devlink: Fix port_type_set function pointer check
[linux-block.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
8842d285 17#include <linux/bvec.h>
1da177e4 18#include <linux/cache.h>
56b17425 19#include <linux/rbtree.h>
51f3d02b 20#include <linux/socket.h>
c1d1b437 21#include <linux/refcount.h>
1da177e4 22
60063497 23#include <linux/atomic.h>
1da177e4
LT
24#include <asm/types.h>
25#include <linux/spinlock.h>
1da177e4 26#include <linux/net.h>
3fc7e8a6 27#include <linux/textsearch.h>
1da177e4 28#include <net/checksum.h>
a80958f4 29#include <linux/rcupdate.h>
b7aa0bf7 30#include <linux/hrtimer.h>
131ea667 31#include <linux/dma-mapping.h>
c8f44aff 32#include <linux/netdev_features.h>
363ec392 33#include <linux/sched.h>
e6017571 34#include <linux/sched/clock.h>
1bd758eb 35#include <net/flow_dissector.h>
a60e3cc7 36#include <linux/splice.h>
72b31f72 37#include <linux/in6.h>
8b10cab6 38#include <linux/if_packet.h>
f70ea018 39#include <net/flow.h>
6a5bcd84 40#include <net/page_pool.h>
261db6c2
JS
41#if IS_ENABLED(CONFIG_NF_CONNTRACK)
42#include <linux/netfilter/nf_conntrack_common.h>
43#endif
1da177e4 44
7a6ae71b
TH
45/* The interface for checksum offload between the stack and networking drivers
46 * is as follows...
47 *
48 * A. IP checksum related features
49 *
50 * Drivers advertise checksum offload capabilities in the features of a device.
db1f00fb
DC
51 * From the stack's point of view these are capabilities offered by the driver.
52 * A driver typically only advertises features that it is capable of offloading
7a6ae71b
TH
53 * to its device.
54 *
55 * The checksum related features are:
56 *
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
62 *
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
db1f00fb 67 * is TCP or UDP. The IPv4 header may contain IP options.
7a6ae71b
TH
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
71 *
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
645f0897 75 * IPv6|UDP where the Next Header field in the IPv6
7a6ae71b
TH
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
81 *
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
db1f00fb 83 * This flag is only used to disable the RX checksum
7a6ae71b
TH
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
87 *
88 * B. Checksumming of received packets by device. Indication of checksum
db1f00fb 89 * verification is set in skb->ip_summed. Possible values are:
78ea85f1
DB
90 *
91 * CHECKSUM_NONE:
92 *
7a6ae71b 93 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
96 *
97 * CHECKSUM_UNNECESSARY:
98 *
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
77cffe23
TH
105 *
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
b4759dcd 113 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
114 *
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
db1f00fb 119 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
77cffe23 120 * two. If the device were only able to verify the UDP checksum and not
db1f00fb 121 * GRE, either because it doesn't support GRE checksum or because GRE
77cffe23
TH
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
78ea85f1
DB
124 *
125 * CHECKSUM_COMPLETE:
126 *
127 * This is the most generic way. The device supplied checksum of the _whole_
db1f00fb 128 * packet as seen by netif_rx() and fills in skb->csum. This means the
78ea85f1
DB
129 * hardware doesn't need to parse L3/L4 headers to implement this.
130 *
b4759dcd
DC
131 * Notes:
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
135 *
136 * CHECKSUM_PARTIAL:
137 *
6edec0e6
TH
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
146 * be verified.
78ea85f1 147 *
7a6ae71b
TH
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
150 *
151 * CHECKSUM_PARTIAL:
152 *
7a6ae71b 153 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 154 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
db1f00fb
DC
157 * offset of the packet, but it should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum -- it is the
7a6ae71b
TH
159 * purview of the stack to validate that csum_start and csum_offset are set
160 * correctly.
161 *
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
166 *
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
78ea85f1 174 *
7a6ae71b 175 * CHECKSUM_NONE:
78ea85f1 176 *
7a6ae71b
TH
177 * The skb was already checksummed by the protocol, or a checksum is not
178 * required.
78ea85f1
DB
179 *
180 * CHECKSUM_UNNECESSARY:
181 *
db1f00fb 182 * This has the same meaning as CHECKSUM_NONE for checksum offload on
7a6ae71b 183 * output.
78ea85f1 184 *
7a6ae71b
TH
185 * CHECKSUM_COMPLETE:
186 * Not used in checksum output. If a driver observes a packet with this value
db1f00fb 187 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
7a6ae71b
TH
188 *
189 * D. Non-IP checksum (CRC) offloads
190 *
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
db1f00fb 193 * will set csum_start and csum_offset accordingly, set ip_summed to
dba00306
DC
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
200 *
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
db1f00fb
DC
204 * accordingly. Note that there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
7a6ae71b 206 * both IP checksum offload and FCOE CRC offload must verify which offload
db1f00fb 207 * is configured for a packet, presumably by inspecting packet headers.
7a6ae71b
TH
208 *
209 * E. Checksumming on output with GSO.
210 *
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
db1f00fb
DC
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
216 * csum_offset are set to refer to the outermost checksum being offloaded
217 * (two offloaded checksums are possible with UDP encapsulation).
78ea85f1
DB
218 */
219
60476372 220/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
221#define CHECKSUM_NONE 0
222#define CHECKSUM_UNNECESSARY 1
223#define CHECKSUM_COMPLETE 2
224#define CHECKSUM_PARTIAL 3
1da177e4 225
77cffe23
TH
226/* Maximum value in skb->csum_level */
227#define SKB_MAX_CSUM_LEVEL 3
228
0bec8c88 229#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 230#define SKB_WITH_OVERHEAD(X) \
deea84b0 231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
232#define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
234#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
236
87fb4b7b
ED
237/* return minimum truesize of one skb containing X bytes of data */
238#define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
241
7999096f 242struct ahash_request;
1da177e4 243struct net_device;
716ea3a7 244struct scatterlist;
9c55e01c 245struct pipe_inode_info;
a8f820aa 246struct iov_iter;
fd11a83d 247struct napi_struct;
d58e468b
PP
248struct bpf_prog;
249union bpf_attr;
df5042f4 250struct skb_ext;
1da177e4 251
34666d46 252#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 253struct nf_bridge_info {
3eaf4025
FW
254 enum {
255 BRNF_PROTO_UNCHANGED,
256 BRNF_PROTO_8021Q,
257 BRNF_PROTO_PPPOE
7fb48c5b 258 } orig_proto:8;
72b1e5e4
FW
259 u8 pkt_otherhost:1;
260 u8 in_prerouting:1;
261 u8 bridged_dnat:1;
411ffb4f 262 __u16 frag_max_size;
bf1ac5ca 263 struct net_device *physindev;
63cdbc06
FW
264
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
7fb48c5b 267 union {
72b1e5e4 268 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
269 __be32 ipv4_daddr;
270 struct in6_addr ipv6_daddr;
72b1e5e4
FW
271
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
275 */
276 char neigh_header[8];
72b31f72 277 };
1da177e4
LT
278};
279#endif
280
95a7233c
PB
281#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
282/* Chain in tc_skb_ext will be used to share the tc chain with
283 * ovs recirc_id. It will be set to the current chain by tc
284 * and read by ovs to recirc_id.
285 */
286struct tc_skb_ext {
287 __u32 chain;
038ebb1a 288 __u16 mru;
d29334c1 289 bool post_ct;
95a7233c
PB
290};
291#endif
292
1da177e4
LT
293struct sk_buff_head {
294 /* These two members must be first. */
295 struct sk_buff *next;
296 struct sk_buff *prev;
297
298 __u32 qlen;
299 spinlock_t lock;
300};
301
302struct sk_buff;
303
9d4dde52
IC
304/* To allow 64K frame to be packed as single skb without frag_list we
305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
306 * buffers which do not start on a page boundary.
307 *
308 * Since GRO uses frags we allocate at least 16 regardless of page
309 * size.
a715dea3 310 */
9d4dde52 311#if (65536/PAGE_SIZE + 1) < 16
eec00954 312#define MAX_SKB_FRAGS 16UL
a715dea3 313#else
9d4dde52 314#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 315#endif
5f74f82e 316extern int sysctl_max_skb_frags;
1da177e4 317
3953c46c
MRL
318/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
319 * segment using its current segmentation instead.
320 */
321#define GSO_BY_FRAGS 0xFFFF
322
8842d285 323typedef struct bio_vec skb_frag_t;
1da177e4 324
161e6137 325/**
7240b60c 326 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
327 * @frag: skb fragment
328 */
9e903e08
ED
329static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330{
b8b576a1 331 return frag->bv_len;
9e903e08
ED
332}
333
161e6137 334/**
7240b60c 335 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
336 * @frag: skb fragment
337 * @size: size of fragment
338 */
9e903e08
ED
339static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
340{
b8b576a1 341 frag->bv_len = size;
9e903e08
ED
342}
343
161e6137 344/**
7240b60c 345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
346 * @frag: skb fragment
347 * @delta: value to add
348 */
9e903e08
ED
349static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
350{
b8b576a1 351 frag->bv_len += delta;
9e903e08
ED
352}
353
161e6137 354/**
7240b60c 355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
356 * @frag: skb fragment
357 * @delta: value to subtract
358 */
9e903e08
ED
359static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
360{
b8b576a1 361 frag->bv_len -= delta;
9e903e08
ED
362}
363
161e6137
PT
364/**
365 * skb_frag_must_loop - Test if %p is a high memory page
366 * @p: fragment's page
367 */
c613c209
WB
368static inline bool skb_frag_must_loop(struct page *p)
369{
370#if defined(CONFIG_HIGHMEM)
29766bcf 371 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
372 return true;
373#endif
374 return false;
375}
376
377/**
378 * skb_frag_foreach_page - loop over pages in a fragment
379 *
380 * @f: skb frag to operate on
1dfa5bd3 381 * @f_off: offset from start of f->bv_page
c613c209
WB
382 * @f_len: length from f_off to loop over
383 * @p: (temp var) current page
384 * @p_off: (temp var) offset from start of current page,
385 * non-zero only on first page.
386 * @p_len: (temp var) length in current page,
387 * < PAGE_SIZE only on first and last page.
388 * @copied: (temp var) length so far, excluding current p_len.
389 *
390 * A fragment can hold a compound page, in which case per-page
391 * operations, notably kmap_atomic, must be called for each
392 * regular page.
393 */
394#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
396 p_off = (f_off) & (PAGE_SIZE - 1), \
397 p_len = skb_frag_must_loop(p) ? \
398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
399 copied = 0; \
400 copied < f_len; \
401 copied += p_len, p++, p_off = 0, \
402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
403
ac45f602
PO
404#define HAVE_HW_TIME_STAMP
405
406/**
d3a21be8 407 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
408 * @hwtstamp: hardware time stamp transformed into duration
409 * since arbitrary point in time
ac45f602
PO
410 *
411 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 412 * skb->tstamp.
ac45f602
PO
413 *
414 * hwtstamps can only be compared against other hwtstamps from
415 * the same device.
416 *
417 * This structure is attached to packets as part of the
418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
419 */
420struct skb_shared_hwtstamps {
421 ktime_t hwtstamp;
ac45f602
PO
422};
423
2244d07b
OH
424/* Definitions for tx_flags in struct skb_shared_info */
425enum {
426 /* generate hardware time stamp */
427 SKBTX_HW_TSTAMP = 1 << 0,
428
e7fd2885 429 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
430 SKBTX_SW_TSTAMP = 1 << 1,
431
432 /* device driver is going to provide hardware time stamp */
433 SKBTX_IN_PROGRESS = 1 << 2,
434
6e3e939f 435 /* generate wifi status information (where possible) */
62b1a8ab 436 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 437
e7fd2885
WB
438 /* generate software time stamp when entering packet scheduling */
439 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
440};
441
e1c8a607 442#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 443 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
444#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
445
06b4feb3
JL
446/* Definitions for flags in struct skb_shared_info */
447enum {
448 /* use zcopy routines */
449 SKBFL_ZEROCOPY_ENABLE = BIT(0),
450
451 /* This indicates at least one fragment might be overwritten
452 * (as in vmsplice(), sendfile() ...)
453 * If we need to compute a TX checksum, we'll need to copy
454 * all frags to avoid possible bad checksum
455 */
456 SKBFL_SHARED_FRAG = BIT(1),
457};
458
459#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
460
a6686f2f
SM
461/*
462 * The callback notifies userspace to release buffers when skb DMA is done in
463 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
464 * The zerocopy_success argument is true if zero copy transmit occurred,
465 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
466 * The ctx field is used to track device context.
467 * The desc field is used to track userspace buffer index.
a6686f2f
SM
468 */
469struct ubuf_info {
36177832
JL
470 void (*callback)(struct sk_buff *, struct ubuf_info *,
471 bool zerocopy_success);
4ab6c99d
WB
472 union {
473 struct {
474 unsigned long desc;
475 void *ctx;
476 };
477 struct {
478 u32 id;
479 u16 len;
480 u16 zerocopy:1;
481 u32 bytelen;
482 };
483 };
c1d1b437 484 refcount_t refcnt;
04c2d33e 485 u8 flags;
a91dbff5
WB
486
487 struct mmpin {
488 struct user_struct *user;
489 unsigned int num_pg;
490 } mmp;
ac45f602
PO
491};
492
52267790
WB
493#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
494
6f89dbce
SV
495int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
496void mm_unaccount_pinned_pages(struct mmpin *mmp);
497
8c793822
JL
498struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
499struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
500 struct ubuf_info *uarg);
52267790 501
8c793822 502void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790 503
8c793822
JL
504void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
505 bool success);
52267790 506
b5947e5d 507int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
508int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
509 struct msghdr *msg, int len,
510 struct ubuf_info *uarg);
511
1da177e4
LT
512/* This data is invariant across clones and lives at
513 * the end of the header data, ie. at skb->end.
514 */
515struct skb_shared_info {
06b4feb3 516 __u8 flags;
de8f3a83
DB
517 __u8 meta_len;
518 __u8 nr_frags;
9f42f126 519 __u8 tx_flags;
7967168c
HX
520 unsigned short gso_size;
521 /* Warning: this field is not always filled in (UFO)! */
522 unsigned short gso_segs;
1da177e4 523 struct sk_buff *frag_list;
ac45f602 524 struct skb_shared_hwtstamps hwtstamps;
7f564528 525 unsigned int gso_type;
09c2d251 526 u32 tskey;
ec7d2f2c
ED
527
528 /*
529 * Warning : all fields before dataref are cleared in __alloc_skb()
530 */
531 atomic_t dataref;
532
69e3c75f
JB
533 /* Intermediate layers must ensure that destructor_arg
534 * remains valid until skb destructor */
535 void * destructor_arg;
a6686f2f 536
fed66381
ED
537 /* must be last field, see pskb_expand_head() */
538 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
539};
540
541/* We divide dataref into two halves. The higher 16 bits hold references
542 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
543 * the entire skb->data. A clone of a headerless skb holds the length of
544 * the header in skb->hdr_len.
1da177e4
LT
545 *
546 * All users must obey the rule that the skb->data reference count must be
547 * greater than or equal to the payload reference count.
548 *
549 * Holding a reference to the payload part means that the user does not
550 * care about modifications to the header part of skb->data.
551 */
552#define SKB_DATAREF_SHIFT 16
553#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
554
d179cd12
DM
555
556enum {
c8753d55
VS
557 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
558 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
559 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
560};
561
7967168c
HX
562enum {
563 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
564
565 /* This indicates the skb is from an untrusted source. */
d9d30adf 566 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
567
568 /* This indicates the tcp segment has CWR set. */
d9d30adf 569 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 570
d9d30adf 571 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 572
d9d30adf 573 SKB_GSO_TCPV6 = 1 << 4,
68c33163 574
d9d30adf 575 SKB_GSO_FCOE = 1 << 5,
73136267 576
d9d30adf 577 SKB_GSO_GRE = 1 << 6,
0d89d203 578
d9d30adf 579 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 580
d9d30adf 581 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 582
d9d30adf 583 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 584
d9d30adf 585 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 586
d9d30adf 587 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 588
d9d30adf 589 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 590
d9d30adf 591 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 592
d9d30adf 593 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 594
d9d30adf 595 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
596
597 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
598
599 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
600
601 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
602};
603
2e07fa9c
ACM
604#if BITS_PER_LONG > 32
605#define NET_SKBUFF_DATA_USES_OFFSET 1
606#endif
607
608#ifdef NET_SKBUFF_DATA_USES_OFFSET
609typedef unsigned int sk_buff_data_t;
610#else
611typedef unsigned char *sk_buff_data_t;
612#endif
613
161e6137 614/**
1da177e4
LT
615 * struct sk_buff - socket buffer
616 * @next: Next buffer in list
617 * @prev: Previous buffer in list
363ec392 618 * @tstamp: Time we arrived/left
d2f273f0
RD
619 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
620 * for retransmit timer
56b17425 621 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 622 * @list: queue head
d84e0bd7 623 * @sk: Socket we are owned by
d2f273f0
RD
624 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
625 * fragmentation management
1da177e4 626 * @dev: Device we arrived on/are leaving by
d2f273f0 627 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 628 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 629 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 630 * @sp: the security path, used for xfrm
1da177e4
LT
631 * @len: Length of actual data
632 * @data_len: Data length
633 * @mac_len: Length of link layer header
334a8132 634 * @hdr_len: writable header length of cloned skb
663ead3b
HX
635 * @csum: Checksum (must include start/offset pair)
636 * @csum_start: Offset from skb->head where checksumming should start
637 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 638 * @priority: Packet queueing priority
60ff7467 639 * @ignore_df: allow local fragmentation
1da177e4 640 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 641 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
642 * @nohdr: Payload reference only, must not modify header
643 * @pkt_type: Packet class
c83c2486 644 * @fclone: skbuff clone status
c83c2486 645 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
646 * @inner_protocol_type: whether the inner protocol is
647 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
648 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
649 * @offload_fwd_mark: Packet was L2-forwarded in hardware
650 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 651 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 652 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
653 * @redirected: packet was redirected by packet classifier
654 * @from_ingress: packet was redirected from the ingress path
31729363
RD
655 * @peeked: this packet has been seen already, so stats have been
656 * done for it, don't do them again
ba9dda3a 657 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
658 * @protocol: Packet protocol from driver
659 * @destructor: Destruct function
e2080072 660 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
6ed6e1c7 661 * @_sk_redir: socket redirection information for skmsg
a9e419dc 662 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 663 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 664 * @skb_iif: ifindex of device we arrived on
1da177e4 665 * @tc_index: Traffic control index
61b905da 666 * @hash: the packet hash
d84e0bd7 667 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
668 * @head_frag: skb was allocated from page fragments,
669 * not allocated by kmalloc() or vmalloc().
