Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
[deliverable/linux.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>
b7aa0bf7 31#include <linux/hrtimer.h>
131ea667 32#include <linux/dma-mapping.h>
c8f44aff 33#include <linux/netdev_features.h>
363ec392 34#include <linux/sched.h>
5203cd28 35#include <net/flow_keys.h>
1da177e4 36
78ea85f1
DB
37/* A. Checksumming of received packets by device.
38 *
39 * CHECKSUM_NONE:
40 *
41 * Device failed to checksum this packet e.g. due to lack of capabilities.
42 * The packet contains full (though not verified) checksum in packet but
43 * not in skb->csum. Thus, skb->csum is undefined in this case.
44 *
45 * CHECKSUM_UNNECESSARY:
46 *
47 * The hardware you're dealing with doesn't calculate the full checksum
48 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
49 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
50 * if their checksums are okay. skb->csum is still undefined in this case
51 * though. It is a bad option, but, unfortunately, nowadays most vendors do
52 * this. Apparently with the secret goal to sell you new devices, when you
53 * will add new protocol to your host, f.e. IPv6 8)
54 *
55 * CHECKSUM_UNNECESSARY is applicable to following protocols:
56 * TCP: IPv6 and IPv4.
57 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
58 * zero UDP checksum for either IPv4 or IPv6, the networking stack
59 * may perform further validation in this case.
60 * GRE: only if the checksum is present in the header.
61 * SCTP: indicates the CRC in SCTP header has been validated.
62 *
63 * skb->csum_level indicates the number of consecutive checksums found in
64 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
65 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
66 * and a device is able to verify the checksums for UDP (possibly zero),
67 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
68 * two. If the device were only able to verify the UDP checksum and not
69 * GRE, either because it doesn't support GRE checksum of because GRE
70 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
71 * not considered in this case).
78ea85f1
DB
72 *
73 * CHECKSUM_COMPLETE:
74 *
75 * This is the most generic way. The device supplied checksum of the _whole_
76 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
77 * hardware doesn't need to parse L3/L4 headers to implement this.
78 *
79 * Note: Even if device supports only some protocols, but is able to produce
80 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
81 *
82 * CHECKSUM_PARTIAL:
83 *
84 * This is identical to the case for output below. This may occur on a packet
85 * received directly from another Linux OS, e.g., a virtualized Linux kernel
86 * on the same host. The packet can be treated in the same way as
87 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
88 * checksum must be filled in by the OS or the hardware.
89 *
90 * B. Checksumming on output.
91 *
92 * CHECKSUM_NONE:
93 *
94 * The skb was already checksummed by the protocol, or a checksum is not
95 * required.
96 *
97 * CHECKSUM_PARTIAL:
98 *
99 * The device is required to checksum the packet as seen by hard_start_xmit()
100 * from skb->csum_start up to the end, and to record/write the checksum at
101 * offset skb->csum_start + skb->csum_offset.
102 *
103 * The device must show its capabilities in dev->features, set up at device
104 * setup time, e.g. netdev_features.h:
105 *
106 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
107 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
108 * IPv4. Sigh. Vendors like this way for an unknown reason.
109 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
110 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
111 * NETIF_F_... - Well, you get the picture.
112 *
113 * CHECKSUM_UNNECESSARY:
114 *
115 * Normally, the device will do per protocol specific checksumming. Protocol
116 * implementations that do not want the NIC to perform the checksum
117 * calculation should use this flag in their outgoing skbs.
118 *
119 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
120 * offload. Correspondingly, the FCoE protocol driver
121 * stack should use CHECKSUM_UNNECESSARY.
122 *
123 * Any questions? No questions, good. --ANK
124 */
125
60476372 126/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
127#define CHECKSUM_NONE 0
128#define CHECKSUM_UNNECESSARY 1
129#define CHECKSUM_COMPLETE 2
130#define CHECKSUM_PARTIAL 3
1da177e4 131
77cffe23
TH
132/* Maximum value in skb->csum_level */
133#define SKB_MAX_CSUM_LEVEL 3
134
0bec8c88 135#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 136#define SKB_WITH_OVERHEAD(X) \
deea84b0 137 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
138#define SKB_MAX_ORDER(X, ORDER) \
139 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
140#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
141#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
142
87fb4b7b
ED
143/* return minimum truesize of one skb containing X bytes of data */
144#define SKB_TRUESIZE(X) ((X) + \
145 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
146 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
147
1da177e4 148struct net_device;
716ea3a7 149struct scatterlist;
9c55e01c 150struct pipe_inode_info;
1da177e4 151
5f79e0f9 152#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
153struct nf_conntrack {
154 atomic_t use;
1da177e4 155};
5f79e0f9 156#endif
1da177e4 157
34666d46 158#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 159struct nf_bridge_info {
bf1ac5ca
ED
160 atomic_t use;
161 unsigned int mask;
162 struct net_device *physindev;
163 struct net_device *physoutdev;
164 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
165};
166#endif
167
1da177e4
LT
168struct sk_buff_head {
169 /* These two members must be first. */
170 struct sk_buff *next;
171 struct sk_buff *prev;
172
173 __u32 qlen;
174 spinlock_t lock;
175};
176
177struct sk_buff;
178
9d4dde52
IC
179/* To allow 64K frame to be packed as single skb without frag_list we
180 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
181 * buffers which do not start on a page boundary.
182 *
183 * Since GRO uses frags we allocate at least 16 regardless of page
184 * size.
a715dea3 185 */
9d4dde52 186#if (65536/PAGE_SIZE + 1) < 16
eec00954 187#define MAX_SKB_FRAGS 16UL
a715dea3 188#else
9d4dde52 189#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 190#endif
1da177e4
LT
191
192typedef struct skb_frag_struct skb_frag_t;
193
194struct skb_frag_struct {
a8605c60
IC
195 struct {
196 struct page *p;
197 } page;
cb4dfe56 198#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
199 __u32 page_offset;
200 __u32 size;
cb4dfe56
ED
201#else
202 __u16 page_offset;
203 __u16 size;
204#endif
1da177e4
LT
205};
206
9e903e08
ED
207static inline unsigned int skb_frag_size(const skb_frag_t *frag)
208{
209 return frag->size;
210}
211
212static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
213{
214 frag->size = size;
215}
216
217static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
218{
219 frag->size += delta;
220}
221
222static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
223{
224 frag->size -= delta;
225}
226
ac45f602
PO
227#define HAVE_HW_TIME_STAMP
228
229/**
d3a21be8 230 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
231 * @hwtstamp: hardware time stamp transformed into duration
232 * since arbitrary point in time
ac45f602
PO
233 *
234 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 235 * skb->tstamp.
ac45f602
PO
236 *
237 * hwtstamps can only be compared against other hwtstamps from
238 * the same device.
239 *
240 * This structure is attached to packets as part of the
241 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
242 */
243struct skb_shared_hwtstamps {
244 ktime_t hwtstamp;
ac45f602
PO
245};
246
2244d07b
OH
247/* Definitions for tx_flags in struct skb_shared_info */
248enum {
249 /* generate hardware time stamp */
250 SKBTX_HW_TSTAMP = 1 << 0,
251
e7fd2885 252 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
253 SKBTX_SW_TSTAMP = 1 << 1,
254
255 /* device driver is going to provide hardware time stamp */
256 SKBTX_IN_PROGRESS = 1 << 2,
257
a6686f2f 258 /* device driver supports TX zero-copy buffers */
62b1a8ab 259 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
260
261 /* generate wifi status information (where possible) */
62b1a8ab 262 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
263
264 /* This indicates at least one fragment might be overwritten
265 * (as in vmsplice(), sendfile() ...)
266 * If we need to compute a TX checksum, we'll need to copy
267 * all frags to avoid possible bad checksum
268 */
269 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
270
271 /* generate software time stamp when entering packet scheduling */
272 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
273
274 /* generate software timestamp on peer data acknowledgment */
275 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
276};
277
e1c8a607
WB
278#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
279 SKBTX_SCHED_TSTAMP | \
280 SKBTX_ACK_TSTAMP)
f24b9be5
WB
281#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
282
a6686f2f
SM
283/*
284 * The callback notifies userspace to release buffers when skb DMA is done in
285 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
286 * The zerocopy_success argument is true if zero copy transmit occurred,
287 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
288 * The ctx field is used to track device context.
289 * The desc field is used to track userspace buffer index.
a6686f2f
SM
290 */
291struct ubuf_info {
e19d6763 292 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 293 void *ctx;
a6686f2f 294 unsigned long desc;
ac45f602
PO
295};
296
1da177e4
LT
297/* This data is invariant across clones and lives at
298 * the end of the header data, ie. at skb->end.
299 */
300struct skb_shared_info {
9f42f126
IC
301 unsigned char nr_frags;
302 __u8 tx_flags;
7967168c
HX
303 unsigned short gso_size;
304 /* Warning: this field is not always filled in (UFO)! */
305 unsigned short gso_segs;
306 unsigned short gso_type;
1da177e4 307 struct sk_buff *frag_list;
ac45f602 308 struct skb_shared_hwtstamps hwtstamps;
09c2d251 309 u32 tskey;
9f42f126 310 __be32 ip6_frag_id;
ec7d2f2c
ED
311
312 /*
313 * Warning : all fields before dataref are cleared in __alloc_skb()
314 */
315 atomic_t dataref;
316
69e3c75f
JB
317 /* Intermediate layers must ensure that destructor_arg
318 * remains valid until skb destructor */
319 void * destructor_arg;
a6686f2f 320
fed66381
ED
321 /* must be last field, see pskb_expand_head() */
322 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
323};
324
325/* We divide dataref into two halves. The higher 16 bits hold references
326 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
327 * the entire skb->data. A clone of a headerless skb holds the length of
328 * the header in skb->hdr_len.
1da177e4
LT
329 *
330 * All users must obey the rule that the skb->data reference count must be
331 * greater than or equal to the payload reference count.
332 *
333 * Holding a reference to the payload part means that the user does not
334 * care about modifications to the header part of skb->data.
335 */
336#define SKB_DATAREF_SHIFT 16
337#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
338
d179cd12
DM
339
340enum {
c8753d55
VS
341 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
342 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
343 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
344 SKB_FCLONE_FREE, /* this companion fclone skb is available */
d179cd12
DM
345};
346
7967168c
HX
347enum {
348 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 349 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
350
351 /* This indicates the skb is from an untrusted source. */
352 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
353
354 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
355 SKB_GSO_TCP_ECN = 1 << 3,
356
357 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
358
359 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
360
361 SKB_GSO_GRE = 1 << 6,
73136267 362
4b28252c 363 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 364
4b28252c 365 SKB_GSO_IPIP = 1 << 8,
cb32f511 366
4b28252c 367 SKB_GSO_SIT = 1 << 9,
61c1db7f 368
4b28252c 369 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
370
371 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 372
4b28252c
TH
373 SKB_GSO_MPLS = 1 << 12,
374
7967168c
HX
375};
376
2e07fa9c
ACM
377#if BITS_PER_LONG > 32
378#define NET_SKBUFF_DATA_USES_OFFSET 1
379#endif
380
381#ifdef NET_SKBUFF_DATA_USES_OFFSET
382typedef unsigned int sk_buff_data_t;
383#else
384typedef unsigned char *sk_buff_data_t;
385#endif
386
363ec392
ED
387/**
388 * struct skb_mstamp - multi resolution time stamps
389 * @stamp_us: timestamp in us resolution
390 * @stamp_jiffies: timestamp in jiffies
391 */
392struct skb_mstamp {
393 union {
394 u64 v64;
395 struct {
396 u32 stamp_us;
397 u32 stamp_jiffies;
398 };
399 };
400};
401
402/**
403 * skb_mstamp_get - get current timestamp
404 * @cl: place to store timestamps
405 */
406static inline void skb_mstamp_get(struct skb_mstamp *cl)
407{
408 u64 val = local_clock();
409
410 do_div(val, NSEC_PER_USEC);
411 cl->stamp_us = (u32)val;
412 cl->stamp_jiffies = (u32)jiffies;
413}
414
415/**
416 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
417 * @t1: pointer to newest sample
418 * @t0: pointer to oldest sample
419 */
420static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
421 const struct skb_mstamp *t0)
422{
423 s32 delta_us = t1->stamp_us - t0->stamp_us;
424 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
425
426 /* If delta_us is negative, this might be because interval is too big,
427 * or local_clock() drift is too big : fallback using jiffies.
