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 * @_sk_redir: socket redirection information for skmsg 660 * @_nfct: Associated connection, if any (with nfctinfo bits) 661 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 662 * @skb_iif: ifindex of device we arrived on 663 * @tc_index: Traffic control index 664 * @hash: the packet hash 665 * @queue_mapping: Queue mapping for multiqueue devices 666 * @head_frag: skb was allocated from page fragments, 667 * not allocated by kmalloc() or vmalloc(). 668 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 669 * @active_extensions: active extensions (skb_ext_id types) 670 * @ndisc_nodetype: router type (from link layer) 671 * @ooo_okay: allow the mapping of a socket to a queue to be changed 672 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 673 * ports. 674 * @sw_hash: indicates hash was computed in software stack 675 * @wifi_acked_valid: wifi_acked was set 676 * @wifi_acked: whether frame was acked on wifi or not 677 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 678 * @encapsulation: indicates the inner headers in the skbuff are valid 679 * @encap_hdr_csum: software checksum is needed 680 * @csum_valid: checksum is already valid 681 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 682 * @csum_complete_sw: checksum was completed by software 683 * @csum_level: indicates the number of consecutive checksums found in 684 * the packet minus one that have been verified as 685 * CHECKSUM_UNNECESSARY (max 3) 686 * @dst_pending_confirm: need to confirm neighbour 687 * @decrypted: Decrypted SKB 688 * @napi_id: id of the NAPI struct this skb came from 689 * @sender_cpu: (aka @napi_id) source CPU in XPS 690 * @secmark: security marking 691 * @mark: Generic packet mark 692 * @reserved_tailroom: (aka @mark) number of bytes of free space available 693 * at the tail of an sk_buff 694 * @vlan_present: VLAN tag is present 695 * @vlan_proto: vlan encapsulation protocol 696 * @vlan_tci: vlan tag control information 697 * @inner_protocol: Protocol (encapsulation) 698 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 699 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 700 * @inner_transport_header: Inner transport layer header (encapsulation) 701 * @inner_network_header: Network layer header (encapsulation) 702 * @inner_mac_header: Link layer header (encapsulation) 703 * @transport_header: Transport layer header 704 * @network_header: Network layer header 705 * @mac_header: Link layer header 706 * @kcov_handle: KCOV remote handle for remote coverage collection 707 * @tail: Tail pointer 708 * @end: End pointer 709 * @head: Head of buffer 710 * @data: Data head pointer 711 * @truesize: Buffer size 712 * @users: User count - see {datagram,tcp}.c 713 * @extensions: allocated extensions, valid if active_extensions is nonzero 714 */ 715 716 struct sk_buff { 717 union { 718 struct { 719 /* These two members must be first. */ 720 struct sk_buff *next; 721 struct sk_buff *prev; 722 723 union { 724 struct net_device *dev; 725 /* Some protocols might use this space to store information, 726 * while device pointer would be NULL. 727 * UDP receive path is one user. 728 */ 729 unsigned long dev_scratch; 730 }; 731 }; 732 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 733 struct list_head list; 734 }; 735 736 union { 737 struct sock *sk; 738 int ip_defrag_offset; 739 }; 740 741 union { 742 ktime_t tstamp; 743 u64 skb_mstamp_ns; /* earliest departure time */ 744 }; 745 /* 746 * This is the control buffer. It is free to use for every 747 * layer. Please put your private variables there. If you 748 * want to keep them across layers you have to do a skb_clone() 749 * first. This is owned by whoever has the skb queued ATM. 750 */ 751 char cb[48] __aligned(8); 752 753 union { 754 struct { 755 unsigned long _skb_refdst; 756 void (*destructor)(struct sk_buff *skb); 757 }; 758 struct list_head tcp_tsorted_anchor; 759 #ifdef CONFIG_NET_SOCK_MSG 760 unsigned long _sk_redir; 761 #endif 762 }; 763 764 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 765 unsigned long _nfct; 766 #endif 767 unsigned int len, 768 data_len; 769 __u16 mac_len, 770 hdr_len; 771 772 /* Following fields are _not_ copied in __copy_skb_header() 773 * Note that queue_mapping is here mostly to fill a hole. 774 */ 775 __u16 queue_mapping; 776 777 /* if you move cloned around you also must adapt those constants */ 778 #ifdef __BIG_ENDIAN_BITFIELD 779 #define CLONED_MASK (1 << 7) 780 #else 781 #define CLONED_MASK 1 782 #endif 783 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset) 784 785 /* private: */ 786 __u8 __cloned_offset[0]; 787 /* public: */ 788 __u8 cloned:1, 789 nohdr:1, 790 fclone:2, 791 peeked:1, 792 head_frag:1, 793 pfmemalloc:1; 794 #ifdef CONFIG_SKB_EXTENSIONS 795 __u8 active_extensions; 796 #endif 797 /* fields enclosed in headers_start/headers_end are copied 798 * using a single memcpy() in __copy_skb_header() 799 */ 800 /* private: */ 801 __u32 headers_start[0]; 802 /* public: */ 803 804 /* if you move pkt_type around you also must adapt those constants */ 805 #ifdef __BIG_ENDIAN_BITFIELD 806 #define PKT_TYPE_MAX (7 << 5) 807 #else 808 #define PKT_TYPE_MAX 7 809 #endif 810 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 811 812 /* private: */ 813 __u8 __pkt_type_offset[0]; 814 /* public: */ 815 __u8 pkt_type:3; 816 __u8 ignore_df:1; 817 __u8 nf_trace:1; 818 __u8 ip_summed:2; 819 __u8 ooo_okay:1; 820 821 __u8 l4_hash:1; 822 __u8 sw_hash:1; 823 __u8 wifi_acked_valid:1; 824 __u8 wifi_acked:1; 825 __u8 no_fcs:1; 826 /* Indicates the inner headers are valid in the skbuff. */ 827 __u8 encapsulation:1; 828 __u8 encap_hdr_csum:1; 829 __u8 csum_valid:1; 830 831 #ifdef __BIG_ENDIAN_BITFIELD 832 #define PKT_VLAN_PRESENT_BIT 7 833 #else 834 #define PKT_VLAN_PRESENT_BIT 0 835 #endif 836 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset) 837 /* private: */ 838 __u8 __pkt_vlan_present_offset[0]; 839 /* public: */ 840 __u8 vlan_present:1; 841 __u8 csum_complete_sw:1; 842 __u8 csum_level:2; 843 __u8 csum_not_inet:1; 844 __u8 dst_pending_confirm:1; 845 #ifdef CONFIG_IPV6_NDISC_NODETYPE 846 __u8 ndisc_nodetype:2; 847 #endif 848 849 __u8 ipvs_property:1; 850 __u8 inner_protocol_type:1; 851 __u8 remcsum_offload:1; 852 #ifdef CONFIG_NET_SWITCHDEV 853 __u8 offload_fwd_mark:1; 854 __u8 offload_l3_fwd_mark:1; 855 #endif 856 #ifdef CONFIG_NET_CLS_ACT 857 __u8 tc_skip_classify:1; 858 __u8 tc_at_ingress:1; 859 #endif 860 #ifdef CONFIG_NET_REDIRECT 861 __u8 redirected:1; 862 __u8 from_ingress:1; 863 #endif 864 #ifdef CONFIG_TLS_DEVICE 865 __u8 decrypted:1; 866 #endif 867 868 #ifdef CONFIG_NET_SCHED 869 __u16 tc_index; /* traffic control index */ 870 #endif 871 872 union { 873 __wsum csum; 874 struct { 875 __u16 csum_start; 876 __u16 csum_offset; 877 }; 878 }; 879 __u32 priority; 880 int skb_iif; 881 __u32 hash; 882 __be16 vlan_proto; 883 __u16 vlan_tci; 884 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 885 union { 886 unsigned int napi_id; 887 unsigned int sender_cpu; 888 }; 889 #endif 890 #ifdef CONFIG_NETWORK_SECMARK 891 __u32 secmark; 892 #endif 893 894 union { 895 __u32 mark; 896 __u32 reserved_tailroom; 897 }; 898 899 union { 900 __be16 inner_protocol; 901 __u8 inner_ipproto; 902 }; 903 904 __u16 inner_transport_header; 905 __u16 inner_network_header; 906 __u16 inner_mac_header; 907 908 __be16 protocol; 909 __u16 transport_header; 910 __u16 network_header; 911 __u16 mac_header; 912 913 #ifdef CONFIG_KCOV 914 u64 kcov_handle; 915 #endif 916 917 /* private: */ 918 __u32 headers_end[0]; 919 /* public: */ 920 921 /* These elements must be at the end, see alloc_skb() for details. */ 922 sk_buff_data_t tail; 923 sk_buff_data_t end; 924 unsigned char *head, 925 *data; 926 unsigned int truesize; 927 refcount_t users; 928 929 #ifdef CONFIG_SKB_EXTENSIONS 930 /* only useable after checking ->active_extensions != 0 */ 931 struct skb_ext *extensions; 932 #endif 933 }; 934 935 #ifdef __KERNEL__ 936 /* 937 * Handling routines are only of interest to the kernel 938 */ 939 940 #define SKB_ALLOC_FCLONE 0x01 941 #define SKB_ALLOC_RX 0x02 942 #define SKB_ALLOC_NAPI 0x04 943 944 /** 945 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 946 * @skb: buffer 947 */ 948 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 949 { 950 return unlikely(skb->pfmemalloc); 951 } 952 953 /* 954 * skb might have a dst pointer attached, refcounted or not. 955 * _skb_refdst low order bit is set if refcount was _not_ taken 956 */ 957 #define SKB_DST_NOREF 1UL 958 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 959 960 /** 961 * skb_dst - returns skb dst_entry 962 * @skb: buffer 963 * 964 * Returns skb dst_entry, regardless of reference taken or not. 965 */ 966 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 967 { 968 /* If refdst was not refcounted, check we still are in a 969 * rcu_read_lock section 970 */ 971 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 972 !rcu_read_lock_held() && 973 !rcu_read_lock_bh_held()); 974 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 975 } 976 977 /** 978 * skb_dst_set - sets skb dst 979 * @skb: buffer 980 * @dst: dst entry 981 * 982 * Sets skb dst, assuming a reference was taken on dst and should 983 * be released by skb_dst_drop() 984 */ 985 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 986 { 987 skb->_skb_refdst = (unsigned long)dst; 988 } 989 990 /** 991 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 992 * @skb: buffer 993 * @dst: dst entry 994 * 995 * Sets skb dst, assuming a reference was not taken on dst. 996 * If dst entry is cached, we do not take reference and dst_release 997 * will be avoided by refdst_drop. If dst entry is not cached, we take 998 * reference, so that last dst_release can destroy the dst immediately. 999 */ 1000 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 1001 { 1002 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1003 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1004 } 1005 1006 /** 1007 * skb_dst_is_noref - Test if skb dst isn't refcounted 1008 * @skb: buffer 1009 */ 1010 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1011 { 1012 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1013 } 1014 1015 /** 1016 * skb_rtable - Returns the skb &rtable 1017 * @skb: buffer 1018 */ 1019 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1020 { 1021 return (struct rtable *)skb_dst(skb); 1022 } 1023 1024 /* For mangling skb->pkt_type from user space side from applications 1025 * such as nft, tc, etc, we only allow a conservative subset of 1026 * possible pkt_types to be set. 1027 */ 1028 static inline bool skb_pkt_type_ok(u32 ptype) 1029 { 1030 return ptype <= PACKET_OTHERHOST; 1031 } 1032 1033 /** 1034 * skb_napi_id - Returns the skb's NAPI id 1035 * @skb: buffer 1036 */ 1037 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1038 { 1039 #ifdef CONFIG_NET_RX_BUSY_POLL 1040 return skb->napi_id; 1041 #else 1042 return 0; 1043 #endif 1044 } 1045 1046 /** 1047 * skb_unref - decrement the skb's reference count 1048 * @skb: buffer 1049 * 1050 * Returns true if we can free the skb. 1051 */ 1052 static inline bool skb_unref(struct sk_buff *skb) 1053 { 1054 if (unlikely(!skb)) 1055 return false; 1056 if (likely(refcount_read(&skb->users) == 1)) 1057 smp_rmb(); 1058 else if (likely(!refcount_dec_and_test(&skb->users))) 1059 return false; 1060 1061 return true; 1062 } 1063 1064 void skb_release_head_state(struct sk_buff *skb); 1065 void kfree_skb(struct sk_buff *skb); 1066 void kfree_skb_list(struct sk_buff *segs); 1067 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1068 void skb_tx_error(struct sk_buff *skb); 1069 1070 #ifdef CONFIG_TRACEPOINTS 1071 void consume_skb(struct sk_buff *skb); 1072 #else 1073 static inline void consume_skb(struct sk_buff *skb) 1074 { 1075 return kfree_skb(skb); 1076 } 1077 #endif 1078 1079 void __consume_stateless_skb(struct sk_buff *skb); 1080 void __kfree_skb(struct sk_buff *skb); 1081 extern struct kmem_cache *skbuff_head_cache; 1082 1083 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1084 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1085 bool *fragstolen, int *delta_truesize); 1086 1087 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1088 int node); 1089 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1090 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1091 struct sk_buff *build_skb_around(struct sk_buff *skb, 1092 void *data, unsigned int frag_size); 1093 1094 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1095 1096 /** 1097 * alloc_skb - allocate a network buffer 1098 * @size: size to allocate 1099 * @priority: allocation mask 1100 * 1101 * This function is a convenient wrapper around __alloc_skb(). 1102 */ 1103 static inline struct sk_buff *alloc_skb(unsigned int size, 1104 gfp_t priority) 1105 { 1106 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1107 } 1108 1109 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1110 unsigned long data_len, 1111 int max_page_order, 1112 int *errcode, 1113 gfp_t gfp_mask); 1114 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1115 1116 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1117 struct sk_buff_fclones { 1118 struct sk_buff skb1; 1119 1120 struct sk_buff skb2; 1121 1122 refcount_t fclone_ref; 1123 }; 1124 1125 /** 1126 * skb_fclone_busy - check if fclone is busy 1127 * @sk: socket 1128 * @skb: buffer 1129 * 1130 * Returns true if skb is a fast clone, and its clone is not freed. 1131 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1132 * so we also check that this didnt happen. 1133 */ 1134 static inline bool skb_fclone_busy(const struct sock *sk, 1135 const struct sk_buff *skb) 1136 { 1137 const struct sk_buff_fclones *fclones; 1138 1139 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1140 1141 return skb->fclone == SKB_FCLONE_ORIG && 1142 refcount_read(&fclones->fclone_ref) > 1 && 1143 READ_ONCE(fclones->skb2.sk) == sk; 1144 } 1145 1146 /** 1147 * alloc_skb_fclone - allocate a network buffer from fclone cache 1148 * @size: size to allocate 1149 * @priority: allocation mask 1150 * 1151 * This function is a convenient wrapper around __alloc_skb(). 1152 */ 1153 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1154 gfp_t priority) 1155 { 1156 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1157 } 1158 1159 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1160 void skb_headers_offset_update(struct sk_buff *skb, int off); 1161 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1162 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1163 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1164 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1165 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1166 gfp_t gfp_mask, bool fclone); 1167 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1168 gfp_t gfp_mask) 1169 { 1170 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1171 } 1172 1173 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1174 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1175 unsigned int headroom); 1176 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1177 int newtailroom, gfp_t priority); 1178 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1179 int offset, int len); 1180 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1181 int offset, int len); 1182 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1183 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1184 1185 /** 1186 * skb_pad - zero pad the tail of an skb 1187 * @skb: buffer to pad 1188 * @pad: space to pad 1189 * 1190 * Ensure that a buffer is followed by a padding area that is zero 1191 * filled. Used by network drivers which may DMA or transfer data 1192 * beyond the buffer end onto the wire. 1193 * 1194 * May return error in out of memory cases. The skb is freed on error. 