail_pointer(skb, len);
1da177e4
LT
1758}
1759
7965bd4d 1760void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1761
1762static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1763{
3cc0e873
HX
1764 if (skb->data_len)
1765 return ___pskb_trim(skb, len);
1766 __skb_trim(skb, len);
1767 return 0;
1da177e4
LT
1768}
1769
1770static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1771{
1772 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1773}
1774
e9fa4f7b
HX
1775/**
1776 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1777 * @skb: buffer to alter
1778 * @len: new length
1779 *
1780 * This is identical to pskb_trim except that the caller knows that
1781 * the skb is not cloned so we should never get an error due to out-
1782 * of-memory.
1783 */
1784static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1785{
1786 int err = pskb_trim(skb, len);
1787 BUG_ON(err);
1788}
1789
1da177e4
LT
1790/**
1791 * skb_orphan - orphan a buffer
1792 * @skb: buffer to orphan
1793 *
1794 * If a buffer currently has an owner then we call the owner's
1795 * destructor function and make the @skb unowned. The buffer continues
1796 * to exist but is no longer charged to its former owner.
1797 */
1798static inline void skb_orphan(struct sk_buff *skb)
1799{
c34a7612 1800 if (skb->destructor) {
1da177e4 1801 skb->destructor(skb);
c34a7612
ED
1802 skb->destructor = NULL;
1803 skb->sk = NULL;
376c7311
ED
1804 } else {
1805 BUG_ON(skb->sk);
c34a7612 1806 }
1da177e4
LT
1807}
1808
a353e0ce
MT
1809/**
1810 * skb_orphan_frags - orphan the frags contained in a buffer
1811 * @skb: buffer to orphan frags from
1812 * @gfp_mask: allocation mask for replacement pages
1813 *
1814 * For each frag in the SKB which needs a destructor (i.e. has an
1815 * owner) create a copy of that frag and release the original
1816 * page by calling the destructor.
1817 */
1818static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1819{
1820 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1821 return 0;
1822 return skb_copy_ubufs(skb, gfp_mask);
1823}
1824
1da177e4
LT
1825/**
1826 * __skb_queue_purge - empty a list
1827 * @list: list to empty
1828 *
1829 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1830 * the list and one reference dropped. This function does not take the
1831 * list lock and the caller must hold the relevant locks to use it.
1832 */
7965bd4d 1833void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
1834static inline void __skb_queue_purge(struct sk_buff_head *list)
1835{
1836 struct sk_buff *skb;
1837 while ((skb = __skb_dequeue(list)) != NULL)
1838 kfree_skb(skb);
1839}
1840
e5e67305
AD
1841#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1842#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1843#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1844
7965bd4d 1845void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 1846
7965bd4d
JP
1847struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1848 gfp_t gfp_mask);
8af27456
CH
1849
1850/**
1851 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1852 * @dev: network device to receive on
1853 * @length: length to allocate
1854 *
1855 * Allocate a new &sk_buff and assign it a usage count of one. The
1856 * buffer has unspecified headroom built in. Users should allocate
1857 * the headroom they think they need without accounting for the
1858 * built in space. The built in space is used for optimisations.
1859 *
1860 * %NULL is returned if there is no free memory. Although this function
1861 * allocates memory it can be called from an interrupt.
1862 */
1863static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 1864 unsigned int length)
8af27456
CH
1865{
1866 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1867}
1868
6f532612
ED
1869/* legacy helper around __netdev_alloc_skb() */
1870static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1871 gfp_t gfp_mask)
1872{
1873 return __netdev_alloc_skb(NULL, length, gfp_mask);
1874}
1875
1876/* legacy helper around netdev_alloc_skb() */
1877static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1878{
1879 return netdev_alloc_skb(NULL, length);
1880}
1881
1882
4915a0de
ED
1883static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1884 unsigned int length, gfp_t gfp)
61321bbd 1885{
4915a0de 1886 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
1887
1888 if (NET_IP_ALIGN && skb)
1889 skb_reserve(skb, NET_IP_ALIGN);
1890 return skb;
1891}
1892
4915a0de
ED
1893static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1894 unsigned int length)
1895{
1896 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1897}
1898
bc6fc9fa
FF
1899/**
1900 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
1901 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1902 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1903 * @order: size of the allocation
1904 *
1905 * Allocate a new page.