428 */
429 if (delta_us <= 0 ||
430 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
431
432 delta_us = jiffies_to_usecs(delta_jiffies);
433
434 return delta_us;
435}
436
437
1da177e4
LT
438/**
439 * struct sk_buff - socket buffer
440 * @next: Next buffer in list
441 * @prev: Previous buffer in list
363ec392 442 * @tstamp: Time we arrived/left
d84e0bd7 443 * @sk: Socket we are owned by
1da177e4 444 * @dev: Device we arrived on/are leaving by
d84e0bd7 445 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 446 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 447 * @sp: the security path, used for xfrm
1da177e4
LT
448 * @len: Length of actual data
449 * @data_len: Data length
450 * @mac_len: Length of link layer header
334a8132 451 * @hdr_len: writable header length of cloned skb
663ead3b
HX
452 * @csum: Checksum (must include start/offset pair)
453 * @csum_start: Offset from skb->head where checksumming should start
454 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 455 * @priority: Packet queueing priority
60ff7467 456 * @ignore_df: allow local fragmentation
1da177e4 457 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 458 * @ip_summed: Driver fed us an IP checksum
1da177e4 459 * @nohdr: Payload reference only, must not modify header
d84e0bd7 460 * @nfctinfo: Relationship of this skb to the connection
1da177e4 461 * @pkt_type: Packet class
c83c2486 462 * @fclone: skbuff clone status
c83c2486 463 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
464 * @peeked: this packet has been seen already, so stats have been
465 * done for it, don't do them again
ba9dda3a 466 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
467 * @protocol: Packet protocol from driver
468 * @destructor: Destruct function
469 * @nfct: Associated connection, if any
1da177e4 470 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 471 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
472 * @tc_index: Traffic control index
473 * @tc_verd: traffic control verdict
61b905da 474 * @hash: the packet hash
d84e0bd7 475 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 476 * @xmit_more: More SKBs are pending for this queue
553a5672 477 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 478 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 479 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 480 * ports.
a3b18ddb 481 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
482 * @wifi_acked_valid: wifi_acked was set
483 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 484 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 485 * @napi_id: id of the NAPI struct this skb came from
984bc16c 486 * @secmark: security marking
d84e0bd7
DB
487 * @mark: Generic packet mark
488 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 489 * @vlan_proto: vlan encapsulation protocol
6aa895b0 490 * @vlan_tci: vlan tag control information
0d89d203 491 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
492 * @inner_transport_header: Inner transport layer header (encapsulation)
493 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 494 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
495 * @transport_header: Transport layer header
496 * @network_header: Network layer header
497 * @mac_header: Link layer header
498 * @tail: Tail pointer
499 * @end: End pointer
500 * @head: Head of buffer
501 * @data: Data head pointer
502 * @truesize: Buffer size
503 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
504 */
505
506struct sk_buff {
507 /* These two members must be first. */
508 struct sk_buff *next;
509 struct sk_buff *prev;
510
363ec392
ED
511 union {
512 ktime_t tstamp;
513 struct skb_mstamp skb_mstamp;
514 };
da3f5cf1
FF
515
516 struct sock *sk;
1da177e4 517 struct net_device *dev;
1da177e4 518
1da177e4
LT
519 /*
520 * This is the control buffer. It is free to use for every
521 * layer. Please put your private variables there. If you
522 * want to keep them across layers you have to do a skb_clone()
523 * first. This is owned by whoever has the skb queued ATM.
524 */
da3f5cf1 525 char cb[48] __aligned(8);
1da177e4 526
7fee226a 527 unsigned long _skb_refdst;
b1937227 528 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
529#ifdef CONFIG_XFRM
530 struct sec_path *sp;
b1937227
ED
531#endif
532#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
533 struct nf_conntrack *nfct;
534#endif
85224844 535#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 536 struct nf_bridge_info *nf_bridge;
da3f5cf1 537#endif
1da177e4 538 unsigned int len,
334a8132
PM
539 data_len;
540 __u16 mac_len,
541 hdr_len;
b1937227
ED
542
543 /* Following fields are _not_ copied in __copy_skb_header()
544 * Note that queue_mapping is here mostly to fill a hole.
545 */
fe55f6d5 546 kmemcheck_bitfield_begin(flags1);
b1937227
ED
547 __u16 queue_mapping;
548 __u8 cloned:1,
6869c4d8 549 nohdr:1,
b84f4cc9 550 fclone:2,
a59322be 551 peeked:1,
b1937227
ED
552 head_frag:1,
553 xmit_more:1;
554 /* one bit hole */
fe55f6d5 555 kmemcheck_bitfield_end(flags1);
4031ae6e 556
b1937227
ED
557 /* fields enclosed in headers_start/headers_end are copied
558 * using a single memcpy() in __copy_skb_header()
559 */
560 __u32 headers_start[0];
4031ae6e 561
233577a2
HFS
562/* if you move pkt_type around you also must adapt those constants */
563#ifdef __BIG_ENDIAN_BITFIELD
564#define PKT_TYPE_MAX (7 << 5)
565#else
566#define PKT_TYPE_MAX 7
1da177e4 567#endif
233577a2 568#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 569
233577a2 570 __u8 __pkt_type_offset[0];
b1937227 571 __u8 pkt_type:3;
c93bdd0e 572 __u8 pfmemalloc:1;
b1937227
ED
573 __u8 ignore_df:1;
574 __u8 nfctinfo:3;
575
576 __u8 nf_trace:1;
577 __u8 ip_summed:2;
3853b584 578 __u8 ooo_okay:1;
61b905da 579 __u8 l4_hash:1;
a3b18ddb 580 __u8 sw_hash:1;
6e3e939f
JB
581 __u8 wifi_acked_valid:1;
582 __u8 wifi_acked:1;
b1937227 583
3bdc0eba 584 __u8 no_fcs:1;
77cffe23 585 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 586 __u8 encapsulation:1;
7e2b10c1 587 __u8 encap_hdr_csum:1;
5d0c2b95 588 __u8 csum_valid:1;
7e3cead5 589 __u8 csum_complete_sw:1;
b1937227
ED
590 __u8 csum_level:2;
591 __u8 csum_bad:1;
fe55f6d5 592
b1937227
ED
593#ifdef CONFIG_IPV6_NDISC_NODETYPE
594 __u8 ndisc_nodetype:2;
595#endif
596 __u8 ipvs_property:1;
8bce6d7d
TH
597 __u8 inner_protocol_type:1;
598 /* 4 or 6 bit hole */
b1937227
ED
599
600#ifdef CONFIG_NET_SCHED
601 __u16 tc_index; /* traffic control index */
602#ifdef CONFIG_NET_CLS_ACT
603 __u16 tc_verd; /* traffic control verdict */
604#endif
605#endif
fe55f6d5 606
b1937227
ED
607 union {
608 __wsum csum;
609 struct {
610 __u16 csum_start;
611 __u16 csum_offset;
612 };
613 };
614 __u32 priority;
615 int skb_iif;
616 __u32 hash;
617 __be16 vlan_proto;
618 __u16 vlan_tci;
7bced397
DW
619#ifdef CONFIG_NET_RX_BUSY_POLL
620 unsigned int napi_id;
97fc2f08 621#endif
984bc16c
JM
622#ifdef CONFIG_NETWORK_SECMARK
623 __u32 secmark;
624#endif
3b885787
NH
625 union {
626 __u32 mark;
627 __u32 dropcount;
16fad69c 628 __u32 reserved_tailroom;
3b885787 629 };
1da177e4 630
8bce6d7d
TH
631 union {
632 __be16 inner_protocol;
633 __u8 inner_ipproto;
634 };
635
1a37e412
SH
636 __u16 inner_transport_header;
637 __u16 inner_network_header;
638 __u16 inner_mac_header;
b1937227
ED
639
640 __be16 protocol;
1a37e412
SH
641 __u16 transport_header;
642 __u16 network_header;
643 __u16 mac_header;
b1937227
ED
644
645 __u32 headers_end[0];
646
1da177e4 647 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 648 sk_buff_data_t tail;
4305b541 649 sk_buff_data_t end;
1da177e4 650 unsigned char *head,
4305b541 651 *data;
27a884dc
ACM
652 unsigned int truesize;
653 atomic_t users;
1da177e4
LT
654};
655
656#ifdef __KERNEL__
657/*
658 * Handling routines are only of interest to the kernel
659 */
660#include <linux/slab.h>
661
1da177e4 662
c93bdd0e
MG
663#define SKB_ALLOC_FCLONE 0x01
664#define SKB_ALLOC_RX 0x02
665
666/* Returns true if the skb was allocated from PFMEMALLOC reserves */
667static inline bool skb_pfmemalloc(const struct sk_buff *skb)
668{
669 return unlikely(skb->pfmemalloc);
670}
671
7fee226a
ED
672/*
673 * skb might have a dst pointer attached, refcounted or not.
674 * _skb_refdst low order bit is set if refcount was _not_ taken
675 */
676#define SKB_DST_NOREF 1UL
677#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
678
679/**
680 * skb_dst - returns skb dst_entry
681 * @skb: buffer
682 *
683 * Returns skb dst_entry, regardless of reference taken or not.
684 */
adf30907
ED
685static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
686{
7fee226a
ED
687 /* If refdst was not refcounted, check we still are in a
688 * rcu_read_lock section
689 */
690 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
691 !rcu_read_lock_held() &&
692 !rcu_read_lock_bh_held());
693 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
694}
695
7fee226a
ED
696/**
697 * skb_dst_set - sets skb dst
698 * @skb: buffer
699 * @dst: dst entry
700 *
701 * Sets skb dst, assuming a reference was taken on dst and should
702 * be released by skb_dst_drop()
703 */
adf30907
ED
704static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
705{
7fee226a
ED
706 skb->_skb_refdst = (unsigned long)dst;
707}
708
7965bd4d
JP
709void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
710 bool force);
932bc4d7
JA
711
712/**
713 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
714 * @skb: buffer
715 * @dst: dst entry
716 *
717 * Sets skb dst, assuming a reference was not taken on dst.
718 * If dst entry is cached, we do not take reference and dst_release
719 * will be avoided by refdst_drop. If dst entry is not cached, we take
720 * reference, so that last dst_release can destroy the dst immediately.
721 */
722static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
723{
724 __skb_dst_set_noref(skb, dst, false);
725}
726
727/**
728 * skb_dst_set_noref_force - sets skb dst, without taking reference
729 * @skb: buffer
730 * @dst: dst entry
731 *
732 * Sets skb dst, assuming a reference was not taken on dst.
733 * No reference is taken and no dst_release will be called. While for
734 * cached dsts deferred reclaim is a basic feature, for entries that are
735 * not cached it is caller's job to guarantee that last dst_release for
736 * provided dst happens when nobody uses it, eg. after a RCU grace period.
737 */
738static inline void skb_dst_set_noref_force(struct sk_buff *skb,
739 struct dst_entry *dst)
740{
741 __skb_dst_set_noref(skb, dst, true);
742}
7fee226a
ED
743
744/**
25985edc 745 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
746 * @skb: buffer
747 */
748static inline bool skb_dst_is_noref(const struct sk_buff *skb)
749{
750 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
751}
752
511c3f92
ED
753static inline struct rtable *skb_rtable(const struct sk_buff *skb)
754{
adf30907 755 return (struct rtable *)skb_dst(skb);
511c3f92
ED
756}
757
7965bd4d
JP
758void kfree_skb(struct sk_buff *skb);
759void kfree_skb_list(struct sk_buff *segs);
760void skb_tx_error(struct sk_buff *skb);
761void consume_skb(struct sk_buff *skb);
762void __kfree_skb(struct sk_buff *skb);
d7e8883c 763extern struct kmem_cache *skbuff_head_cache;
bad43ca8 764
7965bd4d
JP
765void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
766bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
767 bool *fragstolen, int *delta_truesize);
bad43ca8 768
7965bd4d
JP
769struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
770 int node);
771struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 772static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 773 gfp_t priority)
d179cd12 774{
564824b0 775 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
776}
777
2e4e4410
ED
778struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
779 unsigned long data_len,
780 int max_page_order,
781 int *errcode,
782 gfp_t gfp_mask);
783
d0bf4a9e
ED
784/* Layout of fast clones : [skb1][skb2][fclone_ref] */
785struct sk_buff_fclones {
786 struct sk_buff skb1;
787
788 struct sk_buff skb2;
789
790 atomic_t fclone_ref;
791};
792
793/**
794 * skb_fclone_busy - check if fclone is busy
795 * @skb: buffer
796 *
797 * Returns true is skb is a fast clone, and its clone is not freed.
