at v5.11 134 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 (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || 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 __u32 frag_off; 1207}; 1208 1209void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1210 unsigned int to, struct skb_seq_state *st); 1211unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1212 struct skb_seq_state *st); 1213void skb_abort_seq_read(struct skb_seq_state *st); 1214 1215unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1216 unsigned int to, struct ts_config *config); 1217 1218/* 1219 * Packet hash types specify the type of hash in skb_set_hash. 1220 * 1221 * Hash types refer to the protocol layer addresses which are used to 1222 * construct a packet's hash. The hashes are used to differentiate or identify 1223 * flows of the protocol layer for the hash type. Hash types are either 1224 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1225 * 1226 * Properties of hashes: 1227 * 1228 * 1) Two packets in different flows have different hash values 1229 * 2) Two packets in the same flow should have the same hash value 1230 * 1231 * A hash at a higher layer is considered to be more specific. A driver should 1232 * set the most specific hash possible. 1233 * 1234 * A driver cannot indicate a more specific hash than the layer at which a hash 1235 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1236 * 1237 * A driver may indicate a hash level which is less specific than the 1238 * actual layer the hash was computed on. For instance, a hash computed 1239 * at L4 may be considered an L3 hash. This should only be done if the 1240 * driver can't unambiguously determine that the HW computed the hash at 1241 * the higher layer. Note that the "should" in the second property above 1242 * permits this. 1243 */ 1244enum pkt_hash_types { 1245 PKT_HASH_TYPE_NONE, /* Undefined type */ 1246 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1247 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1248 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1249}; 1250 1251static inline void skb_clear_hash(struct sk_buff *skb) 1252{ 1253 skb->hash = 0; 1254 skb->sw_hash = 0; 1255 skb->l4_hash = 0; 1256} 1257 1258static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1259{ 1260 if (!skb->l4_hash) 1261 skb_clear_hash(skb); 1262} 1263 1264static inline void 1265__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1266{ 1267 skb->l4_hash = is_l4; 1268 skb->sw_hash = is_sw; 1269 skb->hash = hash; 1270} 1271 1272static inline void 1273skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1274{ 1275 /* Used by drivers to set hash from HW */ 1276 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1277} 1278 1279static inline void 1280__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1281{ 1282 __skb_set_hash(skb, hash, true, is_l4); 1283} 1284 1285void __skb_get_hash(struct sk_buff *skb); 1286u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1287u32 skb_get_poff(const struct sk_buff *skb); 1288u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1289 const struct flow_keys_basic *keys, int hlen); 1290__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1291 void *data, int hlen_proto); 1292 1293static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1294 int thoff, u8 ip_proto) 1295{ 1296 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1297} 1298 1299void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1300 const struct flow_dissector_key *key, 1301 unsigned int key_count); 1302 1303struct bpf_flow_dissector; 1304bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1305 __be16 proto, int nhoff, int hlen, unsigned int flags); 1306 1307bool __skb_flow_dissect(const struct net *net, 1308 const struct sk_buff *skb, 1309 struct flow_dissector *flow_dissector, 1310 void *target_container, 1311 void *data, __be16 proto, int nhoff, int hlen, 1312 unsigned int flags); 1313 1314static inline bool skb_flow_dissect(const struct sk_buff *skb, 1315 struct flow_dissector *flow_dissector, 1316 void *target_container, unsigned int flags) 1317{ 1318 return __skb_flow_dissect(NULL, skb, flow_dissector, 1319 target_container, NULL, 0, 0, 0, flags); 1320} 1321 1322static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1323 struct flow_keys *flow, 1324 unsigned int flags) 1325{ 1326 memset(flow, 0, sizeof(*flow)); 1327 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1328 flow, NULL, 0, 0, 0, flags); 1329} 1330 1331static inline bool 1332skb_flow_dissect_flow_keys_basic(const struct net *net, 1333 const struct sk_buff *skb, 1334 struct flow_keys_basic *flow, void *data, 1335 __be16 proto, int nhoff, int hlen, 1336 unsigned int flags) 1337{ 1338 memset(flow, 0, sizeof(*flow)); 1339 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1340 data, proto, nhoff, hlen, flags); 1341} 1342 1343void skb_flow_dissect_meta(const struct sk_buff *skb, 1344 struct flow_dissector *flow_dissector, 1345 void *target_container); 1346 1347/* Gets a skb connection tracking info, ctinfo map should be a 1348 * map of mapsize to translate enum ip_conntrack_info states 1349 * to user states. 1350 */ 1351void 1352skb_flow_dissect_ct(const struct sk_buff *skb, 1353 struct flow_dissector *flow_dissector, 1354 void *target_container, 1355 u16 *ctinfo_map, 1356 size_t mapsize); 1357void 1358skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1359 struct flow_dissector *flow_dissector, 1360 void *target_container); 1361 1362void skb_flow_dissect_hash(const struct sk_buff *skb, 1363 struct flow_dissector *flow_dissector, 1364 void *target_container); 1365 1366static inline __u32 skb_get_hash(struct sk_buff *skb) 1367{ 1368 if (!skb->l4_hash && !skb->sw_hash) 1369 __skb_get_hash(skb); 1370 1371 return skb->hash; 1372} 1373 1374static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1375{ 1376 if (!skb->l4_hash && !skb->sw_hash) { 1377 struct flow_keys keys; 1378 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1379 1380 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1381 } 1382 1383 return skb->hash; 1384} 1385 1386__u32 skb_get_hash_perturb(const struct sk_buff *skb, 1387 const siphash_key_t *perturb); 1388 1389static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1390{ 1391 return skb->hash; 1392} 1393 1394static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1395{ 1396 to->hash = from->hash; 1397 to->sw_hash = from->sw_hash; 1398 to->l4_hash = from->l4_hash; 1399}; 1400 1401static inline void skb_copy_decrypted(struct sk_buff *to, 1402 const struct sk_buff *from) 1403{ 1404#ifdef CONFIG_TLS_DEVICE 1405 to->decrypted = from->decrypted; 1406#endif 1407} 1408 1409#ifdef NET_SKBUFF_DATA_USES_OFFSET 1410static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1411{ 1412 return skb->head + skb->end; 1413} 1414 1415static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1416{ 1417 return skb->end; 1418} 1419#else 1420static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1421{ 1422 return skb->end; 1423} 1424 1425static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1426{ 1427 return skb->end - skb->head; 1428} 1429#endif 1430 1431/* Internal */ 1432#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1433 1434static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1435{ 1436 return &skb_shinfo(skb)->hwtstamps; 1437} 1438 1439static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1440{ 1441 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY; 1442 1443 return is_zcopy ? skb_uarg(skb) : NULL; 1444} 1445 1446static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1447 bool *have_ref) 1448{ 1449 if (skb && uarg && !skb_zcopy(skb)) { 1450 if (unlikely(have_ref && *have_ref)) 1451 *have_ref = false; 1452 else 1453 sock_zerocopy_get(uarg); 1454 skb_shinfo(skb)->destructor_arg = uarg; 1455 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG; 1456 } 1457} 1458 1459static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1460{ 1461 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1462 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG; 1463} 1464 1465static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1466{ 1467 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1468} 1469 1470static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1471{ 1472 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1473} 1474 1475/* Release a reference on a zerocopy structure */ 1476static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy) 1477{ 1478 struct ubuf_info *uarg = skb_zcopy(skb); 1479 1480 if (uarg) { 1481 if (skb_zcopy_is_nouarg(skb)) { 1482 /* no notification callback */ 1483 } else if (uarg->callback == sock_zerocopy_callback) { 1484 uarg->zerocopy = uarg->zerocopy && zerocopy; 1485 sock_zerocopy_put(uarg); 1486 } else { 1487 uarg->callback(uarg, zerocopy); 1488 } 1489 1490 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG; 1491 } 1492} 1493 1494/* Abort a zerocopy operation and revert zckey on error in send syscall */ 1495static inline void skb_zcopy_abort(struct sk_buff *skb) 1496{ 1497 struct ubuf_info *uarg = skb_zcopy(skb); 1498 1499 if (uarg) { 1500 sock_zerocopy_put_abort(uarg, false); 1501 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG; 1502 } 1503} 1504 1505static inline void skb_mark_not_on_list(struct sk_buff *skb) 1506{ 1507 skb->next = NULL; 1508} 1509 1510/* Iterate through singly-linked GSO fragments of an skb. */ 1511#define skb_list_walk_safe(first, skb, next_skb) \ 1512 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1513 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1514 1515static inline void skb_list_del_init(struct sk_buff *skb) 1516{ 1517 __list_del_entry(&skb->list); 1518 skb_mark_not_on_list(skb); 1519} 1520 1521/** 1522 * skb_queue_empty - check if a queue is empty 1523 * @list: queue head 1524 * 1525 * Returns true if the queue is empty, false otherwise. 1526 */ 1527static inline int skb_queue_empty(const struct sk_buff_head *list) 1528{ 1529 return list->next == (const struct sk_buff *) list; 1530} 1531 1532/** 1533 * skb_queue_empty_lockless - check if a queue is empty 1534 * @list: queue head 1535 * 1536 * Returns true if the queue is empty, false otherwise. 1537 * This variant can be used in lockless contexts. 1538 */ 1539static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1540{ 1541 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1542} 1543 1544 1545/** 1546 * skb_queue_is_last - check if skb is the last entry in the queue 1547 * @list: queue head 1548 * @skb: buffer 1549 * 1550 * Returns true if @skb is the last buffer on the list. 1551 */ 1552static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1553 const struct sk_buff *skb) 1554{ 1555 return skb->next == (const struct sk_buff *) list; 1556} 1557 1558/** 1559 * skb_queue_is_first - check if skb is the first entry in the queue 1560 * @list: queue head 1561 * @skb: buffer 1562 * 1563 * Returns true if @skb is the first buffer on the list. 1564 */ 1565static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1566 const struct sk_buff *skb) 1567{ 1568 return skb->prev == (const struct sk_buff *) list; 1569} 1570 1571/** 1572 * skb_queue_next - return the next packet in the queue 1573 * @list: queue head 1574 * @skb: current buffer 1575 * 1576 * Return the next packet in @list after @skb. It is only valid to 1577 * call this if skb_queue_is_last() evaluates to false. 1578 */ 1579static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1580 const struct sk_buff *skb) 1581{ 1582 /* This BUG_ON may seem severe, but if we just return then we 1583 * are going to dereference garbage. 1584 */ 1585 BUG_ON(skb_queue_is_last(list, skb)); 1586 return skb->next; 1587} 1588 1589/** 1590 * skb_queue_prev - return the prev packet in the queue 1591 * @list: queue head 1592 * @skb: current buffer 1593 * 1594 * Return the prev packet in @list before @skb. It is only valid to 1595 * call this if skb_queue_is_first() evaluates to false. 1596 */ 1597static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1598 const struct sk_buff *skb) 1599{ 1600 /* This BUG_ON may seem severe, but if we just return then we 1601 * are going to dereference garbage. 1602 */ 1603 BUG_ON(skb_queue_is_first(list, skb)); 1604 return skb->prev; 1605} 1606 1607/** 1608 * skb_get - reference buffer 1609 * @skb: buffer to reference 1610 * 1611 * Makes another reference to a socket buffer and returns a pointer 1612 * to the buffer. 1613 */ 1614static inline struct sk_buff *skb_get(struct sk_buff *skb) 1615{ 1616 refcount_inc(&skb->users); 1617 return skb; 1618} 1619 1620/* 1621 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1622 */ 1623 1624/** 1625 * skb_cloned - is the buffer a clone 1626 * @skb: buffer to check 1627 * 1628 * Returns true if the buffer was generated with skb_clone() and is 1629 * one of multiple shared copies of the buffer. Cloned buffers are 1630 * shared data so must not be written to under normal circumstances. 