1906 *
1907 * %NULL is returned if there is no free memory.
1908*/
1909static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1910 struct sk_buff *skb,
1911 unsigned int order)
1912{
1913 struct page *page;
1914
1915 gfp_mask |= __GFP_COLD;
1916
1917 if (!(gfp_mask & __GFP_NOMEMALLOC))
1918 gfp_mask |= __GFP_MEMALLOC;
1919
1920 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1921 if (skb && page && page->pfmemalloc)
1922 skb->pfmemalloc = true;
1923
1924 return page;
1925}
1926
1927/**
1928 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1929 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1930 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1931 *
1932 * Allocate a new page.
1933 *
1934 * %NULL is returned if there is no free memory.
1935 */
1936static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1937 struct sk_buff *skb)
1938{
1939 return __skb_alloc_pages(gfp_mask, skb, 0);
1940}
1941
1942/**
1943 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1944 * @page: The page that was allocated from skb_alloc_page
1945 * @skb: The skb that may need pfmemalloc set
1946 */
1947static inline void skb_propagate_pfmemalloc(struct page *page,
1948 struct sk_buff *skb)
1949{
1950 if (page && page->pfmemalloc)
1951 skb->pfmemalloc = true;
1952}
1953
131ea667
IC
1954/**
1955 * skb_frag_page - retrieve the page refered to by a paged fragment
1956 * @frag: the paged fragment
1957 *
1958 * Returns the &struct page associated with @frag.
1959 */
1960static inline struct page *skb_frag_page(const skb_frag_t *frag)
1961{
a8605c60 1962 return frag->page.p;
131ea667
IC
1963}
1964
1965/**
1966 * __skb_frag_ref - take an addition reference on a paged fragment.
1967 * @frag: the paged fragment
1968 *
1969 * Takes an additional reference on the paged fragment @frag.
1970 */
1971static inline void __skb_frag_ref(skb_frag_t *frag)
1972{
1973 get_page(skb_frag_page(frag));
1974}
1975
1976/**
1977 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1978 * @skb: the buffer
1979 * @f: the fragment offset.
1980 *
1981 * Takes an additional reference on the @f'th paged fragment of @skb.
1982 */
1983static inline void skb_frag_ref(struct sk_buff *skb, int f)
1984{
1985 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1986}
1987
1988/**
1989 * __skb_frag_unref - release a reference on a paged fragment.
1990 * @frag: the paged fragment
1991 *
1992 * Releases a reference on the paged fragment @frag.
1993 */
1994static inline void __skb_frag_unref(skb_frag_t *frag)
1995{
1996 put_page(skb_frag_page(frag));
1997}
1998
1999/**
2000 * skb_frag_unref - release a reference on a paged fragment of an skb.
2001 * @skb: the buffer
2002 * @f: the fragment offset
2003 *
2004 * Releases a reference on the @f'th paged fragment of @skb.
2005 */
2006static inline void skb_frag_unref(struct sk_buff *skb, int f)
2007{
2008 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2009}
2010
2011/**
2012 * skb_frag_address - gets the address of the data contained in a paged fragment
2013 * @frag: the paged fragment buffer
2014 *
2015 * Returns the address of the data within @frag. The page must already
2016 * be mapped.
2017 */
2018static inline void *skb_frag_address(const skb_frag_t *frag)
2019{
2020 return page_address(skb_frag_page(frag)) + frag->page_offset;
2021}
2022
2023/**
2024 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2025 * @frag: the paged fragment buffer
2026 *
2027 * Returns the address of the data within @frag. Checks that the page
2028 * is mapped and returns %NULL otherwise.
