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