Merge tag 'arm64-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
[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 */
ebcf34f3 560 /* private: */
b1937227 561 __u32 headers_start[0];
ebcf34f3 562 /* public: */
4031ae6e 563
233577a2
HFS
564/* if you move pkt_type around you also must adapt those constants */
565#ifdef __BIG_ENDIAN_BITFIELD
566#define PKT_TYPE_MAX (7 << 5)
567#else
568#define PKT_TYPE_MAX 7
1da177e4 569#endif
233577a2 570#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 571
233577a2 572 __u8 __pkt_type_offset[0];
b1937227 573 __u8 pkt_type:3;
c93bdd0e 574 __u8 pfmemalloc:1;
b1937227
ED
575 __u8 ignore_df:1;
576 __u8 nfctinfo:3;
577
578 __u8 nf_trace:1;
579 __u8 ip_summed:2;
3853b584 580 __u8 ooo_okay:1;
61b905da 581 __u8 l4_hash:1;
a3b18ddb 582 __u8 sw_hash:1;
6e3e939f
JB
583 __u8 wifi_acked_valid:1;
584 __u8 wifi_acked:1;
b1937227 585
3bdc0eba 586 __u8 no_fcs:1;
77cffe23 587 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 588 __u8 encapsulation:1;
7e2b10c1 589 __u8 encap_hdr_csum:1;
5d0c2b95 590 __u8 csum_valid:1;
7e3cead5 591 __u8 csum_complete_sw:1;
b1937227
ED
592 __u8 csum_level:2;
593 __u8 csum_bad:1;
fe55f6d5 594
b1937227
ED
595#ifdef CONFIG_IPV6_NDISC_NODETYPE
596 __u8 ndisc_nodetype:2;
597#endif
598 __u8 ipvs_property:1;
8bce6d7d
TH
599 __u8 inner_protocol_type:1;
600 /* 4 or 6 bit hole */
b1937227
ED
601
602#ifdef CONFIG_NET_SCHED
603 __u16 tc_index; /* traffic control index */
604#ifdef CONFIG_NET_CLS_ACT
605 __u16 tc_verd; /* traffic control verdict */
606#endif
607#endif
fe55f6d5 608
b1937227
ED
609 union {
610 __wsum csum;
611 struct {
612 __u16 csum_start;
613 __u16 csum_offset;
614 };
615 };
616 __u32 priority;
617 int skb_iif;
618 __u32 hash;
619 __be16 vlan_proto;
620 __u16 vlan_tci;
7bced397
DW
621#ifdef CONFIG_NET_RX_BUSY_POLL
622 unsigned int napi_id;
97fc2f08 623#endif
984bc16c
JM
624#ifdef CONFIG_NETWORK_SECMARK
625 __u32 secmark;
626#endif
3b885787
NH
627 union {
628 __u32 mark;
629 __u32 dropcount;
16fad69c 630 __u32 reserved_tailroom;
3b885787 631 };
1da177e4 632
8bce6d7d
TH
633 union {
634 __be16 inner_protocol;
635 __u8 inner_ipproto;
636 };
637
1a37e412
SH
638 __u16 inner_transport_header;
639 __u16 inner_network_header;
640 __u16 inner_mac_header;
b1937227
ED
641
642 __be16 protocol;
1a37e412
SH
643 __u16 transport_header;
644 __u16 network_header;
645 __u16 mac_header;
b1937227 646
ebcf34f3 647 /* private: */
b1937227 648 __u32 headers_end[0];
ebcf34f3 649 /* public: */
b1937227 650
1da177e4 651 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 652 sk_buff_data_t tail;
4305b541 653 sk_buff_data_t end;
1da177e4 654 unsigned char *head,
4305b541 655 *data;
27a884dc
ACM
656 unsigned int truesize;
657 atomic_t users;
1da177e4
LT
658};
659
660#ifdef __KERNEL__
661/*
662 * Handling routines are only of interest to the kernel
663 */
664#include <linux/slab.h>
665
1da177e4 666
c93bdd0e
MG
667#define SKB_ALLOC_FCLONE 0x01
668#define SKB_ALLOC_RX 0x02
669
670/* Returns true if the skb was allocated from PFMEMALLOC reserves */
671static inline bool skb_pfmemalloc(const struct sk_buff *skb)
672{
673 return unlikely(skb->pfmemalloc);
674}
675
7fee226a
ED
676/*
677 * skb might have a dst pointer attached, refcounted or not.
678 * _skb_refdst low order bit is set if refcount was _not_ taken
679 */
680#define SKB_DST_NOREF 1UL
681#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
682
683/**
684 * skb_dst - returns skb dst_entry
685 * @skb: buffer
686 *
687 * Returns skb dst_entry, regardless of reference taken or not.
688 */
adf30907
ED
689static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
690{
7fee226a
ED
691 /* If refdst was not refcounted, check we still are in a
692 * rcu_read_lock section
693 */
694 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
695 !rcu_read_lock_held() &&
696 !rcu_read_lock_bh_held());
697 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
698}
699
7fee226a
ED
700/**
701 * skb_dst_set - sets skb dst
702 * @skb: buffer
703 * @dst: dst entry
704 *
705 * Sets skb dst, assuming a reference was taken on dst and should
706 * be released by skb_dst_drop()
707 */
adf30907
ED
708static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
709{
7fee226a
ED
710 skb->_skb_refdst = (unsigned long)dst;
711}
712
7965bd4d
JP
713void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
714 bool force);
932bc4d7
JA
715
716/**
717 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
718 * @skb: buffer
719 * @dst: dst entry
720 *
721 * Sets skb dst, assuming a reference was not taken on dst.
722 * If dst entry is cached, we do not take reference and dst_release
723 * will be avoided by refdst_drop. If dst entry is not cached, we take
724 * reference, so that last dst_release can destroy the dst immediately.
725 */
726static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
727{
728 __skb_dst_set_noref(skb, dst, false);
729}
730
731/**
732 * skb_dst_set_noref_force - sets skb dst, without taking reference
733 * @skb: buffer
734 * @dst: dst entry
735 *
736 * Sets skb dst, assuming a reference was not taken on dst.
737 * No reference is taken and no dst_release will be called. While for
738 * cached dsts deferred reclaim is a basic feature, for entries that are
739 * not cached it is caller's job to guarantee that last dst_release for
740 * provided dst happens when nobody uses it, eg. after a RCU grace period.
741 */
742static inline void skb_dst_set_noref_force(struct sk_buff *skb,
743 struct dst_entry *dst)
744{
745 __skb_dst_set_noref(skb, dst, true);
746}
7fee226a
ED
747
748/**
25985edc 749 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
750 * @skb: buffer
751 */
752static inline bool skb_dst_is_noref(const struct sk_buff *skb)
753{
754 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
755}
756
511c3f92
ED
757static inline struct rtable *skb_rtable(const struct sk_buff *skb)
758{
adf30907 759 return (struct rtable *)skb_dst(skb);
511c3f92
ED
760}
761
7965bd4d
JP
762void kfree_skb(struct sk_buff *skb);
763void kfree_skb_list(struct sk_buff *segs);
764void skb_tx_error(struct sk_buff *skb);
765void consume_skb(struct sk_buff *skb);
766void __kfree_skb(struct sk_buff *skb);
d7e8883c 767extern struct kmem_cache *skbuff_head_cache;
bad43ca8 768
7965bd4d
JP
769void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
770bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
771 bool *fragstolen, int *delta_truesize);
bad43ca8 772
7965bd4d
JP
773struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
774 int node);
775struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 776static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 777 gfp_t priority)
d179cd12 778{
564824b0 779 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
780}
781
2e4e4410
ED
782struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
783 unsigned long data_len,
784 int max_page_order,
785 int *errcode,
786 gfp_t gfp_mask);
787
d0bf4a9e
ED
788/* Layout of fast clones : [skb1][skb2][fclone_ref] */
789struct sk_buff_fclones {
790 struct sk_buff skb1;
791
792 struct sk_buff skb2;
793
794 atomic_t fclone_ref;
795};
796
797/**
798 * skb_fclone_busy - check if fclone is busy
799 * @skb: buffer
800 *
801 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
802 * Some drivers call skb_orphan() in their ndo_start_xmit(),
803 * so we also check that this didnt happen.
d0bf4a9e 804 */
39bb5e62
ED
805static inline bool skb_fclone_busy(const struct sock *sk,
806 const struct sk_buff *skb)
d0bf4a9e
ED
807{
808 const struct sk_buff_fclones *fclones;
809
810 fclones = container_of(skb, struct sk_buff_fclones, skb1);
811
812 return skb->fclone == SKB_FCLONE_ORIG &&
39bb5e62
ED
813 fclones->skb2.fclone == SKB_FCLONE_CLONE &&
814 fclones->skb2.sk == sk;
d0bf4a9e
ED
815}
816
d179cd12 817static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 818 gfp_t priority)
d179cd12 819{
c93bdd0e 820 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
821}
822
7965bd4d 823struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
824static inline struct sk_buff *alloc_skb_head(gfp_t priority)
825{
826 return __alloc_skb_head(priority, -1);
827}
828
7965bd4d
JP
829struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
830int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
831struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
832struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
833struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
834 gfp_t gfp_mask, bool fclone);
835static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
836 gfp_t gfp_mask)
837{
838 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
839}
7965bd4d
JP
840
841int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
842struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
843 unsigned int headroom);
844struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
845 int newtailroom, gfp_t priority);
25a91d8d
FD
846int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
847 int offset, int len);
7965bd4d
JP
848int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
849 int len);
850int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
851int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 852#define dev_kfree_skb(a) consume_skb(a)
1da177e4 853
7965bd4d
JP
854int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
855 int getfrag(void *from, char *to, int offset,
856 int len, int odd, struct sk_buff *skb),
857 void *from, int length);
e89e9cf5 858
d94d9fee 859struct skb_seq_state {
677e90ed
TG
860 __u32 lower_offset;
861 __u32 upper_offset;
862 __u32 frag_idx;
863 __u32 stepped_offset;
864 struct sk_buff *root_skb;
865 struct sk_buff *cur_skb;
866 __u8 *frag_data;
867};
868
7965bd4d
JP
869void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
870 unsigned int to, struct skb_seq_state *st);
871unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
872 struct skb_seq_state *st);
873void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 874
7965bd4d
JP
875unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
876 unsigned int to, struct ts_config *config,
877 struct ts_state *state);
3fc7e8a6 878
09323cc4
TH
879/*
880 * Packet hash types specify the type of hash in skb_set_hash.