2029 */
2030static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2031{
2032 void *ptr = page_address(skb_frag_page(frag));
2033 if (unlikely(!ptr))
2034 return NULL;
2035
2036 return ptr + frag->page_offset;
2037}
2038
2039/**
2040 * __skb_frag_set_page - sets the page contained in a paged fragment
2041 * @frag: the paged fragment
2042 * @page: the page to set
2043 *
2044 * Sets the fragment @frag to contain @page.
2045 */
2046static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2047{
a8605c60 2048 frag->page.p = page;
131ea667
IC
2049}
2050
2051/**
2052 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2053 * @skb: the buffer
2054 * @f: the fragment offset
2055 * @page: the page to set
2056 *
2057 * Sets the @f'th fragment of @skb to contain @page.
2058 */
2059static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2060 struct page *page)
2061{
2062 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2063}
2064
400dfd3a
ED
2065bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2066
131ea667
IC
2067/**
2068 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2069 * @dev: the device to map the fragment to
131ea667
IC
2070 * @frag: the paged fragment to map
2071 * @offset: the offset within the fragment (starting at the
2072 * fragment's own offset)
2073 * @size: the number of bytes to map
f83347df 2074 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2075 *
2076 * Maps the page associated with @frag to @device.
2077 */
2078static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2079 const skb_frag_t *frag,
2080 size_t offset, size_t size,
2081 enum dma_data_direction dir)
2082{
2083 return dma_map_page(dev, skb_frag_page(frag),
2084 frag->page_offset + offset, size, dir);
2085}
2086
117632e6
ED
2087static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2088 gfp_t gfp_mask)
2089{
2090 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2091}
2092
334a8132
PM
2093/**
2094 * skb_clone_writable - is the header of a clone writable
2095 * @skb: buffer to check
2096 * @len: length up to which to write
2097 *
2098 * Returns true if modifying the header part of the cloned buffer
2099 * does not requires the data to be copied.
2100 */
05bdd2f1 2101static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2102{
2103 return !skb_header_cloned(skb) &&
2104 skb_headroom(skb) + len <= skb->hdr_len;
2105}
2106
d9cc2048
HX
2107static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2108 int cloned)
2109{
2110 int delta = 0;
2111
d9cc2048
HX
2112 if (headroom > skb_headroom(skb))
2113 delta = headroom - skb_headroom(skb);
2114
2115 if (delta || cloned)
2116 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2117 GFP_ATOMIC);
2118 return 0;
2119}
2120
1da177e4
LT
2121/**
2122 * skb_cow - copy header of skb when it is required
2123 * @skb: buffer to cow
2124 * @headroom: needed headroom
2125 *
2126 * If the skb passed lacks sufficient headroom or its data part
2127 * is shared, data is reallocated. If reallocation fails, an error
2128 * is returned and original skb is not changed.
2129 *
2130 * The result is skb with writable area skb->head...skb->tail
2131 * and at least @headroom of space at head.
2132 */
2133static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2134{
d9cc2048
HX
2135 return __skb_cow(skb, headroom, skb_cloned(skb));
2136}
1da177e4 2137
d9cc2048
HX
2138/**
2139 * skb_cow_head - skb_cow but only making the head writable
2140 * @skb: buffer to cow
2141 * @headroom: needed headroom
2142 *
2143 * This function is identical to skb_cow except that we replace the
2144 * skb_cloned check by skb_header_cloned. It should be used when
2145 * you only need to push on some header and do not need to modify
2146 * the data.
2147 */
2148static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2149{
2150 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2151}
2152
2153/**
2154 * skb_padto - pad an skbuff up to a minimal size
2155 * @skb: buffer to pad
2156 * @len: minimal length
2157 *
2158 * Pads up a buffer to ensure the trailing bytes exist and are
2159 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2160 * is untouched. Otherwise it is extended. Returns zero on
2161 * success. The skb is freed on error.