1195 */ 1196 static inline int skb_pad(struct sk_buff *skb, int pad) 1197 { 1198 return __skb_pad(skb, pad, true); 1199 } 1200 #define dev_kfree_skb(a) consume_skb(a) 1201 1202 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1203 int offset, size_t size); 1204 1205 struct skb_seq_state { 1206 __u32 lower_offset; 1207 __u32 upper_offset; 1208 __u32 frag_idx; 1209 __u32 stepped_offset; 1210 struct sk_buff *root_skb; 1211 struct sk_buff *cur_skb; 1212 __u8 *frag_data; 1213 __u32 frag_off; 1214 }; 1215 1216 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1217 unsigned int to, struct skb_seq_state *st); 1218 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1219 struct skb_seq_state *st); 1220 void skb_abort_seq_read(struct skb_seq_state *st); 1221 1222 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1223 unsigned int to, struct ts_config *config); 1224 1225 /* 1226 * Packet hash types specify the type of hash in skb_set_hash. 1227 * 1228 * Hash types refer to the protocol layer addresses which are used to 1229 * construct a packet's hash. The hashes are used to differentiate or identify 1230 * flows of the protocol layer for the hash type. Hash types are either 1231 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1232 * 1233 * Properties of hashes: 1234 * 1235 * 1) Two packets in different flows have different hash values 1236 * 2) Two packets in the same flow should have the same hash value 1237 * 1238 * A hash at a higher layer is considered to be more specific. A driver should 1239 * set the most specific hash possible. 1240 * 1241 * A driver cannot indicate a more specific hash than the layer at which a hash 1242 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1243 * 1244 * A driver may indicate a hash level which is less specific than the 1245 * actual layer the hash was computed on. For instance, a hash computed 1246 * at L4 may be considered an L3 hash. This should only be done if the 1247 * driver can't unambiguously determine that the HW computed the hash at 1248 * the higher layer. Note that the "should" in the second property above 1249 * permits this. 1250 */ 1251 enum pkt_hash_types { 1252 PKT_HASH_TYPE_NONE, /* Undefined type */ 1253 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1254 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1255 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1256 }; 1257 1258 static inline void skb_clear_hash(struct sk_buff *skb) 1259 { 1260 skb->hash = 0; 1261 skb->sw_hash = 0; 1262 skb->l4_hash = 0; 1263 } 1264 1265 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1266 { 1267 if (!skb->l4_hash) 1268 skb_clear_hash(skb); 1269 } 1270 1271 static inline void 1272 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1273 { 1274 skb->l4_hash = is_l4; 1275 skb->sw_hash = is_sw; 1276 skb->hash = hash; 1277 } 1278 1279 static inline void 1280 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1281 { 1282 /* Used by drivers to set hash from HW */ 1283 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1284 } 1285 1286 static inline void 1287 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1288 { 1289 __skb_set_hash(skb, hash, true, is_l4); 1290 } 1291 1292 void __skb_get_hash(struct sk_buff *skb); 1293 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1294 u32 skb_get_poff(const struct sk_buff *skb); 1295 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1296 const struct flow_keys_basic *keys, int hlen); 1297 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1298 const void *data, int hlen_proto); 1299 1300 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1301 int thoff, u8 ip_proto) 1302 { 1303 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1304 } 1305 1306 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1307 const struct flow_dissector_key *key, 1308 unsigned int key_count); 1309 1310 struct bpf_flow_dissector; 1311 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1312 __be16 proto, int nhoff, int hlen, unsigned int flags); 1313 1314 bool __skb_flow_dissect(const struct net *net, 1315 const struct sk_buff *skb, 1316 struct flow_dissector *flow_dissector, 1317 void *target_container, const void *data, 1318 __be16 proto, int nhoff, int hlen, unsigned int flags); 1319 1320 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1321 struct flow_dissector *flow_dissector, 1322 void *target_container, unsigned int flags) 1323 { 1324 return __skb_flow_dissect(NULL, skb, flow_dissector, 1325 target_container, NULL, 0, 0, 0, flags); 1326 } 1327 1328 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1329 struct flow_keys *flow, 1330 unsigned int flags) 1331 { 1332 memset(flow, 0, sizeof(*flow)); 1333 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1334 flow, NULL, 0, 0, 0, flags); 1335 } 1336 1337 static inline bool 1338 skb_flow_dissect_flow_keys_basic(const struct net *net, 1339 const struct sk_buff *skb, 1340 struct flow_keys_basic *flow, 1341 const void *data, __be16 proto, 1342 int nhoff, int hlen, unsigned int flags) 1343 { 1344 memset(flow, 0, sizeof(*flow)); 1345 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1346 data, proto, nhoff, hlen, flags); 1347 } 1348 1349 void skb_flow_dissect_meta(const struct sk_buff *skb, 1350 struct flow_dissector *flow_dissector, 1351 void *target_container); 1352 1353 /* Gets a skb connection tracking info, ctinfo map should be a 1354 * map of mapsize to translate enum ip_conntrack_info states 1355 * to user states. 1356 */ 1357 void 1358 skb_flow_dissect_ct(const struct sk_buff *skb, 1359 struct flow_dissector *flow_dissector, 1360 void *target_container, 1361 u16 *ctinfo_map, size_t mapsize, 1362 bool post_ct); 1363 void 1364 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1365 struct flow_dissector *flow_dissector, 1366 void *target_container); 1367 1368 void skb_flow_dissect_hash(const struct sk_buff *skb, 1369 struct flow_dissector *flow_dissector, 1370 void *target_container); 1371 1372 static inline __u32 skb_get_hash(struct sk_buff *skb) 1373 { 1374 if (!skb->l4_hash && !skb->sw_hash) 1375 __skb_get_hash(skb); 1376 1377 return skb->hash; 1378 } 1379 1380 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1381 { 1382 if (!skb->l4_hash && !skb->sw_hash) { 1383 struct flow_keys keys; 1384 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1385 1386 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1387 } 1388 1389 return skb->hash; 1390 } 1391 1392 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1393 const siphash_key_t *perturb); 1394 1395 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1396 { 1397 return skb->hash; 1398 } 1399 1400 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1401 { 1402 to->hash = from->hash; 1403 to->sw_hash = from->sw_hash; 1404 to->l4_hash = from->l4_hash; 1405 }; 1406 1407 static inline void skb_copy_decrypted(struct sk_buff *to, 1408 const struct sk_buff *from) 1409 { 1410 #ifdef CONFIG_TLS_DEVICE 1411 to->decrypted = from->decrypted; 1412 #endif 1413 } 1414 1415 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1416 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1417 { 1418 return skb->head + skb->end; 1419 } 1420 1421 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1422 { 1423 return skb->end; 1424 } 1425 #else 1426 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1427 { 1428 return skb->end; 1429 } 1430 1431 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1432 { 1433 return skb->end - skb->head; 1434 } 1435 #endif 1436 1437 /* Internal */ 1438 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1439 1440 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1441 { 1442 return &skb_shinfo(skb)->hwtstamps; 1443 } 1444 1445 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1446 { 1447 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1448 1449 return is_zcopy ? skb_uarg(skb) : NULL; 1450 } 1451 1452 static inline void net_zcopy_get(struct ubuf_info *uarg) 1453 { 1454 refcount_inc(&uarg->refcnt); 1455 } 1456 1457 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1458 { 1459 skb_shinfo(skb)->destructor_arg = uarg; 1460 skb_shinfo(skb)->flags |= uarg->flags; 1461 } 1462 1463 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1464 bool *have_ref) 1465 { 1466 if (skb && uarg && !skb_zcopy(skb)) { 1467 if (unlikely(have_ref && *have_ref)) 1468 *have_ref = false; 1469 else 1470 net_zcopy_get(uarg); 1471 skb_zcopy_init(skb, uarg); 1472 } 1473 } 1474 1475 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1476 { 1477 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1478 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1479 } 1480 1481 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1482 { 1483 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1484 } 1485 1486 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1487 { 1488 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1489 } 1490 1491 static inline void net_zcopy_put(struct ubuf_info *uarg) 1492 { 1493 if (uarg) 1494 uarg->callback(NULL, uarg, true); 1495 } 1496 1497 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1498 { 1499 if (uarg) { 1500 if (uarg->callback == msg_zerocopy_callback) 1501 msg_zerocopy_put_abort(uarg, have_uref); 1502 else if (have_uref) 1503 net_zcopy_put(uarg); 1504 } 1505 } 1506 1507 /* Release a reference on a zerocopy structure */ 1508 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1509 { 1510 struct ubuf_info *uarg = skb_zcopy(skb); 1511 1512 if (uarg) { 1513 if (!skb_zcopy_is_nouarg(skb)) 1514 uarg->callback(skb, uarg, zerocopy_success); 1515 1516 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG; 1517 } 1518 } 1519 1520 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1521 { 1522 skb->next = NULL; 1523 } 1524 1525 /* Iterate through singly-linked GSO fragments of an skb. */ 1526 #define skb_list_walk_safe(first, skb, next_skb) \ 1527 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1528 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1529 1530 static inline void skb_list_del_init(struct sk_buff *skb) 1531 { 1532 __list_del_entry(&skb->list); 1533 skb_mark_not_on_list(skb); 1534 } 1535 1536 /** 1537 * skb_queue_empty - check if a queue is empty 1538 * @list: queue head 1539 * 1540 * Returns true if the queue is empty, false otherwise. 1541 */ 1542 static inline int skb_queue_empty(const struct sk_buff_head *list) 1543 { 1544 return list->next == (const struct sk_buff *) list; 1545 } 1546 1547 /** 1548 * skb_queue_empty_lockless - check if a queue is empty 1549 * @list: queue head 1550 * 1551 * Returns true if the queue is empty, false otherwise. 1552 * This variant can be used in lockless contexts. 1553 */ 1554 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1555 { 1556 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1557 } 1558 1559 1560 /** 1561 * skb_queue_is_last - check if skb is the last entry in the queue 1562 * @list: queue head 1563 * @skb: buffer 1564 * 1565 * Returns true if @skb is the last buffer on the list. 1566 */ 1567 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1568 const struct sk_buff *skb) 1569 { 1570 return skb->next == (const struct sk_buff *) list; 1571 } 1572 1573 /** 1574 * skb_queue_is_first - check if skb is the first entry in the queue 1575 * @list: queue head 1576 * @skb: buffer 1577 * 1578 * Returns true if @skb is the first buffer on the list. 1579 */ 1580 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1581 const struct sk_buff *skb) 1582 { 1583 return skb->prev == (const struct sk_buff *) list; 1584 } 1585 1586 /** 1587 * skb_queue_next - return the next packet in the queue 1588 * @list: queue head 1589 * @skb: current buffer 1590 * 1591 * Return the next packet in @list after @skb. It is only valid to 1592 * call this if skb_queue_is_last() evaluates to false. 1593 */ 1594 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1595 const struct sk_buff *skb) 1596 { 1597 /* This BUG_ON may seem severe, but if we just return then we 1598 * are going to dereference garbage. 1599 */ 1600 BUG_ON(skb_queue_is_last(list, skb)); 1601 return skb->next; 1602 } 1603 1604 /** 1605 * skb_queue_prev - return the prev packet in the queue 1606 * @list: queue head 1607 * @skb: current buffer 1608 * 1609 * Return the prev packet in @list before @skb. It is only valid to 1610 * call this if skb_queue_is_first() evaluates to false. 1611 */ 1612 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1613 const struct sk_buff *skb) 1614 { 1615 /* This BUG_ON may seem severe, but if we just return then we 1616 * are going to dereference garbage. 1617 */ 1618 BUG_ON(skb_queue_is_first(list, skb)); 1619 return skb->prev; 1620 } 1621 1622 /** 1623 * skb_get - reference buffer 1624 * @skb: buffer to reference 1625 * 1626 * Makes another reference to a socket buffer and returns a pointer 1627 * to the buffer. 1628 */ 1629 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1630 { 1631 refcount_inc(&skb->users); 1632 return skb; 1633 } 1634 1635 /* 1636 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1637 */ 1638 1639 /** 1640 * skb_cloned - is the buffer a clone 1641 * @skb: buffer to check 1642 * 1643 * Returns true if the buffer was generated with skb_clone() and is 1644 * one of multiple shared copies of the buffer. Cloned buffers are 1645 * shared data so must not be written to under normal circumstances. 1646 */ 1647 static inline int skb_cloned(const struct sk_buff *skb) 1648 { 1649 return skb->cloned && 1650 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1651 } 1652 1653 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1654 { 1655 might_sleep_if(gfpflags_allow_blocking(pri)); 1656 1657 if (skb_cloned(skb)) 1658 return pskb_expand_head(skb, 0, 0, pri); 1659 1660 return 0; 1661 } 1662 1663 /** 1664 * skb_header_cloned - is the header a clone 1665 * @skb: buffer to check 1666 * 1667 * Returns true if modifying the header part of the buffer requires 1668 * the data to be copied. 