881 *
882 * Hash types refer to the protocol layer addresses which are used to
883 * construct a packet's hash. The hashes are used to differentiate or identify
884 * flows of the protocol layer for the hash type. Hash types are either
885 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
886 *
887 * Properties of hashes:
888 *
889 * 1) Two packets in different flows have different hash values
890 * 2) Two packets in the same flow should have the same hash value
891 *
892 * A hash at a higher layer is considered to be more specific. A driver should
893 * set the most specific hash possible.
894 *
895 * A driver cannot indicate a more specific hash than the layer at which a hash
896 * was computed. For instance an L3 hash cannot be set as an L4 hash.
897 *
898 * A driver may indicate a hash level which is less specific than the
899 * actual layer the hash was computed on. For instance, a hash computed
900 * at L4 may be considered an L3 hash. This should only be done if the
901 * driver can't unambiguously determine that the HW computed the hash at
902 * the higher layer. Note that the "should" in the second property above
903 * permits this.
904 */
905enum pkt_hash_types {
906 PKT_HASH_TYPE_NONE, /* Undefined type */
907 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
908 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
909 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
910};
911
912static inline void
913skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
914{
61b905da 915 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 916 skb->sw_hash = 0;
61b905da 917 skb->hash = hash;
09323cc4
TH
918}
919
3958afa1
TH
920void __skb_get_hash(struct sk_buff *skb);
921static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 922{
a3b18ddb 923 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 924 __skb_get_hash(skb);
bfb564e7 925
61b905da 926 return skb->hash;
bfb564e7
KK
927}
928
57bdf7f4
TH
929static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
930{
61b905da 931 return skb->hash;
57bdf7f4
TH
932}
933
7539fadc
TH
934static inline void skb_clear_hash(struct sk_buff *skb)
935{
61b905da 936 skb->hash = 0;
a3b18ddb 937 skb->sw_hash = 0;
61b905da 938 skb->l4_hash = 0;
7539fadc
TH
939}
940
941static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
942{
61b905da 943 if (!skb->l4_hash)
7539fadc
TH
944 skb_clear_hash(skb);
945}
946
3df7a74e
TH
947static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
948{
61b905da 949 to->hash = from->hash;
a3b18ddb 950 to->sw_hash = from->sw_hash;
61b905da 951 to->l4_hash = from->l4_hash;
3df7a74e
TH
952};
953
4305b541
ACM
954#ifdef NET_SKBUFF_DATA_USES_OFFSET
955static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
956{
957 return skb->head + skb->end;
958}
ec47ea82
AD
959
960static inline unsigned int skb_end_offset(const struct sk_buff *skb)
961{
962 return skb->end;
963}
4305b541
ACM
964#else
965static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
966{
967 return skb->end;
968}
ec47ea82
AD
969
970static inline unsigned int skb_end_offset(const struct sk_buff *skb)
971{
972 return skb->end - skb->head;
973}
4305b541
ACM
974#endif
975
1da177e4 976/* Internal */
4305b541 977#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 978
ac45f602
PO
979static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
980{
981 return &skb_shinfo(skb)->hwtstamps;
982}
983
1da177e4
LT
984/**
985 * skb_queue_empty - check if a queue is empty
986 * @list: queue head
987 *
988 * Returns true if the queue is empty, false otherwise.
989 */
990static inline int skb_queue_empty(const struct sk_buff_head *list)
991{
fd44b93c 992 return list->next == (const struct sk_buff *) list;
1da177e4
LT
993}
994
fc7ebb21
DM
995/**
996 * skb_queue_is_last - check if skb is the last entry in the queue
997 * @list: queue head
998 * @skb: buffer
999 *
1000 * Returns true if @skb is the last buffer on the list.
1001 */
1002static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1003 const struct sk_buff *skb)
1004{
fd44b93c 1005 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1006}
1007
832d11c5
IJ
1008/**
1009 * skb_queue_is_first - check if skb is the first entry in the queue
1010 * @list: queue head
1011 * @skb: buffer
1012 *
1013 * Returns true if @skb is the first buffer on the list.
1014 */
1015static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1016 const struct sk_buff *skb)
1017{
fd44b93c 1018 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1019}
1020
249c8b42
DM
1021/**
1022 * skb_queue_next - return the next packet in the queue
1023 * @list: queue head
1024 * @skb: current buffer
1025 *
1026 * Return the next packet in @list after @skb. It is only valid to
1027 * call this if skb_queue_is_last() evaluates to false.
1028 */
1029static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1030 const struct sk_buff *skb)
1031{
1032 /* This BUG_ON may seem severe, but if we just return then we
1033 * are going to dereference garbage.
1034 */
1035 BUG_ON(skb_queue_is_last(list, skb));
1036 return skb->next;
1037}
1038
832d11c5
IJ
1039/**
1040 * skb_queue_prev - return the prev packet in the queue
1041 * @list: queue head
1042 * @skb: current buffer
1043 *
1044 * Return the prev packet in @list before @skb. It is only valid to
1045 * call this if skb_queue_is_first() evaluates to false.
1046 */
1047static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1048 const struct sk_buff *skb)
1049{
1050 /* This BUG_ON may seem severe, but if we just return then we
1051 * are going to dereference garbage.
1052 */
1053 BUG_ON(skb_queue_is_first(list, skb));
1054 return skb->prev;
1055}
1056
1da177e4
LT
1057/**
1058 * skb_get - reference buffer
1059 * @skb: buffer to reference
1060 *
1061 * Makes another reference to a socket buffer and returns a pointer
1062 * to the buffer.
1063 */
1064static inline struct sk_buff *skb_get(struct sk_buff *skb)
1065{
1066 atomic_inc(&skb->users);
1067 return skb;
1068}
1069
1070/*
1071 * If users == 1, we are the only owner and are can avoid redundant
1072 * atomic change.
1073 */
1074
1da177e4
LT
1075/**
1076 * skb_cloned - is the buffer a clone
1077 * @skb: buffer to check
1078 *
1079 * Returns true if the buffer was generated with skb_clone() and is
1080 * one of multiple shared copies of the buffer. Cloned buffers are
1081 * shared data so must not be written to under normal circumstances.
1082 */
1083static inline int skb_cloned(const struct sk_buff *skb)
1084{
1085 return skb->cloned &&
1086 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1087}
1088
14bbd6a5
PS
1089static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1090{
1091 might_sleep_if(pri & __GFP_WAIT);
1092
1093 if (skb_cloned(skb))
1094 return pskb_expand_head(skb, 0, 0, pri);
1095
1096 return 0;
1097}
1098
1da177e4
LT
1099/**
1100 * skb_header_cloned - is the header a clone
1101 * @skb: buffer to check
1102 *
1103 * Returns true if modifying the header part of the buffer requires
1104 * the data to be copied.
1105 */
1106static inline int skb_header_cloned(const struct sk_buff *skb)
1107{
1108 int dataref;
1109
1110 if (!skb->cloned)
1111 return 0;
1112
1113 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1114 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1115 return dataref != 1;
1116}
1117
1118/**
1119 * skb_header_release - release reference to header
1120 * @skb: buffer to operate on
1121 *
1122 * Drop a reference to the header part of the buffer. This is done
1123 * by acquiring a payload reference. You must not read from the header
1124 * part of skb->data after this.
f4a775d1 1125 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1126 */
1127static inline void skb_header_release(struct sk_buff *skb)
1128{
1129 BUG_ON(skb->nohdr);
1130 skb->nohdr = 1;
1131 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1132}
1133
f4a775d1
ED
1134/**
1135 * __skb_header_release - release reference to header
1136 * @skb: buffer to operate on
1137 *
1138 * Variant of skb_header_release() assuming skb is private to caller.
1139 * We can avoid one atomic operation.
1140 */
1141static inline void __skb_header_release(struct sk_buff *skb)
1142{
1143 skb->nohdr = 1;
1144 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1145}
1146
1147
1da177e4
LT
1148/**
1149 * skb_shared - is the buffer shared
1150 * @skb: buffer to check
1151 *
1152 * Returns true if more than one person has a reference to this
1153 * buffer.
1154 */
1155static inline int skb_shared(const struct sk_buff *skb)
1156{
1157 return atomic_read(&skb->users) != 1;
1158}
1159
1160/**
1161 * skb_share_check - check if buffer is shared and if so clone it
1162 * @skb: buffer to check
1163 * @pri: priority for memory allocation
1164 *
1165 * If the buffer is shared the buffer is cloned and the old copy
1166 * drops a reference. A new clone with a single reference is returned.
1167 * If the buffer is not shared the original buffer is returned. When
1168 * being called from interrupt status or with spinlocks held pri must
1169 * be GFP_ATOMIC.
1170 *
1171 * NULL is returned on a memory allocation failure.
1172 */
47061bc4 1173static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1174{
1175 might_sleep_if(pri & __GFP_WAIT);
1176 if (skb_shared(skb)) {
1177 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1178
1179 if (likely(nskb))
1180 consume_skb(skb);
1181 else
1182 kfree_skb(skb);
1da177e4
LT
1183 skb = nskb;
1184 }
1185 return skb;
1186}
1187
1188/*
1189 * Copy shared buffers into a new sk_buff. We effectively do COW on
1190 * packets to handle cases where we have a local reader and forward
1191 * and a couple of other messy ones. The normal one is tcpdumping
1192 * a packet thats being forwarded.
1193 */
1194
1195/**
1196 * skb_unshare - make a copy of a shared buffer
1197 * @skb: buffer to check
1198 * @pri: priority for memory allocation
1199 *
1200 * If the socket buffer is a clone then this function creates a new
1201 * copy of the data, drops a reference count on the old copy and returns
1202 * the new copy with the reference count at 1. If the buffer is not a clone
1203 * the original buffer is returned. When called with a spinlock held or
1204 * from interrupt state @pri must be %GFP_ATOMIC
1205 *
1206 * %NULL is returned on a memory allocation failure.
