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