1669 */ 1670 static inline int skb_header_cloned(const struct sk_buff *skb) 1671 { 1672 int dataref; 1673 1674 if (!skb->cloned) 1675 return 0; 1676 1677 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1678 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1679 return dataref != 1; 1680 } 1681 1682 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1683 { 1684 might_sleep_if(gfpflags_allow_blocking(pri)); 1685 1686 if (skb_header_cloned(skb)) 1687 return pskb_expand_head(skb, 0, 0, pri); 1688 1689 return 0; 1690 } 1691 1692 /** 1693 * __skb_header_release - release reference to header 1694 * @skb: buffer to operate on 1695 */ 1696 static inline void __skb_header_release(struct sk_buff *skb) 1697 { 1698 skb->nohdr = 1; 1699 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1700 } 1701 1702 1703 /** 1704 * skb_shared - is the buffer shared 1705 * @skb: buffer to check 1706 * 1707 * Returns true if more than one person has a reference to this 1708 * buffer. 1709 */ 1710 static inline int skb_shared(const struct sk_buff *skb) 1711 { 1712 return refcount_read(&skb->users) != 1; 1713 } 1714 1715 /** 1716 * skb_share_check - check if buffer is shared and if so clone it 1717 * @skb: buffer to check 1718 * @pri: priority for memory allocation 1719 * 1720 * If the buffer is shared the buffer is cloned and the old copy 1721 * drops a reference. A new clone with a single reference is returned. 1722 * If the buffer is not shared the original buffer is returned. When 1723 * being called from interrupt status or with spinlocks held pri must 1724 * be GFP_ATOMIC. 1725 * 1726 * NULL is returned on a memory allocation failure. 1727 */ 1728 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1729 { 1730 might_sleep_if(gfpflags_allow_blocking(pri)); 1731 if (skb_shared(skb)) { 1732 struct sk_buff *nskb = skb_clone(skb, pri); 1733 1734 if (likely(nskb)) 1735 consume_skb(skb); 1736 else 1737 kfree_skb(skb); 1738 skb = nskb; 1739 } 1740 return skb; 1741 } 1742 1743 /* 1744 * Copy shared buffers into a new sk_buff. We effectively do COW on 1745 * packets to handle cases where we have a local reader and forward 1746 * and a couple of other messy ones. The normal one is tcpdumping 1747 * a packet thats being forwarded. 1748 */ 1749 1750 /** 1751 * skb_unshare - make a copy of a shared buffer 1752 * @skb: buffer to check 1753 * @pri: priority for memory allocation 1754 * 1755 * If the socket buffer is a clone then this function creates a new 1756 * copy of the data, drops a reference count on the old copy and returns 1757 * the new copy with the reference count at 1. If the buffer is not a clone 1758 * the original buffer is returned. When called with a spinlock held or 1759 * from interrupt state @pri must be %GFP_ATOMIC 1760 * 1761 * %NULL is returned on a memory allocation failure. 1762 */ 1763 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1764 gfp_t pri) 1765 { 1766 might_sleep_if(gfpflags_allow_blocking(pri)); 1767 if (skb_cloned(skb)) { 1768 struct sk_buff *nskb = skb_copy(skb, pri); 1769 1770 /* Free our shared copy */ 1771 if (likely(nskb)) 1772 consume_skb(skb); 1773 else 1774 kfree_skb(skb); 1775 skb = nskb; 1776 } 1777 return skb; 1778 } 1779 1780 /** 1781 * skb_peek - peek at the head of an &sk_buff_head 1782 * @list_: list to peek at 1783 * 1784 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1785 * be careful with this one. A peek leaves the buffer on the 1786 * list and someone else may run off with it. You must hold 1787 * the appropriate locks or have a private queue to do this. 1788 * 1789 * Returns %NULL for an empty list or a pointer to the head element. 1790 * The reference count is not incremented and the reference is therefore 1791 * volatile. Use with caution. 1792 */ 1793 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1794 { 1795 struct sk_buff *skb = list_->next; 1796 1797 if (skb == (struct sk_buff *)list_) 1798 skb = NULL; 1799 return skb; 1800 } 1801 1802 /** 1803 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1804 * @list_: list to peek at 1805 * 1806 * Like skb_peek(), but the caller knows that the list is not empty. 1807 */ 1808 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1809 { 1810 return list_->next; 1811 } 1812 1813 /** 1814 * skb_peek_next - peek skb following the given one from a queue 1815 * @skb: skb to start from 1816 * @list_: list to peek at 1817 * 1818 * Returns %NULL when the end of the list is met or a pointer to the 1819 * next element. The reference count is not incremented and the 1820 * reference is therefore volatile. Use with caution. 1821 */ 1822 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1823 const struct sk_buff_head *list_) 1824 { 1825 struct sk_buff *next = skb->next; 1826 1827 if (next == (struct sk_buff *)list_) 1828 next = NULL; 1829 return next; 1830 } 1831 1832 /** 1833 * skb_peek_tail - peek at the tail of an &sk_buff_head 1834 * @list_: list to peek at 1835 * 1836 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1837 * be careful with this one. A peek leaves the buffer on the 1838 * list and someone else may run off with it. You must hold 1839 * the appropriate locks or have a private queue to do this. 1840 * 1841 * Returns %NULL for an empty list or a pointer to the tail element. 1842 * The reference count is not incremented and the reference is therefore 1843 * volatile. Use with caution. 1844 */ 1845 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1846 { 1847 struct sk_buff *skb = READ_ONCE(list_->prev); 1848 1849 if (skb == (struct sk_buff *)list_) 1850 skb = NULL; 1851 return skb; 1852 1853 } 1854 1855 /** 1856 * skb_queue_len - get queue length 1857 * @list_: list to measure 1858 * 1859 * Return the length of an &sk_buff queue. 1860 */ 1861 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1862 { 1863 return list_->qlen; 1864 } 1865 1866 /** 1867 * skb_queue_len_lockless - get queue length 1868 * @list_: list to measure 1869 * 1870 * Return the length of an &sk_buff queue. 1871 * This variant can be used in lockless contexts. 1872 */ 1873 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1874 { 1875 return READ_ONCE(list_->qlen); 1876 } 1877 1878 /** 1879 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1880 * @list: queue to initialize 1881 * 1882 * This initializes only the list and queue length aspects of 1883 * an sk_buff_head object. This allows to initialize the list 1884 * aspects of an sk_buff_head without reinitializing things like 1885 * the spinlock. It can also be used for on-stack sk_buff_head 1886 * objects where the spinlock is known to not be used. 1887 */ 1888 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1889 { 1890 list->prev = list->next = (struct sk_buff *)list; 1891 list->qlen = 0; 1892 } 1893 1894 /* 1895 * This function creates a split out lock class for each invocation; 1896 * this is needed for now since a whole lot of users of the skb-queue 1897 * infrastructure in drivers have different locking usage (in hardirq) 1898 * than the networking core (in softirq only). In the long run either the 1899 * network layer or drivers should need annotation to consolidate the 1900 * main types of usage into 3 classes. 1901 */ 1902 static inline void skb_queue_head_init(struct sk_buff_head *list) 1903 { 1904 spin_lock_init(&list->lock); 1905 __skb_queue_head_init(list); 1906 } 1907 1908 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1909 struct lock_class_key *class) 1910 { 1911 skb_queue_head_init(list); 1912 lockdep_set_class(&list->lock, class); 1913 } 1914 1915 /* 1916 * Insert an sk_buff on a list. 1917 * 1918 * The "__skb_xxxx()" functions are the non-atomic ones that 1919 * can only be called with interrupts disabled. 1920 */ 1921 static inline void __skb_insert(struct sk_buff *newsk, 1922 struct sk_buff *prev, struct sk_buff *next, 1923 struct sk_buff_head *list) 1924 { 1925 /* See skb_queue_empty_lockless() and skb_peek_tail() 1926 * for the opposite READ_ONCE() 1927 */ 1928 WRITE_ONCE(newsk->next, next); 1929 WRITE_ONCE(newsk->prev, prev); 1930 WRITE_ONCE(next->prev, newsk); 1931 WRITE_ONCE(prev->next, newsk); 1932 list->qlen++; 1933 } 1934 1935 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1936 struct sk_buff *prev, 1937 struct sk_buff *next) 1938 { 1939 struct sk_buff *first = list->next; 1940 struct sk_buff *last = list->prev; 1941 1942 WRITE_ONCE(first->prev, prev); 1943 WRITE_ONCE(prev->next, first); 1944 1945 WRITE_ONCE(last->next, next); 1946 WRITE_ONCE(next->prev, last); 1947 } 1948 1949 /** 1950 * skb_queue_splice - join two skb lists, this is designed for stacks 1951 * @list: the new list to add 1952 * @head: the place to add it in the first list 1953 */ 1954 static inline void skb_queue_splice(const struct sk_buff_head *list, 1955 struct sk_buff_head *head) 1956 { 1957 if (!skb_queue_empty(list)) { 1958 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1959 head->qlen += list->qlen; 1960 } 1961 } 1962 1963 /** 1964 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1965 * @list: the new list to add 1966 * @head: the place to add it in the first list 1967 * 1968 * The list at @list is reinitialised 1969 */ 1970 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1971 struct sk_buff_head *head) 1972 { 1973 if (!skb_queue_empty(list)) { 1974 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1975 head->qlen += list->qlen; 1976 __skb_queue_head_init(list); 1977 } 1978 } 1979 1980 /** 1981 * skb_queue_splice_tail - join two skb lists, each list being a queue 1982 * @list: the new list to add 1983 * @head: the place to add it in the first list 1984 */ 1985 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1986 struct sk_buff_head *head) 1987 { 1988 if (!skb_queue_empty(list)) { 1989 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1990 head->qlen += list->qlen; 1991 } 1992 } 1993 1994 /** 1995 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1996 * @list: the new list to add 1997 * @head: the place to add it in the first list 1998 * 1999 * Each of the lists is a queue. 2000 * The list at @list is reinitialised 2001 */ 2002 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2003 struct sk_buff_head *head) 2004 { 2005 if (!skb_queue_empty(list)) { 2006 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2007 head->qlen += list->qlen; 2008 __skb_queue_head_init(list); 2009 } 2010 } 2011 2012 /** 2013 * __skb_queue_after - queue a buffer at the list head 2014 * @list: list to use 2015 * @prev: place after this buffer 2016 * @newsk: buffer to queue 2017 * 2018 * Queue a buffer int the middle of a list. This function takes no locks 2019 * and you must therefore hold required locks before calling it. 2020 * 2021 * A buffer cannot be placed on two lists at the same time. 2022 */ 2023 static inline void __skb_queue_after(struct sk_buff_head *list, 2024 struct sk_buff *prev, 2025 struct sk_buff *newsk) 2026 { 2027 __skb_insert(newsk, prev, prev->next, list); 2028 } 2029 2030 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2031 struct sk_buff_head *list); 2032 2033 static inline void __skb_queue_before(struct sk_buff_head *list, 2034 struct sk_buff *next, 2035 struct sk_buff *newsk) 2036 { 2037 __skb_insert(newsk, next->prev, next, list); 2038 } 2039 2040 /** 2041 * __skb_queue_head - queue a buffer at the list head 2042 * @list: list to use 2043 * @newsk: buffer to queue 2044 * 2045 * Queue a buffer at the start of a list. This function takes no locks 2046 * and you must therefore hold required locks before calling it. 2047 * 2048 * A buffer cannot be placed on two lists at the same time. 2049 */ 2050 static inline void __skb_queue_head(struct sk_buff_head *list, 2051 struct sk_buff *newsk) 2052 { 2053 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2054 } 2055 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2056 2057 /** 2058 * __skb_queue_tail - queue a buffer at the list tail 2059 * @list: list to use 2060 * @newsk: buffer to queue 2061 * 2062 * Queue a buffer at the end of a list. This function takes no locks 2063 * and you must therefore hold required locks before calling it. 2064 * 2065 * A buffer cannot be placed on two lists at the same time. 2066 */ 2067 static inline void __skb_queue_tail(struct sk_buff_head *list, 2068 struct sk_buff *newsk) 2069 { 2070 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2071 } 2072 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2073 2074 /* 2075 * remove sk_buff from list. _Must_ be called atomically, and with 2076 * the list known.. 2077 */ 2078 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2079 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2080 { 2081 struct sk_buff *next, *prev; 2082 2083 WRITE_ONCE(list->qlen, list->qlen - 1); 2084 next = skb->next; 2085 prev = skb->prev; 2086 skb->next = skb->prev = NULL; 2087 WRITE_ONCE(next->prev, prev); 2088 WRITE_ONCE(prev->next, next); 2089 } 2090 2091 /** 2092 * __skb_dequeue - remove from the head of the queue 2093 * @list: list to dequeue from 2094 * 2095 * Remove the head of the list. This function does not take any locks 2096 * so must be used with appropriate locks held only. The head item is 2097 * returned or %NULL if the list is empty. 2098 */ 2099 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2100 { 2101 struct sk_buff *skb = skb_peek(list); 2102 if (skb) 2103 __skb_unlink(skb, list); 2104 return skb; 2105 } 2106 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2107 2108 /** 2109 * __skb_dequeue_tail - remove from the tail of the queue 2110 * @list: list to dequeue from 2111 * 2112 * Remove the tail of the list. This function does not take any locks 2113 * so must be used with appropriate locks held only. The tail item is 2114 * returned or %NULL if the list is empty. 2115 */ 2116 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2117 { 2118 struct sk_buff *skb = skb_peek_tail(list); 2119 if (skb) 2120 __skb_unlink(skb, list); 2121 return skb; 2122 } 2123 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2124 2125 2126 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2127 { 2128 return skb->data_len; 2129 } 2130 2131 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2132 { 2133 return skb->len - skb->data_len; 2134 } 2135 2136 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2137 { 2138 unsigned int i, len = 0; 2139 2140 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2141 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2142 return len; 2143 } 2144 2145 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2146 { 2147 return skb_headlen(skb) + __skb_pagelen(skb); 2148 } 2149 2150 /** 2151 * __skb_fill_page_desc - initialise a paged fragment in an skb 2152 * @skb: buffer containing fragment to be initialised 2153 * @i: paged fragment index to initialise 2154 * @page: the page to use for this fragment 2155 * @off: the offset to the data with @page 2156 * @size: the length of the data 2157 * 2158 * Initialises the @i'th fragment of @skb to point to &size bytes at 2159 * offset @off within @page. 2160 * 2161 * Does not take any additional reference on the fragment. 2162 */ 2163 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2164 struct page *page, int off, int size) 2165 { 2166 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2167 2168 /* 2169 * Propagate page pfmemalloc to the skb if we can. The problem is 2170 * that not all callers have unique ownership of the page but rely 2171 * on page_is_pfmemalloc doing the right thing(tm). 