8b700862 670 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
6a5bcd84
IA
671 * @pp_recycle: mark the packet for recycling instead of freeing (implies
672 * page_pool support on driver)
df5042f4 673 * @active_extensions: active extensions (skb_ext_id types)
553a5672 674 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 675 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 676 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 677 * ports.
a3b18ddb 678 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
679 * @wifi_acked_valid: wifi_acked was set
680 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 681 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
682 * @encapsulation: indicates the inner headers in the skbuff are valid
683 * @encap_hdr_csum: software checksum is needed
684 * @csum_valid: checksum is already valid
dba00306 685 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
686 * @csum_complete_sw: checksum was completed by software
687 * @csum_level: indicates the number of consecutive checksums found in
688 * the packet minus one that have been verified as
689 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 690 * @dst_pending_confirm: need to confirm neighbour
a48d189e 691 * @decrypted: Decrypted SKB
5fc88f93 692 * @slow_gro: state present at GRO time, slower prepare step required
161e6137 693 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 694 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 695 * @secmark: security marking
d84e0bd7 696 * @mark: Generic packet mark
d2f273f0
RD
697 * @reserved_tailroom: (aka @mark) number of bytes of free space available
698 * at the tail of an sk_buff
699 * @vlan_present: VLAN tag is present
86a9bad3 700 * @vlan_proto: vlan encapsulation protocol
6aa895b0 701 * @vlan_tci: vlan tag control information
0d89d203 702 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
703 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
704 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
705 * @inner_transport_header: Inner transport layer header (encapsulation)
706 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 707 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
708 * @transport_header: Transport layer header
709 * @network_header: Network layer header
710 * @mac_header: Link layer header
fa69ee5a 711 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
712 * @tail: Tail pointer
713 * @end: End pointer
714 * @head: Head of buffer
715 * @data: Data head pointer
716 * @truesize: Buffer size
717 * @users: User count - see {datagram,tcp}.c
df5042f4 718 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
719 */
720
721struct sk_buff {
363ec392 722 union {
56b17425
ED
723 struct {
724 /* These two members must be first. */
725 struct sk_buff *next;
726 struct sk_buff *prev;
727
728 union {
bffa72cf
ED
729 struct net_device *dev;
730 /* Some protocols might use this space to store information,
731 * while device pointer would be NULL.
732 * UDP receive path is one user.
733 */
734 unsigned long dev_scratch;
56b17425
ED
735 };
736 };
fa0f5273 737 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 738 struct list_head list;
363ec392 739 };
fa0f5273
PO
740
741 union {
742 struct sock *sk;
743 int ip_defrag_offset;
744 };
1da177e4 745
c84d9490 746 union {
bffa72cf 747 ktime_t tstamp;
d3edd06e 748 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 749 };
1da177e4
LT
750 /*
751 * This is the control buffer. It is free to use for every
752 * layer. Please put your private variables there. If you
753 * want to keep them across layers you have to do a skb_clone()
754 * first. This is owned by whoever has the skb queued ATM.
755 */
da3f5cf1 756 char cb[48] __aligned(8);
1da177e4 757
e2080072
ED
758 union {
759 struct {
760 unsigned long _skb_refdst;
761 void (*destructor)(struct sk_buff *skb);
762 };
763 struct list_head tcp_tsorted_anchor;
e3526bb9
CW
764#ifdef CONFIG_NET_SOCK_MSG
765 unsigned long _sk_redir;
766#endif
e2080072
ED
767 };
768
b1937227 769#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 770 unsigned long _nfct;
da3f5cf1 771#endif
1da177e4 772 unsigned int len,
334a8132
PM
773 data_len;
774 __u16 mac_len,
775 hdr_len;
b1937227
ED
776
777 /* Following fields are _not_ copied in __copy_skb_header()
778 * Note that queue_mapping is here mostly to fill a hole.
779 */
b1937227 780 __u16 queue_mapping;
36bbef52
DB
781
782/* if you move cloned around you also must adapt those constants */
783#ifdef __BIG_ENDIAN_BITFIELD
784#define CLONED_MASK (1 << 7)
785#else
786#define CLONED_MASK 1
787#endif
788#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
789
d2f273f0 790 /* private: */
36bbef52 791 __u8 __cloned_offset[0];
d2f273f0 792 /* public: */
b1937227 793 __u8 cloned:1,
6869c4d8 794 nohdr:1,
b84f4cc9 795 fclone:2,
a59322be 796 peeked:1,
b1937227 797 head_frag:1,
6a5bcd84
IA
798 pfmemalloc:1,
799 pp_recycle:1; /* page_pool recycle indicator */
df5042f4
FW
800#ifdef CONFIG_SKB_EXTENSIONS
801 __u8 active_extensions;
802#endif
6a5bcd84 803
b1937227
ED
804 /* fields enclosed in headers_start/headers_end are copied
805 * using a single memcpy() in __copy_skb_header()
806 */
ebcf34f3 807 /* private: */
b1937227 808 __u32 headers_start[0];
ebcf34f3 809 /* public: */
4031ae6e 810
233577a2
HFS
811/* if you move pkt_type around you also must adapt those constants */
812#ifdef __BIG_ENDIAN_BITFIELD
813#define PKT_TYPE_MAX (7 << 5)
814#else
815#define PKT_TYPE_MAX 7
1da177e4 816#endif
233577a2 817#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 818
d2f273f0 819 /* private: */
233577a2 820 __u8 __pkt_type_offset[0];
d2f273f0 821 /* public: */
b1937227 822 __u8 pkt_type:3;
b1937227 823 __u8 ignore_df:1;
b1937227
ED
824 __u8 nf_trace:1;
825 __u8 ip_summed:2;
3853b584 826 __u8 ooo_okay:1;
8b700862 827
61b905da 828 __u8 l4_hash:1;
a3b18ddb 829 __u8 sw_hash:1;
6e3e939f
JB
830 __u8 wifi_acked_valid:1;
831 __u8 wifi_acked:1;
3bdc0eba 832 __u8 no_fcs:1;
77cffe23 833 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 834 __u8 encapsulation:1;
7e2b10c1 835 __u8 encap_hdr_csum:1;
5d0c2b95 836 __u8 csum_valid:1;
8b700862 837
0c4b2d37
MM
838#ifdef __BIG_ENDIAN_BITFIELD
839#define PKT_VLAN_PRESENT_BIT 7
840#else
841#define PKT_VLAN_PRESENT_BIT 0
842#endif
843#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 844 /* private: */
0c4b2d37 845 __u8 __pkt_vlan_present_offset[0];
d2f273f0 846 /* public: */
0c4b2d37 847 __u8 vlan_present:1;
7e3cead5 848 __u8 csum_complete_sw:1;
b1937227 849 __u8 csum_level:2;
dba00306 850 __u8 csum_not_inet:1;
4ff06203 851 __u8 dst_pending_confirm:1;
b1937227
ED
852#ifdef CONFIG_IPV6_NDISC_NODETYPE
853 __u8 ndisc_nodetype:2;
854#endif
8b700862 855
0c4b2d37 856 __u8 ipvs_property:1;
8bce6d7d 857 __u8 inner_protocol_type:1;
e585f236 858 __u8 remcsum_offload:1;
6bc506b4
IS
859#ifdef CONFIG_NET_SWITCHDEV
860 __u8 offload_fwd_mark:1;
875e8939 861 __u8 offload_l3_fwd_mark:1;
6bc506b4 862#endif
e7246e12
WB
863#ifdef CONFIG_NET_CLS_ACT
864 __u8 tc_skip_classify:1;
8dc07fdb 865 __u8 tc_at_ingress:1;
2c64605b 866#endif
2c64605b 867 __u8 redirected:1;
11941f8a 868#ifdef CONFIG_NET_REDIRECT
2c64605b 869 __u8 from_ingress:1;
e7246e12 870#endif
a48d189e
SB
871#ifdef CONFIG_TLS_DEVICE
872 __u8 decrypted:1;
873#endif
5fc88f93 874 __u8 slow_gro:1;
b1937227
ED
875
876#ifdef CONFIG_NET_SCHED
877 __u16 tc_index; /* traffic control index */
b1937227 878#endif
fe55f6d5 879
b1937227
ED
880 union {
881 __wsum csum;
882 struct {
883 __u16 csum_start;
884 __u16 csum_offset;
885 };
886 };
887 __u32 priority;
888 int skb_iif;
889 __u32 hash;
890 __be16 vlan_proto;
891 __u16 vlan_tci;
2bd82484
ED
892#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
893 union {
894 unsigned int napi_id;
895 unsigned int sender_cpu;
896 };
97fc2f08 897#endif
984bc16c 898#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 899 __u32 secmark;
0c4f691f 900#endif
0c4f691f 901
3b885787
NH
902 union {
903 __u32 mark;
16fad69c 904 __u32 reserved_tailroom;
3b885787 905 };
1da177e4 906
8bce6d7d
TH
907 union {
908 __be16 inner_protocol;
909 __u8 inner_ipproto;
910 };
911
1a37e412
SH
912 __u16 inner_transport_header;
913 __u16 inner_network_header;
914 __u16 inner_mac_header;
b1937227
ED
915
916 __be16 protocol;
1a37e412
SH
917 __u16 transport_header;
918 __u16 network_header;
919 __u16 mac_header;
b1937227 920
fa69ee5a
ME
921#ifdef CONFIG_KCOV
922 u64 kcov_handle;
923#endif
924
ebcf34f3 925 /* private: */
b1937227 926 __u32 headers_end[0];
ebcf34f3 927 /* public: */
b1937227 928
1da177e4 929 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 930 sk_buff_data_t tail;
4305b541 931 sk_buff_data_t end;
1da177e4 932 unsigned char *head,
4305b541 933 *data;
27a884dc 934 unsigned int truesize;
63354797 935 refcount_t users;
df5042f4
FW
936
937#ifdef CONFIG_SKB_EXTENSIONS
938 /* only useable after checking ->active_extensions != 0 */
939 struct skb_ext *extensions;
940#endif
1da177e4
LT
941};
942
943#ifdef __KERNEL__
944/*
945 * Handling routines are only of interest to the kernel
946 */
1da177e4 947
c93bdd0e
MG
948#define SKB_ALLOC_FCLONE 0x01
949#define SKB_ALLOC_RX 0x02
fd11a83d 950#define SKB_ALLOC_NAPI 0x04
c93bdd0e 951
161e6137
PT
952/**
953 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
954 * @skb: buffer
955 */
c93bdd0e
MG
956static inline bool skb_pfmemalloc(const struct sk_buff *skb)
957{
958 return unlikely(skb->pfmemalloc);
959}
960
7fee226a
ED
961/*
962 * skb might have a dst pointer attached, refcounted or not.
963 * _skb_refdst low order bit is set if refcount was _not_ taken
964 */
965#define SKB_DST_NOREF 1UL
966#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
967
968/**
969 * skb_dst - returns skb dst_entry
970 * @skb: buffer
971 *
972 * Returns skb dst_entry, regardless of reference taken or not.
973 */
adf30907
ED
974static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
975{
161e6137 976 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
977 * rcu_read_lock section
978 */
979 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
980 !rcu_read_lock_held() &&
981 !rcu_read_lock_bh_held());
982 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
983}
984
7fee226a
ED
985/**
986 * skb_dst_set - sets skb dst
987 * @skb: buffer
988 * @dst: dst entry
989 *
990 * Sets skb dst, assuming a reference was taken on dst and should
991 * be released by skb_dst_drop()
992 */
adf30907
ED
993static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
994{
8a886b14 995 skb->slow_gro |= !!dst;
7fee226a
ED
996 skb->_skb_refdst = (unsigned long)dst;
997}
998
932bc4d7
JA
999/**
1000 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1001 * @skb: buffer
1002 * @dst: dst entry
1003 *
1004 * Sets skb dst, assuming a reference was not taken on dst.
1005 * If dst entry is cached, we do not take reference and dst_release
1006 * will be avoided by refdst_drop. If dst entry is not cached, we take
1007 * reference, so that last dst_release can destroy the dst immediately.
1008 */
1009static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1010{
dbfc4fb7 1011 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
a432934a 1012 skb->slow_gro |= !!dst;
dbfc4fb7 1013 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1014}
7fee226a
ED
1015
1016/**
25985edc 1017 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1018 * @skb: buffer
1019 */
1020static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1021{
1022 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1023}
1024
161e6137
PT
1025/**
1026 * skb_rtable - Returns the skb &rtable
1027 * @skb: buffer
1028 */
511c3f92
ED
1029static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1030{
adf30907 1031 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1032}
1033
8b10cab6
JHS
1034/* For mangling skb->pkt_type from user space side from applications
1035 * such as nft, tc, etc, we only allow a conservative subset of
1036 * possible pkt_types to be set.
1037*/
1038static inline bool skb_pkt_type_ok(u32 ptype)
1039{
1040 return ptype <= PACKET_OTHERHOST;
1041}
1042
161e6137
PT
1043/**
1044 * skb_napi_id - Returns the skb's NAPI id
1045 * @skb: buffer
1046 */
90b602f8
ML
1047static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1048{
1049#ifdef CONFIG_NET_RX_BUSY_POLL
1050 return skb->napi_id;
1051#else
1052 return 0;
1053#endif
1054}
1055
161e6137
PT
1056/**
1057 * skb_unref - decrement the skb's reference count
1058 * @skb: buffer
1059 *
1060 * Returns true if we can free the skb.
1061 */
3889a803
PA
1062static inline bool skb_unref(struct sk_buff *skb)
1063{
1064 if (unlikely(!skb))
1065 return false;
63354797 1066 if (likely(refcount_read(&skb->users) == 1))
3889a803 1067 smp_rmb();
63354797 1068 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1069 return false;
1070
1071 return true;
1072}
1073
0a463c78 1074void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1075void kfree_skb(struct sk_buff *skb);
1076void kfree_skb_list(struct sk_buff *segs);
6413139d 1077void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1078void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1079
1080#ifdef CONFIG_TRACEPOINTS
7965bd4d 1081void consume_skb(struct sk_buff *skb);
be769db2
HX
1082#else
1083static inline void consume_skb(struct sk_buff *skb)
1084{
1085 return kfree_skb(skb);
1086}
1087#endif
1088
ca2c1418 1089void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1090void __kfree_skb(struct sk_buff *skb);
d7e8883c 1091extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1092
7965bd4d
JP
1093void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1094bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1095 bool *fragstolen, int *delta_truesize);
bad43ca8 1096
7965bd4d
JP
1097struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1098 int node);
2ea2f62c 1099struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1100struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1101struct sk_buff *build_skb_around(struct sk_buff *skb,
1102 void *data, unsigned int frag_size);
161e6137 1103
f450d539
AL
1104struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1105
161e6137
PT
1106/**
1107 * alloc_skb - allocate a network buffer
1108 * @size: size to allocate
1109 * @priority: allocation mask
1110 *
1111 * This function is a convenient wrapper around __alloc_skb().
1112 */
d179cd12 1113static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1114 gfp_t priority)
d179cd12 1115{
564824b0 1116 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1117}
1118
2e4e4410
ED
1119struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1120 unsigned long data_len,
1121 int max_page_order,
1122 int *errcode,
1123 gfp_t gfp_mask);
da29e4b4 1124struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1125
d0bf4a9e
ED
1126/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1127struct sk_buff_fclones {
1128 struct sk_buff skb1;
1129
1130 struct sk_buff skb2;
1131
2638595a 1132 refcount_t fclone_ref;
d0bf4a9e
ED
1133};
1134
1135/**
1136 * skb_fclone_busy - check if fclone is busy
293de7de 1137 * @sk: socket
d0bf4a9e
ED
1138 * @skb: buffer
1139 *
bda13fed 1140 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1141 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1142 * so we also check that this didnt happen.
d0bf4a9e 1143 */
39bb5e62
ED
1144static inline bool skb_fclone_busy(const struct sock *sk,
1145 const struct sk_buff *skb)
d0bf4a9e
ED
1146{
1147 const struct sk_buff_fclones *fclones;
1148
1149 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1150
1151 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1152 refcount_read(&fclones->fclone_ref) > 1 &&
f4dae54e 1153 READ_ONCE(fclones->skb2.sk) == sk;
d0bf4a9e
ED
1154}
1155
161e6137
PT
1156/**
1157 * alloc_skb_fclone - allocate a network buffer from fclone cache
1158 * @size: size to allocate
1159 * @priority: allocation mask
1160 *
1161 * This function is a convenient wrapper around __alloc_skb().