1207 */
e2bf521d 1208static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1209 gfp_t pri)
1da177e4
LT
1210{
1211 might_sleep_if(pri & __GFP_WAIT);
1212 if (skb_cloned(skb)) {
1213 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1214
1215 /* Free our shared copy */
1216 if (likely(nskb))
1217 consume_skb(skb);
1218 else
1219 kfree_skb(skb);
1da177e4
LT
1220 skb = nskb;
1221 }
1222 return skb;
1223}
1224
1225/**
1a5778aa 1226 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1227 * @list_: list to peek at
1228 *
1229 * Peek an &sk_buff. Unlike most other operations you _MUST_
1230 * be careful with this one. A peek leaves the buffer on the
1231 * list and someone else may run off with it. You must hold
1232 * the appropriate locks or have a private queue to do this.
1233 *
1234 * Returns %NULL for an empty list or a pointer to the head element.
1235 * The reference count is not incremented and the reference is therefore
1236 * volatile. Use with caution.
1237 */
05bdd2f1 1238static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1239{
18d07000
ED
1240 struct sk_buff *skb = list_->next;
1241
1242 if (skb == (struct sk_buff *)list_)
1243 skb = NULL;
1244 return skb;
1da177e4
LT
1245}
1246
da5ef6e5
PE
1247/**
1248 * skb_peek_next - peek skb following the given one from a queue
1249 * @skb: skb to start from
1250 * @list_: list to peek at
1251 *
1252 * Returns %NULL when the end of the list is met or a pointer to the
1253 * next element. The reference count is not incremented and the
1254 * reference is therefore volatile. Use with caution.
1255 */
1256static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1257 const struct sk_buff_head *list_)
1258{
1259 struct sk_buff *next = skb->next;
18d07000 1260
da5ef6e5
PE
1261 if (next == (struct sk_buff *)list_)
1262 next = NULL;
1263 return next;
1264}
1265
1da177e4 1266/**
1a5778aa 1267 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1268 * @list_: list to peek at
1269 *
1270 * Peek an &sk_buff. Unlike most other operations you _MUST_
1271 * be careful with this one. A peek leaves the buffer on the
1272 * list and someone else may run off with it. You must hold
1273 * the appropriate locks or have a private queue to do this.
1274 *
1275 * Returns %NULL for an empty list or a pointer to the tail element.
1276 * The reference count is not incremented and the reference is therefore
1277 * volatile. Use with caution.
1278 */
05bdd2f1 1279static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1280{
18d07000
ED
1281 struct sk_buff *skb = list_->prev;
1282
1283 if (skb == (struct sk_buff *)list_)
1284 skb = NULL;
1285 return skb;
1286
1da177e4
LT
1287}
1288
1289/**
1290 * skb_queue_len - get queue length
1291 * @list_: list to measure
1292 *
1293 * Return the length of an &sk_buff queue.
1294 */
1295static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1296{
1297 return list_->qlen;
1298}
1299
67fed459
DM
1300/**
1301 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1302 * @list: queue to initialize
1303 *
1304 * This initializes only the list and queue length aspects of
1305 * an sk_buff_head object. This allows to initialize the list
1306 * aspects of an sk_buff_head without reinitializing things like
1307 * the spinlock. It can also be used for on-stack sk_buff_head
1308 * objects where the spinlock is known to not be used.
1309 */
1310static inline void __skb_queue_head_init(struct sk_buff_head *list)
1311{
1312 list->prev = list->next = (struct sk_buff *)list;
1313 list->qlen = 0;
1314}
1315
76f10ad0
AV
1316/*
1317 * This function creates a split out lock class for each invocation;
1318 * this is needed for now since a whole lot of users of the skb-queue
1319 * infrastructure in drivers have different locking usage (in hardirq)
1320 * than the networking core (in softirq only). In the long run either the
1321 * network layer or drivers should need annotation to consolidate the
1322 * main types of usage into 3 classes.
1323 */
1da177e4
LT
1324static inline void skb_queue_head_init(struct sk_buff_head *list)
1325{
1326 spin_lock_init(&list->lock);
67fed459 1327 __skb_queue_head_init(list);
1da177e4
LT
1328}
1329
c2ecba71
PE
1330static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1331 struct lock_class_key *class)
1332{
1333 skb_queue_head_init(list);
1334 lockdep_set_class(&list->lock, class);
1335}
1336
1da177e4 1337/*
bf299275 1338 * Insert an sk_buff on a list.
1da177e4
LT
1339 *
1340 * The "__skb_xxxx()" functions are the non-atomic ones that
1341 * can only be called with interrupts disabled.
1342 */
7965bd4d
JP
1343void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1344 struct sk_buff_head *list);
bf299275
GR
1345static inline void __skb_insert(struct sk_buff *newsk,
1346 struct sk_buff *prev, struct sk_buff *next,
1347 struct sk_buff_head *list)
1348{
1349 newsk->next = next;
1350 newsk->prev = prev;
1351 next->prev = prev->next = newsk;
1352 list->qlen++;
1353}
1da177e4 1354
67fed459
DM
1355static inline void __skb_queue_splice(const struct sk_buff_head *list,
1356 struct sk_buff *prev,
1357 struct sk_buff *next)
1358{
1359 struct sk_buff *first = list->next;
1360 struct sk_buff *last = list->prev;
1361
1362 first->prev = prev;
1363 prev->next = first;
1364
1365 last->next = next;
1366 next->prev = last;
1367}
1368
1369/**
1370 * skb_queue_splice - join two skb lists, this is designed for stacks
1371 * @list: the new list to add
1372 * @head: the place to add it in the first list
1373 */
1374static inline void skb_queue_splice(const struct sk_buff_head *list,
1375 struct sk_buff_head *head)
1376{
1377 if (!skb_queue_empty(list)) {
1378 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1379 head->qlen += list->qlen;
67fed459
DM
1380 }
1381}
1382
1383/**
d9619496 1384 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1385 * @list: the new list to add
1386 * @head: the place to add it in the first list
1387 *
1388 * The list at @list is reinitialised
1389 */
1390static inline void skb_queue_splice_init(struct sk_buff_head *list,
1391 struct sk_buff_head *head)
1392{
1393 if (!skb_queue_empty(list)) {
1394 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1395 head->qlen += list->qlen;
67fed459
DM
1396 __skb_queue_head_init(list);
1397 }
1398}
1399
1400/**
1401 * skb_queue_splice_tail - join two skb lists, each list being a queue
1402 * @list: the new list to add
1403 * @head: the place to add it in the first list
1404 */
1405static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1406 struct sk_buff_head *head)
1407{
1408 if (!skb_queue_empty(list)) {
1409 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1410 head->qlen += list->qlen;
67fed459
DM
1411 }
1412}
1413
1414/**
d9619496 1415 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1416 * @list: the new list to add
1417 * @head: the place to add it in the first list
1418 *
1419 * Each of the lists is a queue.
1420 * The list at @list is reinitialised
1421 */
1422static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1423 struct sk_buff_head *head)
1424{
1425 if (!skb_queue_empty(list)) {
1426 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1427 head->qlen += list->qlen;
67fed459
DM
1428 __skb_queue_head_init(list);
1429 }
1430}
1431
1da177e4 1432/**
300ce174 1433 * __skb_queue_after - queue a buffer at the list head
1da177e4 1434 * @list: list to use
300ce174 1435 * @prev: place after this buffer
1da177e4
LT
1436 * @newsk: buffer to queue
1437 *
300ce174 1438 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1439 * and you must therefore hold required locks before calling it.
1440 *
1441 * A buffer cannot be placed on two lists at the same time.
1442 */
300ce174
SH
1443static inline void __skb_queue_after(struct sk_buff_head *list,
1444 struct sk_buff *prev,
1445 struct sk_buff *newsk)
1da177e4 1446{
bf299275 1447 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1448}
1449
7965bd4d
JP
1450void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1451 struct sk_buff_head *list);
7de6c033 1452
f5572855
GR
1453static inline void __skb_queue_before(struct sk_buff_head *list,
1454 struct sk_buff *next,
1455 struct sk_buff *newsk)
1456{
1457 __skb_insert(newsk, next->prev, next, list);
1458}
1459
300ce174
SH
1460/**
1461 * __skb_queue_head - queue a buffer at the list head
1462 * @list: list to use
1463 * @newsk: buffer to queue
1464 *
1465 * Queue a buffer at the start of a list. This function takes no locks
1466 * and you must therefore hold required locks before calling it.
1467 *
1468 * A buffer cannot be placed on two lists at the same time.
1469 */
7965bd4d 1470void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1471static inline void __skb_queue_head(struct sk_buff_head *list,
1472 struct sk_buff *newsk)
1473{
1474 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1475}
1476
1da177e4
LT
1477/**
1478 * __skb_queue_tail - queue a buffer at the list tail
1479 * @list: list to use
1480 * @newsk: buffer to queue
1481 *
1482 * Queue a buffer at the end of a list. This function takes no locks
1483 * and you must therefore hold required locks before calling it.
1484 *
1485 * A buffer cannot be placed on two lists at the same time.
1486 */
7965bd4d 1487void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1488static inline void __skb_queue_tail(struct sk_buff_head *list,
1489 struct sk_buff *newsk)
1490{
f5572855 1491 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1492}
1493
1da177e4
LT
1494/*
1495 * remove sk_buff from list. _Must_ be called atomically, and with
1496 * the list known..
1497 */
7965bd4d 1498void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1499static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1500{
1501 struct sk_buff *next, *prev;
1502
1503 list->qlen--;
1504 next = skb->next;
1505 prev = skb->prev;
1506 skb->next = skb->prev = NULL;
1da177e4
LT
1507 next->prev = prev;
1508 prev->next = next;
1509}
1510
f525c06d
GR
1511/**
1512 * __skb_dequeue - remove from the head of the queue
1513 * @list: list to dequeue from
1514 *
1515 * Remove the head of the list. This function does not take any locks
1516 * so must be used with appropriate locks held only. The head item is
1517 * returned or %NULL if the list is empty.
1518 */
7965bd4d 1519struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1520static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1521{
1522 struct sk_buff *skb = skb_peek(list);
1523 if (skb)
1524 __skb_unlink(skb, list);
1525 return skb;
1526}
1da177e4
LT
1527
1528/**
1529 * __skb_dequeue_tail - remove from the tail of the queue
1530 * @list: list to dequeue from
1531 *
1532 * Remove the tail of the list. This function does not take any locks
1533 * so must be used with appropriate locks held only. The tail item is
1534 * returned or %NULL if the list is empty.