1631 */ 1632static inline int skb_cloned(const struct sk_buff *skb) 1633{ 1634 return skb->cloned && 1635 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1636} 1637 1638static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1639{ 1640 might_sleep_if(gfpflags_allow_blocking(pri)); 1641 1642 if (skb_cloned(skb)) 1643 return pskb_expand_head(skb, 0, 0, pri); 1644 1645 return 0; 1646} 1647 1648/** 1649 * skb_header_cloned - is the header a clone 1650 * @skb: buffer to check 1651 * 1652 * Returns true if modifying the header part of the buffer requires 1653 * the data to be copied. 1654 */ 1655static inline int skb_header_cloned(const struct sk_buff *skb) 1656{ 1657 int dataref; 1658 1659 if (!skb->cloned) 1660 return 0; 1661 1662 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1663 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1664 return dataref != 1; 1665} 1666 1667static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1668{ 1669 might_sleep_if(gfpflags_allow_blocking(pri)); 1670 1671 if (skb_header_cloned(skb)) 1672 return pskb_expand_head(skb, 0, 0, pri); 1673 1674 return 0; 1675} 1676 1677/** 1678 * __skb_header_release - release reference to header 1679 * @skb: buffer to operate on 1680 */ 1681static inline void __skb_header_release(struct sk_buff *skb) 1682{ 1683 skb->nohdr = 1; 1684 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1685} 1686 1687 1688/** 1689 * skb_shared - is the buffer shared 1690 * @skb: buffer to check 1691 * 1692 * Returns true if more than one person has a reference to this 1693 * buffer. 1694 */ 1695static inline int skb_shared(const struct sk_buff *skb) 1696{ 1697 return refcount_read(&skb->users) != 1; 1698} 1699 1700/** 1701 * skb_share_check - check if buffer is shared and if so clone it 1702 * @skb: buffer to check 1703 * @pri: priority for memory allocation 1704 * 1705 * If the buffer is shared the buffer is cloned and the old copy 1706 * drops a reference. A new clone with a single reference is returned. 1707 * If the buffer is not shared the original buffer is returned. When 1708 * being called from interrupt status or with spinlocks held pri must 1709 * be GFP_ATOMIC. 1710 * 1711 * NULL is returned on a memory allocation failure. 1712 */ 1713static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1714{ 1715 might_sleep_if(gfpflags_allow_blocking(pri)); 1716 if (skb_shared(skb)) { 1717 struct sk_buff *nskb = skb_clone(skb, pri); 1718 1719 if (likely(nskb)) 1720 consume_skb(skb); 1721 else 1722 kfree_skb(skb); 1723 skb = nskb; 1724 } 1725 return skb; 1726} 1727 1728/* 1729 * Copy shared buffers into a new sk_buff. We effectively do COW on 1730 * packets to handle cases where we have a local reader and forward 1731 * and a couple of other messy ones. The normal one is tcpdumping 1732 * a packet thats being forwarded. 1733 */ 1734 1735/** 1736 * skb_unshare - make a copy of a shared buffer 1737 * @skb: buffer to check 1738 * @pri: priority for memory allocation 1739 * 1740 * If the socket buffer is a clone then this function creates a new 1741 * copy of the data, drops a reference count on the old copy and returns 1742 * the new copy with the reference count at 1. If the buffer is not a clone 1743 * the original buffer is returned. When called with a spinlock held or 1744 * from interrupt state @pri must be %GFP_ATOMIC 1745 * 1746 * %NULL is returned on a memory allocation failure. 1747 */ 1748static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1749 gfp_t pri) 1750{ 1751 might_sleep_if(gfpflags_allow_blocking(pri)); 1752 if (skb_cloned(skb)) { 1753 struct sk_buff *nskb = skb_copy(skb, pri); 1754 1755 /* Free our shared copy */ 1756 if (likely(nskb)) 1757 consume_skb(skb); 1758 else 1759 kfree_skb(skb); 1760 skb = nskb; 1761 } 1762 return skb; 1763} 1764 1765/** 1766 * skb_peek - peek at the head of an &sk_buff_head 1767 * @list_: list to peek at 1768 * 1769 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1770 * be careful with this one. A peek leaves the buffer on the 1771 * list and someone else may run off with it. You must hold 1772 * the appropriate locks or have a private queue to do this. 1773 * 1774 * Returns %NULL for an empty list or a pointer to the head element. 1775 * The reference count is not incremented and the reference is therefore 1776 * volatile. Use with caution. 1777 */ 1778static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1779{ 1780 struct sk_buff *skb = list_->next; 1781 1782 if (skb == (struct sk_buff *)list_) 1783 skb = NULL; 1784 return skb; 1785} 1786 1787/** 1788 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1789 * @list_: list to peek at 1790 * 1791 * Like skb_peek(), but the caller knows that the list is not empty. 1792 */ 1793static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1794{ 1795 return list_->next; 1796} 1797 1798/** 1799 * skb_peek_next - peek skb following the given one from a queue 1800 * @skb: skb to start from 1801 * @list_: list to peek at 1802 * 1803 * Returns %NULL when the end of the list is met or a pointer to the 1804 * next element. The reference count is not incremented and the 1805 * reference is therefore volatile. Use with caution. 1806 */ 1807static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1808 const struct sk_buff_head *list_) 1809{ 1810 struct sk_buff *next = skb->next; 1811 1812 if (next == (struct sk_buff *)list_) 1813 next = NULL; 1814 return next; 1815} 1816 1817/** 1818 * skb_peek_tail - peek at the tail of an &sk_buff_head 1819 * @list_: list to peek at 1820 * 1821 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1822 * be careful with this one. A peek leaves the buffer on the 1823 * list and someone else may run off with it. You must hold 1824 * the appropriate locks or have a private queue to do this. 1825 * 1826 * Returns %NULL for an empty list or a pointer to the tail element. 1827 * The reference count is not incremented and the reference is therefore 1828 * volatile. Use with caution. 1829 */ 1830static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1831{ 1832 struct sk_buff *skb = READ_ONCE(list_->prev); 1833 1834 if (skb == (struct sk_buff *)list_) 1835 skb = NULL; 1836 return skb; 1837 1838} 1839 1840/** 1841 * skb_queue_len - get queue length 1842 * @list_: list to measure 1843 * 1844 * Return the length of an &sk_buff queue. 1845 */ 1846static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1847{ 1848 return list_->qlen; 1849} 1850 1851/** 1852 * skb_queue_len_lockless - get queue length 1853 * @list_: list to measure 1854 * 1855 * Return the length of an &sk_buff queue. 1856 * This variant can be used in lockless contexts. 1857 */ 1858static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1859{ 1860 return READ_ONCE(list_->qlen); 1861} 1862 1863/** 1864 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1865 * @list: queue to initialize 1866 * 1867 * This initializes only the list and queue length aspects of 1868 * an sk_buff_head object. This allows to initialize the list 1869 * aspects of an sk_buff_head without reinitializing things like 1870 * the spinlock. It can also be used for on-stack sk_buff_head 1871 * objects where the spinlock is known to not be used. 1872 */ 1873static inline void __skb_queue_head_init(struct sk_buff_head *list) 1874{ 1875 list->prev = list->next = (struct sk_buff *)list; 1876 list->qlen = 0; 1877} 1878 1879/* 1880 * This function creates a split out lock class for each invocation; 1881 * this is needed for now since a whole lot of users of the skb-queue 1882 * infrastructure in drivers have different locking usage (in hardirq) 1883 * than the networking core (in softirq only). In the long run either the 1884 * network layer or drivers should need annotation to consolidate the 1885 * main types of usage into 3 classes. 1886 */ 1887static inline void skb_queue_head_init(struct sk_buff_head *list) 1888{ 1889 spin_lock_init(&list->lock); 1890 __skb_queue_head_init(list); 1891} 1892 1893static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1894 struct lock_class_key *class) 1895{ 1896 skb_queue_head_init(list); 1897 lockdep_set_class(&list->lock, class); 1898} 1899 1900/* 1901 * Insert an sk_buff on a list. 1902 * 1903 * The "__skb_xxxx()" functions are the non-atomic ones that 1904 * can only be called with interrupts disabled. 1905 */ 1906static inline void __skb_insert(struct sk_buff *newsk, 1907 struct sk_buff *prev, struct sk_buff *next, 1908 struct sk_buff_head *list) 1909{ 1910 /* See skb_queue_empty_lockless() and skb_peek_tail() 1911 * for the opposite READ_ONCE() 1912 */ 1913 WRITE_ONCE(newsk->next, next); 1914 WRITE_ONCE(newsk->prev, prev); 1915 WRITE_ONCE(next->prev, newsk); 1916 WRITE_ONCE(prev->next, newsk); 1917 list->qlen++; 1918} 1919 1920static inline void __skb_queue_splice(const struct sk_buff_head *list, 1921 struct sk_buff *prev, 1922 struct sk_buff *next) 1923{ 1924 struct sk_buff *first = list->next; 1925 struct sk_buff *last = list->prev; 1926 1927 WRITE_ONCE(first->prev, prev); 1928 WRITE_ONCE(prev->next, first); 1929 1930 WRITE_ONCE(last->next, next); 1931 WRITE_ONCE(next->prev, last); 1932} 1933 1934/** 1935 * skb_queue_splice - join two skb lists, this is designed for stacks 1936 * @list: the new list to add 1937 * @head: the place to add it in the first list 1938 */ 1939static inline void skb_queue_splice(const struct sk_buff_head *list, 1940 struct sk_buff_head *head) 1941{ 1942 if (!skb_queue_empty(list)) { 1943 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1944 head->qlen += list->qlen; 1945 } 1946} 1947 1948/** 1949 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1950 * @list: the new list to add 1951 * @head: the place to add it in the first list 1952 * 1953 * The list at @list is reinitialised 1954 */ 1955static inline void skb_queue_splice_init(struct sk_buff_head *list, 1956 struct sk_buff_head *head) 1957{ 1958 if (!skb_queue_empty(list)) { 1959 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1960 head->qlen += list->qlen; 1961 __skb_queue_head_init(list); 1962 } 1963} 1964 1965/** 1966 * skb_queue_splice_tail - join two skb lists, each list being a queue 1967 * @list: the new list to add 1968 * @head: the place to add it in the first list 1969 */ 1970static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1971 struct sk_buff_head *head) 1972{ 1973 if (!skb_queue_empty(list)) { 1974 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1975 head->qlen += list->qlen; 1976 } 1977} 1978 1979/** 1980 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1981 * @list: the new list to add 1982 * @head: the place to add it in the first list 1983 * 1984 * Each of the lists is a queue. 1985 * The list at @list is reinitialised 1986 */ 1987static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 1988 struct sk_buff_head *head) 1989{ 1990 if (!skb_queue_empty(list)) { 1991 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1992 head->qlen += list->qlen; 1993 __skb_queue_head_init(list); 1994 } 1995} 1996 1997/** 1998 * __skb_queue_after - queue a buffer at the list head 1999 * @list: list to use 2000 * @prev: place after this buffer 2001 * @newsk: buffer to queue 2002 * 2003 * Queue a buffer int the middle of a list. This function takes no locks 2004 * and you must therefore hold required locks before calling it. 2005 * 2006 * A buffer cannot be placed on two lists at the same time. 2007 */ 2008static inline void __skb_queue_after(struct sk_buff_head *list, 2009 struct sk_buff *prev, 2010 struct sk_buff *newsk) 2011{ 2012 __skb_insert(newsk, prev, prev->next, list); 2013} 2014 2015void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2016 struct sk_buff_head *list); 2017 2018static inline void __skb_queue_before(struct sk_buff_head *list, 2019 struct sk_buff *next, 2020 struct sk_buff *newsk) 2021{ 2022 __skb_insert(newsk, next->prev, next, list); 2023} 2024 2025/** 2026 * __skb_queue_head - queue a buffer at the list head 2027 * @list: list to use 2028 * @newsk: buffer to queue 2029 * 2030 * Queue a buffer at the start of a list. This function takes no locks 2031 * and you must therefore hold required locks before calling it. 2032 * 2033 * A buffer cannot be placed on two lists at the same time. 2034 */ 2035static inline void __skb_queue_head(struct sk_buff_head *list, 2036 struct sk_buff *newsk) 2037{ 2038 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2039} 2040void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2041 2042/** 2043 * __skb_queue_tail - queue a buffer at the list tail 2044 * @list: list to use 2045 * @newsk: buffer to queue 2046 * 2047 * Queue a buffer at the end of a list. This function takes no locks 2048 * and you must therefore hold required locks before calling it. 2049 * 2050 * A buffer cannot be placed on two lists at the same time. 2051 */ 2052static inline void __skb_queue_tail(struct sk_buff_head *list, 2053 struct sk_buff *newsk) 2054{ 2055 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2056} 2057void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2058 2059/* 2060 * remove sk_buff from list. _Must_ be called atomically, and with 2061 * the list known.. 