at v5.11-rc2 4631 lines 134 kB view raw
1/* SPDX-License-Identifier: GPL-2.0-or-later */ 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> 8 */ 9 10#ifndef _LINUX_SKBUFF_H 11#define _LINUX_SKBUFF_H 12 13#include <linux/kernel.h> 14#include <linux/compiler.h> 15#include <linux/time.h> 16#include <linux/bug.h> 17#include <linux/bvec.h> 18#include <linux/cache.h> 19#include <linux/rbtree.h> 20#include <linux/socket.h> 21#include <linux/refcount.h> 22 23#include <linux/atomic.h> 24#include <asm/types.h> 25#include <linux/spinlock.h> 26#include <linux/net.h> 27#include <linux/textsearch.h> 28#include <net/checksum.h> 29#include <linux/rcupdate.h> 30#include <linux/hrtimer.h> 31#include <linux/dma-mapping.h> 32#include <linux/netdev_features.h> 33#include <linux/sched.h> 34#include <linux/sched/clock.h> 35#include <net/flow_dissector.h> 36#include <linux/splice.h> 37#include <linux/in6.h> 38#include <linux/if_packet.h> 39#include <net/flow.h> 40#if IS_ENABLED(CONFIG_NF_CONNTRACK) 41#include <linux/netfilter/nf_conntrack_common.h> 42#endif 43 44/* The interface for checksum offload between the stack and networking drivers 45 * is as follows... 46 * 47 * A. IP checksum related features 48 * 49 * Drivers advertise checksum offload capabilities in the features of a device. 50 * From the stack's point of view these are capabilities offered by the driver. 51 * A driver typically only advertises features that it is capable of offloading 52 * to its device. 53 * 54 * The checksum related features are: 55 * 56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one 57 * IP (one's complement) checksum for any combination 58 * of protocols or protocol layering. The checksum is 59 * computed and set in a packet per the CHECKSUM_PARTIAL 60 * interface (see below). 61 * 62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain 63 * TCP or UDP packets over IPv4. These are specifically 64 * unencapsulated packets of the form IPv4|TCP or 65 * IPv4|UDP where the Protocol field in the IPv4 header 66 * is TCP or UDP. The IPv4 header may contain IP options. 67 * This feature cannot be set in features for a device 68 * with NETIF_F_HW_CSUM also set. This feature is being 69 * DEPRECATED (see below). 70 * 71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain 72 * TCP or UDP packets over IPv6. These are specifically 73 * unencapsulated packets of the form IPv6|TCP or 74 * IPv6|UDP where the Next Header field in the IPv6 75 * header is either TCP or UDP. IPv6 extension headers 76 * are not supported with this feature. This feature 77 * cannot be set in features for a device with 78 * NETIF_F_HW_CSUM also set. This feature is being 79 * DEPRECATED (see below). 80 * 81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. 82 * This flag is only used to disable the RX checksum 83 * feature for a device. The stack will accept receive 84 * checksum indication in packets received on a device 85 * regardless of whether NETIF_F_RXCSUM is set. 86 * 87 * B. Checksumming of received packets by device. Indication of checksum 88 * verification is set in skb->ip_summed. Possible values are: 89 * 90 * CHECKSUM_NONE: 91 * 92 * Device did not checksum this packet e.g. due to lack of capabilities. 93 * The packet contains full (though not verified) checksum in packet but 94 * not in skb->csum. Thus, skb->csum is undefined in this case. 95 * 96 * CHECKSUM_UNNECESSARY: 97 * 98 * The hardware you're dealing with doesn't calculate the full checksum 99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 101 * if their checksums are okay. skb->csum is still undefined in this case 102 * though. A driver or device must never modify the checksum field in the 103 * packet even if checksum is verified. 104 * 105 * CHECKSUM_UNNECESSARY is applicable to following protocols: 106 * TCP: IPv6 and IPv4. 107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 108 * zero UDP checksum for either IPv4 or IPv6, the networking stack 109 * may perform further validation in this case. 110 * GRE: only if the checksum is present in the header. 111 * SCTP: indicates the CRC in SCTP header has been validated. 112 * FCOE: indicates the CRC in FC frame has been validated. 113 * 114 * skb->csum_level indicates the number of consecutive checksums found in 115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 117 * and a device is able to verify the checksums for UDP (possibly zero), 118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to 119 * two. If the device were only able to verify the UDP checksum and not 120 * GRE, either because it doesn't support GRE checksum or because GRE 121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 122 * not considered in this case). 123 * 124 * CHECKSUM_COMPLETE: 125 * 126 * This is the most generic way. The device supplied checksum of the _whole_ 127 * packet as seen by netif_rx() and fills in skb->csum. This means the 128 * hardware doesn't need to parse L3/L4 headers to implement this. 129 * 130 * Notes: 131 * - Even if device supports only some protocols, but is able to produce 132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 134 * 135 * CHECKSUM_PARTIAL: 136 * 137 * A checksum is set up to be offloaded to a device as described in the 138 * output description for CHECKSUM_PARTIAL. This may occur on a packet 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel 140 * on the same host, or it may be set in the input path in GRO or remote 141 * checksum offload. For the purposes of checksum verification, the checksum 142 * referred to by skb->csum_start + skb->csum_offset and any preceding 143 * checksums in the packet are considered verified. Any checksums in the 144 * packet that are after the checksum being offloaded are not considered to 145 * be verified. 146 * 147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload 148 * in the skb->ip_summed for a packet. Values are: 149 * 150 * CHECKSUM_PARTIAL: 151 * 152 * The driver is required to checksum the packet as seen by hard_start_xmit() 153 * from skb->csum_start up to the end, and to record/write the checksum at 154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the 155 * csum_start and csum_offset values are valid values given the length and 156 * offset of the packet, but it should not attempt to validate that the 157 * checksum refers to a legitimate transport layer checksum -- it is the 158 * purview of the stack to validate that csum_start and csum_offset are set 159 * correctly. 160 * 161 * When the stack requests checksum offload for a packet, the driver MUST 162 * ensure that the checksum is set correctly. A driver can either offload the 163 * checksum calculation to the device, or call skb_checksum_help (in the case 164 * that the device does not support offload for a particular checksum). 165 * 166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of 167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate 168 * checksum offload capability. 169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based 170 * on network device checksumming capabilities: if a packet does not match 171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of 172 * csum_not_inet, see item D.) is called to resolve the checksum. 173 * 174 * CHECKSUM_NONE: 175 * 176 * The skb was already checksummed by the protocol, or a checksum is not 177 * required. 178 * 179 * CHECKSUM_UNNECESSARY: 180 * 181 * This has the same meaning as CHECKSUM_NONE for checksum offload on 182 * output. 183 * 184 * CHECKSUM_COMPLETE: 185 * Not used in checksum output. If a driver observes a packet with this value 186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set. 187 * 188 * D. Non-IP checksum (CRC) offloads 189 * 190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of 191 * offloading the SCTP CRC in a packet. To perform this offload the stack 192 * will set csum_start and csum_offset accordingly, set ip_summed to 193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in 194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c. 195 * A driver that supports both IP checksum offload and SCTP CRC32c offload 196 * must verify which offload is configured for a packet by testing the 197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve 198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 199 * 200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of 201 * offloading the FCOE CRC in a packet. To perform this offload the stack 202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset 203 * accordingly. Note that there is no indication in the skbuff that the 204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 205 * both IP checksum offload and FCOE CRC offload must verify which offload 206 * is configured for a packet, presumably by inspecting packet headers. 207 * 208 * E. Checksumming on output with GSO. 209 * 210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload 211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as 213 * part of the GSO operation is implied. If a checksum is being offloaded 214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and 215 * csum_offset are set to refer to the outermost checksum being offloaded 216 * (two offloaded checksums are possible with UDP encapsulation). 217 */ 218 219/* Don't change this without changing skb_csum_unnecessary! */ 220#define CHECKSUM_NONE 0 221#define CHECKSUM_UNNECESSARY 1 222#define CHECKSUM_COMPLETE 2 223#define CHECKSUM_PARTIAL 3 224 225/* Maximum value in skb->csum_level */ 226#define SKB_MAX_CSUM_LEVEL 3 227 228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 229#define SKB_WITH_OVERHEAD(X) \ 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 231#define SKB_MAX_ORDER(X, ORDER) \ 232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 235 236/* return minimum truesize of one skb containing X bytes of data */ 237#define SKB_TRUESIZE(X) ((X) + \ 238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 240 241struct ahash_request; 242struct net_device; 243struct scatterlist; 244struct pipe_inode_info; 245struct iov_iter; 246struct napi_struct; 247struct bpf_prog; 248union bpf_attr; 249struct skb_ext; 250 251#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 252struct nf_bridge_info { 253 enum { 254 BRNF_PROTO_UNCHANGED, 255 BRNF_PROTO_8021Q, 256 BRNF_PROTO_PPPOE 257 } orig_proto:8; 258 u8 pkt_otherhost:1; 259 u8 in_prerouting:1; 260 u8 bridged_dnat:1; 261 __u16 frag_max_size; 262 struct net_device *physindev; 263 264 /* always valid & non-NULL from FORWARD on, for physdev match */ 265 struct net_device *physoutdev; 266 union { 267 /* prerouting: detect dnat in orig/reply direction */ 268 __be32 ipv4_daddr; 269 struct in6_addr ipv6_daddr; 270 271 /* after prerouting + nat detected: store original source 272 * mac since neigh resolution overwrites it, only used while 273 * skb is out in neigh layer. 274 */ 275 char neigh_header[8]; 276 }; 277}; 278#endif 279 280#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 281/* Chain in tc_skb_ext will be used to share the tc chain with 282 * ovs recirc_id. It will be set to the current chain by tc 283 * and read by ovs to recirc_id. 284 */ 285struct tc_skb_ext { 286 __u32 chain; 287 __u16 mru; 288}; 289#endif 290 291struct sk_buff_head { 292 /* These two members must be first. */ 293 struct sk_buff *next; 294 struct sk_buff *prev; 295 296 __u32 qlen; 297 spinlock_t lock; 298}; 299 300struct sk_buff; 301 302/* To allow 64K frame to be packed as single skb without frag_list we 303 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 304 * buffers which do not start on a page boundary. 305 * 306 * Since GRO uses frags we allocate at least 16 regardless of page 307 * size. 308 */ 309#if (65536/PAGE_SIZE + 1) < 16 310#define MAX_SKB_FRAGS 16UL 311#else 312#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 313#endif 314extern int sysctl_max_skb_frags; 315 316/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 317 * segment using its current segmentation instead. 318 */ 319#define GSO_BY_FRAGS 0xFFFF 320 321typedef struct bio_vec skb_frag_t; 322 323/** 324 * skb_frag_size() - Returns the size of a skb fragment 325 * @frag: skb fragment 326 */ 327static inline unsigned int skb_frag_size(const skb_frag_t *frag) 328{ 329 return frag->bv_len; 330} 331 332/** 333 * skb_frag_size_set() - Sets the size of a skb fragment 334 * @frag: skb fragment 335 * @size: size of fragment 336 */ 337static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 338{ 339 frag->bv_len = size; 340} 341 342/** 343 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 344 * @frag: skb fragment 345 * @delta: value to add 346 */ 347static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 348{ 349 frag->bv_len += delta; 350} 351 352/** 353 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 354 * @frag: skb fragment 355 * @delta: value to subtract 356 */ 357static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 358{ 359 frag->bv_len -= delta; 360} 361 362/** 363 * skb_frag_must_loop - Test if %p is a high memory page 364 * @p: fragment's page 365 */ 366static inline bool skb_frag_must_loop(struct page *p) 367{ 368#if defined(CONFIG_HIGHMEM) 369 if (PageHighMem(p)) 370 return true; 371#endif 372 return false; 373} 374 375/** 376 * skb_frag_foreach_page - loop over pages in a fragment 377 * 378 * @f: skb frag to operate on 379 * @f_off: offset from start of f->bv_page 380 * @f_len: length from f_off to loop over 381 * @p: (temp var) current page 382 * @p_off: (temp var) offset from start of current page, 383 * non-zero only on first page. 384 * @p_len: (temp var) length in current page, 385 * < PAGE_SIZE only on first and last page. 386 * @copied: (temp var) length so far, excluding current p_len. 387 * 388 * A fragment can hold a compound page, in which case per-page 389 * operations, notably kmap_atomic, must be called for each 390 * regular page. 391 */ 392#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 393 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 394 p_off = (f_off) & (PAGE_SIZE - 1), \ 395 p_len = skb_frag_must_loop(p) ? \ 396 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 397 copied = 0; \ 398 copied < f_len; \ 399 copied += p_len, p++, p_off = 0, \ 400 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 401 402#define HAVE_HW_TIME_STAMP 403 404/** 405 * struct skb_shared_hwtstamps - hardware time stamps 406 * @hwtstamp: hardware time stamp transformed into duration 407 * since arbitrary point in time 408 * 409 * Software time stamps generated by ktime_get_real() are stored in 410 * skb->tstamp. 411 * 412 * hwtstamps can only be compared against other hwtstamps from 413 * the same device. 414 * 415 * This structure is attached to packets as part of the 416 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 417 */ 418struct skb_shared_hwtstamps { 419 ktime_t hwtstamp; 420}; 421 422/* Definitions for tx_flags in struct skb_shared_info */ 423enum { 424 /* generate hardware time stamp */ 425 SKBTX_HW_TSTAMP = 1 << 0, 426 427 /* generate software time stamp when queueing packet to NIC */ 428 SKBTX_SW_TSTAMP = 1 << 1, 429 430 /* device driver is going to provide hardware time stamp */ 431 SKBTX_IN_PROGRESS = 1 << 2, 432 433 /* device driver supports TX zero-copy buffers */ 434 SKBTX_DEV_ZEROCOPY = 1 << 3, 435 436 /* generate wifi status information (where possible) */ 437 SKBTX_WIFI_STATUS = 1 << 4, 438 439 /* This indicates at least one fragment might be overwritten 440 * (as in vmsplice(), sendfile() ...) 441 * If we need to compute a TX checksum, we'll need to copy 442 * all frags to avoid possible bad checksum 443 */ 444 SKBTX_SHARED_FRAG = 1 << 5, 445 446 /* generate software time stamp when entering packet scheduling */ 447 SKBTX_SCHED_TSTAMP = 1 << 6, 448}; 449 450#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG) 451#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 452 SKBTX_SCHED_TSTAMP) 453#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 454 455/* 456 * The callback notifies userspace to release buffers when skb DMA is done in 457 * lower device, the skb last reference should be 0 when calling this. 458 * The zerocopy_success argument is true if zero copy transmit occurred, 459 * false on data copy or out of memory error caused by data copy attempt. 460 * The ctx field is used to track device context. 461 * The desc field is used to track userspace buffer index. 462 */ 463struct ubuf_info { 464 void (*callback)(struct ubuf_info *, bool zerocopy_success); 465 union { 466 struct { 467 unsigned long desc; 468 void *ctx; 469 }; 470 struct { 471 u32 id; 472 u16 len; 473 u16 zerocopy:1; 474 u32 bytelen; 475 }; 476 }; 477 refcount_t refcnt; 478 479 struct mmpin { 480 struct user_struct *user; 481 unsigned int num_pg; 482 } mmp; 483}; 484 485#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 486 487int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 488void mm_unaccount_pinned_pages(struct mmpin *mmp); 489 490struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size); 491struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size, 492 struct ubuf_info *uarg); 493 494static inline void sock_zerocopy_get(struct ubuf_info *uarg) 495{ 496 refcount_inc(&uarg->refcnt); 497} 498 499void sock_zerocopy_put(struct ubuf_info *uarg); 500void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 501 502void sock_zerocopy_callback(struct ubuf_info *uarg, bool success); 503 504int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len); 505int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 506 struct msghdr *msg, int len, 507 struct ubuf_info *uarg); 508 509/* This data is invariant across clones and lives at 510 * the end of the header data, ie. at skb->end. 511 */ 512struct skb_shared_info { 513 __u8 __unused; 514 __u8 meta_len; 515 __u8 nr_frags; 516 __u8 tx_flags; 517 unsigned short gso_size; 518 /* Warning: this field is not always filled in (UFO)! */ 519 unsigned short gso_segs; 520 struct sk_buff *frag_list; 521 struct skb_shared_hwtstamps hwtstamps; 522 unsigned int gso_type; 523 u32 tskey; 524 525 /* 526 * Warning : all fields before dataref are cleared in __alloc_skb() 527 */ 528 atomic_t dataref; 529 530 /* Intermediate layers must ensure that destructor_arg 531 * remains valid until skb destructor */ 532 void * destructor_arg; 533 534 /* must be last field, see pskb_expand_head() */ 535 skb_frag_t frags[MAX_SKB_FRAGS]; 536}; 537 538/* We divide dataref into two halves. The higher 16 bits hold references 539 * to the payload part of skb->data. The lower 16 bits hold references to 540 * the entire skb->data. A clone of a headerless skb holds the length of 541 * the header in skb->hdr_len. 542 * 543 * All users must obey the rule that the skb->data reference count must be 544 * greater than or equal to the payload reference count. 