1da177e4
LT
2162 */
2163
5b057c6b 2164static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2165{
2166 unsigned int size = skb->len;
2167 if (likely(size >= len))
5b057c6b 2168 return 0;
987c402a 2169 return skb_pad(skb, len - size);
1da177e4
LT
2170}
2171
2172static inline int skb_add_data(struct sk_buff *skb,
2173 char __user *from, int copy)
2174{
2175 const int off = skb->len;
2176
2177 if (skb->ip_summed == CHECKSUM_NONE) {
2178 int err = 0;
5084205f 2179 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2180 copy, 0, &err);
2181 if (!err) {
2182 skb->csum = csum_block_add(skb->csum, csum, off);
2183 return 0;
2184 }
2185 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2186 return 0;
2187
2188 __skb_trim(skb, off);
2189 return -EFAULT;
2190}
2191
38ba0a65
ED
2192static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2193 const struct page *page, int off)
1da177e4
LT
2194{
2195 if (i) {
9e903e08 2196 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2197
ea2ab693 2198 return page == skb_frag_page(frag) &&
9e903e08 2199 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2200 }
38ba0a65 2201 return false;
1da177e4
LT
2202}
2203
364c6bad
HX
2204static inline int __skb_linearize(struct sk_buff *skb)
2205{
2206 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2207}
2208
1da177e4
LT
2209/**
2210 * skb_linearize - convert paged skb to linear one
2211 * @skb: buffer to linarize
1da177e4
LT
2212 *
2213 * If there is no free memory -ENOMEM is returned, otherwise zero
2214 * is returned and the old skb data released.
2215 */
364c6bad
HX
2216static inline int skb_linearize(struct sk_buff *skb)
2217{
2218 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2219}
2220
cef401de
ED
2221/**
2222 * skb_has_shared_frag - can any frag be overwritten
2223 * @skb: buffer to test
2224 *
2225 * Return true if the skb has at least one frag that might be modified
2226 * by an external entity (as in vmsplice()/sendfile())
2227 */
2228static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2229{
c9af6db4
PS
2230 return skb_is_nonlinear(skb) &&
2231 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2232}
2233
364c6bad
HX
2234/**
2235 * skb_linearize_cow - make sure skb is linear and writable
2236 * @skb: buffer to process
2237 *
2238 * If there is no free memory -ENOMEM is returned, otherwise zero
2239 * is returned and the old skb data released.
2240 */
2241static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2242{
364c6bad
HX
2243 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2244 __skb_linearize(skb) : 0;
1da177e4
LT
2245}
2246
2247/**
2248 * skb_postpull_rcsum - update checksum for received skb after pull
2249 * @skb: buffer to update
2250 * @start: start of data before pull
2251 * @len: length of data pulled
2252 *
2253 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2254 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2255 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2256 */
2257
2258static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2259 const void *start, unsigned int len)
1da177e4 2260{
84fa7933 2261 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2262 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2263}
2264
cbb042f9
HX
2265unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2266
1da177e4
LT
2267/**
2268 * pskb_trim_rcsum - trim received skb and update checksum
2269 * @skb: buffer to trim
2270 * @len: new length
2271 *
2272 * This is exactly the same as pskb_trim except that it ensures the
2273 * checksum of received packets are still valid after the operation.