798 */
799static inline bool skb_fclone_busy(const struct sk_buff *skb)
800{
801 const struct sk_buff_fclones *fclones;
802
803 fclones = container_of(skb, struct sk_buff_fclones, skb1);
804
805 return skb->fclone == SKB_FCLONE_ORIG &&
806 fclones->skb2.fclone == SKB_FCLONE_CLONE;
807}
808
d179cd12 809static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 810 gfp_t priority)
d179cd12 811{
c93bdd0e 812 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
813}
814
7965bd4d 815struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
816static inline struct sk_buff *alloc_skb_head(gfp_t priority)
817{
818 return __alloc_skb_head(priority, -1);
819}
820
7965bd4d
JP
821struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
822int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
823struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
824struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
825struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
826 gfp_t gfp_mask, bool fclone);
827static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
828 gfp_t gfp_mask)
829{
830 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
831}
7965bd4d
JP
832
833int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
834struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
835 unsigned int headroom);
836struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
837 int newtailroom, gfp_t priority);
25a91d8d
FD
838int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
839 int offset, int len);
7965bd4d
JP
840int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
841 int len);
842int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
843int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 844#define dev_kfree_skb(a) consume_skb(a)
1da177e4 845
7965bd4d
JP
846int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
847 int getfrag(void *from, char *to, int offset,
848 int len, int odd, struct sk_buff *skb),
849 void *from, int length);
e89e9cf5 850
d94d9fee 851struct skb_seq_state {
677e90ed
TG
852 __u32 lower_offset;
853 __u32 upper_offset;
854 __u32 frag_idx;
855 __u32 stepped_offset;
856 struct sk_buff *root_skb;
857 struct sk_buff *cur_skb;
858 __u8 *frag_data;
859};
860
7965bd4d
JP
861void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
862 unsigned int to, struct skb_seq_state *st);
863unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
864 struct skb_seq_state *st);
865void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 866
7965bd4d
JP
867unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
868 unsigned int to, struct ts_config *config,
869 struct ts_state *state);
3fc7e8a6 870
09323cc4
TH
871/*
872 * Packet hash types specify the type of hash in skb_set_hash.
873 *
874 * Hash types refer to the protocol layer addresses which are used to
875 * construct a packet's hash. The hashes are used to differentiate or identify
876 * flows of the protocol layer for the hash type. Hash types are either
877 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
878 *
879 * Properties of hashes:
880 *
881 * 1) Two packets in different flows have different hash values
882 * 2) Two packets in the same flow should have the same hash value
883 *
884 * A hash at a higher layer is considered to be more specific. A driver should
885 * set the most specific hash possible.
886 *
887 * A driver cannot indicate a more specific hash than the layer at which a hash
888 * was computed. For instance an L3 hash cannot be set as an L4 hash.
889 *
890 * A driver may indicate a hash level which is less specific than the
891 * actual layer the hash was computed on. For instance, a hash computed
892 * at L4 may be considered an L3 hash. This should only be done if the
893 * driver can't unambiguously determine that the HW computed the hash at
894 * the higher layer. Note that the "should" in the second property above
895 * permits this.
896 */
897enum pkt_hash_types {
898 PKT_HASH_TYPE_NONE, /* Undefined type */
899 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
900 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
901 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
902};
903
904static inline void
905skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
906{
61b905da 907 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 908 skb->sw_hash = 0;
61b905da 909 skb->hash = hash;
09323cc4
TH
910}
911
3958afa1
TH
912void __skb_get_hash(struct sk_buff *skb);
913static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 914{
a3b18ddb 915 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 916 __skb_get_hash(skb);
bfb564e7 917
61b905da 918 return skb->hash;
bfb564e7
KK
919}
920
57bdf7f4
TH
921static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
922{
61b905da 923 return skb->hash;
57bdf7f4
TH
924}
925
7539fadc
TH
926static inline void skb_clear_hash(struct sk_buff *skb)
927{
61b905da 928 skb->hash = 0;
a3b18ddb 929 skb->sw_hash = 0;
61b905da 930 skb->l4_hash = 0;
7539fadc
TH
931}
932
933static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
934{
61b905da 935 if (!skb->l4_hash)
7539fadc
TH
936 skb_clear_hash(skb);
937}
938
3df7a74e
TH
939static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
940{
61b905da 941 to->hash = from->hash;
a3b18ddb 942 to->sw_hash = from->sw_hash;
61b905da 943 to->l4_hash = from->l4_hash;
3df7a74e
TH
944};
945
4305b541
ACM
946#ifdef NET_SKBUFF_DATA_USES_OFFSET
947static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
948{
949 return skb->head + skb->end;
950}
ec47ea82
AD
951
952static inline unsigned int skb_end_offset(const struct sk_buff *skb)
953{
954 return skb->end;
955}
4305b541
ACM
956#else
957static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
958{
959 return skb->end;
960}
ec47ea82
AD
961
962static inline unsigned int skb_end_offset(const struct sk_buff *skb)
963{
964 return skb->end - skb->head;
965}
4305b541
ACM
966#endif
967
1da177e4 968/* Internal */
4305b541 969#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 970
ac45f602
PO
971static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
972{
973 return &skb_shinfo(skb)->hwtstamps;
974}
975
1da177e4
LT
976/**
977 * skb_queue_empty - check if a queue is empty
978 * @list: queue head
979 *
980 * Returns true if the queue is empty, false otherwise.
981 */
982static inline int skb_queue_empty(const struct sk_buff_head *list)
983{
fd44b93c 984 return list->next == (const struct sk_buff *) list;
1da177e4
LT
985}
986
fc7ebb21
DM
987/**
988 * skb_queue_is_last - check if skb is the last entry in the queue
989 * @list: queue head
990 * @skb: buffer
991 *
992 * Returns true if @skb is the last buffer on the list.
993 */
994static inline bool skb_queue_is_last(const struct sk_buff_head *list,
995 const struct sk_buff *skb)
996{
fd44b93c 997 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
998}
999
832d11c5
IJ
1000/**
1001 * skb_queue_is_first - check if skb is the first entry in the queue
1002 * @list: queue head
1003 * @skb: buffer
1004 *
1005 * Returns true if @skb is the first buffer on the list.
1006 */
1007static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1008 const struct sk_buff *skb)
1009{
fd44b93c 1010 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1011}
1012
249c8b42
DM
1013/**
1014 * skb_queue_next - return the next packet in the queue
1015 * @list: queue head
1016 * @skb: current buffer
1017 *
1018 * Return the next packet in @list after @skb. It is only valid to
1019 * call this if skb_queue_is_last() evaluates to false.
1020 */
1021static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1022 const struct sk_buff *skb)
1023{
1024 /* This BUG_ON may seem severe, but if we just return then we
1025 * are going to dereference garbage.
1026 */
1027 BUG_ON(skb_queue_is_last(list, skb));
1028 return skb->next;
1029}
1030
832d11c5
IJ
1031/**
1032 * skb_queue_prev - return the prev packet in the queue
1033 * @list: queue head
1034 * @skb: current buffer
1035 *
1036 * Return the prev packet in @list before @skb. It is only valid to
1037 * call this if skb_queue_is_first() evaluates to false.
1038 */
1039static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1040 const struct sk_buff *skb)
1041{
1042 /* This BUG_ON may seem severe, but if we just return then we
1043 * are going to dereference garbage.
1044 */
1045 BUG_ON(skb_queue_is_first(list, skb));
1046 return skb->prev;
1047}
1048
1da177e4
LT
1049/**
1050 * skb_get - reference buffer
1051 * @skb: buffer to reference
1052 *
1053 * Makes another reference to a socket buffer and returns a pointer
1054 * to the buffer.
1055 */
1056static inline struct sk_buff *skb_get(struct sk_buff *skb)
1057{
1058 atomic_inc(&skb->users);
1059 return skb;
1060}
1061
1062/*
1063 * If users == 1, we are the only owner and are can avoid redundant
1064 * atomic change.
1065 */
1066
1da177e4
LT
1067/**
1068 * skb_cloned - is the buffer a clone
1069 * @skb: buffer to check
1070 *
1071 * Returns true if the buffer was generated with skb_clone() and is
1072 * one of multiple shared copies of the buffer. Cloned buffers are
1073 * shared data so must not be written to under normal circumstances.
1074 */
1075static inline int skb_cloned(const struct sk_buff *skb)
1076{
1077 return skb->cloned &&
1078 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1079}
1080
14bbd6a5
PS
1081static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1082{
1083 might_sleep_if(pri & __GFP_WAIT);
1084
1085 if (skb_cloned(skb))
1086 return pskb_expand_head(skb, 0, 0, pri);
1087
1088 return 0;
1089}
1090
1da177e4
LT
1091/**
1092 * skb_header_cloned - is the header a clone
1093 * @skb: buffer to check
1094 *
1095 * Returns true if modifying the header part of the buffer requires
1096 * the data to be copied.
1097 */
1098static inline int skb_header_cloned(const struct sk_buff *skb)
1099{
1100 int dataref;
1101
1102 if (!skb->cloned)
1103 return 0;
1104
1105 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1106 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1107 return dataref != 1;
1108}
1109
1110/**
1111 * skb_header_release - release reference to header
1112 * @skb: buffer to operate on
1113 *
1114 * Drop a reference to the header part of the buffer. This is done
1115 * by acquiring a payload reference. You must not read from the header
1116 * part of skb->data after this.
f4a775d1 1117 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1118 */
1119static inline void skb_header_release(struct sk_buff *skb)
1120{
1121 BUG_ON(skb->nohdr);
1122 skb->nohdr = 1;
1123 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1124}
1125
f4a775d1
ED
1126/**
1127 * __skb_header_release - release reference to header
1128 * @skb: buffer to operate on
1129 *
1130 * Variant of skb_header_release() assuming skb is private to caller.
1131 * We can avoid one atomic operation.
1132 */
1133static inline void __skb_header_release(struct sk_buff *skb)
1134{
1135 skb->nohdr = 1;
1136 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1137}
1138
1139
1da177e4
LT
1140/**
1141 * skb_shared - is the buffer shared
1142 * @skb: buffer to check
1143 *
1144 * Returns true if more than one person has a reference to this
1145 * buffer.
1146 */
1147static inline int skb_shared(const struct sk_buff *skb)
1148{
1149 return atomic_read(&skb->users) != 1;
1150}
1151
1152/**
1153 * skb_share_check - check if buffer is shared and if so clone it
1154 * @skb: buffer to check
1155 * @pri: priority for memory allocation
1156 *
1157 * If the buffer is shared the buffer is cloned and the old copy
1158 * drops a reference. A new clone with a single reference is returned.
1159 * If the buffer is not shared the original buffer is returned. When
1160 * being called from interrupt status or with spinlocks held pri must
1161 * be GFP_ATOMIC.
1162 *
1163 * NULL is returned on a memory allocation failure.
1164 */
47061bc4 1165static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1166{
1167 might_sleep_if(pri & __GFP_WAIT);
1168 if (skb_shared(skb)) {
1169 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1170
1171 if (likely(nskb))
1172 consume_skb(skb);
1173 else
1174 kfree_skb(skb);
1da177e4
LT
1175 skb = nskb;
1176 }
1177 return skb;
1178}
1179
1180/*
1181 * Copy shared buffers into a new sk_buff. We effectively do COW on
1182 * packets to handle cases where we have a local reader and forward
1183 * and a couple of other messy ones. The normal one is tcpdumping
1184 * a packet thats being forwarded.
1185 */
1186
1187/**
1188 * skb_unshare - make a copy of a shared buffer
1189 * @skb: buffer to check
1190 * @pri: priority for memory allocation
1191 *
1192 * If the socket buffer is a clone then this function creates a new
1193 * copy of the data, drops a reference count on the old copy and returns
1194 * the new copy with the reference count at 1. If the buffer is not a clone
1195 * the original buffer is returned. When called with a spinlock held or
1196 * from interrupt state @pri must be %GFP_ATOMIC
1197 *
1198 * %NULL is returned on a memory allocation failure.