1535 */
7965bd4d 1536struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1537static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1538{
1539 struct sk_buff *skb = skb_peek_tail(list);
1540 if (skb)
1541 __skb_unlink(skb, list);
1542 return skb;
1543}
1544
1545
bdcc0924 1546static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1547{
1548 return skb->data_len;
1549}
1550
1551static inline unsigned int skb_headlen(const struct sk_buff *skb)
1552{
1553 return skb->len - skb->data_len;
1554}
1555
1556static inline int skb_pagelen(const struct sk_buff *skb)
1557{
1558 int i, len = 0;
1559
1560 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1561 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1562 return len + skb_headlen(skb);
1563}
1564
131ea667
IC
1565/**
1566 * __skb_fill_page_desc - initialise a paged fragment in an skb
1567 * @skb: buffer containing fragment to be initialised
1568 * @i: paged fragment index to initialise
1569 * @page: the page to use for this fragment
1570 * @off: the offset to the data with @page
1571 * @size: the length of the data
1572 *
1573 * Initialises the @i'th fragment of @skb to point to &size bytes at
1574 * offset @off within @page.
1575 *
1576 * Does not take any additional reference on the fragment.
1577 */
1578static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1579 struct page *page, int off, int size)
1da177e4
LT
1580{
1581 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1582
c48a11c7
MG
1583 /*
1584 * Propagate page->pfmemalloc to the skb if we can. The problem is
1585 * that not all callers have unique ownership of the page. If
1586 * pfmemalloc is set, we check the mapping as a mapping implies
1587 * page->index is set (index and pfmemalloc share space).
1588 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1589 * do not lose pfmemalloc information as the pages would not be
1590 * allocated using __GFP_MEMALLOC.
1591 */
a8605c60 1592 frag->page.p = page;
1da177e4 1593 frag->page_offset = off;
9e903e08 1594 skb_frag_size_set(frag, size);
cca7af38
PE
1595
1596 page = compound_head(page);
1597 if (page->pfmemalloc && !page->mapping)
1598 skb->pfmemalloc = true;
131ea667
IC
1599}
1600
1601/**
1602 * skb_fill_page_desc - initialise a paged fragment in an skb
1603 * @skb: buffer containing fragment to be initialised
1604 * @i: paged fragment index to initialise
1605 * @page: the page to use for this fragment
1606 * @off: the offset to the data with @page
1607 * @size: the length of the data
1608 *
1609 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1610 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1611 * addition updates @skb such that @i is the last fragment.
1612 *
1613 * Does not take any additional reference on the fragment.
1614 */
1615static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1616 struct page *page, int off, int size)
1617{
1618 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1619 skb_shinfo(skb)->nr_frags = i + 1;
1620}
1621
7965bd4d
JP
1622void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1623 int size, unsigned int truesize);
654bed16 1624
f8e617e1
JW
1625void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1626 unsigned int truesize);
1627
1da177e4 1628#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1629#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1630#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1631
27a884dc
ACM
1632#ifdef NET_SKBUFF_DATA_USES_OFFSET
1633static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1634{
1635 return skb->head + skb->tail;
1636}
1637
1638static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1639{
1640 skb->tail = skb->data - skb->head;
1641}
1642
1643static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1644{
1645 skb_reset_tail_pointer(skb);
1646 skb->tail += offset;
1647}
7cc46190 1648
27a884dc
ACM
1649#else /* NET_SKBUFF_DATA_USES_OFFSET */
1650static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1651{
1652 return skb->tail;
1653}
1654
1655static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1656{
1657 skb->tail = skb->data;
1658}
1659
1660static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1661{
1662 skb->tail = skb->data + offset;
1663}
4305b541 1664
27a884dc
ACM
1665#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1666
1da177e4
LT
1667/*
1668 * Add data to an sk_buff
1669 */
0c7ddf36 1670unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1671unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1672static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1673{
27a884dc 1674 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1675 SKB_LINEAR_ASSERT(skb);
1676 skb->tail += len;
1677 skb->len += len;
1678 return tmp;
1679}
1680
7965bd4d 1681unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1682static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1683{
1684 skb->data -= len;
1685 skb->len += len;
1686 return skb->data;
1687}
1688
7965bd4d 1689unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1690static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1691{
1692 skb->len -= len;
1693 BUG_ON(skb->len < skb->data_len);
1694 return skb->data += len;
1695}
1696
47d29646
DM
1697static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1698{
1699 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1700}
1701
7965bd4d 1702unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1703
1704static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1705{
1706 if (len > skb_headlen(skb) &&
987c402a 1707 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1708 return NULL;
1709 skb->len -= len;
1710 return skb->data += len;
1711}
1712
1713static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1714{
1715 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1716}
1717
1718static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1719{
1720 if (likely(len <= skb_headlen(skb)))
1721 return 1;
1722 if (unlikely(len > skb->len))
1723 return 0;
987c402a 1724 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1725}
1726
1727/**
1728 * skb_headroom - bytes at buffer head
1729 * @skb: buffer to check
1730 *
1731 * Return the number of bytes of free space at the head of an &sk_buff.
1732 */
c2636b4d 1733static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1734{
1735 return skb->data - skb->head;
1736}
1737
1738/**
1739 * skb_tailroom - bytes at buffer end
1740 * @skb: buffer to check
1741 *
1742 * Return the number of bytes of free space at the tail of an sk_buff
1743 */
1744static inline int skb_tailroom(const struct sk_buff *skb)
1745{
4305b541 1746 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1747}
1748
a21d4572
ED
1749/**
1750 * skb_availroom - bytes at buffer end
1751 * @skb: buffer to check
1752 *
1753 * Return the number of bytes of free space at the tail of an sk_buff
1754 * allocated by sk_stream_alloc()
1755 */
1756static inline int skb_availroom(const struct sk_buff *skb)
1757{
16fad69c
ED
1758 if (skb_is_nonlinear(skb))
1759 return 0;
1760
1761 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1762}
1763
1da177e4
LT
1764/**
1765 * skb_reserve - adjust headroom
1766 * @skb: buffer to alter
1767 * @len: bytes to move
1768 *
1769 * Increase the headroom of an empty &sk_buff by reducing the tail
1770 * room. This is only allowed for an empty buffer.
1771 */
8243126c 1772static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1773{
1774 skb->data += len;
1775 skb->tail += len;
1776}
1777
8bce6d7d
TH
1778#define ENCAP_TYPE_ETHER 0
1779#define ENCAP_TYPE_IPPROTO 1
1780
1781static inline void skb_set_inner_protocol(struct sk_buff *skb,
1782 __be16 protocol)
1783{
1784 skb->inner_protocol = protocol;
1785 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1786}
1787
1788static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1789 __u8 ipproto)
1790{
1791 skb->inner_ipproto = ipproto;
1792 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1793}
1794
6a674e9c
JG
1795static inline void skb_reset_inner_headers(struct sk_buff *skb)
1796{
aefbd2b3 1797 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1798 skb->inner_network_header = skb->network_header;
1799 skb->inner_transport_header = skb->transport_header;
1800}
1801
0b5c9db1
JP
1802static inline void skb_reset_mac_len(struct sk_buff *skb)
1803{
1804 skb->mac_len = skb->network_header - skb->mac_header;
1805}
1806
6a674e9c
JG
1807static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1808 *skb)
1809{
1810 return skb->head + skb->inner_transport_header;
1811}
1812
1813static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1814{
1815 skb->inner_transport_header = skb->data - skb->head;
1816}
1817
1818static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1819 const int offset)
1820{
1821 skb_reset_inner_transport_header(skb);
1822 skb->inner_transport_header += offset;
1823}
1824
1825static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1826{
1827 return skb->head + skb->inner_network_header;
1828}
1829
1830static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1831{
1832 skb->inner_network_header = skb->data - skb->head;
1833}
1834
1835static inline void skb_set_inner_network_header(struct sk_buff *skb,
1836 const int offset)
1837{
1838 skb_reset_inner_network_header(skb);
1839 skb->inner_network_header += offset;
1840}
1841
aefbd2b3
PS
1842static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1843{
1844 return skb->head + skb->inner_mac_header;
1845}
1846
1847static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1848{
1849 skb->inner_mac_header = skb->data - skb->head;
1850}
1851
1852static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1853 const int offset)
1854{
1855 skb_reset_inner_mac_header(skb);
1856 skb->inner_mac_header += offset;
1857}
fda55eca
ED
1858static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1859{
35d04610 1860 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1861}
1862
9c70220b
ACM
1863static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1864{
2e07fa9c 1865 return skb->head + skb->transport_header;
9c70220b
ACM
1866}
1867
badff6d0
ACM
1868static inline void skb_reset_transport_header(struct sk_buff *skb)
1869{
2e07fa9c 1870 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1871}
1872
967b05f6
ACM
1873static inline void skb_set_transport_header(struct sk_buff *skb,
1874 const int offset)
1875{
2e07fa9c
ACM
1876 skb_reset_transport_header(skb);
1877 skb->transport_header += offset;
ea2ae17d
ACM
1878}
1879
d56f90a7
ACM
1880static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1881{
2e07fa9c 1882 return skb->head + skb->network_header;
d56f90a7
ACM
1883}
1884
c1d2bbe1
ACM
1885static inline void skb_reset_network_header(struct sk_buff *skb)
1886{
2e07fa9c 1887 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1888}
1889
c14d2450
ACM
1890static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1891{
2e07fa9c
ACM
1892 skb_reset_network_header(skb);
1893 skb->network_header += offset;
c14d2450
ACM
1894}
1895
2e07fa9c 1896static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1897{
2e07fa9c 1898 return skb->head + skb->mac_header;
bbe735e4
ACM
1899}
1900
2e07fa9c 1901static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1902{
35d04610 1903 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1904}
1905
1906static inline void skb_reset_mac_header(struct sk_buff *skb)
1907{
1908 skb->mac_header = skb->data - skb->head;
1909}
1910
1911static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1912{
1913 skb_reset_mac_header(skb);
1914 skb->mac_header += offset;
1915}
1916
0e3da5bb
TT
1917static inline void skb_pop_mac_header(struct sk_buff *skb)
1918{
1919 skb->mac_header = skb->network_header;
1920}
1921
fbbdb8f0
YX
1922static inline void skb_probe_transport_header(struct sk_buff *skb,
1923 const int offset_hint)
1924{
1925 struct flow_keys keys;
1926
1927 if (skb_transport_header_was_set(skb))
1928 return;
1929 else if (skb_flow_dissect(skb, &keys))
1930 skb_set_transport_header(skb, keys.thoff);
1931 else
1932 skb_set_transport_header(skb, offset_hint);
1933}
1934
03606895
ED
1935static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1936{
1937 if (skb_mac_header_was_set(skb)) {
1938 const unsigned char *old_mac = skb_mac_header(skb);
1939
1940 skb_set_mac_header(skb, -skb->mac_len);
1941 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1942 }
1943}
1944
04fb451e
MM
1945static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1946{
1947 return skb->csum_start - skb_headroom(skb);
1948}
1949
2e07fa9c
ACM
1950static inline int skb_transport_offset(const struct sk_buff *skb)
1951{
1952 return skb_transport_header(skb) - skb->data;
1953}
1954
1955static inline u32 skb_network_header_len(const struct sk_buff *skb)
1956{
1957 return skb->transport_header - skb->network_header;
1958}
1959
6a674e9c
JG
1960static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1961{
1962 return skb->inner_transport_header - skb->inner_network_header;
1963}
1964
2e07fa9c
ACM
1965static inline int skb_network_offset(const struct sk_buff *skb)
1966{
1967 return skb_network_header(skb) - skb->data;
1968}
48d49d0c 1969
6a674e9c
JG
1970static inline int skb_inner_network_offset(const struct sk_buff *skb)
1971{
1972 return skb_inner_network_header(skb) - skb->data;
1973}
1974
f9599ce1
CG
1975static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1976{
1977 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1978}
1979
1da177e4
LT
1980/*
1981 * CPUs often take a performance hit when accessing unaligned memory
1982 * locations. The actual performance hit varies, it can be small if the
1983 * hardware handles it or large if we have to take an exception and fix it
1984 * in software.