2274 */
2275
2276static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2277{
0e4e4220 2278 if (likely(len >= skb->len))
1da177e4 2279 return 0;
84fa7933 2280 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2281 skb->ip_summed = CHECKSUM_NONE;
2282 return __pskb_trim(skb, len);
2283}
2284
1da177e4
LT
2285#define skb_queue_walk(queue, skb) \
2286 for (skb = (queue)->next; \
a1e4891f 2287 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2288 skb = skb->next)
2289
46f8914e
JC
2290#define skb_queue_walk_safe(queue, skb, tmp) \
2291 for (skb = (queue)->next, tmp = skb->next; \
2292 skb != (struct sk_buff *)(queue); \
2293 skb = tmp, tmp = skb->next)
2294
1164f52a 2295#define skb_queue_walk_from(queue, skb) \
a1e4891f 2296 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2297 skb = skb->next)
2298
2299#define skb_queue_walk_from_safe(queue, skb, tmp) \
2300 for (tmp = skb->next; \
2301 skb != (struct sk_buff *)(queue); \
2302 skb = tmp, tmp = skb->next)
2303
300ce174
SH
2304#define skb_queue_reverse_walk(queue, skb) \
2305 for (skb = (queue)->prev; \
a1e4891f 2306 skb != (struct sk_buff *)(queue); \
300ce174
SH
2307 skb = skb->prev)
2308
686a2955
DM
2309#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2310 for (skb = (queue)->prev, tmp = skb->prev; \
2311 skb != (struct sk_buff *)(queue); \
2312 skb = tmp, tmp = skb->prev)
2313
2314#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2315 for (tmp = skb->prev; \
2316 skb != (struct sk_buff *)(queue); \
2317 skb = tmp, tmp = skb->prev)
1da177e4 2318
21dc3301 2319static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2320{
2321 return skb_shinfo(skb)->frag_list != NULL;
2322}
2323
2324static inline void skb_frag_list_init(struct sk_buff *skb)
2325{
2326 skb_shinfo(skb)->frag_list = NULL;
2327}
2328
2329static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2330{
2331 frag->next = skb_shinfo(skb)->frag_list;
2332 skb_shinfo(skb)->frag_list = frag;
2333}
2334
2335#define skb_walk_frags(skb, iter) \
2336 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2337
7965bd4d
JP
2338struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2339 int *peeked, int *off, int *err);
2340struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2341 int *err);
2342unsigned int datagram_poll(struct file *file, struct socket *sock,
2343 struct poll_table_struct *wait);
2344int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2345 struct iovec *to, int size);
2346int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2347 struct iovec *iov);
2348int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2349 const struct iovec *from, int from_offset,
2350 int len);
2351int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2352 int offset, size_t count);
2353int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2354 const struct iovec *to, int to_offset,
2355 int size);
2356void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2357void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2358int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2359__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2360 __wsum csum);
2361int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2362int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2363__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2364 int len, __wsum csum);
2365int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2366 struct pipe_inode_info *pipe, unsigned int len,
2367 unsigned int flags);
2368void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2369void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2370int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2371void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2372
2373struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2374
1da177e4
LT
2375static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2376 int len, void *buffer)
2377{
2378 int hlen = skb_headlen(skb);
2379
55820ee2 2380 if (hlen - offset >= len)
1da177e4
LT
2381 return skb->data + offset;
2382
2383 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2384 return NULL;
2385
2386 return buffer;
2387}
2388
d626f62b
ACM
2389static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2390 void *to,
2391 const unsigned int len)
2392{
2393 memcpy(to, skb->data, len);
2394}
2395
2396static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2397 const int offset, void *to,
2398 const unsigned int len)
2399{
2400 memcpy(to, skb->data + offset, len);
2401}
2402
27d7ff46
ACM
2403static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2404 const void *from,
2405 const unsigned int len)
2406{
2407 memcpy(skb->data, from, len);
2408}
2409
2410static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2411 const int offset,
2412 const void *from,
2413 const unsigned int len)
2414{
2415 memcpy(skb->data + offset, from, len);
2416}
2417
7965bd4d 2418void skb_init(void);
1da177e4 2419
ac45f602
PO
2420static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2421{
2422 return skb->tstamp;
2423}
2424
a61bbcf2
PM
2425/**
2426 * skb_get_timestamp - get timestamp from a skb
2427 * @skb: skb to get stamp from
2428 * @stamp: pointer to struct timeval to store stamp in
2429 *
2430 * Timestamps are stored in the skb as offsets to a base timestamp.