1199 */
e2bf521d 1200static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1201 gfp_t pri)
1da177e4
LT
1202{
1203 might_sleep_if(pri & __GFP_WAIT);
1204 if (skb_cloned(skb)) {
1205 struct sk_buff *nskb = skb_copy(skb, pri);
1206 kfree_skb(skb); /* Free our shared copy */
1207 skb = nskb;
1208 }
1209 return skb;
1210}
1211
1212/**
1a5778aa 1213 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1214 * @list_: list to peek at
1215 *
1216 * Peek an &sk_buff. Unlike most other operations you _MUST_
1217 * be careful with this one. A peek leaves the buffer on the
1218 * list and someone else may run off with it. You must hold
1219 * the appropriate locks or have a private queue to do this.
1220 *
1221 * Returns %NULL for an empty list or a pointer to the head element.
1222 * The reference count is not incremented and the reference is therefore
1223 * volatile. Use with caution.
1224 */
05bdd2f1 1225static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1226{
18d07000
ED
1227 struct sk_buff *skb = list_->next;
1228
1229 if (skb == (struct sk_buff *)list_)
1230 skb = NULL;
1231 return skb;
1da177e4
LT
1232}
1233
da5ef6e5
PE
1234/**
1235 * skb_peek_next - peek skb following the given one from a queue
1236 * @skb: skb to start from
1237 * @list_: list to peek at
1238 *
1239 * Returns %NULL when the end of the list is met or a pointer to the
1240 * next element. The reference count is not incremented and the
1241 * reference is therefore volatile. Use with caution.
1242 */
1243static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1244 const struct sk_buff_head *list_)
1245{
1246 struct sk_buff *next = skb->next;
18d07000 1247
da5ef6e5
PE
1248 if (next == (struct sk_buff *)list_)
1249 next = NULL;
1250 return next;
1251}
1252
1da177e4 1253/**
1a5778aa 1254 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1255 * @list_: list to peek at
1256 *
1257 * Peek an &sk_buff. Unlike most other operations you _MUST_
1258 * be careful with this one. A peek leaves the buffer on the
1259 * list and someone else may run off with it. You must hold
1260 * the appropriate locks or have a private queue to do this.
1261 *
1262 * Returns %NULL for an empty list or a pointer to the tail element.
1263 * The reference count is not incremented and the reference is therefore
1264 * volatile. Use with caution.
1265 */
05bdd2f1 1266static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1267{
18d07000
ED
1268 struct sk_buff *skb = list_->prev;
1269
1270 if (skb == (struct sk_buff *)list_)
1271 skb = NULL;
1272 return skb;
1273
1da177e4
LT
1274}
1275
1276/**
1277 * skb_queue_len - get queue length
1278 * @list_: list to measure
1279 *
1280 * Return the length of an &sk_buff queue.
1281 */
1282static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1283{
1284 return list_->qlen;
1285}
1286
67fed459
DM
1287/**
1288 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1289 * @list: queue to initialize
1290 *
1291 * This initializes only the list and queue length aspects of
1292 * an sk_buff_head object. This allows to initialize the list
1293 * aspects of an sk_buff_head without reinitializing things like
1294 * the spinlock. It can also be used for on-stack sk_buff_head
1295 * objects where the spinlock is known to not be used.
1296 */
1297static inline void __skb_queue_head_init(struct sk_buff_head *list)
1298{
1299 list->prev = list->next = (struct sk_buff *)list;
1300 list->qlen = 0;
1301}
1302
76f10ad0
AV
1303/*
1304 * This function creates a split out lock class for each invocation;
1305 * this is needed for now since a whole lot of users of the skb-queue
1306 * infrastructure in drivers have different locking usage (in hardirq)
1307 * than the networking core (in softirq only). In the long run either the
1308 * network layer or drivers should need annotation to consolidate the
1309 * main types of usage into 3 classes.
1310 */
1da177e4
LT
1311static inline void skb_queue_head_init(struct sk_buff_head *list)
1312{
1313 spin_lock_init(&list->lock);
67fed459 1314 __skb_queue_head_init(list);
1da177e4
LT
1315}
1316
c2ecba71
PE
1317static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1318 struct lock_class_key *class)
1319{
1320 skb_queue_head_init(list);
1321 lockdep_set_class(&list->lock, class);
1322}
1323
1da177e4 1324/*
bf299275 1325 * Insert an sk_buff on a list.
1da177e4
LT
1326 *
1327 * The "__skb_xxxx()" functions are the non-atomic ones that
1328 * can only be called with interrupts disabled.
1329 */
7965bd4d
JP
1330void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1331 struct sk_buff_head *list);
bf299275
GR
1332static inline void __skb_insert(struct sk_buff *newsk,
1333 struct sk_buff *prev, struct sk_buff *next,
1334 struct sk_buff_head *list)
1335{
1336 newsk->next = next;
1337 newsk->prev = prev;
1338 next->prev = prev->next = newsk;
1339 list->qlen++;
1340}
1da177e4 1341
67fed459
DM
1342static inline void __skb_queue_splice(const struct sk_buff_head *list,
1343 struct sk_buff *prev,
1344 struct sk_buff *next)
1345{
1346 struct sk_buff *first = list->next;
1347 struct sk_buff *last = list->prev;
1348
1349 first->prev = prev;
1350 prev->next = first;
1351
1352 last->next = next;
1353 next->prev = last;
1354}
1355
1356/**
1357 * skb_queue_splice - join two skb lists, this is designed for stacks
1358 * @list: the new list to add
1359 * @head: the place to add it in the first list
1360 */
1361static inline void skb_queue_splice(const struct sk_buff_head *list,
1362 struct sk_buff_head *head)
1363{
1364 if (!skb_queue_empty(list)) {
1365 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1366 head->qlen += list->qlen;
67fed459
DM
1367 }
1368}
1369
1370/**
d9619496 1371 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1372 * @list: the new list to add
1373 * @head: the place to add it in the first list
1374 *
1375 * The list at @list is reinitialised
1376 */
1377static inline void skb_queue_splice_init(struct sk_buff_head *list,
1378 struct sk_buff_head *head)
1379{
1380 if (!skb_queue_empty(list)) {
1381 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1382 head->qlen += list->qlen;
67fed459
DM
1383 __skb_queue_head_init(list);
1384 }
1385}
1386
1387/**
1388 * skb_queue_splice_tail - join two skb lists, each list being a queue
1389 * @list: the new list to add
1390 * @head: the place to add it in the first list
1391 */
1392static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1393 struct sk_buff_head *head)
1394{
1395 if (!skb_queue_empty(list)) {
1396 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1397 head->qlen += list->qlen;
67fed459
DM
1398 }
1399}
1400
1401/**
d9619496 1402 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1403 * @list: the new list to add
1404 * @head: the place to add it in the first list
1405 *
1406 * Each of the lists is a queue.
1407 * The list at @list is reinitialised
1408 */
1409static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1410 struct sk_buff_head *head)
1411{
1412 if (!skb_queue_empty(list)) {
1413 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1414 head->qlen += list->qlen;
67fed459
DM
1415 __skb_queue_head_init(list);
1416 }
1417}
1418
1da177e4 1419/**
300ce174 1420 * __skb_queue_after - queue a buffer at the list head
1da177e4 1421 * @list: list to use
300ce174 1422 * @prev: place after this buffer
1da177e4
LT
1423 * @newsk: buffer to queue
1424 *
300ce174 1425 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1426 * and you must therefore hold required locks before calling it.
1427 *
1428 * A buffer cannot be placed on two lists at the same time.
1429 */
300ce174
SH
1430static inline void __skb_queue_after(struct sk_buff_head *list,
1431 struct sk_buff *prev,
1432 struct sk_buff *newsk)
1da177e4 1433{
bf299275 1434 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1435}
1436
7965bd4d
JP
1437void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1438 struct sk_buff_head *list);
7de6c033 1439
f5572855
GR
1440static inline void __skb_queue_before(struct sk_buff_head *list,
1441 struct sk_buff *next,
1442 struct sk_buff *newsk)
1443{
1444 __skb_insert(newsk, next->prev, next, list);
1445}
1446
300ce174
SH
1447/**
1448 * __skb_queue_head - queue a buffer at the list head
1449 * @list: list to use
1450 * @newsk: buffer to queue
1451 *
1452 * Queue a buffer at the start of a list. This function takes no locks
1453 * and you must therefore hold required locks before calling it.
1454 *
1455 * A buffer cannot be placed on two lists at the same time.
1456 */
7965bd4d 1457void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1458static inline void __skb_queue_head(struct sk_buff_head *list,
1459 struct sk_buff *newsk)
1460{
1461 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1462}
1463
1da177e4
LT
1464/**
1465 * __skb_queue_tail - queue a buffer at the list tail
1466 * @list: list to use
1467 * @newsk: buffer to queue
1468 *
1469 * Queue a buffer at the end of a list. This function takes no locks
1470 * and you must therefore hold required locks before calling it.
1471 *
1472 * A buffer cannot be placed on two lists at the same time.
1473 */
7965bd4d 1474void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1475static inline void __skb_queue_tail(struct sk_buff_head *list,
1476 struct sk_buff *newsk)
1477{
f5572855 1478 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1479}
1480
1da177e4
LT
1481/*
1482 * remove sk_buff from list. _Must_ be called atomically, and with
1483 * the list known..
1484 */
7965bd4d 1485void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1486static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1487{
1488 struct sk_buff *next, *prev;
1489
1490 list->qlen--;
1491 next = skb->next;
1492 prev = skb->prev;
1493 skb->next = skb->prev = NULL;
1da177e4
LT
1494 next->prev = prev;
1495 prev->next = next;
1496}
1497
f525c06d
GR
1498/**
1499 * __skb_dequeue - remove from the head of the queue
1500 * @list: list to dequeue from
1501 *
1502 * Remove the head of the list. This function does not take any locks
1503 * so must be used with appropriate locks held only. The head item is
1504 * returned or %NULL if the list is empty.
1505 */
7965bd4d 1506struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1507static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1508{
1509 struct sk_buff *skb = skb_peek(list);
1510 if (skb)
1511 __skb_unlink(skb, list);
1512 return skb;
1513}
1da177e4
LT
1514
1515/**
1516 * __skb_dequeue_tail - remove from the tail of the queue
1517 * @list: list to dequeue from
1518 *
1519 * Remove the tail of the list. This function does not take any locks
1520 * so must be used with appropriate locks held only. The tail item is
1521 * returned or %NULL if the list is empty.
1522 */
7965bd4d 1523struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1524static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1525{
1526 struct sk_buff *skb = skb_peek_tail(list);
1527 if (skb)
1528 __skb_unlink(skb, list);
1529 return skb;
1530}
1531
1532
bdcc0924 1533static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1534{
1535 return skb->data_len;
1536}
1537
1538static inline unsigned int skb_headlen(const struct sk_buff *skb)
1539{
1540 return skb->len - skb->data_len;
1541}
1542
1543static inline int skb_pagelen(const struct sk_buff *skb)
1544{
1545 int i, len = 0;
1546
1547 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1548 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1549 return len + skb_headlen(skb);
1550}
1551
131ea667
IC
1552/**
1553 * __skb_fill_page_desc - initialise a paged fragment in an skb
1554 * @skb: buffer containing fragment to be initialised
1555 * @i: paged fragment index to initialise
1556 * @page: the page to use for this fragment
1557 * @off: the offset to the data with @page
1558 * @size: the length of the data
1559 *
1560 * Initialises the @i'th fragment of @skb to point to &size bytes at
1561 * offset @off within @page.
1562 *
1563 * Does not take any additional reference on the fragment.
1564 */
1565static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1566 struct page *page, int off, int size)
1da177e4
LT
1567{
1568 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1569
c48a11c7
MG
1570 /*
1571 * Propagate page->pfmemalloc to the skb if we can. The problem is
1572 * that not all callers have unique ownership of the page. If
1573 * pfmemalloc is set, we check the mapping as a mapping implies
1574 * page->index is set (index and pfmemalloc share space).
1575 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1576 * do not lose pfmemalloc information as the pages would not be
1577 * allocated using __GFP_MEMALLOC.
