1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Definitions for the 'struct sk_buff' memory handlers. 4 * 5 * Authors: 6 * Alan Cox, <[email protected]> 7 * Florian La Roche, <[email protected]> 8 */ 9 10 #ifndef _LINUX_SKBUFF_H 11 #define _LINUX_SKBUFF_H 12 13 #include <linux/kernel.h> 14 #include <linux/compiler.h> 15 #include <linux/time.h> 16 #include <linux/bug.h> 17 #include <linux/bvec.h> 18 #include <linux/cache.h> 19 #include <linux/rbtree.h> 20 #include <linux/socket.h> 21 #include <linux/refcount.h> 22 23 #include <linux/atomic.h> 24 #include <asm/types.h> 25 #include <linux/spinlock.h> 26 #include <linux/net.h> 27 #include <linux/textsearch.h> 28 #include <net/checksum.h> 29 #include <linux/rcupdate.h> 30 #include <linux/hrtimer.h> 31 #include <linux/dma-mapping.h> 32 #include <linux/netdev_features.h> 33 #include <linux/sched.h> 34 #include <linux/sched/clock.h> 35 #include <net/flow_dissector.h> 36 #include <linux/splice.h> 37 #include <linux/in6.h> 38 #include <linux/if_packet.h> 39 #include <net/flow.h> 40 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 41 #include <linux/netfilter/nf_conntrack_common.h> 42 #endif 43 44 /* The interface for checksum offload between the stack and networking drivers 45 * is as follows... 46 * 47 * A. IP checksum related features 48 * 49 * Drivers advertise checksum offload capabilities in the features of a device. 50 * From the stack's point of view these are capabilities offered by the driver. 51 * A driver typically only advertises features that it is capable of offloading 52 * to its device. 53 * 54 * The checksum related features are: 55 * 56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one 57 * IP (one's complement) checksum for any combination 58 * of protocols or protocol layering. The checksum is 59 * computed and set in a packet per the CHECKSUM_PARTIAL 60 * interface (see below). 61 * 62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain 63 * TCP or UDP packets over IPv4. These are specifically 64 * unencapsulated packets of the form IPv4|TCP or 65 * IPv4|UDP where the Protocol field in the IPv4 header 66 * is TCP or UDP. The IPv4 header may contain IP options. 67 * This feature cannot be set in features for a device 68 * with NETIF_F_HW_CSUM also set. This feature is being 69 * DEPRECATED (see below). 70 * 71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain 72 * TCP or UDP packets over IPv6. These are specifically 73 * unencapsulated packets of the form IPv6|TCP or 74 * IPv6|UDP where the Next Header field in the IPv6 75 * header is either TCP or UDP. IPv6 extension headers 76 * are not supported with this feature. This feature 77 * cannot be set in features for a device with 78 * NETIF_F_HW_CSUM also set. This feature is being 79 * DEPRECATED (see below). 80 * 81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. 82 * This flag is only used to disable the RX checksum 83 * feature for a device. The stack will accept receive 84 * checksum indication in packets received on a device 85 * regardless of whether NETIF_F_RXCSUM is set. 86 * 87 * B. Checksumming of received packets by device. Indication of checksum 88 * verification is set in skb->ip_summed. Possible values are: 89 * 90 * CHECKSUM_NONE: 91 * 92 * Device did not checksum this packet e.g. due to lack of capabilities. 93 * The packet contains full (though not verified) checksum in packet but 94 * not in skb->csum. Thus, skb->csum is undefined in this case. 95 * 96 * CHECKSUM_UNNECESSARY: 97 * 98 * The hardware you're dealing with doesn't calculate the full checksum 99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 101 * if their checksums are okay. skb->csum is still undefined in this case 102 * though. A driver or device must never modify the checksum field in the 103 * packet even if checksum is verified. 104 * 105 * CHECKSUM_UNNECESSARY is applicable to following protocols: 106 * TCP: IPv6 and IPv4. 107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 108 * zero UDP checksum for either IPv4 or IPv6, the networking stack 109 * may perform further validation in this case. 110 * GRE: only if the checksum is present in the header. 111 * SCTP: indicates the CRC in SCTP header has been validated. 112 * FCOE: indicates the CRC in FC frame has been validated. 113 * 114 * skb->csum_level indicates the number of consecutive checksums found in 115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 117 * and a device is able to verify the checksums for UDP (possibly zero), 118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to 119 * two. If the device were only able to verify the UDP checksum and not 120 * GRE, either because it doesn't support GRE checksum or because GRE 121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 122 * not considered in this case). 123 * 124 * CHECKSUM_COMPLETE: 125 * 126 * This is the most generic way. The device supplied checksum of the _whole_ 127 * packet as seen by netif_rx() and fills in skb->csum. This means the 128 * hardware doesn't need to parse L3/L4 headers to implement this. 129 * 130 * Notes: 131 * - Even if device supports only some protocols, but is able to produce 132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 134 * 135 * CHECKSUM_PARTIAL: 136 * 137 * A checksum is set up to be offloaded to a device as described in the 138 * output description for CHECKSUM_PARTIAL. This may occur on a packet 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel 140 * on the same host, or it may be set in the input path in GRO or remote 141 * checksum offload. For the purposes of checksum verification, the checksum 142 * referred to by skb->csum_start + skb->csum_offset and any preceding 143 * checksums in the packet are considered verified. Any checksums in the 144 * packet that are after the checksum being offloaded are not considered to 145 * be verified. 146 * 147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload 148 * in the skb->ip_summed for a packet. Values are: 149 * 150 * CHECKSUM_PARTIAL: 151 * 152 * The driver is required to checksum the packet as seen by hard_start_xmit() 153 * from skb->csum_start up to the end, and to record/write the checksum at 154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the 155 * csum_start and csum_offset values are valid values given the length and 156 * offset of the packet, but it should not attempt to validate that the 157 * checksum refers to a legitimate transport layer checksum -- it is the 158 * purview of the stack to validate that csum_start and csum_offset are set 159 * correctly. 160 * 161 * When the stack requests checksum offload for a packet, the driver MUST 162 * ensure that the checksum is set correctly. A driver can either offload the 163 * checksum calculation to the device, or call skb_checksum_help (in the case 164 * that the device does not support offload for a particular checksum). 165 * 166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of 167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate 168 * checksum offload capability. 169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based 170 * on network device checksumming capabilities: if a packet does not match 171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of 172 * csum_not_inet, see item D.) is called to resolve the checksum. 173 * 174 * CHECKSUM_NONE: 175 * 176 * The skb was already checksummed by the protocol, or a checksum is not 177 * required. 178 * 179 * CHECKSUM_UNNECESSARY: 180 * 181 * This has the same meaning as CHECKSUM_NONE for checksum offload on 182 * output. 183 * 184 * CHECKSUM_COMPLETE: 185 * Not used in checksum output. If a driver observes a packet with this value 186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set. 187 * 188 * D. Non-IP checksum (CRC) offloads 189 * 190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of 191 * offloading the SCTP CRC in a packet. To perform this offload the stack 192 * will set csum_start and csum_offset accordingly, set ip_summed to 193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in 194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c. 195 * A driver that supports both IP checksum offload and SCTP CRC32c offload 196 * must verify which offload is configured for a packet by testing the 197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve 198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 199 * 200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of 201 * offloading the FCOE CRC in a packet. To perform this offload the stack 202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset 203 * accordingly. Note that there is no indication in the skbuff that the 204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 205 * both IP checksum offload and FCOE CRC offload must verify which offload 206 * is configured for a packet, presumably by inspecting packet headers. 207 * 208 * E. Checksumming on output with GSO. 209 * 210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload 211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as 213 * part of the GSO operation is implied. If a checksum is being offloaded 214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and 215 * csum_offset are set to refer to the outermost checksum being offloaded 216 * (two offloaded checksums are possible with UDP encapsulation). 217 */ 218 219 /* Don't change this without changing skb_csum_unnecessary! */ 220 #define CHECKSUM_NONE 0 221 #define CHECKSUM_UNNECESSARY 1 222 #define CHECKSUM_COMPLETE 2 223 #define CHECKSUM_PARTIAL 3 224 225 /* Maximum value in skb->csum_level */ 226 #define SKB_MAX_CSUM_LEVEL 3 227 228 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 229 #define SKB_WITH_OVERHEAD(X) \ 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 231 #define SKB_MAX_ORDER(X, ORDER) \ 232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 233 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 234 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 235 236 /* return minimum truesize of one skb containing X bytes of data */ 237 #define SKB_TRUESIZE(X) ((X) + \ 238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 240 241 struct ahash_request; 242 struct net_device; 243 struct scatterlist; 244 struct pipe_inode_info; 245 struct iov_iter; 246 struct napi_struct; 247 struct bpf_prog; 248 union bpf_attr; 249 struct skb_ext; 250 251 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 252 struct nf_bridge_info { 253 enum { 254 BRNF_PROTO_UNCHANGED, 255 BRNF_PROTO_8021Q, 256 BRNF_PROTO_PPPOE 257 } orig_proto:8; 258 u8 pkt_otherhost:1; 259 u8 in_prerouting:1; 260 u8 bridged_dnat:1; 261 __u16 frag_max_size; 262 struct net_device *physindev; 263 264 /* always valid & non-NULL from FORWARD on, for physdev match */ 265 struct net_device *physoutdev; 266 union { 267 /* prerouting: detect dnat in orig/reply direction */ 268 __be32 ipv4_daddr; 269 struct in6_addr ipv6_daddr; 270 271 /* after prerouting + nat detected: store original source 272 * mac since neigh resolution overwrites it, only used while 273 * skb is out in neigh layer. 274 */ 275 char neigh_header[8]; 276 }; 277 }; 278 #endif 279 280 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 281 /* Chain in tc_skb_ext will be used to share the tc chain with 282 * ovs recirc_id. It will be set to the current chain by tc 283 * and read by ovs to recirc_id. 284 */ 285 struct tc_skb_ext { 286 __u32 chain; 287 __u16 mru; 288 }; 289 #endif 290 291 struct sk_buff_head { 292 /* These two members must be first. */ 293 struct sk_buff *next; 294 struct sk_buff *prev; 295 296 __u32 qlen; 297 spinlock_t lock; 298 }; 299 300 struct sk_buff; 301 302 /* To allow 64K frame to be packed as single skb without frag_list we 303 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 304 * buffers which do not start on a page boundary. 305 * 306 * Since GRO uses frags we allocate at least 16 regardless of page 307 * size. 308 */ 309 #if (65536/PAGE_SIZE + 1) < 16 310 #define MAX_SKB_FRAGS 16UL 311 #else 312 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 313 #endif 314 extern int sysctl_max_skb_frags; 315 316 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 317 * segment using its current segmentation instead. 318 */ 319 #define GSO_BY_FRAGS 0xFFFF 320 321 typedef struct bio_vec skb_frag_t; 322 323 /** 324 * skb_frag_size() - Returns the size of a skb fragment 325 * @frag: skb fragment 326 */ 327 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 328 { 329 return frag->bv_len; 330 } 331 332 /** 333 * skb_frag_size_set() - Sets the size of a skb fragment 334 * @frag: skb fragment 335 * @size: size of fragment 336 */ 337 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 338 { 339 frag->bv_len = size; 340 } 341 342 /** 343 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 344 * @frag: skb fragment 345 * @delta: value to add 346 */ 347 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 348 { 349 frag->bv_len += delta; 350 } 351 352 /** 353 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 354 * @frag: skb fragment 355 * @delta: value to subtract 356 */ 357 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 358 { 359 frag->bv_len -= delta; 360 } 361 362 /** 363 * skb_frag_must_loop - Test if %p is a high memory page 364 * @p: fragment's page 365 */ 366 static inline bool skb_frag_must_loop(struct page *p) 367 { 368 #if defined(CONFIG_HIGHMEM) 369 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 370 return true; 371 #endif 372 return false; 373 } 374 375 /** 376 * skb_frag_foreach_page - loop over pages in a fragment 377 * 378 * @f: skb frag to operate on 379 * @f_off: offset from start of f->bv_page 380 * @f_len: length from f_off to loop over 381 * @p: (temp var) current page 382 * @p_off: (temp var) offset from start of current page, 383 * non-zero only on first page. 384 * @p_len: (temp var) length in current page, 385 * < PAGE_SIZE only on first and last page. 386 * @copied: (temp var) length so far, excluding current p_len. 387 * 388 * A fragment can hold a compound page, in which case per-page 389 * operations, notably kmap_atomic, must be called for each 390 * regular page. 391 */ 392 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 393 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 394 p_off = (f_off) & (PAGE_SIZE - 1), \ 395 p_len = skb_frag_must_loop(p) ? \ 396 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 397 copied = 0; \ 398 copied < f_len; \ 399 copied += p_len, p++, p_off = 0, \ 400 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 401 402 #define HAVE_HW_TIME_STAMP 403 404 /** 405 * struct skb_shared_hwtstamps - hardware time stamps 406 * @hwtstamp: hardware time stamp transformed into duration 407 * since arbitrary point in time 408 * 409 * Software time stamps generated by ktime_get_real() are stored in 410 * skb->tstamp. 411 * 412 * hwtstamps can only be compared against other hwtstamps from 413 * the same device. 