2431 * This function converts the offset back to a struct timeval and stores
2432 * it in stamp.
2433 */
ac45f602
PO
2434static inline void skb_get_timestamp(const struct sk_buff *skb,
2435 struct timeval *stamp)
a61bbcf2 2436{
b7aa0bf7 2437 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2438}
2439
ac45f602
PO
2440static inline void skb_get_timestampns(const struct sk_buff *skb,
2441 struct timespec *stamp)
2442{
2443 *stamp = ktime_to_timespec(skb->tstamp);
2444}
2445
b7aa0bf7 2446static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2447{
b7aa0bf7 2448 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2449}
2450
164891aa
SH
2451static inline ktime_t net_timedelta(ktime_t t)
2452{
2453 return ktime_sub(ktime_get_real(), t);
2454}
2455
b9ce204f
IJ
2456static inline ktime_t net_invalid_timestamp(void)
2457{
2458 return ktime_set(0, 0);
2459}
a61bbcf2 2460
7965bd4d 2461void skb_timestamping_init(void);
c1f19b51
RC
2462
2463#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2464
7965bd4d
JP
2465void skb_clone_tx_timestamp(struct sk_buff *skb);
2466bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2467
2468#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2469
2470static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2471{
2472}
2473
2474static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2475{
2476 return false;
2477}
2478
2479#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2480
2481/**
2482 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2483 *
da92b194
RC
2484 * PHY drivers may accept clones of transmitted packets for
2485 * timestamping via their phy_driver.txtstamp method. These drivers
2486 * must call this function to return the skb back to the stack, with
2487 * or without a timestamp.
2488 *
c1f19b51 2489 * @skb: clone of the the original outgoing packet
da92b194 2490 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2491 *
2492 */
2493void skb_complete_tx_timestamp(struct sk_buff *skb,
2494 struct skb_shared_hwtstamps *hwtstamps);
2495
ac45f602
PO
2496/**
2497 * skb_tstamp_tx - queue clone of skb with send time stamps
2498 * @orig_skb: the original outgoing packet
2499 * @hwtstamps: hardware time stamps, may be NULL if not available
2500 *
2501 * If the skb has a socket associated, then this function clones the
2502 * skb (thus sharing the actual data and optional structures), stores
2503 * the optional hardware time stamping information (if non NULL) or
2504 * generates a software time stamp (otherwise), then queues the clone
2505 * to the error queue of the socket. Errors are silently ignored.
2506 */
7965bd4d
JP
2507void skb_tstamp_tx(struct sk_buff *orig_skb,
2508 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2509
4507a715
RC
2510static inline void sw_tx_timestamp(struct sk_buff *skb)
2511{
2244d07b
OH
2512 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2513 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2514 skb_tstamp_tx(skb, NULL);
2515}
2516
2517/**
2518 * skb_tx_timestamp() - Driver hook for transmit timestamping
2519 *
2520 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2521 * function immediately before giving the sk_buff to the MAC hardware.
4507a715
RC
2522 *
2523 * @skb: A socket buffer.
2524 */
2525static inline void skb_tx_timestamp(struct sk_buff *skb)
2526{
c1f19b51 2527 skb_clone_tx_timestamp(skb);
4507a715
RC
2528 sw_tx_timestamp(skb);
2529}
2530
6e3e939f
JB
2531/**
2532 * skb_complete_wifi_ack - deliver skb with wifi status
2533 *
2534 * @skb: the original outgoing packet
2535 * @acked: ack status
2536 *
2537 */
2538void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2539
7965bd4d
JP
2540__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2541__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2542
60476372
HX
2543static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2544{
2545 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2546}
2547
fb286bb2
HX
2548/**
2549 * skb_checksum_complete - Calculate checksum of an entire packet
2550 * @skb: packet to process
2551 *
2552 * This function calculates the checksum over the entire packet plus
2553 * the value of skb->csum. The latter can be used to supply the
2554 * checksum of a pseudo header as used by TCP/UDP. It returns the
2555 * checksum.