2062 */ 2063void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2064static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2065{ 2066 struct sk_buff *next, *prev; 2067 2068 WRITE_ONCE(list->qlen, list->qlen - 1); 2069 next = skb->next; 2070 prev = skb->prev; 2071 skb->next = skb->prev = NULL; 2072 WRITE_ONCE(next->prev, prev); 2073 WRITE_ONCE(prev->next, next); 2074} 2075 2076/** 2077 * __skb_dequeue - remove from the head of the queue 2078 * @list: list to dequeue from 2079 * 2080 * Remove the head of the list. This function does not take any locks 2081 * so must be used with appropriate locks held only. The head item is 2082 * returned or %NULL if the list is empty. 2083 */ 2084static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2085{ 2086 struct sk_buff *skb = skb_peek(list); 2087 if (skb) 2088 __skb_unlink(skb, list); 2089 return skb; 2090} 2091struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2092 2093/** 2094 * __skb_dequeue_tail - remove from the tail of the queue 2095 * @list: list to dequeue from 2096 * 2097 * Remove the tail of the list. This function does not take any locks 2098 * so must be used with appropriate locks held only. The tail item is 2099 * returned or %NULL if the list is empty. 2100 */ 2101static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2102{ 2103 struct sk_buff *skb = skb_peek_tail(list); 2104 if (skb) 2105 __skb_unlink(skb, list); 2106 return skb; 2107} 2108struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2109 2110 2111static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2112{ 2113 return skb->data_len; 2114} 2115 2116static inline unsigned int skb_headlen(const struct sk_buff *skb) 2117{ 2118 return skb->len - skb->data_len; 2119} 2120 2121static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2122{ 2123 unsigned int i, len = 0; 2124 2125 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2126 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2127 return len; 2128} 2129 2130static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2131{ 2132 return skb_headlen(skb) + __skb_pagelen(skb); 2133} 2134 2135/** 2136 * __skb_fill_page_desc - initialise a paged fragment in an skb 2137 * @skb: buffer containing fragment to be initialised 2138 * @i: paged fragment index to initialise 2139 * @page: the page to use for this fragment 2140 * @off: the offset to the data with @page 2141 * @size: the length of the data 2142 * 2143 * Initialises the @i'th fragment of @skb to point to &size bytes at 2144 * offset @off within @page. 2145 * 2146 * Does not take any additional reference on the fragment. 2147 */ 2148static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2149 struct page *page, int off, int size) 2150{ 2151 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2152 2153 /* 2154 * Propagate page pfmemalloc to the skb if we can. The problem is 2155 * that not all callers have unique ownership of the page but rely 2156 * on page_is_pfmemalloc doing the right thing(tm). 2157 */ 2158 frag->bv_page = page; 2159 frag->bv_offset = off; 2160 skb_frag_size_set(frag, size); 2161 2162 page = compound_head(page); 2163 if (page_is_pfmemalloc(page)) 2164 skb->pfmemalloc = true; 2165} 2166 2167/** 2168 * skb_fill_page_desc - initialise a paged fragment in an skb 2169 * @skb: buffer containing fragment to be initialised 2170 * @i: paged fragment index to initialise 2171 * @page: the page to use for this fragment 2172 * @off: the offset to the data with @page 2173 * @size: the length of the data 2174 * 2175 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2176 * @skb to point to @size bytes at offset @off within @page. In 2177 * addition updates @skb such that @i is the last fragment. 2178 * 2179 * Does not take any additional reference on the fragment. 2180 */ 2181static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2182 struct page *page, int off, int size) 2183{ 2184 __skb_fill_page_desc(skb, i, page, off, size); 2185 skb_shinfo(skb)->nr_frags = i + 1; 2186} 2187 2188void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2189 int size, unsigned int truesize); 2190 2191void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2192 unsigned int truesize); 2193 2194#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2195 2196#ifdef NET_SKBUFF_DATA_USES_OFFSET 2197static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2198{ 2199 return skb->head + skb->tail; 2200} 2201 2202static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2203{ 2204 skb->tail = skb->data - skb->head; 2205} 2206 2207static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2208{ 2209 skb_reset_tail_pointer(skb); 2210 skb->tail += offset; 2211} 2212 2213#else /* NET_SKBUFF_DATA_USES_OFFSET */ 2214static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2215{ 2216 return skb->tail; 2217} 2218 2219static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2220{ 2221 skb->tail = skb->data; 2222} 2223 2224static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2225{ 2226 skb->tail = skb->data + offset; 2227} 2228 2229#endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2230 2231/* 2232 * Add data to an sk_buff 2233 */ 2234void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2235void *skb_put(struct sk_buff *skb, unsigned int len); 2236static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2237{ 2238 void *tmp = skb_tail_pointer(skb); 2239 SKB_LINEAR_ASSERT(skb); 2240 skb->tail += len; 2241 skb->len += len; 2242 return tmp; 2243} 2244 2245static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2246{ 2247 void *tmp = __skb_put(skb, len); 2248 2249 memset(tmp, 0, len); 2250 return tmp; 2251} 2252 2253static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2254 unsigned int len) 2255{ 2256 void *tmp = __skb_put(skb, len); 2257 2258 memcpy(tmp, data, len); 2259 return tmp; 2260} 2261 2262static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2263{ 2264 *(u8 *)__skb_put(skb, 1) = val; 2265} 2266 2267static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2268{ 2269 void *tmp = skb_put(skb, len); 2270 2271 memset(tmp, 0, len); 2272 2273 return tmp; 2274} 2275 2276static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2277 unsigned int len) 2278{ 2279 void *tmp = skb_put(skb, len); 2280 2281 memcpy(tmp, data, len); 2282 2283 return tmp; 2284} 2285 2286static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2287{ 2288 *(u8 *)skb_put(skb, 1) = val; 2289} 2290 2291void *skb_push(struct sk_buff *skb, unsigned int len); 2292static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2293{ 2294 skb->data -= len; 2295 skb->len += len; 2296 return skb->data; 2297} 2298 2299void *skb_pull(struct sk_buff *skb, unsigned int len); 2300static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2301{ 2302 skb->len -= len; 2303 BUG_ON(skb->len < skb->data_len); 2304 return skb->data += len; 2305} 2306 2307static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2308{ 2309 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2310} 2311 2312void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2313 2314static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2315{ 2316 if (len > skb_headlen(skb) && 2317 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2318 return NULL; 2319 skb->len -= len; 2320 return skb->data += len; 2321} 2322 2323static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2324{ 2325 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2326} 2327 2328static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2329{ 2330 if (likely(len <= skb_headlen(skb))) 2331 return true; 2332 if (unlikely(len > skb->len)) 2333 return false; 2334 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2335} 2336 2337void skb_condense(struct sk_buff *skb); 2338 2339/** 2340 * skb_headroom - bytes at buffer head 2341 * @skb: buffer to check 2342 * 2343 * Return the number of bytes of free space at the head of an &sk_buff. 2344 */ 2345static inline unsigned int skb_headroom(const struct sk_buff *skb) 2346{ 2347 return skb->data - skb->head; 2348} 2349 2350/** 2351 * skb_tailroom - bytes at buffer end 2352 * @skb: buffer to check 2353 * 2354 * Return the number of bytes of free space at the tail of an sk_buff 2355 */ 2356static inline int skb_tailroom(const struct sk_buff *skb) 2357{ 2358 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2359} 2360 2361/** 2362 * skb_availroom - bytes at buffer end 2363 * @skb: buffer to check 2364 * 2365 * Return the number of bytes of free space at the tail of an sk_buff 2366 * allocated by sk_stream_alloc() 2367 */ 2368static inline int skb_availroom(const struct sk_buff *skb) 2369{ 2370 if (skb_is_nonlinear(skb)) 2371 return 0; 2372 2373 return skb->end - skb->tail - skb->reserved_tailroom; 2374} 2375 2376/** 2377 * skb_reserve - adjust headroom 2378 * @skb: buffer to alter 2379 * @len: bytes to move 2380 * 2381 * Increase the headroom of an empty &sk_buff by reducing the tail 2382 * room. This is only allowed for an empty buffer. 2383 */ 2384static inline void skb_reserve(struct sk_buff *skb, int len) 2385{ 2386 skb->data += len; 2387 skb->tail += len; 2388} 2389 2390/** 2391 * skb_tailroom_reserve - adjust reserved_tailroom 2392 * @skb: buffer to alter 2393 * @mtu: maximum amount of headlen permitted 2394 * @needed_tailroom: minimum amount of reserved_tailroom 2395 * 2396 * Set reserved_tailroom so that headlen can be as large as possible but 2397 * not larger than mtu and tailroom cannot be smaller than 2398 * needed_tailroom. 2399 * The required headroom should already have been reserved before using 2400 * this function. 2401 */ 2402static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2403 unsigned int needed_tailroom) 2404{ 2405 SKB_LINEAR_ASSERT(skb); 2406 if (mtu < skb_tailroom(skb) - needed_tailroom) 2407 /* use at most mtu */ 2408 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2409 else 2410 /* use up to all available space */ 2411 skb->reserved_tailroom = needed_tailroom; 2412} 2413 2414#define ENCAP_TYPE_ETHER 0 2415#define ENCAP_TYPE_IPPROTO 1 2416 2417static inline void skb_set_inner_protocol(struct sk_buff *skb, 2418 __be16 protocol) 2419{ 2420 skb->inner_protocol = protocol; 2421 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2422} 2423 2424static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2425 __u8 ipproto) 2426{ 2427 skb->inner_ipproto = ipproto; 2428 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2429} 2430 2431static inline void skb_reset_inner_headers(struct sk_buff *skb) 2432{ 2433 skb->inner_mac_header = skb->mac_header; 2434 skb->inner_network_header = skb->network_header; 2435 skb->inner_transport_header = skb->transport_header; 2436} 2437 2438static inline void skb_reset_mac_len(struct sk_buff *skb) 2439{ 2440 skb->mac_len = skb->network_header - skb->mac_header; 2441} 2442 2443static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2444 *skb) 2445{ 2446 return skb->head + skb->inner_transport_header; 2447} 2448 2449static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2450{ 2451 return skb_inner_transport_header(skb) - skb->data; 2452} 2453 2454static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2455{ 2456 skb->inner_transport_header = skb->data - skb->head; 2457} 2458 2459static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2460 const int offset) 2461{ 2462 skb_reset_inner_transport_header(skb); 2463 skb->inner_transport_header += offset; 2464} 2465 2466static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2467{ 2468 return skb->head + skb->inner_network_header; 2469} 2470 2471static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2472{ 2473 skb->inner_network_header = skb->data - skb->head; 2474} 2475 2476static inline void skb_set_inner_network_header(struct sk_buff *skb, 2477 const int offset) 2478{ 2479 skb_reset_inner_network_header(skb); 2480 skb->inner_network_header += offset; 2481} 2482 2483static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2484{ 2485 return skb->head + skb->inner_mac_header; 2486} 2487 2488static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2489{ 2490 skb->inner_mac_header = skb->data - skb->head; 2491} 2492 2493static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2494 const int offset) 2495{ 2496 skb_reset_inner_mac_header(skb); 2497 skb->inner_mac_header += offset; 2498} 2499static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2500{ 2501 return skb->transport_header != (typeof(skb->transport_header))~0U; 2502} 2503 2504static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2505{ 2506 return skb->head + skb->transport_header; 2507} 2508 2509static