1985 *
1986 * Since an ethernet header is 14 bytes network drivers often end up with
1987 * the IP header at an unaligned offset. The IP header can be aligned by
1988 * shifting the start of the packet by 2 bytes. Drivers should do this
1989 * with:
1990 *
8660c124 1991 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1992 *
1993 * The downside to this alignment of the IP header is that the DMA is now
1994 * unaligned. On some architectures the cost of an unaligned DMA is high
1995 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1996 *
1da177e4
LT
1997 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1998 * to be overridden.
1999 */
2000#ifndef NET_IP_ALIGN
2001#define NET_IP_ALIGN 2
2002#endif
2003
025be81e
AB
2004/*
2005 * The networking layer reserves some headroom in skb data (via
2006 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2007 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2008 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2009 *
2010 * Unfortunately this headroom changes the DMA alignment of the resulting
2011 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2012 * on some architectures. An architecture can override this value,
2013 * perhaps setting it to a cacheline in size (since that will maintain
2014 * cacheline alignment of the DMA). It must be a power of 2.
2015 *
d6301d3d 2016 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2017 * headroom, you should not reduce this.
5933dd2f
ED
2018 *
2019 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2020 * to reduce average number of cache lines per packet.
2021 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2022 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2023 */
2024#ifndef NET_SKB_PAD
5933dd2f 2025#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2026#endif
2027
7965bd4d 2028int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2029
2030static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2031{
c4264f27 2032 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2033 WARN_ON(1);
2034 return;
2035 }
27a884dc
ACM
2036 skb->len = len;
2037 skb_set_tail_pointer(skb, len);
1da177e4
LT
2038}
2039
7965bd4d 2040void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2041
2042static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2043{
3cc0e873
HX
2044 if (skb->data_len)
2045 return ___pskb_trim(skb, len);
2046 __skb_trim(skb, len);
2047 return 0;
1da177e4
LT
2048}
2049
2050static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2051{
2052 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2053}
2054
e9fa4f7b
HX
2055/**
2056 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2057 * @skb: buffer to alter
2058 * @len: new length
2059 *
2060 * This is identical to pskb_trim except that the caller knows that
2061 * the skb is not cloned so we should never get an error due to out-
2062 * of-memory.
2063 */
2064static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2065{
2066 int err = pskb_trim(skb, len);
2067 BUG_ON(err);
2068}
2069
1da177e4
LT
2070/**
2071 * skb_orphan - orphan a buffer
2072 * @skb: buffer to orphan
2073 *
2074 * If a buffer currently has an owner then we call the owner's
2075 * destructor function and make the @skb unowned. The buffer continues
2076 * to exist but is no longer charged to its former owner.
2077 */
2078static inline void skb_orphan(struct sk_buff *skb)
2079{
c34a7612 2080 if (skb->destructor) {
1da177e4 2081 skb->destructor(skb);
c34a7612
ED
2082 skb->destructor = NULL;
2083 skb->sk = NULL;
376c7311
ED
2084 } else {
2085 BUG_ON(skb->sk);
c34a7612 2086 }
1da177e4
LT
2087}
2088
a353e0ce
MT
2089/**
2090 * skb_orphan_frags - orphan the frags contained in a buffer
2091 * @skb: buffer to orphan frags from
2092 * @gfp_mask: allocation mask for replacement pages
2093 *
2094 * For each frag in the SKB which needs a destructor (i.e. has an
2095 * owner) create a copy of that frag and release the original
2096 * page by calling the destructor.
2097 */
2098static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2099{
2100 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2101 return 0;
2102 return skb_copy_ubufs(skb, gfp_mask);
2103}
2104
1da177e4
LT
2105/**
2106 * __skb_queue_purge - empty a list
2107 * @list: list to empty
2108 *
2109 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2110 * the list and one reference dropped. This function does not take the
2111 * list lock and the caller must hold the relevant locks to use it.
2112 */
7965bd4d 2113void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2114static inline void __skb_queue_purge(struct sk_buff_head *list)
2115{
2116 struct sk_buff *skb;
2117 while ((skb = __skb_dequeue(list)) != NULL)
2118 kfree_skb(skb);
2119}
2120
e5e67305
AD
2121#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2122#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2123#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2124
7965bd4d 2125void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2126
7965bd4d
JP
2127struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2128 gfp_t gfp_mask);
8af27456
CH
2129
2130/**
2131 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2132 * @dev: network device to receive on
2133 * @length: length to allocate
2134 *
2135 * Allocate a new &sk_buff and assign it a usage count of one. The
2136 * buffer has unspecified headroom built in. Users should allocate
2137 * the headroom they think they need without accounting for the
2138 * built in space. The built in space is used for optimisations.
2139 *
2140 * %NULL is returned if there is no free memory. Although this function
2141 * allocates memory it can be called from an interrupt.
2142 */
2143static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2144 unsigned int length)
8af27456
CH
2145{
2146 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2147}
2148
6f532612
ED
2149/* legacy helper around __netdev_alloc_skb() */
2150static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2151 gfp_t gfp_mask)
2152{
2153 return __netdev_alloc_skb(NULL, length, gfp_mask);
2154}
2155
2156/* legacy helper around netdev_alloc_skb() */
2157static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2158{
2159 return netdev_alloc_skb(NULL, length);
2160}
2161
2162
4915a0de
ED
2163static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2164 unsigned int length, gfp_t gfp)
61321bbd 2165{
4915a0de 2166 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2167
2168 if (NET_IP_ALIGN && skb)
2169 skb_reserve(skb, NET_IP_ALIGN);
2170 return skb;
2171}
2172
4915a0de
ED
2173static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2174 unsigned int length)
2175{
2176 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2177}
2178
bc6fc9fa
FF
2179/**
2180 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2181 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2182 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2183 * @order: size of the allocation
2184 *
2185 * Allocate a new page.
2186 *
2187 * %NULL is returned if there is no free memory.
2188*/
2189static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2190 struct sk_buff *skb,
2191 unsigned int order)
2192{
2193 struct page *page;
2194
2195 gfp_mask |= __GFP_COLD;
2196
2197 if (!(gfp_mask & __GFP_NOMEMALLOC))
2198 gfp_mask |= __GFP_MEMALLOC;
2199
2200 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2201 if (skb && page && page->pfmemalloc)
2202 skb->pfmemalloc = true;
2203
2204 return page;
2205}
2206
2207/**
2208 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2209 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2210 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2211 *
2212 * Allocate a new page.
2213 *
2214 * %NULL is returned if there is no free memory.
2215 */
2216static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2217 struct sk_buff *skb)
2218{
2219 return __skb_alloc_pages(gfp_mask, skb, 0);
2220}
2221
2222/**
2223 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2224 * @page: The page that was allocated from skb_alloc_page
2225 * @skb: The skb that may need pfmemalloc set
2226 */
2227static inline void skb_propagate_pfmemalloc(struct page *page,
2228 struct sk_buff *skb)
2229{
2230 if (page && page->pfmemalloc)
2231 skb->pfmemalloc = true;
2232}
2233
131ea667 2234/**
e227867f 2235 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2236 * @frag: the paged fragment
2237 *
2238 * Returns the &struct page associated with @frag.
2239 */
2240static inline struct page *skb_frag_page(const skb_frag_t *frag)
2241{
a8605c60 2242 return frag->page.p;
131ea667
IC
2243}
2244
2245/**
2246 * __skb_frag_ref - take an addition reference on a paged fragment.
2247 * @frag: the paged fragment
2248 *
2249 * Takes an additional reference on the paged fragment @frag.
2250 */
2251static inline void __skb_frag_ref(skb_frag_t *frag)
2252{
2253 get_page(skb_frag_page(frag));
2254}
2255
2256/**
2257 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2258 * @skb: the buffer
2259 * @f: the fragment offset.
2260 *
2261 * Takes an additional reference on the @f'th paged fragment of @skb.
2262 */
2263static inline void skb_frag_ref(struct sk_buff *skb, int f)
2264{
2265 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2266}
2267
2268/**
2269 * __skb_frag_unref - release a reference on a paged fragment.
2270 * @frag: the paged fragment
2271 *
2272 * Releases a reference on the paged fragment @frag.
2273 */
2274static inline void __skb_frag_unref(skb_frag_t *frag)
2275{
2276 put_page(skb_frag_page(frag));
2277}
2278
2279/**
2280 * skb_frag_unref - release a reference on a paged fragment of an skb.