1162 */
d179cd12 1163static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1164 gfp_t priority)
d179cd12 1165{
c93bdd0e 1166 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1167}
1168
7965bd4d 1169struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1170void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1171int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1172struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1173void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1174struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1175struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1176 gfp_t gfp_mask, bool fclone);
1177static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1178 gfp_t gfp_mask)
1179{
1180 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1181}
7965bd4d
JP
1182
1183int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1184struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1185 unsigned int headroom);
f1260ff1 1186struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
7965bd4d
JP
1187struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1188 int newtailroom, gfp_t priority);
48a1df65
JD
1189int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1190 int offset, int len);
1191int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1192 int offset, int len);
7965bd4d 1193int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1194int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1195
1196/**
1197 * skb_pad - zero pad the tail of an skb
1198 * @skb: buffer to pad
1199 * @pad: space to pad
1200 *
1201 * Ensure that a buffer is followed by a padding area that is zero
1202 * filled. Used by network drivers which may DMA or transfer data
1203 * beyond the buffer end onto the wire.
1204 *
1205 * May return error in out of memory cases. The skb is freed on error.
1206 */
1207static inline int skb_pad(struct sk_buff *skb, int pad)
1208{
1209 return __skb_pad(skb, pad, true);
1210}
ead2ceb0 1211#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1212
be12a1fe
HFS
1213int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1214 int offset, size_t size);
1215
d94d9fee 1216struct skb_seq_state {
677e90ed
TG
1217 __u32 lower_offset;
1218 __u32 upper_offset;
1219 __u32 frag_idx;
1220 __u32 stepped_offset;
1221 struct sk_buff *root_skb;
1222 struct sk_buff *cur_skb;
1223 __u8 *frag_data;
97550f6f 1224 __u32 frag_off;
677e90ed
TG
1225};
1226
7965bd4d
JP
1227void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1228 unsigned int to, struct skb_seq_state *st);
1229unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1230 struct skb_seq_state *st);
1231void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1232
7965bd4d 1233unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1234 unsigned int to, struct ts_config *config);
3fc7e8a6 1235
09323cc4
TH
1236/*
1237 * Packet hash types specify the type of hash in skb_set_hash.
1238 *
1239 * Hash types refer to the protocol layer addresses which are used to
1240 * construct a packet's hash. The hashes are used to differentiate or identify
1241 * flows of the protocol layer for the hash type. Hash types are either
1242 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1243 *
1244 * Properties of hashes:
1245 *
1246 * 1) Two packets in different flows have different hash values
1247 * 2) Two packets in the same flow should have the same hash value
1248 *
1249 * A hash at a higher layer is considered to be more specific. A driver should
1250 * set the most specific hash possible.
1251 *
1252 * A driver cannot indicate a more specific hash than the layer at which a hash
1253 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1254 *
1255 * A driver may indicate a hash level which is less specific than the
1256 * actual layer the hash was computed on. For instance, a hash computed
1257 * at L4 may be considered an L3 hash. This should only be done if the
1258 * driver can't unambiguously determine that the HW computed the hash at
1259 * the higher layer. Note that the "should" in the second property above
1260 * permits this.
1261 */
1262enum pkt_hash_types {
1263 PKT_HASH_TYPE_NONE, /* Undefined type */
1264 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1265 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1266 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1267};
1268
bcc83839 1269static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1270{
bcc83839 1271 skb->hash = 0;
a3b18ddb 1272 skb->sw_hash = 0;
bcc83839
TH
1273 skb->l4_hash = 0;
1274}
1275
1276static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1277{
1278 if (!skb->l4_hash)
1279 skb_clear_hash(skb);
1280}
1281
1282static inline void
1283__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1284{
1285 skb->l4_hash = is_l4;
1286 skb->sw_hash = is_sw;
61b905da 1287 skb->hash = hash;
09323cc4
TH
1288}
1289
bcc83839
TH
1290static inline void
1291skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1292{
1293 /* Used by drivers to set hash from HW */
1294 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1295}
1296
1297static inline void
1298__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1299{
1300 __skb_set_hash(skb, hash, true, is_l4);
1301}
1302
e5276937 1303void __skb_get_hash(struct sk_buff *skb);
b917783c 1304u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937 1305u32 skb_get_poff(const struct sk_buff *skb);
f96533cd 1306u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
72a338bc 1307 const struct flow_keys_basic *keys, int hlen);
e5276937 1308__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
f96533cd 1309 const void *data, int hlen_proto);
e5276937
TH
1310
1311static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1312 int thoff, u8 ip_proto)
1313{
1314 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1315}
1316
1317void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1318 const struct flow_dissector_key *key,
1319 unsigned int key_count);
1320
089b19a9
SF
1321struct bpf_flow_dissector;
1322bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1323 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1324
3cbf4ffb
SF
1325bool __skb_flow_dissect(const struct net *net,
1326 const struct sk_buff *skb,
e5276937 1327 struct flow_dissector *flow_dissector,
f96533cd
AL
1328 void *target_container, const void *data,
1329 __be16 proto, int nhoff, int hlen, unsigned int flags);
e5276937
TH
1330
1331static inline bool skb_flow_dissect(const struct sk_buff *skb,
1332 struct flow_dissector *flow_dissector,
cd79a238 1333 void *target_container, unsigned int flags)
e5276937 1334{
3cbf4ffb
SF
1335 return __skb_flow_dissect(NULL, skb, flow_dissector,
1336 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1337}
1338
1339static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1340 struct flow_keys *flow,
1341 unsigned int flags)
e5276937
TH
1342{
1343 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1344 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1345 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1346}
1347
72a338bc 1348static inline bool
3cbf4ffb
SF
1349skb_flow_dissect_flow_keys_basic(const struct net *net,
1350 const struct sk_buff *skb,
f96533cd
AL
1351 struct flow_keys_basic *flow,
1352 const void *data, __be16 proto,
1353 int nhoff, int hlen, unsigned int flags)
e5276937
TH
1354{
1355 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1356 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1357 data, proto, nhoff, hlen, flags);
e5276937
TH
1358}
1359
82828b88
JP
1360void skb_flow_dissect_meta(const struct sk_buff *skb,
1361 struct flow_dissector *flow_dissector,
1362 void *target_container);
1363
75a56758 1364/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1365 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1366 * to user states.
1367 */
1368void
1369skb_flow_dissect_ct(const struct sk_buff *skb,
1370 struct flow_dissector *flow_dissector,
1371 void *target_container,
7baf2429 1372 u16 *ctinfo_map, size_t mapsize,
1373 bool post_ct);
62b32379
SH
1374void
1375skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1376 struct flow_dissector *flow_dissector,
1377 void *target_container);
1378
0cb09aff
AL
1379void skb_flow_dissect_hash(const struct sk_buff *skb,
1380 struct flow_dissector *flow_dissector,
1381 void *target_container);
1382
3958afa1 1383static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1384{
a3b18ddb 1385 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1386 __skb_get_hash(skb);
bfb564e7 1387
61b905da 1388 return skb->hash;
bfb564e7
KK
1389}
1390
20a17bf6 1391static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1392{
c6cc1ca7
TH
1393 if (!skb->l4_hash && !skb->sw_hash) {
1394 struct flow_keys keys;
de4c1f8b 1395 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1396
de4c1f8b 1397 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1398 }
f70ea018
TH
1399
1400 return skb->hash;
1401}
1402
55667441
ED
1403__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1404 const siphash_key_t *perturb);
50fb7992 1405
57bdf7f4
TH
1406static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1407{
61b905da 1408 return skb->hash;
57bdf7f4
TH
1409}
1410
3df7a74e
TH
1411static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1412{
61b905da 1413 to->hash = from->hash;
a3b18ddb 1414 to->sw_hash = from->sw_hash;
61b905da 1415 to->l4_hash = from->l4_hash;
3df7a74e
TH
1416};
1417
41477662
JK
1418static inline void skb_copy_decrypted(struct sk_buff *to,
1419 const struct sk_buff *from)
1420{
1421#ifdef CONFIG_TLS_DEVICE
1422 to->decrypted = from->decrypted;
1423#endif
1424}
1425
4305b541
ACM
1426#ifdef NET_SKBUFF_DATA_USES_OFFSET
1427static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1428{
1429 return skb->head + skb->end;
1430}
ec47ea82
AD
1431
1432static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1433{
1434 return skb->end;
1435}
4305b541
ACM
1436#else
1437static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1438{
1439 return skb->end;
1440}
ec47ea82
AD
1441
1442static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1443{
1444 return skb->end - skb->head;
1445}
4305b541
ACM
1446#endif
1447
1da177e4 1448/* Internal */
4305b541 1449#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1450
ac45f602
PO
1451static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1452{
1453 return &skb_shinfo(skb)->hwtstamps;
1454}
1455
52267790
WB
1456static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1457{
06b4feb3 1458 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1459
1460 return is_zcopy ? skb_uarg(skb) : NULL;
1461}
1462
8e044917 1463static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1464{
1465 refcount_inc(&uarg->refcnt);
1466}
1467
9ee5e5ad
JL
1468static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1469{
1470 skb_shinfo(skb)->destructor_arg = uarg;
1471 skb_shinfo(skb)->flags |= uarg->flags;
1472}
1473
52900d22
WB
1474static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1475 bool *have_ref)
52267790
WB
1476{
1477 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1478 if (unlikely(have_ref && *have_ref))
1479 *have_ref = false;
1480 else
8e044917 1481 net_zcopy_get(uarg);
9ee5e5ad 1482 skb_zcopy_init(skb, uarg);
52267790
WB
1483 }
1484}
1485
5cd8d46e
WB
1486static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1487{
1488 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1489 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1490}
1491
1492static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1493{
1494 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1495}
1496
1497static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1498{
1499 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1500}
1501
8e044917 1502static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1503{
1504 if (uarg)
36177832 1505 uarg->callback(NULL, uarg, true);
59776362
JL
1506}
1507
8e044917 1508static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1509{
1510 if (uarg) {
8c793822
JL
1511 if (uarg->callback == msg_zerocopy_callback)
1512 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1513 else if (have_uref)
8e044917 1514 net_zcopy_put(uarg);
236a6b1c
JL
1515 }
1516}
1517
52267790 1518/* Release a reference on a zerocopy structure */
36177832 1519static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1520{
1521 struct ubuf_info *uarg = skb_zcopy(skb);
1522
1523 if (uarg) {
36177832
JL
1524 if (!skb_zcopy_is_nouarg(skb))
1525 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1526
06b4feb3 1527 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
52267790
WB
1528 }
1529}
1530
a8305bff
DM
1531static inline void skb_mark_not_on_list(struct sk_buff *skb)
1532{
1533 skb->next = NULL;
1534}
1535
dcfea72e 1536/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1537#define skb_list_walk_safe(first, skb, next_skb) \
1538 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1539 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1540
992cba7e
DM
1541static inline void skb_list_del_init(struct sk_buff *skb)
1542{
1543 __list_del_entry(&skb->list);
1544 skb_mark_not_on_list(skb);
1545}
1546
1da177e4
LT
1547/**
1548 * skb_queue_empty - check if a queue is empty
1549 * @list: queue head
1550 *
1551 * Returns true if the queue is empty, false otherwise.
1552 */
1553static inline int skb_queue_empty(const struct sk_buff_head *list)
1554{
fd44b93c 1555 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1556}
1557
d7d16a89
ED
1558/**
1559 * skb_queue_empty_lockless - check if a queue is empty
1560 * @list: queue head
1561 *
1562 * Returns true if the queue is empty, false otherwise.
1563 * This variant can be used in lockless contexts.
1564 */
1565static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1566{
1567 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1568}
1569
1570
fc7ebb21
DM
1571/**
1572 * skb_queue_is_last - check if skb is the last entry in the queue
1573 * @list: queue head
1574 * @skb: buffer
1575 *
1576 * Returns true if @skb is the last buffer on the list.
1577 */
1578static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1579 const struct sk_buff *skb)
1580{
fd44b93c 1581 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1582}
1583
832d11c5
IJ
1584/**
1585 * skb_queue_is_first - check if skb is the first entry in the queue
1586 * @list: queue head
1587 * @skb: buffer
1588 *
1589 * Returns true if @skb is the first buffer on the list.
1590 */
1591static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1592 const struct sk_buff *skb)
1593{
fd44b93c 1594 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1595}
1596
249c8b42
DM
1597/**
1598 * skb_queue_next - return the next packet in the queue
1599 * @list: queue head
1600 * @skb: current buffer
1601 *
1602 * Return the next packet in @list after @skb. It is only valid to
1603 * call this if skb_queue_is_last() evaluates to false.
1604 */
1605static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1606 const struct sk_buff *skb)
1607{
1608 /* This BUG_ON may seem severe, but if we just return then we
1609 * are going to dereference garbage.
1610 */
1611 BUG_ON(skb_queue_is_last(list, skb));
1612 return skb->next;
1613}
1614
832d11c5
IJ
1615/**
1616 * skb_queue_prev - return the prev packet in the queue
1617 * @list: queue head
1618 * @skb: current buffer
1619 *
1620 * Return the prev packet in @list before @skb. It is only valid to
1621 * call this if skb_queue_is_first() evaluates to false.
1622 */
1623static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1624 const struct sk_buff *skb)
1625{
1626 /* This BUG_ON may seem severe, but if we just return then we
1627 * are going to dereference garbage.
1628 */
1629 BUG_ON(skb_queue_is_first(list, skb));
1630 return skb->prev;
1631}
1632
1da177e4
LT
1633/**
1634 * skb_get - reference buffer
1635 * @skb: buffer to reference
1636 *
1637 * Makes another reference to a socket buffer and returns a pointer
1638 * to the buffer.
1639 */
1640static inline struct sk_buff *skb_get(struct sk_buff *skb)
1641{
63354797 1642 refcount_inc(&skb->users);
1da177e4
LT
1643 return skb;
1644}
1645
1646/*
f8821f96 1647 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1648 */
1649
1da177e4
LT
1650/**
1651 * skb_cloned - is the buffer a clone
1652 * @skb: buffer to check
1653 *
1654 * Returns true if the buffer was generated with skb_clone() and is
1655 * one of multiple shared copies of the buffer. Cloned buffers are
1656 * shared data so must not be written to under normal circumstances.
1657 */
1658static inline int skb_cloned(const struct sk_buff *skb)
1659{
1660 return skb->cloned &&
1661 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1662}
1663
14bbd6a5
PS
1664static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1665{
d0164adc 1666 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1667
1668 if (skb_cloned(skb))
1669 return pskb_expand_head(skb, 0, 0, pri);
1670
1671 return 0;
1672}
1673
1da177e4
LT
1674/**
1675 * skb_header_cloned - is the header a clone
1676 * @skb: buffer to check
1677 *
1678 * Returns true if modifying the header part of the buffer requires
1679 * the data to be copied.
1680 */
1681static inline int skb_header_cloned(const struct sk_buff *skb)
1682{
1683 int dataref;
1684
1685 if (!skb->cloned)
1686 return 0;
1687
1688 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1689 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1690 return dataref != 1;
1691}
1692
9580bf2e
ED
1693static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1694{
1695 might_sleep_if(gfpflags_allow_blocking(pri));
1696
1697 if (skb_header_cloned(skb))
1698 return pskb_expand_head(skb, 0, 0, pri);
1699
1700 return 0;
1701}
1702
f4a775d1
ED
1703/**
1704 * __skb_header_release - release reference to header
1705 * @skb: buffer to operate on
f4a775d1
ED
1706 */
1707static inline void __skb_header_release(struct sk_buff *skb)
1708{
1709 skb->nohdr = 1;
1710 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1711}
1712
1713
1da177e4
LT
1714/**
1715 * skb_shared - is the buffer shared
1716 * @skb: buffer to check
1717 *
1718 * Returns true if more than one person has a reference to this
1719 * buffer.
1720 */
1721static inline int skb_shared(const struct sk_buff *skb)
1722{
63354797 1723 return refcount_read(&skb->users) != 1;
1da177e4
LT
1724}
1725
1726/**
1727 * skb_share_check - check if buffer is shared and if so clone it
1728 * @skb: buffer to check
1729 * @pri: priority for memory allocation
1730 *
1731 * If the buffer is shared the buffer is cloned and the old copy
1732 * drops a reference. A new clone with a single reference is returned.
1733 * If the buffer is not shared the original buffer is returned. When
1734 * being called from interrupt status or with spinlocks held pri must
1735 * be GFP_ATOMIC.
1736 *
1737 * NULL is returned on a memory allocation failure.