inline void skb_reset_transport_header(struct sk_buff *skb) 2510{ 2511 skb->transport_header = skb->data - skb->head; 2512} 2513 2514static inline void skb_set_transport_header(struct sk_buff *skb, 2515 const int offset) 2516{ 2517 skb_reset_transport_header(skb); 2518 skb->transport_header += offset; 2519} 2520 2521static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2522{ 2523 return skb->head + skb->network_header; 2524} 2525 2526static inline void skb_reset_network_header(struct sk_buff *skb) 2527{ 2528 skb->network_header = skb->data - skb->head; 2529} 2530 2531static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2532{ 2533 skb_reset_network_header(skb); 2534 skb->network_header += offset; 2535} 2536 2537static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2538{ 2539 return skb->head + skb->mac_header; 2540} 2541 2542static inline int skb_mac_offset(const struct sk_buff *skb) 2543{ 2544 return skb_mac_header(skb) - skb->data; 2545} 2546 2547static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2548{ 2549 return skb->network_header - skb->mac_header; 2550} 2551 2552static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2553{ 2554 return skb->mac_header != (typeof(skb->mac_header))~0U; 2555} 2556 2557static inline void skb_unset_mac_header(struct sk_buff *skb) 2558{ 2559 skb->mac_header = (typeof(skb->mac_header))~0U; 2560} 2561 2562static inline void skb_reset_mac_header(struct sk_buff *skb) 2563{ 2564 skb->mac_header = skb->data - skb->head; 2565} 2566 2567static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2568{ 2569 skb_reset_mac_header(skb); 2570 skb->mac_header += offset; 2571} 2572 2573static inline void skb_pop_mac_header(struct sk_buff *skb) 2574{ 2575 skb->mac_header = skb->network_header; 2576} 2577 2578static inline void skb_probe_transport_header(struct sk_buff *skb) 2579{ 2580 struct flow_keys_basic keys; 2581 2582 if (skb_transport_header_was_set(skb)) 2583 return; 2584 2585 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2586 NULL, 0, 0, 0, 0)) 2587 skb_set_transport_header(skb, keys.control.thoff); 2588} 2589 2590static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2591{ 2592 if (skb_mac_header_was_set(skb)) { 2593 const unsigned char *old_mac = skb_mac_header(skb); 2594 2595 skb_set_mac_header(skb, -skb->mac_len); 2596 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2597 } 2598} 2599 2600static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2601{ 2602 return skb->csum_start - skb_headroom(skb); 2603} 2604 2605static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2606{ 2607 return skb->head + skb->csum_start; 2608} 2609 2610static inline int skb_transport_offset(const struct sk_buff *skb) 2611{ 2612 return skb_transport_header(skb) - skb->data; 2613} 2614 2615static inline u32 skb_network_header_len(const struct sk_buff *skb) 2616{ 2617 return skb->transport_header - skb->network_header; 2618} 2619 2620static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2621{ 2622 return skb->inner_transport_header - skb->inner_network_header; 2623} 2624 2625static inline int skb_network_offset(const struct sk_buff *skb) 2626{ 2627 return skb_network_header(skb) - skb->data; 2628} 2629 2630static inline int skb_inner_network_offset(const struct sk_buff *skb) 2631{ 2632 return skb_inner_network_header(skb) - skb->data; 2633} 2634 2635static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2636{ 2637 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2638} 2639 2640/* 2641 * CPUs often take a performance hit when accessing unaligned memory 2642 * locations. The actual performance hit varies, it can be small if the 2643 * hardware handles it or large if we have to take an exception and fix it 2644 * in software. 2645 * 2646 * Since an ethernet header is 14 bytes network drivers often end up with 2647 * the IP header at an unaligned offset. The IP header can be aligned by 2648 * shifting the start of the packet by 2 bytes. Drivers should do this 2649 * with: 2650 * 2651 * skb_reserve(skb, NET_IP_ALIGN); 2652 * 2653 * The downside to this alignment of the IP header is that the DMA is now 2654 * unaligned. On some architectures the cost of an unaligned DMA is high 2655 * and this cost outweighs the gains made by aligning the IP header. 2656 * 2657 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2658 * to be overridden. 2659 */ 2660#ifndef NET_IP_ALIGN 2661#define NET_IP_ALIGN 2 2662#endif 2663 2664/* 2665 * The networking layer reserves some headroom in skb data (via 2666 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2667 * the header has to grow. In the default case, if the header has to grow 2668 * 32 bytes or less we avoid the reallocation. 2669 * 2670 * Unfortunately this headroom changes the DMA alignment of the resulting 2671 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2672 * on some architectures. An architecture can override this value, 2673 * perhaps setting it to a cacheline in size (since that will maintain 2674 * cacheline alignment of the DMA). It must be a power of 2. 2675 * 2676 * Various parts of the networking layer expect at least 32 bytes of 2677 * headroom, you should not reduce this. 2678 * 2679 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2680 * to reduce average number of cache lines per packet. 2681 * get_rps_cpu() for example only access one 64 bytes aligned block : 2682 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2683 */ 2684#ifndef NET_SKB_PAD 2685#define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2686#endif 2687 2688int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2689 2690static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2691{ 2692 if (WARN_ON(skb_is_nonlinear(skb))) 2693 return; 2694 skb->len = len; 2695 skb_set_tail_pointer(skb, len); 2696} 2697 2698static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2699{ 2700 __skb_set_length(skb, len); 2701} 2702 2703void skb_trim(struct sk_buff *skb, unsigned int len); 2704 2705static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2706{ 2707 if (skb->data_len) 2708 return ___pskb_trim(skb, len); 2709 __skb_trim(skb, len); 2710 return 0; 2711} 2712 2713static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2714{ 2715 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2716} 2717 2718/** 2719 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2720 * @skb: buffer to alter 2721 * @len: new length 2722 * 2723 * This is identical to pskb_trim except that the caller knows that 2724 * the skb is not cloned so we should never get an error due to out- 2725 * of-memory. 2726 */ 2727static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2728{ 2729 int err = pskb_trim(skb, len); 2730 BUG_ON(err); 2731} 2732 2733static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2734{ 2735 unsigned int diff = len - skb->len; 2736 2737 if (skb_tailroom(skb) < diff) { 2738 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2739 GFP_ATOMIC); 2740 if (ret) 2741 return ret; 2742 } 2743 __skb_set_length(skb, len); 2744 return 0; 2745} 2746 2747/** 2748 * skb_orphan - orphan a buffer 2749 * @skb: buffer to orphan 2750 * 2751 * If a buffer currently has an owner then we call the owner's 2752 * destructor function and make the @skb unowned. The buffer continues 2753 * to exist but is no longer charged to its former owner. 2754 */ 2755static inline void skb_orphan(struct sk_buff *skb) 2756{ 2757 if (skb->destructor) { 2758 skb->destructor(skb); 2759 skb->destructor = NULL; 2760 skb->sk = NULL; 2761 } else { 2762 BUG_ON(skb->sk); 2763 } 2764} 2765 2766/** 2767 * skb_orphan_frags - orphan the frags contained in a buffer 2768 * @skb: buffer to orphan frags from 2769 * @gfp_mask: allocation mask for replacement pages 2770 * 2771 * For each frag in the SKB which needs a destructor (i.e. has an 2772 * owner) create a copy of that frag and release the original 2773 * page by calling the destructor. 2774 */ 2775static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2776{ 2777 if (likely(!skb_zcopy(skb))) 2778 return 0; 2779 if (!skb_zcopy_is_nouarg(skb) && 2780 skb_uarg(skb)->callback == sock_zerocopy_callback) 2781 return 0; 2782 return skb_copy_ubufs(skb, gfp_mask); 2783} 2784 2785/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2786static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2787{ 2788 if (likely(!skb_zcopy(skb))) 2789 return 0; 2790 return skb_copy_ubufs(skb, gfp_mask); 2791} 2792 2793/** 2794 * __skb_queue_purge - empty a list 2795 * @list: list to empty 2796 * 2797 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2798 * the list and one reference dropped. This function does not take the 2799 * list lock and the caller must hold the relevant locks to use it. 2800 */ 2801static inline void __skb_queue_purge(struct sk_buff_head *list) 2802{ 2803 struct sk_buff *skb; 2804 while ((skb = __skb_dequeue(list)) != NULL) 2805 kfree_skb(skb); 2806} 2807void skb_queue_purge(struct sk_buff_head *list); 2808 2809unsigned int skb_rbtree_purge(struct rb_root *root); 2810 2811void *netdev_alloc_frag(unsigned int fragsz); 2812 2813struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2814 gfp_t gfp_mask); 2815 2816/** 2817 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2818 * @dev: network device to receive on 2819 * @length: length to allocate 2820 * 2821 * Allocate a new &sk_buff and assign it a usage count of one. The 2822 * buffer has unspecified headroom built in. Users should allocate 2823 * the headroom they think they need without accounting for the 2824 * built in space. The built in space is used for optimisations. 2825 * 2826 * %NULL is returned if there is no free memory. Although this function 2827 * allocates memory it can be called from an interrupt. 2828 */ 2829static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2830 unsigned int length) 2831{ 2832 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2833} 2834 2835/* legacy helper around __netdev_alloc_skb() */ 2836static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2837 gfp_t gfp_mask) 2838{ 2839 return __netdev_alloc_skb(NULL, length, gfp_mask); 2840} 2841 2842/* legacy helper around netdev_alloc_skb() */ 2843static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2844{ 2845 return netdev_alloc_skb(NULL, length); 2846} 2847 2848 2849static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2850 unsigned int length, gfp_t gfp) 2851{ 2852 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2853 2854 if (NET_IP_ALIGN && skb) 2855 skb_reserve(skb, NET_IP_ALIGN); 2856 return skb; 2857} 2858 2859static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2860 unsigned int length) 2861{ 2862 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2863} 2864 2865static inline void skb_free_frag(void *addr) 2866{ 2867 page_frag_free(addr); 2868} 2869 2870void *napi_alloc_frag(unsigned int fragsz); 2871struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2872 unsigned int length, gfp_t gfp_mask); 2873static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2874 unsigned int length) 2875{ 2876 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2877} 2878void napi_consume_skb(struct sk_buff *skb, int budget); 2879 2880void __kfree_skb_flush(void); 2881void __kfree_skb_defer(struct sk_buff *skb); 2882 2883/** 2884 * __dev_alloc_pages - allocate page for network Rx 2885 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2886 * @order: size of the allocation 2887 * 2888 * Allocate a new page. 2889 * 2890 * %NULL is returned if there is no free memory. 2891*/ 2892static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2893 unsigned int order) 2894{ 2895 /* This piece of code contains several assumptions. 2896 * 1. This is for device Rx, therefor a cold page is preferred. 2897 * 2. The expectation is the user wants a compound page. 2898 * 3. If requesting a order 0 page it will not be compound 2899 * due to the check to see if order has a value in prep_new_page 2900 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2901 * code in gfp_to_alloc_flags that should be enforcing this. 2902 */ 2903 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2904 2905 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2906} 2907 2908static inline struct page *dev_alloc_pages(unsigned int order) 2909{ 2910 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2911} 2912 2913/** 2914 * __dev_alloc_page - allocate a page for network Rx 2915 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2916 * 2917 * Allocate a new page. 2918 * 2919 * %NULL is returned if there is no free memory. 