2556 *
2557 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2558 * this function can be used to verify that checksum on received
2559 * packets. In that case the function should return zero if the
2560 * checksum is correct. In particular, this function will return zero
2561 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2562 * hardware has already verified the correctness of the checksum.
2563 */
4381ca3c 2564static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2565{
60476372
HX
2566 return skb_csum_unnecessary(skb) ?
2567 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2568}
2569
5f79e0f9 2570#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2571void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2572static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2573{
2574 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2575 nf_conntrack_destroy(nfct);
1da177e4
LT
2576}
2577static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2578{
2579 if (nfct)
2580 atomic_inc(&nfct->use);
2581}
2fc72c7b
KK
2582#endif
2583#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
9fb9cbb1
YK
2584static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2585{
2586 if (skb)
2587 atomic_inc(&skb->users);
2588}
2589static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2590{
2591 if (skb)
2592 kfree_skb(skb);
2593}
2594#endif
1da177e4
LT
2595#ifdef CONFIG_BRIDGE_NETFILTER
2596static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2597{
2598 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2599 kfree(nf_bridge);
2600}
2601static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2602{
2603 if (nf_bridge)
2604 atomic_inc(&nf_bridge->use);
2605}
2606#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2607static inline void nf_reset(struct sk_buff *skb)
2608{
5f79e0f9 2609#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2610 nf_conntrack_put(skb->nfct);
2611 skb->nfct = NULL;
2fc72c7b
KK
2612#endif
2613#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
a193a4ab
PM
2614 nf_conntrack_put_reasm(skb->nfct_reasm);
2615 skb->nfct_reasm = NULL;
2616#endif
2617#ifdef CONFIG_BRIDGE_NETFILTER
2618 nf_bridge_put(skb->nf_bridge);
2619 skb->nf_bridge = NULL;
2620#endif
2621}
2622
124dff01
PM
2623static inline void nf_reset_trace(struct sk_buff *skb)
2624{
130549fe
G
2625#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2626 skb->nf_trace = 0;
2627#endif
a193a4ab
PM
2628}
2629
edda553c
YK
2630/* Note: This doesn't put any conntrack and bridge info in dst. */
2631static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2632{
5f79e0f9 2633#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2634 dst->nfct = src->nfct;
2635 nf_conntrack_get(src->nfct);
2636 dst->nfctinfo = src->nfctinfo;
2fc72c7b
KK
2637#endif
2638#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
edda553c
YK
2639 dst->nfct_reasm = src->nfct_reasm;
2640 nf_conntrack_get_reasm(src->nfct_reasm);
2641#endif
2642#ifdef CONFIG_BRIDGE_NETFILTER
2643 dst->nf_bridge = src->nf_bridge;
2644 nf_bridge_get(src->nf_bridge);
2645#endif
2646}
2647
e7ac05f3
YK
2648static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2649{
e7ac05f3 2650#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2651 nf_conntrack_put(dst->nfct);
2fc72c7b
KK
2652#endif
2653#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
e7ac05f3
YK
2654 nf_conntrack_put_reasm(dst->nfct_reasm);
2655#endif
2656#ifdef CONFIG_BRIDGE_NETFILTER
2657 nf_bridge_put(dst->nf_bridge);
2658#endif
2659 __nf_copy(dst, src);
2660}
2661
984bc16c
JM
2662#ifdef CONFIG_NETWORK_SECMARK
2663static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2664{
2665 to->secmark = from->secmark;
2666}
2667
2668static inline void skb_init_secmark(struct sk_buff *skb)
2669{
2670 skb->secmark = 0;
2671}
2672#else
2673static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2674{ }
2675
2676static inline void skb_init_secmark(struct sk_buff *skb)
2677{ }
2678#endif
2679
f25f4e44
PWJ
2680static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2681{
f25f4e44 2682 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2683}
2684
9247744e 2685static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2686{
4e3ab47a 2687 return skb->queue_mapping;