414 * 415 * This structure is attached to packets as part of the 416 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 417 */ 418 struct skb_shared_hwtstamps { 419 ktime_t hwtstamp; 420 }; 421 422 /* Definitions for tx_flags in struct skb_shared_info */ 423 enum { 424 /* generate hardware time stamp */ 425 SKBTX_HW_TSTAMP = 1 << 0, 426 427 /* generate software time stamp when queueing packet to NIC */ 428 SKBTX_SW_TSTAMP = 1 << 1, 429 430 /* device driver is going to provide hardware time stamp */ 431 SKBTX_IN_PROGRESS = 1 << 2, 432 433 /* generate wifi status information (where possible) */ 434 SKBTX_WIFI_STATUS = 1 << 4, 435 436 /* generate software time stamp when entering packet scheduling */ 437 SKBTX_SCHED_TSTAMP = 1 << 6, 438 }; 439 440 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 441 SKBTX_SCHED_TSTAMP) 442 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 443 444 /* Definitions for flags in struct skb_shared_info */ 445 enum { 446 /* use zcopy routines */ 447 SKBFL_ZEROCOPY_ENABLE = BIT(0), 448 449 /* This indicates at least one fragment might be overwritten 450 * (as in vmsplice(), sendfile() ...) 451 * If we need to compute a TX checksum, we'll need to copy 452 * all frags to avoid possible bad checksum 453 */ 454 SKBFL_SHARED_FRAG = BIT(1), 455 }; 456 457 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 458 459 /* 460 * The callback notifies userspace to release buffers when skb DMA is done in 461 * lower device, the skb last reference should be 0 when calling this. 462 * The zerocopy_success argument is true if zero copy transmit occurred, 463 * false on data copy or out of memory error caused by data copy attempt. 464 * The ctx field is used to track device context. 465 * The desc field is used to track userspace buffer index. 466 */ 467 struct ubuf_info { 468 void (*callback)(struct sk_buff *, struct ubuf_info *, 469 bool zerocopy_success); 470 union { 471 struct { 472 unsigned long desc; 473 void *ctx; 474 }; 475 struct { 476 u32 id; 477 u16 len; 478 u16 zerocopy:1; 479 u32 bytelen; 480 }; 481 }; 482 refcount_t refcnt; 483 u8 flags; 484 485 struct mmpin { 486 struct user_struct *user; 487 unsigned int num_pg; 488 } mmp; 489 }; 490 491 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 492 493 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 494 void mm_unaccount_pinned_pages(struct mmpin *mmp); 495 496 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size); 497 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 498 struct ubuf_info *uarg); 499 500 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 501 502 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 503 bool success); 504 505 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len); 506 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 507 struct msghdr *msg, int len, 508 struct ubuf_info *uarg); 509 510 /* This data is invariant across clones and lives at 511 * the end of the header data, ie. at skb->end. 512 */ 513 struct skb_shared_info { 514 __u8 flags; 515 __u8 meta_len; 516 __u8 nr_frags; 517 __u8 tx_flags; 518 unsigned short gso_size; 519 /* Warning: this field is not always filled in (UFO)! */ 520 unsigned short gso_segs; 521 struct sk_buff *frag_list; 522 struct skb_shared_hwtstamps hwtstamps; 523 unsigned int gso_type; 524 u32 tskey; 525 526 /* 527 * Warning : all fields before dataref are cleared in __alloc_skb() 528 */ 529 atomic_t dataref; 530 531 /* Intermediate layers must ensure that destructor_arg 532 * remains valid until skb destructor */ 533 void * destructor_arg; 534 535 /* must be last field, see pskb_expand_head() */ 536 skb_frag_t frags[MAX_SKB_FRAGS]; 537 }; 538 539 /* We divide dataref into two halves. The higher 16 bits hold references 540 * to the payload part of skb->data. The lower 16 bits hold references to 541 * the entire skb->data. A clone of a headerless skb holds the length of 542 * the header in skb->hdr_len. 543 * 544 * All users must obey the rule that the skb->data reference count must be 545 * greater than or equal to the payload reference count. 546 * 547 * Holding a reference to the payload part means that the user does not 548 * care about modifications to the header part of skb->data. 549 */ 550 #define SKB_DATAREF_SHIFT 16 551 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 552 553 554 enum { 555 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 556 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 557 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 558 }; 559 560 enum { 561 SKB_GSO_TCPV4 = 1 << 0, 562 563 /* This indicates the skb is from an untrusted source. */ 564 SKB_GSO_DODGY = 1 << 1, 565 566 /* This indicates the tcp segment has CWR set. */ 567 SKB_GSO_TCP_ECN = 1 << 2, 568 569 SKB_GSO_TCP_FIXEDID = 1 << 3, 570 571 SKB_GSO_TCPV6 = 1 << 4, 572 573 SKB_GSO_FCOE = 1 << 5, 574 575 SKB_GSO_GRE = 1 << 6, 576 577 SKB_GSO_GRE_CSUM = 1 << 7, 578 579 SKB_GSO_IPXIP4 = 1 << 8, 580 581 SKB_GSO_IPXIP6 = 1 << 9, 582 583 SKB_GSO_UDP_TUNNEL = 1 << 10, 584 585 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 586 587 SKB_GSO_PARTIAL = 1 << 12, 588 589 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 590 591 SKB_GSO_SCTP = 1 << 14, 592 593 SKB_GSO_ESP = 1 << 15, 594 595 SKB_GSO_UDP = 1 << 16, 596 597 SKB_GSO_UDP_L4 = 1 << 17, 598 599 SKB_GSO_FRAGLIST = 1 << 18, 600 }; 601 602 #if BITS_PER_LONG > 32 603 #define NET_SKBUFF_DATA_USES_OFFSET 1 604 #endif 605 606 #ifdef NET_SKBUFF_DATA_USES_OFFSET 607 typedef unsigned int sk_buff_data_t; 608 #else 609 typedef unsigned char *sk_buff_data_t; 610 #endif 611 612 /** 613 * struct sk_buff - socket buffer 614 * @next: Next buffer in list 615 * @prev: Previous buffer in list 616 * @tstamp: Time we arrived/left 617 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 618 * for retransmit timer 619 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 620 * @list: queue head 621 * @sk: Socket we are owned by 622 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 623 * fragmentation management 624 * @dev: Device we arrived on/are leaving by 625 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 626 * @cb: Control buffer. Free for use by every layer. Put private vars here 627 * @_skb_refdst: destination entry (with norefcount bit) 628 * @sp: the security path, used for xfrm 629 * @len: Length of actual data 630 * @data_len: Data length 631 * @mac_len: Length of link layer header 632 * @hdr_len: writable header length of cloned skb 633 * @csum: Checksum (must include start/offset pair) 634 * @csum_start: Offset from skb->head where checksumming should start 635 * @csum_offset: Offset from csum_start where checksum should be stored 636 * @priority: Packet queueing priority 637 * @ignore_df: allow local fragmentation 638 * @cloned: Head may be cloned (check refcnt to be sure) 639 * @ip_summed: Driver fed us an IP checksum 640 * @nohdr: Payload reference only, must not modify header 641 * @pkt_type: Packet class 642 * @fclone: skbuff clone status 643 * @ipvs_property: skbuff is owned by ipvs 644 * @inner_protocol_type: whether the inner protocol is 645 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 646 * @remcsum_offload: remote checksum offload is enabled 647 * @offload_fwd_mark: Packet was L2-forwarded in hardware 648 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 649 * @tc_skip_classify: do not classify packet. set by IFB device 650 * @tc_at_ingress: used within tc_classify to distinguish in/egress 651 * @redirected: packet was redirected by packet classifier 652 * @from_ingress: packet was redirected from the ingress path 653 * @peeked: this packet has been seen already, so stats have been 654 * done for it, don't do them again 655 * @nf_trace: netfilter packet trace flag 656 * @protocol: Packet protocol from driver 657 * @destructor: Destruct function 658 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 659 * @_nfct: Associated connection, if any (with nfctinfo bits) 660 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 661 * @skb_iif: ifindex of device we arrived on 662 * @tc_index: Traffic control index 663 * @hash: the packet hash 664 * @queue_mapping: Queue mapping for multiqueue devices 665 * @head_frag: skb was allocated from page fragments, 666 * not allocated by kmalloc() or vmalloc(). 667 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 668 * @active_extensions: active extensions (skb_ext_id types) 669 * @ndisc_nodetype: router type (from link layer) 670 * @ooo_okay: allow the mapping of a socket to a queue to be changed 671 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 672 * ports. 673 * @sw_hash: indicates hash was computed in software stack 674 * @wifi_acked_valid: wifi_acked was set 675 * @wifi_acked: whether frame was acked on wifi or not 676 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 677 * @encapsulation: indicates the inner headers in the skbuff are valid 678 * @encap_hdr_csum: software checksum is needed 679 * @csum_valid: checksum is already valid 680 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 681 * @csum_complete_sw: checksum was completed by software 682 * @csum_level: indicates the number of consecutive checksums found in 683 * the packet minus one that have been verified as 684 * CHECKSUM_UNNECESSARY (max 3) 685 * @dst_pending_confirm: need to confirm neighbour 686 * @decrypted: Decrypted SKB 687 * @napi_id: id of the NAPI struct this skb came from 688 * @sender_cpu: (aka @napi_id) source CPU in XPS 689 * @secmark: security marking 690 * @mark: Generic packet mark 691 * @reserved_tailroom: (aka @mark) number of bytes of free space available 692 * at the tail of an sk_buff 693 * @vlan_present: VLAN tag is present 694 * @vlan_proto: vlan encapsulation protocol 695 * @vlan_tci: vlan tag control information 696 * @inner_protocol: Protocol (encapsulation) 697 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 698 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 699 * @inner_transport_header: Inner transport layer header (encapsulation) 700 * @inner_network_header: Network layer header (encapsulation) 701 * @inner_mac_header: Link layer header (encapsulation) 702 * @transport_header: Transport layer header 703 * @network_header: Network layer header 704 * @mac_header: Link layer header 705 * @kcov_handle: KCOV remote handle for remote coverage collection 706 * @tail: Tail pointer 707 * @end: End pointer 708 * @head: Head of buffer 709 * @data: Data head pointer 710 * @truesize: Buffer size 711 * @users: User count - see {datagram,tcp}.c 712 * @extensions: allocated extensions, valid if active_extensions is nonzero 713 */ 714 715 struct sk_buff { 716 union { 717 struct { 718 /* These two members must be first. */ 719 struct sk_buff *next; 720 struct sk_buff *prev; 721 722 union { 723 struct net_device *dev; 724 /* Some protocols might use this space to store information, 725 * while device pointer would be NULL. 726 * UDP receive path is one user. 727 */ 728 unsigned long dev_scratch; 729 }; 730 }; 731 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 732 struct list_head list; 733 }; 734 735 union { 736 struct sock *sk; 737 int ip_defrag_offset; 738 }; 739 740 union { 741 ktime_t tstamp; 742 u64 skb_mstamp_ns; /* earliest departure time */ 743 }; 744 /* 745 * This is the control buffer. It is free to use for every 746 * layer. Please put your private variables there. If you 747 * want to keep them across layers you have to do a skb_clone() 748 * first. This is owned by whoever has the skb queued ATM. 749 */ 750 char cb[48] __aligned(8); 751 752 union { 753 struct { 754 unsigned long _skb_refdst; 755 void (*destructor)(struct sk_buff *skb); 756 }; 757 struct list_head tcp_tsorted_anchor; 758 }; 759 760 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 761 unsigned long _nfct; 762 #endif 763 unsigned int len, 764 data_len; 765 __u16 mac_len, 766 hdr_len; 767 768 /* Following fields are _not_ copied in __copy_skb_header() 769 * Note that queue_mapping is here mostly to fill a hole. 770 */ 771 __u16 queue_mapping; 772 773 /* if you move cloned around you also must adapt those constants */ 774 #ifdef __BIG_ENDIAN_BITFIELD 775 #define CLONED_MASK (1 << 7) 776 #else 777 #define CLONED_MASK 1 778 #endif 779 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset) 780 781 /* private: */ 782 __u8 __cloned_offset[0]; 783 /* public: */ 784 __u8 cloned:1, 785 nohdr:1, 786 fclone:2, 787 peeked:1, 788 head_frag:1, 789 pfmemalloc:1; 790 #ifdef CONFIG_SKB_EXTENSIONS 791 __u8 active_extensions; 792 #endif 793 /* fields enclosed in headers_start/headers_end are copied 794 * using a single memcpy() in __copy_skb_header() 795 */ 796 /* private: */ 797 __u32 headers_start[0]; 798 /* public: */ 799 800 /* if you move pkt_type around you also must adapt those constants */ 801 #ifdef __BIG_ENDIAN_BITFIELD 802 #define PKT_TYPE_MAX (7 << 5) 803 #else 804 #define PKT_TYPE_MAX 7 805 #endif 806 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 807 808 /* private: */ 809 __u8 __pkt_type_offset[0]; 810 /* public: */ 811 __u8 pkt_type:3; 812 __u8 ignore_df:1; 813 __u8 nf_trace:1; 814 __u8 ip_summed:2; 815 __u8 ooo_okay:1; 816 817 __u8 l4_hash:1; 818 __u8 sw_hash:1; 819 __u8 wifi_acked_valid:1; 820 __u8 wifi_acked:1; 821 __u8 no_fcs:1; 822 /* Indicates the inner headers are valid in the skbuff. */ 823 __u8 encapsulation:1; 824 __u8 encap_hdr_csum:1; 825 __u8 csum_valid:1; 826 827 #ifdef __BIG_ENDIAN_BITFIELD 828 #define PKT_VLAN_PRESENT_BIT 7 829 #else 830 #define PKT_VLAN_PRESENT_BIT 0 831 #endif 832 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset) 833 /* private: */ 834 __u8 __pkt_vlan_present_offset[0]; 835 /* public: */ 836 __u8 vlan_present:1; 837 __u8 csum_complete_sw:1; 838 __u8 csum_level:2; 839 __u8 csum_not_inet:1; 840 __u8 dst_pending_confirm:1; 841 #ifdef CONFIG_IPV6_NDISC_NODETYPE 842 __u8 ndisc_nodetype:2; 843 #endif 844 845 __u8 ipvs_property:1; 846 __u8 inner_protocol_type:1; 847 __u8 remcsum_offload:1; 848 #ifdef CONFIG_NET_SWITCHDEV 849 __u8 offload_fwd_mark:1; 850 __u8 offload_l3_fwd_mark:1; 851 #endif 852 #ifdef CONFIG_NET_CLS_ACT 853 __u8 tc_skip_classify:1; 854 __u8 tc_at_ingress:1; 855 #endif 856 #ifdef CONFIG_NET_REDIRECT 857 __u8 redirected:1; 858 __u8 from_ingress:1; 859 #endif 860 #ifdef CONFIG_TLS_DEVICE 861 __u8 decrypted:1; 862 #endif 863 864 #ifdef CONFIG_NET_SCHED 865 __u16 tc_index; /* traffic control index */ 866 #endif 867 868 union { 869 __wsum csum; 870 struct { 871 __u16 csum_start; 872 __u16 csum_offset; 873 }; 874 }; 875 __u32 priority; 876 int skb_iif; 877 __u32 hash; 878 __be16 vlan_proto; 879 __u16 vlan_tci; 880 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 881 union { 882 unsigned int napi_id; 883 unsigned int sender_cpu; 884 }; 885 #endif 886 #ifdef CONFIG_NETWORK_SECMARK 887 __u32 secmark; 888 #endif 889 890 union { 891 __u32 mark; 892 __u32 reserved_tailroom; 893 }; 894 895 union { 896 __be16 inner_protocol; 897 __u8 inner_ipproto; 898 }; 899 900 __u16 inner_transport_header; 901 __u16 inner_network_header; 902 __u16 inner_mac_header; 903 904 __be16 protocol; 905 __u16 transport_header; 906 __u16 network_header; 907 __u16 mac_header; 908 909 #ifdef CONFIG_KCOV 910 u64 kcov_handle; 