545 * 546 * Holding a reference to the payload part means that the user does not 547 * care about modifications to the header part of skb->data. 548 */ 549#define SKB_DATAREF_SHIFT 16 550#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 551 552 553enum { 554 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 555 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 556 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 557}; 558 559enum { 560 SKB_GSO_TCPV4 = 1 << 0, 561 562 /* This indicates the skb is from an untrusted source. */ 563 SKB_GSO_DODGY = 1 << 1, 564 565 /* This indicates the tcp segment has CWR set. */ 566 SKB_GSO_TCP_ECN = 1 << 2, 567 568 SKB_GSO_TCP_FIXEDID = 1 << 3, 569 570 SKB_GSO_TCPV6 = 1 << 4, 571 572 SKB_GSO_FCOE = 1 << 5, 573 574 SKB_GSO_GRE = 1 << 6, 575 576 SKB_GSO_GRE_CSUM = 1 << 7, 577 578 SKB_GSO_IPXIP4 = 1 << 8, 579 580 SKB_GSO_IPXIP6 = 1 << 9, 581 582 SKB_GSO_UDP_TUNNEL = 1 << 10, 583 584 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 585 586 SKB_GSO_PARTIAL = 1 << 12, 587 588 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 589 590 SKB_GSO_SCTP = 1 << 14, 591 592 SKB_GSO_ESP = 1 << 15, 593 594 SKB_GSO_UDP = 1 << 16, 595 596 SKB_GSO_UDP_L4 = 1 << 17, 597 598 SKB_GSO_FRAGLIST = 1 << 18, 599}; 600 601#if BITS_PER_LONG > 32 602#define NET_SKBUFF_DATA_USES_OFFSET 1 603#endif 604 605#ifdef NET_SKBUFF_DATA_USES_OFFSET 606typedef unsigned int sk_buff_data_t; 607#else 608typedef unsigned char *sk_buff_data_t; 609#endif 610 611/** 612 * struct sk_buff - socket buffer 613 * @next: Next buffer in list 614 * @prev: Previous buffer in list 615 * @tstamp: Time we arrived/left 616 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 617 * for retransmit timer 618 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 619 * @list: queue head 620 * @sk: Socket we are owned by 621 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 622 * fragmentation management 623 * @dev: Device we arrived on/are leaving by 624 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 625 * @cb: Control buffer. Free for use by every layer. Put private vars here 626 * @_skb_refdst: destination entry (with norefcount bit) 627 * @sp: the security path, used for xfrm 628 * @len: Length of actual data 629 * @data_len: Data length 630 * @mac_len: Length of link layer header 631 * @hdr_len: writable header length of cloned skb 632 * @csum: Checksum (must include start/offset pair) 633 * @csum_start: Offset from skb->head where checksumming should start 634 * @csum_offset: Offset from csum_start where checksum should be stored 635 * @priority: Packet queueing priority 636 * @ignore_df: allow local fragmentation 637 * @cloned: Head may be cloned (check refcnt to be sure) 638 * @ip_summed: Driver fed us an IP checksum 639 * @nohdr: Payload reference only, must not modify header 640 * @pkt_type: Packet class 641 * @fclone: skbuff clone status 642 * @ipvs_property: skbuff is owned by ipvs 643 * @inner_protocol_type: whether the inner protocol is 644 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 645 * @remcsum_offload: remote checksum offload is enabled 646 * @offload_fwd_mark: Packet was L2-forwarded in hardware 647 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 648 * @tc_skip_classify: do not classify packet. set by IFB device 649 * @tc_at_ingress: used within tc_classify to distinguish in/egress 650 * @redirected: packet was redirected by packet classifier 651 * @from_ingress: packet was redirected from the ingress path 652 * @peeked: this packet has been seen already, so stats have been 653 * done for it, don't do them again 654 * @nf_trace: netfilter packet trace flag 655 * @protocol: Packet protocol from driver 656 * @destructor: Destruct function 657 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 658 * @_nfct: Associated connection, if any (with nfctinfo bits) 659 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 660 * @skb_iif: ifindex of device we arrived on 661 * @tc_index: Traffic control index 662 * @hash: the packet hash 663 * @queue_mapping: Queue mapping for multiqueue devices 664 * @head_frag: skb was allocated from page fragments, 665 * not allocated by kmalloc() or vmalloc(). 666 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 667 * @active_extensions: active extensions (skb_ext_id types) 668 * @ndisc_nodetype: router type (from link layer) 669 * @ooo_okay: allow the mapping of a socket to a queue to be changed 670 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 671 * ports. 672 * @sw_hash: indicates hash was computed in software stack 673 * @wifi_acked_valid: wifi_acked was set 674 * @wifi_acked: whether frame was acked on wifi or not 675 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 676 * @encapsulation: indicates the inner headers in the skbuff are valid 677 * @encap_hdr_csum: software checksum is needed 678 * @csum_valid: checksum is already valid 679 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 680 * @csum_complete_sw: checksum was completed by software 681 * @csum_level: indicates the number of consecutive checksums found in 682 * the packet minus one that have been verified as 683 * CHECKSUM_UNNECESSARY (max 3) 684 * @dst_pending_confirm: need to confirm neighbour 685 * @decrypted: Decrypted SKB 686 * @napi_id: id of the NAPI struct this skb came from 687 * @sender_cpu: (aka @napi_id) source CPU in XPS 688 * @secmark: security marking 689 * @mark: Generic packet mark 690 * @reserved_tailroom: (aka @mark) number of bytes of free space available 691 * at the tail of an sk_buff 692 * @vlan_present: VLAN tag is present 693 * @vlan_proto: vlan encapsulation protocol 694 * @vlan_tci: vlan tag control information 695 * @inner_protocol: Protocol (encapsulation) 696 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 697 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 698 * @inner_transport_header: Inner transport layer header (encapsulation) 699 * @inner_network_header: Network layer header (encapsulation) 700 * @inner_mac_header: Link layer header (encapsulation) 701 * @transport_header: Transport layer header 702 * @network_header: Network layer header 703 * @mac_header: Link layer header 704 * @kcov_handle: KCOV remote handle for remote coverage collection 705 * @tail: Tail pointer 706 * @end: End pointer 707 * @head: Head of buffer 708 * @data: Data head pointer 709 * @truesize: Buffer size 710 * @users: User count - see {datagram,tcp}.c 711 * @extensions: allocated extensions, valid if active_extensions is nonzero 712 */ 713 714struct sk_buff { 715 union { 716 struct { 717 /* These two members must be first. */ 718 struct sk_buff *next; 719 struct sk_buff *prev; 720 721 union { 722 struct net_device *dev; 723 /* Some protocols might use this space to store information, 724 * while device pointer would be NULL. 725 * UDP receive path is one user. 726 */ 727 unsigned long dev_scratch; 728 }; 729 }; 730 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 731 struct list_head list; 732 }; 733 734 union { 735 struct sock *sk; 736 int ip_defrag_offset; 737 }; 738 739 union { 740 ktime_t tstamp; 741 u64 skb_mstamp_ns; /* earliest departure time */ 742 }; 743 /* 744 * This is the control buffer. It is free to use for every 745 * layer. Please put your private variables there. If you 746 * want to keep them across layers you have to do a skb_clone() 747 * first. This is owned by whoever has the skb queued ATM. 748 */ 749 char cb[48] __aligned(8); 750 751 union { 752 struct { 753 unsigned long _skb_refdst; 754 void (*destructor)(struct sk_buff *skb); 755 }; 756 struct list_head tcp_tsorted_anchor; 757 }; 758 759#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 760 unsigned long _nfct; 761#endif 762 unsigned int len, 763 data_len; 764 __u16 mac_len, 765 hdr_len; 766 767 /* Following fields are _not_ copied in __copy_skb_header() 768 * Note that queue_mapping is here mostly to fill a hole. 769 */ 770 __u16 queue_mapping; 771 772/* if you move cloned around you also must adapt those constants */ 773#ifdef __BIG_ENDIAN_BITFIELD 774#define CLONED_MASK (1 << 7) 775#else 776#define CLONED_MASK 1 777#endif 778#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset) 779 780 /* private: */ 781 __u8 __cloned_offset[0]; 782 /* public: */ 783 __u8 cloned:1, 784 nohdr:1, 785 fclone:2, 786 peeked:1, 787 head_frag:1, 788 pfmemalloc:1; 789#ifdef CONFIG_SKB_EXTENSIONS 790 __u8 active_extensions; 791#endif 792 /* fields enclosed in headers_start/headers_end are copied 793 * using a single memcpy() in __copy_skb_header() 794 */ 795 /* private: */ 796 __u32 headers_start[0]; 797 /* public: */ 798 799/* if you move pkt_type around you also must adapt those constants */ 800#ifdef __BIG_ENDIAN_BITFIELD 801#define PKT_TYPE_MAX (7 << 5) 802#else 803#define PKT_TYPE_MAX 7 804#endif 805#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 806 807 /* private: */ 808 __u8 __pkt_type_offset[0]; 809 /* public: */ 810 __u8 pkt_type:3; 811 __u8 ignore_df:1; 812 __u8 nf_trace:1; 813 __u8 ip_summed:2; 814 __u8 ooo_okay:1; 815 816 __u8 l4_hash:1; 817 __u8 sw_hash:1; 818 __u8 wifi_acked_valid:1; 819 __u8 wifi_acked:1; 820 __u8 no_fcs:1; 821 /* Indicates the inner headers are valid in the skbuff. */ 822 __u8 encapsulation:1; 823 __u8 encap_hdr_csum:1; 824 __u8 csum_valid:1; 825 826#ifdef __BIG_ENDIAN_BITFIELD 827#define PKT_VLAN_PRESENT_BIT 7 828#else 829#define PKT_VLAN_PRESENT_BIT 0 830#endif 831#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset) 832 /* private: */ 833 __u8 __pkt_vlan_present_offset[0]; 834 /* public: */ 835 __u8 vlan_present:1; 836 __u8 csum_complete_sw:1; 837 __u8 csum_level:2; 838 __u8 csum_not_inet:1; 839 __u8 dst_pending_confirm:1; 840#ifdef CONFIG_IPV6_NDISC_NODETYPE 841 __u8 ndisc_nodetype:2; 842#endif 843 844 __u8 ipvs_property:1; 845 __u8 inner_protocol_type:1; 846 __u8 remcsum_offload:1; 847#ifdef CONFIG_NET_SWITCHDEV 848 __u8 offload_fwd_mark:1; 849 __u8 offload_l3_fwd_mark:1; 850#endif 851#ifdef CONFIG_NET_CLS_ACT 852 __u8 tc_skip_classify:1; 853 __u8 tc_at_ingress:1; 854#endif 855#ifdef CONFIG_NET_REDIRECT 856 __u8 redirected:1; 857 __u8 from_ingress:1; 858#endif 859#ifdef CONFIG_TLS_DEVICE 860 __u8 decrypted:1; 861#endif 862 863#ifdef CONFIG_NET_SCHED 864 __u16 tc_index; /* traffic control index */ 865#endif 866 867 union { 868 __wsum csum; 869 struct { 870 __u16 csum_start; 871 __u16 csum_offset; 872 }; 873 }; 874 __u32 priority; 875 int skb_iif; 876 __u32 hash; 877 __be16 vlan_proto; 878 __u16 vlan_tci; 879#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 880 union { 881 unsigned int napi_id; 882 unsigned int sender_cpu; 883 }; 884#endif 885#ifdef CONFIG_NETWORK_SECMARK 886 __u32 secmark; 887#endif 888 889 union { 890 __u32 mark; 891 __u32 reserved_tailroom; 892 }; 893 894 union { 895 __be16 inner_protocol; 896 __u8 inner_ipproto; 897 }; 898 899 __u16 inner_transport_header; 900 __u16 inner_network_header; 901 __u16 inner_mac_header; 902 903 __be16 protocol; 904 __u16 transport_header; 905 __u16 network_header; 906 __u16 mac_header; 907 908#ifdef CONFIG_KCOV 909 u64 kcov_handle; 910#endif 911 912 /* private: */ 913 __u32 headers_end[0]; 914 /* public: */ 915 916 /* These elements must be at the end, see alloc_skb() for details. */ 917 sk_buff_data_t tail; 918 sk_buff_data_t end; 919 unsigned char *head, 920 *data; 921 unsigned int truesize; 922 refcount_t users; 923 924#ifdef CONFIG_SKB_EXTENSIONS 925 /* only useable after checking ->active_extensions != 0 */ 926 struct skb_ext *extensions; 927#endif 928}; 929 930#ifdef __KERNEL__ 931/* 932 * Handling routines are only of interest to the kernel 933 */ 934 935#define SKB_ALLOC_FCLONE 0x01 936#define SKB_ALLOC_RX 0x02 937#define SKB_ALLOC_NAPI 0x04 938 939/** 940 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 941 * @skb: buffer 942 */ 943static inline bool skb_pfmemalloc(const struct sk_buff *skb) 944{ 945 return unlikely(skb->pfmemalloc); 946} 947 948/* 949 * skb might have a dst pointer attached, refcounted or not. 950 * _skb_refdst low order bit is set if refcount was _not_ taken 951 */ 952#define SKB_DST_NOREF 1UL 953#define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 954 955/** 956 * skb_dst - returns skb dst_entry 957 * @skb: buffer 958 * 959 * Returns skb dst_entry, regardless of reference taken or not. 960 */ 961static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 962{ 963 /* If refdst was not refcounted, check we still are in a 964 * rcu_read_lock section 965 */ 966 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 967 !rcu_read_lock_held() && 968 !rcu_read_lock_bh_held()); 969 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 970} 971 972/** 973 * skb_dst_set - sets skb dst 974 * @skb: buffer 975 * @dst: dst entry 976 * 977 * Sets skb dst, assuming a reference was taken on dst and should 978 * be released by skb_dst_drop() 979 */ 980static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 981{ 982 skb->_skb_refdst = (unsigned long)dst; 983} 984 985/** 986 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 987 * @skb: buffer 988 * @dst: dst entry 989 * 990 * Sets skb dst, assuming a reference was not taken on dst. 991 * If dst entry is cached, we do not take reference and dst_release 992 * will be avoided by refdst_drop. If dst entry is not cached, we take 993 * reference, so that last dst_release can destroy the dst immediately. 994 */ 995static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 996{ 997 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 998 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 999} 1000 1001/** 1002 * skb_dst_is_noref - Test if skb dst isn't refcounted 1003 * @skb: buffer 1004 */ 1005static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1006{ 1007 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1008} 1009 1010/** 1011 * skb_rtable - Returns the skb &rtable 1012 * @skb: buffer 1013 */ 1014static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1015{ 1016 return (struct rtable *)skb_dst(skb); 1017} 1018 1019/* For mangling skb->pkt_type from user space side from applications 1020 * such as nft, tc, etc, we only allow a conservative subset of 1021 * possible pkt_types to be set. 1022*/ 1023static inline bool skb_pkt_type_ok(u32 ptype) 1024{ 1025 return ptype <= PACKET_OTHERHOST; 1026} 1027 1028/** 1029 * skb_napi_id - Returns the skb's NAPI id 1030 * @skb: buffer 1031 */ 1032static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1033{ 1034#ifdef CONFIG_NET_RX_BUSY_POLL 1035 return skb->napi_id; 1036#else 1037 return 0; 1038#endif 1039} 1040 1041/** 1042 * skb_unref - decrement the skb's reference count 1043 * @skb: buffer 1044 * 1045 * Returns true if we can free the skb. 1046 */ 1047static inline bool skb_unref(struct sk_buff *skb) 1048{ 1049 if (unlikely(!skb)) 1050 return false; 1051 if (likely(refcount_read(&skb->users) == 1)) 1052 smp_rmb(); 1053 else if (likely(!refcount_dec_and_test(&skb->users))) 1054 return false; 1055 1056 return true; 1057} 1058 1059void skb_release_head_state(struct sk_buff *skb); 1060void kfree_skb(struct sk_buff *skb); 1061void kfree_skb_list(struct sk_buff *segs); 1062void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1063void skb_tx_error(struct sk_buff *skb); 1064 1065#ifdef CONFIG_TRACEPOINTS 1066void consume_skb(struct sk_buff *skb); 1067#else 1068static inline void consume_skb(struct sk_buff *skb) 1069{ 1070 return kfree_skb(skb); 1071} 1072#endif 1073 1074void __consume_stateless_skb(struct sk_buff *skb); 1075void __kfree_skb(struct sk_buff *skb); 1076extern struct kmem_cache *skbuff_head_cache; 1077 1078void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1079bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1080 bool *fragstolen, int *delta_truesize); 1081 1082struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1083 int node); 1084struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1085struct sk_buff *build_skb(void *data, unsigned int frag_size); 1086struct sk_buff *build_skb_around(struct sk_buff *skb, 1087 void *data, unsigned int frag_size); 1088 1089/** 1090 * alloc_skb - allocate a network buffer 1091 * @size: size to allocate 1092 * @priority: allocation mask 1093 * 1094 * This function is a convenient wrapper around __alloc_skb(). 1095 */ 1096static inline struct sk_buff *alloc_skb(unsigned int size, 1097 gfp_t priority) 1098{ 1099 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1100} 1101 1102struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1103 unsigned long data_len, 1104 int max_page_order, 1105 int *errcode, 1106 gfp_t gfp_mask); 1107struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1108 1109/* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1110struct sk_buff_fclones { 1111 struct sk_buff skb1; 1112 1113 struct sk_buff skb2; 1114 1115 refcount_t fclone_ref; 1116}; 1117 1118/** 1119 * skb_fclone_busy - check if fclone is busy 1120 * @sk: socket 1121 * @skb: buffer 1122 * 1123 * Returns true if skb is a fast clone, and its clone is not freed. 