1578 */
a8605c60 1579 frag->page.p = page;
1da177e4 1580 frag->page_offset = off;
9e903e08 1581 skb_frag_size_set(frag, size);
cca7af38
PE
1582
1583 page = compound_head(page);
1584 if (page->pfmemalloc && !page->mapping)
1585 skb->pfmemalloc = true;
131ea667
IC
1586}
1587
1588/**
1589 * skb_fill_page_desc - initialise a paged fragment in an skb
1590 * @skb: buffer containing fragment to be initialised
1591 * @i: paged fragment index to initialise
1592 * @page: the page to use for this fragment
1593 * @off: the offset to the data with @page
1594 * @size: the length of the data
1595 *
1596 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1597 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1598 * addition updates @skb such that @i is the last fragment.
1599 *
1600 * Does not take any additional reference on the fragment.
1601 */
1602static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1603 struct page *page, int off, int size)
1604{
1605 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1606 skb_shinfo(skb)->nr_frags = i + 1;
1607}
1608
7965bd4d
JP
1609void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1610 int size, unsigned int truesize);
654bed16 1611
f8e617e1
JW
1612void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1613 unsigned int truesize);
1614
1da177e4 1615#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1616#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1617#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1618
27a884dc
ACM
1619#ifdef NET_SKBUFF_DATA_USES_OFFSET
1620static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1621{
1622 return skb->head + skb->tail;
1623}
1624
1625static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1626{
1627 skb->tail = skb->data - skb->head;
1628}
1629
1630static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1631{
1632 skb_reset_tail_pointer(skb);
1633 skb->tail += offset;
1634}
7cc46190 1635
27a884dc
ACM
1636#else /* NET_SKBUFF_DATA_USES_OFFSET */
1637static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1638{
1639 return skb->tail;
1640}
1641
1642static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1643{
1644 skb->tail = skb->data;
1645}
1646
1647static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1648{
1649 skb->tail = skb->data + offset;
1650}
4305b541 1651
27a884dc
ACM
1652#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1653
1da177e4
LT
1654/*
1655 * Add data to an sk_buff
1656 */
0c7ddf36 1657unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1658unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1659static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1660{
27a884dc 1661 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1662 SKB_LINEAR_ASSERT(skb);
1663 skb->tail += len;
1664 skb->len += len;
1665 return tmp;
1666}
1667
7965bd4d 1668unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1669static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1670{
1671 skb->data -= len;
1672 skb->len += len;
1673 return skb->data;
1674}
1675
7965bd4d 1676unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1677static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1678{
1679 skb->len -= len;
1680 BUG_ON(skb->len < skb->data_len);
1681 return skb->data += len;
1682}
1683
47d29646
DM
1684static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1685{
1686 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1687}
1688
7965bd4d 1689unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1690
1691static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1692{
1693 if (len > skb_headlen(skb) &&
987c402a 1694 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1695 return NULL;
1696 skb->len -= len;
1697 return skb->data += len;
1698}
1699
1700static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1701{
1702 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1703}
1704
1705static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1706{
1707 if (likely(len <= skb_headlen(skb)))
1708 return 1;
1709 if (unlikely(len > skb->len))
1710 return 0;
987c402a 1711 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1712}
1713
1714/**
1715 * skb_headroom - bytes at buffer head
1716 * @skb: buffer to check
1717 *
1718 * Return the number of bytes of free space at the head of an &sk_buff.
1719 */
c2636b4d 1720static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1721{
1722 return skb->data - skb->head;
1723}
1724
1725/**
1726 * skb_tailroom - bytes at buffer end
1727 * @skb: buffer to check
1728 *
1729 * Return the number of bytes of free space at the tail of an sk_buff
1730 */
1731static inline int skb_tailroom(const struct sk_buff *skb)
1732{
4305b541 1733 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1734}
1735
a21d4572
ED
1736/**
1737 * skb_availroom - bytes at buffer end
1738 * @skb: buffer to check
1739 *
1740 * Return the number of bytes of free space at the tail of an sk_buff
1741 * allocated by sk_stream_alloc()
1742 */
1743static inline int skb_availroom(const struct sk_buff *skb)
1744{
16fad69c
ED
1745 if (skb_is_nonlinear(skb))
1746 return 0;
1747
1748 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1749}
1750
1da177e4
LT
1751/**
1752 * skb_reserve - adjust headroom
1753 * @skb: buffer to alter
1754 * @len: bytes to move
1755 *
1756 * Increase the headroom of an empty &sk_buff by reducing the tail
1757 * room. This is only allowed for an empty buffer.
1758 */
8243126c 1759static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1760{
1761 skb->data += len;
1762 skb->tail += len;
1763}
1764
8bce6d7d
TH
1765#define ENCAP_TYPE_ETHER 0
1766#define ENCAP_TYPE_IPPROTO 1
1767
1768static inline void skb_set_inner_protocol(struct sk_buff *skb,
1769 __be16 protocol)
1770{
1771 skb->inner_protocol = protocol;
1772 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1773}
1774
1775static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1776 __u8 ipproto)
1777{
1778 skb->inner_ipproto = ipproto;
1779 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1780}
1781
6a674e9c
JG
1782static inline void skb_reset_inner_headers(struct sk_buff *skb)
1783{
aefbd2b3 1784 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1785 skb->inner_network_header = skb->network_header;
1786 skb->inner_transport_header = skb->transport_header;
1787}
1788
0b5c9db1
JP
1789static inline void skb_reset_mac_len(struct sk_buff *skb)
1790{
1791 skb->mac_len = skb->network_header - skb->mac_header;
1792}
1793
6a674e9c
JG
1794static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1795 *skb)
1796{
1797 return skb->head + skb->inner_transport_header;
1798}
1799
1800static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1801{
1802 skb->inner_transport_header = skb->data - skb->head;
1803}
1804
1805static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1806 const int offset)
1807{
1808 skb_reset_inner_transport_header(skb);
1809 skb->inner_transport_header += offset;
1810}
1811
1812static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1813{
1814 return skb->head + skb->inner_network_header;
1815}
1816
1817static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1818{
1819 skb->inner_network_header = skb->data - skb->head;
1820}
1821
1822static inline void skb_set_inner_network_header(struct sk_buff *skb,
1823 const int offset)
1824{
1825 skb_reset_inner_network_header(skb);
1826 skb->inner_network_header += offset;
1827}
1828
aefbd2b3
PS
1829static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1830{
1831 return skb->head + skb->inner_mac_header;
1832}
1833
1834static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1835{
1836 skb->inner_mac_header = skb->data - skb->head;
1837}
1838
1839static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1840 const int offset)
1841{
1842 skb_reset_inner_mac_header(skb);
1843 skb->inner_mac_header += offset;
1844}
fda55eca
ED
1845static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1846{
35d04610 1847 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1848}
1849
9c70220b
ACM
1850static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1851{
2e07fa9c 1852 return skb->head + skb->transport_header;
9c70220b
ACM
1853}
1854
badff6d0
ACM
1855static inline void skb_reset_transport_header(struct sk_buff *skb)
1856{
2e07fa9c 1857 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1858}
1859
967b05f6
ACM
1860static inline void skb_set_transport_header(struct sk_buff *skb,
1861 const int offset)
1862{
2e07fa9c
ACM
1863 skb_reset_transport_header(skb);
1864 skb->transport_header += offset;
ea2ae17d
ACM
1865}
1866
d56f90a7
ACM
1867static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1868{
2e07fa9c 1869 return skb->head + skb->network_header;
d56f90a7
ACM
1870}
1871
c1d2bbe1
ACM
1872static inline void skb_reset_network_header(struct sk_buff *skb)
1873{
2e07fa9c 1874 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1875}
1876
c14d2450
ACM
1877static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1878{
2e07fa9c
ACM
1879 skb_reset_network_header(skb);
1880 skb->network_header += offset;
c14d2450
ACM
1881}
1882
2e07fa9c 1883static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1884{
2e07fa9c 1885 return skb->head + skb->mac_header;
bbe735e4
ACM
1886}
1887
2e07fa9c 1888static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1889{
35d04610 1890 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1891}
1892
1893static inline void skb_reset_mac_header(struct sk_buff *skb)
1894{
1895 skb->mac_header = skb->data - skb->head;
1896}
1897
1898static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1899{
1900 skb_reset_mac_header(skb);
1901 skb->mac_header += offset;
1902}
1903
0e3da5bb
TT
1904static inline void skb_pop_mac_header(struct sk_buff *skb)
1905{
1906 skb->mac_header = skb->network_header;
1907}
1908
fbbdb8f0
YX
1909static inline void skb_probe_transport_header(struct sk_buff *skb,
1910 const int offset_hint)
1911{
1912 struct flow_keys keys;
1913
1914 if (skb_transport_header_was_set(skb))
1915 return;
1916 else if (skb_flow_dissect(skb, &keys))
1917 skb_set_transport_header(skb, keys.thoff);
1918 else
1919 skb_set_transport_header(skb, offset_hint);
1920}
1921
03606895
ED
1922static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1923{
1924 if (skb_mac_header_was_set(skb)) {
1925 const unsigned char *old_mac = skb_mac_header(skb);
1926
1927 skb_set_mac_header(skb, -skb->mac_len);
1928 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1929 }
1930}
1931
04fb451e
MM
1932static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1933{
1934 return skb->csum_start - skb_headroom(skb);
1935}
1936
2e07fa9c
ACM
1937static inline int skb_transport_offset(const struct sk_buff *skb)
1938{
1939 return skb_transport_header(skb) - skb->data;
1940}
1941
1942static inline u32 skb_network_header_len(const struct sk_buff *skb)
1943{
1944 return skb->transport_header - skb->network_header;
1945}
1946
6a674e9c
JG
1947static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1948{
1949 return skb->inner_transport_header - skb->inner_network_header;
1950}
1951
2e07fa9c
ACM
1952static inline int skb_network_offset(const struct sk_buff *skb)
1953{
1954 return skb_network_header(skb) - skb->data;
1955}
48d49d0c 1956
6a674e9c
JG
1957static inline int skb_inner_network_offset(const struct sk_buff *skb)
1958{
1959 return skb_inner_network_header(skb) - skb->data;
1960}
1961
f9599ce1
CG
1962static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1963{
1964 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1965}
1966
1da177e4
LT
1967/*
1968 * CPUs often take a performance hit when accessing unaligned memory
1969 * locations. The actual performance hit varies, it can be small if the
1970 * hardware handles it or large if we have to take an exception and fix it
1971 * in software.
1972 *
1973 * Since an ethernet header is 14 bytes network drivers often end up with
1974 * the IP header at an unaligned offset. The IP header can be aligned by
1975 * shifting the start of the packet by 2 bytes. Drivers should do this
1976 * with:
1977 *
8660c124 1978 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1979 *
1980 * The downside to this alignment of the IP header is that the DMA is now
1981 * unaligned. On some architectures the cost of an unaligned DMA is high
1982 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1983 *
1da177e4
LT
1984 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1985 * to be overridden.
1986 */
1987#ifndef NET_IP_ALIGN
1988#define NET_IP_ALIGN 2
1989#endif
1990
025be81e
AB
1991/*
1992 * The networking layer reserves some headroom in skb data (via
1993 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1994 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1995 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1996 *
1997 * Unfortunately this headroom changes the DMA alignment of the resulting
1998 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1999 * on some architectures. An architecture can override this value,
2000 * perhaps setting it to a cacheline in size (since that will maintain
2001 * cacheline alignment of the DMA). It must be a power of 2.
2002 *
d6301d3d 2003 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2004 * headroom, you should not reduce this.
5933dd2f
ED
2005 *
2006 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2007 * to reduce average number of cache lines per packet.
2008 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2009 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2010 */
2011#ifndef NET_SKB_PAD
5933dd2f 2012#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2013#endif
2014
7965bd4d 2015int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2016
2017static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2018{
c4264f27 2019 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2020 WARN_ON(1);
2021 return;
2022 }
27a884dc
ACM
2023 skb->len = len;
2024 skb_set_tail_pointer(skb, len);
1da177e4
LT
2025}
2026
7965bd4d 2027void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2028
2029static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2030{
3cc0e873
HX
2031 if (skb->data_len)
2032 return ___pskb_trim(skb, len);
2033 __skb_trim(skb, len);
2034 return 0;
1da177e4
LT
2035}
2036
2037static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2038{
2039 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2040}
2041
e9fa4f7b
HX
2042/**
2043 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2044 * @skb: buffer to alter
2045 * @len: new length
2046 *
2047 * This is identical to pskb_trim except that the caller knows that
2048 * the skb is not cloned so we should never get an error due to out-
2049 * of-memory.