1738 */
47061bc4 1739static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1740{
d0164adc 1741 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1742 if (skb_shared(skb)) {
1743 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1744
1745 if (likely(nskb))
1746 consume_skb(skb);
1747 else
1748 kfree_skb(skb);
1da177e4
LT
1749 skb = nskb;
1750 }
1751 return skb;
1752}
1753
1754/*
1755 * Copy shared buffers into a new sk_buff. We effectively do COW on
1756 * packets to handle cases where we have a local reader and forward
1757 * and a couple of other messy ones. The normal one is tcpdumping
1758 * a packet thats being forwarded.
1759 */
1760
1761/**
1762 * skb_unshare - make a copy of a shared buffer
1763 * @skb: buffer to check
1764 * @pri: priority for memory allocation
1765 *
1766 * If the socket buffer is a clone then this function creates a new
1767 * copy of the data, drops a reference count on the old copy and returns
1768 * the new copy with the reference count at 1. If the buffer is not a clone
1769 * the original buffer is returned. When called with a spinlock held or
1770 * from interrupt state @pri must be %GFP_ATOMIC
1771 *
1772 * %NULL is returned on a memory allocation failure.
1773 */
e2bf521d 1774static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1775 gfp_t pri)
1da177e4 1776{
d0164adc 1777 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1778 if (skb_cloned(skb)) {
1779 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1780
1781 /* Free our shared copy */
1782 if (likely(nskb))
1783 consume_skb(skb);
1784 else
1785 kfree_skb(skb);
1da177e4
LT
1786 skb = nskb;
1787 }
1788 return skb;
1789}
1790
1791/**
1a5778aa 1792 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1793 * @list_: list to peek at
1794 *
1795 * Peek an &sk_buff. Unlike most other operations you _MUST_
1796 * be careful with this one. A peek leaves the buffer on the
1797 * list and someone else may run off with it. You must hold
1798 * the appropriate locks or have a private queue to do this.
1799 *
1800 * Returns %NULL for an empty list or a pointer to the head element.
1801 * The reference count is not incremented and the reference is therefore
1802 * volatile. Use with caution.
1803 */
05bdd2f1 1804static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1805{
18d07000
ED
1806 struct sk_buff *skb = list_->next;
1807
1808 if (skb == (struct sk_buff *)list_)
1809 skb = NULL;
1810 return skb;
1da177e4
LT
1811}
1812
8b69bd7d
DM
1813/**
1814 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1815 * @list_: list to peek at
1816 *
1817 * Like skb_peek(), but the caller knows that the list is not empty.
1818 */
1819static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1820{
1821 return list_->next;
1822}
1823
da5ef6e5
PE
1824/**
1825 * skb_peek_next - peek skb following the given one from a queue
1826 * @skb: skb to start from
1827 * @list_: list to peek at
1828 *
1829 * Returns %NULL when the end of the list is met or a pointer to the
1830 * next element. The reference count is not incremented and the
1831 * reference is therefore volatile. Use with caution.
1832 */
1833static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1834 const struct sk_buff_head *list_)
1835{
1836 struct sk_buff *next = skb->next;
18d07000 1837
da5ef6e5
PE
1838 if (next == (struct sk_buff *)list_)
1839 next = NULL;
1840 return next;
1841}
1842
1da177e4 1843/**
1a5778aa 1844 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1845 * @list_: list to peek at
1846 *
1847 * Peek an &sk_buff. Unlike most other operations you _MUST_
1848 * be careful with this one. A peek leaves the buffer on the
1849 * list and someone else may run off with it. You must hold
1850 * the appropriate locks or have a private queue to do this.
1851 *
1852 * Returns %NULL for an empty list or a pointer to the tail element.
1853 * The reference count is not incremented and the reference is therefore
1854 * volatile. Use with caution.
1855 */
05bdd2f1 1856static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1857{
f8cc62ca 1858 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1859
1860 if (skb == (struct sk_buff *)list_)
1861 skb = NULL;
1862 return skb;
1863
1da177e4
LT
1864}
1865
1866/**
1867 * skb_queue_len - get queue length
1868 * @list_: list to measure
1869 *
1870 * Return the length of an &sk_buff queue.
1871 */
1872static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1873{
1874 return list_->qlen;
1875}
1876
86b18aaa
QC
1877/**
1878 * skb_queue_len_lockless - get queue length
1879 * @list_: list to measure
1880 *
1881 * Return the length of an &sk_buff queue.
1882 * This variant can be used in lockless contexts.
1883 */
1884static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1885{
1886 return READ_ONCE(list_->qlen);
1887}
1888
67fed459
DM
1889/**
1890 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1891 * @list: queue to initialize
1892 *
1893 * This initializes only the list and queue length aspects of
1894 * an sk_buff_head object. This allows to initialize the list
1895 * aspects of an sk_buff_head without reinitializing things like
1896 * the spinlock. It can also be used for on-stack sk_buff_head
1897 * objects where the spinlock is known to not be used.
1898 */
1899static inline void __skb_queue_head_init(struct sk_buff_head *list)
1900{
1901 list->prev = list->next = (struct sk_buff *)list;
1902 list->qlen = 0;
1903}
1904
76f10ad0
AV
1905/*
1906 * This function creates a split out lock class for each invocation;
1907 * this is needed for now since a whole lot of users of the skb-queue
1908 * infrastructure in drivers have different locking usage (in hardirq)
1909 * than the networking core (in softirq only). In the long run either the
1910 * network layer or drivers should need annotation to consolidate the
1911 * main types of usage into 3 classes.
1912 */
1da177e4
LT
1913static inline void skb_queue_head_init(struct sk_buff_head *list)
1914{
1915 spin_lock_init(&list->lock);
67fed459 1916 __skb_queue_head_init(list);
1da177e4
LT
1917}
1918
c2ecba71
PE
1919static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1920 struct lock_class_key *class)
1921{
1922 skb_queue_head_init(list);
1923 lockdep_set_class(&list->lock, class);
1924}
1925
1da177e4 1926/*
bf299275 1927 * Insert an sk_buff on a list.
1da177e4
LT
1928 *
1929 * The "__skb_xxxx()" functions are the non-atomic ones that
1930 * can only be called with interrupts disabled.
1931 */
bf299275
GR
1932static inline void __skb_insert(struct sk_buff *newsk,
1933 struct sk_buff *prev, struct sk_buff *next,
1934 struct sk_buff_head *list)
1935{
f8cc62ca
ED
1936 /* See skb_queue_empty_lockless() and skb_peek_tail()
1937 * for the opposite READ_ONCE()
1938 */
d7d16a89
ED
1939 WRITE_ONCE(newsk->next, next);
1940 WRITE_ONCE(newsk->prev, prev);
1941 WRITE_ONCE(next->prev, newsk);
1942 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1943 list->qlen++;
1944}
1da177e4 1945
67fed459
DM
1946static inline void __skb_queue_splice(const struct sk_buff_head *list,
1947 struct sk_buff *prev,
1948 struct sk_buff *next)
1949{
1950 struct sk_buff *first = list->next;
1951 struct sk_buff *last = list->prev;
1952
d7d16a89
ED
1953 WRITE_ONCE(first->prev, prev);
1954 WRITE_ONCE(prev->next, first);
67fed459 1955
d7d16a89
ED
1956 WRITE_ONCE(last->next, next);
1957 WRITE_ONCE(next->prev, last);
67fed459
DM
1958}
1959
1960/**
1961 * skb_queue_splice - join two skb lists, this is designed for stacks
1962 * @list: the new list to add
1963 * @head: the place to add it in the first list
1964 */
1965static inline void skb_queue_splice(const struct sk_buff_head *list,
1966 struct sk_buff_head *head)
1967{
1968 if (!skb_queue_empty(list)) {
1969 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1970 head->qlen += list->qlen;
67fed459
DM
1971 }
1972}
1973
1974/**
d9619496 1975 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1976 * @list: the new list to add
1977 * @head: the place to add it in the first list
1978 *
1979 * The list at @list is reinitialised
1980 */
1981static inline void skb_queue_splice_init(struct sk_buff_head *list,
1982 struct sk_buff_head *head)
1983{
1984 if (!skb_queue_empty(list)) {
1985 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1986 head->qlen += list->qlen;
67fed459
DM
1987 __skb_queue_head_init(list);
1988 }
1989}
1990
1991/**
1992 * skb_queue_splice_tail - join two skb lists, each list being a queue
1993 * @list: the new list to add
1994 * @head: the place to add it in the first list
1995 */
1996static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1997 struct sk_buff_head *head)
1998{
1999 if (!skb_queue_empty(list)) {
2000 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2001 head->qlen += list->qlen;
67fed459
DM
2002 }
2003}
2004
2005/**
d9619496 2006 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
2007 * @list: the new list to add
2008 * @head: the place to add it in the first list
2009 *
2010 * Each of the lists is a queue.
2011 * The list at @list is reinitialised
2012 */
2013static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2014 struct sk_buff_head *head)
2015{
2016 if (!skb_queue_empty(list)) {
2017 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2018 head->qlen += list->qlen;
67fed459
DM
2019 __skb_queue_head_init(list);
2020 }
2021}
2022
1da177e4 2023/**
300ce174 2024 * __skb_queue_after - queue a buffer at the list head
1da177e4 2025 * @list: list to use
300ce174 2026 * @prev: place after this buffer
1da177e4
LT
2027 * @newsk: buffer to queue
2028 *
300ce174 2029 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2030 * and you must therefore hold required locks before calling it.
2031 *
2032 * A buffer cannot be placed on two lists at the same time.
2033 */
300ce174
SH
2034static inline void __skb_queue_after(struct sk_buff_head *list,
2035 struct sk_buff *prev,
2036 struct sk_buff *newsk)
1da177e4 2037{
bf299275 2038 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
2039}
2040
7965bd4d
JP
2041void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2042 struct sk_buff_head *list);
7de6c033 2043
f5572855
GR
2044static inline void __skb_queue_before(struct sk_buff_head *list,
2045 struct sk_buff *next,
2046 struct sk_buff *newsk)
2047{
2048 __skb_insert(newsk, next->prev, next, list);
2049}
2050
300ce174
SH
2051/**
2052 * __skb_queue_head - queue a buffer at the list head
2053 * @list: list to use
2054 * @newsk: buffer to queue
2055 *
2056 * Queue a buffer at the start of a list. This function takes no locks
2057 * and you must therefore hold required locks before calling it.
2058 *
2059 * A buffer cannot be placed on two lists at the same time.
2060 */
300ce174
SH
2061static inline void __skb_queue_head(struct sk_buff_head *list,
2062 struct sk_buff *newsk)
2063{
2064 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2065}
4ea7b0cf 2066void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2067
1da177e4
LT
2068/**
2069 * __skb_queue_tail - queue a buffer at the list tail
2070 * @list: list to use
2071 * @newsk: buffer to queue
2072 *
2073 * Queue a buffer at the end of a list. This function takes no locks
2074 * and you must therefore hold required locks before calling it.
2075 *
2076 * A buffer cannot be placed on two lists at the same time.
2077 */
1da177e4
LT
2078static inline void __skb_queue_tail(struct sk_buff_head *list,
2079 struct sk_buff *newsk)
2080{
f5572855 2081 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2082}
4ea7b0cf 2083void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2084
1da177e4
LT
2085/*
2086 * remove sk_buff from list. _Must_ be called atomically, and with
2087 * the list known..
2088 */
7965bd4d 2089void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2090static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2091{
2092 struct sk_buff *next, *prev;
2093
86b18aaa 2094 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2095 next = skb->next;
2096 prev = skb->prev;
2097 skb->next = skb->prev = NULL;
d7d16a89
ED
2098 WRITE_ONCE(next->prev, prev);
2099 WRITE_ONCE(prev->next, next);
1da177e4
LT
2100}
2101
f525c06d
GR
2102/**
2103 * __skb_dequeue - remove from the head of the queue
2104 * @list: list to dequeue from
2105 *
2106 * Remove the head of the list. This function does not take any locks
2107 * so must be used with appropriate locks held only. The head item is
2108 * returned or %NULL if the list is empty.
2109 */
f525c06d
GR
2110static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2111{
2112 struct sk_buff *skb = skb_peek(list);
2113 if (skb)
2114 __skb_unlink(skb, list);
2115 return skb;
2116}
4ea7b0cf 2117struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2118
2119/**
2120 * __skb_dequeue_tail - remove from the tail of the queue
2121 * @list: list to dequeue from
2122 *
2123 * Remove the tail of the list. This function does not take any locks
2124 * so must be used with appropriate locks held only. The tail item is
2125 * returned or %NULL if the list is empty.
2126 */
1da177e4
LT
2127static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2128{
2129 struct sk_buff *skb = skb_peek_tail(list);
2130 if (skb)
2131 __skb_unlink(skb, list);
2132 return skb;
2133}
4ea7b0cf 2134struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2135
2136
bdcc0924 2137static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2138{
2139 return skb->data_len;
2140}
2141
2142static inline unsigned int skb_headlen(const struct sk_buff *skb)
2143{
2144 return skb->len - skb->data_len;
2145}
2146
3ece7826 2147static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2148{
c72d8cda 2149 unsigned int i, len = 0;
1da177e4 2150
c72d8cda 2151 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2152 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2153 return len;
2154}
2155
2156static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2157{
2158 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2159}
2160
131ea667
IC
2161/**
2162 * __skb_fill_page_desc - initialise a paged fragment in an skb
2163 * @skb: buffer containing fragment to be initialised
2164 * @i: paged fragment index to initialise
2165 * @page: the page to use for this fragment
2166 * @off: the offset to the data with @page
2167 * @size: the length of the data
2168 *
2169 * Initialises the @i'th fragment of @skb to point to &size bytes at
2170 * offset @off within @page.
2171 *
2172 * Does not take any additional reference on the fragment.
2173 */
2174static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2175 struct page *page, int off, int size)
1da177e4
LT
2176{
2177 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2178
c48a11c7 2179 /*
2f064f34
MH
2180 * Propagate page pfmemalloc to the skb if we can. The problem is
2181 * that not all callers have unique ownership of the page but rely
2182 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2183 */
1dfa5bd3 2184 frag->bv_page = page;
65c84f14 2185 frag->bv_offset = off;
9e903e08 2186 skb_frag_size_set(frag, size);
cca7af38
PE
2187
2188 page = compound_head(page);
2f064f34 2189 if (page_is_pfmemalloc(page))
cca7af38 2190 skb->pfmemalloc = true;
131ea667
IC
2191}
2192
2193/**
2194 * skb_fill_page_desc - initialise a paged fragment in an skb
2195 * @skb: buffer containing fragment to be initialised
2196 * @i: paged fragment index to initialise
2197 * @page: the page to use for this fragment
2198 * @off: the offset to the data with @page
2199 * @size: the length of the data
2200 *
2201 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2202 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2203 * addition updates @skb such that @i is the last fragment.
2204 *
2205 * Does not take any additional reference on the fragment.
2206 */
2207static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2208 struct page *page, int off, int size)
2209{
2210 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2211 skb_shinfo(skb)->nr_frags = i + 1;
2212}
2213
7965bd4d
JP
2214void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2215 int size, unsigned int truesize);
654bed16 2216
f8e617e1
JW
2217void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2218 unsigned int truesize);
2219
1da177e4
LT
2220#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2221
27a884dc
ACM
2222#ifdef NET_SKBUFF_DATA_USES_OFFSET
2223static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2224{
2225 return skb->head + skb->tail;
2226}
2227
2228static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2229{
2230 skb->tail = skb->data - skb->head;
2231}
2232
2233static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2234{
2235 skb_reset_tail_pointer(skb);
2236 skb->tail += offset;
2237}
7cc46190 2238
27a884dc
ACM
2239#else /* NET_SKBUFF_DATA_USES_OFFSET */
2240static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2241{
2242 return skb->tail;
2243}
2244
2245static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2246{
2247 skb->tail = skb->data;
2248}
2249
2250static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2251{
2252 skb->tail = skb->data + offset;
2253}
4305b541 2254
27a884dc
ACM
2255#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2256
1da177e4
LT
2257/*
2258 * Add data to an sk_buff
2259 */
4df864c1
JB
2260void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2261void *skb_put(struct sk_buff *skb, unsigned int len);
2262static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2263{
4df864c1 2264 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2265 SKB_LINEAR_ASSERT(skb);
2266 skb->tail += len;
2267 skb->len += len;
2268 return tmp;
2269}
2270
de77b966 2271static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2272{
2273 void *tmp = __skb_put(skb, len);
2274
2275 memset(tmp, 0, len);
2276 return tmp;
2277}
2278
2279static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2280 unsigned int len)
2281{
2282 void *tmp = __skb_put(skb, len);
2283
2284 memcpy(tmp, data, len);
2285 return tmp;
2286}
2287
2288static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2289{
2290 *(u8 *)__skb_put(skb, 1) = val;
2291}
2292
83ad357d 2293static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2294{
83ad357d 2295 void *tmp = skb_put(skb, len);
e45a79da
JB
2296
2297 memset(tmp, 0, len);
2298
2299 return tmp;
2300}
2301
59ae1d12
JB
2302static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2303 unsigned int len)
2304{
2305 void *tmp = skb_put(skb, len);
2306
2307 memcpy(tmp, data, len);
2308
2309 return tmp;
2310}
2311
634fef61
JB
2312static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2313{
2314 *(u8 *)skb_put(skb, 1) = val;
2315}
2316
d58ff351
JB
2317void *skb_push(struct sk_buff *skb, unsigned int len);
2318static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2319{
2320 skb->data -= len;
2321 skb->len += len;
2322 return skb->data;
2323}
2324
af72868b
JB
2325void *skb_pull(struct sk_buff *skb, unsigned int len);
2326static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2327{
2328 skb->len -= len;
2329 BUG_ON(skb->len < skb->data_len);
2330 return skb->data += len;
2331}
2332
af72868b 2333static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2334{
2335 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2336}
2337
af72868b 2338void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2339
af72868b 2340static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2341{
2342 if (len > skb_headlen(skb) &&
987c402a 2343 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2344 return NULL;
2345 skb->len -= len;
2346 return skb->data += len;
2347}
2348
af72868b 2349static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2350{
2351 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2352}
2353
b9df4fd7 2354static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2355{
2356 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2357 return true;
1da177e4 2358 if (unlikely(len > skb->len))
b9df4fd7 2359 return false;
987c402a 2360 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2361}
2362
c8c8b127
ED
2363void skb_condense(struct sk_buff *skb);
2364
1da177e4
LT
2365/**
2366 * skb_headroom - bytes at buffer head
2367 * @skb: buffer to check
2368 *
2369 * Return the number of bytes of free space at the head of an &sk_buff.