1124 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1125 * so we also check that this didnt happen. 1126 */ 1127static inline bool skb_fclone_busy(const struct sock *sk, 1128 const struct sk_buff *skb) 1129{ 1130 const struct sk_buff_fclones *fclones; 1131 1132 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1133 1134 return skb->fclone == SKB_FCLONE_ORIG && 1135 refcount_read(&fclones->fclone_ref) > 1 && 1136 fclones->skb2.sk == sk; 1137} 1138 1139/** 1140 * alloc_skb_fclone - allocate a network buffer from fclone cache 1141 * @size: size to allocate 1142 * @priority: allocation mask 1143 * 1144 * This function is a convenient wrapper around __alloc_skb(). 1145 */ 1146static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1147 gfp_t priority) 1148{ 1149 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1150} 1151 1152struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1153void skb_headers_offset_update(struct sk_buff *skb, int off); 1154int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1155struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1156void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1157struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1158struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1159 gfp_t gfp_mask, bool fclone); 1160static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1161 gfp_t gfp_mask) 1162{ 1163 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1164} 1165 1166int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1167struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1168 unsigned int headroom); 1169struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1170 int newtailroom, gfp_t priority); 1171int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1172 int offset, int len); 1173int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1174 int offset, int len); 1175int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1176int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1177 1178/** 1179 * skb_pad - zero pad the tail of an skb 1180 * @skb: buffer to pad 1181 * @pad: space to pad 1182 * 1183 * Ensure that a buffer is followed by a padding area that is zero 1184 * filled. Used by network drivers which may DMA or transfer data 1185 * beyond the buffer end onto the wire. 1186 * 1187 * May return error in out of memory cases. The skb is freed on error. 1188 */ 1189static inline int skb_pad(struct sk_buff *skb, int pad) 1190{ 1191 return __skb_pad(skb, pad, true); 1192} 1193#define dev_kfree_skb(a) consume_skb(a) 1194 1195int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1196 int offset, size_t size); 1197 1198struct skb_seq_state { 1199 __u32 lower_offset; 1200 __u32 upper_offset; 1201 __u32 frag_idx; 1202 __u32 stepped_offset; 1203 struct sk_buff *root_skb; 1204 struct sk_buff *cur_skb; 1205 __u8 *frag_data; 1206}; 1207 1208void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1209 unsigned int to, struct skb_seq_state *st); 1210unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1211 struct skb_seq_state *st); 1212void skb_abort_seq_read(struct skb_seq_state *st); 1213 1214unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1215 unsigned int to, struct ts_config *config); 1216 1217/* 1218 * Packet hash types specify the type of hash in skb_set_hash. 1219 * 1220 * Hash types refer to the protocol layer addresses which are used to 1221 * construct a packet's hash. The hashes are used to differentiate or identify 1222 * flows of the protocol layer for the hash type. Hash types are either 1223 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1224 * 1225 * Properties of hashes: 1226 * 1227 * 1) Two packets in different flows have different hash values 1228 * 2) Two packets in the same flow should have the same hash value 1229 * 1230 * A hash at a higher layer is considered to be more specific. A driver should 1231 * set the most specific hash possible. 1232 * 1233 * A driver cannot indicate a more specific hash than the layer at which a hash 1234 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1235 * 1236 * A driver may indicate a hash level which is less specific than the 1237 * actual layer the hash was computed on. For instance, a hash computed 1238 * at L4 may be considered an L3 hash. This should only be done if the 1239 * driver can't unambiguously determine that the HW computed the hash at 1240 * the higher layer. Note that the "should" in the second property above 1241 * permits this. 1242 */ 1243enum pkt_hash_types { 1244 PKT_HASH_TYPE_NONE, /* Undefined type */ 1245 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1246 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1247 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1248}; 1249 1250static inline void skb_clear_hash(struct sk_buff *skb) 1251{ 1252 skb->hash = 0; 1253 skb->sw_hash = 0; 1254 skb->l4_hash = 0; 1255} 1256 1257static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1258{ 1259 if (!skb->l4_hash) 1260 skb_clear_hash(skb); 1261} 1262 1263static inline void 1264__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1265{ 1266 skb->l4_hash = is_l4; 1267 skb->sw_hash = is_sw; 1268 skb->hash = hash; 1269} 1270 1271static inline void 1272skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1273{ 1274 /* Used by drivers to set hash from HW */ 1275 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1276} 1277 1278static inline void 1279__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1280{ 1281 __skb_set_hash(skb, hash, true, is_l4); 1282} 1283 1284void __skb_get_hash(struct sk_buff *skb); 1285u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1286u32 skb_get_poff(const struct sk_buff *skb); 1287u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1288 const struct flow_keys_basic *keys, int hlen); 1289__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1290 void *data, int hlen_proto); 1291 1292static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1293 int thoff, u8 ip_proto) 1294{ 1295 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1296} 1297 1298void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1299 const struct flow_dissector_key *key, 1300 unsigned int key_count); 1301 1302struct bpf_flow_dissector; 1303bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1304 __be16 proto, int nhoff, int hlen, unsigned int flags); 1305 1306bool __skb_flow_dissect(const struct net *net, 1307 const struct sk_buff *skb, 1308 struct flow_dissector *flow_dissector, 1309 void *target_container, 1310 void *data, __be16 proto, int nhoff, int hlen, 1311 unsigned int flags); 1312 1313static inline bool skb_flow_dissect(const struct sk_buff *skb, 1314 struct flow_dissector *flow_dissector, 1315 void *target_container, unsigned int flags) 1316{ 1317 return __skb_flow_dissect(NULL, skb, flow_dissector, 1318 target_container, NULL, 0, 0, 0, flags); 1319} 1320 1321static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1322 struct flow_keys *flow, 1323 unsigned int flags) 1324{ 1325 memset(flow, 0, sizeof(*flow)); 1326 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1327 flow, NULL, 0, 0, 0, flags); 1328} 1329 1330static inline bool 1331skb_flow_dissect_flow_keys_basic(const struct net *net, 1332 const struct sk_buff *skb, 1333 struct flow_keys_basic *flow, void *data, 1334 __be16 proto, int nhoff, int hlen, 1335 unsigned int flags) 1336{ 1337 memset(flow, 0, sizeof(*flow)); 1338 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1339 data, proto, nhoff, hlen, flags); 1340} 1341 1342void skb_flow_dissect_meta(const struct sk_buff *skb, 1343 struct flow_dissector *flow_dissector, 1344 void *target_container); 1345 1346/* Gets a skb connection tracking info, ctinfo map should be a 1347 * map of mapsize to translate enum ip_conntrack_info states 1348 * to user states. 1349 */ 1350void 1351skb_flow_dissect_ct(const struct sk_buff *skb, 1352 struct flow_dissector *flow_dissector, 1353 void *target_container, 1354 u16 *ctinfo_map, 1355 size_t mapsize); 1356void 1357skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1358 struct flow_dissector *flow_dissector, 1359 void *target_container); 1360 1361void skb_flow_dissect_hash(const struct sk_buff *skb, 1362 struct flow_dissector *flow_dissector, 1363 void *target_container); 1364 1365static inline __u32 skb_get_hash(struct sk_buff *skb) 1366{ 1367 if (!skb->l4_hash && !skb->sw_hash) 1368 __skb_get_hash(skb); 1369 1370 return skb->hash; 1371} 1372 1373static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1374{ 1375 if (!skb->l4_hash && !skb->sw_hash) { 1376 struct flow_keys keys; 1377 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1378 1379 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1380 } 1381 1382 return skb->hash; 1383} 1384 1385__u32 skb_get_hash_perturb(const struct sk_buff *skb, 1386 const siphash_key_t *perturb); 1387 1388static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1389{ 1390 return skb->hash; 1391} 1392 1393static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1394{ 1395 to->hash = from->hash; 1396 to->sw_hash = from->sw_hash; 1397 to->l4_hash = from->l4_hash; 1398}; 1399 1400static inline void skb_copy_decrypted(struct sk_buff *to, 1401 const struct sk_buff *from) 1402{ 1403#ifdef CONFIG_TLS_DEVICE 1404 to->decrypted = from->decrypted; 1405#endif 1406} 1407 1408#ifdef NET_SKBUFF_DATA_USES_OFFSET 1409static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1410{ 1411 return skb->head + skb->end; 1412} 1413 1414static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1415{ 1416 return skb->end; 1417} 1418#else 1419static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1420{ 1421 return skb->end; 1422} 1423 1424static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1425{ 1426 return skb->end - skb->head; 1427} 1428#endif 1429 1430/* Internal */ 1431#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1432 1433static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1434{ 1435 return &skb_shinfo(skb)->hwtstamps; 1436} 1437 1438static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1439{ 1440 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY; 1441 1442 return is_zcopy ? skb_uarg(skb) : NULL; 1443} 1444 1445static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1446 bool *have_ref) 1447{ 1448 if (skb && uarg && !skb_zcopy(skb)) { 1449 if (unlikely(have_ref && *have_ref)) 1450 *have_ref = false; 1451 else 1452 sock_zerocopy_get(uarg); 1453 skb_shinfo(skb)->destructor_arg = uarg; 1454 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG; 1455 } 1456} 1457 1458static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1459{ 1460 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1461 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG; 1462} 1463 1464static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1465{ 1466 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1467} 1468 1469static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1470{ 1471 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1472} 1473 1474/* Release a reference on a zerocopy structure */ 1475static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy) 1476{ 1477 struct ubuf_info *uarg = skb_zcopy(skb); 1478 1479 if (uarg) { 1480 if (skb_zcopy_is_nouarg(skb)) { 1481 /* no notification callback */ 1482 } else if (uarg->callback == sock_zerocopy_callback) { 1483 uarg->zerocopy = uarg->zerocopy && zerocopy; 1484 sock_zerocopy_put(uarg); 1485 } else { 1486 uarg->callback(uarg, zerocopy); 1487 } 1488 1489 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG; 1490 } 1491} 1492 1493/* Abort a zerocopy operation and revert zckey on error in send syscall */ 1494static inline void skb_zcopy_abort(struct sk_buff *skb) 1495{ 1496 struct ubuf_info *uarg = skb_zcopy(skb); 1497 1498 if (uarg) { 1499 sock_zerocopy_put_abort(uarg, false); 1500 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG; 1501 } 1502} 1503 1504static inline void skb_mark_not_on_list(struct sk_buff *skb) 1505{ 1506 skb->next = NULL; 1507} 1508 1509/* Iterate through singly-linked GSO fragments of an skb. */ 1510#define skb_list_walk_safe(first, skb, next_skb) \ 1511 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1512 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1513 1514static inline void skb_list_del_init(struct sk_buff *skb) 1515{ 1516 __list_del_entry(&skb->list); 1517 skb_mark_not_on_list(skb); 1518} 1519 1520/** 1521 * skb_queue_empty - check if a queue is empty 1522 * @list: queue head 1523 * 1524 * Returns true if the queue is empty, false otherwise. 1525 */ 1526static inline int skb_queue_empty(const struct sk_buff_head *list) 1527{ 1528 return list->next == (const struct sk_buff *) list; 1529} 1530 1531/** 1532 * skb_queue_empty_lockless - check if a queue is empty 1533 * @list: queue head 1534 * 1535 * Returns true if the queue is empty, false otherwise. 1536 * This variant can be used in lockless contexts. 1537 */ 1538static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1539{ 1540 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1541} 1542 1543 1544/** 1545 * skb_queue_is_last - check if skb is the last entry in the queue 1546 * @list: queue head 1547 * @skb: buffer 1548 * 1549 * Returns true if @skb is the last buffer on the list. 1550 */ 1551static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1552 const struct sk_buff *skb) 1553{ 1554 return skb->next == (const struct sk_buff *) list; 1555} 1556 1557/** 1558 * skb_queue_is_first - check if skb is the first entry in the queue 1559 * @list: queue head 1560 * @skb: buffer 1561 * 1562 * Returns true if @skb is the first buffer on the list. 1563 */ 1564static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1565 const struct sk_buff *skb) 1566{ 1567 return skb->prev == (const struct sk_buff *) list; 1568} 1569 1570/** 1571 * skb_queue_next - return the next packet in the queue 1572 * @list: queue head 1573 * @skb: current buffer 1574 * 1575 * Return the next packet in @list after @skb. It is only valid to 1576 * call this if skb_queue_is_last() evaluates to false. 1577 */ 1578static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1579 const struct sk_buff *skb) 1580{ 1581 /* This BUG_ON may seem severe, but if we just return then we 1582 * are going to dereference garbage. 1583 */ 1584 BUG_ON(skb_queue_is_last(list, skb)); 1585 return skb->next; 1586} 1587 1588/** 1589 * skb_queue_prev - return the prev packet in the queue 1590 * @list: queue head 1591 * @skb: current buffer 1592 * 1593 * Return the prev packet in @list before @skb. It is only valid to 1594 * call this if skb_queue_is_first() evaluates to false. 1595 */ 1596static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1597 const struct sk_buff *skb) 1598{ 1599 /* This BUG_ON may seem severe, but if we just return then we 1600 * are going to dereference garbage. 1601 */ 1602 BUG_ON(skb_queue_is_first(list, skb)); 1603 return skb->prev; 1604} 1605 1606/** 1607 * skb_get - reference buffer 1608 * @skb: buffer to reference 1609 * 1610 * Makes another reference to a socket buffer and returns a pointer 1611 * to the buffer. 1612 */ 1613static inline struct sk_buff *skb_get(struct sk_buff *skb) 1614{ 1615 refcount_inc(&skb->users); 1616 return skb; 1617} 1618 1619/* 1620 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1621 */ 1622 1623/** 1624 * skb_cloned - is the buffer a clone 1625 * @skb: buffer to check 1626 * 1627 * Returns true if the buffer was generated with skb_clone() and is 1628 * one of multiple shared copies of the buffer. Cloned buffers are 1629 * shared data so must not be written to under normal circumstances. 1630 */ 1631static inline int skb_cloned(const struct sk_buff *skb) 1632{ 1633 return skb->cloned && 1634 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1635} 1636 1637static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1638{ 1639 might_sleep_if(gfpflags_allow_blocking(pri)); 1640 1641 if (skb_cloned(skb)) 1642 return pskb_expand_head(skb, 0, 0, pri); 1643 1644 return 0; 1645} 1646 1647/** 1648 * skb_header_cloned - is the header a clone 1649 * @skb: buffer to check 1650 * 1651 * Returns true if modifying the header part of the buffer requires 1652 * the data to be copied. 1653 */ 1654static inline int skb_header_cloned(const struct sk_buff *skb) 1655{ 1656 int dataref; 1657 1658 if (!skb->cloned) 1659 return 0; 1660 1661 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1662 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1663 return dataref != 1; 1664} 1665 1666static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1667{ 1668 might_sleep_if(gfpflags_allow_blocking(pri)); 1669 1670 if (skb_header_cloned(skb)) 1671 return pskb_expand_head(skb, 0, 0, pri); 1672 1673 return 0; 1674} 1675 1676/** 1677 * __skb_header_release - release reference to header 1678 * @skb: buffer to operate on 1679 */ 1680static inline void __skb_header_release(struct sk_buff *skb) 1681{ 1682 skb->nohdr = 1; 1683 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1684} 1685 1686 1687/** 1688 * skb_shared - is the buffer shared 1689 * @skb: buffer to check 1690 * 1691 * Returns true if more than one person has a reference to this 1692 * buffer. 