2050 */
2051static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2052{
2053 int err = pskb_trim(skb, len);
2054 BUG_ON(err);
2055}
2056
1da177e4
LT
2057/**
2058 * skb_orphan - orphan a buffer
2059 * @skb: buffer to orphan
2060 *
2061 * If a buffer currently has an owner then we call the owner's
2062 * destructor function and make the @skb unowned. The buffer continues
2063 * to exist but is no longer charged to its former owner.
2064 */
2065static inline void skb_orphan(struct sk_buff *skb)
2066{
c34a7612 2067 if (skb->destructor) {
1da177e4 2068 skb->destructor(skb);
c34a7612
ED
2069 skb->destructor = NULL;
2070 skb->sk = NULL;
376c7311
ED
2071 } else {
2072 BUG_ON(skb->sk);
c34a7612 2073 }
1da177e4
LT
2074}
2075
a353e0ce
MT
2076/**
2077 * skb_orphan_frags - orphan the frags contained in a buffer
2078 * @skb: buffer to orphan frags from
2079 * @gfp_mask: allocation mask for replacement pages
2080 *
2081 * For each frag in the SKB which needs a destructor (i.e. has an
2082 * owner) create a copy of that frag and release the original
2083 * page by calling the destructor.
2084 */
2085static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2086{
2087 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2088 return 0;
2089 return skb_copy_ubufs(skb, gfp_mask);
2090}
2091
1da177e4
LT
2092/**
2093 * __skb_queue_purge - empty a list
2094 * @list: list to empty
2095 *
2096 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2097 * the list and one reference dropped. This function does not take the
2098 * list lock and the caller must hold the relevant locks to use it.
2099 */
7965bd4d 2100void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2101static inline void __skb_queue_purge(struct sk_buff_head *list)
2102{
2103 struct sk_buff *skb;
2104 while ((skb = __skb_dequeue(list)) != NULL)
2105 kfree_skb(skb);
2106}
2107
e5e67305
AD
2108#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2109#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2110#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2111
7965bd4d 2112void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2113
7965bd4d
JP
2114struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2115 gfp_t gfp_mask);
8af27456
CH
2116
2117/**
2118 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2119 * @dev: network device to receive on
2120 * @length: length to allocate
2121 *
2122 * Allocate a new &sk_buff and assign it a usage count of one. The
2123 * buffer has unspecified headroom built in. Users should allocate
2124 * the headroom they think they need without accounting for the
2125 * built in space. The built in space is used for optimisations.
2126 *
2127 * %NULL is returned if there is no free memory. Although this function
2128 * allocates memory it can be called from an interrupt.
2129 */
2130static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2131 unsigned int length)
8af27456
CH
2132{
2133 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2134}
2135
6f532612
ED
2136/* legacy helper around __netdev_alloc_skb() */
2137static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2138 gfp_t gfp_mask)
2139{
2140 return __netdev_alloc_skb(NULL, length, gfp_mask);
2141}
2142
2143/* legacy helper around netdev_alloc_skb() */
2144static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2145{
2146 return netdev_alloc_skb(NULL, length);
2147}
2148
2149
4915a0de
ED
2150static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2151 unsigned int length, gfp_t gfp)
61321bbd 2152{
4915a0de 2153 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2154
2155 if (NET_IP_ALIGN && skb)
2156 skb_reserve(skb, NET_IP_ALIGN);
2157 return skb;
2158}
2159
4915a0de
ED
2160static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2161 unsigned int length)
2162{
2163 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2164}
2165
bc6fc9fa
FF
2166/**
2167 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2168 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2169 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2170 * @order: size of the allocation
2171 *
2172 * Allocate a new page.
2173 *
2174 * %NULL is returned if there is no free memory.
2175*/
2176static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2177 struct sk_buff *skb,
2178 unsigned int order)
2179{
2180 struct page *page;
2181
2182 gfp_mask |= __GFP_COLD;
2183
2184 if (!(gfp_mask & __GFP_NOMEMALLOC))
2185 gfp_mask |= __GFP_MEMALLOC;
2186
2187 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2188 if (skb && page && page->pfmemalloc)
2189 skb->pfmemalloc = true;
2190
2191 return page;
2192}
2193
2194/**
2195 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2196 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2197 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2198 *
2199 * Allocate a new page.
2200 *
2201 * %NULL is returned if there is no free memory.
2202 */
2203static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2204 struct sk_buff *skb)
2205{
2206 return __skb_alloc_pages(gfp_mask, skb, 0);
2207}
2208
2209/**
2210 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2211 * @page: The page that was allocated from skb_alloc_page
2212 * @skb: The skb that may need pfmemalloc set
2213 */
2214static inline void skb_propagate_pfmemalloc(struct page *page,
2215 struct sk_buff *skb)
2216{
2217 if (page && page->pfmemalloc)
2218 skb->pfmemalloc = true;
2219}
2220
131ea667 2221/**
e227867f 2222 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2223 * @frag: the paged fragment
2224 *
2225 * Returns the &struct page associated with @frag.
2226 */
2227static inline struct page *skb_frag_page(const skb_frag_t *frag)
2228{
a8605c60 2229 return frag->page.p;
131ea667
IC
2230}
2231
2232/**
2233 * __skb_frag_ref - take an addition reference on a paged fragment.
2234 * @frag: the paged fragment
2235 *
2236 * Takes an additional reference on the paged fragment @frag.
2237 */
2238static inline void __skb_frag_ref(skb_frag_t *frag)
2239{
2240 get_page(skb_frag_page(frag));
2241}
2242
2243/**
2244 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2245 * @skb: the buffer
2246 * @f: the fragment offset.
2247 *
2248 * Takes an additional reference on the @f'th paged fragment of @skb.
2249 */
2250static inline void skb_frag_ref(struct sk_buff *skb, int f)
2251{
2252 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2253}
2254
2255/**
2256 * __skb_frag_unref - release a reference on a paged fragment.
2257 * @frag: the paged fragment
2258 *
2259 * Releases a reference on the paged fragment @frag.
2260 */
2261static inline void __skb_frag_unref(skb_frag_t *frag)
2262{
2263 put_page(skb_frag_page(frag));
2264}
2265
2266/**
2267 * skb_frag_unref - release a reference on a paged fragment of an skb.
2268 * @skb: the buffer
2269 * @f: the fragment offset
2270 *
2271 * Releases a reference on the @f'th paged fragment of @skb.
2272 */
2273static inline void skb_frag_unref(struct sk_buff *skb, int f)
2274{
2275 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2276}
2277
2278/**
2279 * skb_frag_address - gets the address of the data contained in a paged fragment
2280 * @frag: the paged fragment buffer
2281 *
2282 * Returns the address of the data within @frag. The page must already
2283 * be mapped.
2284 */
2285static inline void *skb_frag_address(const skb_frag_t *frag)
2286{
2287 return page_address(skb_frag_page(frag)) + frag->page_offset;
2288}
2289
2290/**
2291 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2292 * @frag: the paged fragment buffer
2293 *
2294 * Returns the address of the data within @frag. Checks that the page
2295 * is mapped and returns %NULL otherwise.
2296 */
2297static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2298{
2299 void *ptr = page_address(skb_frag_page(frag));
2300 if (unlikely(!ptr))
2301 return NULL;
2302
2303 return ptr + frag->page_offset;
2304}
2305
2306/**
2307 * __skb_frag_set_page - sets the page contained in a paged fragment
2308 * @frag: the paged fragment
2309 * @page: the page to set
2310 *
2311 * Sets the fragment @frag to contain @page.
2312 */
2313static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2314{
a8605c60 2315 frag->page.p = page;
131ea667
IC
2316}
2317
2318/**
2319 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2320 * @skb: the buffer
2321 * @f: the fragment offset
2322 * @page: the page to set
2323 *
2324 * Sets the @f'th fragment of @skb to contain @page.
2325 */
2326static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2327 struct page *page)
2328{
2329 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2330}
2331
400dfd3a
ED
2332bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2333
131ea667
IC
2334/**
2335 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2336 * @dev: the device to map the fragment to
131ea667
IC
2337 * @frag: the paged fragment to map
2338 * @offset: the offset within the fragment (starting at the
2339 * fragment's own offset)
2340 * @size: the number of bytes to map
f83347df 2341 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2342 *
2343 * Maps the page associated with @frag to @device.
2344 */
2345static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2346 const skb_frag_t *frag,
2347 size_t offset, size_t size,
2348 enum dma_data_direction dir)
2349{
2350 return dma_map_page(dev, skb_frag_page(frag),
2351 frag->page_offset + offset, size, dir);
2352}
2353
117632e6
ED
2354static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2355 gfp_t gfp_mask)
2356{
2357 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2358}
2359
bad93e9d
OP
2360
2361static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2362 gfp_t gfp_mask)
2363{
2364 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2365}
2366
2367
334a8132
PM
2368/**
2369 * skb_clone_writable - is the header of a clone writable
2370 * @skb: buffer to check
2371 * @len: length up to which to write
2372 *
2373 * Returns true if modifying the header part of the cloned buffer
2374 * does not requires the data to be copied.
2375 */
05bdd2f1 2376static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2377{
2378 return !skb_header_cloned(skb) &&
2379 skb_headroom(skb) + len <= skb->hdr_len;
2380}
2381
d9cc2048
HX
2382static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2383 int cloned)
2384{
2385 int delta = 0;
2386
d9cc2048
HX
2387 if (headroom > skb_headroom(skb))
2388 delta = headroom - skb_headroom(skb);
2389
2390 if (delta || cloned)
2391 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2392 GFP_ATOMIC);
2393 return 0;
2394}
2395
1da177e4
LT
2396/**
2397 * skb_cow - copy header of skb when it is required
2398 * @skb: buffer to cow
2399 * @headroom: needed headroom
2400 *
2401 * If the skb passed lacks sufficient headroom or its data part
2402 * is shared, data is reallocated. If reallocation fails, an error
2403 * is returned and original skb is not changed.
2404 *
2405 * The result is skb with writable area skb->head...skb->tail
2406 * and at least @headroom of space at head.
2407 */
2408static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2409{
d9cc2048
HX
2410 return __skb_cow(skb, headroom, skb_cloned(skb));
2411}
1da177e4 2412
d9cc2048
HX
2413/**
2414 * skb_cow_head - skb_cow but only making the head writable
2415 * @skb: buffer to cow
2416 * @headroom: needed headroom
2417 *
2418 * This function is identical to skb_cow except that we replace the
2419 * skb_cloned check by skb_header_cloned. It should be used when
2420 * you only need to push on some header and do not need to modify
2421 * the data.
2422 */
2423static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2424{
2425 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2426}
2427
2428/**
2429 * skb_padto - pad an skbuff up to a minimal size
2430 * @skb: buffer to pad
2431 * @len: minimal length
2432 *
2433 * Pads up a buffer to ensure the trailing bytes exist and are
2434 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2435 * is untouched. Otherwise it is extended. Returns zero on
2436 * success. The skb is freed on error.
1da177e4
LT
2437 */
2438
5b057c6b 2439static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2440{
2441 unsigned int size = skb->len;
2442 if (likely(size >= len))
5b057c6b 2443 return 0;
987c402a 2444 return skb_pad(skb, len - size);
1da177e4
LT
2445}
2446
2447static inline int skb_add_data(struct sk_buff *skb,
2448 char __user *from, int copy)
2449{
2450 const int off = skb->len;
2451
2452 if (skb->ip_summed == CHECKSUM_NONE) {
2453 int err = 0;
5084205f 2454 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2455 copy, 0, &err);
2456 if (!err) {
2457 skb->csum = csum_block_add(skb->csum, csum, off);
2458 return 0;
2459 }
2460 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2461 return 0;
2462
2463 __skb_trim(skb, off);
2464 return -EFAULT;
2465}
2466
38ba0a65
ED
2467static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2468 const struct page *page, int off)
1da177e4
LT
2469{
2470 if (i) {
9e903e08 2471 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2472
ea2ab693 2473 return page == skb_frag_page(frag) &&
9e903e08 2474 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2475 }
38ba0a65 2476 return false;
1da177e4
LT
2477}
2478
364c6bad
HX
2479static inline int __skb_linearize(struct sk_buff *skb)
2480{
2481 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2482}
2483
1da177e4
LT
2484/**
2485 * skb_linearize - convert paged skb to linear one
2486 * @skb: buffer to linarize
1da177e4
LT
2487 *
2488 * If there is no free memory -ENOMEM is returned, otherwise zero
2489 * is returned and the old skb data released.