911 #endif 912 913 /* private: */ 914 __u32 headers_end[0]; 915 /* public: */ 916 917 /* These elements must be at the end, see alloc_skb() for details. */ 918 sk_buff_data_t tail; 919 sk_buff_data_t end; 920 unsigned char *head, 921 *data; 922 unsigned int truesize; 923 refcount_t users; 924 925 #ifdef CONFIG_SKB_EXTENSIONS 926 /* only useable after checking ->active_extensions != 0 */ 927 struct skb_ext *extensions; 928 #endif 929 }; 930 931 #ifdef __KERNEL__ 932 /* 933 * Handling routines are only of interest to the kernel 934 */ 935 936 #define SKB_ALLOC_FCLONE 0x01 937 #define SKB_ALLOC_RX 0x02 938 #define SKB_ALLOC_NAPI 0x04 939 940 /** 941 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 942 * @skb: buffer 943 */ 944 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 945 { 946 return unlikely(skb->pfmemalloc); 947 } 948 949 /* 950 * skb might have a dst pointer attached, refcounted or not. 951 * _skb_refdst low order bit is set if refcount was _not_ taken 952 */ 953 #define SKB_DST_NOREF 1UL 954 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 955 956 /** 957 * skb_dst - returns skb dst_entry 958 * @skb: buffer 959 * 960 * Returns skb dst_entry, regardless of reference taken or not. 961 */ 962 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 963 { 964 /* If refdst was not refcounted, check we still are in a 965 * rcu_read_lock section 966 */ 967 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 968 !rcu_read_lock_held() && 969 !rcu_read_lock_bh_held()); 970 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 971 } 972 973 /** 974 * skb_dst_set - sets skb dst 975 * @skb: buffer 976 * @dst: dst entry 977 * 978 * Sets skb dst, assuming a reference was taken on dst and should 979 * be released by skb_dst_drop() 980 */ 981 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 982 { 983 skb->_skb_refdst = (unsigned long)dst; 984 } 985 986 /** 987 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 988 * @skb: buffer 989 * @dst: dst entry 990 * 991 * Sets skb dst, assuming a reference was not taken on dst. 992 * If dst entry is cached, we do not take reference and dst_release 993 * will be avoided by refdst_drop. If dst entry is not cached, we take 994 * reference, so that last dst_release can destroy the dst immediately. 995 */ 996 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 997 { 998 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 999 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1000 } 1001 1002 /** 1003 * skb_dst_is_noref - Test if skb dst isn't refcounted 1004 * @skb: buffer 1005 */ 1006 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1007 { 1008 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1009 } 1010 1011 /** 1012 * skb_rtable - Returns the skb &rtable 1013 * @skb: buffer 1014 */ 1015 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1016 { 1017 return (struct rtable *)skb_dst(skb); 1018 } 1019 1020 /* For mangling skb->pkt_type from user space side from applications 1021 * such as nft, tc, etc, we only allow a conservative subset of 1022 * possible pkt_types to be set. 1023 */ 1024 static inline bool skb_pkt_type_ok(u32 ptype) 1025 { 1026 return ptype <= PACKET_OTHERHOST; 1027 } 1028 1029 /** 1030 * skb_napi_id - Returns the skb's NAPI id 1031 * @skb: buffer 1032 */ 1033 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1034 { 1035 #ifdef CONFIG_NET_RX_BUSY_POLL 1036 return skb->napi_id; 1037 #else 1038 return 0; 1039 #endif 1040 } 1041 1042 /** 1043 * skb_unref - decrement the skb's reference count 1044 * @skb: buffer 1045 * 1046 * Returns true if we can free the skb. 1047 */ 1048 static inline bool skb_unref(struct sk_buff *skb) 1049 { 1050 if (unlikely(!skb)) 1051 return false; 1052 if (likely(refcount_read(&skb->users) == 1)) 1053 smp_rmb(); 1054 else if (likely(!refcount_dec_and_test(&skb->users))) 1055 return false; 1056 1057 return true; 1058 } 1059 1060 void skb_release_head_state(struct sk_buff *skb); 1061 void kfree_skb(struct sk_buff *skb); 1062 void kfree_skb_list(struct sk_buff *segs); 1063 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1064 void skb_tx_error(struct sk_buff *skb); 1065 1066 #ifdef CONFIG_TRACEPOINTS 1067 void consume_skb(struct sk_buff *skb); 1068 #else 1069 static inline void consume_skb(struct sk_buff *skb) 1070 { 1071 return kfree_skb(skb); 1072 } 1073 #endif 1074 1075 void __consume_stateless_skb(struct sk_buff *skb); 1076 void __kfree_skb(struct sk_buff *skb); 1077 extern struct kmem_cache *skbuff_head_cache; 1078 1079 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1080 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1081 bool *fragstolen, int *delta_truesize); 1082 1083 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1084 int node); 1085 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1086 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1087 struct sk_buff *build_skb_around(struct sk_buff *skb, 1088 void *data, unsigned int frag_size); 1089 1090 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1091 1092 /** 1093 * alloc_skb - allocate a network buffer 1094 * @size: size to allocate 1095 * @priority: allocation mask 1096 * 1097 * This function is a convenient wrapper around __alloc_skb(). 1098 */ 1099 static inline struct sk_buff *alloc_skb(unsigned int size, 1100 gfp_t priority) 1101 { 1102 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1103 } 1104 1105 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1106 unsigned long data_len, 1107 int max_page_order, 1108 int *errcode, 1109 gfp_t gfp_mask); 1110 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1111 1112 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1113 struct sk_buff_fclones { 1114 struct sk_buff skb1; 1115 1116 struct sk_buff skb2; 1117 1118 refcount_t fclone_ref; 1119 }; 1120 1121 /** 1122 * skb_fclone_busy - check if fclone is busy 1123 * @sk: socket 1124 * @skb: buffer 1125 * 1126 * Returns true if skb is a fast clone, and its clone is not freed. 1127 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1128 * so we also check that this didnt happen. 1129 */ 1130 static inline bool skb_fclone_busy(const struct sock *sk, 1131 const struct sk_buff *skb) 1132 { 1133 const struct sk_buff_fclones *fclones; 1134 1135 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1136 1137 return skb->fclone == SKB_FCLONE_ORIG && 1138 refcount_read(&fclones->fclone_ref) > 1 && 1139 fclones->skb2.sk == sk; 1140 } 1141 1142 /** 1143 * alloc_skb_fclone - allocate a network buffer from fclone cache 1144 * @size: size to allocate 1145 * @priority: allocation mask 1146 * 1147 * This function is a convenient wrapper around __alloc_skb(). 1148 */ 1149 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1150 gfp_t priority) 1151 { 1152 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1153 } 1154 1155 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1156 void skb_headers_offset_update(struct sk_buff *skb, int off); 1157 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1158 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1159 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1160 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1161 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1162 gfp_t gfp_mask, bool fclone); 1163 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1164 gfp_t gfp_mask) 1165 { 1166 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1167 } 1168 1169 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1170 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1171 unsigned int headroom); 1172 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1173 int newtailroom, gfp_t priority); 1174 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1175 int offset, int len); 1176 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1177 int offset, int len); 1178 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1179 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1180 1181 /** 1182 * skb_pad - zero pad the tail of an skb 1183 * @skb: buffer to pad 1184 * @pad: space to pad 1185 * 1186 * Ensure that a buffer is followed by a padding area that is zero 1187 * filled. Used by network drivers which may DMA or transfer data 1188 * beyond the buffer end onto the wire. 1189 * 1190 * May return error in out of memory cases. The skb is freed on error. 1191 */ 1192 static inline int skb_pad(struct sk_buff *skb, int pad) 1193 { 1194 return __skb_pad(skb, pad, true); 1195 } 1196 #define dev_kfree_skb(a) consume_skb(a) 1197 1198 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1199 int offset, size_t size); 1200 1201 struct skb_seq_state { 1202 __u32 lower_offset; 1203 __u32 upper_offset; 1204 __u32 frag_idx; 1205 __u32 stepped_offset; 1206 struct sk_buff *root_skb; 1207 struct sk_buff *cur_skb; 1208 __u8 *frag_data; 1209 __u32 frag_off; 1210 }; 1211 1212 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1213 unsigned int to, struct skb_seq_state *st); 1214 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1215 struct skb_seq_state *st); 1216 void skb_abort_seq_read(struct skb_seq_state *st); 1217 1218 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1219 unsigned int to, struct ts_config *config); 1220 1221 /* 1222 * Packet hash types specify the type of hash in skb_set_hash. 1223 * 1224 * Hash types refer to the protocol layer addresses which are used to 1225 * construct a packet's hash. The hashes are used to differentiate or identify 1226 * flows of the protocol layer for the hash type. Hash types are either 1227 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1228 * 1229 * Properties of hashes: 1230 * 1231 * 1) Two packets in different flows have different hash values 1232 * 2) Two packets in the same flow should have the same hash value 1233 * 1234 * A hash at a higher layer is considered to be more specific. A driver should 1235 * set the most specific hash possible. 1236 * 1237 * A driver cannot indicate a more specific hash than the layer at which a hash 1238 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1239 * 1240 * A driver may indicate a hash level which is less specific than the 1241 * actual layer the hash was computed on. For instance, a hash computed 1242 * at L4 may be considered an L3 hash. This should only be done if the 1243 * driver can't unambiguously determine that the HW computed the hash at 1244 * the higher layer. Note that the "should" in the second property above 1245 * permits this. 1246 */ 1247 enum pkt_hash_types { 1248 PKT_HASH_TYPE_NONE, /* Undefined type */ 1249 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1250 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1251 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1252 }; 1253 1254 static inline void skb_clear_hash(struct sk_buff *skb) 1255 { 1256 skb->hash = 0; 1257 skb->sw_hash = 0; 1258 skb->l4_hash = 0; 1259 } 1260 1261 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1262 { 1263 if (!skb->l4_hash) 1264 skb_clear_hash(skb); 1265 } 1266 1267 static inline void 1268 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1269 { 1270 skb->l4_hash = is_l4; 1271 skb->sw_hash = is_sw; 1272 skb->hash = hash; 1273 } 1274 1275 static inline void 1276 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1277 { 1278 /* Used by drivers to set hash from HW */ 1279 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1280 } 1281 1282 static inline void 1283 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1284 { 1285 __skb_set_hash(skb, hash, true, is_l4); 1286 } 1287 1288 void __skb_get_hash(struct sk_buff *skb); 1289 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1290 u32 skb_get_poff(const struct sk_buff *skb); 1291 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1292 const struct flow_keys_basic *keys, int hlen); 1293 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1294 void *data, int hlen_proto); 1295 1296 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1297 int thoff, u8 ip_proto) 1298 { 1299 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1300 } 1301 1302 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1303 const struct flow_dissector_key *key, 1304 unsigned int key_count); 1305 1306 struct bpf_flow_dissector; 1307 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1308 __be16 proto, int nhoff, int hlen, unsigned int flags); 1309 1310 bool __skb_flow_dissect(const struct net *net, 1311 const struct sk_buff *skb, 1312 struct flow_dissector *flow_dissector, 1313 void *target_container, 1314 void *data, __be16 proto, int nhoff, int hlen, 1315 unsigned int flags); 1316 1317 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1318 struct flow_dissector *flow_dissector, 1319 void *target_container, unsigned int flags) 1320 { 1321 return __skb_flow_dissect(NULL, skb, flow_dissector, 1322 target_container, NULL, 0, 0, 0, flags); 1323 } 1324 1325 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1326 struct flow_keys *flow, 1327 unsigned int flags) 1328 { 1329 memset(flow, 0, sizeof(*flow)); 1330 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1331 flow, NULL, 0, 0, 0, flags); 1332 } 1333 1334 static inline bool 1335 skb_flow_dissect_flow_keys_basic(const struct net *net, 1336 const struct sk_buff *skb, 1337 struct flow_keys_basic *flow, void *data, 1338 __be16 proto, int nhoff, int hlen, 1339 unsigned int flags) 1340 { 1341 memset(flow, 0, sizeof(*flow)); 1342 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1343 data, proto, nhoff, hlen, flags); 1344 } 1345 1346 void skb_flow_dissect_meta(const struct sk_buff *skb, 1347 struct flow_dissector *flow_dissector, 1348 void *target_container); 1349 1350 /* Gets a skb connection tracking info, ctinfo map should be a 1351 * map of mapsize to translate enum ip_conntrack_info states 1352 * to user states. 