1693 */ 1694static inline int skb_shared(const struct sk_buff *skb) 1695{ 1696 return refcount_read(&skb->users) != 1; 1697} 1698 1699/** 1700 * skb_share_check - check if buffer is shared and if so clone it 1701 * @skb: buffer to check 1702 * @pri: priority for memory allocation 1703 * 1704 * If the buffer is shared the buffer is cloned and the old copy 1705 * drops a reference. A new clone with a single reference is returned. 1706 * If the buffer is not shared the original buffer is returned. When 1707 * being called from interrupt status or with spinlocks held pri must 1708 * be GFP_ATOMIC. 1709 * 1710 * NULL is returned on a memory allocation failure. 1711 */ 1712static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1713{ 1714 might_sleep_if(gfpflags_allow_blocking(pri)); 1715 if (skb_shared(skb)) { 1716 struct sk_buff *nskb = skb_clone(skb, pri); 1717 1718 if (likely(nskb)) 1719 consume_skb(skb); 1720 else 1721 kfree_skb(skb); 1722 skb = nskb; 1723 } 1724 return skb; 1725} 1726 1727/* 1728 * Copy shared buffers into a new sk_buff. We effectively do COW on 1729 * packets to handle cases where we have a local reader and forward 1730 * and a couple of other messy ones. The normal one is tcpdumping 1731 * a packet thats being forwarded. 1732 */ 1733 1734/** 1735 * skb_unshare - make a copy of a shared buffer 1736 * @skb: buffer to check 1737 * @pri: priority for memory allocation 1738 * 1739 * If the socket buffer is a clone then this function creates a new 1740 * copy of the data, drops a reference count on the old copy and returns 1741 * the new copy with the reference count at 1. If the buffer is not a clone 1742 * the original buffer is returned. When called with a spinlock held or 1743 * from interrupt state @pri must be %GFP_ATOMIC 1744 * 1745 * %NULL is returned on a memory allocation failure. 1746 */ 1747static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1748 gfp_t pri) 1749{ 1750 might_sleep_if(gfpflags_allow_blocking(pri)); 1751 if (skb_cloned(skb)) { 1752 struct sk_buff *nskb = skb_copy(skb, pri); 1753 1754 /* Free our shared copy */ 1755 if (likely(nskb)) 1756 consume_skb(skb); 1757 else 1758 kfree_skb(skb); 1759 skb = nskb; 1760 } 1761 return skb; 1762} 1763 1764/** 1765 * skb_peek - peek at the head of an &sk_buff_head 1766 * @list_: list to peek at 1767 * 1768 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1769 * be careful with this one. A peek leaves the buffer on the 1770 * list and someone else may run off with it. You must hold 1771 * the appropriate locks or have a private queue to do this. 1772 * 1773 * Returns %NULL for an empty list or a pointer to the head element. 1774 * The reference count is not incremented and the reference is therefore 1775 * volatile. Use with caution. 1776 */ 1777static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1778{ 1779 struct sk_buff *skb = list_->next; 1780 1781 if (skb == (struct sk_buff *)list_) 1782 skb = NULL; 1783 return skb; 1784} 1785 1786/** 1787 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1788 * @list_: list to peek at 1789 * 1790 * Like skb_peek(), but the caller knows that the list is not empty. 1791 */ 1792static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1793{ 1794 return list_->next; 1795} 1796 1797/** 1798 * skb_peek_next - peek skb following the given one from a queue 1799 * @skb: skb to start from 1800 * @list_: list to peek at 1801 * 1802 * Returns %NULL when the end of the list is met or a pointer to the 1803 * next element. The reference count is not incremented and the 1804 * reference is therefore volatile. Use with caution. 1805 */ 1806static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1807 const struct sk_buff_head *list_) 1808{ 1809 struct sk_buff *next = skb->next; 1810 1811 if (next == (struct sk_buff *)list_) 1812 next = NULL; 1813 return next; 1814} 1815 1816/** 1817 * skb_peek_tail - peek at the tail of an &sk_buff_head 1818 * @list_: list to peek at 1819 * 1820 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1821 * be careful with this one. A peek leaves the buffer on the 1822 * list and someone else may run off with it. You must hold 1823 * the appropriate locks or have a private queue to do this. 1824 * 1825 * Returns %NULL for an empty list or a pointer to the tail element. 1826 * The reference count is not incremented and the reference is therefore 1827 * volatile. Use with caution. 1828 */ 1829static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1830{ 1831 struct sk_buff *skb = READ_ONCE(list_->prev); 1832 1833 if (skb == (struct sk_buff *)list_) 1834 skb = NULL; 1835 return skb; 1836 1837} 1838 1839/** 1840 * skb_queue_len - get queue length 1841 * @list_: list to measure 1842 * 1843 * Return the length of an &sk_buff queue. 1844 */ 1845static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1846{ 1847 return list_->qlen; 1848} 1849 1850/** 1851 * skb_queue_len_lockless - get queue length 1852 * @list_: list to measure 1853 * 1854 * Return the length of an &sk_buff queue. 1855 * This variant can be used in lockless contexts. 1856 */ 1857static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1858{ 1859 return READ_ONCE(list_->qlen); 1860} 1861 1862/** 1863 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1864 * @list: queue to initialize 1865 * 1866 * This initializes only the list and queue length aspects of 1867 * an sk_buff_head object. This allows to initialize the list 1868 * aspects of an sk_buff_head without reinitializing things like 1869 * the spinlock. It can also be used for on-stack sk_buff_head 1870 * objects where the spinlock is known to not be used. 1871 */ 1872static inline void __skb_queue_head_init(struct sk_buff_head *list) 1873{ 1874 list->prev = list->next = (struct sk_buff *)list; 1875 list->qlen = 0; 1876} 1877 1878/* 1879 * This function creates a split out lock class for each invocation; 1880 * this is needed for now since a whole lot of users of the skb-queue 1881 * infrastructure in drivers have different locking usage (in hardirq) 1882 * than the networking core (in softirq only). In the long run either the 1883 * network layer or drivers should need annotation to consolidate the 1884 * main types of usage into 3 classes. 1885 */ 1886static inline void skb_queue_head_init(struct sk_buff_head *list) 1887{ 1888 spin_lock_init(&list->lock); 1889 __skb_queue_head_init(list); 1890} 1891 1892static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1893 struct lock_class_key *class) 1894{ 1895 skb_queue_head_init(list); 1896 lockdep_set_class(&list->lock, class); 1897} 1898 1899/* 1900 * Insert an sk_buff on a list. 1901 * 1902 * The "__skb_xxxx()" functions are the non-atomic ones that 1903 * can only be called with interrupts disabled. 1904 */ 1905static inline void __skb_insert(struct sk_buff *newsk, 1906 struct sk_buff *prev, struct sk_buff *next, 1907 struct sk_buff_head *list) 1908{ 1909 /* See skb_queue_empty_lockless() and skb_peek_tail() 1910 * for the opposite READ_ONCE() 1911 */ 1912 WRITE_ONCE(newsk->next, next); 1913 WRITE_ONCE(newsk->prev, prev); 1914 WRITE_ONCE(next->prev, newsk); 1915 WRITE_ONCE(prev->next, newsk); 1916 list->qlen++; 1917} 1918 1919static inline void __skb_queue_splice(const struct sk_buff_head *list, 1920 struct sk_buff *prev, 1921 struct sk_buff *next) 1922{ 1923 struct sk_buff *first = list->next; 1924 struct sk_buff *last = list->prev; 1925 1926 WRITE_ONCE(first->prev, prev); 1927 WRITE_ONCE(prev->next, first); 1928 1929 WRITE_ONCE(last->next, next); 1930 WRITE_ONCE(next->prev, last); 1931} 1932 1933/** 1934 * skb_queue_splice - join two skb lists, this is designed for stacks 1935 * @list: the new list to add 1936 * @head: the place to add it in the first list 1937 */ 1938static inline void skb_queue_splice(const struct sk_buff_head *list, 1939 struct sk_buff_head *head) 1940{ 1941 if (!skb_queue_empty(list)) { 1942 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1943 head->qlen += list->qlen; 1944 } 1945} 1946 1947/** 1948 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1949 * @list: the new list to add 1950 * @head: the place to add it in the first list 1951 * 1952 * The list at @list is reinitialised 1953 */ 1954static inline void skb_queue_splice_init(struct sk_buff_head *list, 1955 struct sk_buff_head *head) 1956{ 1957 if (!skb_queue_empty(list)) { 1958 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1959 head->qlen += list->qlen; 1960 __skb_queue_head_init(list); 1961 } 1962} 1963 1964/** 1965 * skb_queue_splice_tail - join two skb lists, each list being a queue 1966 * @list: the new list to add 1967 * @head: the place to add it in the first list 1968 */ 1969static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1970 struct sk_buff_head *head) 1971{ 1972 if (!skb_queue_empty(list)) { 1973 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1974 head->qlen += list->qlen; 1975 } 1976} 1977 1978/** 1979 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1980 * @list: the new list to add 1981 * @head: the place to add it in the first list 1982 * 1983 * Each of the lists is a queue. 1984 * The list at @list is reinitialised 1985 */ 1986static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 1987 struct sk_buff_head *head) 1988{ 1989 if (!skb_queue_empty(list)) { 1990 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1991 head->qlen += list->qlen; 1992 __skb_queue_head_init(list); 1993 } 1994} 1995 1996/** 1997 * __skb_queue_after - queue a buffer at the list head 1998 * @list: list to use 1999 * @prev: place after this buffer 2000 * @newsk: buffer to queue 2001 * 2002 * Queue a buffer int the middle of a list. This function takes no locks 2003 * and you must therefore hold required locks before calling it. 2004 * 2005 * A buffer cannot be placed on two lists at the same time. 2006 */ 2007static inline void __skb_queue_after(struct sk_buff_head *list, 2008 struct sk_buff *prev, 2009 struct sk_buff *newsk) 2010{ 2011 __skb_insert(newsk, prev, prev->next, list); 2012} 2013 2014void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2015 struct sk_buff_head *list); 2016 2017static inline void __skb_queue_before(struct sk_buff_head *list, 2018 struct sk_buff *next, 2019 struct sk_buff *newsk) 2020{ 2021 __skb_insert(newsk, next->prev, next, list); 2022} 2023 2024/** 2025 * __skb_queue_head - queue a buffer at the list head 2026 * @list: list to use 2027 * @newsk: buffer to queue 2028 * 2029 * Queue a buffer at the start of a list. This function takes no locks 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 */ 2034static inline void __skb_queue_head(struct sk_buff_head *list, 2035 struct sk_buff *newsk) 2036{ 2037 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2038} 2039void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2040 2041/** 2042 * __skb_queue_tail - queue a buffer at the list tail 2043 * @list: list to use 2044 * @newsk: buffer to queue 2045 * 2046 * Queue a buffer at the end of a list. This function takes no locks 2047 * and you must therefore hold required locks before calling it. 2048 * 2049 * A buffer cannot be placed on two lists at the same time. 2050 */ 2051static inline void __skb_queue_tail(struct sk_buff_head *list, 2052 struct sk_buff *newsk) 2053{ 2054 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2055} 2056void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2057 2058/* 2059 * remove sk_buff from list. _Must_ be called atomically, and with 2060 * the list known.. 2061 */ 2062void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2063static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2064{ 2065 struct sk_buff *next, *prev; 2066 2067 WRITE_ONCE(list->qlen, list->qlen - 1); 2068 next = skb->next; 2069 prev = skb->prev; 2070 skb->next = skb->prev = NULL; 2071 WRITE_ONCE(next->prev, prev); 2072 WRITE_ONCE(prev->next, next); 2073} 2074 2075/** 2076 * __skb_dequeue - remove from the head of the queue 2077 * @list: list to dequeue from 2078 * 2079 * Remove the head of the list. This function does not take any locks 2080 * so must be used with appropriate locks held only. The head item is 2081 * returned or %NULL if the list is empty. 2082 */ 2083static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2084{ 2085 struct sk_buff *skb = skb_peek(list); 2086 if (skb) 2087 __skb_unlink(skb, list); 2088 return skb; 2089} 2090struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2091 2092/** 2093 * __skb_dequeue_tail - remove from the tail of the queue 2094 * @list: list to dequeue from 2095 * 2096 * Remove the tail of the list. This function does not take any locks 2097 * so must be used with appropriate locks held only. The tail item is 2098 * returned or %NULL if the list is empty. 2099 */ 2100static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2101{ 2102 struct sk_buff *skb = skb_peek_tail(list); 2103 if (skb) 2104 __skb_unlink(skb, list); 2105 return skb; 2106} 2107struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2108 2109 2110static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2111{ 2112 return skb->data_len; 2113} 2114 2115static inline unsigned int skb_headlen(const struct sk_buff *skb) 2116{ 2117 return skb->len - skb->data_len; 2118} 2119 2120static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2121{ 2122 unsigned int i, len = 0; 2123 2124 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2125 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2126 return len; 2127} 2128 2129static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2130{ 2131 return skb_headlen(skb) + __skb_pagelen(skb); 2132} 2133 2134/** 2135 * __skb_fill_page_desc - initialise a paged fragment in an skb 2136 * @skb: buffer containing fragment to be initialised 2137 * @i: paged fragment index to initialise 2138 * @page: the page to use for this fragment 2139 * @off: the offset to the data with @page 2140 * @size: the length of the data 2141 * 2142 * Initialises the @i'th fragment of @skb to point to &size bytes at 2143 * offset @off within @page. 2144 * 2145 * Does not take any additional reference on the fragment. 2146 */ 2147static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2148 struct page *page, int off, int size) 2149{ 2150 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2151 2152 /* 2153 * Propagate page pfmemalloc to the skb if we can. The problem is 2154 * that not all callers have unique ownership of the page but rely 2155 * on page_is_pfmemalloc doing the right thing(tm). 2156 */ 2157 frag->bv_page = page; 2158 frag->bv_offset = off; 2159 skb_frag_size_set(frag, size); 2160 2161 page = compound_head(page); 2162 if (page_is_pfmemalloc(page)) 2163 skb->pfmemalloc = true; 2164} 2165 2166/** 2167 * skb_fill_page_desc - initialise a paged fragment in an skb 2168 * @skb: buffer containing fragment to be initialised 2169 * @i: paged fragment index to initialise 2170 * @page: the page to use for this fragment 2171 * @off: the offset to the data with @page 2172 * @size: the length of the data 2173 * 2174 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2175 * @skb to point to @size bytes at offset @off within @page. In 2176 * addition updates @skb such that @i is the last fragment. 2177 * 2178 * Does not take any additional reference on the fragment. 