2172 */ 2173 frag->bv_page = page; 2174 frag->bv_offset = off; 2175 skb_frag_size_set(frag, size); 2176 2177 page = compound_head(page); 2178 if (page_is_pfmemalloc(page)) 2179 skb->pfmemalloc = true; 2180 } 2181 2182 /** 2183 * skb_fill_page_desc - initialise a paged fragment in an skb 2184 * @skb: buffer containing fragment to be initialised 2185 * @i: paged fragment index to initialise 2186 * @page: the page to use for this fragment 2187 * @off: the offset to the data with @page 2188 * @size: the length of the data 2189 * 2190 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2191 * @skb to point to @size bytes at offset @off within @page. In 2192 * addition updates @skb such that @i is the last fragment. 2193 * 2194 * Does not take any additional reference on the fragment. 2195 */ 2196 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2197 struct page *page, int off, int size) 2198 { 2199 __skb_fill_page_desc(skb, i, page, off, size); 2200 skb_shinfo(skb)->nr_frags = i + 1; 2201 } 2202 2203 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2204 int size, unsigned int truesize); 2205 2206 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2207 unsigned int truesize); 2208 2209 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2210 2211 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2212 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2213 { 2214 return skb->head + skb->tail; 2215 } 2216 2217 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2218 { 2219 skb->tail = skb->data - skb->head; 2220 } 2221 2222 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2223 { 2224 skb_reset_tail_pointer(skb); 2225 skb->tail += offset; 2226 } 2227 2228 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2229 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2230 { 2231 return skb->tail; 2232 } 2233 2234 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2235 { 2236 skb->tail = skb->data; 2237 } 2238 2239 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2240 { 2241 skb->tail = skb->data + offset; 2242 } 2243 2244 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2245 2246 /* 2247 * Add data to an sk_buff 2248 */ 2249 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2250 void *skb_put(struct sk_buff *skb, unsigned int len); 2251 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2252 { 2253 void *tmp = skb_tail_pointer(skb); 2254 SKB_LINEAR_ASSERT(skb); 2255 skb->tail += len; 2256 skb->len += len; 2257 return tmp; 2258 } 2259 2260 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2261 { 2262 void *tmp = __skb_put(skb, len); 2263 2264 memset(tmp, 0, len); 2265 return tmp; 2266 } 2267 2268 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2269 unsigned int len) 2270 { 2271 void *tmp = __skb_put(skb, len); 2272 2273 memcpy(tmp, data, len); 2274 return tmp; 2275 } 2276 2277 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2278 { 2279 *(u8 *)__skb_put(skb, 1) = val; 2280 } 2281 2282 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2283 { 2284 void *tmp = skb_put(skb, len); 2285 2286 memset(tmp, 0, len); 2287 2288 return tmp; 2289 } 2290 2291 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2292 unsigned int len) 2293 { 2294 void *tmp = skb_put(skb, len); 2295 2296 memcpy(tmp, data, len); 2297 2298 return tmp; 2299 } 2300 2301 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2302 { 2303 *(u8 *)skb_put(skb, 1) = val; 2304 } 2305 2306 void *skb_push(struct sk_buff *skb, unsigned int len); 2307 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2308 { 2309 skb->data -= len; 2310 skb->len += len; 2311 return skb->data; 2312 } 2313 2314 void *skb_pull(struct sk_buff *skb, unsigned int len); 2315 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2316 { 2317 skb->len -= len; 2318 BUG_ON(skb->len < skb->data_len); 2319 return skb->data += len; 2320 } 2321 2322 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2323 { 2324 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2325 } 2326 2327 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2328 2329 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2330 { 2331 if (len > skb_headlen(skb) && 2332 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2333 return NULL; 2334 skb->len -= len; 2335 return skb->data += len; 2336 } 2337 2338 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2339 { 2340 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2341 } 2342 2343 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2344 { 2345 if (likely(len <= skb_headlen(skb))) 2346 return true; 2347 if (unlikely(len > skb->len)) 2348 return false; 2349 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2350 } 2351 2352 void skb_condense(struct sk_buff *skb); 2353 2354 /** 2355 * skb_headroom - bytes at buffer head 2356 * @skb: buffer to check 2357 * 2358 * Return the number of bytes of free space at the head of an &sk_buff. 2359 */ 2360 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2361 { 2362 return skb->data - skb->head; 2363 } 2364 2365 /** 2366 * skb_tailroom - bytes at buffer end 2367 * @skb: buffer to check 2368 * 2369 * Return the number of bytes of free space at the tail of an sk_buff 2370 */ 2371 static inline int skb_tailroom(const struct sk_buff *skb) 2372 { 2373 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2374 } 2375 2376 /** 2377 * skb_availroom - bytes at buffer end 2378 * @skb: buffer to check 2379 * 2380 * Return the number of bytes of free space at the tail of an sk_buff 2381 * allocated by sk_stream_alloc() 2382 */ 2383 static inline int skb_availroom(const struct sk_buff *skb) 2384 { 2385 if (skb_is_nonlinear(skb)) 2386 return 0; 2387 2388 return skb->end - skb->tail - skb->reserved_tailroom; 2389 } 2390 2391 /** 2392 * skb_reserve - adjust headroom 2393 * @skb: buffer to alter 2394 * @len: bytes to move 2395 * 2396 * Increase the headroom of an empty &sk_buff by reducing the tail 2397 * room. This is only allowed for an empty buffer. 2398 */ 2399 static inline void skb_reserve(struct sk_buff *skb, int len) 2400 { 2401 skb->data += len; 2402 skb->tail += len; 2403 } 2404 2405 /** 2406 * skb_tailroom_reserve - adjust reserved_tailroom 2407 * @skb: buffer to alter 2408 * @mtu: maximum amount of headlen permitted 2409 * @needed_tailroom: minimum amount of reserved_tailroom 2410 * 2411 * Set reserved_tailroom so that headlen can be as large as possible but 2412 * not larger than mtu and tailroom cannot be smaller than 2413 * needed_tailroom. 2414 * The required headroom should already have been reserved before using 2415 * this function. 2416 */ 2417 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2418 unsigned int needed_tailroom) 2419 { 2420 SKB_LINEAR_ASSERT(skb); 2421 if (mtu < skb_tailroom(skb) - needed_tailroom) 2422 /* use at most mtu */ 2423 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2424 else 2425 /* use up to all available space */ 2426 skb->reserved_tailroom = needed_tailroom; 2427 } 2428 2429 #define ENCAP_TYPE_ETHER 0 2430 #define ENCAP_TYPE_IPPROTO 1 2431 2432 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2433 __be16 protocol) 2434 { 2435 skb->inner_protocol = protocol; 2436 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2437 } 2438 2439 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2440 __u8 ipproto) 2441 { 2442 skb->inner_ipproto = ipproto; 2443 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2444 } 2445 2446 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2447 { 2448 skb->inner_mac_header = skb->mac_header; 2449 skb->inner_network_header = skb->network_header; 2450 skb->inner_transport_header = skb->transport_header; 2451 } 2452 2453 static inline void skb_reset_mac_len(struct sk_buff *skb) 2454 { 2455 skb->mac_len = skb->network_header - skb->mac_header; 2456 } 2457 2458 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2459 *skb) 2460 { 2461 return skb->head + skb->inner_transport_header; 2462 } 2463 2464 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2465 { 2466 return skb_inner_transport_header(skb) - skb->data; 2467 } 2468 2469 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2470 { 2471 skb->inner_transport_header = skb->data - skb->head; 2472 } 2473 2474 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2475 const int offset) 2476 { 2477 skb_reset_inner_transport_header(skb); 2478 skb->inner_transport_header += offset; 2479 } 2480 2481 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2482 { 2483 return skb->head + skb->inner_network_header; 2484 } 2485 2486 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2487 { 2488 skb->inner_network_header = skb->data - skb->head; 2489 } 2490 2491 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2492 const int offset) 2493 { 2494 skb_reset_inner_network_header(skb); 2495 skb->inner_network_header += offset; 2496 } 2497 2498 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2499 { 2500 return skb->head + skb->inner_mac_header; 2501 } 2502 2503 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2504 { 2505 skb->inner_mac_header = skb->data - skb->head; 2506 } 2507 2508 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2509 const int offset) 2510 { 2511 skb_reset_inner_mac_header(skb); 2512 skb->inner_mac_header += offset; 2513 } 2514 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2515 { 2516 return skb->transport_header != (typeof(skb->transport_header))~0U; 2517 } 2518 2519 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2520 { 2521 return skb->head + skb->transport_header; 2522 } 2523 2524 static inline void skb_reset_transport_header(struct sk_buff *skb) 2525 { 2526 skb->transport_header = skb->data - skb->head; 2527 } 2528 2529 static inline void skb_set_transport_header(struct sk_buff *skb, 2530 const int offset) 2531 { 2532 skb_reset_transport_header(skb); 2533 skb->transport_header += offset; 2534 } 2535 2536 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2537 { 2538 return skb->head + skb->network_header; 2539 } 2540 2541 static inline void skb_reset_network_header(struct sk_buff *skb) 2542 { 2543 skb->network_header = skb->data - skb->head; 2544 } 2545 2546 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2547 { 2548 skb_reset_network_header(skb); 2549 skb->network_header += offset; 2550 } 2551 2552 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2553 { 2554 return skb->head + skb->mac_header; 2555 } 2556 2557 static inline int skb_mac_offset(const struct sk_buff *skb) 2558 { 2559 return skb_mac_header(skb) - skb->data; 2560 } 2561 2562 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2563 { 2564 return skb->network_header - skb->mac_header; 2565 } 2566 2567 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2568 { 2569 return skb->mac_header != (typeof(skb->mac_header))~0U; 2570 } 2571 2572 static inline void skb_unset_mac_header(struct sk_buff *skb) 2573 { 2574 skb->mac_header = (typeof(skb->mac_header))~0U; 2575 } 2576 2577 static inline void skb_reset_mac_header(struct sk_buff *skb) 2578 { 2579 skb->mac_header = skb->data - skb->head; 2580 } 2581 2582 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2583 { 2584 skb_reset_mac_header(skb); 2585 skb->mac_header += offset; 2586 } 2587 2588 static inline void skb_pop_mac_header(struct sk_buff *skb) 2589 { 2590 skb->mac_header = skb->network_header; 2591 } 2592 2593 static inline void skb_probe_transport_header(struct sk_buff *skb) 2594 { 2595 struct flow_keys_basic keys; 2596 2597 if (skb_transport_header_was_set(skb)) 2598 return; 2599 2600 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2601 NULL, 0, 0, 0, 0)) 2602 skb_set_transport_header(skb, keys.control.thoff); 2603 } 2604 2605 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2606 { 2607 if (skb_mac_header_was_set(skb)) { 2608 const unsigned char *old_mac = skb_mac_header(skb); 2609 2610 skb_set_mac_header(skb, -skb->mac_len); 2611 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2612 } 2613 } 2614 2615 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2616 { 2617 return skb->csum_start - skb_headroom(skb); 2618 } 2619 2620 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2621 { 2622 return skb->head + skb->csum_start; 2623 } 2624 2625 static inline int skb_transport_offset(const struct sk_buff *skb) 2626 { 2627 return skb_transport_header(skb) - skb->data; 2628 } 2629 2630 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2631 { 2632 return skb->transport_header - skb->network_header; 2633 } 2634 2635 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2636 { 2637 return skb->inner_transport_header - skb->inner_network_header; 2638 } 2639 2640 static inline int skb_network_offset(const struct sk_buff *skb) 2641 { 2642 return skb_network_header(skb) - skb->data; 2643 } 2644 2645 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2646 { 2647 return skb_inner_network_header(skb) - skb->data; 2648 } 2649 2650 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2651 { 2652 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2653 } 2654 2655 /* 2656 * CPUs often take a performance hit when accessing unaligned memory 2657 * locations. The actual performance hit varies, it can be small if the 2658 * hardware handles it or large if we have to take an exception and fix it 2659 * in software. 2660 * 2661 * Since an ethernet header is 14 bytes network drivers often end up with 2662 * the IP header at an unaligned offset. The IP header can be aligned by 2663 * shifting the start of the packet by 2 bytes. Drivers should do this 2664 * with: 2665 * 2666 * skb_reserve(skb, NET_IP_ALIGN); 2667 * 2668 * The downside to this alignment of the IP header is that the DMA is now 2669 * unaligned. On some architectures the cost of an unaligned DMA is high 2670 * and this cost outweighs the gains made by aligning the IP header. 2671 * 2672 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2673 * to be overridden. 2674 */ 2675 #ifndef NET_IP_ALIGN 2676 #define NET_IP_ALIGN 2 2677 #endif 2678 2679 /* 2680 * The networking layer reserves some headroom in skb data (via 2681 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2682 * the header has to grow. In the default case, if the header has to grow 2683 * 32 bytes or less we avoid the reallocation. 2684 * 2685 * Unfortunately this headroom changes the DMA alignment of the resulting 2686 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2687 * on some architectures. An architecture can override this value, 2688 * perhaps setting it to a cacheline in size (since that will maintain 2689 * cacheline alignment of the DMA). It must be a power of 2. 2690 * 2691 * Various parts of the networking layer expect at least 32 bytes of 2692 * headroom, you should not reduce this. 2693 * 2694 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2695 * to reduce average number of cache lines per packet. 