2370 */
c2636b4d 2371static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2372{
2373 return skb->data - skb->head;
2374}
2375
2376/**
2377 * skb_tailroom - bytes at buffer end
2378 * @skb: buffer to check
2379 *
2380 * Return the number of bytes of free space at the tail of an sk_buff
2381 */
2382static inline int skb_tailroom(const struct sk_buff *skb)
2383{
4305b541 2384 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2385}
2386
a21d4572
ED
2387/**
2388 * skb_availroom - bytes at buffer end
2389 * @skb: buffer to check
2390 *
2391 * Return the number of bytes of free space at the tail of an sk_buff
2392 * allocated by sk_stream_alloc()
2393 */
2394static inline int skb_availroom(const struct sk_buff *skb)
2395{
16fad69c
ED
2396 if (skb_is_nonlinear(skb))
2397 return 0;
2398
2399 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2400}
2401
1da177e4
LT
2402/**
2403 * skb_reserve - adjust headroom
2404 * @skb: buffer to alter
2405 * @len: bytes to move
2406 *
2407 * Increase the headroom of an empty &sk_buff by reducing the tail
2408 * room. This is only allowed for an empty buffer.
2409 */
8243126c 2410static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2411{
2412 skb->data += len;
2413 skb->tail += len;
2414}
2415
1837b2e2
BP
2416/**
2417 * skb_tailroom_reserve - adjust reserved_tailroom
2418 * @skb: buffer to alter
2419 * @mtu: maximum amount of headlen permitted
2420 * @needed_tailroom: minimum amount of reserved_tailroom
2421 *
2422 * Set reserved_tailroom so that headlen can be as large as possible but
2423 * not larger than mtu and tailroom cannot be smaller than
2424 * needed_tailroom.
2425 * The required headroom should already have been reserved before using
2426 * this function.
2427 */
2428static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2429 unsigned int needed_tailroom)
2430{
2431 SKB_LINEAR_ASSERT(skb);
2432 if (mtu < skb_tailroom(skb) - needed_tailroom)
2433 /* use at most mtu */
2434 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2435 else
2436 /* use up to all available space */
2437 skb->reserved_tailroom = needed_tailroom;
2438}
2439
8bce6d7d
TH
2440#define ENCAP_TYPE_ETHER 0
2441#define ENCAP_TYPE_IPPROTO 1
2442
2443static inline void skb_set_inner_protocol(struct sk_buff *skb,
2444 __be16 protocol)
2445{
2446 skb->inner_protocol = protocol;
2447 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2448}
2449
2450static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2451 __u8 ipproto)
2452{
2453 skb->inner_ipproto = ipproto;
2454 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2455}
2456
6a674e9c
JG
2457static inline void skb_reset_inner_headers(struct sk_buff *skb)
2458{
aefbd2b3 2459 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2460 skb->inner_network_header = skb->network_header;
2461 skb->inner_transport_header = skb->transport_header;
2462}
2463
0b5c9db1
JP
2464static inline void skb_reset_mac_len(struct sk_buff *skb)
2465{
2466 skb->mac_len = skb->network_header - skb->mac_header;
2467}
2468
6a674e9c
JG
2469static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2470 *skb)
2471{
2472 return skb->head + skb->inner_transport_header;
2473}
2474
55dc5a9f
TH
2475static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2476{
2477 return skb_inner_transport_header(skb) - skb->data;
2478}
2479
6a674e9c
JG
2480static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2481{
2482 skb->inner_transport_header = skb->data - skb->head;
2483}
2484
2485static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2486 const int offset)
2487{
2488 skb_reset_inner_transport_header(skb);
2489 skb->inner_transport_header += offset;
2490}
2491
2492static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2493{
2494 return skb->head + skb->inner_network_header;
2495}
2496
2497static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2498{
2499 skb->inner_network_header = skb->data - skb->head;
2500}
2501
2502static inline void skb_set_inner_network_header(struct sk_buff *skb,
2503 const int offset)
2504{
2505 skb_reset_inner_network_header(skb);
2506 skb->inner_network_header += offset;
2507}
2508
aefbd2b3
PS
2509static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2510{
2511 return skb->head + skb->inner_mac_header;
2512}
2513
2514static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2515{
2516 skb->inner_mac_header = skb->data - skb->head;
2517}
2518
2519static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2520 const int offset)
2521{
2522 skb_reset_inner_mac_header(skb);
2523 skb->inner_mac_header += offset;
2524}
fda55eca
ED
2525static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2526{
35d04610 2527 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2528}
2529
9c70220b
ACM
2530static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2531{
2e07fa9c 2532 return skb->head + skb->transport_header;
9c70220b
ACM
2533}
2534
badff6d0
ACM
2535static inline void skb_reset_transport_header(struct sk_buff *skb)
2536{
2e07fa9c 2537 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2538}
2539
967b05f6
ACM
2540static inline void skb_set_transport_header(struct sk_buff *skb,
2541 const int offset)
2542{
2e07fa9c
ACM
2543 skb_reset_transport_header(skb);
2544 skb->transport_header += offset;
ea2ae17d
ACM
2545}
2546
d56f90a7
ACM
2547static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2548{
2e07fa9c 2549 return skb->head + skb->network_header;
d56f90a7
ACM
2550}
2551
c1d2bbe1
ACM
2552static inline void skb_reset_network_header(struct sk_buff *skb)
2553{
2e07fa9c 2554 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2555}
2556
c14d2450
ACM
2557static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2558{
2e07fa9c
ACM
2559 skb_reset_network_header(skb);
2560 skb->network_header += offset;
c14d2450
ACM
2561}
2562
2e07fa9c 2563static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2564{
2e07fa9c 2565 return skb->head + skb->mac_header;
bbe735e4
ACM
2566}
2567
ea6da4fd
AV
2568static inline int skb_mac_offset(const struct sk_buff *skb)
2569{
2570 return skb_mac_header(skb) - skb->data;
2571}
2572
0daf4349
DB
2573static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2574{
2575 return skb->network_header - skb->mac_header;
2576}
2577
2e07fa9c 2578static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2579{
35d04610 2580 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2581}
2582
b4ab3141
DB
2583static inline void skb_unset_mac_header(struct sk_buff *skb)
2584{
2585 skb->mac_header = (typeof(skb->mac_header))~0U;
2586}
2587
2e07fa9c
ACM
2588static inline void skb_reset_mac_header(struct sk_buff *skb)
2589{
2590 skb->mac_header = skb->data - skb->head;
2591}
2592
2593static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2594{
2595 skb_reset_mac_header(skb);
2596 skb->mac_header += offset;
2597}
2598
0e3da5bb
TT
2599static inline void skb_pop_mac_header(struct sk_buff *skb)
2600{
2601 skb->mac_header = skb->network_header;
2602}
2603
d2aa125d 2604static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2605{
72a338bc 2606 struct flow_keys_basic keys;
fbbdb8f0
YX
2607
2608 if (skb_transport_header_was_set(skb))
2609 return;
72a338bc 2610
3cbf4ffb
SF
2611 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2612 NULL, 0, 0, 0, 0))
42aecaa9 2613 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2614}
2615
03606895
ED
2616static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2617{
2618 if (skb_mac_header_was_set(skb)) {
2619 const unsigned char *old_mac = skb_mac_header(skb);
2620
2621 skb_set_mac_header(skb, -skb->mac_len);
2622 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2623 }
2624}
2625
04fb451e
MM
2626static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2627{
2628 return skb->csum_start - skb_headroom(skb);
2629}
2630
08b64fcc
AD
2631static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2632{
2633 return skb->head + skb->csum_start;
2634}
2635
2e07fa9c
ACM
2636static inline int skb_transport_offset(const struct sk_buff *skb)
2637{
2638 return skb_transport_header(skb) - skb->data;
2639}
2640
2641static inline u32 skb_network_header_len(const struct sk_buff *skb)
2642{
2643 return skb->transport_header - skb->network_header;
2644}
2645
6a674e9c
JG
2646static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2647{
2648 return skb->inner_transport_header - skb->inner_network_header;
2649}
2650
2e07fa9c
ACM
2651static inline int skb_network_offset(const struct sk_buff *skb)
2652{
2653 return skb_network_header(skb) - skb->data;
2654}
48d49d0c 2655
6a674e9c
JG
2656static inline int skb_inner_network_offset(const struct sk_buff *skb)
2657{
2658 return skb_inner_network_header(skb) - skb->data;
2659}
2660
f9599ce1
CG
2661static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2662{
2663 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2664}
2665
1da177e4
LT
2666/*
2667 * CPUs often take a performance hit when accessing unaligned memory
2668 * locations. The actual performance hit varies, it can be small if the
2669 * hardware handles it or large if we have to take an exception and fix it
2670 * in software.
2671 *
2672 * Since an ethernet header is 14 bytes network drivers often end up with
2673 * the IP header at an unaligned offset. The IP header can be aligned by
2674 * shifting the start of the packet by 2 bytes. Drivers should do this
2675 * with:
2676 *
8660c124 2677 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2678 *
2679 * The downside to this alignment of the IP header is that the DMA is now
2680 * unaligned. On some architectures the cost of an unaligned DMA is high
2681 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2682 *
1da177e4
LT
2683 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2684 * to be overridden.
2685 */
2686#ifndef NET_IP_ALIGN
2687#define NET_IP_ALIGN 2
2688#endif
2689
025be81e
AB
2690/*
2691 * The networking layer reserves some headroom in skb data (via
2692 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2693 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2694 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2695 *
2696 * Unfortunately this headroom changes the DMA alignment of the resulting
2697 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2698 * on some architectures. An architecture can override this value,
2699 * perhaps setting it to a cacheline in size (since that will maintain
2700 * cacheline alignment of the DMA). It must be a power of 2.
2701 *
d6301d3d 2702 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2703 * headroom, you should not reduce this.
5933dd2f
ED
2704 *
2705 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2706 * to reduce average number of cache lines per packet.
645f0897 2707 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2708 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2709 */
2710#ifndef NET_SKB_PAD
5933dd2f 2711#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2712#endif
2713
7965bd4d 2714int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2715
5293efe6 2716static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2717{
5e1abdc3 2718 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2719 return;
27a884dc
ACM
2720 skb->len = len;
2721 skb_set_tail_pointer(skb, len);
1da177e4
LT
2722}
2723
5293efe6
DB
2724static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2725{
2726 __skb_set_length(skb, len);
2727}
2728
7965bd4d 2729void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2730
2731static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2732{
3cc0e873
HX
2733 if (skb->data_len)
2734 return ___pskb_trim(skb, len);
2735 __skb_trim(skb, len);
2736 return 0;
1da177e4
LT
2737}
2738
2739static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2740{
2741 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2742}
2743
e9fa4f7b
HX
2744/**
2745 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2746 * @skb: buffer to alter
2747 * @len: new length
2748 *
2749 * This is identical to pskb_trim except that the caller knows that
2750 * the skb is not cloned so we should never get an error due to out-
2751 * of-memory.
2752 */
2753static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2754{
2755 int err = pskb_trim(skb, len);
2756 BUG_ON(err);
2757}
2758
5293efe6
DB
2759static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2760{
2761 unsigned int diff = len - skb->len;
2762
2763 if (skb_tailroom(skb) < diff) {
2764 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2765 GFP_ATOMIC);
2766 if (ret)
2767 return ret;
2768 }
2769 __skb_set_length(skb, len);
2770 return 0;
2771}
2772
1da177e4
LT
2773/**
2774 * skb_orphan - orphan a buffer
2775 * @skb: buffer to orphan
2776 *
2777 * If a buffer currently has an owner then we call the owner's
2778 * destructor function and make the @skb unowned. The buffer continues
2779 * to exist but is no longer charged to its former owner.
2780 */
2781static inline void skb_orphan(struct sk_buff *skb)
2782{
c34a7612 2783 if (skb->destructor) {
1da177e4 2784 skb->destructor(skb);
c34a7612
ED
2785 skb->destructor = NULL;
2786 skb->sk = NULL;
376c7311
ED
2787 } else {
2788 BUG_ON(skb->sk);
c34a7612 2789 }
1da177e4
LT
2790}
2791
a353e0ce
MT
2792/**
2793 * skb_orphan_frags - orphan the frags contained in a buffer
2794 * @skb: buffer to orphan frags from
2795 * @gfp_mask: allocation mask for replacement pages
2796 *
2797 * For each frag in the SKB which needs a destructor (i.e. has an
2798 * owner) create a copy of that frag and release the original
2799 * page by calling the destructor.
2800 */
2801static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2802{
1f8b977a
WB
2803 if (likely(!skb_zcopy(skb)))
2804 return 0;
185ce5c3 2805 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 2806 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
2807 return 0;
2808 return skb_copy_ubufs(skb, gfp_mask);
2809}
2810
2811/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2812static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2813{
2814 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2815 return 0;
2816 return skb_copy_ubufs(skb, gfp_mask);
2817}
2818
1da177e4
LT
2819/**
2820 * __skb_queue_purge - empty a list
2821 * @list: list to empty
2822 *
2823 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2824 * the list and one reference dropped. This function does not take the
2825 * list lock and the caller must hold the relevant locks to use it.
2826 */
1da177e4
LT
2827static inline void __skb_queue_purge(struct sk_buff_head *list)
2828{
2829 struct sk_buff *skb;
2830 while ((skb = __skb_dequeue(list)) != NULL)
2831 kfree_skb(skb);
2832}
4ea7b0cf 2833void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2834
385114de 2835unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2836
3f6e687d
KH
2837void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2838
2839/**
2840 * netdev_alloc_frag - allocate a page fragment
2841 * @fragsz: fragment size
2842 *
2843 * Allocates a frag from a page for receive buffer.
2844 * Uses GFP_ATOMIC allocations.
2845 */
2846static inline void *netdev_alloc_frag(unsigned int fragsz)
2847{
2848 return __netdev_alloc_frag_align(fragsz, ~0u);
2849}
2850
2851static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2852 unsigned int align)
2853{
2854 WARN_ON_ONCE(!is_power_of_2(align));
2855 return __netdev_alloc_frag_align(fragsz, -align);
2856}
1da177e4 2857
7965bd4d
JP
2858struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2859 gfp_t gfp_mask);
8af27456
CH
2860
2861/**
2862 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2863 * @dev: network device to receive on
2864 * @length: length to allocate
2865 *
2866 * Allocate a new &sk_buff and assign it a usage count of one. The
2867 * buffer has unspecified headroom built in. Users should allocate
2868 * the headroom they think they need without accounting for the
2869 * built in space. The built in space is used for optimisations.
2870 *
2871 * %NULL is returned if there is no free memory. Although this function
2872 * allocates memory it can be called from an interrupt.
2873 */
2874static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2875 unsigned int length)
8af27456
CH
2876{
2877 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2878}
2879
6f532612
ED
2880/* legacy helper around __netdev_alloc_skb() */
2881static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2882 gfp_t gfp_mask)
2883{
2884 return __netdev_alloc_skb(NULL, length, gfp_mask);
2885}
2886
2887/* legacy helper around netdev_alloc_skb() */
2888static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2889{
2890 return netdev_alloc_skb(NULL, length);
2891}
2892
2893
4915a0de
ED
2894static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2895 unsigned int length, gfp_t gfp)
61321bbd 2896{
4915a0de 2897 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2898
2899 if (NET_IP_ALIGN && skb)
2900 skb_reserve(skb, NET_IP_ALIGN);
2901 return skb;
2902}
2903
4915a0de
ED
2904static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2905 unsigned int length)
2906{
2907 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2908}
2909
181edb2b
AD
2910static inline void skb_free_frag(void *addr)
2911{
8c2dd3e4 2912 page_frag_free(addr);
181edb2b
AD
2913}
2914
3f6e687d
KH
2915void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2916
2917static inline void *napi_alloc_frag(unsigned int fragsz)
2918{
2919 return __napi_alloc_frag_align(fragsz, ~0u);
2920}
2921
2922static inline void *napi_alloc_frag_align(unsigned int fragsz,
2923 unsigned int align)
2924{
2925 WARN_ON_ONCE(!is_power_of_2(align));
2926 return __napi_alloc_frag_align(fragsz, -align);
2927}
2928
fd11a83d
AD
2929struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2930 unsigned int length, gfp_t gfp_mask);
2931static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2932 unsigned int length)
2933{
2934 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2935}
795bb1c0
JDB
2936void napi_consume_skb(struct sk_buff *skb, int budget);
2937
9243adfc 2938void napi_skb_free_stolen_head(struct sk_buff *skb);
15fad714 2939void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2940
71dfda58
AD
2941/**
2942 * __dev_alloc_pages - allocate page for network Rx
2943 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2944 * @order: size of the allocation
2945 *
2946 * Allocate a new page.