4e3ab47a
PE
2688}
2689
f25f4e44
PWJ
2690static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2691{
f25f4e44 2692 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2693}
2694
d5a9e24a
DM
2695static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2696{
2697 skb->queue_mapping = rx_queue + 1;
2698}
2699
9247744e 2700static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
2701{
2702 return skb->queue_mapping - 1;
2703}
2704
9247744e 2705static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 2706{
a02cec21 2707 return skb->queue_mapping != 0;
d5a9e24a
DM
2708}
2709
7965bd4d
JP
2710u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2711 unsigned int num_tx_queues);
9247744e 2712
def8b4fa
AD
2713static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2714{
0b3d8e08 2715#ifdef CONFIG_XFRM
def8b4fa 2716 return skb->sp;
def8b4fa 2717#else
def8b4fa 2718 return NULL;
def8b4fa 2719#endif
0b3d8e08 2720}
def8b4fa 2721
68c33163
PS
2722/* Keeps track of mac header offset relative to skb->head.
2723 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2724 * For non-tunnel skb it points to skb_mac_header() and for
2725 * tunnel skb it points to outer mac header. */
2726struct skb_gso_cb {
2727 int mac_offset;
2728};
2729#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2730
2731static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2732{
2733 return (skb_mac_header(inner_skb) - inner_skb->head) -
2734 SKB_GSO_CB(inner_skb)->mac_offset;
2735}
2736
1e2bd517
PS
2737static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2738{
2739 int new_headroom, headroom;
2740 int ret;
2741
2742 headroom = skb_headroom(skb);
2743 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2744 if (ret)
2745 return ret;
2746
2747 new_headroom = skb_headroom(skb);
2748 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2749 return 0;
2750}
2751
bdcc0924 2752static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
2753{
2754 return skb_shinfo(skb)->gso_size;
2755}
2756
36a8f39e 2757/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 2758static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
2759{
2760 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2761}
2762
7965bd4d 2763void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
2764
2765static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2766{
2767 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2768 * wanted then gso_type will be set. */
05bdd2f1
ED
2769 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2770
b78462eb
AD
2771 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2772 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
2773 __skb_warn_lro_forwarding(skb);
2774 return true;
2775 }
2776 return false;
2777}
2778
35fc92a9
HX
2779static inline void skb_forward_csum(struct sk_buff *skb)
2780{
2781 /* Unfortunately we don't support this one. Any brave souls? */
2782 if (skb->ip_summed == CHECKSUM_COMPLETE)
2783 skb->ip_summed = CHECKSUM_NONE;
2784}
2785
bc8acf2c
ED
2786/**
2787 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2788 * @skb: skb to check
2789 *
2790 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2791 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2792 * use this helper, to document places where we make this assertion.
2793 */
05bdd2f1 2794static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
2795{
2796#ifdef DEBUG
2797 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2798#endif
2799}
2800
f35d9d8a 2801bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 2802
f77668dc
DB
2803u32 __skb_get_poff(const struct sk_buff *skb);
2804
3a7c1ee4
AD
2805/**
2806 * skb_head_is_locked - Determine if the skb->head is locked down
2807 * @skb: skb to check
2808 *
2809 * The head on skbs build around a head frag can be removed if they are
2810 * not cloned. This function returns true if the skb head is locked down
2811 * due to either being allocated via kmalloc, or by being a clone with
2812 * multiple references to the head.
2813 */
2814static inline bool skb_head_is_locked(const struct sk_buff *skb)
2815{
2816 return !skb->head_frag || skb_cloned(skb);
2817}
1da177e4
LT
2818#endif /* __KERNEL__ */
2819#endif /* _LINUX_SKBUFF_H */