2179 */ 2180static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2181 struct page *page, int off, int size) 2182{ 2183 __skb_fill_page_desc(skb, i, page, off, size); 2184 skb_shinfo(skb)->nr_frags = i + 1; 2185} 2186 2187void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2188 int size, unsigned int truesize); 2189 2190void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2191 unsigned int truesize); 2192 2193#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2194 2195#ifdef NET_SKBUFF_DATA_USES_OFFSET 2196static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2197{ 2198 return skb->head + skb->tail; 2199} 2200 2201static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2202{ 2203 skb->tail = skb->data - skb->head; 2204} 2205 2206static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2207{ 2208 skb_reset_tail_pointer(skb); 2209 skb->tail += offset; 2210} 2211 2212#else /* NET_SKBUFF_DATA_USES_OFFSET */ 2213static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2214{ 2215 return skb->tail; 2216} 2217 2218static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2219{ 2220 skb->tail = skb->data; 2221} 2222 2223static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2224{ 2225 skb->tail = skb->data + offset; 2226} 2227 2228#endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2229 2230/* 2231 * Add data to an sk_buff 2232 */ 2233void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2234void *skb_put(struct sk_buff *skb, unsigned int len); 2235static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2236{ 2237 void *tmp = skb_tail_pointer(skb); 2238 SKB_LINEAR_ASSERT(skb); 2239 skb->tail += len; 2240 skb->len += len; 2241 return tmp; 2242} 2243 2244static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2245{ 2246 void *tmp = __skb_put(skb, len); 2247 2248 memset(tmp, 0, len); 2249 return tmp; 2250} 2251 2252static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2253 unsigned int len) 2254{ 2255 void *tmp = __skb_put(skb, len); 2256 2257 memcpy(tmp, data, len); 2258 return tmp; 2259} 2260 2261static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2262{ 2263 *(u8 *)__skb_put(skb, 1) = val; 2264} 2265 2266static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2267{ 2268 void *tmp = skb_put(skb, len); 2269 2270 memset(tmp, 0, len); 2271 2272 return tmp; 2273} 2274 2275static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2276 unsigned int len) 2277{ 2278 void *tmp = skb_put(skb, len); 2279 2280 memcpy(tmp, data, len); 2281 2282 return tmp; 2283} 2284 2285static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2286{ 2287 *(u8 *)skb_put(skb, 1) = val; 2288} 2289 2290void *skb_push(struct sk_buff *skb, unsigned int len); 2291static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2292{ 2293 skb->data -= len; 2294 skb->len += len; 2295 return skb->data; 2296} 2297 2298void *skb_pull(struct sk_buff *skb, unsigned int len); 2299static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2300{ 2301 skb->len -= len; 2302 BUG_ON(skb->len < skb->data_len); 2303 return skb->data += len; 2304} 2305 2306static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2307{ 2308 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2309} 2310 2311void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2312 2313static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2314{ 2315 if (len > skb_headlen(skb) && 2316 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2317 return NULL; 2318 skb->len -= len; 2319 return skb->data += len; 2320} 2321 2322static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2323{ 2324 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2325} 2326 2327static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2328{ 2329 if (likely(len <= skb_headlen(skb))) 2330 return true; 2331 if (unlikely(len > skb->len)) 2332 return false; 2333 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2334} 2335 2336void skb_condense(struct sk_buff *skb); 2337 2338/** 2339 * skb_headroom - bytes at buffer head 2340 * @skb: buffer to check 2341 * 2342 * Return the number of bytes of free space at the head of an &sk_buff. 2343 */ 2344static inline unsigned int skb_headroom(const struct sk_buff *skb) 2345{ 2346 return skb->data - skb->head; 2347} 2348 2349/** 2350 * skb_tailroom - bytes at buffer end 2351 * @skb: buffer to check 2352 * 2353 * Return the number of bytes of free space at the tail of an sk_buff 2354 */ 2355static inline int skb_tailroom(const struct sk_buff *skb) 2356{ 2357 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2358} 2359 2360/** 2361 * skb_availroom - bytes at buffer end 2362 * @skb: buffer to check 2363 * 2364 * Return the number of bytes of free space at the tail of an sk_buff 2365 * allocated by sk_stream_alloc() 2366 */ 2367static inline int skb_availroom(const struct sk_buff *skb) 2368{ 2369 if (skb_is_nonlinear(skb)) 2370 return 0; 2371 2372 return skb->end - skb->tail - skb->reserved_tailroom; 2373} 2374 2375/** 2376 * skb_reserve - adjust headroom 2377 * @skb: buffer to alter 2378 * @len: bytes to move 2379 * 2380 * Increase the headroom of an empty &sk_buff by reducing the tail 2381 * room. This is only allowed for an empty buffer. 2382 */ 2383static inline void skb_reserve(struct sk_buff *skb, int len) 2384{ 2385 skb->data += len; 2386 skb->tail += len; 2387} 2388 2389/** 2390 * skb_tailroom_reserve - adjust reserved_tailroom 2391 * @skb: buffer to alter 2392 * @mtu: maximum amount of headlen permitted 2393 * @needed_tailroom: minimum amount of reserved_tailroom 2394 * 2395 * Set reserved_tailroom so that headlen can be as large as possible but 2396 * not larger than mtu and tailroom cannot be smaller than 2397 * needed_tailroom. 2398 * The required headroom should already have been reserved before using 2399 * this function. 2400 */ 2401static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2402 unsigned int needed_tailroom) 2403{ 2404 SKB_LINEAR_ASSERT(skb); 2405 if (mtu < skb_tailroom(skb) - needed_tailroom) 2406 /* use at most mtu */ 2407 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2408 else 2409 /* use up to all available space */ 2410 skb->reserved_tailroom = needed_tailroom; 2411} 2412 2413#define ENCAP_TYPE_ETHER 0 2414#define ENCAP_TYPE_IPPROTO 1 2415 2416static inline void skb_set_inner_protocol(struct sk_buff *skb, 2417 __be16 protocol) 2418{ 2419 skb->inner_protocol = protocol; 2420 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2421} 2422 2423static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2424 __u8 ipproto) 2425{ 2426 skb->inner_ipproto = ipproto; 2427 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2428} 2429 2430static inline void skb_reset_inner_headers(struct sk_buff *skb) 2431{ 2432 skb->inner_mac_header = skb->mac_header; 2433 skb->inner_network_header = skb->network_header; 2434 skb->inner_transport_header = skb->transport_header; 2435} 2436 2437static inline void skb_reset_mac_len(struct sk_buff *skb) 2438{ 2439 skb->mac_len = skb->network_header - skb->mac_header; 2440} 2441 2442static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2443 *skb) 2444{ 2445 return skb->head + skb->inner_transport_header; 2446} 2447 2448static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2449{ 2450 return skb_inner_transport_header(skb) - skb->data; 2451} 2452 2453static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2454{ 2455 skb->inner_transport_header = skb->data - skb->head; 2456} 2457 2458static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2459 const int offset) 2460{ 2461 skb_reset_inner_transport_header(skb); 2462 skb->inner_transport_header += offset; 2463} 2464 2465static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2466{ 2467 return skb->head + skb->inner_network_header; 2468} 2469 2470static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2471{ 2472 skb->inner_network_header = skb->data - skb->head; 2473} 2474 2475static inline void skb_set_inner_network_header(struct sk_buff *skb, 2476 const int offset) 2477{ 2478 skb_reset_inner_network_header(skb); 2479 skb->inner_network_header += offset; 2480} 2481 2482static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2483{ 2484 return skb->head + skb->inner_mac_header; 2485} 2486 2487static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2488{ 2489 skb->inner_mac_header = skb->data - skb->head; 2490} 2491 2492static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2493 const int offset) 2494{ 2495 skb_reset_inner_mac_header(skb); 2496 skb->inner_mac_header += offset; 2497} 2498static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2499{ 2500 return skb->transport_header != (typeof(skb->transport_header))~0U; 2501} 2502 2503static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2504{ 2505 return skb->head + skb->transport_header; 2506} 2507 2508static inline void skb_reset_transport_header(struct sk_buff *skb) 2509{ 2510 skb->transport_header = skb->data - skb->head; 2511} 2512 2513static inline void skb_set_transport_header(struct sk_buff *skb, 2514 const int offset) 2515{ 2516 skb_reset_transport_header(skb); 2517 skb->transport_header += offset; 2518} 2519 2520static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2521{ 2522 return skb->head + skb->network_header; 2523} 2524 2525static inline void skb_reset_network_header(struct sk_buff *skb) 2526{ 2527 skb->network_header = skb->data - skb->head; 2528} 2529 2530static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2531{ 2532 skb_reset_network_header(skb); 2533 skb->network_header += offset; 2534} 2535 2536static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2537{ 2538 return skb->head + skb->mac_header; 2539} 2540 2541static inline int skb_mac_offset(const struct sk_buff *skb) 2542{ 2543 return skb_mac_header(skb) - skb->data; 2544} 2545 2546static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2547{ 2548 return skb->network_header - skb->mac_header; 2549} 2550 2551static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2552{ 2553 return skb->mac_header != (typeof(skb->mac_header))~0U; 2554} 2555 2556static inline void skb_unset_mac_header(struct sk_buff *skb) 2557{ 2558 skb->mac_header = (typeof(skb->mac_header))~0U; 2559} 2560 2561static inline void skb_reset_mac_header(struct sk_buff *skb) 2562{ 2563 skb->mac_header = skb->data - skb->head; 2564} 2565 2566static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2567{ 2568 skb_reset_mac_header(skb); 2569 skb->mac_header += offset; 2570} 2571 2572static inline void skb_pop_mac_header(struct sk_buff *skb) 2573{ 2574 skb->mac_header = skb->network_header; 2575} 2576 2577static inline void skb_probe_transport_header(struct sk_buff *skb) 2578{ 2579 struct flow_keys_basic keys; 2580 2581 if (skb_transport_header_was_set(skb)) 2582 return; 2583 2584 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2585 NULL, 0, 0, 0, 0)) 2586 skb_set_transport_header(skb, keys.control.thoff); 2587} 2588 2589static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2590{ 2591 if (skb_mac_header_was_set(skb)) { 2592 const unsigned char *old_mac = skb_mac_header(skb); 2593 2594 skb_set_mac_header(skb, -skb->mac_len); 2595 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2596 } 2597} 2598 2599static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2600{ 2601 return skb->csum_start - skb_headroom(skb); 2602} 2603 2604static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2605{ 2606 return skb->head + skb->csum_start; 2607} 2608 2609static inline int skb_transport_offset(const struct sk_buff *skb) 2610{ 2611 return skb_transport_header(skb) - skb->data; 2612} 2613 2614static inline u32 skb_network_header_len(const struct sk_buff *skb) 2615{ 2616 return skb->transport_header - skb->network_header; 2617} 2618 2619static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2620{ 2621 return skb->inner_transport_header - skb->inner_network_header; 2622} 2623 2624static inline int skb_network_offset(const struct sk_buff *skb) 2625{ 2626 return skb_network_header(skb) - skb->data; 2627} 2628 2629static inline int skb_inner_network_offset(const struct sk_buff *skb) 2630{ 2631 return skb_inner_network_header(skb) - skb->data; 2632} 2633 2634static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2635{ 2636 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2637} 2638 2639/* 2640 * CPUs often take a performance hit when accessing unaligned memory 2641 * locations. The actual performance hit varies, it can be small if the 2642 * hardware handles it or large if we have to take an exception and fix it 2643 * in software. 2644 * 2645 * Since an ethernet header is 14 bytes network drivers often end up with 2646 * the IP header at an unaligned offset. The IP header can be aligned by 2647 * shifting the start of the packet by 2 bytes. Drivers should do this 2648 * with: 2649 * 2650 * skb_reserve(skb, NET_IP_ALIGN); 2651 * 2652 * The downside to this alignment of the IP header is that the DMA is now 2653 * unaligned. On some architectures the cost of an unaligned DMA is high 2654 * and this cost outweighs the gains made by aligning the IP header. 2655 * 2656 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2657 * to be overridden. 2658 */ 2659#ifndef NET_IP_ALIGN 2660#define NET_IP_ALIGN 2 2661#endif 2662 2663/* 2664 * The networking layer reserves some headroom in skb data (via 2665 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2666 * the header has to grow. In the default case, if the header has to grow 2667 * 32 bytes or less we avoid the reallocation. 2668 * 2669 * Unfortunately this headroom changes the DMA alignment of the resulting 2670 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2671 * on some architectures. An architecture can override this value, 2672 * perhaps setting it to a cacheline in size (since that will maintain 2673 * cacheline alignment of the DMA). It must be a power of 2. 2674 * 2675 * Various parts of the networking layer expect at least 32 bytes of 2676 * headroom, you should not reduce this. 2677 * 2678 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2679 * to reduce average number of cache lines per packet. 2680 * get_rps_cpu() for example only access one 64 bytes aligned block : 2681 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2682 */ 2683#ifndef NET_SKB_PAD 2684#define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2685#endif 2686 2687int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2688 2689static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2690{ 2691 if (WARN_ON(skb_is_nonlinear(skb))) 2692 return; 2693 skb->len = len; 2694 skb_set_tail_pointer(skb, len); 2695} 2696 2697static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2698{ 2699 __skb_set_length(skb, len); 2700} 2701 2702void skb_trim(struct sk_buff *skb, unsigned int len); 2703 2704static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2705{ 2706 if (skb->data_len) 2707 return ___pskb_trim(skb, len); 2708 __skb_trim(skb, len); 2709 return 0; 2710} 2711 2712static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2713{ 2714 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2715} 2716 2717/** 2718 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2719 * @skb: buffer to alter 2720 * @len: new length 2721 * 2722 * This is identical to pskb_trim except that the caller knows that 2723 * the skb is not cloned so we should never get an error due to out- 2724 * of-memory. 2725 */ 2726static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2727{ 2728 int err = pskb_trim(skb, len); 2729 BUG_ON(err); 2730} 2731 2732static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2733{ 2734 unsigned int diff = len - skb->len; 2735 2736 if (skb_tailroom(skb) < diff) { 2737 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2738 GFP_ATOMIC); 2739 if (ret) 2740 return ret; 2741 } 2742 __skb_set_length(skb, len); 2743 return 0; 2744} 2745 2746/** 2747 * skb_orphan - orphan a buffer 2748 * @skb: buffer to orphan 2749 * 2750 * If a buffer currently has an owner then we call the owner's 2751 * destructor function and make the @skb unowned. The buffer continues 2752 * to exist but is no longer charged to its former owner. 2753 */ 2754static inline void skb_orphan(struct sk_buff *skb) 2755{ 2756 if (skb->destructor) { 2757 skb->destructor(skb); 2758 skb->destructor = NULL; 2759 skb->sk = NULL; 2760 } else { 2761 BUG_ON(skb->sk); 2762 } 2763} 2764 2765/** 2766 * skb_orphan_frags - orphan the frags contained in a buffer 2767 * @skb: buffer to orphan frags from 2768 * @gfp_mask: allocation mask for replacement pages 2769 * 2770 * For each frag in the SKB which needs a destructor (i.e. has an 2771 * owner) create a copy of that frag and release the original 2772 * page by calling the destructor. 2773 */ 2774static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2775{ 2776 if (likely(!skb_zcopy(skb))) 2777 return 0; 2778 if (!skb_zcopy_is_nouarg(skb) && 2779 skb_uarg(skb)->callback == sock_zerocopy_callback) 2780 return 0; 2781 return skb_copy_ubufs(skb, gfp_mask); 2782} 2783 2784/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2785static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2786{ 2787 if (likely(!skb_zcopy(skb))) 2788 return 0; 2789 return skb_copy_ubufs(skb, gfp_mask); 2790} 2791 2792/** 2793 * __skb_queue_purge - empty a list 2794 * @list: list to empty 2795 * 2796 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2797 * the list and one reference dropped. This function does not take the 2798 * list lock and the caller must hold the relevant locks to use it. 2799 */ 2800static inline void __skb_queue_purge(struct sk_buff_head *list) 2801{ 2802 struct sk_buff *skb; 2803 while ((skb = __skb_dequeue(list)) != NULL) 2804 kfree_skb(skb); 2805} 2806void skb_queue_purge(struct sk_buff_head *list); 2807 2808unsigned int skb_rbtree_purge(struct rb_root *root); 2809 2810void *netdev_alloc_frag(unsigned int fragsz); 2811 2812struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2813 gfp_t gfp_mask); 2814 2815/** 2816 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2817 * @dev: network device to receive on 2818 * @length: length to allocate 2819 * 2820 * Allocate a new &sk_buff and assign it a usage count of one. The 2821 * buffer has unspecified headroom built in. Users should allocate 2822 * the headroom they think they need without accounting for the 2823 * built in space. The built in space is used for optimisations. 2824 * 2825 * %NULL is returned if there is no free memory. Although this function 2826 * allocates memory it can be called from an interrupt. 2827 */ 2828static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2829 unsigned int length) 2830{ 2831 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2832} 2833 2834/* legacy helper around __netdev_alloc_skb() */ 2835static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2836 gfp_t gfp_mask) 2837{ 2838 return __netdev_alloc_skb(NULL, length, gfp_mask); 2839} 2840 2841/* legacy helper around netdev_alloc_skb() */ 2842static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2843{ 2844 return netdev_alloc_skb(NULL, length); 2845} 2846 2847 2848static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2849 unsigned int length, gfp_t gfp) 2850{ 2851 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2852 2853 if (NET_IP_ALIGN && skb) 2854 skb_reserve(skb, NET_IP_ALIGN); 2855 return skb; 2856} 2857 2858static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2859 unsigned int length) 2860{ 2861 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2862} 2863 2864static inline void skb_free_frag(void *addr) 2865{ 2866 page_frag_free(addr); 2867} 2868 2869void *napi_alloc_frag(unsigned int fragsz); 2870struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2871 unsigned int length, gfp_t gfp_mask); 2872static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2873 unsigned int length) 2874{ 2875 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2876} 2877void napi_consume_skb(struct sk_buff *skb, int budget); 2878 2879void __kfree_skb_flush(void); 2880void __kfree_skb_defer(struct sk_buff *skb); 2881 2882/** 2883 * __dev_alloc_pages - allocate page for network Rx 2884 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2885 * @order: size of the allocation 2886 * 2887 * Allocate a new page. 2888 * 2889 * %NULL is returned if there is no free memory. 