1353 */ 1354 void 1355 skb_flow_dissect_ct(const struct sk_buff *skb, 1356 struct flow_dissector *flow_dissector, 1357 void *target_container, 1358 u16 *ctinfo_map, size_t mapsize, 1359 bool post_ct); 1360 void 1361 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1362 struct flow_dissector *flow_dissector, 1363 void *target_container); 1364 1365 void skb_flow_dissect_hash(const struct sk_buff *skb, 1366 struct flow_dissector *flow_dissector, 1367 void *target_container); 1368 1369 static inline __u32 skb_get_hash(struct sk_buff *skb) 1370 { 1371 if (!skb->l4_hash && !skb->sw_hash) 1372 __skb_get_hash(skb); 1373 1374 return skb->hash; 1375 } 1376 1377 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1378 { 1379 if (!skb->l4_hash && !skb->sw_hash) { 1380 struct flow_keys keys; 1381 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1382 1383 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1384 } 1385 1386 return skb->hash; 1387 } 1388 1389 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1390 const siphash_key_t *perturb); 1391 1392 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1393 { 1394 return skb->hash; 1395 } 1396 1397 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1398 { 1399 to->hash = from->hash; 1400 to->sw_hash = from->sw_hash; 1401 to->l4_hash = from->l4_hash; 1402 }; 1403 1404 static inline void skb_copy_decrypted(struct sk_buff *to, 1405 const struct sk_buff *from) 1406 { 1407 #ifdef CONFIG_TLS_DEVICE 1408 to->decrypted = from->decrypted; 1409 #endif 1410 } 1411 1412 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1413 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1414 { 1415 return skb->head + skb->end; 1416 } 1417 1418 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1419 { 1420 return skb->end; 1421 } 1422 #else 1423 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1424 { 1425 return skb->end; 1426 } 1427 1428 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1429 { 1430 return skb->end - skb->head; 1431 } 1432 #endif 1433 1434 /* Internal */ 1435 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1436 1437 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1438 { 1439 return &skb_shinfo(skb)->hwtstamps; 1440 } 1441 1442 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1443 { 1444 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1445 1446 return is_zcopy ? skb_uarg(skb) : NULL; 1447 } 1448 1449 static inline void net_zcopy_get(struct ubuf_info *uarg) 1450 { 1451 refcount_inc(&uarg->refcnt); 1452 } 1453 1454 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1455 { 1456 skb_shinfo(skb)->destructor_arg = uarg; 1457 skb_shinfo(skb)->flags |= uarg->flags; 1458 } 1459 1460 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1461 bool *have_ref) 1462 { 1463 if (skb && uarg && !skb_zcopy(skb)) { 1464 if (unlikely(have_ref && *have_ref)) 1465 *have_ref = false; 1466 else 1467 net_zcopy_get(uarg); 1468 skb_zcopy_init(skb, uarg); 1469 } 1470 } 1471 1472 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1473 { 1474 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1475 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1476 } 1477 1478 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1479 { 1480 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1481 } 1482 1483 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1484 { 1485 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1486 } 1487 1488 static inline void net_zcopy_put(struct ubuf_info *uarg) 1489 { 1490 if (uarg) 1491 uarg->callback(NULL, uarg, true); 1492 } 1493 1494 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1495 { 1496 if (uarg) { 1497 if (uarg->callback == msg_zerocopy_callback) 1498 msg_zerocopy_put_abort(uarg, have_uref); 1499 else if (have_uref) 1500 net_zcopy_put(uarg); 1501 } 1502 } 1503 1504 /* Release a reference on a zerocopy structure */ 1505 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1506 { 1507 struct ubuf_info *uarg = skb_zcopy(skb); 1508 1509 if (uarg) { 1510 if (!skb_zcopy_is_nouarg(skb)) 1511 uarg->callback(skb, uarg, zerocopy_success); 1512 1513 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG; 1514 } 1515 } 1516 1517 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1518 { 1519 skb->next = NULL; 1520 } 1521 1522 /* Iterate through singly-linked GSO fragments of an skb. */ 1523 #define skb_list_walk_safe(first, skb, next_skb) \ 1524 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1525 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1526 1527 static inline void skb_list_del_init(struct sk_buff *skb) 1528 { 1529 __list_del_entry(&skb->list); 1530 skb_mark_not_on_list(skb); 1531 } 1532 1533 /** 1534 * skb_queue_empty - check if a queue is empty 1535 * @list: queue head 1536 * 1537 * Returns true if the queue is empty, false otherwise. 1538 */ 1539 static inline int skb_queue_empty(const struct sk_buff_head *list) 1540 { 1541 return list->next == (const struct sk_buff *) list; 1542 } 1543 1544 /** 1545 * skb_queue_empty_lockless - check if a queue is empty 1546 * @list: queue head 1547 * 1548 * Returns true if the queue is empty, false otherwise. 1549 * This variant can be used in lockless contexts. 1550 */ 1551 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1552 { 1553 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1554 } 1555 1556 1557 /** 1558 * skb_queue_is_last - check if skb is the last entry in the queue 1559 * @list: queue head 1560 * @skb: buffer 1561 * 1562 * Returns true if @skb is the last buffer on the list. 1563 */ 1564 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1565 const struct sk_buff *skb) 1566 { 1567 return skb->next == (const struct sk_buff *) list; 1568 } 1569 1570 /** 1571 * skb_queue_is_first - check if skb is the first entry in the queue 1572 * @list: queue head 1573 * @skb: buffer 1574 * 1575 * Returns true if @skb is the first buffer on the list. 1576 */ 1577 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1578 const struct sk_buff *skb) 1579 { 1580 return skb->prev == (const struct sk_buff *) list; 1581 } 1582 1583 /** 1584 * skb_queue_next - return the next packet in the queue 1585 * @list: queue head 1586 * @skb: current buffer 1587 * 1588 * Return the next packet in @list after @skb. It is only valid to 1589 * call this if skb_queue_is_last() evaluates to false. 1590 */ 1591 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1592 const struct sk_buff *skb) 1593 { 1594 /* This BUG_ON may seem severe, but if we just return then we 1595 * are going to dereference garbage. 1596 */ 1597 BUG_ON(skb_queue_is_last(list, skb)); 1598 return skb->next; 1599 } 1600 1601 /** 1602 * skb_queue_prev - return the prev packet in the queue 1603 * @list: queue head 1604 * @skb: current buffer 1605 * 1606 * Return the prev packet in @list before @skb. It is only valid to 1607 * call this if skb_queue_is_first() evaluates to false. 1608 */ 1609 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1610 const struct sk_buff *skb) 1611 { 1612 /* This BUG_ON may seem severe, but if we just return then we 1613 * are going to dereference garbage. 1614 */ 1615 BUG_ON(skb_queue_is_first(list, skb)); 1616 return skb->prev; 1617 } 1618 1619 /** 1620 * skb_get - reference buffer 1621 * @skb: buffer to reference 1622 * 1623 * Makes another reference to a socket buffer and returns a pointer 1624 * to the buffer. 1625 */ 1626 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1627 { 1628 refcount_inc(&skb->users); 1629 return skb; 1630 } 1631 1632 /* 1633 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1634 */ 1635 1636 /** 1637 * skb_cloned - is the buffer a clone 1638 * @skb: buffer to check 1639 * 1640 * Returns true if the buffer was generated with skb_clone() and is 1641 * one of multiple shared copies of the buffer. Cloned buffers are 1642 * shared data so must not be written to under normal circumstances. 1643 */ 1644 static inline int skb_cloned(const struct sk_buff *skb) 1645 { 1646 return skb->cloned && 1647 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1648 } 1649 1650 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1651 { 1652 might_sleep_if(gfpflags_allow_blocking(pri)); 1653 1654 if (skb_cloned(skb)) 1655 return pskb_expand_head(skb, 0, 0, pri); 1656 1657 return 0; 1658 } 1659 1660 /** 1661 * skb_header_cloned - is the header a clone 1662 * @skb: buffer to check 1663 * 1664 * Returns true if modifying the header part of the buffer requires 1665 * the data to be copied. 1666 */ 1667 static inline int skb_header_cloned(const struct sk_buff *skb) 1668 { 1669 int dataref; 1670 1671 if (!skb->cloned) 1672 return 0; 1673 1674 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1675 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1676 return dataref != 1; 1677 } 1678 1679 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1680 { 1681 might_sleep_if(gfpflags_allow_blocking(pri)); 1682 1683 if (skb_header_cloned(skb)) 1684 return pskb_expand_head(skb, 0, 0, pri); 1685 1686 return 0; 1687 } 1688 1689 /** 1690 * __skb_header_release - release reference to header 1691 * @skb: buffer to operate on 1692 */ 1693 static inline void __skb_header_release(struct sk_buff *skb) 1694 { 1695 skb->nohdr = 1; 1696 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1697 } 1698 1699 1700 /** 1701 * skb_shared - is the buffer shared 1702 * @skb: buffer to check 1703 * 1704 * Returns true if more than one person has a reference to this 1705 * buffer. 1706 */ 1707 static inline int skb_shared(const struct sk_buff *skb) 1708 { 1709 return refcount_read(&skb->users) != 1; 1710 } 1711 1712 /** 1713 * skb_share_check - check if buffer is shared and if so clone it 1714 * @skb: buffer to check 1715 * @pri: priority for memory allocation 1716 * 1717 * If the buffer is shared the buffer is cloned and the old copy 1718 * drops a reference. A new clone with a single reference is returned. 1719 * If the buffer is not shared the original buffer is returned. When 1720 * being called from interrupt status or with spinlocks held pri must 1721 * be GFP_ATOMIC. 1722 * 1723 * NULL is returned on a memory allocation failure. 1724 */ 1725 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1726 { 1727 might_sleep_if(gfpflags_allow_blocking(pri)); 1728 if (skb_shared(skb)) { 1729 struct sk_buff *nskb = skb_clone(skb, pri); 1730 1731 if (likely(nskb)) 1732 consume_skb(skb); 1733 else 1734 kfree_skb(skb); 1735 skb = nskb; 1736 } 1737 return skb; 1738 } 1739 1740 /* 1741 * Copy shared buffers into a new sk_buff. We effectively do COW on 1742 * packets to handle cases where we have a local reader and forward 1743 * and a couple of other messy ones. The normal one is tcpdumping 1744 * a packet thats being forwarded. 1745 */ 1746 1747 /** 1748 * skb_unshare - make a copy of a shared buffer 1749 * @skb: buffer to check 1750 * @pri: priority for memory allocation 1751 * 1752 * If the socket buffer is a clone then this function creates a new 1753 * copy of the data, drops a reference count on the old copy and returns 1754 * the new copy with the reference count at 1. If the buffer is not a clone 1755 * the original buffer is returned. When called with a spinlock held or 1756 * from interrupt state @pri must be %GFP_ATOMIC 1757 * 1758 * %NULL is returned on a memory allocation failure. 1759 */ 1760 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1761 gfp_t pri) 1762 { 1763 might_sleep_if(gfpflags_allow_blocking(pri)); 1764 if (skb_cloned(skb)) { 1765 struct sk_buff *nskb = skb_copy(skb, pri); 1766 1767 /* Free our shared copy */ 1768 if (likely(nskb)) 1769 consume_skb(skb); 1770 else 1771 kfree_skb(skb); 1772 skb = nskb; 1773 } 1774 return skb; 1775 } 1776 1777 /** 1778 * skb_peek - peek at the head of an &sk_buff_head 1779 * @list_: list to peek at 1780 * 1781 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1782 * be careful with this one. A peek leaves the buffer on the 1783 * list and someone else may run off with it. You must hold 1784 * the appropriate locks or have a private queue to do this. 1785 * 1786 * Returns %NULL for an empty list or a pointer to the head element. 1787 * The reference count is not incremented and the reference is therefore 1788 * volatile. Use with caution. 1789 */ 1790 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1791 { 1792 struct sk_buff *skb = list_->next; 1793 1794 if (skb == (struct sk_buff *)list_) 1795 skb = NULL; 1796 return skb; 1797 } 1798 1799 /** 1800 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1801 * @list_: list to peek at 1802 * 1803 * Like skb_peek(), but the caller knows that the list is not empty. 1804 */ 1805 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1806 { 1807 return list_->next; 1808 } 1809 1810 /** 1811 * skb_peek_next - peek skb following the given one from a queue 1812 * @skb: skb to start from 1813 * @list_: list to peek at 1814 * 1815 * Returns %NULL when the end of the list is met or a pointer to the 1816 * next element. The reference count is not incremented and the 1817 * reference is therefore volatile. Use with caution. 1818 */ 1819 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1820 const struct sk_buff_head *list_) 1821 { 1822 struct sk_buff *next = skb->next; 1823 1824 if (next == (struct sk_buff *)list_) 1825 next = NULL; 1826 return next; 1827 } 1828 1829 /** 1830 * skb_peek_tail - peek at the tail of an &sk_buff_head 1831 * @list_: list to peek at 1832 * 1833 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1834 * be careful with this one. A peek leaves the buffer on the 1835 * list and someone else may run off with it. You must hold 1836 * the appropriate locks or have a private queue to do this. 1837 * 1838 * Returns %NULL for an empty list or a pointer to the tail element. 1839 * The reference count is not incremented and the reference is therefore 1840 * volatile. Use with caution. 1841 */ 1842 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1843 { 1844 struct sk_buff *skb = READ_ONCE(list_->prev); 1845 1846 if (skb == (struct sk_buff *)list_) 1847 skb = NULL; 1848 return skb; 1849 1850 } 1851 1852 /** 1853 * skb_queue_len - get queue length 1854 * @list_: list to measure 1855 * 1856 * Return the length of an &sk_buff queue. 1857 */ 1858 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1859 { 1860 return list_->qlen; 1861 } 1862 1863 /** 1864 * skb_queue_len_lockless - get queue length 1865 * @list_: list to measure 1866 * 1867 * Return the length of an &sk_buff queue. 1868 * This variant can be used in lockless contexts. 1869 */ 1870 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1871 { 1872 return READ_ONCE(list_->qlen); 1873 } 1874 1875 /** 1876 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1877 * @list: queue to initialize 1878 * 1879 * This initializes only the list and queue length aspects of 1880 * an sk_buff_head object. This allows to initialize the list 1881 * aspects of an sk_buff_head without reinitializing things like 1882 * the spinlock. It can also be used for on-stack sk_buff_head 1883 * objects where the spinlock is known to not be used. 1884 */ 1885 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1886 { 1887 list->prev = list->next = (struct sk_buff *)list; 1888 list->qlen = 0; 1889 } 1890 1891 /* 1892 * This function creates a split out lock class for each invocation; 1893 * this is needed for now since a whole lot of users of the skb-queue 1894 * infrastructure in drivers have different locking usage (in hardirq) 1895 * than the networking core (in softirq only). In the long run either the 1896 * network layer or drivers should need annotation to consolidate the 1897 * main types of usage into 3 classes. 1898 */ 1899 static inline void skb_queue_head_init(struct sk_buff_head *list) 1900 { 1901 spin_lock_init(&list->lock); 1902 __skb_queue_head_init(list); 1903 } 1904 1905 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1906 struct lock_class_key *class) 1907 { 1908 skb_queue_head_init(list); 1909 lockdep_set_class(&list->lock, class); 1910 } 1911 1912 /* 1913 * Insert an sk_buff on a list. 1914 * 1915 * The "__skb_xxxx()" functions are the non-atomic ones that 1916 * can only be called with interrupts disabled. 