2920 */ 2921static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2922{ 2923 return __dev_alloc_pages(gfp_mask, 0); 2924} 2925 2926static inline struct page *dev_alloc_page(void) 2927{ 2928 return dev_alloc_pages(0); 2929} 2930 2931/** 2932 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2933 * @page: The page that was allocated from skb_alloc_page 2934 * @skb: The skb that may need pfmemalloc set 2935 */ 2936static inline void skb_propagate_pfmemalloc(struct page *page, 2937 struct sk_buff *skb) 2938{ 2939 if (page_is_pfmemalloc(page)) 2940 skb->pfmemalloc = true; 2941} 2942 2943/** 2944 * skb_frag_off() - Returns the offset of a skb fragment 2945 * @frag: the paged fragment 2946 */ 2947static inline unsigned int skb_frag_off(const skb_frag_t *frag) 2948{ 2949 return frag->bv_offset; 2950} 2951 2952/** 2953 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 2954 * @frag: skb fragment 2955 * @delta: value to add 2956 */ 2957static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 2958{ 2959 frag->bv_offset += delta; 2960} 2961 2962/** 2963 * skb_frag_off_set() - Sets the offset of a skb fragment 2964 * @frag: skb fragment 2965 * @offset: offset of fragment 2966 */ 2967static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 2968{ 2969 frag->bv_offset = offset; 2970} 2971 2972/** 2973 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 2974 * @fragto: skb fragment where offset is set 2975 * @fragfrom: skb fragment offset is copied from 2976 */ 2977static inline void skb_frag_off_copy(skb_frag_t *fragto, 2978 const skb_frag_t *fragfrom) 2979{ 2980 fragto->bv_offset = fragfrom->bv_offset; 2981} 2982 2983/** 2984 * skb_frag_page - retrieve the page referred to by a paged fragment 2985 * @frag: the paged fragment 2986 * 2987 * Returns the &struct page associated with @frag. 2988 */ 2989static inline struct page *skb_frag_page(const skb_frag_t *frag) 2990{ 2991 return frag->bv_page; 2992} 2993 2994/** 2995 * __skb_frag_ref - take an addition reference on a paged fragment. 2996 * @frag: the paged fragment 2997 * 2998 * Takes an additional reference on the paged fragment @frag. 2999 */ 3000static inline void __skb_frag_ref(skb_frag_t *frag) 3001{ 3002 get_page(skb_frag_page(frag)); 3003} 3004 3005/** 3006 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3007 * @skb: the buffer 3008 * @f: the fragment offset. 3009 * 3010 * Takes an additional reference on the @f'th paged fragment of @skb. 3011 */ 3012static inline void skb_frag_ref(struct sk_buff *skb, int f) 3013{ 3014 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3015} 3016 3017/** 3018 * __skb_frag_unref - release a reference on a paged fragment. 3019 * @frag: the paged fragment 3020 * 3021 * Releases a reference on the paged fragment @frag. 3022 */ 3023static inline void __skb_frag_unref(skb_frag_t *frag) 3024{ 3025 put_page(skb_frag_page(frag)); 3026} 3027 3028/** 3029 * skb_frag_unref - release a reference on a paged fragment of an skb. 3030 * @skb: the buffer 3031 * @f: the fragment offset 3032 * 3033 * Releases a reference on the @f'th paged fragment of @skb. 3034 */ 3035static inline void skb_frag_unref(struct sk_buff *skb, int f) 3036{ 3037 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 3038} 3039 3040/** 3041 * skb_frag_address - gets the address of the data contained in a paged fragment 3042 * @frag: the paged fragment buffer 3043 * 3044 * Returns the address of the data within @frag. The page must already 3045 * be mapped. 3046 */ 3047static inline void *skb_frag_address(const skb_frag_t *frag) 3048{ 3049 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3050} 3051 3052/** 3053 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3054 * @frag: the paged fragment buffer 3055 * 3056 * Returns the address of the data within @frag. Checks that the page 3057 * is mapped and returns %NULL otherwise. 3058 */ 3059static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3060{ 3061 void *ptr = page_address(skb_frag_page(frag)); 3062 if (unlikely(!ptr)) 3063 return NULL; 3064 3065 return ptr + skb_frag_off(frag); 3066} 3067 3068/** 3069 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3070 * @fragto: skb fragment where page is set 3071 * @fragfrom: skb fragment page is copied from 3072 */ 3073static inline void skb_frag_page_copy(skb_frag_t *fragto, 3074 const skb_frag_t *fragfrom) 3075{ 3076 fragto->bv_page = fragfrom->bv_page; 3077} 3078 3079/** 3080 * __skb_frag_set_page - sets the page contained in a paged fragment 3081 * @frag: the paged fragment 3082 * @page: the page to set 3083 * 3084 * Sets the fragment @frag to contain @page. 3085 */ 3086static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3087{ 3088 frag->bv_page = page; 3089} 3090 3091/** 3092 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3093 * @skb: the buffer 3094 * @f: the fragment offset 3095 * @page: the page to set 3096 * 3097 * Sets the @f'th fragment of @skb to contain @page. 3098 */ 3099static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3100 struct page *page) 3101{ 3102 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3103} 3104 3105bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3106 3107/** 3108 * skb_frag_dma_map - maps a paged fragment via the DMA API 3109 * @dev: the device to map the fragment to 3110 * @frag: the paged fragment to map 3111 * @offset: the offset within the fragment (starting at the 3112 * fragment's own offset) 3113 * @size: the number of bytes to map 3114 * @dir: the direction of the mapping (``PCI_DMA_*``) 3115 * 3116 * Maps the page associated with @frag to @device. 3117 */ 3118static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3119 const skb_frag_t *frag, 3120 size_t offset, size_t size, 3121 enum dma_data_direction dir) 3122{ 3123 return dma_map_page(dev, skb_frag_page(frag), 3124 skb_frag_off(frag) + offset, size, dir); 3125} 3126 3127static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3128 gfp_t gfp_mask) 3129{ 3130 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3131} 3132 3133 3134static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3135 gfp_t gfp_mask) 3136{ 3137 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3138} 3139 3140 3141/** 3142 * skb_clone_writable - is the header of a clone writable 3143 * @skb: buffer to check 3144 * @len: length up to which to write 3145 * 3146 * Returns true if modifying the header part of the cloned buffer 3147 * does not requires the data to be copied. 3148 */ 3149static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3150{ 3151 return !skb_header_cloned(skb) && 3152 skb_headroom(skb) + len <= skb->hdr_len; 3153} 3154 3155static inline int skb_try_make_writable(struct sk_buff *skb, 3156 unsigned int write_len) 3157{ 3158 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3159 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3160} 3161 3162static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3163 int cloned) 3164{ 3165 int delta = 0; 3166 3167 if (headroom > skb_headroom(skb)) 3168 delta = headroom - skb_headroom(skb); 3169 3170 if (delta || cloned) 3171 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3172 GFP_ATOMIC); 3173 return 0; 3174} 3175 3176/** 3177 * skb_cow - copy header of skb when it is required 3178 * @skb: buffer to cow 3179 * @headroom: needed headroom 3180 * 3181 * If the skb passed lacks sufficient headroom or its data part 3182 * is shared, data is reallocated. If reallocation fails, an error 3183 * is returned and original skb is not changed. 3184 * 3185 * The result is skb with writable area skb->head...skb->tail 3186 * and at least @headroom of space at head. 3187 */ 3188static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3189{ 3190 return __skb_cow(skb, headroom, skb_cloned(skb)); 3191} 3192 3193/** 3194 * skb_cow_head - skb_cow but only making the head writable 3195 * @skb: buffer to cow 3196 * @headroom: needed headroom 3197 * 3198 * This function is identical to skb_cow except that we replace the 3199 * skb_cloned check by skb_header_cloned. It should be used when 3200 * you only need to push on some header and do not need to modify 3201 * the data. 3202 */ 3203static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3204{ 3205 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3206} 3207 3208/** 3209 * skb_padto - pad an skbuff up to a minimal size 3210 * @skb: buffer to pad 3211 * @len: minimal length 3212 * 3213 * Pads up a buffer to ensure the trailing bytes exist and are 3214 * blanked. If the buffer already contains sufficient data it 3215 * is untouched. Otherwise it is extended. Returns zero on 3216 * success. The skb is freed on error. 3217 */ 3218static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3219{ 3220 unsigned int size = skb->len; 3221 if (likely(size >= len)) 3222 return 0; 3223 return skb_pad(skb, len - size); 3224} 3225 3226/** 3227 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3228 * @skb: buffer to pad 3229 * @len: minimal length 3230 * @free_on_error: free buffer on error 3231 * 3232 * Pads up a buffer to ensure the trailing bytes exist and are 3233 * blanked. If the buffer already contains sufficient data it 3234 * is untouched. Otherwise it is extended. Returns zero on 3235 * success. The skb is freed on error if @free_on_error is true. 3236 */ 3237static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3238 unsigned int len, 3239 bool free_on_error) 3240{ 3241 unsigned int size = skb->len; 3242 3243 if (unlikely(size < len)) { 3244 len -= size; 3245 if (__skb_pad(skb, len, free_on_error)) 3246 return -ENOMEM; 3247 __skb_put(skb, len); 3248 } 3249 return 0; 3250} 3251 3252/** 3253 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3254 * @skb: buffer to pad 3255 * @len: minimal length 3256 * 3257 * Pads up a buffer to ensure the trailing bytes exist and are 3258 * blanked. If the buffer already contains sufficient data it 3259 * is untouched. Otherwise it is extended. Returns zero on 3260 * success. The skb is freed on error. 3261 */ 3262static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3263{ 3264 return __skb_put_padto(skb, len, true); 3265} 3266 3267static inline int skb_add_data(struct sk_buff *skb, 3268 struct iov_iter *from, int copy) 3269{ 3270 const int off = skb->len; 3271 3272 if (skb->ip_summed == CHECKSUM_NONE) { 3273 __wsum csum = 0; 3274 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3275 &csum, from)) { 3276 skb->csum = csum_block_add(skb->csum, csum, off); 3277 return 0; 3278 } 3279 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3280 return 0; 3281 3282 __skb_trim(skb, off); 3283 return -EFAULT; 3284} 3285 3286static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3287 const struct page *page, int off) 3288{ 3289 if (skb_zcopy(skb)) 3290 return false; 3291 if (i) { 3292 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3293 3294 return page == skb_frag_page(frag) && 3295 off == skb_frag_off(frag) + skb_frag_size(frag); 3296 } 3297 return false; 3298} 3299 3300static inline int __skb_linearize(struct sk_buff *skb) 3301{ 3302 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3303} 3304 3305/** 3306 * skb_linearize - convert paged skb to linear one 3307 * @skb: buffer to linarize 3308 * 3309 * If there is no free memory -ENOMEM is returned, otherwise zero 3310 * is returned and the old skb data released. 3311 */ 3312static inline int skb_linearize(struct sk_buff *skb) 3313{ 3314 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3315} 3316 3317/** 3318 * skb_has_shared_frag - can any frag be overwritten 3319 * @skb: buffer to test 3320 * 3321 * Return true if the skb has at least one frag that might be modified 3322 * by an external entity (as in vmsplice()/sendfile()) 3323 */ 3324static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3325{ 3326 return skb_is_nonlinear(skb) && 3327 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG; 3328} 3329 3330/** 3331 * skb_linearize_cow - make sure skb is linear and writable 3332 * @skb: buffer to process 3333 * 3334 * If there is no free memory -ENOMEM is returned, otherwise zero 3335 * is returned and the old skb data released. 3336 */ 3337static inline int skb_linearize_cow(struct sk_buff *skb) 3338{ 3339 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3340 __skb_linearize(skb) : 0; 3341} 3342 3343static __always_inline void 3344__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3345 unsigned int off) 3346{ 3347 if (skb->ip_summed == CHECKSUM_COMPLETE) 3348 skb->csum = csum_block_sub(skb->csum, 3349 csum_partial(start, len, 0), off); 3350 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3351 skb_checksum_start_offset(skb) < 0) 3352 skb->ip_summed = CHECKSUM_NONE; 3353} 3354 3355/** 3356 * skb_postpull_rcsum - update checksum for received skb after pull 3357 * @skb: buffer to update 3358 * @start: start of data before pull 3359 * @len: length of data pulled 3360 * 3361 * After doing a pull on a received packet, you need to call this to 3362 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3363 * CHECKSUM_NONE so that it can be recomputed from scratch. 