2890*/ 2891static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2892 unsigned int order) 2893{ 2894 /* This piece of code contains several assumptions. 2895 * 1. This is for device Rx, therefor a cold page is preferred. 2896 * 2. The expectation is the user wants a compound page. 2897 * 3. If requesting a order 0 page it will not be compound 2898 * due to the check to see if order has a value in prep_new_page 2899 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2900 * code in gfp_to_alloc_flags that should be enforcing this. 2901 */ 2902 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2903 2904 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2905} 2906 2907static inline struct page *dev_alloc_pages(unsigned int order) 2908{ 2909 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2910} 2911 2912/** 2913 * __dev_alloc_page - allocate a page for network Rx 2914 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2915 * 2916 * Allocate a new page. 2917 * 2918 * %NULL is returned if there is no free memory. 2919 */ 2920static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2921{ 2922 return __dev_alloc_pages(gfp_mask, 0); 2923} 2924 2925static inline struct page *dev_alloc_page(void) 2926{ 2927 return dev_alloc_pages(0); 2928} 2929 2930/** 2931 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2932 * @page: The page that was allocated from skb_alloc_page 2933 * @skb: The skb that may need pfmemalloc set 2934 */ 2935static inline void skb_propagate_pfmemalloc(struct page *page, 2936 struct sk_buff *skb) 2937{ 2938 if (page_is_pfmemalloc(page)) 2939 skb->pfmemalloc = true; 2940} 2941 2942/** 2943 * skb_frag_off() - Returns the offset of a skb fragment 2944 * @frag: the paged fragment 2945 */ 2946static inline unsigned int skb_frag_off(const skb_frag_t *frag) 2947{ 2948 return frag->bv_offset; 2949} 2950 2951/** 2952 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 2953 * @frag: skb fragment 2954 * @delta: value to add 2955 */ 2956static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 2957{ 2958 frag->bv_offset += delta; 2959} 2960 2961/** 2962 * skb_frag_off_set() - Sets the offset of a skb fragment 2963 * @frag: skb fragment 2964 * @offset: offset of fragment 2965 */ 2966static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 2967{ 2968 frag->bv_offset = offset; 2969} 2970 2971/** 2972 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 2973 * @fragto: skb fragment where offset is set 2974 * @fragfrom: skb fragment offset is copied from 2975 */ 2976static inline void skb_frag_off_copy(skb_frag_t *fragto, 2977 const skb_frag_t *fragfrom) 2978{ 2979 fragto->bv_offset = fragfrom->bv_offset; 2980} 2981 2982/** 2983 * skb_frag_page - retrieve the page referred to by a paged fragment 2984 * @frag: the paged fragment 2985 * 2986 * Returns the &struct page associated with @frag. 2987 */ 2988static inline struct page *skb_frag_page(const skb_frag_t *frag) 2989{ 2990 return frag->bv_page; 2991} 2992 2993/** 2994 * __skb_frag_ref - take an addition reference on a paged fragment. 2995 * @frag: the paged fragment 2996 * 2997 * Takes an additional reference on the paged fragment @frag. 2998 */ 2999static inline void __skb_frag_ref(skb_frag_t *frag) 3000{ 3001 get_page(skb_frag_page(frag)); 3002} 3003 3004/** 3005 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3006 * @skb: the buffer 3007 * @f: the fragment offset. 3008 * 3009 * Takes an additional reference on the @f'th paged fragment of @skb. 3010 */ 3011static inline void skb_frag_ref(struct sk_buff *skb, int f) 3012{ 3013 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3014} 3015 3016/** 3017 * __skb_frag_unref - release a reference on a paged fragment. 3018 * @frag: the paged fragment 3019 * 3020 * Releases a reference on the paged fragment @frag. 3021 */ 3022static inline void __skb_frag_unref(skb_frag_t *frag) 3023{ 3024 put_page(skb_frag_page(frag)); 3025} 3026 3027/** 3028 * skb_frag_unref - release a reference on a paged fragment of an skb. 3029 * @skb: the buffer 3030 * @f: the fragment offset 3031 * 3032 * Releases a reference on the @f'th paged fragment of @skb. 3033 */ 3034static inline void skb_frag_unref(struct sk_buff *skb, int f) 3035{ 3036 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 3037} 3038 3039/** 3040 * skb_frag_address - gets the address of the data contained in a paged fragment 3041 * @frag: the paged fragment buffer 3042 * 3043 * Returns the address of the data within @frag. The page must already 3044 * be mapped. 3045 */ 3046static inline void *skb_frag_address(const skb_frag_t *frag) 3047{ 3048 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3049} 3050 3051/** 3052 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3053 * @frag: the paged fragment buffer 3054 * 3055 * Returns the address of the data within @frag. Checks that the page 3056 * is mapped and returns %NULL otherwise. 3057 */ 3058static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3059{ 3060 void *ptr = page_address(skb_frag_page(frag)); 3061 if (unlikely(!ptr)) 3062 return NULL; 3063 3064 return ptr + skb_frag_off(frag); 3065} 3066 3067/** 3068 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3069 * @fragto: skb fragment where page is set 3070 * @fragfrom: skb fragment page is copied from 3071 */ 3072static inline void skb_frag_page_copy(skb_frag_t *fragto, 3073 const skb_frag_t *fragfrom) 3074{ 3075 fragto->bv_page = fragfrom->bv_page; 3076} 3077 3078/** 3079 * __skb_frag_set_page - sets the page contained in a paged fragment 3080 * @frag: the paged fragment 3081 * @page: the page to set 3082 * 3083 * Sets the fragment @frag to contain @page. 3084 */ 3085static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3086{ 3087 frag->bv_page = page; 3088} 3089 3090/** 3091 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3092 * @skb: the buffer 3093 * @f: the fragment offset 3094 * @page: the page to set 3095 * 3096 * Sets the @f'th fragment of @skb to contain @page. 3097 */ 3098static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3099 struct page *page) 3100{ 3101 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3102} 3103 3104bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3105 3106/** 3107 * skb_frag_dma_map - maps a paged fragment via the DMA API 3108 * @dev: the device to map the fragment to 3109 * @frag: the paged fragment to map 3110 * @offset: the offset within the fragment (starting at the 3111 * fragment's own offset) 3112 * @size: the number of bytes to map 3113 * @dir: the direction of the mapping (``PCI_DMA_*``) 3114 * 3115 * Maps the page associated with @frag to @device. 3116 */ 3117static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3118 const skb_frag_t *frag, 3119 size_t offset, size_t size, 3120 enum dma_data_direction dir) 3121{ 3122 return dma_map_page(dev, skb_frag_page(frag), 3123 skb_frag_off(frag) + offset, size, dir); 3124} 3125 3126static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3127 gfp_t gfp_mask) 3128{ 3129 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3130} 3131 3132 3133static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3134 gfp_t gfp_mask) 3135{ 3136 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3137} 3138 3139 3140/** 3141 * skb_clone_writable - is the header of a clone writable 3142 * @skb: buffer to check 3143 * @len: length up to which to write 3144 * 3145 * Returns true if modifying the header part of the cloned buffer 3146 * does not requires the data to be copied. 3147 */ 3148static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3149{ 3150 return !skb_header_cloned(skb) && 3151 skb_headroom(skb) + len <= skb->hdr_len; 3152} 3153 3154static inline int skb_try_make_writable(struct sk_buff *skb, 3155 unsigned int write_len) 3156{ 3157 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3158 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3159} 3160 3161static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3162 int cloned) 3163{ 3164 int delta = 0; 3165 3166 if (headroom > skb_headroom(skb)) 3167 delta = headroom - skb_headroom(skb); 3168 3169 if (delta || cloned) 3170 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3171 GFP_ATOMIC); 3172 return 0; 3173} 3174 3175/** 3176 * skb_cow - copy header of skb when it is required 3177 * @skb: buffer to cow 3178 * @headroom: needed headroom 3179 * 3180 * If the skb passed lacks sufficient headroom or its data part 3181 * is shared, data is reallocated. If reallocation fails, an error 3182 * is returned and original skb is not changed. 3183 * 3184 * The result is skb with writable area skb->head...skb->tail 3185 * and at least @headroom of space at head. 3186 */ 3187static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3188{ 3189 return __skb_cow(skb, headroom, skb_cloned(skb)); 3190} 3191 3192/** 3193 * skb_cow_head - skb_cow but only making the head writable 3194 * @skb: buffer to cow 3195 * @headroom: needed headroom 3196 * 3197 * This function is identical to skb_cow except that we replace the 3198 * skb_cloned check by skb_header_cloned. It should be used when 3199 * you only need to push on some header and do not need to modify 3200 * the data. 3201 */ 3202static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3203{ 3204 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3205} 3206 3207/** 3208 * skb_padto - pad an skbuff up to a minimal size 3209 * @skb: buffer to pad 3210 * @len: minimal length 3211 * 3212 * Pads up a buffer to ensure the trailing bytes exist and are 3213 * blanked. If the buffer already contains sufficient data it 3214 * is untouched. Otherwise it is extended. Returns zero on 3215 * success. The skb is freed on error. 3216 */ 3217static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3218{ 3219 unsigned int size = skb->len; 3220 if (likely(size >= len)) 3221 return 0; 3222 return skb_pad(skb, len - size); 3223} 3224 3225/** 3226 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3227 * @skb: buffer to pad 3228 * @len: minimal length 3229 * @free_on_error: free buffer on error 3230 * 3231 * Pads up a buffer to ensure the trailing bytes exist and are 3232 * blanked. If the buffer already contains sufficient data it 3233 * is untouched. Otherwise it is extended. Returns zero on 3234 * success. The skb is freed on error if @free_on_error is true. 3235 */ 3236static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3237 unsigned int len, 3238 bool free_on_error) 3239{ 3240 unsigned int size = skb->len; 3241 3242 if (unlikely(size < len)) { 3243 len -= size; 3244 if (__skb_pad(skb, len, free_on_error)) 3245 return -ENOMEM; 3246 __skb_put(skb, len); 3247 } 3248 return 0; 3249} 3250 3251/** 3252 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3253 * @skb: buffer to pad 3254 * @len: minimal length 3255 * 3256 * Pads up a buffer to ensure the trailing bytes exist and are 3257 * blanked. If the buffer already contains sufficient data it 3258 * is untouched. Otherwise it is extended. Returns zero on 3259 * success. The skb is freed on error. 3260 */ 3261static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3262{ 3263 return __skb_put_padto(skb, len, true); 3264} 3265 3266static inline int skb_add_data(struct sk_buff *skb, 3267 struct iov_iter *from, int copy) 3268{ 3269 const int off = skb->len; 3270 3271 if (skb->ip_summed == CHECKSUM_NONE) { 3272 __wsum csum = 0; 3273 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3274 &csum, from)) { 3275 skb->csum = csum_block_add(skb->csum, csum, off); 3276 return 0; 3277 } 3278 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3279 return 0; 3280 3281 __skb_trim(skb, off); 3282 return -EFAULT; 3283} 3284 3285static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3286 const struct page *page, int off) 3287{ 3288 if (skb_zcopy(skb)) 3289 return false; 3290 if (i) { 3291 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3292 3293 return page == skb_frag_page(frag) && 3294 off == skb_frag_off(frag) + skb_frag_size(frag); 3295 } 3296 return false; 3297} 3298 3299static inline int __skb_linearize(struct sk_buff *skb) 3300{ 3301 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3302} 3303 3304/** 3305 * skb_linearize - convert paged skb to linear one 3306 * @skb: buffer to linarize 3307 * 3308 * If there is no free memory -ENOMEM is returned, otherwise zero 3309 * is returned and the old skb data released. 3310 */ 3311static inline int skb_linearize(struct sk_buff *skb) 3312{ 3313 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3314} 3315 3316/** 3317 * skb_has_shared_frag - can any frag be overwritten 3318 * @skb: buffer to test 3319 * 3320 * Return true if the skb has at least one frag that might be modified 3321 * by an external entity (as in vmsplice()/sendfile()) 3322 */ 3323static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3324{ 3325 return skb_is_nonlinear(skb) && 3326 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG; 3327} 3328 3329/** 3330 * skb_linearize_cow - make sure skb is linear and writable 3331 * @skb: buffer to process 3332 * 3333 * If there is no free memory -ENOMEM is returned, otherwise zero 3334 * is returned and the old skb data released. 3335 */ 3336static inline int skb_linearize_cow(struct sk_buff *skb) 3337{ 3338 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3339 __skb_linearize(skb) : 0; 3340} 3341 3342static __always_inline void 3343__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3344 unsigned int off) 3345{ 3346 if (skb->ip_summed == CHECKSUM_COMPLETE) 3347 skb->csum = csum_block_sub(skb->csum, 3348 csum_partial(start, len, 0), off); 3349 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3350 skb_checksum_start_offset(skb) < 0) 3351 skb->ip_summed = CHECKSUM_NONE; 3352} 3353 3354/** 3355 * skb_postpull_rcsum - update checksum for received skb after pull 3356 * @skb: buffer to update 3357 * @start: start of data before pull 3358 * @len: length of data pulled 3359 * 3360 * After doing a pull on a received packet, you need to call this to 3361 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3362 * CHECKSUM_NONE so that it can be recomputed from scratch. 3363 */ 3364static inline void skb_postpull_rcsum(struct sk_buff *skb, 3365 const void *start, unsigned int len) 3366{ 3367 __skb_postpull_rcsum(skb, start, len, 0); 3368} 3369 3370static __always_inline void 3371__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3372 unsigned int off) 3373{ 3374 if (skb->ip_summed == CHECKSUM_COMPLETE) 3375 skb->csum = csum_block_add(skb->csum, 3376 csum_partial(start, len, 0), off); 3377} 3378 3379/** 3380 * skb_postpush_rcsum - update checksum for received skb after push 3381 * @skb: buffer to update 3382 * @start: start of data after push 3383 * @len: length of data pushed 3384 * 3385 * After doing a push on a received packet, you need to call this to 3386 * update the CHECKSUM_COMPLETE checksum. 3387 */ 3388static inline void skb_postpush_rcsum(struct sk_buff *skb, 3389 const void *start, unsigned int len) 3390{ 3391 __skb_postpush_rcsum(skb, start, len, 0); 3392} 3393 3394void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3395 3396/** 3397 * skb_push_rcsum - push skb and update receive checksum 3398 * @skb: buffer to update 3399 * @len: length of data pulled 3400 * 3401 * This function performs an skb_push on the packet and updates 3402 * the CHECKSUM_COMPLETE checksum. It should be used on 3403 * receive path processing instead of skb_push unless you know 3404 * that the checksum difference is zero (e.g., a valid IP header) 3405 * or you are setting ip_summed to CHECKSUM_NONE. 3406 */ 3407static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3408{ 3409 skb_push(skb, len); 3410 skb_postpush_rcsum(skb, skb->data, len); 3411 return skb->data; 3412} 3413 3414int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3415/** 3416 * pskb_trim_rcsum - trim received skb and update checksum 3417 * @skb: buffer to trim 3418 * @len: new length 3419 * 3420 * This is exactly the same as pskb_trim except that it ensures the 3421 * checksum of received packets are still valid after the operation. 3422 * It can change skb pointers. 