1917 */ 1918 static inline void __skb_insert(struct sk_buff *newsk, 1919 struct sk_buff *prev, struct sk_buff *next, 1920 struct sk_buff_head *list) 1921 { 1922 /* See skb_queue_empty_lockless() and skb_peek_tail() 1923 * for the opposite READ_ONCE() 1924 */ 1925 WRITE_ONCE(newsk->next, next); 1926 WRITE_ONCE(newsk->prev, prev); 1927 WRITE_ONCE(next->prev, newsk); 1928 WRITE_ONCE(prev->next, newsk); 1929 list->qlen++; 1930 } 1931 1932 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1933 struct sk_buff *prev, 1934 struct sk_buff *next) 1935 { 1936 struct sk_buff *first = list->next; 1937 struct sk_buff *last = list->prev; 1938 1939 WRITE_ONCE(first->prev, prev); 1940 WRITE_ONCE(prev->next, first); 1941 1942 WRITE_ONCE(last->next, next); 1943 WRITE_ONCE(next->prev, last); 1944 } 1945 1946 /** 1947 * skb_queue_splice - join two skb lists, this is designed for stacks 1948 * @list: the new list to add 1949 * @head: the place to add it in the first list 1950 */ 1951 static inline void skb_queue_splice(const struct sk_buff_head *list, 1952 struct sk_buff_head *head) 1953 { 1954 if (!skb_queue_empty(list)) { 1955 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1956 head->qlen += list->qlen; 1957 } 1958 } 1959 1960 /** 1961 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1962 * @list: the new list to add 1963 * @head: the place to add it in the first list 1964 * 1965 * The list at @list is reinitialised 1966 */ 1967 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1968 struct sk_buff_head *head) 1969 { 1970 if (!skb_queue_empty(list)) { 1971 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1972 head->qlen += list->qlen; 1973 __skb_queue_head_init(list); 1974 } 1975 } 1976 1977 /** 1978 * skb_queue_splice_tail - join two skb lists, each list being a queue 1979 * @list: the new list to add 1980 * @head: the place to add it in the first list 1981 */ 1982 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1983 struct sk_buff_head *head) 1984 { 1985 if (!skb_queue_empty(list)) { 1986 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1987 head->qlen += list->qlen; 1988 } 1989 } 1990 1991 /** 1992 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1993 * @list: the new list to add 1994 * @head: the place to add it in the first list 1995 * 1996 * Each of the lists is a queue. 1997 * The list at @list is reinitialised 1998 */ 1999 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2000 struct sk_buff_head *head) 2001 { 2002 if (!skb_queue_empty(list)) { 2003 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2004 head->qlen += list->qlen; 2005 __skb_queue_head_init(list); 2006 } 2007 } 2008 2009 /** 2010 * __skb_queue_after - queue a buffer at the list head 2011 * @list: list to use 2012 * @prev: place after this buffer 2013 * @newsk: buffer to queue 2014 * 2015 * Queue a buffer int the middle of a list. This function takes no locks 2016 * and you must therefore hold required locks before calling it. 2017 * 2018 * A buffer cannot be placed on two lists at the same time. 2019 */ 2020 static inline void __skb_queue_after(struct sk_buff_head *list, 2021 struct sk_buff *prev, 2022 struct sk_buff *newsk) 2023 { 2024 __skb_insert(newsk, prev, prev->next, list); 2025 } 2026 2027 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2028 struct sk_buff_head *list); 2029 2030 static inline void __skb_queue_before(struct sk_buff_head *list, 2031 struct sk_buff *next, 2032 struct sk_buff *newsk) 2033 { 2034 __skb_insert(newsk, next->prev, next, list); 2035 } 2036 2037 /** 2038 * __skb_queue_head - queue a buffer at the list head 2039 * @list: list to use 2040 * @newsk: buffer to queue 2041 * 2042 * Queue a buffer at the start of a list. This function takes no locks 2043 * and you must therefore hold required locks before calling it. 2044 * 2045 * A buffer cannot be placed on two lists at the same time. 2046 */ 2047 static inline void __skb_queue_head(struct sk_buff_head *list, 2048 struct sk_buff *newsk) 2049 { 2050 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2051 } 2052 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2053 2054 /** 2055 * __skb_queue_tail - queue a buffer at the list tail 2056 * @list: list to use 2057 * @newsk: buffer to queue 2058 * 2059 * Queue a buffer at the end of a list. This function takes no locks 2060 * and you must therefore hold required locks before calling it. 2061 * 2062 * A buffer cannot be placed on two lists at the same time. 2063 */ 2064 static inline void __skb_queue_tail(struct sk_buff_head *list, 2065 struct sk_buff *newsk) 2066 { 2067 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2068 } 2069 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2070 2071 /* 2072 * remove sk_buff from list. _Must_ be called atomically, and with 2073 * the list known.. 2074 */ 2075 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2076 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2077 { 2078 struct sk_buff *next, *prev; 2079 2080 WRITE_ONCE(list->qlen, list->qlen - 1); 2081 next = skb->next; 2082 prev = skb->prev; 2083 skb->next = skb->prev = NULL; 2084 WRITE_ONCE(next->prev, prev); 2085 WRITE_ONCE(prev->next, next); 2086 } 2087 2088 /** 2089 * __skb_dequeue - remove from the head of the queue 2090 * @list: list to dequeue from 2091 * 2092 * Remove the head of the list. This function does not take any locks 2093 * so must be used with appropriate locks held only. The head item is 2094 * returned or %NULL if the list is empty. 2095 */ 2096 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2097 { 2098 struct sk_buff *skb = skb_peek(list); 2099 if (skb) 2100 __skb_unlink(skb, list); 2101 return skb; 2102 } 2103 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2104 2105 /** 2106 * __skb_dequeue_tail - remove from the tail of the queue 2107 * @list: list to dequeue from 2108 * 2109 * Remove the tail of the list. This function does not take any locks 2110 * so must be used with appropriate locks held only. The tail item is 2111 * returned or %NULL if the list is empty. 2112 */ 2113 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2114 { 2115 struct sk_buff *skb = skb_peek_tail(list); 2116 if (skb) 2117 __skb_unlink(skb, list); 2118 return skb; 2119 } 2120 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2121 2122 2123 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2124 { 2125 return skb->data_len; 2126 } 2127 2128 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2129 { 2130 return skb->len - skb->data_len; 2131 } 2132 2133 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2134 { 2135 unsigned int i, len = 0; 2136 2137 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2138 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2139 return len; 2140 } 2141 2142 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2143 { 2144 return skb_headlen(skb) + __skb_pagelen(skb); 2145 } 2146 2147 /** 2148 * __skb_fill_page_desc - initialise a paged fragment in an skb 2149 * @skb: buffer containing fragment to be initialised 2150 * @i: paged fragment index to initialise 2151 * @page: the page to use for this fragment 2152 * @off: the offset to the data with @page 2153 * @size: the length of the data 2154 * 2155 * Initialises the @i'th fragment of @skb to point to &size bytes at 2156 * offset @off within @page. 2157 * 2158 * Does not take any additional reference on the fragment. 2159 */ 2160 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2161 struct page *page, int off, int size) 2162 { 2163 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2164 2165 /* 2166 * Propagate page pfmemalloc to the skb if we can. The problem is 2167 * that not all callers have unique ownership of the page but rely 2168 * on page_is_pfmemalloc doing the right thing(tm). 2169 */ 2170 frag->bv_page = page; 2171 frag->bv_offset = off; 2172 skb_frag_size_set(frag, size); 2173 2174 page = compound_head(page); 2175 if (page_is_pfmemalloc(page)) 2176 skb->pfmemalloc = true; 2177 } 2178 2179 /** 2180 * skb_fill_page_desc - initialise a paged fragment in an skb 2181 * @skb: buffer containing fragment to be initialised 2182 * @i: paged fragment index to initialise 2183 * @page: the page to use for this fragment 2184 * @off: the offset to the data with @page 2185 * @size: the length of the data 2186 * 2187 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2188 * @skb to point to @size bytes at offset @off within @page. In 2189 * addition updates @skb such that @i is the last fragment. 2190 * 2191 * Does not take any additional reference on the fragment. 2192 */ 2193 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2194 struct page *page, int off, int size) 2195 { 2196 __skb_fill_page_desc(skb, i, page, off, size); 2197 skb_shinfo(skb)->nr_frags = i + 1; 2198 } 2199 2200 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2201 int size, unsigned int truesize); 2202 2203 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2204 unsigned int truesize); 2205 2206 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2207 2208 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2209 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2210 { 2211 return skb->head + skb->tail; 2212 } 2213 2214 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2215 { 2216 skb->tail = skb->data - skb->head; 2217 } 2218 2219 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2220 { 2221 skb_reset_tail_pointer(skb); 2222 skb->tail += offset; 2223 } 2224 2225 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2226 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2227 { 2228 return skb->tail; 2229 } 2230 2231 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2232 { 2233 skb->tail = skb->data; 2234 } 2235 2236 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2237 { 2238 skb->tail = skb->data + offset; 2239 } 2240 2241 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2242 2243 /* 2244 * Add data to an sk_buff 2245 */ 2246 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2247 void *skb_put(struct sk_buff *skb, unsigned int len); 2248 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2249 { 2250 void *tmp = skb_tail_pointer(skb); 2251 SKB_LINEAR_ASSERT(skb); 2252 skb->tail += len; 2253 skb->len += len; 2254 return tmp; 2255 } 2256 2257 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2258 { 2259 void *tmp = __skb_put(skb, len); 2260 2261 memset(tmp, 0, len); 2262 return tmp; 2263 } 2264 2265 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2266 unsigned int len) 2267 { 2268 void *tmp = __skb_put(skb, len); 2269 2270 memcpy(tmp, data, len); 2271 return tmp; 2272 } 2273 2274 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2275 { 2276 *(u8 *)__skb_put(skb, 1) = val; 2277 } 2278 2279 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2280 { 2281 void *tmp = skb_put(skb, len); 2282 2283 memset(tmp, 0, len); 2284 2285 return tmp; 2286 } 2287 2288 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2289 unsigned int len) 2290 { 2291 void *tmp = skb_put(skb, len); 2292 2293 memcpy(tmp, data, len); 2294 2295 return tmp; 2296 } 2297 2298 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2299 { 2300 *(u8 *)skb_put(skb, 1) = val; 2301 } 2302 2303 void *skb_push(struct sk_buff *skb, unsigned int len); 2304 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2305 { 2306 skb->data -= len; 2307 skb->len += len; 2308 return skb->data; 2309 } 2310 2311 void *skb_pull(struct sk_buff *skb, unsigned int len); 2312 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2313 { 2314 skb->len -= len; 2315 BUG_ON(skb->len < skb->data_len); 2316 return skb->data += len; 2317 } 2318 2319 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2320 { 2321 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2322 } 2323 2324 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2325 2326 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2327 { 2328 if (len > skb_headlen(skb) && 2329 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2330 return NULL; 2331 skb->len -= len; 2332 return skb->data += len; 2333 } 2334 2335 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2336 { 2337 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2338 } 2339 2340 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2341 { 2342 if (likely(len <= skb_headlen(skb))) 2343 return true; 2344 if (unlikely(len > skb->len)) 2345 return false; 2346 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2347 } 2348 2349 void skb_condense(struct sk_buff *skb); 2350 2351 /** 2352 * skb_headroom - bytes at buffer head 2353 * @skb: buffer to check 2354 * 2355 * Return the number of bytes of free space at the head of an &sk_buff. 2356 */ 2357 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2358 { 2359 return skb->data - skb->head; 2360 } 2361 2362 /** 2363 * skb_tailroom - bytes at buffer end 2364 * @skb: buffer to check 2365 * 2366 * Return the number of bytes of free space at the tail of an sk_buff 2367 */ 2368 static inline int skb_tailroom(const struct sk_buff *skb) 2369 { 2370 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2371 } 2372 2373 /** 2374 * skb_availroom - bytes at buffer end 2375 * @skb: buffer to check 2376 * 2377 * Return the number of bytes of free space at the tail of an sk_buff 2378 * allocated by sk_stream_alloc() 2379 */ 2380 static inline int skb_availroom(const struct sk_buff *skb) 2381 { 2382 if (skb_is_nonlinear(skb)) 2383 return 0; 2384 2385 return skb->end - skb->tail - skb->reserved_tailroom; 2386 } 2387 2388 /** 2389 * skb_reserve - adjust headroom 2390 * @skb: buffer to alter 2391 * @len: bytes to move 2392 * 2393 * Increase the headroom of an empty &sk_buff by reducing the tail 2394 * room. This is only allowed for an empty buffer. 2395 */ 2396 static inline void skb_reserve(struct sk_buff *skb, int len) 2397 { 2398 skb->data += len; 2399 skb->tail += len; 2400 } 2401 2402 /** 2403 * skb_tailroom_reserve - adjust reserved_tailroom 2404 * @skb: buffer to alter 2405 * @mtu: maximum amount of headlen permitted 2406 * @needed_tailroom: minimum amount of reserved_tailroom 2407 * 2408 * Set reserved_tailroom so that headlen can be as large as possible but 2409 * not larger than mtu and tailroom cannot be smaller than 2410 * needed_tailroom. 2411 * The required headroom should already have been reserved before using 2412 * this function. 