2281 * @skb: the buffer
2282 * @f: the fragment offset
2283 *
2284 * Releases a reference on the @f'th paged fragment of @skb.
2285 */
2286static inline void skb_frag_unref(struct sk_buff *skb, int f)
2287{
2288 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2289}
2290
2291/**
2292 * skb_frag_address - gets the address of the data contained in a paged fragment
2293 * @frag: the paged fragment buffer
2294 *
2295 * Returns the address of the data within @frag. The page must already
2296 * be mapped.
2297 */
2298static inline void *skb_frag_address(const skb_frag_t *frag)
2299{
2300 return page_address(skb_frag_page(frag)) + frag->page_offset;
2301}
2302
2303/**
2304 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2305 * @frag: the paged fragment buffer
2306 *
2307 * Returns the address of the data within @frag. Checks that the page
2308 * is mapped and returns %NULL otherwise.
2309 */
2310static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2311{
2312 void *ptr = page_address(skb_frag_page(frag));
2313 if (unlikely(!ptr))
2314 return NULL;
2315
2316 return ptr + frag->page_offset;
2317}
2318
2319/**
2320 * __skb_frag_set_page - sets the page contained in a paged fragment
2321 * @frag: the paged fragment
2322 * @page: the page to set
2323 *
2324 * Sets the fragment @frag to contain @page.
2325 */
2326static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2327{
a8605c60 2328 frag->page.p = page;
131ea667
IC
2329}
2330
2331/**
2332 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2333 * @skb: the buffer
2334 * @f: the fragment offset
2335 * @page: the page to set
2336 *
2337 * Sets the @f'th fragment of @skb to contain @page.
2338 */
2339static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2340 struct page *page)
2341{
2342 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2343}
2344
400dfd3a
ED
2345bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2346
131ea667
IC
2347/**
2348 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2349 * @dev: the device to map the fragment to
131ea667
IC
2350 * @frag: the paged fragment to map
2351 * @offset: the offset within the fragment (starting at the
2352 * fragment's own offset)
2353 * @size: the number of bytes to map
f83347df 2354 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2355 *
2356 * Maps the page associated with @frag to @device.
2357 */
2358static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2359 const skb_frag_t *frag,
2360 size_t offset, size_t size,
2361 enum dma_data_direction dir)
2362{
2363 return dma_map_page(dev, skb_frag_page(frag),
2364 frag->page_offset + offset, size, dir);
2365}
2366
117632e6
ED
2367static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2368 gfp_t gfp_mask)
2369{
2370 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2371}
2372
bad93e9d
OP
2373
2374static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2375 gfp_t gfp_mask)
2376{
2377 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2378}
2379
2380
334a8132
PM
2381/**
2382 * skb_clone_writable - is the header of a clone writable
2383 * @skb: buffer to check
2384 * @len: length up to which to write
2385 *
2386 * Returns true if modifying the header part of the cloned buffer
2387 * does not requires the data to be copied.
2388 */
05bdd2f1 2389static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2390{
2391 return !skb_header_cloned(skb) &&
2392 skb_headroom(skb) + len <= skb->hdr_len;
2393}
2394
d9cc2048
HX
2395static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2396 int cloned)
2397{
2398 int delta = 0;
2399
d9cc2048
HX
2400 if (headroom > skb_headroom(skb))
2401 delta = headroom - skb_headroom(skb);
2402
2403 if (delta || cloned)
2404 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2405 GFP_ATOMIC);
2406 return 0;
2407}
2408
1da177e4
LT
2409/**
2410 * skb_cow - copy header of skb when it is required
2411 * @skb: buffer to cow
2412 * @headroom: needed headroom
2413 *
2414 * If the skb passed lacks sufficient headroom or its data part
2415 * is shared, data is reallocated. If reallocation fails, an error
2416 * is returned and original skb is not changed.
2417 *
2418 * The result is skb with writable area skb->head...skb->tail
2419 * and at least @headroom of space at head.
2420 */
2421static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2422{
d9cc2048
HX
2423 return __skb_cow(skb, headroom, skb_cloned(skb));
2424}
1da177e4 2425
d9cc2048
HX
2426/**
2427 * skb_cow_head - skb_cow but only making the head writable
2428 * @skb: buffer to cow
2429 * @headroom: needed headroom
2430 *
2431 * This function is identical to skb_cow except that we replace the
2432 * skb_cloned check by skb_header_cloned. It should be used when
2433 * you only need to push on some header and do not need to modify
2434 * the data.
2435 */
2436static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2437{
2438 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2439}
2440
2441/**
2442 * skb_padto - pad an skbuff up to a minimal size
2443 * @skb: buffer to pad
2444 * @len: minimal length
2445 *
2446 * Pads up a buffer to ensure the trailing bytes exist and are
2447 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2448 * is untouched. Otherwise it is extended. Returns zero on
2449 * success. The skb is freed on error.
1da177e4
LT
2450 */
2451
5b057c6b 2452static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2453{
2454 unsigned int size = skb->len;
2455 if (likely(size >= len))
5b057c6b 2456 return 0;
987c402a 2457 return skb_pad(skb, len - size);
1da177e4
LT
2458}
2459
2460static inline int skb_add_data(struct sk_buff *skb,
2461 char __user *from, int copy)
2462{
2463 const int off = skb->len;
2464
2465 if (skb->ip_summed == CHECKSUM_NONE) {
2466 int err = 0;
5084205f 2467 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2468 copy, 0, &err);
2469 if (!err) {
2470 skb->csum = csum_block_add(skb->csum, csum, off);
2471 return 0;
2472 }
2473 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2474 return 0;
2475
2476 __skb_trim(skb, off);
2477 return -EFAULT;
2478}
2479
38ba0a65
ED
2480static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2481 const struct page *page, int off)
1da177e4
LT
2482{
2483 if (i) {
9e903e08 2484 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2485
ea2ab693 2486 return page == skb_frag_page(frag) &&
9e903e08 2487 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2488 }
38ba0a65 2489 return false;
1da177e4
LT
2490}
2491
364c6bad
HX
2492static inline int __skb_linearize(struct sk_buff *skb)
2493{
2494 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2495}
2496
1da177e4
LT
2497/**
2498 * skb_linearize - convert paged skb to linear one
2499 * @skb: buffer to linarize
1da177e4
LT
2500 *
2501 * If there is no free memory -ENOMEM is returned, otherwise zero
2502 * is returned and the old skb data released.
2503 */
364c6bad
HX
2504static inline int skb_linearize(struct sk_buff *skb)
2505{
2506 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2507}
2508
cef401de
ED
2509/**
2510 * skb_has_shared_frag - can any frag be overwritten
2511 * @skb: buffer to test
2512 *
2513 * Return true if the skb has at least one frag that might be modified
2514 * by an external entity (as in vmsplice()/sendfile())
2515 */
2516static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2517{
c9af6db4
PS
2518 return skb_is_nonlinear(skb) &&
2519 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2520}
2521
364c6bad
HX
2522/**
2523 * skb_linearize_cow - make sure skb is linear and writable
2524 * @skb: buffer to process
2525 *
2526 * If there is no free memory -ENOMEM is returned, otherwise zero
2527 * is returned and the old skb data released.
2528 */
2529static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2530{
364c6bad
HX
2531 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2532 __skb_linearize(skb) : 0;
1da177e4
LT
2533}
2534
2535/**
2536 * skb_postpull_rcsum - update checksum for received skb after pull
2537 * @skb: buffer to update
2538 * @start: start of data before pull
2539 * @len: length of data pulled
2540 *
2541 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2542 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2543 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2544 */
2545
2546static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2547 const void *start, unsigned int len)
1da177e4 2548{
84fa7933 2549 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2550 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2551}
2552
cbb042f9
HX
2553unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2554
7ce5a27f
DM
2555/**
2556 * pskb_trim_rcsum - trim received skb and update checksum
2557 * @skb: buffer to trim
2558 * @len: new length
2559 *
2560 * This is exactly the same as pskb_trim except that it ensures the
2561 * checksum of received packets are still valid after the operation.
2562 */
2563
2564static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2565{
2566 if (likely(len >= skb->len))
2567 return 0;
2568 if (skb->ip_summed == CHECKSUM_COMPLETE)
2569 skb->ip_summed = CHECKSUM_NONE;
2570 return __pskb_trim(skb, len);
2571}
2572
1da177e4
LT
2573#define skb_queue_walk(queue, skb) \
2574 for (skb = (queue)->next; \
a1e4891f 2575 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2576 skb = skb->next)
2577
46f8914e
JC
2578#define skb_queue_walk_safe(queue, skb, tmp) \
2579 for (skb = (queue)->next, tmp = skb->next; \
2580 skb != (struct sk_buff *)(queue); \
2581 skb = tmp, tmp = skb->next)
2582
1164f52a 2583#define skb_queue_walk_from(queue, skb) \
a1e4891f 2584 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2585 skb = skb->next)
2586
2587#define skb_queue_walk_from_safe(queue, skb, tmp) \
2588 for (tmp = skb->next; \
2589 skb != (struct sk_buff *)(queue); \
2590 skb = tmp, tmp = skb->next)
2591
300ce174
SH
2592#define skb_queue_reverse_walk(queue, skb) \
2593 for (skb = (queue)->prev; \
a1e4891f 2594 skb != (struct sk_buff *)(queue); \
300ce174
SH
2595 skb = skb->prev)
2596
686a2955
DM
2597#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2598 for (skb = (queue)->prev, tmp = skb->prev; \
2599 skb != (struct sk_buff *)(queue); \
2600 skb = tmp, tmp = skb->prev)
2601
2602#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2603 for (tmp = skb->prev; \
2604 skb != (struct sk_buff *)(queue); \
2605 skb = tmp, tmp = skb->prev)
1da177e4 2606
21dc3301 2607static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2608{
2609 return skb_shinfo(skb)->frag_list != NULL;
2610}
2611
2612static inline void skb_frag_list_init(struct sk_buff *skb)
2613{
2614 skb_shinfo(skb)->frag_list = NULL;
2615}
2616
2617static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2618{
2619 frag->next = skb_shinfo(skb)->frag_list;
2620 skb_shinfo(skb)->frag_list = frag;
2621}
2622
2623#define skb_walk_frags(skb, iter) \
2624 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2625
7965bd4d
JP
2626struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2627 int *peeked, int *off, int *err);
2628struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2629 int *err);
2630unsigned int datagram_poll(struct file *file, struct socket *sock,
2631 struct poll_table_struct *wait);
2632int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2633 struct iovec *to, int size);
2634int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2635 struct iovec *iov);
2636int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2637 const struct iovec *from, int from_offset,
2638 int len);
2639int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2640 int offset, size_t count);
2641int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2642 const struct iovec *to, int to_offset,
2643 int size);
2644void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2645void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2646int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2647int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2648int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2649__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2650 int len, __wsum csum);
2651int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2652 struct pipe_inode_info *pipe, unsigned int len,
2653 unsigned int flags);
2654void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2655unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2656int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2657 int len, int hlen);
7965bd4d
JP
2658void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2659int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2660void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2661unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2662struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2663struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2664
2817a336
DB
2665struct skb_checksum_ops {
2666 __wsum (*update)(const void *mem, int len, __wsum wsum);
2667 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2668};
2669
2670__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2671 __wsum csum, const struct skb_checksum_ops *ops);
2672__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2673 __wsum csum);
2674
690e36e7
DM
2675static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2676 int len, void *data, int hlen, void *buffer)
1da177e4 2677{
55820ee2 2678 if (hlen - offset >= len)
690e36e7 2679 return data + offset;
1da177e4 2680
690e36e7
DM
2681 if (!skb ||
2682 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2683 return NULL;
2684
2685 return buffer;
2686}
2687
690e36e7
DM
2688static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2689 int len, void *buffer)
2690{
2691 return __skb_header_pointer(skb, offset, len, skb->data,
2692 skb_headlen(skb), buffer);
2693}
2694
4262e5cc
DB
2695/**
2696 * skb_needs_linearize - check if we need to linearize a given skb
2697 * depending on the given device features.