2947 *
2948 * %NULL is returned if there is no free memory.
2949*/
2950static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2951 unsigned int order)
2952{
2953 /* This piece of code contains several assumptions.
2954 * 1. This is for device Rx, therefor a cold page is preferred.
2955 * 2. The expectation is the user wants a compound page.
2956 * 3. If requesting a order 0 page it will not be compound
2957 * due to the check to see if order has a value in prep_new_page
2958 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2959 * code in gfp_to_alloc_flags that should be enforcing this.
2960 */
453f85d4 2961 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2962
2963 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2964}
2965
2966static inline struct page *dev_alloc_pages(unsigned int order)
2967{
95829b3a 2968 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2969}
2970
2971/**
2972 * __dev_alloc_page - allocate a page for network Rx
2973 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2974 *
2975 * Allocate a new page.
2976 *
2977 * %NULL is returned if there is no free memory.
2978 */
2979static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2980{
2981 return __dev_alloc_pages(gfp_mask, 0);
2982}
2983
2984static inline struct page *dev_alloc_page(void)
2985{
95829b3a 2986 return dev_alloc_pages(0);
71dfda58
AD
2987}
2988
bc38f30f
AL
2989/**
2990 * dev_page_is_reusable - check whether a page can be reused for network Rx
2991 * @page: the page to test
2992 *
2993 * A page shouldn't be considered for reusing/recycling if it was allocated
2994 * under memory pressure or at a distant memory node.
2995 *
2996 * Returns false if this page should be returned to page allocator, true
2997 * otherwise.
2998 */
2999static inline bool dev_page_is_reusable(const struct page *page)
3000{
3001 return likely(page_to_nid(page) == numa_mem_id() &&
3002 !page_is_pfmemalloc(page));
3003}
3004
0614002b
MG
3005/**
3006 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3007 * @page: The page that was allocated from skb_alloc_page
3008 * @skb: The skb that may need pfmemalloc set
3009 */
48f971c9
AL
3010static inline void skb_propagate_pfmemalloc(const struct page *page,
3011 struct sk_buff *skb)
0614002b 3012{
2f064f34 3013 if (page_is_pfmemalloc(page))
0614002b
MG
3014 skb->pfmemalloc = true;
3015}
3016
7240b60c
JL
3017/**
3018 * skb_frag_off() - Returns the offset of a skb fragment
3019 * @frag: the paged fragment
3020 */
3021static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3022{
65c84f14 3023 return frag->bv_offset;
7240b60c
JL
3024}
3025
3026/**
3027 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3028 * @frag: skb fragment
3029 * @delta: value to add
3030 */
3031static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3032{
65c84f14 3033 frag->bv_offset += delta;
7240b60c
JL
3034}
3035
3036/**
3037 * skb_frag_off_set() - Sets the offset of a skb fragment
3038 * @frag: skb fragment
3039 * @offset: offset of fragment
3040 */
3041static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3042{
65c84f14 3043 frag->bv_offset = offset;
7240b60c
JL
3044}
3045
3046/**
3047 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3048 * @fragto: skb fragment where offset is set
3049 * @fragfrom: skb fragment offset is copied from
3050 */
3051static inline void skb_frag_off_copy(skb_frag_t *fragto,
3052 const skb_frag_t *fragfrom)
3053{
65c84f14 3054 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
3055}
3056
131ea667 3057/**
e227867f 3058 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
3059 * @frag: the paged fragment
3060 *
3061 * Returns the &struct page associated with @frag.
3062 */
3063static inline struct page *skb_frag_page(const skb_frag_t *frag)
3064{
1dfa5bd3 3065 return frag->bv_page;
131ea667
IC
3066}
3067
3068/**
3069 * __skb_frag_ref - take an addition reference on a paged fragment.
3070 * @frag: the paged fragment
3071 *
3072 * Takes an additional reference on the paged fragment @frag.
3073 */
3074static inline void __skb_frag_ref(skb_frag_t *frag)
3075{
3076 get_page(skb_frag_page(frag));
3077}
3078
3079/**
3080 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3081 * @skb: the buffer
3082 * @f: the fragment offset.
3083 *
3084 * Takes an additional reference on the @f'th paged fragment of @skb.
3085 */
3086static inline void skb_frag_ref(struct sk_buff *skb, int f)
3087{
3088 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3089}
3090
3091/**
3092 * __skb_frag_unref - release a reference on a paged fragment.
3093 * @frag: the paged fragment
c420c989 3094 * @recycle: recycle the page if allocated via page_pool
131ea667 3095 *
c420c989
MC
3096 * Releases a reference on the paged fragment @frag
3097 * or recycles the page via the page_pool API.
131ea667 3098 */
c420c989 3099static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
131ea667 3100{
6a5bcd84
IA
3101 struct page *page = skb_frag_page(frag);
3102
3103#ifdef CONFIG_PAGE_POOL
3104 if (recycle && page_pool_return_skb_page(page))
3105 return;
3106#endif
3107 put_page(page);
131ea667
IC
3108}
3109
3110/**
3111 * skb_frag_unref - release a reference on a paged fragment of an skb.
3112 * @skb: the buffer
3113 * @f: the fragment offset
3114 *
3115 * Releases a reference on the @f'th paged fragment of @skb.
3116 */
3117static inline void skb_frag_unref(struct sk_buff *skb, int f)
3118{
6a5bcd84 3119 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
131ea667
IC
3120}
3121
3122/**
3123 * skb_frag_address - gets the address of the data contained in a paged fragment
3124 * @frag: the paged fragment buffer
3125 *
3126 * Returns the address of the data within @frag. The page must already
3127 * be mapped.
3128 */
3129static inline void *skb_frag_address(const skb_frag_t *frag)
3130{
7240b60c 3131 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3132}
3133
3134/**
3135 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3136 * @frag: the paged fragment buffer
3137 *
3138 * Returns the address of the data within @frag. Checks that the page
3139 * is mapped and returns %NULL otherwise.
3140 */
3141static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3142{
3143 void *ptr = page_address(skb_frag_page(frag));
3144 if (unlikely(!ptr))
3145 return NULL;
3146
7240b60c
JL
3147 return ptr + skb_frag_off(frag);
3148}
3149
3150/**
3151 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3152 * @fragto: skb fragment where page is set
3153 * @fragfrom: skb fragment page is copied from
3154 */
3155static inline void skb_frag_page_copy(skb_frag_t *fragto,
3156 const skb_frag_t *fragfrom)
3157{
3158 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3159}
3160
3161/**
3162 * __skb_frag_set_page - sets the page contained in a paged fragment
3163 * @frag: the paged fragment
3164 * @page: the page to set
3165 *
3166 * Sets the fragment @frag to contain @page.
3167 */
3168static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3169{
1dfa5bd3 3170 frag->bv_page = page;
131ea667
IC
3171}
3172
3173/**
3174 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3175 * @skb: the buffer
3176 * @f: the fragment offset
3177 * @page: the page to set
3178 *
3179 * Sets the @f'th fragment of @skb to contain @page.
3180 */
3181static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3182 struct page *page)
3183{
3184 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3185}
3186
400dfd3a
ED
3187bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3188
131ea667
IC
3189/**
3190 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3191 * @dev: the device to map the fragment to
131ea667
IC
3192 * @frag: the paged fragment to map
3193 * @offset: the offset within the fragment (starting at the
3194 * fragment's own offset)
3195 * @size: the number of bytes to map
771b00a8 3196 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3197 *
3198 * Maps the page associated with @frag to @device.
3199 */
3200static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3201 const skb_frag_t *frag,
3202 size_t offset, size_t size,
3203 enum dma_data_direction dir)
3204{
3205 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3206 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3207}
3208
117632e6
ED
3209static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3210 gfp_t gfp_mask)
3211{
3212 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3213}
3214
bad93e9d
OP
3215
3216static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3217 gfp_t gfp_mask)
3218{
3219 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3220}
3221
3222
334a8132
PM
3223/**
3224 * skb_clone_writable - is the header of a clone writable
3225 * @skb: buffer to check
3226 * @len: length up to which to write
3227 *
3228 * Returns true if modifying the header part of the cloned buffer
3229 * does not requires the data to be copied.
3230 */
05bdd2f1 3231static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3232{
3233 return !skb_header_cloned(skb) &&
3234 skb_headroom(skb) + len <= skb->hdr_len;
3235}
3236
3697649f
DB
3237static inline int skb_try_make_writable(struct sk_buff *skb,
3238 unsigned int write_len)
3239{
3240 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3241 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3242}
3243
d9cc2048
HX
3244static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3245 int cloned)
3246{
3247 int delta = 0;
3248
d9cc2048
HX
3249 if (headroom > skb_headroom(skb))
3250 delta = headroom - skb_headroom(skb);
3251
3252 if (delta || cloned)
3253 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3254 GFP_ATOMIC);
3255 return 0;
3256}
3257
1da177e4
LT
3258/**
3259 * skb_cow - copy header of skb when it is required
3260 * @skb: buffer to cow
3261 * @headroom: needed headroom
3262 *
3263 * If the skb passed lacks sufficient headroom or its data part
3264 * is shared, data is reallocated. If reallocation fails, an error
3265 * is returned and original skb is not changed.
3266 *
3267 * The result is skb with writable area skb->head...skb->tail
3268 * and at least @headroom of space at head.
3269 */
3270static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3271{
d9cc2048
HX
3272 return __skb_cow(skb, headroom, skb_cloned(skb));
3273}
1da177e4 3274
d9cc2048
HX
3275/**
3276 * skb_cow_head - skb_cow but only making the head writable
3277 * @skb: buffer to cow
3278 * @headroom: needed headroom
3279 *
3280 * This function is identical to skb_cow except that we replace the
3281 * skb_cloned check by skb_header_cloned. It should be used when
3282 * you only need to push on some header and do not need to modify
3283 * the data.
3284 */
3285static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3286{
3287 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3288}
3289
3290/**
3291 * skb_padto - pad an skbuff up to a minimal size
3292 * @skb: buffer to pad
3293 * @len: minimal length
3294 *
3295 * Pads up a buffer to ensure the trailing bytes exist and are
3296 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3297 * is untouched. Otherwise it is extended. Returns zero on
3298 * success. The skb is freed on error.
1da177e4 3299 */
5b057c6b 3300static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3301{
3302 unsigned int size = skb->len;
3303 if (likely(size >= len))
5b057c6b 3304 return 0;
987c402a 3305 return skb_pad(skb, len - size);
1da177e4
LT
3306}
3307
9c0c1124 3308/**
4ea7b0cf 3309 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3310 * @skb: buffer to pad
3311 * @len: minimal length
cd0a137a 3312 * @free_on_error: free buffer on error
9c0c1124
AD
3313 *
3314 * Pads up a buffer to ensure the trailing bytes exist and are
3315 * blanked. If the buffer already contains sufficient data it
3316 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3317 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3318 */
4a009cb0
ED
3319static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3320 unsigned int len,
3321 bool free_on_error)
9c0c1124
AD
3322{
3323 unsigned int size = skb->len;
3324
3325 if (unlikely(size < len)) {
3326 len -= size;
cd0a137a 3327 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3328 return -ENOMEM;
3329 __skb_put(skb, len);
3330 }
3331 return 0;
3332}
3333
cd0a137a
FF
3334/**
3335 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3336 * @skb: buffer to pad
3337 * @len: minimal length
3338 *
3339 * Pads up a buffer to ensure the trailing bytes exist and are
3340 * blanked. If the buffer already contains sufficient data it
3341 * is untouched. Otherwise it is extended. Returns zero on
3342 * success. The skb is freed on error.
3343 */
4a009cb0 3344static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3345{
3346 return __skb_put_padto(skb, len, true);
3347}
3348
1da177e4 3349static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3350 struct iov_iter *from, int copy)
1da177e4
LT
3351{
3352 const int off = skb->len;
3353
3354 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3355 __wsum csum = 0;
15e6cb46
AV
3356 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3357 &csum, from)) {
1da177e4
LT
3358 skb->csum = csum_block_add(skb->csum, csum, off);
3359 return 0;
3360 }
15e6cb46 3361 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3362 return 0;
3363
3364 __skb_trim(skb, off);
3365 return -EFAULT;
3366}
3367
38ba0a65
ED
3368static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3369 const struct page *page, int off)
1da177e4 3370{
1f8b977a
WB
3371 if (skb_zcopy(skb))
3372 return false;
1da177e4 3373 if (i) {
d8e18a51 3374 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3375
ea2ab693 3376 return page == skb_frag_page(frag) &&
7240b60c 3377 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3378 }
38ba0a65 3379 return false;
1da177e4
LT
3380}
3381
364c6bad
HX
3382static inline int __skb_linearize(struct sk_buff *skb)
3383{
3384 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3385}
3386
1da177e4
LT
3387/**
3388 * skb_linearize - convert paged skb to linear one
3389 * @skb: buffer to linarize
1da177e4
LT
3390 *
3391 * If there is no free memory -ENOMEM is returned, otherwise zero
3392 * is returned and the old skb data released.
3393 */
364c6bad
HX
3394static inline int skb_linearize(struct sk_buff *skb)
3395{
3396 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3397}
3398
cef401de
ED
3399/**
3400 * skb_has_shared_frag - can any frag be overwritten
3401 * @skb: buffer to test
3402 *
3403 * Return true if the skb has at least one frag that might be modified
3404 * by an external entity (as in vmsplice()/sendfile())
3405 */
3406static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3407{
c9af6db4 3408 return skb_is_nonlinear(skb) &&
06b4feb3 3409 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3410}
3411
364c6bad
HX
3412/**
3413 * skb_linearize_cow - make sure skb is linear and writable
3414 * @skb: buffer to process
3415 *
3416 * If there is no free memory -ENOMEM is returned, otherwise zero
3417 * is returned and the old skb data released.
3418 */
3419static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3420{
364c6bad
HX
3421 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3422 __skb_linearize(skb) : 0;
1da177e4
LT
3423}
3424
479ffccc
DB
3425static __always_inline void
3426__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3427 unsigned int off)
3428{
3429 if (skb->ip_summed == CHECKSUM_COMPLETE)
3430 skb->csum = csum_block_sub(skb->csum,
3431 csum_partial(start, len, 0), off);
3432 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3433 skb_checksum_start_offset(skb) < 0)
3434 skb->ip_summed = CHECKSUM_NONE;
3435}
3436
1da177e4
LT
3437/**
3438 * skb_postpull_rcsum - update checksum for received skb after pull
3439 * @skb: buffer to update
3440 * @start: start of data before pull
3441 * @len: length of data pulled
3442 *
3443 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3444 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3445 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3446 */
1da177e4 3447static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3448 const void *start, unsigned int len)
1da177e4 3449{
479ffccc 3450 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3451}
3452
479ffccc
DB
3453static __always_inline void
3454__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3455 unsigned int off)
3456{
3457 if (skb->ip_summed == CHECKSUM_COMPLETE)
3458 skb->csum = csum_block_add(skb->csum,
3459 csum_partial(start, len, 0), off);
3460}
cbb042f9 3461
479ffccc
DB
3462/**
3463 * skb_postpush_rcsum - update checksum for received skb after push
3464 * @skb: buffer to update
3465 * @start: start of data after push
3466 * @len: length of data pushed
3467 *
3468 * After doing a push on a received packet, you need to call this to
3469 * update the CHECKSUM_COMPLETE checksum.
3470 */
f8ffad69
DB
3471static inline void skb_postpush_rcsum(struct sk_buff *skb,
3472 const void *start, unsigned int len)
3473{
479ffccc 3474 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3475}
3476
af72868b 3477void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3478
82a31b92
WC
3479/**
3480 * skb_push_rcsum - push skb and update receive checksum
3481 * @skb: buffer to update
3482 * @len: length of data pulled
3483 *
3484 * This function performs an skb_push on the packet and updates
3485 * the CHECKSUM_COMPLETE checksum. It should be used on
3486 * receive path processing instead of skb_push unless you know
3487 * that the checksum difference is zero (e.g., a valid IP header)
3488 * or you are setting ip_summed to CHECKSUM_NONE.
3489 */
d58ff351 3490static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3491{
3492 skb_push(skb, len);
3493 skb_postpush_rcsum(skb, skb->data, len);
3494 return skb->data;
3495}
3496
88078d98 3497int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3498/**
3499 * pskb_trim_rcsum - trim received skb and update checksum
3500 * @skb: buffer to trim
3501 * @len: new length
3502 *
3503 * This is exactly the same as pskb_trim except that it ensures the
3504 * checksum of received packets are still valid after the operation.
6c57f045 3505 * It can change skb pointers.