3364 */ 3365static inline void skb_postpull_rcsum(struct sk_buff *skb, 3366 const void *start, unsigned int len) 3367{ 3368 __skb_postpull_rcsum(skb, start, len, 0); 3369} 3370 3371static __always_inline void 3372__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3373 unsigned int off) 3374{ 3375 if (skb->ip_summed == CHECKSUM_COMPLETE) 3376 skb->csum = csum_block_add(skb->csum, 3377 csum_partial(start, len, 0), off); 3378} 3379 3380/** 3381 * skb_postpush_rcsum - update checksum for received skb after push 3382 * @skb: buffer to update 3383 * @start: start of data after push 3384 * @len: length of data pushed 3385 * 3386 * After doing a push on a received packet, you need to call this to 3387 * update the CHECKSUM_COMPLETE checksum. 3388 */ 3389static inline void skb_postpush_rcsum(struct sk_buff *skb, 3390 const void *start, unsigned int len) 3391{ 3392 __skb_postpush_rcsum(skb, start, len, 0); 3393} 3394 3395void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3396 3397/** 3398 * skb_push_rcsum - push skb and update receive checksum 3399 * @skb: buffer to update 3400 * @len: length of data pulled 3401 * 3402 * This function performs an skb_push on the packet and updates 3403 * the CHECKSUM_COMPLETE checksum. It should be used on 3404 * receive path processing instead of skb_push unless you know 3405 * that the checksum difference is zero (e.g., a valid IP header) 3406 * or you are setting ip_summed to CHECKSUM_NONE. 3407 */ 3408static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3409{ 3410 skb_push(skb, len); 3411 skb_postpush_rcsum(skb, skb->data, len); 3412 return skb->data; 3413} 3414 3415int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3416/** 3417 * pskb_trim_rcsum - trim received skb and update checksum 3418 * @skb: buffer to trim 3419 * @len: new length 3420 * 3421 * This is exactly the same as pskb_trim except that it ensures the 3422 * checksum of received packets are still valid after the operation. 3423 * It can change skb pointers. 3424 */ 3425 3426static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3427{ 3428 if (likely(len >= skb->len)) 3429 return 0; 3430 return pskb_trim_rcsum_slow(skb, len); 3431} 3432 3433static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3434{ 3435 if (skb->ip_summed == CHECKSUM_COMPLETE) 3436 skb->ip_summed = CHECKSUM_NONE; 3437 __skb_trim(skb, len); 3438 return 0; 3439} 3440 3441static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3442{ 3443 if (skb->ip_summed == CHECKSUM_COMPLETE) 3444 skb->ip_summed = CHECKSUM_NONE; 3445 return __skb_grow(skb, len); 3446} 3447 3448#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3449#define skb_rb_first(root) rb_to_skb(rb_first(root)) 3450#define skb_rb_last(root) rb_to_skb(rb_last(root)) 3451#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3452#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3453 3454#define skb_queue_walk(queue, skb) \ 3455 for (skb = (queue)->next; \ 3456 skb != (struct sk_buff *)(queue); \ 3457 skb = skb->next) 3458 3459#define skb_queue_walk_safe(queue, skb, tmp) \ 3460 for (skb = (queue)->next, tmp = skb->next; \ 3461 skb != (struct sk_buff *)(queue); \ 3462 skb = tmp, tmp = skb->next) 3463 3464#define skb_queue_walk_from(queue, skb) \ 3465 for (; skb != (struct sk_buff *)(queue); \ 3466 skb = skb->next) 3467 3468#define skb_rbtree_walk(skb, root) \ 3469 for (skb = skb_rb_first(root); skb != NULL; \ 3470 skb = skb_rb_next(skb)) 3471 3472#define skb_rbtree_walk_from(skb) \ 3473 for (; skb != NULL; \ 3474 skb = skb_rb_next(skb)) 3475 3476#define skb_rbtree_walk_from_safe(skb, tmp) \ 3477 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3478 skb = tmp) 3479 3480#define skb_queue_walk_from_safe(queue, skb, tmp) \ 3481 for (tmp = skb->next; \ 3482 skb != (struct sk_buff *)(queue); \ 3483 skb = tmp, tmp = skb->next) 3484 3485#define skb_queue_reverse_walk(queue, skb) \ 3486 for (skb = (queue)->prev; \ 3487 skb != (struct sk_buff *)(queue); \ 3488 skb = skb->prev) 3489 3490#define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3491 for (skb = (queue)->prev, tmp = skb->prev; \ 3492 skb != (struct sk_buff *)(queue); \ 3493 skb = tmp, tmp = skb->prev) 3494 3495#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3496 for (tmp = skb->prev; \ 3497 skb != (struct sk_buff *)(queue); \ 3498 skb = tmp, tmp = skb->prev) 3499 3500static inline bool skb_has_frag_list(const struct sk_buff *skb) 3501{ 3502 return skb_shinfo(skb)->frag_list != NULL; 3503} 3504 3505static inline void skb_frag_list_init(struct sk_buff *skb) 3506{ 3507 skb_shinfo(skb)->frag_list = NULL; 3508} 3509 3510#define skb_walk_frags(skb, iter) \ 3511 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3512 3513 3514int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3515 int *err, long *timeo_p, 3516 const struct sk_buff *skb); 3517struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3518 struct sk_buff_head *queue, 3519 unsigned int flags, 3520 int *off, int *err, 3521 struct sk_buff **last); 3522struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3523 struct sk_buff_head *queue, 3524 unsigned int flags, int *off, int *err, 3525 struct sk_buff **last); 3526struct sk_buff *__skb_recv_datagram(struct sock *sk, 3527 struct sk_buff_head *sk_queue, 3528 unsigned int flags, int *off, int *err); 3529struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3530 int *err); 3531__poll_t datagram_poll(struct file *file, struct socket *sock, 3532 struct poll_table_struct *wait); 3533int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3534 struct iov_iter *to, int size); 3535static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3536 struct msghdr *msg, int size) 3537{ 3538 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3539} 3540int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3541 struct msghdr *msg); 3542int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3543 struct iov_iter *to, int len, 3544 struct ahash_request *hash); 3545int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3546 struct iov_iter *from, int len); 3547int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3548void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3549void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3550static inline void skb_free_datagram_locked(struct sock *sk, 3551 struct sk_buff *skb) 3552{ 3553 __skb_free_datagram_locked(sk, skb, 0); 3554} 3555int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3556int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3557int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3558__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3559 int len); 3560int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3561 struct pipe_inode_info *pipe, unsigned int len, 3562 unsigned int flags); 3563int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3564 int len); 3565void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3566unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3567int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3568 int len, int hlen); 3569void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3570int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3571void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3572bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3573bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3574struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3575struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3576 unsigned int offset); 3577struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3578int skb_ensure_writable(struct sk_buff *skb, int write_len); 3579int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3580int skb_vlan_pop(struct sk_buff *skb); 3581int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3582int skb_eth_pop(struct sk_buff *skb); 3583int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3584 const unsigned char *src); 3585int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3586 int mac_len, bool ethernet); 3587int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3588 bool ethernet); 3589int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3590int skb_mpls_dec_ttl(struct sk_buff *skb); 3591struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3592 gfp_t gfp); 3593 3594static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3595{ 3596 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3597} 3598 3599static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3600{ 3601 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3602} 3603 3604struct skb_checksum_ops { 3605 __wsum (*update)(const void *mem, int len, __wsum wsum); 3606 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3607}; 3608 3609extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3610 3611__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3612 __wsum csum, const struct skb_checksum_ops *ops); 3613__wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3614 __wsum csum); 3615 3616static inline void * __must_check 3617__skb_header_pointer(const struct sk_buff *skb, int offset, 3618 int len, void *data, int hlen, void *buffer) 3619{ 3620 if (hlen - offset >= len) 3621 return data + offset; 3622 3623 if (!skb || 3624 skb_copy_bits(skb, offset, buffer, len) < 0) 3625 return NULL; 3626 3627 return buffer; 3628} 3629 3630static inline void * __must_check 3631skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3632{ 3633 return __skb_header_pointer(skb, offset, len, skb->data, 3634 skb_headlen(skb), buffer); 3635} 3636 3637/** 3638 * skb_needs_linearize - check if we need to linearize a given skb 3639 * depending on the given device features. 3640 * @skb: socket buffer to check 3641 * @features: net device features 3642 * 3643 * Returns true if either: 3644 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3645 * 2. skb is fragmented and the device does not support SG. 3646 */ 3647static inline bool skb_needs_linearize(struct sk_buff *skb, 3648 netdev_features_t features) 3649{ 3650 return skb_is_nonlinear(skb) && 3651 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3652 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3653} 3654 3655static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3656 void *to, 3657 const unsigned int len) 3658{ 3659 memcpy(to, skb->data, len); 3660} 3661 3662static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3663 const int offset, void *to, 3664 const unsigned int len) 3665{ 3666 memcpy(to, skb->data + offset, len); 3667} 3668 3669static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3670 const void *from, 3671 const unsigned int len) 3672{ 3673 memcpy(skb->data, from, len); 3674} 3675 3676static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3677 const int offset, 3678 const void *from, 3679 const unsigned int len) 3680{ 3681 memcpy(skb->data + offset, from, len); 3682} 3683 3684void skb_init(void); 3685 3686static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3687{ 3688 return skb->tstamp; 3689} 3690 3691/** 3692 * skb_get_timestamp - get timestamp from a skb 3693 * @skb: skb to get stamp from 3694 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3695 * 3696 * Timestamps are stored in the skb as offsets to a base timestamp. 3697 * This function converts the offset back to a struct timeval and stores 3698 * it in stamp. 3699 */ 3700static inline void skb_get_timestamp(const struct sk_buff *skb, 3701 struct __kernel_old_timeval *stamp) 3702{ 3703 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3704} 3705 3706static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3707 struct __kernel_sock_timeval *stamp) 3708{ 3709 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3710 3711 stamp->tv_sec = ts.tv_sec; 3712 stamp->tv_usec = ts.tv_nsec / 1000; 3713} 3714 3715static inline void skb_get_timestampns(const struct sk_buff *skb, 3716 struct __kernel_old_timespec *stamp) 3717{ 3718 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3719 3720 stamp->tv_sec = ts.tv_sec; 3721 stamp->tv_nsec = ts.tv_nsec; 3722} 3723 3724static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3725 struct __kernel_timespec *stamp) 3726{ 3727 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3728 3729 stamp->tv_sec = ts.tv_sec; 3730 stamp->tv_nsec = ts.tv_nsec; 3731} 3732 3733static inline void __net_timestamp(struct sk_buff *skb) 3734{ 3735 skb->tstamp = ktime_get_real(); 3736} 3737 3738static inline ktime_t net_timedelta(ktime_t t) 3739{ 3740 return ktime_sub(ktime_get_real(), t); 3741} 3742 3743static inline ktime_t net_invalid_timestamp(void) 3744{ 3745 return 0; 3746} 3747 3748static inline u8 skb_metadata_len(const struct sk_buff *skb) 3749{ 3750 return skb_shinfo(skb)->meta_len; 3751} 3752 3753static inline void *skb_metadata_end(const struct sk_buff *skb) 3754{ 3755 return skb_mac_header(skb); 3756} 3757 3758static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3759 const struct sk_buff *skb_b, 3760 u8 meta_len) 3761{ 3762 const void *a = skb_metadata_end(skb_a); 3763 const void *b = skb_metadata_end(skb_b); 3764 /* Using more efficient varaiant than plain call to memcmp(). */ 3765#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3766 u64 diffs = 0; 3767 3768 switch (meta_len) { 3769#define __it(x, op) (x -= sizeof(u##op)) 3770#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3771 case 32: diffs |= __it_diff(a, b, 64); 3772 fallthrough; 3773 case 24: diffs |= __it_diff(a, b, 64); 3774 fallthrough; 3775 case 16: diffs |= __it_diff(a, b, 64); 3776 fallthrough; 3777 case 8: diffs |= __it_diff(a, b, 64); 3778 break; 3779 case 28: diffs |= __it_diff(a, b, 64); 3780 fallthrough; 3781 case 20: diffs |= __it_diff(a, b, 64); 3782 fallthrough; 3783 case 12: diffs |= __it_diff(a, b, 64); 3784 fallthrough; 3785 case 4: diffs |= __it_diff(a, b, 32); 3786 break; 3787 } 3788 return diffs; 3789#else 3790 return memcmp(a - meta_len, b - meta_len, meta_len); 3791#endif 3792} 3793 3794static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3795 const struct sk_buff *skb_b) 3796{ 3797 u8 len_a = skb_metadata_len(skb_a); 3798 u8 len_b = skb_metadata_len(skb_b); 3799 3800 if (!