2696 * get_rps_cpu() for example only access one 64 bytes aligned block : 2697 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2698 */ 2699 #ifndef NET_SKB_PAD 2700 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2701 #endif 2702 2703 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2704 2705 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2706 { 2707 if (WARN_ON(skb_is_nonlinear(skb))) 2708 return; 2709 skb->len = len; 2710 skb_set_tail_pointer(skb, len); 2711 } 2712 2713 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2714 { 2715 __skb_set_length(skb, len); 2716 } 2717 2718 void skb_trim(struct sk_buff *skb, unsigned int len); 2719 2720 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2721 { 2722 if (skb->data_len) 2723 return ___pskb_trim(skb, len); 2724 __skb_trim(skb, len); 2725 return 0; 2726 } 2727 2728 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2729 { 2730 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2731 } 2732 2733 /** 2734 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2735 * @skb: buffer to alter 2736 * @len: new length 2737 * 2738 * This is identical to pskb_trim except that the caller knows that 2739 * the skb is not cloned so we should never get an error due to out- 2740 * of-memory. 2741 */ 2742 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2743 { 2744 int err = pskb_trim(skb, len); 2745 BUG_ON(err); 2746 } 2747 2748 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2749 { 2750 unsigned int diff = len - skb->len; 2751 2752 if (skb_tailroom(skb) < diff) { 2753 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2754 GFP_ATOMIC); 2755 if (ret) 2756 return ret; 2757 } 2758 __skb_set_length(skb, len); 2759 return 0; 2760 } 2761 2762 /** 2763 * skb_orphan - orphan a buffer 2764 * @skb: buffer to orphan 2765 * 2766 * If a buffer currently has an owner then we call the owner's 2767 * destructor function and make the @skb unowned. The buffer continues 2768 * to exist but is no longer charged to its former owner. 2769 */ 2770 static inline void skb_orphan(struct sk_buff *skb) 2771 { 2772 if (skb->destructor) { 2773 skb->destructor(skb); 2774 skb->destructor = NULL; 2775 skb->sk = NULL; 2776 } else { 2777 BUG_ON(skb->sk); 2778 } 2779 } 2780 2781 /** 2782 * skb_orphan_frags - orphan the frags contained in a buffer 2783 * @skb: buffer to orphan frags from 2784 * @gfp_mask: allocation mask for replacement pages 2785 * 2786 * For each frag in the SKB which needs a destructor (i.e. has an 2787 * owner) create a copy of that frag and release the original 2788 * page by calling the destructor. 2789 */ 2790 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2791 { 2792 if (likely(!skb_zcopy(skb))) 2793 return 0; 2794 if (!skb_zcopy_is_nouarg(skb) && 2795 skb_uarg(skb)->callback == msg_zerocopy_callback) 2796 return 0; 2797 return skb_copy_ubufs(skb, gfp_mask); 2798 } 2799 2800 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2801 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2802 { 2803 if (likely(!skb_zcopy(skb))) 2804 return 0; 2805 return skb_copy_ubufs(skb, gfp_mask); 2806 } 2807 2808 /** 2809 * __skb_queue_purge - empty a list 2810 * @list: list to empty 2811 * 2812 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2813 * the list and one reference dropped. This function does not take the 2814 * list lock and the caller must hold the relevant locks to use it. 2815 */ 2816 static inline void __skb_queue_purge(struct sk_buff_head *list) 2817 { 2818 struct sk_buff *skb; 2819 while ((skb = __skb_dequeue(list)) != NULL) 2820 kfree_skb(skb); 2821 } 2822 void skb_queue_purge(struct sk_buff_head *list); 2823 2824 unsigned int skb_rbtree_purge(struct rb_root *root); 2825 2826 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2827 2828 /** 2829 * netdev_alloc_frag - allocate a page fragment 2830 * @fragsz: fragment size 2831 * 2832 * Allocates a frag from a page for receive buffer. 2833 * Uses GFP_ATOMIC allocations. 2834 */ 2835 static inline void *netdev_alloc_frag(unsigned int fragsz) 2836 { 2837 return __netdev_alloc_frag_align(fragsz, ~0u); 2838 } 2839 2840 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 2841 unsigned int align) 2842 { 2843 WARN_ON_ONCE(!is_power_of_2(align)); 2844 return __netdev_alloc_frag_align(fragsz, -align); 2845 } 2846 2847 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2848 gfp_t gfp_mask); 2849 2850 /** 2851 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2852 * @dev: network device to receive on 2853 * @length: length to allocate 2854 * 2855 * Allocate a new &sk_buff and assign it a usage count of one. The 2856 * buffer has unspecified headroom built in. Users should allocate 2857 * the headroom they think they need without accounting for the 2858 * built in space. The built in space is used for optimisations. 2859 * 2860 * %NULL is returned if there is no free memory. Although this function 2861 * allocates memory it can be called from an interrupt. 2862 */ 2863 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2864 unsigned int length) 2865 { 2866 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2867 } 2868 2869 /* legacy helper around __netdev_alloc_skb() */ 2870 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2871 gfp_t gfp_mask) 2872 { 2873 return __netdev_alloc_skb(NULL, length, gfp_mask); 2874 } 2875 2876 /* legacy helper around netdev_alloc_skb() */ 2877 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2878 { 2879 return netdev_alloc_skb(NULL, length); 2880 } 2881 2882 2883 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2884 unsigned int length, gfp_t gfp) 2885 { 2886 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2887 2888 if (NET_IP_ALIGN && skb) 2889 skb_reserve(skb, NET_IP_ALIGN); 2890 return skb; 2891 } 2892 2893 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2894 unsigned int length) 2895 { 2896 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2897 } 2898 2899 static inline void skb_free_frag(void *addr) 2900 { 2901 page_frag_free(addr); 2902 } 2903 2904 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2905 2906 static inline void *napi_alloc_frag(unsigned int fragsz) 2907 { 2908 return __napi_alloc_frag_align(fragsz, ~0u); 2909 } 2910 2911 static inline void *napi_alloc_frag_align(unsigned int fragsz, 2912 unsigned int align) 2913 { 2914 WARN_ON_ONCE(!is_power_of_2(align)); 2915 return __napi_alloc_frag_align(fragsz, -align); 2916 } 2917 2918 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2919 unsigned int length, gfp_t gfp_mask); 2920 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2921 unsigned int length) 2922 { 2923 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2924 } 2925 void napi_consume_skb(struct sk_buff *skb, int budget); 2926 2927 void napi_skb_free_stolen_head(struct sk_buff *skb); 2928 void __kfree_skb_defer(struct sk_buff *skb); 2929 2930 /** 2931 * __dev_alloc_pages - allocate page for network Rx 2932 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2933 * @order: size of the allocation 2934 * 2935 * Allocate a new page. 2936 * 2937 * %NULL is returned if there is no free memory. 2938 */ 2939 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2940 unsigned int order) 2941 { 2942 /* This piece of code contains several assumptions. 2943 * 1. This is for device Rx, therefor a cold page is preferred. 2944 * 2. The expectation is the user wants a compound page. 2945 * 3. If requesting a order 0 page it will not be compound 2946 * due to the check to see if order has a value in prep_new_page 2947 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2948 * code in gfp_to_alloc_flags that should be enforcing this. 2949 */ 2950 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2951 2952 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2953 } 2954 2955 static inline struct page *dev_alloc_pages(unsigned int order) 2956 { 2957 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2958 } 2959 2960 /** 2961 * __dev_alloc_page - allocate a page for network Rx 2962 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2963 * 2964 * Allocate a new page. 2965 * 2966 * %NULL is returned if there is no free memory. 2967 */ 2968 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2969 { 2970 return __dev_alloc_pages(gfp_mask, 0); 2971 } 2972 2973 static inline struct page *dev_alloc_page(void) 2974 { 2975 return dev_alloc_pages(0); 2976 } 2977 2978 /** 2979 * dev_page_is_reusable - check whether a page can be reused for network Rx 2980 * @page: the page to test 2981 * 2982 * A page shouldn't be considered for reusing/recycling if it was allocated 2983 * under memory pressure or at a distant memory node. 2984 * 2985 * Returns false if this page should be returned to page allocator, true 2986 * otherwise. 2987 */ 2988 static inline bool dev_page_is_reusable(const struct page *page) 2989 { 2990 return likely(page_to_nid(page) == numa_mem_id() && 2991 !page_is_pfmemalloc(page)); 2992 } 2993 2994 /** 2995 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2996 * @page: The page that was allocated from skb_alloc_page 2997 * @skb: The skb that may need pfmemalloc set 2998 */ 2999 static inline void skb_propagate_pfmemalloc(const struct page *page, 3000 struct sk_buff *skb) 3001 { 3002 if (page_is_pfmemalloc(page)) 3003 skb->pfmemalloc = true; 3004 } 3005 3006 /** 3007 * skb_frag_off() - Returns the offset of a skb fragment 3008 * @frag: the paged fragment 3009 */ 3010 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3011 { 3012 return frag->bv_offset; 3013 } 3014 3015 /** 3016 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3017 * @frag: skb fragment 3018 * @delta: value to add 3019 */ 3020 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3021 { 3022 frag->bv_offset += delta; 3023 } 3024 3025 /** 3026 * skb_frag_off_set() - Sets the offset of a skb fragment 3027 * @frag: skb fragment 3028 * @offset: offset of fragment 3029 */ 3030 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3031 { 3032 frag->bv_offset = offset; 3033 } 3034 3035 /** 3036 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3037 * @fragto: skb fragment where offset is set 3038 * @fragfrom: skb fragment offset is copied from 3039 */ 3040 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3041 const skb_frag_t *fragfrom) 3042 { 3043 fragto->bv_offset = fragfrom->bv_offset; 3044 } 3045 3046 /** 3047 * skb_frag_page - retrieve the page referred to by a paged fragment 3048 * @frag: the paged fragment 3049 * 3050 * Returns the &struct page associated with @frag. 3051 */ 3052 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3053 { 3054 return frag->bv_page; 3055 } 3056 3057 /** 3058 * __skb_frag_ref - take an addition reference on a paged fragment. 3059 * @frag: the paged fragment 3060 * 3061 * Takes an additional reference on the paged fragment @frag. 3062 */ 3063 static inline void __skb_frag_ref(skb_frag_t *frag) 3064 { 3065 get_page(skb_frag_page(frag)); 3066 } 3067 3068 /** 3069 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3070 * @skb: the buffer 3071 * @f: the fragment offset. 3072 * 3073 * Takes an additional reference on the @f'th paged fragment of @skb. 3074 */ 3075 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3076 { 3077 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3078 } 3079 3080 /** 3081 * __skb_frag_unref - release a reference on a paged fragment. 3082 * @frag: the paged fragment 3083 * 3084 * Releases a reference on the paged fragment @frag. 3085 */ 3086 static inline void __skb_frag_unref(skb_frag_t *frag) 3087 { 3088 put_page(skb_frag_page(frag)); 3089 } 3090 3091 /** 3092 * skb_frag_unref - release a reference on a paged fragment of an skb. 3093 * @skb: the buffer 3094 * @f: the fragment offset 3095 * 3096 * Releases a reference on the @f'th paged fragment of @skb. 3097 */ 3098 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3099 { 3100 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 3101 } 3102 3103 /** 3104 * skb_frag_address - gets the address of the data contained in a paged fragment 3105 * @frag: the paged fragment buffer 3106 * 3107 * Returns the address of the data within @frag. The page must already 3108 * be mapped. 3109 */ 3110 static inline void *skb_frag_address(const skb_frag_t *frag) 3111 { 3112 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3113 } 3114 3115 /** 3116 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3117 * @frag: the paged fragment buffer 3118 * 3119 * Returns the address of the data within @frag. Checks that the page 3120 * is mapped and returns %NULL otherwise. 3121 */ 3122 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3123 { 3124 void *ptr = page_address(skb_frag_page(frag)); 3125 if (unlikely(!ptr)) 3126 return NULL; 3127 3128 return ptr + skb_frag_off(frag); 3129 } 3130 3131 /** 3132 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3133 * @fragto: skb fragment where page is set 3134 * @fragfrom: skb fragment page is copied from 3135 */ 3136 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3137 const skb_frag_t *fragfrom) 3138 { 3139 fragto->bv_page = fragfrom->bv_page; 3140 } 3141 3142 /** 3143 * __skb_frag_set_page - sets the page contained in a paged fragment 3144 * @frag: the paged fragment 3145 * @page: the page to set 3146 * 3147 * Sets the fragment @frag to contain @page. 3148 */ 3149 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3150 { 3151 frag->bv_page = page; 3152 } 3153 3154 /** 3155 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3156 * @skb: the buffer 3157 * @f: the fragment offset 3158 * @page: the page to set 3159 * 3160 * Sets the @f'th fragment of @skb to contain @page. 3161 */ 3162 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3163 struct page *page) 3164 { 3165 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3166 } 3167 3168 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3169 3170 /** 3171 * skb_frag_dma_map - maps a paged fragment via the DMA API 3172 * @dev: the device to map the fragment to 3173 * @frag: the paged fragment to map 3174 * @offset: the offset within the fragment (starting at the 3175 * fragment's own offset) 3176 * @size: the number of bytes to map 3177 * @dir: the direction of the mapping (``PCI_DMA_*``) 3178 * 3179 * Maps the page associated with @frag to @device. 3180 */ 3181 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3182 const skb_frag_t *frag, 3183 size_t offset, size_t size, 3184 enum dma_data_direction dir) 3185 { 3186 return dma_map_page(dev, skb_frag_page(frag), 3187 skb_frag_off(frag) + offset, size, dir); 3188 } 3189 3190 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3191 gfp_t gfp_mask) 3192 { 3193 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3194 } 3195 3196 3197 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3198 gfp_t gfp_mask) 3199 { 3200 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3201 } 3202 3203 3204 /** 3205 * skb_clone_writable - is the header of a clone writable 3206 * @skb: buffer to check 3207 * @len: length up to which to write 3208 * 3209 * Returns true if modifying the header part of the cloned buffer 3210 * does not requires the data to be copied. 