7ce5a27f
DM
3506 */
3507
3508static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3509{
3510 if (likely(len >= skb->len))
3511 return 0;
88078d98 3512 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3513}
3514
5293efe6
DB
3515static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3516{
3517 if (skb->ip_summed == CHECKSUM_COMPLETE)
3518 skb->ip_summed = CHECKSUM_NONE;
3519 __skb_trim(skb, len);
3520 return 0;
3521}
3522
3523static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3524{
3525 if (skb->ip_summed == CHECKSUM_COMPLETE)
3526 skb->ip_summed = CHECKSUM_NONE;
3527 return __skb_grow(skb, len);
3528}
3529
18a4c0ea
ED
3530#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3531#define skb_rb_first(root) rb_to_skb(rb_first(root))
3532#define skb_rb_last(root) rb_to_skb(rb_last(root))
3533#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3534#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3535
1da177e4
LT
3536#define skb_queue_walk(queue, skb) \
3537 for (skb = (queue)->next; \
a1e4891f 3538 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3539 skb = skb->next)
3540
46f8914e
JC
3541#define skb_queue_walk_safe(queue, skb, tmp) \
3542 for (skb = (queue)->next, tmp = skb->next; \
3543 skb != (struct sk_buff *)(queue); \
3544 skb = tmp, tmp = skb->next)
3545
1164f52a 3546#define skb_queue_walk_from(queue, skb) \
a1e4891f 3547 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3548 skb = skb->next)
3549
18a4c0ea
ED
3550#define skb_rbtree_walk(skb, root) \
3551 for (skb = skb_rb_first(root); skb != NULL; \
3552 skb = skb_rb_next(skb))
3553
3554#define skb_rbtree_walk_from(skb) \
3555 for (; skb != NULL; \
3556 skb = skb_rb_next(skb))
3557
3558#define skb_rbtree_walk_from_safe(skb, tmp) \
3559 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3560 skb = tmp)
3561
1164f52a
DM
3562#define skb_queue_walk_from_safe(queue, skb, tmp) \
3563 for (tmp = skb->next; \
3564 skb != (struct sk_buff *)(queue); \
3565 skb = tmp, tmp = skb->next)
3566
300ce174
SH
3567#define skb_queue_reverse_walk(queue, skb) \
3568 for (skb = (queue)->prev; \
a1e4891f 3569 skb != (struct sk_buff *)(queue); \
300ce174
SH
3570 skb = skb->prev)
3571
686a2955
DM
3572#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3573 for (skb = (queue)->prev, tmp = skb->prev; \
3574 skb != (struct sk_buff *)(queue); \
3575 skb = tmp, tmp = skb->prev)
3576
3577#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3578 for (tmp = skb->prev; \
3579 skb != (struct sk_buff *)(queue); \
3580 skb = tmp, tmp = skb->prev)
1da177e4 3581
21dc3301 3582static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3583{
3584 return skb_shinfo(skb)->frag_list != NULL;
3585}
3586
3587static inline void skb_frag_list_init(struct sk_buff *skb)
3588{
3589 skb_shinfo(skb)->frag_list = NULL;
3590}
3591
ee039871
DM
3592#define skb_walk_frags(skb, iter) \
3593 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3594
ea3793ee 3595
b50b0580
SD
3596int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3597 int *err, long *timeo_p,
ea3793ee 3598 const struct sk_buff *skb);
65101aec
PA
3599struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3600 struct sk_buff_head *queue,
3601 unsigned int flags,
fd69c399 3602 int *off, int *err,
65101aec 3603 struct sk_buff **last);
b50b0580
SD
3604struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3605 struct sk_buff_head *queue,
e427cad6 3606 unsigned int flags, int *off, int *err,
ea3793ee 3607 struct sk_buff **last);
b50b0580
SD
3608struct sk_buff *__skb_recv_datagram(struct sock *sk,
3609 struct sk_buff_head *sk_queue,
e427cad6 3610 unsigned int flags, int *off, int *err);
7965bd4d
JP
3611struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3612 int *err);
a11e1d43
LT
3613__poll_t datagram_poll(struct file *file, struct socket *sock,
3614 struct poll_table_struct *wait);
c0371da6
AV
3615int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3616 struct iov_iter *to, int size);
51f3d02b
DM
3617static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3618 struct msghdr *msg, int size)
3619{
e5a4b0bb 3620 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3621}
e5a4b0bb
AV
3622int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3623 struct msghdr *msg);
65d69e25
SG
3624int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3625 struct iov_iter *to, int len,
3626 struct ahash_request *hash);
3a654f97
AV
3627int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3628 struct iov_iter *from, int len);
3a654f97 3629int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3630void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3631void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3632static inline void skb_free_datagram_locked(struct sock *sk,
3633 struct sk_buff *skb)
3634{
3635 __skb_free_datagram_locked(sk, skb, 0);
3636}
7965bd4d 3637int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3638int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3639int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3640__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 3641 int len);
a60e3cc7 3642int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3643 struct pipe_inode_info *pipe, unsigned int len,
25869262 3644 unsigned int flags);
20bf50de
TH
3645int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3646 int len);
0739cd28 3647int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
7965bd4d 3648void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3649unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3650int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3651 int len, int hlen);
7965bd4d
JP
3652void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3653int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3654void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3655bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3656bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3657struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3658struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3659 unsigned int offset);
0d5501c1 3660struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3661int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3662int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3663int skb_vlan_pop(struct sk_buff *skb);
3664int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
3665int skb_eth_pop(struct sk_buff *skb);
3666int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3667 const unsigned char *src);
fa4e0f88 3668int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3669 int mac_len, bool ethernet);
040b5cfb
MV
3670int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3671 bool ethernet);
d27cf5c5 3672int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3673int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3674struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3675 gfp_t gfp);
20380731 3676
6ce8e9ce
AV
3677static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3678{
3073f070 3679 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3680}
3681
7eab8d9e
AV
3682static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3683{
e5a4b0bb 3684 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3685}
3686
2817a336
DB
3687struct skb_checksum_ops {
3688 __wsum (*update)(const void *mem, int len, __wsum wsum);
3689 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3690};
3691
9617813d
DC
3692extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3693
2817a336
DB
3694__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3695 __wsum csum, const struct skb_checksum_ops *ops);
3696__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3697 __wsum csum);
3698
1e98a0f0 3699static inline void * __must_check
e3305138
AL
3700__skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3701 const void *data, int hlen, void *buffer)
1da177e4 3702{
d206121f 3703 if (likely(hlen - offset >= len))
e3305138 3704 return (void *)data + offset;
1da177e4 3705
d206121f 3706 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
1da177e4
LT
3707 return NULL;
3708
3709 return buffer;
3710}
3711
1e98a0f0
ED
3712static inline void * __must_check
3713skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3714{
3715 return __skb_header_pointer(skb, offset, len, skb->data,
3716 skb_headlen(skb), buffer);
3717}
3718
4262e5cc
DB
3719/**
3720 * skb_needs_linearize - check if we need to linearize a given skb
3721 * depending on the given device features.
3722 * @skb: socket buffer to check
3723 * @features: net device features
3724 *
3725 * Returns true if either:
3726 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3727 * 2. skb is fragmented and the device does not support SG.
3728 */
3729static inline bool skb_needs_linearize(struct sk_buff *skb,
3730 netdev_features_t features)
3731{
3732 return skb_is_nonlinear(skb) &&
3733 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3734 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3735}
3736
d626f62b
ACM
3737static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3738 void *to,
3739 const unsigned int len)
3740{
3741 memcpy(to, skb->data, len);
3742}
3743
3744static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3745 const int offset, void *to,
3746 const unsigned int len)
3747{
3748 memcpy(to, skb->data + offset, len);
3749}
3750
27d7ff46
ACM
3751static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3752 const void *from,
3753 const unsigned int len)
3754{
3755 memcpy(skb->data, from, len);
3756}
3757
3758static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3759 const int offset,
3760 const void *from,
3761 const unsigned int len)
3762{
3763 memcpy(skb->data + offset, from, len);
3764}
3765
7965bd4d 3766void skb_init(void);
1da177e4 3767
ac45f602
PO
3768static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3769{
3770 return skb->tstamp;
3771}
3772
a61bbcf2
PM
3773/**
3774 * skb_get_timestamp - get timestamp from a skb
3775 * @skb: skb to get stamp from
13c6ee2a 3776 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3777 *
3778 * Timestamps are stored in the skb as offsets to a base timestamp.
3779 * This function converts the offset back to a struct timeval and stores
3780 * it in stamp.
3781 */
ac45f602 3782static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3783 struct __kernel_old_timeval *stamp)
a61bbcf2 3784{
13c6ee2a 3785 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3786}
3787
887feae3
DD
3788static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3789 struct __kernel_sock_timeval *stamp)
3790{
3791 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3792
3793 stamp->tv_sec = ts.tv_sec;
3794 stamp->tv_usec = ts.tv_nsec / 1000;
3795}
3796
ac45f602 3797static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3798 struct __kernel_old_timespec *stamp)
ac45f602 3799{
df1b4ba9
AB
3800 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3801
3802 stamp->tv_sec = ts.tv_sec;
3803 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3804}
3805
887feae3
DD
3806static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3807 struct __kernel_timespec *stamp)
3808{
3809 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3810
3811 stamp->tv_sec = ts.tv_sec;
3812 stamp->tv_nsec = ts.tv_nsec;
3813}
3814
b7aa0bf7 3815static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3816{
b7aa0bf7 3817 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3818}
3819
164891aa
SH
3820static inline ktime_t net_timedelta(ktime_t t)
3821{
3822 return ktime_sub(ktime_get_real(), t);
3823}
3824
b9ce204f
IJ
3825static inline ktime_t net_invalid_timestamp(void)
3826{
8b0e1953 3827 return 0;
b9ce204f 3828}
a61bbcf2 3829
de8f3a83
DB
3830static inline u8 skb_metadata_len(const struct sk_buff *skb)
3831{
3832 return skb_shinfo(skb)->meta_len;
3833}
3834
3835static inline void *skb_metadata_end(const struct sk_buff *skb)
3836{
3837 return skb_mac_header(skb);
3838}
3839
3840static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3841 const struct sk_buff *skb_b,
3842 u8 meta_len)
3843{
3844 const void *a = skb_metadata_end(skb_a);
3845 const void *b = skb_metadata_end(skb_b);
3846 /* Using more efficient varaiant than plain call to memcmp(). */
3847#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3848 u64 diffs = 0;
3849
3850 switch (meta_len) {
3851#define __it(x, op) (x -= sizeof(u##op))
3852#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3853 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3854 fallthrough;
de8f3a83 3855 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3856 fallthrough;
de8f3a83 3857 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3858 fallthrough;
de8f3a83
DB
3859 case 8: diffs |= __it_diff(a, b, 64);
3860 break;
3861 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3862 fallthrough;
de8f3a83 3863 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3864 fallthrough;
de8f3a83 3865 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3866 fallthrough;
de8f3a83
DB
3867 case 4: diffs |= __it_diff(a, b, 32);
3868 break;
3869 }
3870 return diffs;
3871#else
3872 return memcmp(a - meta_len, b - meta_len, meta_len);
3873#endif
3874}
3875
3876static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3877 const struct sk_buff *skb_b)
3878{
3879 u8 len_a = skb_metadata_len(skb_a);
3880 u8 len_b = skb_metadata_len(skb_b);
3881
3882 if (!(len_a | len_b))
3883 return false;
3884
3885 return len_a != len_b ?
3886 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3887}
3888
3889static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3890{
3891 skb_shinfo(skb)->meta_len = meta_len;
3892}
3893
3894static inline void skb_metadata_clear(struct sk_buff *skb)
3895{
3896 skb_metadata_set(skb, 0);
3897}
3898
62bccb8c
AD
3899struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3900
c1f19b51
RC
3901#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3902
7965bd4d
JP
3903void skb_clone_tx_timestamp(struct sk_buff *skb);
3904bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3905
3906#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3907
3908static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3909{
3910}
3911
3912static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3913{
3914 return false;
3915}
3916
3917#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3918
3919/**
3920 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3921 *
da92b194
RC
3922 * PHY drivers may accept clones of transmitted packets for
3923 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3924 * must call this function to return the skb back to the stack with a
3925 * timestamp.
da92b194 3926 *
2ff17117 3927 * @skb: clone of the original outgoing packet
7a76a021 3928 * @hwtstamps: hardware time stamps
c1f19b51
RC
3929 *
3930 */
3931void skb_complete_tx_timestamp(struct sk_buff *skb,
3932 struct skb_shared_hwtstamps *hwtstamps);
3933
e7ed11ee 3934void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
3935 struct skb_shared_hwtstamps *hwtstamps,
3936 struct sock *sk, int tstype);
3937
ac45f602
PO
3938/**
3939 * skb_tstamp_tx - queue clone of skb with send time stamps
3940 * @orig_skb: the original outgoing packet
3941 * @hwtstamps: hardware time stamps, may be NULL if not available
3942 *
3943 * If the skb has a socket associated, then this function clones the
3944 * skb (thus sharing the actual data and optional structures), stores
3945 * the optional hardware time stamping information (if non NULL) or
3946 * generates a software time stamp (otherwise), then queues the clone
3947 * to the error queue of the socket. Errors are silently ignored.
3948 */
7965bd4d
JP
3949void skb_tstamp_tx(struct sk_buff *orig_skb,
3950 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3951
4507a715
RC
3952/**
3953 * skb_tx_timestamp() - Driver hook for transmit timestamping
3954 *
3955 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3956 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3957 *
73409f3b
DM
3958 * Specifically, one should make absolutely sure that this function is
3959 * called before TX completion of this packet can trigger. Otherwise
3960 * the packet could potentially already be freed.
3961 *
4507a715
RC
3962 * @skb: A socket buffer.
3963 */
3964static inline void skb_tx_timestamp(struct sk_buff *skb)
3965{
c1f19b51 3966 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3967 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3968 skb_tstamp_tx(skb, NULL);
4507a715
RC
3969}
3970
6e3e939f
JB
3971/**
3972 * skb_complete_wifi_ack - deliver skb with wifi status
3973 *
3974 * @skb: the original outgoing packet
3975 * @acked: ack status
3976 *
3977 */
3978void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3979
7965bd4d
JP
3980__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3981__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3982
60476372
HX
3983static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3984{
6edec0e6
TH
3985 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3986 skb->csum_valid ||
3987 (skb->ip_summed == CHECKSUM_PARTIAL &&
3988 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3989}
3990
fb286bb2
HX
3991/**
3992 * skb_checksum_complete - Calculate checksum of an entire packet
3993 * @skb: packet to process
3994 *
3995 * This function calculates the checksum over the entire packet plus
3996 * the value of skb->csum. The latter can be used to supply the
3997 * checksum of a pseudo header as used by TCP/UDP. It returns the
3998 * checksum.
3999 *
4000 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4001 * this function can be used to verify that checksum on received
4002 * packets. In that case the function should return zero if the
4003 * checksum is correct. In particular, this function will return zero
4004 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4005 * hardware has already verified the correctness of the checksum.
4006 */
4381ca3c 4007static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 4008{
60476372
HX
4009 return skb_csum_unnecessary(skb) ?
4010 0 : __skb_checksum_complete(skb);
fb286bb2
HX
4011}
4012
77cffe23
TH
4013static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4014{
4015 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4016 if (skb->csum_level == 0)
4017 skb->ip_summed = CHECKSUM_NONE;
4018 else
4019 skb->csum_level--;
4020 }
4021}
4022
4023static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4024{
4025 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4026 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4027 skb->csum_level++;
4028 } else if (skb->ip_summed == CHECKSUM_NONE) {
4029 skb->ip_summed = CHECKSUM_UNNECESSARY;
4030 skb->csum_level = 0;
4031 }
4032}
4033
836e66c2
DB
4034static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4035{
4036 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4037 skb->ip_summed = CHECKSUM_NONE;
4038 skb->csum_level = 0;
4039 }
4040}
4041
76ba0aae
TH
4042/* Check if we need to perform checksum complete validation.
4043 *
4044 * Returns true if checksum complete is needed, false otherwise
4045 * (either checksum is unnecessary or zero checksum is allowed).
4046 */
4047static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4048 bool zero_okay,
4049 __sum16 check)
4050{
5d0c2b95
TH
4051 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4052 skb->csum_valid = 1;
77cffe23 4053 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
4054 return false;
4055 }
4056
4057 return true;
4058}
4059
da279887 4060/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
4061 * in checksum_init.
4062 */
4063#define CHECKSUM_BREAK 76
4064
4e18b9ad
TH
4065/* Unset checksum-complete
4066 *
4067 * Unset checksum complete can be done when packet is being modified
4068 * (uncompressed for instance) and checksum-complete value is
4069 * invalidated.
4070 */
4071static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4072{
4073 if (skb->ip_summed == CHECKSUM_COMPLETE)
4074 skb->ip_summed = CHECKSUM_NONE;
4075}
4076
76ba0aae
TH
4077/* Validate (init) checksum based on checksum complete.
4078 *
4079 * Return values:
4080 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4081 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4082 * checksum is stored in skb->csum for use in __skb_checksum_complete
4083 * non-zero: value of invalid checksum
4084 *
4085 */
4086static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4087 bool complete,
4088 __wsum psum)
4089{
4090 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4091 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4092 skb->csum_valid = 1;
76ba0aae
TH
4093 return 0;
4094 }
4095 }
4096
4097 skb->csum = psum;
4098
5d0c2b95
TH
4099 if (complete || skb->len <= CHECKSUM_BREAK) {
4100 __sum16 csum;
4101
4102 csum = __skb_checksum_complete(skb);
4103 skb->csum_valid = !csum;
4104 return csum;
4105 }
76ba0aae
TH
4106
4107 return 0;
4108}
4109
4110static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4111{
4112 return 0;
4113}
4114
4115/* Perform checksum validate (init). Note that this is a macro since we only
4116 * want to calculate the pseudo header which is an input function if necessary.