2490 */
364c6bad
HX
2491static inline int skb_linearize(struct sk_buff *skb)
2492{
2493 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2494}
2495
cef401de
ED
2496/**
2497 * skb_has_shared_frag - can any frag be overwritten
2498 * @skb: buffer to test
2499 *
2500 * Return true if the skb has at least one frag that might be modified
2501 * by an external entity (as in vmsplice()/sendfile())
2502 */
2503static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2504{
c9af6db4
PS
2505 return skb_is_nonlinear(skb) &&
2506 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2507}
2508
364c6bad
HX
2509/**
2510 * skb_linearize_cow - make sure skb is linear and writable
2511 * @skb: buffer to process
2512 *
2513 * If there is no free memory -ENOMEM is returned, otherwise zero
2514 * is returned and the old skb data released.
2515 */
2516static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2517{
364c6bad
HX
2518 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2519 __skb_linearize(skb) : 0;
1da177e4
LT
2520}
2521
2522/**
2523 * skb_postpull_rcsum - update checksum for received skb after pull
2524 * @skb: buffer to update
2525 * @start: start of data before pull
2526 * @len: length of data pulled
2527 *
2528 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2529 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2530 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2531 */
2532
2533static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2534 const void *start, unsigned int len)
1da177e4 2535{
84fa7933 2536 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2537 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2538}
2539
cbb042f9
HX
2540unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2541
7ce5a27f
DM
2542/**
2543 * pskb_trim_rcsum - trim received skb and update checksum
2544 * @skb: buffer to trim
2545 * @len: new length
2546 *
2547 * This is exactly the same as pskb_trim except that it ensures the
2548 * checksum of received packets are still valid after the operation.
2549 */
2550
2551static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2552{
2553 if (likely(len >= skb->len))
2554 return 0;
2555 if (skb->ip_summed == CHECKSUM_COMPLETE)
2556 skb->ip_summed = CHECKSUM_NONE;
2557 return __pskb_trim(skb, len);
2558}
2559
1da177e4
LT
2560#define skb_queue_walk(queue, skb) \
2561 for (skb = (queue)->next; \
a1e4891f 2562 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2563 skb = skb->next)
2564
46f8914e
JC
2565#define skb_queue_walk_safe(queue, skb, tmp) \
2566 for (skb = (queue)->next, tmp = skb->next; \
2567 skb != (struct sk_buff *)(queue); \
2568 skb = tmp, tmp = skb->next)
2569
1164f52a 2570#define skb_queue_walk_from(queue, skb) \
a1e4891f 2571 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2572 skb = skb->next)
2573
2574#define skb_queue_walk_from_safe(queue, skb, tmp) \
2575 for (tmp = skb->next; \
2576 skb != (struct sk_buff *)(queue); \
2577 skb = tmp, tmp = skb->next)
2578
300ce174
SH
2579#define skb_queue_reverse_walk(queue, skb) \
2580 for (skb = (queue)->prev; \
a1e4891f 2581 skb != (struct sk_buff *)(queue); \
300ce174
SH
2582 skb = skb->prev)
2583
686a2955
DM
2584#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2585 for (skb = (queue)->prev, tmp = skb->prev; \
2586 skb != (struct sk_buff *)(queue); \
2587 skb = tmp, tmp = skb->prev)
2588
2589#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2590 for (tmp = skb->prev; \
2591 skb != (struct sk_buff *)(queue); \
2592 skb = tmp, tmp = skb->prev)
1da177e4 2593
21dc3301 2594static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2595{
2596 return skb_shinfo(skb)->frag_list != NULL;
2597}
2598
2599static inline void skb_frag_list_init(struct sk_buff *skb)
2600{
2601 skb_shinfo(skb)->frag_list = NULL;
2602}
2603
2604static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2605{
2606 frag->next = skb_shinfo(skb)->frag_list;
2607 skb_shinfo(skb)->frag_list = frag;
2608}
2609
2610#define skb_walk_frags(skb, iter) \
2611 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2612
7965bd4d
JP
2613struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2614 int *peeked, int *off, int *err);
2615struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2616 int *err);
2617unsigned int datagram_poll(struct file *file, struct socket *sock,
2618 struct poll_table_struct *wait);
2619int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2620 struct iovec *to, int size);
2621int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2622 struct iovec *iov);
2623int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2624 const struct iovec *from, int from_offset,
2625 int len);
2626int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2627 int offset, size_t count);
2628int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2629 const struct iovec *to, int to_offset,
2630 int size);
2631void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2632void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2633int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2634int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2635int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2636__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2637 int len, __wsum csum);
2638int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2639 struct pipe_inode_info *pipe, unsigned int len,
2640 unsigned int flags);
2641void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2642unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2643int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2644 int len, int hlen);
7965bd4d
JP
2645void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2646int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2647void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2648unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2649struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2650struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2651
2817a336
DB
2652struct skb_checksum_ops {
2653 __wsum (*update)(const void *mem, int len, __wsum wsum);
2654 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2655};
2656
2657__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2658 __wsum csum, const struct skb_checksum_ops *ops);
2659__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2660 __wsum csum);
2661
690e36e7
DM
2662static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2663 int len, void *data, int hlen, void *buffer)
1da177e4 2664{
55820ee2 2665 if (hlen - offset >= len)
690e36e7 2666 return data + offset;
1da177e4 2667
690e36e7
DM
2668 if (!skb ||
2669 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2670 return NULL;
2671
2672 return buffer;
2673}
2674
690e36e7
DM
2675static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2676 int len, void *buffer)
2677{
2678 return __skb_header_pointer(skb, offset, len, skb->data,
2679 skb_headlen(skb), buffer);
2680}
2681
4262e5cc
DB
2682/**
2683 * skb_needs_linearize - check if we need to linearize a given skb
2684 * depending on the given device features.
2685 * @skb: socket buffer to check
2686 * @features: net device features
2687 *
2688 * Returns true if either:
2689 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2690 * 2. skb is fragmented and the device does not support SG.
2691 */
2692static inline bool skb_needs_linearize(struct sk_buff *skb,
2693 netdev_features_t features)
2694{
2695 return skb_is_nonlinear(skb) &&
2696 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2697 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2698}
2699
d626f62b
ACM
2700static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2701 void *to,
2702 const unsigned int len)
2703{
2704 memcpy(to, skb->data, len);
2705}
2706
2707static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2708 const int offset, void *to,
2709 const unsigned int len)
2710{
2711 memcpy(to, skb->data + offset, len);
2712}
2713
27d7ff46
ACM
2714static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2715 const void *from,
2716 const unsigned int len)
2717{
2718 memcpy(skb->data, from, len);
2719}
2720
2721static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2722 const int offset,
2723 const void *from,
2724 const unsigned int len)
2725{
2726 memcpy(skb->data + offset, from, len);
2727}
2728
7965bd4d 2729void skb_init(void);
1da177e4 2730
ac45f602
PO
2731static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2732{
2733 return skb->tstamp;
2734}
2735
a61bbcf2
PM
2736/**
2737 * skb_get_timestamp - get timestamp from a skb
2738 * @skb: skb to get stamp from
2739 * @stamp: pointer to struct timeval to store stamp in
2740 *
2741 * Timestamps are stored in the skb as offsets to a base timestamp.
2742 * This function converts the offset back to a struct timeval and stores
2743 * it in stamp.
2744 */
ac45f602
PO
2745static inline void skb_get_timestamp(const struct sk_buff *skb,
2746 struct timeval *stamp)
a61bbcf2 2747{
b7aa0bf7 2748 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2749}
2750
ac45f602
PO
2751static inline void skb_get_timestampns(const struct sk_buff *skb,
2752 struct timespec *stamp)
2753{
2754 *stamp = ktime_to_timespec(skb->tstamp);
2755}
2756
b7aa0bf7 2757static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2758{
b7aa0bf7 2759 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2760}
2761
164891aa
SH
2762static inline ktime_t net_timedelta(ktime_t t)
2763{
2764 return ktime_sub(ktime_get_real(), t);
2765}
2766
b9ce204f
IJ
2767static inline ktime_t net_invalid_timestamp(void)
2768{
2769 return ktime_set(0, 0);
2770}
a61bbcf2 2771
62bccb8c
AD
2772struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2773
c1f19b51
RC
2774#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2775
7965bd4d
JP
2776void skb_clone_tx_timestamp(struct sk_buff *skb);
2777bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2778
2779#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2780
2781static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2782{
2783}
2784
2785static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2786{
2787 return false;
2788}
2789
2790#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2791
2792/**
2793 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2794 *
da92b194
RC
2795 * PHY drivers may accept clones of transmitted packets for
2796 * timestamping via their phy_driver.txtstamp method. These drivers
2797 * must call this function to return the skb back to the stack, with
2798 * or without a timestamp.
2799 *
c1f19b51 2800 * @skb: clone of the the original outgoing packet
da92b194 2801 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2802 *
2803 */
2804void skb_complete_tx_timestamp(struct sk_buff *skb,
2805 struct skb_shared_hwtstamps *hwtstamps);
2806
e7fd2885
WB
2807void __skb_tstamp_tx(struct sk_buff *orig_skb,
2808 struct skb_shared_hwtstamps *hwtstamps,
2809 struct sock *sk, int tstype);
2810
ac45f602
PO
2811/**
2812 * skb_tstamp_tx - queue clone of skb with send time stamps
2813 * @orig_skb: the original outgoing packet
2814 * @hwtstamps: hardware time stamps, may be NULL if not available
2815 *
2816 * If the skb has a socket associated, then this function clones the
2817 * skb (thus sharing the actual data and optional structures), stores
2818 * the optional hardware time stamping information (if non NULL) or
2819 * generates a software time stamp (otherwise), then queues the clone
2820 * to the error queue of the socket. Errors are silently ignored.
2821 */
7965bd4d
JP
2822void skb_tstamp_tx(struct sk_buff *orig_skb,
2823 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2824
4507a715
RC
2825static inline void sw_tx_timestamp(struct sk_buff *skb)
2826{
2244d07b
OH
2827 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2828 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2829 skb_tstamp_tx(skb, NULL);
2830}
2831
2832/**
2833 * skb_tx_timestamp() - Driver hook for transmit timestamping
2834 *
2835 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2836 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2837 *
73409f3b
DM
2838 * Specifically, one should make absolutely sure that this function is
2839 * called before TX completion of this packet can trigger. Otherwise
2840 * the packet could potentially already be freed.
2841 *
4507a715
RC
2842 * @skb: A socket buffer.
2843 */
2844static inline void skb_tx_timestamp(struct sk_buff *skb)
2845{
c1f19b51 2846 skb_clone_tx_timestamp(skb);
4507a715
RC
2847 sw_tx_timestamp(skb);
2848}
2849
6e3e939f
JB
2850/**
2851 * skb_complete_wifi_ack - deliver skb with wifi status
2852 *
2853 * @skb: the original outgoing packet
2854 * @acked: ack status
2855 *
2856 */
2857void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2858
7965bd4d
JP
2859__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2860__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2861
60476372
HX
2862static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2863{
5d0c2b95 2864 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2865}
2866
fb286bb2
HX
2867/**
2868 * skb_checksum_complete - Calculate checksum of an entire packet
2869 * @skb: packet to process
2870 *
2871 * This function calculates the checksum over the entire packet plus
2872 * the value of skb->csum. The latter can be used to supply the
2873 * checksum of a pseudo header as used by TCP/UDP. It returns the
2874 * checksum.
2875 *
2876 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2877 * this function can be used to verify that checksum on received
2878 * packets. In that case the function should return zero if the
2879 * checksum is correct. In particular, this function will return zero
2880 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2881 * hardware has already verified the correctness of the checksum.
2882 */
4381ca3c 2883static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2884{
60476372
HX
2885 return skb_csum_unnecessary(skb) ?