3211 */ 3212 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3213 { 3214 return !skb_header_cloned(skb) && 3215 skb_headroom(skb) + len <= skb->hdr_len; 3216 } 3217 3218 static inline int skb_try_make_writable(struct sk_buff *skb, 3219 unsigned int write_len) 3220 { 3221 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3222 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3223 } 3224 3225 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3226 int cloned) 3227 { 3228 int delta = 0; 3229 3230 if (headroom > skb_headroom(skb)) 3231 delta = headroom - skb_headroom(skb); 3232 3233 if (delta || cloned) 3234 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3235 GFP_ATOMIC); 3236 return 0; 3237 } 3238 3239 /** 3240 * skb_cow - copy header of skb when it is required 3241 * @skb: buffer to cow 3242 * @headroom: needed headroom 3243 * 3244 * If the skb passed lacks sufficient headroom or its data part 3245 * is shared, data is reallocated. If reallocation fails, an error 3246 * is returned and original skb is not changed. 3247 * 3248 * The result is skb with writable area skb->head...skb->tail 3249 * and at least @headroom of space at head. 3250 */ 3251 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3252 { 3253 return __skb_cow(skb, headroom, skb_cloned(skb)); 3254 } 3255 3256 /** 3257 * skb_cow_head - skb_cow but only making the head writable 3258 * @skb: buffer to cow 3259 * @headroom: needed headroom 3260 * 3261 * This function is identical to skb_cow except that we replace the 3262 * skb_cloned check by skb_header_cloned. It should be used when 3263 * you only need to push on some header and do not need to modify 3264 * the data. 3265 */ 3266 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3267 { 3268 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3269 } 3270 3271 /** 3272 * skb_padto - pad an skbuff up to a minimal size 3273 * @skb: buffer to pad 3274 * @len: minimal length 3275 * 3276 * Pads up a buffer to ensure the trailing bytes exist and are 3277 * blanked. If the buffer already contains sufficient data it 3278 * is untouched. Otherwise it is extended. Returns zero on 3279 * success. The skb is freed on error. 3280 */ 3281 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3282 { 3283 unsigned int size = skb->len; 3284 if (likely(size >= len)) 3285 return 0; 3286 return skb_pad(skb, len - size); 3287 } 3288 3289 /** 3290 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3291 * @skb: buffer to pad 3292 * @len: minimal length 3293 * @free_on_error: free buffer on error 3294 * 3295 * Pads up a buffer to ensure the trailing bytes exist and are 3296 * blanked. If the buffer already contains sufficient data it 3297 * is untouched. Otherwise it is extended. Returns zero on 3298 * success. The skb is freed on error if @free_on_error is true. 3299 */ 3300 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3301 unsigned int len, 3302 bool free_on_error) 3303 { 3304 unsigned int size = skb->len; 3305 3306 if (unlikely(size < len)) { 3307 len -= size; 3308 if (__skb_pad(skb, len, free_on_error)) 3309 return -ENOMEM; 3310 __skb_put(skb, len); 3311 } 3312 return 0; 3313 } 3314 3315 /** 3316 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3317 * @skb: buffer to pad 3318 * @len: minimal length 3319 * 3320 * Pads up a buffer to ensure the trailing bytes exist and are 3321 * blanked. If the buffer already contains sufficient data it 3322 * is untouched. Otherwise it is extended. Returns zero on 3323 * success. The skb is freed on error. 3324 */ 3325 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3326 { 3327 return __skb_put_padto(skb, len, true); 3328 } 3329 3330 static inline int skb_add_data(struct sk_buff *skb, 3331 struct iov_iter *from, int copy) 3332 { 3333 const int off = skb->len; 3334 3335 if (skb->ip_summed == CHECKSUM_NONE) { 3336 __wsum csum = 0; 3337 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3338 &csum, from)) { 3339 skb->csum = csum_block_add(skb->csum, csum, off); 3340 return 0; 3341 } 3342 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3343 return 0; 3344 3345 __skb_trim(skb, off); 3346 return -EFAULT; 3347 } 3348 3349 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3350 const struct page *page, int off) 3351 { 3352 if (skb_zcopy(skb)) 3353 return false; 3354 if (i) { 3355 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3356 3357 return page == skb_frag_page(frag) && 3358 off == skb_frag_off(frag) + skb_frag_size(frag); 3359 } 3360 return false; 3361 } 3362 3363 static inline int __skb_linearize(struct sk_buff *skb) 3364 { 3365 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3366 } 3367 3368 /** 3369 * skb_linearize - convert paged skb to linear one 3370 * @skb: buffer to linarize 3371 * 3372 * If there is no free memory -ENOMEM is returned, otherwise zero 3373 * is returned and the old skb data released. 3374 */ 3375 static inline int skb_linearize(struct sk_buff *skb) 3376 { 3377 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3378 } 3379 3380 /** 3381 * skb_has_shared_frag - can any frag be overwritten 3382 * @skb: buffer to test 3383 * 3384 * Return true if the skb has at least one frag that might be modified 3385 * by an external entity (as in vmsplice()/sendfile()) 3386 */ 3387 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3388 { 3389 return skb_is_nonlinear(skb) && 3390 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3391 } 3392 3393 /** 3394 * skb_linearize_cow - make sure skb is linear and writable 3395 * @skb: buffer to process 3396 * 3397 * If there is no free memory -ENOMEM is returned, otherwise zero 3398 * is returned and the old skb data released. 3399 */ 3400 static inline int skb_linearize_cow(struct sk_buff *skb) 3401 { 3402 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3403 __skb_linearize(skb) : 0; 3404 } 3405 3406 static __always_inline void 3407 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3408 unsigned int off) 3409 { 3410 if (skb->ip_summed == CHECKSUM_COMPLETE) 3411 skb->csum = csum_block_sub(skb->csum, 3412 csum_partial(start, len, 0), off); 3413 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3414 skb_checksum_start_offset(skb) < 0) 3415 skb->ip_summed = CHECKSUM_NONE; 3416 } 3417 3418 /** 3419 * skb_postpull_rcsum - update checksum for received skb after pull 3420 * @skb: buffer to update 3421 * @start: start of data before pull 3422 * @len: length of data pulled 3423 * 3424 * After doing a pull on a received packet, you need to call this to 3425 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3426 * CHECKSUM_NONE so that it can be recomputed from scratch. 3427 */ 3428 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3429 const void *start, unsigned int len) 3430 { 3431 __skb_postpull_rcsum(skb, start, len, 0); 3432 } 3433 3434 static __always_inline void 3435 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3436 unsigned int off) 3437 { 3438 if (skb->ip_summed == CHECKSUM_COMPLETE) 3439 skb->csum = csum_block_add(skb->csum, 3440 csum_partial(start, len, 0), off); 3441 } 3442 3443 /** 3444 * skb_postpush_rcsum - update checksum for received skb after push 3445 * @skb: buffer to update 3446 * @start: start of data after push 3447 * @len: length of data pushed 3448 * 3449 * After doing a push on a received packet, you need to call this to 3450 * update the CHECKSUM_COMPLETE checksum. 3451 */ 3452 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3453 const void *start, unsigned int len) 3454 { 3455 __skb_postpush_rcsum(skb, start, len, 0); 3456 } 3457 3458 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3459 3460 /** 3461 * skb_push_rcsum - push skb and update receive checksum 3462 * @skb: buffer to update 3463 * @len: length of data pulled 3464 * 3465 * This function performs an skb_push on the packet and updates 3466 * the CHECKSUM_COMPLETE checksum. It should be used on 3467 * receive path processing instead of skb_push unless you know 3468 * that the checksum difference is zero (e.g., a valid IP header) 3469 * or you are setting ip_summed to CHECKSUM_NONE. 3470 */ 3471 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3472 { 3473 skb_push(skb, len); 3474 skb_postpush_rcsum(skb, skb->data, len); 3475 return skb->data; 3476 } 3477 3478 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3479 /** 3480 * pskb_trim_rcsum - trim received skb and update checksum 3481 * @skb: buffer to trim 3482 * @len: new length 3483 * 3484 * This is exactly the same as pskb_trim except that it ensures the 3485 * checksum of received packets are still valid after the operation. 3486 * It can change skb pointers. 3487 */ 3488 3489 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3490 { 3491 if (likely(len >= skb->len)) 3492 return 0; 3493 return pskb_trim_rcsum_slow(skb, len); 3494 } 3495 3496 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3497 { 3498 if (skb->ip_summed == CHECKSUM_COMPLETE) 3499 skb->ip_summed = CHECKSUM_NONE; 3500 __skb_trim(skb, len); 3501 return 0; 3502 } 3503 3504 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3505 { 3506 if (skb->ip_summed == CHECKSUM_COMPLETE) 3507 skb->ip_summed = CHECKSUM_NONE; 3508 return __skb_grow(skb, len); 3509 } 3510 3511 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3512 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3513 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3514 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3515 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3516 3517 #define skb_queue_walk(queue, skb) \ 3518 for (skb = (queue)->next; \ 3519 skb != (struct sk_buff *)(queue); \ 3520 skb = skb->next) 3521 3522 #define skb_queue_walk_safe(queue, skb, tmp) \ 3523 for (skb = (queue)->next, tmp = skb->next; \ 3524 skb != (struct sk_buff *)(queue); \ 3525 skb = tmp, tmp = skb->next) 3526 3527 #define skb_queue_walk_from(queue, skb) \ 3528 for (; skb != (struct sk_buff *)(queue); \ 3529 skb = skb->next) 3530 3531 #define skb_rbtree_walk(skb, root) \ 3532 for (skb = skb_rb_first(root); skb != NULL; \ 3533 skb = skb_rb_next(skb)) 3534 3535 #define skb_rbtree_walk_from(skb) \ 3536 for (; skb != NULL; \ 3537 skb = skb_rb_next(skb)) 3538 3539 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3540 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3541 skb = tmp) 3542 3543 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3544 for (tmp = skb->next; \ 3545 skb != (struct sk_buff *)(queue); \ 3546 skb = tmp, tmp = skb->next) 3547 3548 #define skb_queue_reverse_walk(queue, skb) \ 3549 for (skb = (queue)->prev; \ 3550 skb != (struct sk_buff *)(queue); \ 3551 skb = skb->prev) 3552 3553 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3554 for (skb = (queue)->prev, tmp = skb->prev; \ 3555 skb != (struct sk_buff *)(queue); \ 3556 skb = tmp, tmp = skb->prev) 3557 3558 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3559 for (tmp = skb->prev; \ 3560 skb != (struct sk_buff *)(queue); \ 3561 skb = tmp, tmp = skb->prev) 3562 3563 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3564 { 3565 return skb_shinfo(skb)->frag_list != NULL; 3566 } 3567 3568 static inline void skb_frag_list_init(struct sk_buff *skb) 3569 { 3570 skb_shinfo(skb)->frag_list = NULL; 3571 } 3572 3573 #define skb_walk_frags(skb, iter) \ 3574 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3575 3576 3577 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3578 int *err, long *timeo_p, 3579 const struct sk_buff *skb); 3580 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3581 struct sk_buff_head *queue, 3582 unsigned int flags, 3583 int *off, int *err, 3584 struct sk_buff **last); 3585 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3586 struct sk_buff_head *queue, 3587 unsigned int flags, int *off, int *err, 3588 struct sk_buff **last); 3589 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3590 struct sk_buff_head *sk_queue, 3591 unsigned int flags, int *off, int *err); 3592 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3593 int *err); 3594 __poll_t datagram_poll(struct file *file, struct socket *sock, 3595 struct poll_table_struct *wait); 3596 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3597 struct iov_iter *to, int size); 3598 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3599 struct msghdr *msg, int size) 3600 { 3601 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3602 } 3603 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3604 struct msghdr *msg); 3605 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3606 struct iov_iter *to, int len, 3607 struct ahash_request *hash); 3608 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3609 struct iov_iter *from, int len); 3610 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3611 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3612 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3613 static inline void skb_free_datagram_locked(struct sock *sk, 3614 struct sk_buff *skb) 3615 { 3616 __skb_free_datagram_locked(sk, skb, 0); 3617 } 3618 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3619 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3620 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3621 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3622 int len); 3623 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3624 struct pipe_inode_info *pipe, unsigned int len, 3625 unsigned int flags); 3626 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3627 int len); 3628 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3629 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3630 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3631 int len, int hlen); 3632 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3633 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3634 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3635 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3636 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3637 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3638 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3639 unsigned int offset); 3640 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3641 int skb_ensure_writable(struct sk_buff *skb, int write_len); 3642 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3643 int skb_vlan_pop(struct sk_buff *skb); 3644 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3645 int skb_eth_pop(struct sk_buff *skb); 3646 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3647 const unsigned char *src); 3648 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3649 int mac_len, bool ethernet); 3650 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3651 bool ethernet); 3652 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3653 int skb_mpls_dec_ttl(struct sk_buff *skb); 3654 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3655 gfp_t gfp); 3656 3657 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3658 { 3659 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3660 } 3661 3662 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3663 { 3664 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3665 } 3666 3667 struct skb_checksum_ops { 3668 __wsum (*update)(const void *mem, int len, __wsum wsum); 3669 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3670 }; 3671 3672 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3673 3674 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3675 __wsum csum, const struct skb_checksum_ops *ops); 3676 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3677 __wsum csum); 3678 3679 static inline void * __must_check 3680 __skb_header_pointer(const struct sk_buff *skb, int offset, int len, 3681 const void *data, int hlen, void *buffer) 3682 { 3683 if (likely(hlen - offset >= len)) 3684 return (void *)data + offset; 3685 3686 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) 3687 return NULL; 3688 3689 return buffer; 3690 } 3691 3692 static inline void * __must_check 3693 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3694 { 3695 return __skb_header_pointer(skb, offset, len, skb->data, 3696 skb_headlen(skb), buffer); 3697 } 3698 3699 /** 3700 * skb_needs_linearize - check if we need to linearize a given skb 3701 * depending on the given device features. 