2413 */ 2414 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2415 unsigned int needed_tailroom) 2416 { 2417 SKB_LINEAR_ASSERT(skb); 2418 if (mtu < skb_tailroom(skb) - needed_tailroom) 2419 /* use at most mtu */ 2420 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2421 else 2422 /* use up to all available space */ 2423 skb->reserved_tailroom = needed_tailroom; 2424 } 2425 2426 #define ENCAP_TYPE_ETHER 0 2427 #define ENCAP_TYPE_IPPROTO 1 2428 2429 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2430 __be16 protocol) 2431 { 2432 skb->inner_protocol = protocol; 2433 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2434 } 2435 2436 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2437 __u8 ipproto) 2438 { 2439 skb->inner_ipproto = ipproto; 2440 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2441 } 2442 2443 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2444 { 2445 skb->inner_mac_header = skb->mac_header; 2446 skb->inner_network_header = skb->network_header; 2447 skb->inner_transport_header = skb->transport_header; 2448 } 2449 2450 static inline void skb_reset_mac_len(struct sk_buff *skb) 2451 { 2452 skb->mac_len = skb->network_header - skb->mac_header; 2453 } 2454 2455 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2456 *skb) 2457 { 2458 return skb->head + skb->inner_transport_header; 2459 } 2460 2461 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2462 { 2463 return skb_inner_transport_header(skb) - skb->data; 2464 } 2465 2466 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2467 { 2468 skb->inner_transport_header = skb->data - skb->head; 2469 } 2470 2471 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2472 const int offset) 2473 { 2474 skb_reset_inner_transport_header(skb); 2475 skb->inner_transport_header += offset; 2476 } 2477 2478 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2479 { 2480 return skb->head + skb->inner_network_header; 2481 } 2482 2483 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2484 { 2485 skb->inner_network_header = skb->data - skb->head; 2486 } 2487 2488 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2489 const int offset) 2490 { 2491 skb_reset_inner_network_header(skb); 2492 skb->inner_network_header += offset; 2493 } 2494 2495 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2496 { 2497 return skb->head + skb->inner_mac_header; 2498 } 2499 2500 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2501 { 2502 skb->inner_mac_header = skb->data - skb->head; 2503 } 2504 2505 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2506 const int offset) 2507 { 2508 skb_reset_inner_mac_header(skb); 2509 skb->inner_mac_header += offset; 2510 } 2511 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2512 { 2513 return skb->transport_header != (typeof(skb->transport_header))~0U; 2514 } 2515 2516 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2517 { 2518 return skb->head + skb->transport_header; 2519 } 2520 2521 static inline void skb_reset_transport_header(struct sk_buff *skb) 2522 { 2523 skb->transport_header = skb->data - skb->head; 2524 } 2525 2526 static inline void skb_set_transport_header(struct sk_buff *skb, 2527 const int offset) 2528 { 2529 skb_reset_transport_header(skb); 2530 skb->transport_header += offset; 2531 } 2532 2533 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2534 { 2535 return skb->head + skb->network_header; 2536 } 2537 2538 static inline void skb_reset_network_header(struct sk_buff *skb) 2539 { 2540 skb->network_header = skb->data - skb->head; 2541 } 2542 2543 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2544 { 2545 skb_reset_network_header(skb); 2546 skb->network_header += offset; 2547 } 2548 2549 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2550 { 2551 return skb->head + skb->mac_header; 2552 } 2553 2554 static inline int skb_mac_offset(const struct sk_buff *skb) 2555 { 2556 return skb_mac_header(skb) - skb->data; 2557 } 2558 2559 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2560 { 2561 return skb->network_header - skb->mac_header; 2562 } 2563 2564 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2565 { 2566 return skb->mac_header != (typeof(skb->mac_header))~0U; 2567 } 2568 2569 static inline void skb_unset_mac_header(struct sk_buff *skb) 2570 { 2571 skb->mac_header = (typeof(skb->mac_header))~0U; 2572 } 2573 2574 static inline void skb_reset_mac_header(struct sk_buff *skb) 2575 { 2576 skb->mac_header = skb->data - skb->head; 2577 } 2578 2579 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2580 { 2581 skb_reset_mac_header(skb); 2582 skb->mac_header += offset; 2583 } 2584 2585 static inline void skb_pop_mac_header(struct sk_buff *skb) 2586 { 2587 skb->mac_header = skb->network_header; 2588 } 2589 2590 static inline void skb_probe_transport_header(struct sk_buff *skb) 2591 { 2592 struct flow_keys_basic keys; 2593 2594 if (skb_transport_header_was_set(skb)) 2595 return; 2596 2597 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2598 NULL, 0, 0, 0, 0)) 2599 skb_set_transport_header(skb, keys.control.thoff); 2600 } 2601 2602 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2603 { 2604 if (skb_mac_header_was_set(skb)) { 2605 const unsigned char *old_mac = skb_mac_header(skb); 2606 2607 skb_set_mac_header(skb, -skb->mac_len); 2608 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2609 } 2610 } 2611 2612 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2613 { 2614 return skb->csum_start - skb_headroom(skb); 2615 } 2616 2617 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2618 { 2619 return skb->head + skb->csum_start; 2620 } 2621 2622 static inline int skb_transport_offset(const struct sk_buff *skb) 2623 { 2624 return skb_transport_header(skb) - skb->data; 2625 } 2626 2627 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2628 { 2629 return skb->transport_header - skb->network_header; 2630 } 2631 2632 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2633 { 2634 return skb->inner_transport_header - skb->inner_network_header; 2635 } 2636 2637 static inline int skb_network_offset(const struct sk_buff *skb) 2638 { 2639 return skb_network_header(skb) - skb->data; 2640 } 2641 2642 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2643 { 2644 return skb_inner_network_header(skb) - skb->data; 2645 } 2646 2647 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2648 { 2649 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2650 } 2651 2652 /* 2653 * CPUs often take a performance hit when accessing unaligned memory 2654 * locations. The actual performance hit varies, it can be small if the 2655 * hardware handles it or large if we have to take an exception and fix it 2656 * in software. 2657 * 2658 * Since an ethernet header is 14 bytes network drivers often end up with 2659 * the IP header at an unaligned offset. The IP header can be aligned by 2660 * shifting the start of the packet by 2 bytes. Drivers should do this 2661 * with: 2662 * 2663 * skb_reserve(skb, NET_IP_ALIGN); 2664 * 2665 * The downside to this alignment of the IP header is that the DMA is now 2666 * unaligned. On some architectures the cost of an unaligned DMA is high 2667 * and this cost outweighs the gains made by aligning the IP header. 2668 * 2669 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2670 * to be overridden. 2671 */ 2672 #ifndef NET_IP_ALIGN 2673 #define NET_IP_ALIGN 2 2674 #endif 2675 2676 /* 2677 * The networking layer reserves some headroom in skb data (via 2678 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2679 * the header has to grow. In the default case, if the header has to grow 2680 * 32 bytes or less we avoid the reallocation. 2681 * 2682 * Unfortunately this headroom changes the DMA alignment of the resulting 2683 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2684 * on some architectures. An architecture can override this value, 2685 * perhaps setting it to a cacheline in size (since that will maintain 2686 * cacheline alignment of the DMA). It must be a power of 2. 2687 * 2688 * Various parts of the networking layer expect at least 32 bytes of 2689 * headroom, you should not reduce this. 2690 * 2691 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2692 * to reduce average number of cache lines per packet. 2693 * get_rps_cpu() for example only access one 64 bytes aligned block : 2694 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2695 */ 2696 #ifndef NET_SKB_PAD 2697 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2698 #endif 2699 2700 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2701 2702 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2703 { 2704 if (WARN_ON(skb_is_nonlinear(skb))) 2705 return; 2706 skb->len = len; 2707 skb_set_tail_pointer(skb, len); 2708 } 2709 2710 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2711 { 2712 __skb_set_length(skb, len); 2713 } 2714 2715 void skb_trim(struct sk_buff *skb, unsigned int len); 2716 2717 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2718 { 2719 if (skb->data_len) 2720 return ___pskb_trim(skb, len); 2721 __skb_trim(skb, len); 2722 return 0; 2723 } 2724 2725 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2726 { 2727 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2728 } 2729 2730 /** 2731 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2732 * @skb: buffer to alter 2733 * @len: new length 2734 * 2735 * This is identical to pskb_trim except that the caller knows that 2736 * the skb is not cloned so we should never get an error due to out- 2737 * of-memory. 2738 */ 2739 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2740 { 2741 int err = pskb_trim(skb, len); 2742 BUG_ON(err); 2743 } 2744 2745 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2746 { 2747 unsigned int diff = len - skb->len; 2748 2749 if (skb_tailroom(skb) < diff) { 2750 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2751 GFP_ATOMIC); 2752 if (ret) 2753 return ret; 2754 } 2755 __skb_set_length(skb, len); 2756 return 0; 2757 } 2758 2759 /** 2760 * skb_orphan - orphan a buffer 2761 * @skb: buffer to orphan 2762 * 2763 * If a buffer currently has an owner then we call the owner's 2764 * destructor function and make the @skb unowned. The buffer continues 2765 * to exist but is no longer charged to its former owner. 2766 */ 2767 static inline void skb_orphan(struct sk_buff *skb) 2768 { 2769 if (skb->destructor) { 2770 skb->destructor(skb); 2771 skb->destructor = NULL; 2772 skb->sk = NULL; 2773 } else { 2774 BUG_ON(skb->sk); 2775 } 2776 } 2777 2778 /** 2779 * skb_orphan_frags - orphan the frags contained in a buffer 2780 * @skb: buffer to orphan frags from 2781 * @gfp_mask: allocation mask for replacement pages 2782 * 2783 * For each frag in the SKB which needs a destructor (i.e. has an 2784 * owner) create a copy of that frag and release the original 2785 * page by calling the destructor. 2786 */ 2787 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2788 { 2789 if (likely(!skb_zcopy(skb))) 2790 return 0; 2791 if (!skb_zcopy_is_nouarg(skb) && 2792 skb_uarg(skb)->callback == msg_zerocopy_callback) 2793 return 0; 2794 return skb_copy_ubufs(skb, gfp_mask); 2795 } 2796 2797 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2798 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2799 { 2800 if (likely(!skb_zcopy(skb))) 2801 return 0; 2802 return skb_copy_ubufs(skb, gfp_mask); 2803 } 2804 2805 /** 2806 * __skb_queue_purge - empty a list 2807 * @list: list to empty 2808 * 2809 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2810 * the list and one reference dropped. This function does not take the 2811 * list lock and the caller must hold the relevant locks to use it. 2812 */ 2813 static inline void __skb_queue_purge(struct sk_buff_head *list) 2814 { 2815 struct sk_buff *skb; 2816 while ((skb = __skb_dequeue(list)) != NULL) 2817 kfree_skb(skb); 2818 } 2819 void skb_queue_purge(struct sk_buff_head *list); 2820 2821 unsigned int skb_rbtree_purge(struct rb_root *root); 2822 2823 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2824 2825 /** 2826 * netdev_alloc_frag - allocate a page fragment 2827 * @fragsz: fragment size 2828 * 2829 * Allocates a frag from a page for receive buffer. 2830 * Uses GFP_ATOMIC allocations. 2831 */ 2832 static inline void *netdev_alloc_frag(unsigned int fragsz) 2833 { 2834 return __netdev_alloc_frag_align(fragsz, ~0u); 2835 } 2836 2837 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 2838 unsigned int align) 2839 { 2840 WARN_ON_ONCE(!is_power_of_2(align)); 2841 return __netdev_alloc_frag_align(fragsz, -align); 2842 } 2843 2844 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2845 gfp_t gfp_mask); 2846 2847 /** 2848 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2849 * @dev: network device to receive on 2850 * @length: length to allocate 2851 * 2852 * Allocate a new &sk_buff and assign it a usage count of one. The 2853 * buffer has unspecified headroom built in. Users should allocate 2854 * the headroom they think they need without accounting for the 2855 * built in space. The built in space is used for optimisations. 2856 * 2857 * %NULL is returned if there is no free memory. Although this function 2858 * allocates memory it can be called from an interrupt. 2859 */ 2860 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2861 unsigned int length) 2862 { 2863 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2864 } 2865 2866 /* legacy helper around __netdev_alloc_skb() */ 2867 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2868 gfp_t gfp_mask) 2869 { 2870 return __netdev_alloc_skb(NULL, length, gfp_mask); 2871 } 2872 2873 /* legacy helper around netdev_alloc_skb() */ 2874 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2875 { 2876 return netdev_alloc_skb(NULL, length); 2877 } 2878 2879 2880 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2881 unsigned int length, gfp_t gfp) 2882 { 2883 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2884 2885 if (NET_IP_ALIGN && skb) 2886 skb_reserve(skb, NET_IP_ALIGN); 2887 return skb; 2888 } 2889 2890 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2891 unsigned int length) 2892 { 2893 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2894 } 2895 2896 static inline void skb_free_frag(void *addr) 2897 { 2898 page_frag_free(addr); 2899 } 2900 2901 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2902 2903 static inline void *napi_alloc_frag(unsigned int fragsz) 2904 { 2905 return __napi_alloc_frag_align(fragsz, ~0u); 2906 } 2907 2908 static inline void *napi_alloc_frag_align(unsigned int fragsz, 2909 unsigned int align) 2910 { 2911 WARN_ON_ONCE(!is_power_of_2(align)); 2912 return __napi_alloc_frag_align(fragsz, -align); 2913 } 2914 2915 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2916 unsigned int length, gfp_t gfp_mask); 2917 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2918 unsigned int length) 2919 { 2920 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2921 } 2922 void napi_consume_skb(struct sk_buff *skb, int budget); 2923 2924 void napi_skb_free_stolen_head(struct sk_buff *skb); 2925 void __kfree_skb_defer(struct sk_buff *skb); 2926 2927 /** 2928 * __dev_alloc_pages - allocate page for network Rx 2929 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2930 * @order: size of the allocation 2931 * 2932 * Allocate a new page. 2933 * 2934 * %NULL is returned if there is no free memory. 2935 */ 2936 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2937 unsigned int order) 2938 { 2939 /* This piece of code contains several assumptions. 2940 * 1. This is for device Rx, therefor a cold page is preferred. 