4117 * First we try to validate without any computation (checksum unnecessary) and
4118 * then calculate based on checksum complete calling the function to compute
4119 * pseudo header.
4120 *
4121 * Return values:
4122 * 0: checksum is validated or try to in skb_checksum_complete
4123 * non-zero: value of invalid checksum
4124 */
4125#define __skb_checksum_validate(skb, proto, complete, \
4126 zero_okay, check, compute_pseudo) \
4127({ \
4128 __sum16 __ret = 0; \
5d0c2b95 4129 skb->csum_valid = 0; \
76ba0aae
TH
4130 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4131 __ret = __skb_checksum_validate_complete(skb, \
4132 complete, compute_pseudo(skb, proto)); \
4133 __ret; \
4134})
4135
4136#define skb_checksum_init(skb, proto, compute_pseudo) \
4137 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4138
4139#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4140 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4141
4142#define skb_checksum_validate(skb, proto, compute_pseudo) \
4143 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4144
4145#define skb_checksum_validate_zero_check(skb, proto, check, \
4146 compute_pseudo) \
096a4cfa 4147 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4148
4149#define skb_checksum_simple_validate(skb) \
4150 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4151
d96535a1
TH
4152static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4153{
219f1d79 4154 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4155}
4156
e4aa33ad 4157static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4158{
4159 skb->csum = ~pseudo;
4160 skb->ip_summed = CHECKSUM_COMPLETE;
4161}
4162
e4aa33ad 4163#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4164do { \
4165 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4166 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4167} while (0)
4168
15e2396d
TH
4169static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4170 u16 start, u16 offset)
4171{
4172 skb->ip_summed = CHECKSUM_PARTIAL;
4173 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4174 skb->csum_offset = offset - start;
4175}
4176
dcdc8994
TH
4177/* Update skbuf and packet to reflect the remote checksum offload operation.
4178 * When called, ptr indicates the starting point for skb->csum when
4179 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4180 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4181 */
4182static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4183 int start, int offset, bool nopartial)
dcdc8994
TH
4184{
4185 __wsum delta;
4186
15e2396d
TH
4187 if (!nopartial) {
4188 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4189 return;
4190 }
4191
dcdc8994
TH
4192 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4193 __skb_checksum_complete(skb);
4194 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4195 }
4196
4197 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4198
4199 /* Adjust skb->csum since we changed the packet */
4200 skb->csum = csum_add(skb->csum, delta);
4201}
4202
cb9c6836
FW
4203static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4204{
4205#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4206 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4207#else
4208 return NULL;
4209#endif
4210}
4211
261db6c2 4212static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4213{
261db6c2
JS
4214#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4215 return skb->_nfct;
4216#else
4217 return 0UL;
4218#endif
1da177e4 4219}
261db6c2
JS
4220
4221static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4222{
261db6c2 4223#if IS_ENABLED(CONFIG_NF_CONNTRACK)
5fc88f93 4224 skb->slow_gro |= !!nfct;
261db6c2 4225 skb->_nfct = nfct;
2fc72c7b 4226#endif
261db6c2 4227}
df5042f4
FW
4228
4229#ifdef CONFIG_SKB_EXTENSIONS
4230enum skb_ext_id {
4231#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4232 SKB_EXT_BRIDGE_NF,
4165079b
FW
4233#endif
4234#ifdef CONFIG_XFRM
4235 SKB_EXT_SEC_PATH,
95a7233c
PB
4236#endif
4237#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4238 TC_SKB_EXT,
3ee17bc7
MM
4239#endif
4240#if IS_ENABLED(CONFIG_MPTCP)
4241 SKB_EXT_MPTCP,
df5042f4
FW
4242#endif
4243 SKB_EXT_NUM, /* must be last */
4244};
4245
4246/**
4247 * struct skb_ext - sk_buff extensions
4248 * @refcnt: 1 on allocation, deallocated on 0
4249 * @offset: offset to add to @data to obtain extension address
4250 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4251 * @data: start of extension data, variable sized
4252 *
4253 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4254 * to use 'u8' types while allowing up to 2kb worth of extension data.
4255 */
4256struct skb_ext {
4257 refcount_t refcnt;
4258 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4259 u8 chunks; /* same */
5c91aa1d 4260 char data[] __aligned(8);
df5042f4
FW
4261};
4262
4930f483 4263struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4264void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4265 struct skb_ext *ext);
df5042f4
FW
4266void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4267void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4268void __skb_ext_put(struct skb_ext *ext);
4269
4270static inline void skb_ext_put(struct sk_buff *skb)
4271{
4272 if (skb->active_extensions)
4273 __skb_ext_put(skb->extensions);
4274}
4275
df5042f4
FW
4276static inline void __skb_ext_copy(struct sk_buff *dst,
4277 const struct sk_buff *src)
4278{
4279 dst->active_extensions = src->active_extensions;
4280
4281 if (src->active_extensions) {
4282 struct skb_ext *ext = src->extensions;
4283
4284 refcount_inc(&ext->refcnt);
4285 dst->extensions = ext;
4286 }
4287}
4288
4289static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4290{
4291 skb_ext_put(dst);
4292 __skb_ext_copy(dst, src);
4293}
4294
4295static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4296{
4297 return !!ext->offset[i];
4298}
4299
4300static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4301{
4302 return skb->active_extensions & (1 << id);
4303}
4304
4305static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4306{
4307 if (skb_ext_exist(skb, id))
4308 __skb_ext_del(skb, id);
4309}
4310
4311static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4312{
4313 if (skb_ext_exist(skb, id)) {
4314 struct skb_ext *ext = skb->extensions;
4315
4316 return (void *)ext + (ext->offset[id] << 3);
4317 }
4318
4319 return NULL;
4320}
174e2381
FW
4321
4322static inline void skb_ext_reset(struct sk_buff *skb)
4323{
4324 if (unlikely(skb->active_extensions)) {
4325 __skb_ext_put(skb->extensions);
4326 skb->active_extensions = 0;
4327 }
4328}
677bf08c
FW
4329
4330static inline bool skb_has_extensions(struct sk_buff *skb)
4331{
4332 return unlikely(skb->active_extensions);
4333}
df5042f4
FW
4334#else
4335static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4336static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4337static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4338static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4339static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4340static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4341#endif /* CONFIG_SKB_EXTENSIONS */
4342
895b5c9f 4343static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4344{
5f79e0f9 4345#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4346 nf_conntrack_put(skb_nfct(skb));
4347 skb->_nfct = 0;
2fc72c7b 4348#endif
a193a4ab
PM
4349}
4350
124dff01
PM
4351static inline void nf_reset_trace(struct sk_buff *skb)
4352{
478b360a 4353#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4354 skb->nf_trace = 0;
4355#endif
a193a4ab
PM
4356}
4357
2b5ec1a5
YY
4358static inline void ipvs_reset(struct sk_buff *skb)
4359{
4360#if IS_ENABLED(CONFIG_IP_VS)
4361 skb->ipvs_property = 0;
4362#endif
4363}
4364
de8bda1d 4365/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4366static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4367 bool copy)
edda553c 4368{
5f79e0f9 4369#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4370 dst->_nfct = src->_nfct;
4371 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4372#endif
478b360a 4373#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4374 if (copy)
4375 dst->nf_trace = src->nf_trace;
478b360a 4376#endif
edda553c
YK
4377}
4378
e7ac05f3
YK
4379static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4380{
e7ac05f3 4381#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4382 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4383#endif
5fc88f93 4384 dst->slow_gro = src->slow_gro;
b1937227 4385 __nf_copy(dst, src, true);
e7ac05f3
YK
4386}
4387
984bc16c
JM
4388#ifdef CONFIG_NETWORK_SECMARK
4389static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4390{
4391 to->secmark = from->secmark;
4392}
4393
4394static inline void skb_init_secmark(struct sk_buff *skb)
4395{
4396 skb->secmark = 0;
4397}
4398#else
4399static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4400{ }
4401
4402static inline void skb_init_secmark(struct sk_buff *skb)
4403{ }
4404#endif
4405
7af8f4ca
FW
4406static inline int secpath_exists(const struct sk_buff *skb)
4407{
4408#ifdef CONFIG_XFRM
4165079b 4409 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4410#else
4411 return 0;
4412#endif
4413}
4414
574f7194
EB
4415static inline bool skb_irq_freeable(const struct sk_buff *skb)
4416{
4417 return !skb->destructor &&
7af8f4ca 4418 !secpath_exists(skb) &&
cb9c6836 4419 !skb_nfct(skb) &&
574f7194
EB
4420 !skb->_skb_refdst &&
4421 !skb_has_frag_list(skb);
4422}
4423
f25f4e44
PWJ
4424static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4425{
f25f4e44 4426 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4427}
4428
9247744e 4429static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4430{
4e3ab47a 4431 return skb->queue_mapping;
4e3ab47a
PE
4432}
4433
f25f4e44
PWJ
4434static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4435{
f25f4e44 4436 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4437}
4438
d5a9e24a
DM
4439static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4440{
4441 skb->queue_mapping = rx_queue + 1;
4442}
4443
9247744e 4444static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4445{
4446 return skb->queue_mapping - 1;
4447}
4448
9247744e 4449static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4450{
a02cec21 4451 return skb->queue_mapping != 0;
d5a9e24a
DM
4452}
4453
4ff06203
JA
4454static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4455{
4456 skb->dst_pending_confirm = val;
4457}
4458
4459static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4460{
4461 return skb->dst_pending_confirm != 0;
4462}
4463
2294be0f 4464static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4465{
0b3d8e08 4466#ifdef CONFIG_XFRM
4165079b 4467 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4468#else
def8b4fa 4469 return NULL;
def8b4fa 4470#endif
0b3d8e08 4471}
def8b4fa 4472
68c33163
PS
4473/* Keeps track of mac header offset relative to skb->head.
4474 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4475 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4476 * tunnel skb it points to outer mac header.
4477 * Keeps track of level of encapsulation of network headers.
4478 */
68c33163 4479struct skb_gso_cb {
802ab55a
AD
4480 union {
4481 int mac_offset;
4482 int data_offset;
4483 };
3347c960 4484 int encap_level;
76443456 4485 __wsum csum;
7e2b10c1 4486 __u16 csum_start;
68c33163 4487};
a08e7fd9
CZ
4488#define SKB_GSO_CB_OFFSET 32
4489#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4490
4491static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4492{
4493 return (skb_mac_header(inner_skb) - inner_skb->head) -
4494 SKB_GSO_CB(inner_skb)->mac_offset;
4495}
4496
1e2bd517
PS
4497static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4498{
4499 int new_headroom, headroom;
4500 int ret;
4501
4502 headroom = skb_headroom(skb);
4503 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4504 if (ret)
4505 return ret;
4506
4507 new_headroom = skb_headroom(skb);
4508 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4509 return 0;
4510}
4511
08b64fcc
AD
4512static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4513{
4514 /* Do not update partial checksums if remote checksum is enabled. */
4515 if (skb->remcsum_offload)
4516 return;
4517
4518 SKB_GSO_CB(skb)->csum = res;
4519 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4520}
4521
7e2b10c1
TH
4522/* Compute the checksum for a gso segment. First compute the checksum value
4523 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4524 * then add in skb->csum (checksum from csum_start to end of packet).
4525 * skb->csum and csum_start are then updated to reflect the checksum of the
4526 * resultant packet starting from the transport header-- the resultant checksum
4527 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4528 * header.
4529 */
4530static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4531{
76443456
AD
4532 unsigned char *csum_start = skb_transport_header(skb);
4533 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4534 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4535
76443456
AD
4536 SKB_GSO_CB(skb)->csum = res;
4537 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4538
76443456 4539 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4540}
4541
bdcc0924 4542static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4543{
4544 return skb_shinfo(skb)->gso_size;
4545}
4546
36a8f39e 4547/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4548static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4549{
4550 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4551}
4552
d02f51cb
DA
4553/* Note: Should be called only if skb_is_gso(skb) is true */
4554static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4555{
4556 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4557}
4558
4c3024de 4559/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4560static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4561{
4c3024de 4562 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4563}
4564
5293efe6
DB
4565static inline void skb_gso_reset(struct sk_buff *skb)
4566{
4567 skb_shinfo(skb)->gso_size = 0;
4568 skb_shinfo(skb)->gso_segs = 0;
4569 skb_shinfo(skb)->gso_type = 0;
4570}
4571
d02f51cb
DA
4572static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4573 u16 increment)
4574{
4575 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4576 return;
4577 shinfo->gso_size += increment;
4578}
4579
4580static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4581 u16 decrement)
4582{
4583 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4584 return;
4585 shinfo->gso_size -= decrement;
4586}
4587
7965bd4d 4588void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4589
4590static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4591{
4592 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4593 * wanted then gso_type will be set. */
05bdd2f1
ED
4594 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4595
b78462eb
AD
4596 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4597 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4598 __skb_warn_lro_forwarding(skb);
4599 return true;
4600 }
4601 return false;
4602}
4603
35fc92a9
HX
4604static inline void skb_forward_csum(struct sk_buff *skb)
4605{
4606 /* Unfortunately we don't support this one. Any brave souls? */
4607 if (skb->ip_summed == CHECKSUM_COMPLETE)
4608 skb->ip_summed = CHECKSUM_NONE;
4609}
4610
bc8acf2c
ED
4611/**
4612 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4613 * @skb: skb to check
4614 *
4615 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4616 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4617 * use this helper, to document places where we make this assertion.
4618 */
05bdd2f1 4619static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4620{
4621#ifdef DEBUG
4622 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4623#endif
4624}
4625
f35d9d8a 4626bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4627
ed1f50c3 4628int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4629struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4630 unsigned int transport_len,
4631 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4632
3a7c1ee4
AD
4633/**
4634 * skb_head_is_locked - Determine if the skb->head is locked down
4635 * @skb: skb to check
4636 *
4637 * The head on skbs build around a head frag can be removed if they are
4638 * not cloned. This function returns true if the skb head is locked down
4639 * due to either being allocated via kmalloc, or by being a clone with
4640 * multiple references to the head.
4641 */
4642static inline bool skb_head_is_locked(const struct sk_buff *skb)
4643{
4644 return !skb->head_frag || skb_cloned(skb);
4645}
fe6cc55f 4646
179bc67f
EC
4647/* Local Checksum Offload.
4648 * Compute outer checksum based on the assumption that the
4649 * inner checksum will be offloaded later.
d0dcde64 4650 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4651 * explanation of how this works.
179bc67f
EC
4652 * Fill in outer checksum adjustment (e.g. with sum of outer
4653 * pseudo-header) before calling.
4654 * Also ensure that inner checksum is in linear data area.
4655 */
4656static inline __wsum lco_csum(struct sk_buff *skb)
4657{
9e74a6da
AD
4658 unsigned char *csum_start = skb_checksum_start(skb);
4659 unsigned char *l4_hdr = skb_transport_header(skb);
4660 __wsum partial;
179bc67f
EC
4661
4662 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4663 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4664 skb->csum_offset));
4665
179bc67f 4666 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4667 * adjustment filled in by caller) and return result.
179bc67f 4668 */
9e74a6da 4669 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4670}
4671
2c64605b
PNA
4672static inline bool skb_is_redirected(const struct sk_buff *skb)
4673{
2c64605b 4674 return skb->redirected;
2c64605b
PNA
4675}
4676
4677static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4678{
2c64605b 4679 skb->redirected = 1;
11941f8a 4680#ifdef CONFIG_NET_REDIRECT
2c64605b
PNA
4681 skb->from_ingress = from_ingress;
4682 if (skb->from_ingress)
4683 skb->tstamp = 0;
4684#endif
4685}
4686
4687static inline void skb_reset_redirect(struct sk_buff *skb)
4688{
2c64605b 4689 skb->redirected = 0;
2c64605b
PNA
4690}
4691
fa821170
XL
4692static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4693{
4694 return skb->csum_not_inet;
4695}
4696
6370cc3b
AN
4697static inline void skb_set_kcov_handle(struct sk_buff *skb,
4698 const u64 kcov_handle)
4699{
fa69ee5a
ME
4700#ifdef CONFIG_KCOV
4701 skb->kcov_handle = kcov_handle;
4702#endif
6370cc3b
AN
4703}
4704
4705static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4706{
fa69ee5a
ME
4707#ifdef CONFIG_KCOV
4708 return skb->kcov_handle;
6370cc3b 4709#else
fa69ee5a
ME
4710 return 0;
4711#endif
4712}
6370cc3b 4713
6a5bcd84
IA
4714#ifdef CONFIG_PAGE_POOL
4715static inline void skb_mark_for_recycle(struct sk_buff *skb, struct page *page,
4716 struct page_pool *pp)
4717{
4718 skb->pp_recycle = 1;
4719 page_pool_store_mem_info(page, pp);
4720}
4721#endif
4722
4723static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4724{
4725 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4726 return false;
4727 return page_pool_return_skb_page(virt_to_page(data));
4728}
4729
1da177e4
LT
4730#endif /* __KERNEL__ */
4731#endif /* _LINUX_SKBUFF_H */