(len_a | len_b)) 3801 return false; 3802 3803 return len_a != len_b ? 3804 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3805} 3806 3807static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3808{ 3809 skb_shinfo(skb)->meta_len = meta_len; 3810} 3811 3812static inline void skb_metadata_clear(struct sk_buff *skb) 3813{ 3814 skb_metadata_set(skb, 0); 3815} 3816 3817struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3818 3819#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3820 3821void skb_clone_tx_timestamp(struct sk_buff *skb); 3822bool skb_defer_rx_timestamp(struct sk_buff *skb); 3823 3824#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3825 3826static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3827{ 3828} 3829 3830static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3831{ 3832 return false; 3833} 3834 3835#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3836 3837/** 3838 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3839 * 3840 * PHY drivers may accept clones of transmitted packets for 3841 * timestamping via their phy_driver.txtstamp method. These drivers 3842 * must call this function to return the skb back to the stack with a 3843 * timestamp. 3844 * 3845 * @skb: clone of the original outgoing packet 3846 * @hwtstamps: hardware time stamps 3847 * 3848 */ 3849void skb_complete_tx_timestamp(struct sk_buff *skb, 3850 struct skb_shared_hwtstamps *hwtstamps); 3851 3852void __skb_tstamp_tx(struct sk_buff *orig_skb, 3853 struct skb_shared_hwtstamps *hwtstamps, 3854 struct sock *sk, int tstype); 3855 3856/** 3857 * skb_tstamp_tx - queue clone of skb with send time stamps 3858 * @orig_skb: the original outgoing packet 3859 * @hwtstamps: hardware time stamps, may be NULL if not available 3860 * 3861 * If the skb has a socket associated, then this function clones the 3862 * skb (thus sharing the actual data and optional structures), stores 3863 * the optional hardware time stamping information (if non NULL) or 3864 * generates a software time stamp (otherwise), then queues the clone 3865 * to the error queue of the socket. Errors are silently ignored. 3866 */ 3867void skb_tstamp_tx(struct sk_buff *orig_skb, 3868 struct skb_shared_hwtstamps *hwtstamps); 3869 3870/** 3871 * skb_tx_timestamp() - Driver hook for transmit timestamping 3872 * 3873 * Ethernet MAC Drivers should call this function in their hard_xmit() 3874 * function immediately before giving the sk_buff to the MAC hardware. 3875 * 3876 * Specifically, one should make absolutely sure that this function is 3877 * called before TX completion of this packet can trigger. Otherwise 3878 * the packet could potentially already be freed. 3879 * 3880 * @skb: A socket buffer. 3881 */ 3882static inline void skb_tx_timestamp(struct sk_buff *skb) 3883{ 3884 skb_clone_tx_timestamp(skb); 3885 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3886 skb_tstamp_tx(skb, NULL); 3887} 3888 3889/** 3890 * skb_complete_wifi_ack - deliver skb with wifi status 3891 * 3892 * @skb: the original outgoing packet 3893 * @acked: ack status 3894 * 3895 */ 3896void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3897 3898__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3899__sum16 __skb_checksum_complete(struct sk_buff *skb); 3900 3901static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3902{ 3903 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3904 skb->csum_valid || 3905 (skb->ip_summed == CHECKSUM_PARTIAL && 3906 skb_checksum_start_offset(skb) >= 0)); 3907} 3908 3909/** 3910 * skb_checksum_complete - Calculate checksum of an entire packet 3911 * @skb: packet to process 3912 * 3913 * This function calculates the checksum over the entire packet plus 3914 * the value of skb->csum. The latter can be used to supply the 3915 * checksum of a pseudo header as used by TCP/UDP. It returns the 3916 * checksum. 3917 * 3918 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 3919 * this function can be used to verify that checksum on received 3920 * packets. In that case the function should return zero if the 3921 * checksum is correct. In particular, this function will return zero 3922 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 3923 * hardware has already verified the correctness of the checksum. 3924 */ 3925static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 3926{ 3927 return skb_csum_unnecessary(skb) ? 3928 0 : __skb_checksum_complete(skb); 3929} 3930 3931static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 3932{ 3933 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3934 if (skb->csum_level == 0) 3935 skb->ip_summed = CHECKSUM_NONE; 3936 else 3937 skb->csum_level--; 3938 } 3939} 3940 3941static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 3942{ 3943 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3944 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 3945 skb->csum_level++; 3946 } else if (skb->ip_summed == CHECKSUM_NONE) { 3947 skb->ip_summed = CHECKSUM_UNNECESSARY; 3948 skb->csum_level = 0; 3949 } 3950} 3951 3952static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 3953{ 3954 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3955 skb->ip_summed = CHECKSUM_NONE; 3956 skb->csum_level = 0; 3957 } 3958} 3959 3960/* Check if we need to perform checksum complete validation. 3961 * 3962 * Returns true if checksum complete is needed, false otherwise 3963 * (either checksum is unnecessary or zero checksum is allowed). 3964 */ 3965static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 3966 bool zero_okay, 3967 __sum16 check) 3968{ 3969 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 3970 skb->csum_valid = 1; 3971 __skb_decr_checksum_unnecessary(skb); 3972 return false; 3973 } 3974 3975 return true; 3976} 3977 3978/* For small packets <= CHECKSUM_BREAK perform checksum complete directly 3979 * in checksum_init. 3980 */ 3981#define CHECKSUM_BREAK 76 3982 3983/* Unset checksum-complete 3984 * 3985 * Unset checksum complete can be done when packet is being modified 3986 * (uncompressed for instance) and checksum-complete value is 3987 * invalidated. 3988 */ 3989static inline void skb_checksum_complete_unset(struct sk_buff *skb) 3990{ 3991 if (skb->ip_summed == CHECKSUM_COMPLETE) 3992 skb->ip_summed = CHECKSUM_NONE; 3993} 3994 3995/* Validate (init) checksum based on checksum complete. 3996 * 3997 * Return values: 3998 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 3999 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4000 * checksum is stored in skb->csum for use in __skb_checksum_complete 4001 * non-zero: value of invalid checksum 4002 * 4003 */ 4004static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4005 bool complete, 4006 __wsum psum) 4007{ 4008 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4009 if (!csum_fold(csum_add(psum, skb->csum))) { 4010 skb->csum_valid = 1; 4011 return 0; 4012 } 4013 } 4014 4015 skb->csum = psum; 4016 4017 if (complete || skb->len <= CHECKSUM_BREAK) { 4018 __sum16 csum; 4019 4020 csum = __skb_checksum_complete(skb); 4021 skb->csum_valid = !csum; 4022 return csum; 4023 } 4024 4025 return 0; 4026} 4027 4028static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4029{ 4030 return 0; 4031} 4032 4033/* Perform checksum validate (init). Note that this is a macro since we only 4034 * want to calculate the pseudo header which is an input function if necessary. 4035 * First we try to validate without any computation (checksum unnecessary) and 4036 * then calculate based on checksum complete calling the function to compute 4037 * pseudo header. 4038 * 4039 * Return values: 4040 * 0: checksum is validated or try to in skb_checksum_complete 4041 * non-zero: value of invalid checksum 4042 */ 4043#define __skb_checksum_validate(skb, proto, complete, \ 4044 zero_okay, check, compute_pseudo) \ 4045({ \ 4046 __sum16 __ret = 0; \ 4047 skb->csum_valid = 0; \ 4048 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4049 __ret = __skb_checksum_validate_complete(skb, \ 4050 complete, compute_pseudo(skb, proto)); \ 4051 __ret; \ 4052}) 4053 4054#define skb_checksum_init(skb, proto, compute_pseudo) \ 4055 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4056 4057#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4058 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4059 4060#define skb_checksum_validate(skb, proto, compute_pseudo) \ 4061 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4062 4063#define skb_checksum_validate_zero_check(skb, proto, check, \ 4064 compute_pseudo) \ 4065 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4066 4067#define skb_checksum_simple_validate(skb) \ 4068 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4069 4070static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4071{ 4072 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4073} 4074 4075static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4076{ 4077 skb->csum = ~pseudo; 4078 skb->ip_summed = CHECKSUM_COMPLETE; 4079} 4080 4081#define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4082do { \ 4083 if (__skb_checksum_convert_check(skb)) \ 4084 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4085} while (0) 4086 4087static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4088 u16 start, u16 offset) 4089{ 4090 skb->ip_summed = CHECKSUM_PARTIAL; 4091 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4092 skb->csum_offset = offset - start; 4093} 4094 4095/* Update skbuf and packet to reflect the remote checksum offload operation. 4096 * When called, ptr indicates the starting point for skb->csum when 4097 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4098 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4099 */ 4100static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4101 int start, int offset, bool nopartial) 4102{ 4103 __wsum delta; 4104 4105 if (!nopartial) { 4106 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4107 return; 4108 } 4109 4110 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4111 __skb_checksum_complete(skb); 4112 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4113 } 4114 4115 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4116 4117 /* Adjust skb->csum since we changed the packet */ 4118 skb->csum = csum_add(skb->csum, delta); 4119} 4120 4121static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4122{ 4123#if IS_ENABLED(CONFIG_NF_CONNTRACK) 4124 return (void *)(skb->_nfct & NFCT_PTRMASK); 4125#else 4126 return NULL; 4127#endif 4128} 4129 4130static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4131{ 4132#if IS_ENABLED(CONFIG_NF_CONNTRACK) 4133 return skb->_nfct; 4134#else 4135 return 0UL; 4136#endif 4137} 4138 4139static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4140{ 4141#if IS_ENABLED(CONFIG_NF_CONNTRACK) 4142 skb->_nfct = nfct; 4143#endif 4144} 4145 4146#ifdef CONFIG_SKB_EXTENSIONS 4147enum skb_ext_id { 4148#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4149 SKB_EXT_BRIDGE_NF, 4150#endif 4151#ifdef CONFIG_XFRM 4152 SKB_EXT_SEC_PATH, 4153#endif 4154#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4155 TC_SKB_EXT, 4156#endif 4157#if IS_ENABLED(CONFIG_MPTCP) 4158 SKB_EXT_MPTCP, 4159#endif 4160 SKB_EXT_NUM, /* must be last */ 4161}; 4162 4163/** 4164 * struct skb_ext - sk_buff extensions 4165 * @refcnt: 1 on allocation, deallocated on 0 4166 * @offset: offset to add to @data to obtain extension address 4167 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4168 * @data: start of extension data, variable sized 4169 * 4170 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4171 * to use 'u8' types while allowing up to 2kb worth of extension data. 4172 */ 4173struct skb_ext { 4174 refcount_t refcnt; 4175 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4176 u8 chunks; /* same */ 4177 char data[] __aligned(8); 4178}; 