3702 * @skb: socket buffer to check 3703 * @features: net device features 3704 * 3705 * Returns true if either: 3706 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3707 * 2. skb is fragmented and the device does not support SG. 3708 */ 3709 static inline bool skb_needs_linearize(struct sk_buff *skb, 3710 netdev_features_t features) 3711 { 3712 return skb_is_nonlinear(skb) && 3713 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3714 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3715 } 3716 3717 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3718 void *to, 3719 const unsigned int len) 3720 { 3721 memcpy(to, skb->data, len); 3722 } 3723 3724 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3725 const int offset, void *to, 3726 const unsigned int len) 3727 { 3728 memcpy(to, skb->data + offset, len); 3729 } 3730 3731 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3732 const void *from, 3733 const unsigned int len) 3734 { 3735 memcpy(skb->data, from, len); 3736 } 3737 3738 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3739 const int offset, 3740 const void *from, 3741 const unsigned int len) 3742 { 3743 memcpy(skb->data + offset, from, len); 3744 } 3745 3746 void skb_init(void); 3747 3748 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3749 { 3750 return skb->tstamp; 3751 } 3752 3753 /** 3754 * skb_get_timestamp - get timestamp from a skb 3755 * @skb: skb to get stamp from 3756 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3757 * 3758 * Timestamps are stored in the skb as offsets to a base timestamp. 3759 * This function converts the offset back to a struct timeval and stores 3760 * it in stamp. 3761 */ 3762 static inline void skb_get_timestamp(const struct sk_buff *skb, 3763 struct __kernel_old_timeval *stamp) 3764 { 3765 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3766 } 3767 3768 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3769 struct __kernel_sock_timeval *stamp) 3770 { 3771 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3772 3773 stamp->tv_sec = ts.tv_sec; 3774 stamp->tv_usec = ts.tv_nsec / 1000; 3775 } 3776 3777 static inline void skb_get_timestampns(const struct sk_buff *skb, 3778 struct __kernel_old_timespec *stamp) 3779 { 3780 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3781 3782 stamp->tv_sec = ts.tv_sec; 3783 stamp->tv_nsec = ts.tv_nsec; 3784 } 3785 3786 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3787 struct __kernel_timespec *stamp) 3788 { 3789 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3790 3791 stamp->tv_sec = ts.tv_sec; 3792 stamp->tv_nsec = ts.tv_nsec; 3793 } 3794 3795 static inline void __net_timestamp(struct sk_buff *skb) 3796 { 3797 skb->tstamp = ktime_get_real(); 3798 } 3799 3800 static inline ktime_t net_timedelta(ktime_t t) 3801 { 3802 return ktime_sub(ktime_get_real(), t); 3803 } 3804 3805 static inline ktime_t net_invalid_timestamp(void) 3806 { 3807 return 0; 3808 } 3809 3810 static inline u8 skb_metadata_len(const struct sk_buff *skb) 3811 { 3812 return skb_shinfo(skb)->meta_len; 3813 } 3814 3815 static inline void *skb_metadata_end(const struct sk_buff *skb) 3816 { 3817 return skb_mac_header(skb); 3818 } 3819 3820 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3821 const struct sk_buff *skb_b, 3822 u8 meta_len) 3823 { 3824 const void *a = skb_metadata_end(skb_a); 3825 const void *b = skb_metadata_end(skb_b); 3826 /* Using more efficient varaiant than plain call to memcmp(). */ 3827 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3828 u64 diffs = 0; 3829 3830 switch (meta_len) { 3831 #define __it(x, op) (x -= sizeof(u##op)) 3832 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3833 case 32: diffs |= __it_diff(a, b, 64); 3834 fallthrough; 3835 case 24: diffs |= __it_diff(a, b, 64); 3836 fallthrough; 3837 case 16: diffs |= __it_diff(a, b, 64); 3838 fallthrough; 3839 case 8: diffs |= __it_diff(a, b, 64); 3840 break; 3841 case 28: diffs |= __it_diff(a, b, 64); 3842 fallthrough; 3843 case 20: diffs |= __it_diff(a, b, 64); 3844 fallthrough; 3845 case 12: diffs |= __it_diff(a, b, 64); 3846 fallthrough; 3847 case 4: diffs |= __it_diff(a, b, 32); 3848 break; 3849 } 3850 return diffs; 3851 #else 3852 return memcmp(a - meta_len, b - meta_len, meta_len); 3853 #endif 3854 } 3855 3856 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3857 const struct sk_buff *skb_b) 3858 { 3859 u8 len_a = skb_metadata_len(skb_a); 3860 u8 len_b = skb_metadata_len(skb_b); 3861 3862 if (!(len_a | len_b)) 3863 return false; 3864 3865 return len_a != len_b ? 3866 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3867 } 3868 3869 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3870 { 3871 skb_shinfo(skb)->meta_len = meta_len; 3872 } 3873 3874 static inline void skb_metadata_clear(struct sk_buff *skb) 3875 { 3876 skb_metadata_set(skb, 0); 3877 } 3878 3879 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3880 3881 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3882 3883 void skb_clone_tx_timestamp(struct sk_buff *skb); 3884 bool skb_defer_rx_timestamp(struct sk_buff *skb); 3885 3886 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3887 3888 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3889 { 3890 } 3891 3892 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3893 { 3894 return false; 3895 } 3896 3897 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3898 3899 /** 3900 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3901 * 3902 * PHY drivers may accept clones of transmitted packets for 3903 * timestamping via their phy_driver.txtstamp method. These drivers 3904 * must call this function to return the skb back to the stack with a 3905 * timestamp. 3906 * 3907 * @skb: clone of the original outgoing packet 3908 * @hwtstamps: hardware time stamps 3909 * 3910 */ 3911 void skb_complete_tx_timestamp(struct sk_buff *skb, 3912 struct skb_shared_hwtstamps *hwtstamps); 3913 3914 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 3915 struct skb_shared_hwtstamps *hwtstamps, 3916 struct sock *sk, int tstype); 3917 3918 /** 3919 * skb_tstamp_tx - queue clone of skb with send time stamps 3920 * @orig_skb: the original outgoing packet 3921 * @hwtstamps: hardware time stamps, may be NULL if not available 3922 * 3923 * If the skb has a socket associated, then this function clones the 3924 * skb (thus sharing the actual data and optional structures), stores 3925 * the optional hardware time stamping information (if non NULL) or 3926 * generates a software time stamp (otherwise), then queues the clone 3927 * to the error queue of the socket. Errors are silently ignored. 3928 */ 3929 void skb_tstamp_tx(struct sk_buff *orig_skb, 3930 struct skb_shared_hwtstamps *hwtstamps); 3931 3932 /** 3933 * skb_tx_timestamp() - Driver hook for transmit timestamping 3934 * 3935 * Ethernet MAC Drivers should call this function in their hard_xmit() 3936 * function immediately before giving the sk_buff to the MAC hardware. 3937 * 3938 * Specifically, one should make absolutely sure that this function is 3939 * called before TX completion of this packet can trigger. Otherwise 3940 * the packet could potentially already be freed. 3941 * 3942 * @skb: A socket buffer. 3943 */ 3944 static inline void skb_tx_timestamp(struct sk_buff *skb) 3945 { 3946 skb_clone_tx_timestamp(skb); 3947 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3948 skb_tstamp_tx(skb, NULL); 3949 } 3950 3951 /** 3952 * skb_complete_wifi_ack - deliver skb with wifi status 3953 * 3954 * @skb: the original outgoing packet 3955 * @acked: ack status 3956 * 3957 */ 3958 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3959 3960 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3961 __sum16 __skb_checksum_complete(struct sk_buff *skb); 3962 3963 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3964 { 3965 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3966 skb->csum_valid || 3967 (skb->ip_summed == CHECKSUM_PARTIAL && 3968 skb_checksum_start_offset(skb) >= 0)); 3969 } 3970 3971 /** 3972 * skb_checksum_complete - Calculate checksum of an entire packet 3973 * @skb: packet to process 3974 * 3975 * This function calculates the checksum over the entire packet plus 3976 * the value of skb->csum. The latter can be used to supply the 3977 * checksum of a pseudo header as used by TCP/UDP. It returns the 3978 * checksum. 3979 * 3980 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 3981 * this function can be used to verify that checksum on received 3982 * packets. In that case the function should return zero if the 3983 * checksum is correct. In particular, this function will return zero 3984 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 3985 * hardware has already verified the correctness of the checksum. 3986 */ 3987 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 3988 { 3989 return skb_csum_unnecessary(skb) ? 3990 0 : __skb_checksum_complete(skb); 3991 } 3992 3993 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 3994 { 3995 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3996 if (skb->csum_level == 0) 3997 skb->ip_summed = CHECKSUM_NONE; 3998 else 3999 skb->csum_level--; 4000 } 4001 } 4002 4003 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4004 { 4005 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4006 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4007 skb->csum_level++; 4008 } else if (skb->ip_summed == CHECKSUM_NONE) { 4009 skb->ip_summed = CHECKSUM_UNNECESSARY; 4010 skb->csum_level = 0; 4011 } 4012 } 4013 4014 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4015 { 4016 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4017 skb->ip_summed = CHECKSUM_NONE; 4018 skb->csum_level = 0; 4019 } 4020 } 4021 4022 /* Check if we need to perform checksum complete validation. 4023 * 4024 * Returns true if checksum complete is needed, false otherwise 4025 * (either checksum is unnecessary or zero checksum is allowed). 4026 */ 4027 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4028 bool zero_okay, 4029 __sum16 check) 4030 { 4031 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4032 skb->csum_valid = 1; 4033 __skb_decr_checksum_unnecessary(skb); 4034 return false; 4035 } 4036 4037 return true; 4038 } 4039 4040 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4041 * in checksum_init. 4042 */ 4043 #define CHECKSUM_BREAK 76 4044 4045 /* Unset checksum-complete 4046 * 4047 * Unset checksum complete can be done when packet is being modified 4048 * (uncompressed for instance) and checksum-complete value is 4049 * invalidated. 4050 */ 4051 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4052 { 4053 if (skb->ip_summed == CHECKSUM_COMPLETE) 4054 skb->ip_summed = CHECKSUM_NONE; 4055 } 4056 4057 /* Validate (init) checksum based on checksum complete. 4058 * 4059 * Return values: 4060 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4061 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4062 * checksum is stored in skb->csum for use in __skb_checksum_complete 4063 * non-zero: value of invalid checksum 4064 * 4065 */ 4066 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4067 bool complete, 4068 __wsum psum) 4069 { 4070 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4071 if (!csum_fold(csum_add(psum, skb->csum))) { 4072 skb->csum_valid = 1; 4073 return 0; 4074 } 4075 } 4076 4077 skb->csum = psum; 4078 4079 if (complete || skb->len <= CHECKSUM_BREAK) { 4080 __sum16 csum; 4081 4082 csum = __skb_checksum_complete(skb); 4083 skb->csum_valid = !csum; 4084 return csum; 4085 } 4086 4087 return 0; 4088 } 4089 4090 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4091 { 4092 return 0; 4093 } 4094 4095 /* Perform checksum validate (init). Note that this is a macro since we only 4096 * want to calculate the pseudo header which is an input function if necessary. 4097 * First we try to validate without any computation (checksum unnecessary) and 4098 * then calculate based on checksum complete calling the function to compute 4099 * pseudo header. 4100 * 4101 * Return values: 4102 * 0: checksum is validated or try to in skb_checksum_complete 4103 * non-zero: value of invalid checksum 4104 */ 4105 #define __skb_checksum_validate(skb, proto, complete, \ 4106 zero_okay, check, compute_pseudo) \ 4107 ({ \ 4108 __sum16 __ret = 0; \ 4109 skb->csum_valid = 0; \ 4110 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4111 __ret = __skb_checksum_validate_complete(skb, \ 4112 complete, compute_pseudo(skb, proto)); \ 4113 __ret; \ 4114 }) 4115 4116 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4117 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4118 4119 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4120 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4121 4122 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4123 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4124 4125 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4126 compute_pseudo) \ 4127 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4128 4129 #define skb_checksum_simple_validate(skb) \ 4130 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4131 4132 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4133 { 4134 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4135 } 4136 4137 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4138 { 4139 skb->csum = ~pseudo; 4140 skb->ip_summed = CHECKSUM_COMPLETE; 4141 } 4142 4143 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4144 do { \ 4145 if (__skb_checksum_convert_check(skb)) \ 4146 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4147 } while (0) 4148 4149 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4150 u16 start, u16 offset) 4151 { 4152 skb->ip_summed = CHECKSUM_PARTIAL; 4153 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4154 skb->csum_offset = offset - start; 4155 } 4156 4157 /* Update skbuf and packet to reflect the remote checksum offload operation. 