2886 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2887}
2888
77cffe23
TH
2889static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2890{
2891 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2892 if (skb->csum_level == 0)
2893 skb->ip_summed = CHECKSUM_NONE;
2894 else
2895 skb->csum_level--;
2896 }
2897}
2898
2899static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2900{
2901 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2902 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2903 skb->csum_level++;
2904 } else if (skb->ip_summed == CHECKSUM_NONE) {
2905 skb->ip_summed = CHECKSUM_UNNECESSARY;
2906 skb->csum_level = 0;
2907 }
2908}
2909
5a212329
TH
2910static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2911{
2912 /* Mark current checksum as bad (typically called from GRO
2913 * path). In the case that ip_summed is CHECKSUM_NONE
2914 * this must be the first checksum encountered in the packet.
2915 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2916 * checksum after the last one validated. For UDP, a zero
2917 * checksum can not be marked as bad.
2918 */
2919
2920 if (skb->ip_summed == CHECKSUM_NONE ||
2921 skb->ip_summed == CHECKSUM_UNNECESSARY)
2922 skb->csum_bad = 1;
2923}
2924
76ba0aae
TH
2925/* Check if we need to perform checksum complete validation.
2926 *
2927 * Returns true if checksum complete is needed, false otherwise
2928 * (either checksum is unnecessary or zero checksum is allowed).
2929 */
2930static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2931 bool zero_okay,
2932 __sum16 check)
2933{
5d0c2b95
TH
2934 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2935 skb->csum_valid = 1;
77cffe23 2936 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2937 return false;
2938 }
2939
2940 return true;
2941}
2942
2943/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2944 * in checksum_init.
2945 */
2946#define CHECKSUM_BREAK 76
2947
2948/* Validate (init) checksum based on checksum complete.
2949 *
2950 * Return values:
2951 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2952 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2953 * checksum is stored in skb->csum for use in __skb_checksum_complete
2954 * non-zero: value of invalid checksum
2955 *
2956 */
2957static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2958 bool complete,
2959 __wsum psum)
2960{
2961 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2962 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2963 skb->csum_valid = 1;
76ba0aae
TH
2964 return 0;
2965 }
5a212329
TH
2966 } else if (skb->csum_bad) {
2967 /* ip_summed == CHECKSUM_NONE in this case */
2968 return 1;
76ba0aae
TH
2969 }
2970
2971 skb->csum = psum;
2972
5d0c2b95
TH
2973 if (complete || skb->len <= CHECKSUM_BREAK) {
2974 __sum16 csum;
2975
2976 csum = __skb_checksum_complete(skb);
2977 skb->csum_valid = !csum;
2978 return csum;
2979 }
76ba0aae
TH
2980
2981 return 0;
2982}
2983
2984static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2985{
2986 return 0;
2987}
2988
2989/* Perform checksum validate (init). Note that this is a macro since we only
2990 * want to calculate the pseudo header which is an input function if necessary.
2991 * First we try to validate without any computation (checksum unnecessary) and
2992 * then calculate based on checksum complete calling the function to compute
2993 * pseudo header.
2994 *
2995 * Return values:
2996 * 0: checksum is validated or try to in skb_checksum_complete
2997 * non-zero: value of invalid checksum
2998 */
2999#define __skb_checksum_validate(skb, proto, complete, \
3000 zero_okay, check, compute_pseudo) \
3001({ \
3002 __sum16 __ret = 0; \
5d0c2b95 3003 skb->csum_valid = 0; \
76ba0aae
TH
3004 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3005 __ret = __skb_checksum_validate_complete(skb, \
3006 complete, compute_pseudo(skb, proto)); \
3007 __ret; \
3008})
3009
3010#define skb_checksum_init(skb, proto, compute_pseudo) \
3011 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3012
3013#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3014 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3015
3016#define skb_checksum_validate(skb, proto, compute_pseudo) \
3017 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3018
3019#define skb_checksum_validate_zero_check(skb, proto, check, \
3020 compute_pseudo) \
3021 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3022
3023#define skb_checksum_simple_validate(skb) \
3024 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3025
d96535a1
TH
3026static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3027{
3028 return (skb->ip_summed == CHECKSUM_NONE &&
3029 skb->csum_valid && !skb->csum_bad);
3030}
3031
3032static inline void __skb_checksum_convert(struct sk_buff *skb,
3033 __sum16 check, __wsum pseudo)
3034{
3035 skb->csum = ~pseudo;
3036 skb->ip_summed = CHECKSUM_COMPLETE;
3037}
3038
3039#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3040do { \
3041 if (__skb_checksum_convert_check(skb)) \
3042 __skb_checksum_convert(skb, check, \
3043 compute_pseudo(skb, proto)); \
3044} while (0)
3045
5f79e0f9 3046#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3047void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3048static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3049{
3050 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3051 nf_conntrack_destroy(nfct);
1da177e4
LT
3052}
3053static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3054{
3055 if (nfct)
3056 atomic_inc(&nfct->use);
3057}
2fc72c7b 3058#endif
34666d46 3059#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3060static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3061{
3062 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3063 kfree(nf_bridge);
3064}
3065static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3066{
3067 if (nf_bridge)
3068 atomic_inc(&nf_bridge->use);
3069}
3070#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3071static inline void nf_reset(struct sk_buff *skb)
3072{
5f79e0f9 3073#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3074 nf_conntrack_put(skb->nfct);
3075 skb->nfct = NULL;
2fc72c7b 3076#endif
34666d46 3077#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3078 nf_bridge_put(skb->nf_bridge);
3079 skb->nf_bridge = NULL;
3080#endif
3081}
3082
124dff01
PM
3083static inline void nf_reset_trace(struct sk_buff *skb)
3084{
478b360a 3085#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3086 skb->nf_trace = 0;
3087#endif
a193a4ab
PM
3088}
3089
edda553c 3090/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3091static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3092 bool copy)
edda553c 3093{
5f79e0f9 3094#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3095 dst->nfct = src->nfct;
3096 nf_conntrack_get(src->nfct);
b1937227
ED
3097 if (copy)
3098 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3099#endif
34666d46 3100#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3101 dst->nf_bridge = src->nf_bridge;
3102 nf_bridge_get(src->nf_bridge);
3103#endif
478b360a 3104#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3105 if (copy)
3106 dst->nf_trace = src->nf_trace;
478b360a 3107#endif
edda553c
YK
3108}
3109
e7ac05f3
YK
3110static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3111{
e7ac05f3 3112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3113 nf_conntrack_put(dst->nfct);
2fc72c7b 3114#endif
34666d46 3115#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3116 nf_bridge_put(dst->nf_bridge);
3117#endif
b1937227 3118 __nf_copy(dst, src, true);
e7ac05f3
YK
3119}
3120
984bc16c
JM
3121#ifdef CONFIG_NETWORK_SECMARK
3122static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3123{
3124 to->secmark = from->secmark;
3125}
3126
3127static inline void skb_init_secmark(struct sk_buff *skb)
3128{
3129 skb->secmark = 0;
3130}
3131#else
3132static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3133{ }
3134
3135static inline void skb_init_secmark(struct sk_buff *skb)
3136{ }
3137#endif
3138
574f7194
EB
3139static inline bool skb_irq_freeable(const struct sk_buff *skb)
3140{
3141 return !skb->destructor &&
3142#if IS_ENABLED(CONFIG_XFRM)
3143 !skb->sp &&
3144#endif
3145#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3146 !skb->nfct &&
3147#endif
3148 !skb->_skb_refdst &&
3149 !skb_has_frag_list(skb);
3150}
3151
f25f4e44
PWJ
3152static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3153{
f25f4e44 3154 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3155}
3156
9247744e 3157static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3158{
4e3ab47a 3159 return skb->queue_mapping;
4e3ab47a
PE
3160}
3161
f25f4e44
PWJ
3162static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3163{
f25f4e44 3164 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3165}
3166
d5a9e24a
DM
3167static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3168{
3169 skb->queue_mapping = rx_queue + 1;
3170}
3171
9247744e 3172static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3173{
3174 return skb->queue_mapping - 1;
3175}
3176
9247744e 3177static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3178{
a02cec21 3179 return skb->queue_mapping != 0;
d5a9e24a
DM
3180}
3181
0e001614 3182u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3183 unsigned int num_tx_queues);
9247744e 3184
def8b4fa
AD
3185static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3186{
0b3d8e08 3187#ifdef CONFIG_XFRM
def8b4fa 3188 return skb->sp;
def8b4fa 3189#else
def8b4fa 3190 return NULL;
def8b4fa 3191#endif
0b3d8e08 3192}
def8b4fa 3193
68c33163
PS
3194/* Keeps track of mac header offset relative to skb->head.
3195 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3196 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3197 * tunnel skb it points to outer mac header.
3198 * Keeps track of level of encapsulation of network headers.
3199 */
68c33163 3200struct skb_gso_cb {
3347c960
ED
3201 int mac_offset;
3202 int encap_level;
7e2b10c1 3203 __u16 csum_start;
68c33163
PS
3204};
3205#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3206
3207static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3208{
3209 return (skb_mac_header(inner_skb) - inner_skb->head) -
3210 SKB_GSO_CB(inner_skb)->mac_offset;
3211}
3212
1e2bd517
PS
3213static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3214{
3215 int new_headroom, headroom;
3216 int ret;
3217
3218 headroom = skb_headroom(skb);
3219 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3220 if (ret)
3221 return ret;
3222
3223 new_headroom = skb_headroom(skb);
3224 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3225 return 0;
3226}
3227
7e2b10c1
TH
3228/* Compute the checksum for a gso segment. First compute the checksum value
3229 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3230 * then add in skb->csum (checksum from csum_start to end of packet).
3231 * skb->csum and csum_start are then updated to reflect the checksum of the
3232 * resultant packet starting from the transport header-- the resultant checksum
3233 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3234 * header.
3235 */
3236static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3237{
3238 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3239 skb_transport_offset(skb);
3240 __u16 csum;
3241
3242 csum = csum_fold(csum_partial(skb_transport_header(skb),
3243 plen, skb->csum));
3244 skb->csum = res;
3245 SKB_GSO_CB(skb)->csum_start -= plen;
3246
3247 return csum;
3248}
3249
bdcc0924 3250static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3251{
3252 return skb_shinfo(skb)->gso_size;
3253}
3254
36a8f39e 3255/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3256static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3257{
3258 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3259}
3260
7965bd4d 3261void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3262
3263static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3264{
3265 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3266 * wanted then gso_type will be set. */
05bdd2f1
ED
3267 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3268
b78462eb
AD
3269 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3270 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3271 __skb_warn_lro_forwarding(skb);
3272 return true;
3273 }
3274 return false;
3275}
3276
35fc92a9
HX
3277static inline void skb_forward_csum(struct sk_buff *skb)
3278{
3279 /* Unfortunately we don't support this one. Any brave souls? */
3280 if (skb->ip_summed == CHECKSUM_COMPLETE)
3281 skb->ip_summed = CHECKSUM_NONE;
3282}
3283
bc8acf2c
ED
3284/**
3285 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3286 * @skb: skb to check
3287 *
3288 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3289 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3290 * use this helper, to document places where we make this assertion.
3291 */
05bdd2f1 3292static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3293{
3294#ifdef DEBUG
3295 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3296#endif
3297}
3298
f35d9d8a 3299bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3300
ed1f50c3
PD
3301int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3302
56193d1b
AD
3303u32 skb_get_poff(const struct sk_buff *skb);
3304u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3305 const struct flow_keys *keys, int hlen);
f77668dc 3306
3a7c1ee4
AD
3307/**
3308 * skb_head_is_locked - Determine if the skb->head is locked down
3309 * @skb: skb to check
3310 *
3311 * The head on skbs build around a head frag can be removed if they are
3312 * not cloned. This function returns true if the skb head is locked down
3313 * due to either being allocated via kmalloc, or by being a clone with
3314 * multiple references to the head.
3315 */
3316static inline bool skb_head_is_locked(const struct sk_buff *skb)
3317{
3318 return !skb->head_frag || skb_cloned(skb);
3319}
fe6cc55f
FW
3320
3321/**
3322 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3323 *
3324 * @skb: GSO skb
3325 *
3326 * skb_gso_network_seglen is used to determine the real size of the
3327 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3328 *
3329 * The MAC/L2 header is not accounted for.
3330 */
3331static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3332{
3333 unsigned int hdr_len = skb_transport_header(skb) -
3334 skb_network_header(skb);
3335 return hdr_len + skb_gso_transport_seglen(skb);
3336}
1da177e4
LT
3337#endif /* __KERNEL__ */
3338#endif /* _LINUX_SKBUFF_H */
This page took 1.566744 seconds and 5 git commands to generate.