4179 4180struct skb_ext *__skb_ext_alloc(gfp_t flags); 4181void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4182 struct skb_ext *ext); 4183void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4184void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4185void __skb_ext_put(struct skb_ext *ext); 4186 4187static inline void skb_ext_put(struct sk_buff *skb) 4188{ 4189 if (skb->active_extensions) 4190 __skb_ext_put(skb->extensions); 4191} 4192 4193static inline void __skb_ext_copy(struct sk_buff *dst, 4194 const struct sk_buff *src) 4195{ 4196 dst->active_extensions = src->active_extensions; 4197 4198 if (src->active_extensions) { 4199 struct skb_ext *ext = src->extensions; 4200 4201 refcount_inc(&ext->refcnt); 4202 dst->extensions = ext; 4203 } 4204} 4205 4206static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4207{ 4208 skb_ext_put(dst); 4209 __skb_ext_copy(dst, src); 4210} 4211 4212static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4213{ 4214 return !!ext->offset[i]; 4215} 4216 4217static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4218{ 4219 return skb->active_extensions & (1 << id); 4220} 4221 4222static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4223{ 4224 if (skb_ext_exist(skb, id)) 4225 __skb_ext_del(skb, id); 4226} 4227 4228static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4229{ 4230 if (skb_ext_exist(skb, id)) { 4231 struct skb_ext *ext = skb->extensions; 4232 4233 return (void *)ext + (ext->offset[id] << 3); 4234 } 4235 4236 return NULL; 4237} 4238 4239static inline void skb_ext_reset(struct sk_buff *skb) 4240{ 4241 if (unlikely(skb->active_extensions)) { 4242 __skb_ext_put(skb->extensions); 4243 skb->active_extensions = 0; 4244 } 4245} 4246 4247static inline bool skb_has_extensions(struct sk_buff *skb) 4248{ 4249 return unlikely(skb->active_extensions); 4250} 4251#else 4252static inline void skb_ext_put(struct sk_buff *skb) {} 4253static inline void skb_ext_reset(struct sk_buff *skb) {} 4254static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4255static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4256static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4257static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4258#endif /* CONFIG_SKB_EXTENSIONS */ 4259 4260static inline void nf_reset_ct(struct sk_buff *skb) 4261{ 4262#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4263 nf_conntrack_put(skb_nfct(skb)); 4264 skb->_nfct = 0; 4265#endif 4266} 4267 4268static inline void nf_reset_trace(struct sk_buff *skb) 4269{ 4270#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4271 skb->nf_trace = 0; 4272#endif 4273} 4274 4275static inline void ipvs_reset(struct sk_buff *skb) 4276{ 4277#if IS_ENABLED(CONFIG_IP_VS) 4278 skb->ipvs_property = 0; 4279#endif 4280} 4281 4282/* Note: This doesn't put any conntrack info in dst. */ 4283static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4284 bool copy) 4285{ 4286#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4287 dst->_nfct = src->_nfct; 4288 nf_conntrack_get(skb_nfct(src)); 4289#endif 4290#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4291 if (copy) 4292 dst->nf_trace = src->nf_trace; 4293#endif 4294} 4295 4296static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4297{ 4298#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4299 nf_conntrack_put(skb_nfct(dst)); 4300#endif 4301 __nf_copy(dst, src, true); 4302} 4303 4304#ifdef CONFIG_NETWORK_SECMARK 4305static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4306{ 4307 to->secmark = from->secmark; 4308} 4309 4310static inline void skb_init_secmark(struct sk_buff *skb) 4311{ 4312 skb->secmark = 0; 4313} 4314#else 4315static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4316{ } 4317 4318static inline void skb_init_secmark(struct sk_buff *skb) 4319{ } 4320#endif 4321 4322static inline int secpath_exists(const struct sk_buff *skb) 4323{ 4324#ifdef CONFIG_XFRM 4325 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4326#else 4327 return 0; 4328#endif 4329} 4330 4331static inline bool skb_irq_freeable(const struct sk_buff *skb) 4332{ 4333 return !skb->destructor && 4334 !secpath_exists(skb) && 4335 !skb_nfct(skb) && 4336 !skb->_skb_refdst && 4337 !skb_has_frag_list(skb); 4338} 4339 4340static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4341{ 4342 skb->queue_mapping = queue_mapping; 4343} 4344 4345static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4346{ 4347 return skb->queue_mapping; 4348} 4349 4350static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4351{ 4352 to->queue_mapping = from->queue_mapping; 4353} 4354 4355static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4356{ 4357 skb->queue_mapping = rx_queue + 1; 4358} 4359 4360static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4361{ 4362 return skb->queue_mapping - 1; 4363} 4364 4365static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4366{ 4367 return skb->queue_mapping != 0; 4368} 4369 4370static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4371{ 4372 skb->dst_pending_confirm = val; 4373} 4374 4375static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4376{ 4377 return skb->dst_pending_confirm != 0; 4378} 4379 4380static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4381{ 4382#ifdef CONFIG_XFRM 4383 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4384#else 4385 return NULL; 4386#endif 4387} 4388 4389/* Keeps track of mac header offset relative to skb->head. 4390 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4391 * For non-tunnel skb it points to skb_mac_header() and for 4392 * tunnel skb it points to outer mac header. 4393 * Keeps track of level of encapsulation of network headers. 4394 */ 4395struct skb_gso_cb { 4396 union { 4397 int mac_offset; 4398 int data_offset; 4399 }; 4400 int encap_level; 4401 __wsum csum; 4402 __u16 csum_start; 4403}; 4404#define SKB_GSO_CB_OFFSET 32 4405#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4406 4407static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4408{ 4409 return (skb_mac_header(inner_skb) - inner_skb->head) - 4410 SKB_GSO_CB(inner_skb)->mac_offset; 4411} 4412 4413static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4414{ 4415 int new_headroom, headroom; 4416 int ret; 4417 4418 headroom = skb_headroom(skb); 4419 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4420 if (ret) 4421 return ret; 4422 4423 new_headroom = skb_headroom(skb); 4424 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4425 return 0; 4426} 4427 4428static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4429{ 4430 /* Do not update partial checksums if remote checksum is enabled. */ 4431 if (skb->remcsum_offload) 4432 return; 4433 4434 SKB_GSO_CB(skb)->csum = res; 4435 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4436} 4437 4438/* Compute the checksum for a gso segment. First compute the checksum value 4439 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4440 * then add in skb->csum (checksum from csum_start to end of packet). 4441 * skb->csum and csum_start are then updated to reflect the checksum of the 4442 * resultant packet starting from the transport header-- the resultant checksum 4443 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4444 * header. 4445 */ 4446static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4447{ 4448 unsigned char *csum_start = skb_transport_header(skb); 4449 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4450 __wsum partial = SKB_GSO_CB(skb)->csum; 4451 4452 SKB_GSO_CB(skb)->csum = res; 4453 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4454 4455 return csum_fold(csum_partial(csum_start, plen, partial)); 4456} 4457 4458static inline bool skb_is_gso(const struct sk_buff *skb) 4459{ 4460 return skb_shinfo(skb)->gso_size; 4461} 4462 4463/* Note: Should be called only if skb_is_gso(skb) is true */ 4464static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4465{ 4466 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4467} 4468 4469/* Note: Should be called only if skb_is_gso(skb) is true */ 4470static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4471{ 4472 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4473} 4474 4475/* Note: Should be called only if skb_is_gso(skb) is true */ 4476static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4477{ 4478 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4479} 4480 4481static inline void skb_gso_reset(struct sk_buff *skb) 4482{ 4483 skb_shinfo(skb)->gso_size = 0; 4484 skb_shinfo(skb)->gso_segs = 0; 4485 skb_shinfo(skb)->gso_type = 0; 4486} 4487 4488static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4489 u16 increment) 4490{ 4491 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4492 return; 4493 shinfo->gso_size += increment; 4494} 4495 4496static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4497 u16 decrement) 4498{ 4499 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4500 return; 4501 shinfo->gso_size -= decrement; 4502} 4503 4504void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4505 4506static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4507{ 4508 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4509 * wanted then gso_type will be set. */ 4510 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4511 4512 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4513 unlikely(shinfo->gso_type == 0)) { 4514 __skb_warn_lro_forwarding(skb); 4515 return true; 4516 } 4517 return false; 4518} 4519 4520static inline void skb_forward_csum(struct sk_buff *skb) 4521{ 4522 /* Unfortunately we don't support this one. Any brave souls? */ 4523 if (skb->ip_summed == CHECKSUM_COMPLETE) 4524 skb->ip_summed = CHECKSUM_NONE; 4525} 4526 4527/** 4528 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4529 * @skb: skb to check 4530 * 4531 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4532 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4533 * use this helper, to document places where we make this assertion. 4534 */ 4535static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4536{ 4537#ifdef DEBUG 4538 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4539#endif 4540} 4541 4542bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4543 4544int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4545struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4546 unsigned int transport_len, 4547 __sum16(*skb_chkf)(struct sk_buff *skb)); 4548 4549/** 4550 * skb_head_is_locked - Determine if the skb->head is locked down 4551 * @skb: skb to check 4552 * 4553 * The head on skbs build around a head frag can be removed if they are 4554 * not cloned. This function returns true if the skb head is locked down 4555 * due to either being allocated via kmalloc, or by being a clone with 4556 * multiple references to the head. 4557 */ 4558static inline bool skb_head_is_locked(const struct sk_buff *skb) 4559{ 4560 return !skb->head_frag || skb_cloned(skb); 4561} 4562 4563/* Local Checksum Offload. 4564 * Compute outer checksum based on the assumption that the 4565 * inner checksum will be offloaded later. 4566 * See Documentation/networking/checksum-offloads.rst for 4567 * explanation of how this works. 4568 * Fill in outer checksum adjustment (e.g. with sum of outer 4569 * pseudo-header) before calling. 4570 * Also ensure that inner checksum is in linear data area. 4571 */ 4572static inline __wsum lco_csum(struct sk_buff *skb) 4573{ 4574 unsigned char *csum_start = skb_checksum_start(skb); 4575 unsigned char *l4_hdr = skb_transport_header(skb); 4576 __wsum partial; 4577 4578 /* Start with complement of inner checksum adjustment */ 4579 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4580 skb->csum_offset)); 4581 4582 /* Add in checksum of our headers (incl. outer checksum 4583 * adjustment filled in by caller) and return result. 4584 */ 4585 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4586} 4587 4588static inline bool skb_is_redirected(const struct sk_buff *skb) 4589{ 4590#ifdef CONFIG_NET_REDIRECT 4591 return skb->redirected; 4592#else 4593 return false; 4594#endif 4595} 4596 4597static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4598{ 4599#ifdef CONFIG_NET_REDIRECT 4600 skb->redirected = 1; 4601 skb->from_ingress = from_ingress; 4602 if (skb->from_ingress) 4603 skb->tstamp = 0; 4604#endif 4605} 4606 4607static inline void skb_reset_redirect(struct sk_buff *skb) 4608{ 4609#ifdef CONFIG_NET_REDIRECT 4610 skb->redirected = 0; 4611#endif 4612} 4613 4614static inline void skb_set_kcov_handle(struct sk_buff *skb, 4615 const u64 kcov_handle) 4616{ 4617#ifdef CONFIG_KCOV 4618 skb->kcov_handle = kcov_handle; 4619#endif 4620} 4621 4622static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4623{ 4624#ifdef CONFIG_KCOV 4625 return skb->kcov_handle; 4626#else 4627 return 0; 4628#endif 4629} 4630 4631#endif /* __KERNEL__ */ 4632#endif /* _LINUX_SKBUFF_H */