4158 * When called, ptr indicates the starting point for skb->csum when 4159 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4160 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4161 */ 4162 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4163 int start, int offset, bool nopartial) 4164 { 4165 __wsum delta; 4166 4167 if (!nopartial) { 4168 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4169 return; 4170 } 4171 4172 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4173 __skb_checksum_complete(skb); 4174 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4175 } 4176 4177 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4178 4179 /* Adjust skb->csum since we changed the packet */ 4180 skb->csum = csum_add(skb->csum, delta); 4181 } 4182 4183 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4184 { 4185 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4186 return (void *)(skb->_nfct & NFCT_PTRMASK); 4187 #else 4188 return NULL; 4189 #endif 4190 } 4191 4192 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4193 { 4194 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4195 return skb->_nfct; 4196 #else 4197 return 0UL; 4198 #endif 4199 } 4200 4201 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4202 { 4203 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4204 skb->_nfct = nfct; 4205 #endif 4206 } 4207 4208 #ifdef CONFIG_SKB_EXTENSIONS 4209 enum skb_ext_id { 4210 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4211 SKB_EXT_BRIDGE_NF, 4212 #endif 4213 #ifdef CONFIG_XFRM 4214 SKB_EXT_SEC_PATH, 4215 #endif 4216 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4217 TC_SKB_EXT, 4218 #endif 4219 #if IS_ENABLED(CONFIG_MPTCP) 4220 SKB_EXT_MPTCP, 4221 #endif 4222 SKB_EXT_NUM, /* must be last */ 4223 }; 4224 4225 /** 4226 * struct skb_ext - sk_buff extensions 4227 * @refcnt: 1 on allocation, deallocated on 0 4228 * @offset: offset to add to @data to obtain extension address 4229 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4230 * @data: start of extension data, variable sized 4231 * 4232 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4233 * to use 'u8' types while allowing up to 2kb worth of extension data. 4234 */ 4235 struct skb_ext { 4236 refcount_t refcnt; 4237 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4238 u8 chunks; /* same */ 4239 char data[] __aligned(8); 4240 }; 4241 4242 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4243 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4244 struct skb_ext *ext); 4245 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4246 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4247 void __skb_ext_put(struct skb_ext *ext); 4248 4249 static inline void skb_ext_put(struct sk_buff *skb) 4250 { 4251 if (skb->active_extensions) 4252 __skb_ext_put(skb->extensions); 4253 } 4254 4255 static inline void __skb_ext_copy(struct sk_buff *dst, 4256 const struct sk_buff *src) 4257 { 4258 dst->active_extensions = src->active_extensions; 4259 4260 if (src->active_extensions) { 4261 struct skb_ext *ext = src->extensions; 4262 4263 refcount_inc(&ext->refcnt); 4264 dst->extensions = ext; 4265 } 4266 } 4267 4268 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4269 { 4270 skb_ext_put(dst); 4271 __skb_ext_copy(dst, src); 4272 } 4273 4274 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4275 { 4276 return !!ext->offset[i]; 4277 } 4278 4279 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4280 { 4281 return skb->active_extensions & (1 << id); 4282 } 4283 4284 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4285 { 4286 if (skb_ext_exist(skb, id)) 4287 __skb_ext_del(skb, id); 4288 } 4289 4290 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4291 { 4292 if (skb_ext_exist(skb, id)) { 4293 struct skb_ext *ext = skb->extensions; 4294 4295 return (void *)ext + (ext->offset[id] << 3); 4296 } 4297 4298 return NULL; 4299 } 4300 4301 static inline void skb_ext_reset(struct sk_buff *skb) 4302 { 4303 if (unlikely(skb->active_extensions)) { 4304 __skb_ext_put(skb->extensions); 4305 skb->active_extensions = 0; 4306 } 4307 } 4308 4309 static inline bool skb_has_extensions(struct sk_buff *skb) 4310 { 4311 return unlikely(skb->active_extensions); 4312 } 4313 #else 4314 static inline void skb_ext_put(struct sk_buff *skb) {} 4315 static inline void skb_ext_reset(struct sk_buff *skb) {} 4316 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4317 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4318 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4319 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4320 #endif /* CONFIG_SKB_EXTENSIONS */ 4321 4322 static inline void nf_reset_ct(struct sk_buff *skb) 4323 { 4324 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4325 nf_conntrack_put(skb_nfct(skb)); 4326 skb->_nfct = 0; 4327 #endif 4328 } 4329 4330 static inline void nf_reset_trace(struct sk_buff *skb) 4331 { 4332 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4333 skb->nf_trace = 0; 4334 #endif 4335 } 4336 4337 static inline void ipvs_reset(struct sk_buff *skb) 4338 { 4339 #if IS_ENABLED(CONFIG_IP_VS) 4340 skb->ipvs_property = 0; 4341 #endif 4342 } 4343 4344 /* Note: This doesn't put any conntrack info in dst. */ 4345 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4346 bool copy) 4347 { 4348 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4349 dst->_nfct = src->_nfct; 4350 nf_conntrack_get(skb_nfct(src)); 4351 #endif 4352 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4353 if (copy) 4354 dst->nf_trace = src->nf_trace; 4355 #endif 4356 } 4357 4358 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4359 { 4360 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4361 nf_conntrack_put(skb_nfct(dst)); 4362 #endif 4363 __nf_copy(dst, src, true); 4364 } 4365 4366 #ifdef CONFIG_NETWORK_SECMARK 4367 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4368 { 4369 to->secmark = from->secmark; 4370 } 4371 4372 static inline void skb_init_secmark(struct sk_buff *skb) 4373 { 4374 skb->secmark = 0; 4375 } 4376 #else 4377 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4378 { } 4379 4380 static inline void skb_init_secmark(struct sk_buff *skb) 4381 { } 4382 #endif 4383 4384 static inline int secpath_exists(const struct sk_buff *skb) 4385 { 4386 #ifdef CONFIG_XFRM 4387 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4388 #else 4389 return 0; 4390 #endif 4391 } 4392 4393 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4394 { 4395 return !skb->destructor && 4396 !secpath_exists(skb) && 4397 !skb_nfct(skb) && 4398 !skb->_skb_refdst && 4399 !skb_has_frag_list(skb); 4400 } 4401 4402 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4403 { 4404 skb->queue_mapping = queue_mapping; 4405 } 4406 4407 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4408 { 4409 return skb->queue_mapping; 4410 } 4411 4412 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4413 { 4414 to->queue_mapping = from->queue_mapping; 4415 } 4416 4417 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4418 { 4419 skb->queue_mapping = rx_queue + 1; 4420 } 4421 4422 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4423 { 4424 return skb->queue_mapping - 1; 4425 } 4426 4427 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4428 { 4429 return skb->queue_mapping != 0; 4430 } 4431 4432 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4433 { 4434 skb->dst_pending_confirm = val; 4435 } 4436 4437 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4438 { 4439 return skb->dst_pending_confirm != 0; 4440 } 4441 4442 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4443 { 4444 #ifdef CONFIG_XFRM 4445 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4446 #else 4447 return NULL; 4448 #endif 4449 } 4450 4451 /* Keeps track of mac header offset relative to skb->head. 4452 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4453 * For non-tunnel skb it points to skb_mac_header() and for 4454 * tunnel skb it points to outer mac header. 4455 * Keeps track of level of encapsulation of network headers. 4456 */ 4457 struct skb_gso_cb { 4458 union { 4459 int mac_offset; 4460 int data_offset; 4461 }; 4462 int encap_level; 4463 __wsum csum; 4464 __u16 csum_start; 4465 }; 4466 #define SKB_GSO_CB_OFFSET 32 4467 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4468 4469 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4470 { 4471 return (skb_mac_header(inner_skb) - inner_skb->head) - 4472 SKB_GSO_CB(inner_skb)->mac_offset; 4473 } 4474 4475 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4476 { 4477 int new_headroom, headroom; 4478 int ret; 4479 4480 headroom = skb_headroom(skb); 4481 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4482 if (ret) 4483 return ret; 4484 4485 new_headroom = skb_headroom(skb); 4486 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4487 return 0; 4488 } 4489 4490 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4491 { 4492 /* Do not update partial checksums if remote checksum is enabled. */ 4493 if (skb->remcsum_offload) 4494 return; 4495 4496 SKB_GSO_CB(skb)->csum = res; 4497 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4498 } 4499 4500 /* Compute the checksum for a gso segment. First compute the checksum value 4501 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4502 * then add in skb->csum (checksum from csum_start to end of packet). 4503 * skb->csum and csum_start are then updated to reflect the checksum of the 4504 * resultant packet starting from the transport header-- the resultant checksum 4505 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4506 * header. 4507 */ 4508 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4509 { 4510 unsigned char *csum_start = skb_transport_header(skb); 4511 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4512 __wsum partial = SKB_GSO_CB(skb)->csum; 4513 4514 SKB_GSO_CB(skb)->csum = res; 4515 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4516 4517 return csum_fold(csum_partial(csum_start, plen, partial)); 4518 } 4519 4520 static inline bool skb_is_gso(const struct sk_buff *skb) 4521 { 4522 return skb_shinfo(skb)->gso_size; 4523 } 4524 4525 /* Note: Should be called only if skb_is_gso(skb) is true */ 4526 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4527 { 4528 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4529 } 4530 4531 /* Note: Should be called only if skb_is_gso(skb) is true */ 4532 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4533 { 4534 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4535 } 4536 4537 /* Note: Should be called only if skb_is_gso(skb) is true */ 4538 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4539 { 4540 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4541 } 4542 4543 static inline void skb_gso_reset(struct sk_buff *skb) 4544 { 4545 skb_shinfo(skb)->gso_size = 0; 4546 skb_shinfo(skb)->gso_segs = 0; 4547 skb_shinfo(skb)->gso_type = 0; 4548 } 4549 4550 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4551 u16 increment) 4552 { 4553 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4554 return; 4555 shinfo->gso_size += increment; 4556 } 4557 4558 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4559 u16 decrement) 4560 { 4561 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4562 return; 4563 shinfo->gso_size -= decrement; 4564 } 4565 4566 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4567 4568 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4569 { 4570 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4571 * wanted then gso_type will be set. */ 4572 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4573 4574 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4575 unlikely(shinfo->gso_type == 0)) { 4576 __skb_warn_lro_forwarding(skb); 4577 return true; 4578 } 4579 return false; 4580 } 4581 4582 static inline void skb_forward_csum(struct sk_buff *skb) 4583 { 4584 /* Unfortunately we don't support this one. Any brave souls? */ 4585 if (skb->ip_summed == CHECKSUM_COMPLETE) 4586 skb->ip_summed = CHECKSUM_NONE; 4587 } 4588 4589 /** 4590 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4591 * @skb: skb to check 4592 * 4593 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4594 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4595 * use this helper, to document places where we make this assertion. 4596 */ 4597 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4598 { 4599 #ifdef DEBUG 4600 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4601 #endif 4602 } 4603 4604 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4605 4606 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4607 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4608 unsigned int transport_len, 4609 __sum16(*skb_chkf)(struct sk_buff *skb)); 4610 4611 /** 4612 * skb_head_is_locked - Determine if the skb->head is locked down 4613 * @skb: skb to check 4614 * 4615 * The head on skbs build around a head frag can be removed if they are 4616 * not cloned. This function returns true if the skb head is locked down 4617 * due to either being allocated via kmalloc, or by being a clone with 4618 * multiple references to the head. 4619 */ 4620 static inline bool skb_head_is_locked(const struct sk_buff *skb) 4621 { 4622 return !skb->head_frag || skb_cloned(skb); 4623 } 4624 4625 /* Local Checksum Offload. 4626 * Compute outer checksum based on the assumption that the 4627 * inner checksum will be offloaded later. 4628 * See Documentation/networking/checksum-offloads.rst for 4629 * explanation of how this works. 4630 * Fill in outer checksum adjustment (e.g. with sum of outer 4631 * pseudo-header) before calling. 4632 * Also ensure that inner checksum is in linear data area. 4633 */ 4634 static inline __wsum lco_csum(struct sk_buff *skb) 4635 { 4636 unsigned char *csum_start = skb_checksum_start(skb); 4637 unsigned char *l4_hdr = skb_transport_header(skb); 4638 __wsum partial; 4639 4640 /* Start with complement of inner checksum adjustment */ 4641 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4642 skb->csum_offset)); 4643 4644 /* Add in checksum of our headers (incl. outer checksum 4645 * adjustment filled in by caller) and return result. 4646 */ 4647 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4648 } 4649 4650 static inline bool skb_is_redirected(const struct sk_buff *skb) 4651 { 4652 #ifdef CONFIG_NET_REDIRECT 4653 return skb->redirected; 4654 #else 4655 return false; 4656 #endif 4657 } 4658 4659 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4660 { 4661 #ifdef CONFIG_NET_REDIRECT 4662 skb->redirected = 1; 4663 skb->from_ingress = from_ingress; 4664 if (skb->from_ingress) 4665 skb->tstamp = 0; 4666 #endif 4667 } 4668 4669 static inline void skb_reset_redirect(struct sk_buff *skb) 4670 { 4671 #ifdef CONFIG_NET_REDIRECT 4672 skb->redirected = 0; 4673 #endif 4674 } 4675 4676 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 4677 { 4678 return skb->csum_not_inet; 4679 } 4680 4681 static inline void skb_set_kcov_handle(struct sk_buff *skb, 4682 const u64 kcov_handle) 4683 { 4684 #ifdef CONFIG_KCOV 4685 skb->kcov_handle = kcov_handle; 4686 #endif 4687 } 4688 4689 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4690 { 4691 #ifdef CONFIG_KCOV 4692 return skb->kcov_handle; 4693 #else 4694 return 0; 4695 #endif 4696 } 4697 4698 #endif /* __KERNEL__ */ 4699 #endif /* _LINUX_SKBUFF_H */ 4700