2941 * 2. The expectation is the user wants a compound page. 2942 * 3. If requesting a order 0 page it will not be compound 2943 * due to the check to see if order has a value in prep_new_page 2944 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2945 * code in gfp_to_alloc_flags that should be enforcing this. 2946 */ 2947 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2948 2949 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2950 } 2951 2952 static inline struct page *dev_alloc_pages(unsigned int order) 2953 { 2954 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2955 } 2956 2957 /** 2958 * __dev_alloc_page - allocate a page for network Rx 2959 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2960 * 2961 * Allocate a new page. 2962 * 2963 * %NULL is returned if there is no free memory. 2964 */ 2965 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2966 { 2967 return __dev_alloc_pages(gfp_mask, 0); 2968 } 2969 2970 static inline struct page *dev_alloc_page(void) 2971 { 2972 return dev_alloc_pages(0); 2973 } 2974 2975 /** 2976 * dev_page_is_reusable - check whether a page can be reused for network Rx 2977 * @page: the page to test 2978 * 2979 * A page shouldn't be considered for reusing/recycling if it was allocated 2980 * under memory pressure or at a distant memory node. 2981 * 2982 * Returns false if this page should be returned to page allocator, true 2983 * otherwise. 2984 */ 2985 static inline bool dev_page_is_reusable(const struct page *page) 2986 { 2987 return likely(page_to_nid(page) == numa_mem_id() && 2988 !page_is_pfmemalloc(page)); 2989 } 2990 2991 /** 2992 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2993 * @page: The page that was allocated from skb_alloc_page 2994 * @skb: The skb that may need pfmemalloc set 2995 */ 2996 static inline void skb_propagate_pfmemalloc(const struct page *page, 2997 struct sk_buff *skb) 2998 { 2999 if (page_is_pfmemalloc(page)) 3000 skb->pfmemalloc = true; 3001 } 3002 3003 /** 3004 * skb_frag_off() - Returns the offset of a skb fragment 3005 * @frag: the paged fragment 3006 */ 3007 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3008 { 3009 return frag->bv_offset; 3010 } 3011 3012 /** 3013 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3014 * @frag: skb fragment 3015 * @delta: value to add 3016 */ 3017 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3018 { 3019 frag->bv_offset += delta; 3020 } 3021 3022 /** 3023 * skb_frag_off_set() - Sets the offset of a skb fragment 3024 * @frag: skb fragment 3025 * @offset: offset of fragment 3026 */ 3027 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3028 { 3029 frag->bv_offset = offset; 3030 } 3031 3032 /** 3033 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3034 * @fragto: skb fragment where offset is set 3035 * @fragfrom: skb fragment offset is copied from 3036 */ 3037 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3038 const skb_frag_t *fragfrom) 3039 { 3040 fragto->bv_offset = fragfrom->bv_offset; 3041 } 3042 3043 /** 3044 * skb_frag_page - retrieve the page referred to by a paged fragment 3045 * @frag: the paged fragment 3046 * 3047 * Returns the &struct page associated with @frag. 3048 */ 3049 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3050 { 3051 return frag->bv_page; 3052 } 3053 3054 /** 3055 * __skb_frag_ref - take an addition reference on a paged fragment. 3056 * @frag: the paged fragment 3057 * 3058 * Takes an additional reference on the paged fragment @frag. 3059 */ 3060 static inline void __skb_frag_ref(skb_frag_t *frag) 3061 { 3062 get_page(skb_frag_page(frag)); 3063 } 3064 3065 /** 3066 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3067 * @skb: the buffer 3068 * @f: the fragment offset. 3069 * 3070 * Takes an additional reference on the @f'th paged fragment of @skb. 3071 */ 3072 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3073 { 3074 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3075 } 3076 3077 /** 3078 * __skb_frag_unref - release a reference on a paged fragment. 3079 * @frag: the paged fragment 3080 * 3081 * Releases a reference on the paged fragment @frag. 3082 */ 3083 static inline void __skb_frag_unref(skb_frag_t *frag) 3084 { 3085 put_page(skb_frag_page(frag)); 3086 } 3087 3088 /** 3089 * skb_frag_unref - release a reference on a paged fragment of an skb. 3090 * @skb: the buffer 3091 * @f: the fragment offset 3092 * 3093 * Releases a reference on the @f'th paged fragment of @skb. 3094 */ 3095 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3096 { 3097 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 3098 } 3099 3100 /** 3101 * skb_frag_address - gets the address of the data contained in a paged fragment 3102 * @frag: the paged fragment buffer 3103 * 3104 * Returns the address of the data within @frag. The page must already 3105 * be mapped. 3106 */ 3107 static inline void *skb_frag_address(const skb_frag_t *frag) 3108 { 3109 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3110 } 3111 3112 /** 3113 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3114 * @frag: the paged fragment buffer 3115 * 3116 * Returns the address of the data within @frag. Checks that the page 3117 * is mapped and returns %NULL otherwise. 3118 */ 3119 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3120 { 3121 void *ptr = page_address(skb_frag_page(frag)); 3122 if (unlikely(!ptr)) 3123 return NULL; 3124 3125 return ptr + skb_frag_off(frag); 3126 } 3127 3128 /** 3129 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3130 * @fragto: skb fragment where page is set 3131 * @fragfrom: skb fragment page is copied from 3132 */ 3133 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3134 const skb_frag_t *fragfrom) 3135 { 3136 fragto->bv_page = fragfrom->bv_page; 3137 } 3138 3139 /** 3140 * __skb_frag_set_page - sets the page contained in a paged fragment 3141 * @frag: the paged fragment 3142 * @page: the page to set 3143 * 3144 * Sets the fragment @frag to contain @page. 3145 */ 3146 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3147 { 3148 frag->bv_page = page; 3149 } 3150 3151 /** 3152 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3153 * @skb: the buffer 3154 * @f: the fragment offset 3155 * @page: the page to set 3156 * 3157 * Sets the @f'th fragment of @skb to contain @page. 3158 */ 3159 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3160 struct page *page) 3161 { 3162 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3163 } 3164 3165 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3166 3167 /** 3168 * skb_frag_dma_map - maps a paged fragment via the DMA API 3169 * @dev: the device to map the fragment to 3170 * @frag: the paged fragment to map 3171 * @offset: the offset within the fragment (starting at the 3172 * fragment's own offset) 3173 * @size: the number of bytes to map 3174 * @dir: the direction of the mapping (``PCI_DMA_*``) 3175 * 3176 * Maps the page associated with @frag to @device. 3177 */ 3178 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3179 const skb_frag_t *frag, 3180 size_t offset, size_t size, 3181 enum dma_data_direction dir) 3182 { 3183 return dma_map_page(dev, skb_frag_page(frag), 3184 skb_frag_off(frag) + offset, size, dir); 3185 } 3186 3187 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3188 gfp_t gfp_mask) 3189 { 3190 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3191 } 3192 3193 3194 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3195 gfp_t gfp_mask) 3196 { 3197 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3198 } 3199 3200 3201 /** 3202 * skb_clone_writable - is the header of a clone writable 3203 * @skb: buffer to check 3204 * @len: length up to which to write 3205 * 3206 * Returns true if modifying the header part of the cloned buffer 3207 * does not requires the data to be copied. 3208 */ 3209 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3210 { 3211 return !skb_header_cloned(skb) && 3212 skb_headroom(skb) + len <= skb->hdr_len; 3213 } 3214 3215 static inline int skb_try_make_writable(struct sk_buff *skb, 3216 unsigned int write_len) 3217 { 3218 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3219 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3220 } 3221 3222 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3223 int cloned) 3224 { 3225 int delta = 0; 3226 3227 if (headroom > skb_headroom(skb)) 3228 delta = headroom - skb_headroom(skb); 3229 3230 if (delta || cloned) 3231 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3232 GFP_ATOMIC); 3233 return 0; 3234 } 3235 3236 /** 3237 * skb_cow - copy header of skb when it is required 3238 * @skb: buffer to cow 3239 * @headroom: needed headroom 3240 * 3241 * If the skb passed lacks sufficient headroom or its data part 3242 * is shared, data is reallocated. If reallocation fails, an error 3243 * is returned and original skb is not changed. 3244 * 3245 * The result is skb with writable area skb->head...skb->tail 3246 * and at least @headroom of space at head. 3247 */ 3248 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3249 { 3250 return __skb_cow(skb, headroom, skb_cloned(skb)); 3251 } 3252 3253 /** 3254 * skb_cow_head - skb_cow but only making the head writable 3255 * @skb: buffer to cow 3256 * @headroom: needed headroom 3257 * 3258 * This function is identical to skb_cow except that we replace the 3259 * skb_cloned check by skb_header_cloned. It should be used when 3260 * you only need to push on some header and do not need to modify 3261 * the data. 3262 */ 3263 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3264 { 3265 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3266 } 3267 3268 /** 3269 * skb_padto - pad an skbuff up to a minimal size 3270 * @skb: buffer to pad 3271 * @len: minimal length 3272 * 3273 * Pads up a buffer to ensure the trailing bytes exist and are 3274 * blanked. If the buffer already contains sufficient data it 3275 * is untouched. Otherwise it is extended. Returns zero on 3276 * success. The skb is freed on error. 3277 */ 3278 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3279 { 3280 unsigned int size = skb->len; 3281 if (likely(size >= len)) 3282 return 0; 3283 return skb_pad(skb, len - size); 3284 } 3285 3286 /** 3287 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3288 * @skb: buffer to pad 3289 * @len: minimal length 3290 * @free_on_error: free buffer on error 3291 * 3292 * Pads up a buffer to ensure the trailing bytes exist and are 3293 * blanked. If the buffer already contains sufficient data it 3294 * is untouched. Otherwise it is extended. Returns zero on 3295 * success. The skb is freed on error if @free_on_error is true. 3296 */ 3297 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3298 unsigned int len, 3299 bool free_on_error) 3300 { 3301 unsigned int size = skb->len; 3302 3303 if (unlikely(size < len)) { 3304 len -= size; 3305 if (__skb_pad(skb, len, free_on_error)) 3306 return -ENOMEM; 3307 __skb_put(skb, len); 3308 } 3309 return 0; 3310 } 3311 3312 /** 3313 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3314 * @skb: buffer to pad 3315 * @len: minimal length 3316 * 3317 * Pads up a buffer to ensure the trailing bytes exist and are 3318 * blanked. If the buffer already contains sufficient data it 3319 * is untouched. Otherwise it is extended. Returns zero on 3320 * success. The skb is freed on error. 3321 */ 3322 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3323 { 3324 return __skb_put_padto(skb, len, true); 3325 } 3326 3327 static inline int skb_add_data(struct sk_buff *skb, 3328 struct iov_iter *from, int copy) 3329 { 3330 const int off = skb->len; 3331 3332 if (skb->ip_summed == CHECKSUM_NONE) { 3333 __wsum csum = 0; 3334 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3335 &csum, from)) { 3336 skb->csum = csum_block_add(skb->csum, csum, off); 3337 return 0; 3338 } 3339 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3340 return 0; 3341 3342 __skb_trim(skb, off); 3343 return -EFAULT; 3344 } 3345 3346 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3347 const struct page *page, int off) 3348 { 3349 if (skb_zcopy(skb)) 3350 return false; 3351 if (i) { 3352 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3353 3354 return page == skb_frag_page(frag) && 3355 off == skb_frag_off(frag) + skb_frag_size(frag); 3356 } 3357 return false; 3358 } 3359 3360 static inline int __skb_linearize(struct sk_buff *skb) 3361 { 3362 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3363 } 3364 3365 /** 3366 * skb_linearize - convert paged skb to linear one 3367 * @skb: buffer to linarize 3368 * 3369 * If there is no free memory -ENOMEM is returned, otherwise zero 3370 * is returned and the old skb data released. 3371 */ 3372 static inline int skb_linearize(struct sk_buff *skb) 3373 { 3374 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3375 } 3376 3377 /** 3378 * skb_has_shared_frag - can any frag be overwritten 3379 * @skb: buffer to test 3380 * 3381 * Return true if the skb has at least one frag that might be modified 3382 * by an external entity (as in vmsplice()/sendfile()) 3383 */ 3384 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3385 { 3386 return skb_is_nonlinear(skb) && 3387 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3388 } 3389 3390 /** 3391 * skb_linearize_cow - make sure skb is linear and writable 3392 * @skb: buffer to process 3393 * 3394 * If there is no free memory -ENOMEM is returned, otherwise zero 3395 * is returned and the old skb data released. 3396 */ 3397 static inline int skb_linearize_cow(struct sk_buff *skb) 3398 { 3399 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3400 __skb_linearize(skb) : 0; 3401 } 3402 3403 static __always_inline void 3404 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3405 unsigned int off) 3406 { 3407 if (skb->ip_summed == CHECKSUM_COMPLETE) 3408 skb->csum = csum_block_sub(skb->csum, 3409 csum_partial(start, len, 0), off); 3410 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3411 skb_checksum_start_offset(skb) < 0) 3412 skb->ip_summed = CHECKSUM_NONE; 3413 } 3414 3415 /** 3416 * skb_postpull_rcsum - update checksum for received skb after pull 3417 * @skb: buffer to update 3418 * @start: start of data before pull 3419 * @len: length of data pulled 3420 * 3421 * After doing a pull on a received packet, you need to call this to 3422 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3423 * CHECKSUM_NONE so that it can be recomputed from scratch. 3424 */ 3425 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3426 const void *start, unsigned int len) 3427 { 3428 __skb_postpull_rcsum(skb, start, len, 0); 3429 } 3430 3431 static __always_inline void 3432 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3433 unsigned int off) 3434 { 3435 if (skb->ip_summed == CHECKSUM_COMPLETE) 3436 skb->csum = csum_block_add(skb->csum, 3437 csum_partial(start, len, 0), off); 3438 } 3439 3440 /** 3441 * skb_postpush_rcsum - update checksum for received skb after push 3442 * @skb: buffer to update 3443 * @start: start of data after push 3444 * @len: length of data pushed 3445 * 3446 * After doing a push on a received packet, you need to call this to 3447 * update the CHECKSUM_COMPLETE checksum. 