2698 * @skb: socket buffer to check
2699 * @features: net device features
2700 *
2701 * Returns true if either:
2702 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2703 * 2. skb is fragmented and the device does not support SG.
2704 */
2705static inline bool skb_needs_linearize(struct sk_buff *skb,
2706 netdev_features_t features)
2707{
2708 return skb_is_nonlinear(skb) &&
2709 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2710 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2711}
2712
d626f62b
ACM
2713static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2714 void *to,
2715 const unsigned int len)
2716{
2717 memcpy(to, skb->data, len);
2718}
2719
2720static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2721 const int offset, void *to,
2722 const unsigned int len)
2723{
2724 memcpy(to, skb->data + offset, len);
2725}
2726
27d7ff46
ACM
2727static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2728 const void *from,
2729 const unsigned int len)
2730{
2731 memcpy(skb->data, from, len);
2732}
2733
2734static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2735 const int offset,
2736 const void *from,
2737 const unsigned int len)
2738{
2739 memcpy(skb->data + offset, from, len);
2740}
2741
7965bd4d 2742void skb_init(void);
1da177e4 2743
ac45f602
PO
2744static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2745{
2746 return skb->tstamp;
2747}
2748
a61bbcf2
PM
2749/**
2750 * skb_get_timestamp - get timestamp from a skb
2751 * @skb: skb to get stamp from
2752 * @stamp: pointer to struct timeval to store stamp in
2753 *
2754 * Timestamps are stored in the skb as offsets to a base timestamp.
2755 * This function converts the offset back to a struct timeval and stores
2756 * it in stamp.
2757 */
ac45f602
PO
2758static inline void skb_get_timestamp(const struct sk_buff *skb,
2759 struct timeval *stamp)
a61bbcf2 2760{
b7aa0bf7 2761 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2762}
2763
ac45f602
PO
2764static inline void skb_get_timestampns(const struct sk_buff *skb,
2765 struct timespec *stamp)
2766{
2767 *stamp = ktime_to_timespec(skb->tstamp);
2768}
2769
b7aa0bf7 2770static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2771{
b7aa0bf7 2772 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2773}
2774
164891aa
SH
2775static inline ktime_t net_timedelta(ktime_t t)
2776{
2777 return ktime_sub(ktime_get_real(), t);
2778}
2779
b9ce204f
IJ
2780static inline ktime_t net_invalid_timestamp(void)
2781{
2782 return ktime_set(0, 0);
2783}
a61bbcf2 2784
62bccb8c
AD
2785struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2786
c1f19b51
RC
2787#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2788
7965bd4d
JP
2789void skb_clone_tx_timestamp(struct sk_buff *skb);
2790bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2791
2792#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2793
2794static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2795{
2796}
2797
2798static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2799{
2800 return false;
2801}
2802
2803#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2804
2805/**
2806 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2807 *
da92b194
RC
2808 * PHY drivers may accept clones of transmitted packets for
2809 * timestamping via their phy_driver.txtstamp method. These drivers
2810 * must call this function to return the skb back to the stack, with
2811 * or without a timestamp.
2812 *
c1f19b51 2813 * @skb: clone of the the original outgoing packet
da92b194 2814 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2815 *
2816 */
2817void skb_complete_tx_timestamp(struct sk_buff *skb,
2818 struct skb_shared_hwtstamps *hwtstamps);
2819
e7fd2885
WB
2820void __skb_tstamp_tx(struct sk_buff *orig_skb,
2821 struct skb_shared_hwtstamps *hwtstamps,
2822 struct sock *sk, int tstype);
2823
ac45f602
PO
2824/**
2825 * skb_tstamp_tx - queue clone of skb with send time stamps
2826 * @orig_skb: the original outgoing packet
2827 * @hwtstamps: hardware time stamps, may be NULL if not available
2828 *
2829 * If the skb has a socket associated, then this function clones the
2830 * skb (thus sharing the actual data and optional structures), stores
2831 * the optional hardware time stamping information (if non NULL) or
2832 * generates a software time stamp (otherwise), then queues the clone
2833 * to the error queue of the socket. Errors are silently ignored.
2834 */
7965bd4d
JP
2835void skb_tstamp_tx(struct sk_buff *orig_skb,
2836 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2837
4507a715
RC
2838static inline void sw_tx_timestamp(struct sk_buff *skb)
2839{
2244d07b
OH
2840 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2841 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2842 skb_tstamp_tx(skb, NULL);
2843}
2844
2845/**
2846 * skb_tx_timestamp() - Driver hook for transmit timestamping
2847 *
2848 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2849 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2850 *
73409f3b
DM
2851 * Specifically, one should make absolutely sure that this function is
2852 * called before TX completion of this packet can trigger. Otherwise
2853 * the packet could potentially already be freed.
2854 *
4507a715
RC
2855 * @skb: A socket buffer.
2856 */
2857static inline void skb_tx_timestamp(struct sk_buff *skb)
2858{
c1f19b51 2859 skb_clone_tx_timestamp(skb);
4507a715
RC
2860 sw_tx_timestamp(skb);
2861}
2862
6e3e939f
JB
2863/**
2864 * skb_complete_wifi_ack - deliver skb with wifi status
2865 *
2866 * @skb: the original outgoing packet
2867 * @acked: ack status
2868 *
2869 */
2870void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2871
7965bd4d
JP
2872__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2873__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2874
60476372
HX
2875static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2876{
5d0c2b95 2877 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2878}
2879
fb286bb2
HX
2880/**
2881 * skb_checksum_complete - Calculate checksum of an entire packet
2882 * @skb: packet to process
2883 *
2884 * This function calculates the checksum over the entire packet plus
2885 * the value of skb->csum. The latter can be used to supply the
2886 * checksum of a pseudo header as used by TCP/UDP. It returns the
2887 * checksum.
2888 *
2889 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2890 * this function can be used to verify that checksum on received
2891 * packets. In that case the function should return zero if the
2892 * checksum is correct. In particular, this function will return zero
2893 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2894 * hardware has already verified the correctness of the checksum.
2895 */
4381ca3c 2896static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2897{
60476372
HX
2898 return skb_csum_unnecessary(skb) ?
2899 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2900}
2901
77cffe23
TH
2902static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2903{
2904 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2905 if (skb->csum_level == 0)
2906 skb->ip_summed = CHECKSUM_NONE;
2907 else
2908 skb->csum_level--;
2909 }
2910}
2911
2912static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2913{
2914 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2915 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2916 skb->csum_level++;
2917 } else if (skb->ip_summed == CHECKSUM_NONE) {
2918 skb->ip_summed = CHECKSUM_UNNECESSARY;
2919 skb->csum_level = 0;
2920 }
2921}
2922
5a212329
TH
2923static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2924{
2925 /* Mark current checksum as bad (typically called from GRO
2926 * path). In the case that ip_summed is CHECKSUM_NONE
2927 * this must be the first checksum encountered in the packet.
2928 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2929 * checksum after the last one validated. For UDP, a zero
2930 * checksum can not be marked as bad.
2931 */
2932
2933 if (skb->ip_summed == CHECKSUM_NONE ||
2934 skb->ip_summed == CHECKSUM_UNNECESSARY)
2935 skb->csum_bad = 1;
2936}
2937
76ba0aae
TH
2938/* Check if we need to perform checksum complete validation.
2939 *
2940 * Returns true if checksum complete is needed, false otherwise
2941 * (either checksum is unnecessary or zero checksum is allowed).
2942 */
2943static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2944 bool zero_okay,
2945 __sum16 check)
2946{
5d0c2b95
TH
2947 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2948 skb->csum_valid = 1;
77cffe23 2949 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2950 return false;
2951 }
2952
2953 return true;
2954}
2955
2956/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2957 * in checksum_init.
2958 */
2959#define CHECKSUM_BREAK 76
2960
2961/* Validate (init) checksum based on checksum complete.
2962 *
2963 * Return values:
2964 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2965 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2966 * checksum is stored in skb->csum for use in __skb_checksum_complete
2967 * non-zero: value of invalid checksum
2968 *
2969 */
2970static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2971 bool complete,
2972 __wsum psum)
2973{
2974 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2975 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2976 skb->csum_valid = 1;
76ba0aae
TH
2977 return 0;
2978 }
5a212329
TH
2979 } else if (skb->csum_bad) {
2980 /* ip_summed == CHECKSUM_NONE in this case */
2981 return 1;
76ba0aae
TH
2982 }
2983
2984 skb->csum = psum;
2985
5d0c2b95
TH
2986 if (complete || skb->len <= CHECKSUM_BREAK) {
2987 __sum16 csum;
2988
2989 csum = __skb_checksum_complete(skb);
2990 skb->csum_valid = !csum;
2991 return csum;
2992 }
76ba0aae
TH
2993
2994 return 0;
2995}
2996
2997static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2998{
2999 return 0;
3000}
3001
3002/* Perform checksum validate (init). Note that this is a macro since we only
3003 * want to calculate the pseudo header which is an input function if necessary.