3423 */ 3424 3425static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3426{ 3427 if (likely(len >= skb->len)) 3428 return 0; 3429 return pskb_trim_rcsum_slow(skb, len); 3430} 3431 3432static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3433{ 3434 if (skb->ip_summed == CHECKSUM_COMPLETE) 3435 skb->ip_summed = CHECKSUM_NONE; 3436 __skb_trim(skb, len); 3437 return 0; 3438} 3439 3440static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3441{ 3442 if (skb->ip_summed == CHECKSUM_COMPLETE) 3443 skb->ip_summed = CHECKSUM_NONE; 3444 return __skb_grow(skb, len); 3445} 3446 3447#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3448#define skb_rb_first(root) rb_to_skb(rb_first(root)) 3449#define skb_rb_last(root) rb_to_skb(rb_last(root)) 3450#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3451#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3452 3453#define skb_queue_walk(queue, skb) \ 3454 for (skb = (queue)->next; \ 3455 skb != (struct sk_buff *)(queue); \ 3456 skb = skb->next) 3457 3458#define skb_queue_walk_safe(queue, skb, tmp) \ 3459 for (skb = (queue)->next, tmp = skb->next; \ 3460 skb != (struct sk_buff *)(queue); \ 3461 skb = tmp, tmp = skb->next) 3462 3463#define skb_queue_walk_from(queue, skb) \ 3464 for (; skb != (struct sk_buff *)(queue); \ 3465 skb = skb->next) 3466 3467#define skb_rbtree_walk(skb, root) \ 3468 for (skb = skb_rb_first(root); skb != NULL; \ 3469 skb = skb_rb_next(skb)) 3470 3471#define skb_rbtree_walk_from(skb) \ 3472 for (; skb != NULL; \ 3473 skb = skb_rb_next(skb)) 3474 3475#define skb_rbtree_walk_from_safe(skb, tmp) \ 3476 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3477 skb = tmp) 3478 3479#define skb_queue_walk_from_safe(queue, skb, tmp) \ 3480 for (tmp = skb->next; \ 3481 skb != (struct sk_buff *)(queue); \ 3482 skb = tmp, tmp = skb->next) 3483 3484#define skb_queue_reverse_walk(queue, skb) \ 3485 for (skb = (queue)->prev; \ 3486 skb != (struct sk_buff *)(queue); \ 3487 skb = skb->prev) 3488 3489#define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3490 for (skb = (queue)->prev, tmp = skb->prev; \ 3491 skb != (struct sk_buff *)(queue); \ 3492 skb = tmp, tmp = skb->prev) 3493 3494#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3495 for (tmp = skb->prev; \ 3496 skb != (struct sk_buff *)(queue); \ 3497 skb = tmp, tmp = skb->prev) 3498 3499static inline bool skb_has_frag_list(const struct sk_buff *skb) 3500{ 3501 return skb_shinfo(skb)->frag_list != NULL; 3502} 3503 3504static inline void skb_frag_list_init(struct sk_buff *skb) 3505{ 3506 skb_shinfo(skb)->frag_list = NULL; 3507} 3508 3509#define skb_walk_frags(skb, iter) \ 3510 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3511 3512 3513int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3514 int *err, long *timeo_p, 3515 const struct sk_buff *skb); 3516struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3517 struct sk_buff_head *queue, 3518 unsigned int flags, 3519 int *off, int *err, 3520 struct sk_buff **last); 3521struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3522 struct sk_buff_head *queue, 3523 unsigned int flags, int *off, int *err, 3524 struct sk_buff **last); 3525struct sk_buff *__skb_recv_datagram(struct sock *sk, 3526 struct sk_buff_head *sk_queue, 3527 unsigned int flags, int *off, int *err); 3528struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3529 int *err); 3530__poll_t datagram_poll(struct file *file, struct socket *sock, 3531 struct poll_table_struct *wait); 3532int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3533 struct iov_iter *to, int size); 3534static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3535 struct msghdr *msg, int size) 3536{ 3537 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3538} 3539int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3540 struct msghdr *msg); 3541int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3542 struct iov_iter *to, int len, 3543 struct ahash_request *hash); 3544int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3545 struct iov_iter *from, int len); 3546int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3547void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3548void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3549static inline void skb_free_datagram_locked(struct sock *sk, 3550 struct sk_buff *skb) 3551{ 3552 __skb_free_datagram_locked(sk, skb, 0); 3553} 3554int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3555int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3556int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3557__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3558 int len); 3559int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3560 struct pipe_inode_info *pipe, unsigned int len, 3561 unsigned int flags); 3562int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3563 int len); 3564void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3565unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3566int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3567 int len, int hlen); 3568void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3569int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3570void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3571bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3572bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3573struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3574struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3575 unsigned int offset); 3576struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3577int skb_ensure_writable(struct sk_buff *skb, int write_len); 3578int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3579int skb_vlan_pop(struct sk_buff *skb); 3580int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3581int skb_eth_pop(struct sk_buff *skb); 3582int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3583 const unsigned char *src); 3584int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3585 int mac_len, bool ethernet); 3586int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3587 bool ethernet); 3588int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3589int skb_mpls_dec_ttl(struct sk_buff *skb); 3590struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3591 gfp_t gfp); 3592 3593static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3594{ 3595 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3596} 3597 3598static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3599{ 3600 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3601} 3602 3603struct skb_checksum_ops { 3604 __wsum (*update)(const void *mem, int len, __wsum wsum); 3605 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3606}; 3607 3608extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3609 3610__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3611 __wsum csum, const struct skb_checksum_ops *ops); 3612__wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3613 __wsum csum); 3614 3615static inline void * __must_check 3616__skb_header_pointer(const struct sk_buff *skb, int offset, 3617 int len, void *data, int hlen, void *buffer) 3618{ 3619 if (hlen - offset >= len) 3620 return data + offset; 3621 3622 if (!skb || 3623 skb_copy_bits(skb, offset, buffer, len) < 0) 3624 return NULL; 3625 3626 return buffer; 3627} 3628 3629static inline void * __must_check 3630skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3631{ 3632 return __skb_header_pointer(skb, offset, len, skb->data, 3633 skb_headlen(skb), buffer); 3634} 3635 3636/** 3637 * skb_needs_linearize - check if we need to linearize a given skb 3638 * depending on the given device features. 3639 * @skb: socket buffer to check 3640 * @features: net device features 3641 * 3642 * Returns true if either: 3643 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3644 * 2. skb is fragmented and the device does not support SG. 3645 */ 3646static inline bool skb_needs_linearize(struct sk_buff *skb, 3647 netdev_features_t features) 3648{ 3649 return skb_is_nonlinear(skb) && 3650 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3651 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3652} 3653 3654static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3655 void *to, 3656 const unsigned int len) 3657{ 3658 memcpy(to, skb->data, len); 3659} 3660 3661static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3662 const int offset, void *to, 3663 const unsigned int len) 3664{ 3665 memcpy(to, skb->data + offset, len); 3666} 3667 3668static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3669 const void *from, 3670 const unsigned int len) 3671{ 3672 memcpy(skb->data, from, len); 3673} 3674 3675static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3676 const int offset, 3677 const void *from, 3678 const unsigned int len) 3679{ 3680 memcpy(skb->data + offset, from, len); 3681} 3682 3683void skb_init(void); 3684 3685static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3686{ 3687 return skb->tstamp; 3688} 3689 3690/** 3691 * skb_get_timestamp - get timestamp from a skb 3692 * @skb: skb to get stamp from 3693 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3694 * 3695 * Timestamps are stored in the skb as offsets to a base timestamp. 3696 * This function converts the offset back to a struct timeval and stores 3697 * it in stamp. 3698 */ 3699static inline void skb_get_timestamp(const struct sk_buff *skb, 3700 struct __kernel_old_timeval *stamp) 3701{ 3702 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3703} 3704 3705static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3706 struct __kernel_sock_timeval *stamp) 3707{ 3708 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3709 3710 stamp->tv_sec = ts.tv_sec; 3711 stamp->tv_usec = ts.tv_nsec / 1000; 3712} 3713 3714static inline void skb_get_timestampns(const struct sk_buff *skb, 3715 struct __kernel_old_timespec *stamp) 3716{ 3717 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3718 3719 stamp->tv_sec = ts.tv_sec; 3720 stamp->tv_nsec = ts.tv_nsec; 3721} 3722 3723static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3724 struct __kernel_timespec *stamp) 3725{ 3726 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3727 3728 stamp->tv_sec = ts.tv_sec; 3729 stamp->tv_nsec = ts.tv_nsec; 3730} 3731 3732static inline void __net_timestamp(struct sk_buff *skb) 3733{ 3734 skb->tstamp = ktime_get_real(); 3735} 3736 3737static inline ktime_t net_timedelta(ktime_t t) 3738{ 3739 return ktime_sub(ktime_get_real(), t); 3740} 3741 3742static inline ktime_t net_invalid_timestamp(void) 3743{ 3744 return 0; 3745} 3746 3747static inline u8 skb_metadata_len(const struct sk_buff *skb) 3748{ 3749 return skb_shinfo(skb)->meta_len; 3750} 3751 3752static inline void *skb_metadata_end(const struct sk_buff *skb) 3753{ 3754 return skb_mac_header(skb); 3755} 3756 3757static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3758 const struct sk_buff *skb_b, 3759 u8 meta_len) 3760{ 3761 const void *a = skb_metadata_end(skb_a); 3762 const void *b = skb_metadata_end(skb_b); 3763 /* Using more efficient varaiant than plain call to memcmp(). */ 3764#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3765 u64 diffs = 0; 3766 3767 switch (meta_len) { 3768#define __it(x, op) (x -= sizeof(u##op)) 3769#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3770 case 32: diffs |= __it_diff(a, b, 64); 3771 fallthrough; 3772 case 24: diffs |= __it_diff(a, b, 64); 3773 fallthrough; 3774 case 16: diffs |= __it_diff(a, b, 64); 3775 fallthrough; 3776 case 8: diffs |= __it_diff(a, b, 64); 3777 break; 3778 case 28: diffs |= __it_diff(a, b, 64); 3779 fallthrough; 3780 case 20: diffs |= __it_diff(a, b, 64); 3781 fallthrough; 3782 case 12: diffs |= __it_diff(a, b, 64); 3783 fallthrough; 3784 case 4: diffs |= __it_diff(a, b, 32); 3785 break; 3786 } 3787 return diffs; 3788#else 3789 return memcmp(a - meta_len, b - meta_len, meta_len); 3790#endif 3791} 3792 3793static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3794 const struct sk_buff *skb_b) 3795{ 3796 u8 len_a = skb_metadata_len(skb_a); 3797 u8 len_b = skb_metadata_len(skb_b); 3798 3799 if (!(len_a | len_b)) 3800 return false; 3801 3802 return len_a != len_b ? 3803 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3804} 3805 3806static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3807{ 3808 skb_shinfo(skb)->meta_len = meta_len; 3809} 3810 3811static inline void skb_metadata_clear(struct sk_buff *skb) 3812{ 3813 skb_metadata_set(skb, 0); 3814} 3815 3816struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3817 3818#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3819 3820void skb_clone_tx_timestamp(struct sk_buff *skb); 3821bool skb_defer_rx_timestamp(struct sk_buff *skb); 3822 3823#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3824 3825static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3826{ 3827} 3828 3829static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3830{ 3831 return false; 3832} 3833 3834#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3835 3836/** 3837 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3838 * 3839 * PHY drivers may accept clones of transmitted packets for 3840 * timestamping via their phy_driver.txtstamp method. These drivers 3841 * must call this function to return the skb back to the stack with a 3842 * timestamp. 3843 * 3844 * @skb: clone of the original outgoing packet 3845 * @hwtstamps: hardware time stamps 3846 * 3847 */ 3848void skb_complete_tx_timestamp(struct sk_buff *skb, 3849 struct skb_shared_hwtstamps *hwtstamps); 3850 3851void __skb_tstamp_tx(struct sk_buff *orig_skb, 3852 struct skb_shared_hwtstamps *hwtstamps, 3853 struct sock *sk, int tstype); 3854 3855/** 3856 * skb_tstamp_tx - queue clone of skb with send time stamps 3857 * @orig_skb: the original outgoing packet 3858 * @hwtstamps: hardware time stamps, may be NULL if not available 3859 * 3860 * If the skb has a socket associated, then this function clones the 3861 * skb (thus sharing the actual data and optional structures), stores 3862 * the optional hardware time stamping information (if non NULL) or 3863 * generates a software time stamp (otherwise), then queues the clone 3864 * to the error queue of the socket. Errors are silently ignored. 3865 */ 3866void skb_tstamp_tx(struct sk_buff *orig_skb, 3867 struct skb_shared_hwtstamps *hwtstamps); 3868 3869/** 3870 * skb_tx_timestamp() - Driver hook for transmit timestamping 3871 * 3872 * Ethernet MAC Drivers should call this function in their hard_xmit() 3873 * function immediately before giving the sk_buff to the MAC hardware. 3874 * 3875 * Specifically, one should make absolutely sure that this function is 3876 * called before TX completion of this packet can trigger. Otherwise 3877 * the packet could potentially already be freed. 3878 * 3879 * @skb: A socket buffer. 3880 */ 3881static inline void skb_tx_timestamp(struct sk_buff *skb) 3882{ 3883 skb_clone_tx_timestamp(skb); 3884 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3885 skb_tstamp_tx(skb, NULL); 3886} 3887 3888/** 3889 * skb_complete_wifi_ack - deliver skb with wifi status 3890 * 3891 * @skb: the original outgoing packet 3892 * @acked: ack status 3893 * 3894 */ 3895void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3896 3897__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3898__sum16 __skb_checksum_complete(struct sk_buff *skb); 3899 3900static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3901{ 3902 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3903 skb->csum_valid || 3904 (skb->ip_summed == CHECKSUM_PARTIAL && 3905 skb_checksum_start_offset(skb) >= 0)); 3906} 3907 3908/** 3909 * skb_checksum_complete - Calculate checksum of an entire packet 3910 * @skb: packet to process 3911 * 3912 * This function calculates the checksum over the entire packet plus 3913 * the value of skb->csum. The latter can be used to supply the 3914 * checksum of a pseudo header as used by TCP/UDP. It returns the 3915 * checksum. 3916 * 3917 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 3918 * this function can be used to verify that checksum on received 3919 * packets. In that case the function should return zero if the 3920 * checksum is correct. In particular, this function will return zero 3921 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 3922 * hardware has already verified the correctness of the checksum. 3923 */ 3924static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 3925{ 3926 return skb_csum_unnecessary(skb) ? 3927 0 : __skb_checksum_complete(skb); 3928} 3929 3930static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 3931{ 3932 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3933 if (skb->csum_level == 0) 3934 skb->ip_summed = CHECKSUM_NONE; 3935 else 3936 skb->csum_level--; 3937 } 3938} 3939 3940static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 3941{ 3942 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3943 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 3944 skb->csum_level++; 3945 } else if (skb->ip_summed == CHECKSUM_NONE) { 3946 skb->ip_summed = CHECKSUM_UNNECESSARY; 3947 skb->csum_level = 0; 3948 } 3949} 3950 3951static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 3952{ 3953 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3954 skb->ip_summed = CHECKSUM_NONE; 3955 skb->csum_level = 0; 3956 } 3957} 3958 3959/* Check if we need to perform checksum complete validation. 3960 * 3961 * Returns true if checksum complete is needed, false otherwise 3962 * (either checksum is unnecessary or zero checksum is allowed). 3963 */ 3964static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 3965 bool zero_okay, 3966 __sum16 check) 3967{ 3968 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 3969 skb->csum_valid = 1; 3970 __skb_decr_checksum_unnecessary(skb); 3971 return false; 3972 } 3973 3974 return true; 3975} 3976 3977/* For small packets <= CHECKSUM_BREAK perform checksum complete directly 3978 * in checksum_init. 3979 */ 3980#define CHECKSUM_BREAK 76 3981 3982/* Unset checksum-complete 3983 * 3984 * Unset checksum complete can be done when packet is being modified 3985 * (uncompressed for instance) and checksum-complete value is 3986 * invalidated. 3987 */ 3988static inline void skb_checksum_complete_unset(struct sk_buff *skb) 3989{ 3990 if (skb->ip_summed == CHECKSUM_COMPLETE) 3991 skb->ip_summed = CHECKSUM_NONE; 3992} 3993 3994/* Validate (init) checksum based on checksum complete. 3995 * 3996 * Return values: 3997 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 3998 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 3999 * checksum is stored in skb->csum for use in __skb_checksum_complete 4000 * non-zero: value of invalid checksum 4001 * 4002 */ 4003static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4004 bool complete, 4005 __wsum psum) 4006{ 4007 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4008 if (!csum_fold(csum_add(psum, skb->csum))) { 4009 skb->csum_valid = 1; 4010 return 0; 4011 } 4012 } 4013 4014 skb->csum = psum; 4015 4016 if (complete || skb->len <= CHECKSUM_BREAK) { 4017 __sum16 csum; 4018 4019 csum = __skb_checksum_complete(skb); 4020 skb->csum_valid = !csum; 4021 return csum; 4022 } 4023 4024 return 0; 4025} 4026 4027static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4028{ 4029 return 0; 4030} 4031 4032/* Perform checksum validate (init). Note that this is a macro since we only 4033 * want to calculate the pseudo header which is an input function if necessary. 4034 * First we try to validate without any computation (checksum unnecessary) and 4035 * then calculate based on checksum complete calling the function to compute 4036 * pseudo header. 4037 * 4038 * Return values: 4039 * 0: checksum is validated or try to in skb_checksum_complete 4040 * non-zero: value of invalid checksum 4041 */ 4042#define __skb_checksum_validate(skb, proto, complete, \ 4043 zero_okay, check, compute_pseudo) \ 4044({ \ 4045 __sum16 __ret = 0; \ 4046 skb->csum_valid = 0; \ 4047 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4048 __ret = __skb_checksum_validate_complete(skb, \ 4049 complete, compute_pseudo(skb, proto)); \ 4050 __ret; \ 4051}) 4052 4053#define skb_checksum_init(skb, proto, compute_pseudo) \ 4054 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4055 4056#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4057 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4058 4059#define skb_checksum_validate(skb, proto, compute_pseudo) \ 4060 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4061 4062#define skb_checksum_validate_zero_check(skb, proto, check, \ 4063 compute_pseudo) \ 4064 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4065 4066#define skb_checksum_simple_validate(skb) \ 4067 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4068 4069static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4070{ 4071 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4072} 4073 4074static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4075{ 4076 skb->csum = ~pseudo; 4077 skb->ip_summed = CHECKSUM_COMPLETE; 4078} 4079 4080#define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4081do { \ 4082 if (__skb_checksum_convert_check(skb)) \ 4083 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4084} while (0) 4085 4086static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4087 u16 start, u16 offset) 4088{ 4089 skb->ip_summed = CHECKSUM_PARTIAL; 4090 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4091 skb->csum_offset = offset - start; 4092} 4093 4094/* Update skbuf and packet to reflect the remote checksum offload operation. 4095 * When called, ptr indicates the starting point for skb->csum when 4096 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4097 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4098 */ 4099static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4100 int start, int offset, bool nopartial) 4101{ 4102 __wsum delta; 4103 4104 if (!nopartial) { 4105 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4106 return; 4107 } 4108 4109 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4110 __skb_checksum_complete(skb); 4111 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4112 } 4113 4114 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4115 4116 /* Adjust skb->csum since we changed the packet */ 4117 skb->csum = csum_add(skb->csum, delta); 4118} 4119 4120static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4121{ 4122#if IS_ENABLED(CONFIG_NF_CONNTRACK) 4123 return (void *)(skb->_nfct & NFCT_PTRMASK); 4124#else 4125 return NULL; 4126#endif 4127} 4128 4129static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4130{ 4131#if IS_ENABLED(CONFIG_NF_CONNTRACK) 4132 return skb->_nfct; 4133#else 4134 return 0UL; 4135#endif 4136} 4137 4138static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4139{ 4140#if IS_ENABLED(CONFIG_NF_CONNTRACK) 4141 skb->_nfct = nfct; 4142#endif 4143} 4144 4145#ifdef CONFIG_SKB_EXTENSIONS 4146enum skb_ext_id { 4147#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4148 SKB_EXT_BRIDGE_NF, 4149#endif 4150#ifdef CONFIG_XFRM 4151 SKB_EXT_SEC_PATH, 4152#endif 4153#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4154 TC_SKB_EXT, 4155#endif 4156#if IS_ENABLED(CONFIG_MPTCP) 4157 SKB_EXT_MPTCP, 4158#endif 4159 SKB_EXT_NUM, /* must be last */ 4160}; 4161 4162/** 4163 * struct skb_ext - sk_buff extensions 4164 * @refcnt: 1 on allocation, deallocated on 0 4165 * @offset: offset to add to @data to obtain extension address 4166 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4167 * @data: start of extension data, variable sized 4168 * 4169 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4170 * to use 'u8' types while allowing up to 2kb worth of extension data. 4171 */ 4172struct skb_ext { 4173 refcount_t refcnt; 4174 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4175 u8 chunks; /* same */ 4176 char data[] __aligned(8); 4177}; 4178 4179struct skb_ext *__skb_ext_alloc(gfp_t flags); 4180void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4181 struct skb_ext *ext); 4182void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4183void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4184void __skb_ext_put(struct skb_ext *ext); 4185 4186static inline void skb_ext_put(struct sk_buff *skb) 4187{ 4188 if (skb->active_extensions) 4189 __skb_ext_put(skb->extensions); 4190} 4191 4192static inline void __skb_ext_copy(struct sk_buff *dst, 4193 const struct sk_buff *src) 4194{ 4195 dst->active_extensions = src->active_extensions; 4196 4197 if (src->active_extensions) { 4198 struct skb_ext *ext = src->extensions; 4199 4200 refcount_inc(&ext->refcnt); 4201 dst->extensions = ext; 4202 } 4203} 4204 4205static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4206{ 4207 skb_ext_put(dst); 4208 __skb_ext_copy(dst, src); 4209} 4210 4211static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4212{ 4213 return !!ext->offset[i]; 4214} 4215 4216static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4217{ 4218 return skb->active_extensions & (1 << id); 4219} 4220 4221static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4222{ 4223 if (skb_ext_exist(skb, id)) 4224 __skb_ext_del(skb, id); 4225} 4226 4227static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4228{ 4229 if (skb_ext_exist(skb, id)) { 4230 struct skb_ext *ext = skb->extensions; 4231 4232 return (void *)ext + (ext->offset[id] << 3); 4233 } 4234 4235 return NULL; 4236} 4237 4238static inline void skb_ext_reset(struct sk_buff *skb) 4239{ 4240 if (unlikely(skb->active_extensions)) { 4241 __skb_ext_put(skb->extensions); 4242 skb->active_extensions = 0; 4243 } 4244} 4245 4246static inline bool skb_has_extensions(struct sk_buff *skb) 4247{ 4248 return unlikely(skb->active_extensions); 4249} 4250#else 4251static inline void skb_ext_put(struct sk_buff *skb) {} 4252static inline void skb_ext_reset(struct sk_buff *skb) {} 4253static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4254static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4255static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4256static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4257#endif /* CONFIG_SKB_EXTENSIONS */ 4258 4259static inline void nf_reset_ct(struct sk_buff *skb) 4260{ 4261#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4262 nf_conntrack_put(skb_nfct(skb)); 4263 skb->_nfct = 0; 4264#endif 4265} 4266 4267static inline void nf_reset_trace(struct sk_buff *skb) 4268{ 4269#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4270 skb->nf_trace = 0; 4271#endif 4272} 4273 4274static inline void ipvs_reset(struct sk_buff *skb) 4275{ 4276#if IS_ENABLED(CONFIG_IP_VS) 4277 skb->ipvs_property = 0; 4278#endif 4279} 4280 4281/* Note: This doesn't put any conntrack info in dst. */ 4282static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4283 bool copy) 4284{ 4285#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4286 dst->_nfct = src->_nfct; 4287 nf_conntrack_get(skb_nfct(src)); 4288#endif 4289#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4290 if (copy) 4291 dst->nf_trace = src->nf_trace; 4292#endif 4293} 4294 4295static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4296{ 4297#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4298 nf_conntrack_put(skb_nfct(dst)); 4299#endif 4300 __nf_copy(dst, src, true); 4301} 4302 4303#ifdef CONFIG_NETWORK_SECMARK 4304static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4305{ 4306 to->secmark = from->secmark; 4307} 4308 4309static inline void skb_init_secmark(struct sk_buff *skb) 4310{ 4311 skb->secmark = 0; 4312} 4313#else 4314static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4315{ } 4316 4317static inline void skb_init_secmark(struct sk_buff *skb) 4318{ } 4319#endif 4320 4321static inline int secpath_exists(const struct sk_buff *skb) 4322{ 4323#ifdef CONFIG_XFRM 4324 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4325#else 4326 return 0; 4327#endif 4328} 4329 4330static inline bool skb_irq_freeable(const struct sk_buff *skb) 4331{ 4332 return !skb->destructor && 4333 !secpath_exists(skb) && 4334 !skb_nfct(skb) && 4335 !skb->_skb_refdst && 4336 !skb_has_frag_list(skb); 4337} 4338 4339static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4340{ 4341 skb->queue_mapping = queue_mapping; 4342} 4343 4344static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4345{ 4346 return skb->queue_mapping; 4347} 4348 4349static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4350{ 4351 to->queue_mapping = from->queue_mapping; 4352} 4353 4354static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4355{ 4356 skb->queue_mapping = rx_queue + 1; 4357} 4358 4359static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4360{ 4361 return skb->queue_mapping - 1; 4362} 4363 4364static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4365{ 4366 return skb->queue_mapping != 0; 4367} 4368 4369static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4370{ 4371 skb->dst_pending_confirm = val; 4372} 4373 4374static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4375{ 4376 return skb->dst_pending_confirm != 0; 4377} 4378 4379static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4380{ 4381#ifdef CONFIG_XFRM 4382 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4383#else 4384 return NULL; 4385#endif 4386} 4387 4388/* Keeps track of mac header offset relative to skb->head. 4389 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4390 * For non-tunnel skb it points to skb_mac_header() and for 4391 * tunnel skb it points to outer mac header. 4392 * Keeps track of level of encapsulation of network headers. 4393 */ 4394struct skb_gso_cb { 4395 union { 4396 int mac_offset; 4397 int data_offset; 4398 }; 4399 int encap_level; 4400 __wsum csum; 4401 __u16 csum_start; 4402}; 4403#define SKB_GSO_CB_OFFSET 32 4404#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4405 4406static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4407{ 4408 return (skb_mac_header(inner_skb) - inner_skb->head) - 4409 SKB_GSO_CB(inner_skb)->mac_offset; 4410} 4411 4412static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4413{ 4414 int new_headroom, headroom; 4415 int ret; 4416 4417 headroom = skb_headroom(skb); 4418 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4419 if (ret) 4420 return ret; 4421 4422 new_headroom = skb_headroom(skb); 4423 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4424 return 0; 4425} 4426 4427static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4428{ 4429 /* Do not update partial checksums if remote checksum is enabled. */ 4430 if (skb->remcsum_offload) 4431 return; 4432 4433 SKB_GSO_CB(skb)->csum = res; 4434 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4435} 4436 4437/* Compute the checksum for a gso segment. First compute the checksum value 4438 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4439 * then add in skb->csum (checksum from csum_start to end of packet). 4440 * skb->csum and csum_start are then updated to reflect the checksum of the 4441 * resultant packet starting from the transport header-- the resultant checksum 4442 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4443 * header. 4444 */ 4445static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4446{ 4447 unsigned char *csum_start = skb_transport_header(skb); 4448 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4449 __wsum partial = SKB_GSO_CB(skb)->csum; 4450 4451 SKB_GSO_CB(skb)->csum = res; 4452 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4453 4454 return csum_fold(csum_partial(csum_start, plen, partial)); 4455} 4456 4457static inline bool skb_is_gso(const struct sk_buff *skb) 4458{ 4459 return skb_shinfo(skb)->gso_size; 4460} 4461 4462/* Note: Should be called only if skb_is_gso(skb) is true */ 4463static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4464{ 4465 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4466} 4467 4468/* Note: Should be called only if skb_is_gso(skb) is true */ 4469static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4470{ 4471 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4472} 4473 4474/* Note: Should be called only if skb_is_gso(skb) is true */ 4475static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4476{ 4477 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4478} 4479 4480static inline void skb_gso_reset(struct sk_buff *skb) 4481{ 4482 skb_shinfo(skb)->gso_size = 0; 4483 skb_shinfo(skb)->gso_segs = 0; 4484 skb_shinfo(skb)->gso_type = 0; 4485} 4486 4487static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4488 u16 increment) 4489{ 4490 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4491 return; 4492 shinfo->gso_size += increment; 4493} 4494 4495static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4496 u16 decrement) 4497{ 4498 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4499 return; 4500 shinfo->gso_size -= decrement; 4501} 4502 4503void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4504 4505static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4506{ 4507 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4508 * wanted then gso_type will be set. */ 4509 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4510 4511 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4512 unlikely(shinfo->gso_type == 0)) { 4513 __skb_warn_lro_forwarding(skb); 4514 return true; 4515 } 4516 return false; 4517} 4518 4519static inline void skb_forward_csum(struct sk_buff *skb) 4520{ 4521 /* Unfortunately we don't support this one. Any brave souls? */ 4522 if (skb->ip_summed == CHECKSUM_COMPLETE) 4523 skb->ip_summed = CHECKSUM_NONE; 4524} 4525 4526/** 4527 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4528 * @skb: skb to check 4529 * 4530 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4531 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4532 * use this helper, to document places where we make this assertion. 4533 */ 4534static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4535{ 4536#ifdef DEBUG 4537 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4538#endif 4539} 4540 4541bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4542 4543int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4544struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4545 unsigned int transport_len, 4546 __sum16(*skb_chkf)(struct sk_buff *skb)); 4547 4548/** 4549 * skb_head_is_locked - Determine if the skb->head is locked down 4550 * @skb: skb to check 4551 * 4552 * The head on skbs build around a head frag can be removed if they are 4553 * not cloned. This function returns true if the skb head is locked down 4554 * due to either being allocated via kmalloc, or by being a clone with 4555 * multiple references to the head. 4556 */ 4557static inline bool skb_head_is_locked(const struct sk_buff *skb) 4558{ 4559 return !skb->head_frag || skb_cloned(skb); 4560} 4561 4562/* Local Checksum Offload. 4563 * Compute outer checksum based on the assumption that the 4564 * inner checksum will be offloaded later. 4565 * See Documentation/networking/checksum-offloads.rst for 4566 * explanation of how this works. 4567 * Fill in outer checksum adjustment (e.g. with sum of outer 4568 * pseudo-header) before calling. 4569 * Also ensure that inner checksum is in linear data area. 4570 */ 4571static inline __wsum lco_csum(struct sk_buff *skb) 4572{ 4573 unsigned char *csum_start = skb_checksum_start(skb); 4574 unsigned char *l4_hdr = skb_transport_header(skb); 4575 __wsum partial; 4576 4577 /* Start with complement of inner checksum adjustment */ 4578 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4579 skb->csum_offset)); 4580 4581 /* Add in checksum of our headers (incl. outer checksum 4582 * adjustment filled in by caller) and return result. 4583 */ 4584 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4585} 4586 4587static inline bool skb_is_redirected(const struct sk_buff *skb) 4588{ 4589#ifdef CONFIG_NET_REDIRECT 4590 return skb->redirected; 4591#else 4592 return false; 4593#endif 4594} 4595 4596static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4597{ 4598#ifdef CONFIG_NET_REDIRECT 4599 skb->redirected = 1; 4600 skb->from_ingress = from_ingress; 4601 if (skb->from_ingress) 4602 skb->tstamp = 0; 4603#endif 4604} 4605 4606static inline void skb_reset_redirect(struct sk_buff *skb) 4607{ 4608#ifdef CONFIG_NET_REDIRECT 4609 skb->redirected = 0; 4610#endif 4611} 4612 4613static inline void skb_set_kcov_handle(struct sk_buff *skb, 4614 const u64 kcov_handle) 4615{ 4616#ifdef CONFIG_KCOV 4617 skb->kcov_handle = kcov_handle; 4618#endif 4619} 4620 4621static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4622{ 4623#ifdef CONFIG_KCOV 4624 return skb->kcov_handle; 4625#else 4626 return 0; 4627#endif 4628} 4629 4630#endif /* __KERNEL__ */ 4631#endif /* _LINUX_SKBUFF_H */