3004 * First we try to validate without any computation (checksum unnecessary) and
3005 * then calculate based on checksum complete calling the function to compute
3006 * pseudo header.
3007 *
3008 * Return values:
3009 * 0: checksum is validated or try to in skb_checksum_complete
3010 * non-zero: value of invalid checksum
3011 */
3012#define __skb_checksum_validate(skb, proto, complete, \
3013 zero_okay, check, compute_pseudo) \
3014({ \
3015 __sum16 __ret = 0; \
5d0c2b95 3016 skb->csum_valid = 0; \
76ba0aae
TH
3017 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3018 __ret = __skb_checksum_validate_complete(skb, \
3019 complete, compute_pseudo(skb, proto)); \
3020 __ret; \
3021})
3022
3023#define skb_checksum_init(skb, proto, compute_pseudo) \
3024 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3025
3026#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3027 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3028
3029#define skb_checksum_validate(skb, proto, compute_pseudo) \
3030 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3031
3032#define skb_checksum_validate_zero_check(skb, proto, check, \
3033 compute_pseudo) \
3034 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3035
3036#define skb_checksum_simple_validate(skb) \
3037 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3038
d96535a1
TH
3039static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3040{
3041 return (skb->ip_summed == CHECKSUM_NONE &&
3042 skb->csum_valid && !skb->csum_bad);
3043}
3044
3045static inline void __skb_checksum_convert(struct sk_buff *skb,
3046 __sum16 check, __wsum pseudo)
3047{
3048 skb->csum = ~pseudo;
3049 skb->ip_summed = CHECKSUM_COMPLETE;
3050}
3051
3052#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3053do { \
3054 if (__skb_checksum_convert_check(skb)) \
3055 __skb_checksum_convert(skb, check, \
3056 compute_pseudo(skb, proto)); \
3057} while (0)
3058
5f79e0f9 3059#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3060void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3061static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3062{
3063 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3064 nf_conntrack_destroy(nfct);
1da177e4
LT
3065}
3066static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3067{
3068 if (nfct)
3069 atomic_inc(&nfct->use);
3070}
2fc72c7b 3071#endif
34666d46 3072#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3073static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3074{
3075 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3076 kfree(nf_bridge);
3077}
3078static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3079{
3080 if (nf_bridge)
3081 atomic_inc(&nf_bridge->use);
3082}
3083#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3084static inline void nf_reset(struct sk_buff *skb)
3085{
5f79e0f9 3086#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3087 nf_conntrack_put(skb->nfct);
3088 skb->nfct = NULL;
2fc72c7b 3089#endif
34666d46 3090#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3091 nf_bridge_put(skb->nf_bridge);
3092 skb->nf_bridge = NULL;
3093#endif
3094}
3095
124dff01
PM
3096static inline void nf_reset_trace(struct sk_buff *skb)
3097{
478b360a 3098#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3099 skb->nf_trace = 0;
3100#endif
a193a4ab
PM
3101}
3102
edda553c 3103/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3104static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3105 bool copy)
edda553c 3106{
5f79e0f9 3107#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3108 dst->nfct = src->nfct;
3109 nf_conntrack_get(src->nfct);
b1937227
ED
3110 if (copy)
3111 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3112#endif
34666d46 3113#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3114 dst->nf_bridge = src->nf_bridge;
3115 nf_bridge_get(src->nf_bridge);
3116#endif
478b360a 3117#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3118 if (copy)
3119 dst->nf_trace = src->nf_trace;
478b360a 3120#endif
edda553c
YK
3121}
3122
e7ac05f3
YK
3123static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3124{
e7ac05f3 3125#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3126 nf_conntrack_put(dst->nfct);
2fc72c7b 3127#endif
34666d46 3128#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3129 nf_bridge_put(dst->nf_bridge);
3130#endif
b1937227 3131 __nf_copy(dst, src, true);
e7ac05f3
YK
3132}
3133
984bc16c
JM
3134#ifdef CONFIG_NETWORK_SECMARK
3135static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3136{
3137 to->secmark = from->secmark;
3138}
3139
3140static inline void skb_init_secmark(struct sk_buff *skb)
3141{
3142 skb->secmark = 0;
3143}
3144#else
3145static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3146{ }
3147
3148static inline void skb_init_secmark(struct sk_buff *skb)
3149{ }
3150#endif
3151
574f7194
EB
3152static inline bool skb_irq_freeable(const struct sk_buff *skb)
3153{
3154 return !skb->destructor &&
3155#if IS_ENABLED(CONFIG_XFRM)
3156 !skb->sp &&
3157#endif
3158#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3159 !skb->nfct &&
3160#endif
3161 !skb->_skb_refdst &&
3162 !skb_has_frag_list(skb);
3163}
3164
f25f4e44
PWJ
3165static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3166{
f25f4e44 3167 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3168}
3169
9247744e 3170static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3171{
4e3ab47a 3172 return skb->queue_mapping;
4e3ab47a
PE
3173}
3174
f25f4e44
PWJ
3175static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3176{
f25f4e44 3177 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3178}
3179
d5a9e24a
DM
3180static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3181{
3182 skb->queue_mapping = rx_queue + 1;
3183}
3184
9247744e 3185static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3186{
3187 return skb->queue_mapping - 1;
3188}
3189
9247744e 3190static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3191{
a02cec21 3192 return skb->queue_mapping != 0;
d5a9e24a
DM
3193}
3194
0e001614 3195u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3196 unsigned int num_tx_queues);
9247744e 3197
def8b4fa
AD
3198static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3199{
0b3d8e08 3200#ifdef CONFIG_XFRM
def8b4fa 3201 return skb->sp;
def8b4fa 3202#else
def8b4fa 3203 return NULL;
def8b4fa 3204#endif
0b3d8e08 3205}
def8b4fa 3206
68c33163
PS
3207/* Keeps track of mac header offset relative to skb->head.
3208 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3209 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3210 * tunnel skb it points to outer mac header.
3211 * Keeps track of level of encapsulation of network headers.
3212 */
68c33163 3213struct skb_gso_cb {
3347c960
ED
3214 int mac_offset;
3215 int encap_level;
7e2b10c1 3216 __u16 csum_start;
68c33163
PS
3217};
3218#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3219
3220static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3221{
3222 return (skb_mac_header(inner_skb) - inner_skb->head) -
3223 SKB_GSO_CB(inner_skb)->mac_offset;
3224}
3225
1e2bd517
PS
3226static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3227{
3228 int new_headroom, headroom;
3229 int ret;
3230
3231 headroom = skb_headroom(skb);
3232 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3233 if (ret)
3234 return ret;
3235
3236 new_headroom = skb_headroom(skb);
3237 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3238 return 0;
3239}
3240
7e2b10c1
TH
3241/* Compute the checksum for a gso segment. First compute the checksum value
3242 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3243 * then add in skb->csum (checksum from csum_start to end of packet).
3244 * skb->csum and csum_start are then updated to reflect the checksum of the
3245 * resultant packet starting from the transport header-- the resultant checksum
3246 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3247 * header.
3248 */
3249static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3250{
3251 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3252 skb_transport_offset(skb);
3253 __u16 csum;
3254
3255 csum = csum_fold(csum_partial(skb_transport_header(skb),
3256 plen, skb->csum));
3257 skb->csum = res;
3258 SKB_GSO_CB(skb)->csum_start -= plen;
3259
3260 return csum;
3261}
3262
bdcc0924 3263static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3264{
3265 return skb_shinfo(skb)->gso_size;
3266}
3267
36a8f39e 3268/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3269static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3270{
3271 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3272}
3273
7965bd4d 3274void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3275
3276static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3277{
3278 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3279 * wanted then gso_type will be set. */
05bdd2f1
ED
3280 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3281
b78462eb
AD
3282 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3283 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3284 __skb_warn_lro_forwarding(skb);
3285 return true;
3286 }
3287 return false;
3288}
3289
35fc92a9
HX
3290static inline void skb_forward_csum(struct sk_buff *skb)
3291{
3292 /* Unfortunately we don't support this one. Any brave souls? */
3293 if (skb->ip_summed == CHECKSUM_COMPLETE)
3294 skb->ip_summed = CHECKSUM_NONE;
3295}
3296
bc8acf2c
ED
3297/**
3298 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3299 * @skb: skb to check
3300 *
3301 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3302 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3303 * use this helper, to document places where we make this assertion.
3304 */
05bdd2f1 3305static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3306{
3307#ifdef DEBUG
3308 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3309#endif
3310}
3311
f35d9d8a 3312bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3313
ed1f50c3
PD
3314int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3315
56193d1b
AD
3316u32 skb_get_poff(const struct sk_buff *skb);
3317u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3318 const struct flow_keys *keys, int hlen);
f77668dc 3319
3a7c1ee4
AD
3320/**
3321 * skb_head_is_locked - Determine if the skb->head is locked down
3322 * @skb: skb to check
3323 *
3324 * The head on skbs build around a head frag can be removed if they are
3325 * not cloned. This function returns true if the skb head is locked down
3326 * due to either being allocated via kmalloc, or by being a clone with
3327 * multiple references to the head.
3328 */
3329static inline bool skb_head_is_locked(const struct sk_buff *skb)
3330{
3331 return !skb->head_frag || skb_cloned(skb);
3332}
fe6cc55f
FW
3333
3334/**
3335 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3336 *
3337 * @skb: GSO skb
3338 *
3339 * skb_gso_network_seglen is used to determine the real size of the
3340 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3341 *
3342 * The MAC/L2 header is not accounted for.
3343 */
3344static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3345{
3346 unsigned int hdr_len = skb_transport_header(skb) -
3347 skb_network_header(skb);
3348 return hdr_len + skb_gso_transport_seglen(skb);
3349}
1da177e4
LT
3350#endif /* __KERNEL__ */
3351#endif /* _LINUX_SKBUFF_H */
This page took 2.165461 seconds and 5 git commands to generate.