3448 */ 3449 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3450 const void *start, unsigned int len) 3451 { 3452 __skb_postpush_rcsum(skb, start, len, 0); 3453 } 3454 3455 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3456 3457 /** 3458 * skb_push_rcsum - push skb and update receive checksum 3459 * @skb: buffer to update 3460 * @len: length of data pulled 3461 * 3462 * This function performs an skb_push on the packet and updates 3463 * the CHECKSUM_COMPLETE checksum. It should be used on 3464 * receive path processing instead of skb_push unless you know 3465 * that the checksum difference is zero (e.g., a valid IP header) 3466 * or you are setting ip_summed to CHECKSUM_NONE. 3467 */ 3468 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3469 { 3470 skb_push(skb, len); 3471 skb_postpush_rcsum(skb, skb->data, len); 3472 return skb->data; 3473 } 3474 3475 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3476 /** 3477 * pskb_trim_rcsum - trim received skb and update checksum 3478 * @skb: buffer to trim 3479 * @len: new length 3480 * 3481 * This is exactly the same as pskb_trim except that it ensures the 3482 * checksum of received packets are still valid after the operation. 3483 * It can change skb pointers. 3484 */ 3485 3486 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3487 { 3488 if (likely(len >= skb->len)) 3489 return 0; 3490 return pskb_trim_rcsum_slow(skb, len); 3491 } 3492 3493 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3494 { 3495 if (skb->ip_summed == CHECKSUM_COMPLETE) 3496 skb->ip_summed = CHECKSUM_NONE; 3497 __skb_trim(skb, len); 3498 return 0; 3499 } 3500 3501 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3502 { 3503 if (skb->ip_summed == CHECKSUM_COMPLETE) 3504 skb->ip_summed = CHECKSUM_NONE; 3505 return __skb_grow(skb, len); 3506 } 3507 3508 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3509 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3510 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3511 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3512 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3513 3514 #define skb_queue_walk(queue, skb) \ 3515 for (skb = (queue)->next; \ 3516 skb != (struct sk_buff *)(queue); \ 3517 skb = skb->next) 3518 3519 #define skb_queue_walk_safe(queue, skb, tmp) \ 3520 for (skb = (queue)->next, tmp = skb->next; \ 3521 skb != (struct sk_buff *)(queue); \ 3522 skb = tmp, tmp = skb->next) 3523 3524 #define skb_queue_walk_from(queue, skb) \ 3525 for (; skb != (struct sk_buff *)(queue); \ 3526 skb = skb->next) 3527 3528 #define skb_rbtree_walk(skb, root) \ 3529 for (skb = skb_rb_first(root); skb != NULL; \ 3530 skb = skb_rb_next(skb)) 3531 3532 #define skb_rbtree_walk_from(skb) \ 3533 for (; skb != NULL; \ 3534 skb = skb_rb_next(skb)) 3535 3536 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3537 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3538 skb = tmp) 3539 3540 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3541 for (tmp = skb->next; \ 3542 skb != (struct sk_buff *)(queue); \ 3543 skb = tmp, tmp = skb->next) 3544 3545 #define skb_queue_reverse_walk(queue, skb) \ 3546 for (skb = (queue)->prev; \ 3547 skb != (struct sk_buff *)(queue); \ 3548 skb = skb->prev) 3549 3550 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3551 for (skb = (queue)->prev, tmp = skb->prev; \ 3552 skb != (struct sk_buff *)(queue); \ 3553 skb = tmp, tmp = skb->prev) 3554 3555 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3556 for (tmp = skb->prev; \ 3557 skb != (struct sk_buff *)(queue); \ 3558 skb = tmp, tmp = skb->prev) 3559 3560 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3561 { 3562 return skb_shinfo(skb)->frag_list != NULL; 3563 } 3564 3565 static inline void skb_frag_list_init(struct sk_buff *skb) 3566 { 3567 skb_shinfo(skb)->frag_list = NULL; 3568 } 3569 3570 #define skb_walk_frags(skb, iter) \ 3571 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3572 3573 3574 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3575 int *err, long *timeo_p, 3576 const struct sk_buff *skb); 3577 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3578 struct sk_buff_head *queue, 3579 unsigned int flags, 3580 int *off, int *err, 3581 struct sk_buff **last); 3582 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3583 struct sk_buff_head *queue, 3584 unsigned int flags, int *off, int *err, 3585 struct sk_buff **last); 3586 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3587 struct sk_buff_head *sk_queue, 3588 unsigned int flags, int *off, int *err); 3589 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3590 int *err); 3591 __poll_t datagram_poll(struct file *file, struct socket *sock, 3592 struct poll_table_struct *wait); 3593 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3594 struct iov_iter *to, int size); 3595 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3596 struct msghdr *msg, int size) 3597 { 3598 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3599 } 3600 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3601 struct msghdr *msg); 3602 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3603 struct iov_iter *to, int len, 3604 struct ahash_request *hash); 3605 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3606 struct iov_iter *from, int len); 3607 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3608 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3609 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3610 static inline void skb_free_datagram_locked(struct sock *sk, 3611 struct sk_buff *skb) 3612 { 3613 __skb_free_datagram_locked(sk, skb, 0); 3614 } 3615 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3616 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3617 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3618 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3619 int len); 3620 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3621 struct pipe_inode_info *pipe, unsigned int len, 3622 unsigned int flags); 3623 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3624 int len); 3625 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3626 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3627 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3628 int len, int hlen); 3629 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3630 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3631 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3632 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3633 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3634 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3635 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3636 unsigned int offset); 3637 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3638 int skb_ensure_writable(struct sk_buff *skb, int write_len); 3639 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3640 int skb_vlan_pop(struct sk_buff *skb); 3641 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3642 int skb_eth_pop(struct sk_buff *skb); 3643 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3644 const unsigned char *src); 3645 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3646 int mac_len, bool ethernet); 3647 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3648 bool ethernet); 3649 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3650 int skb_mpls_dec_ttl(struct sk_buff *skb); 3651 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3652 gfp_t gfp); 3653 3654 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3655 { 3656 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3657 } 3658 3659 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3660 { 3661 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3662 } 3663 3664 struct skb_checksum_ops { 3665 __wsum (*update)(const void *mem, int len, __wsum wsum); 3666 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3667 }; 3668 3669 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3670 3671 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3672 __wsum csum, const struct skb_checksum_ops *ops); 3673 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3674 __wsum csum); 3675 3676 static inline void * __must_check 3677 __skb_header_pointer(const struct sk_buff *skb, int offset, 3678 int len, void *data, int hlen, void *buffer) 3679 { 3680 if (hlen - offset >= len) 3681 return data + offset; 3682 3683 if (!skb || 3684 skb_copy_bits(skb, offset, buffer, len) < 0) 3685 return NULL; 3686 3687 return buffer; 3688 } 3689 3690 static inline void * __must_check 3691 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3692 { 3693 return __skb_header_pointer(skb, offset, len, skb->data, 3694 skb_headlen(skb), buffer); 3695 } 3696 3697 /** 3698 * skb_needs_linearize - check if we need to linearize a given skb 3699 * depending on the given device features. 3700 * @skb: socket buffer to check 3701 * @features: net device features 3702 * 3703 * Returns true if either: 3704 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3705 * 2. skb is fragmented and the device does not support SG. 3706 */ 3707 static inline bool skb_needs_linearize(struct sk_buff *skb, 3708 netdev_features_t features) 3709 { 3710 return skb_is_nonlinear(skb) && 3711 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3712 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3713 } 3714 3715 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3716 void *to, 3717 const unsigned int len) 3718 { 3719 memcpy(to, skb->data, len); 3720 } 3721 3722 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3723 const int offset, void *to, 3724 const unsigned int len) 3725 { 3726 memcpy(to, skb->data + offset, len); 3727 } 3728 3729 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3730 const void *from, 3731 const unsigned int len) 3732 { 3733 memcpy(skb->data, from, len); 3734 } 3735 3736 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3737 const int offset, 3738 const void *from, 3739 const unsigned int len) 3740 { 3741 memcpy(skb->data + offset, from, len); 3742 } 3743 3744 void skb_init(void); 3745 3746 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3747 { 3748 return skb->tstamp; 3749 } 3750 3751 /** 3752 * skb_get_timestamp - get timestamp from a skb 3753 * @skb: skb to get stamp from 3754 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3755 * 3756 * Timestamps are stored in the skb as offsets to a base timestamp. 3757 * This function converts the offset back to a struct timeval and stores 3758 * it in stamp. 3759 */ 3760 static inline void skb_get_timestamp(const struct sk_buff *skb, 3761 struct __kernel_old_timeval *stamp) 3762 { 3763 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3764 } 3765 3766 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3767 struct __kernel_sock_timeval *stamp) 3768 { 3769 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3770 3771 stamp->tv_sec = ts.tv_sec; 3772 stamp->tv_usec = ts.tv_nsec / 1000; 3773 } 3774 3775 static inline void skb_get_timestampns(const struct sk_buff *skb, 3776 struct __kernel_old_timespec *stamp) 3777 { 3778 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3779 3780 stamp->tv_sec = ts.tv_sec; 3781 stamp->tv_nsec = ts.tv_nsec; 3782 } 3783 3784 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3785 struct __kernel_timespec *stamp) 3786 { 3787 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3788 3789 stamp->tv_sec = ts.tv_sec; 3790 stamp->tv_nsec = ts.tv_nsec; 3791 } 3792 3793 static inline void __net_timestamp(struct sk_buff *skb) 3794 { 3795 skb->tstamp = ktime_get_real(); 3796 } 3797 3798 static inline ktime_t net_timedelta(ktime_t t) 3799 { 3800 return ktime_sub(ktime_get_real(), t); 3801 } 3802 3803 static inline ktime_t net_invalid_timestamp(void) 3804 { 3805 return 0; 3806 } 3807 3808 static inline u8 skb_metadata_len(const struct sk_buff *skb) 3809 { 3810 return skb_shinfo(skb)->meta_len; 3811 } 3812 3813 static inline void *skb_metadata_end(const struct sk_buff *skb) 3814 { 3815 return skb_mac_header(skb); 3816 } 3817 3818 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3819 const struct sk_buff *skb_b, 3820 u8 meta_len) 3821 { 3822 const void *a = skb_metadata_end(skb_a); 3823 const void *b = skb_metadata_end(skb_b); 3824 /* Using more efficient varaiant than plain call to memcmp(). */ 3825 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3826 u64 diffs = 0; 3827 3828 switch (meta_len) { 3829 #define __it(x, op) (x -= sizeof(u##op)) 3830 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3831 case 32: diffs |= __it_diff(a, b, 64); 3832 fallthrough; 3833 case 24: diffs |= __it_diff(a, b, 64); 3834 fallthrough; 3835 case 16: diffs |= __it_diff(a, b, 64); 3836 fallthrough; 3837 case 8: diffs |= __it_diff(a, b, 64); 3838 break; 3839 case 28: diffs |= __it_diff(a, b, 64); 3840 fallthrough; 3841 case 20: diffs |= __it_diff(a, b, 64); 3842 fallthrough; 3843 case 12: diffs |= __it_diff(a, b, 64); 3844 fallthrough; 3845 case 4: diffs |= __it_diff(a, b, 32); 3846 break; 3847 } 3848 return diffs; 3849 #else 3850 return memcmp(a - meta_len, b - meta_len, meta_len); 3851 #endif 3852 } 3853 3854 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3855 const struct sk_buff *skb_b) 3856 { 3857 u8 len_a = skb_metadata_len(skb_a); 3858 u8 len_b = skb_metadata_len(skb_b); 3859 3860 if (!(len_a | len_b)) 3861 return false; 3862 3863 return len_a != len_b ? 3864 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3865 } 3866 3867 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3868 { 3869 skb_shinfo(skb)->meta_len = meta_len; 3870 } 3871 3872 static inline void skb_metadata_clear(struct sk_buff *skb) 3873 { 3874 skb_metadata_set(skb, 0); 3875 } 3876 3877 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3878 3879 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3880 3881 void skb_clone_tx_timestamp(struct sk_buff *skb); 3882 bool skb_defer_rx_timestamp(struct sk_buff *skb); 3883 3884 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3885 3886 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3887 { 3888 } 3889 3890 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3891 { 3892 return false; 3893 } 3894 3895 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3896 3897 /** 3898 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3899 * 3900 * PHY drivers may accept clones of transmitted packets for 3901 * timestamping via their phy_driver.txtstamp method. These drivers 3902 * must call this function to return the skb back to the stack with a 3903 * timestamp. 3904 * 3905 * @skb: clone of the original outgoing packet 3906 * @hwtstamps: hardware time stamps 3907 * 3908 */ 3909 void skb_complete_tx_timestamp(struct sk_buff *skb, 3910 struct skb_shared_hwtstamps *hwtstamps); 3911 3912 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 3913 struct skb_shared_hwtstamps *hwtstamps, 3914 struct sock *sk, int tstype); 3915 3916 /** 3917 * skb_tstamp_tx - queue clone of skb with send time stamps 3918 * @orig_skb: the original outgoing packet 3919 * @hwtstamps: hardware time stamps, may be NULL if not available 3920 * 3921 * If the skb has a socket associated, then this function clones the 3922 * skb (thus sharing the actual data and optional structures), stores 3923 * the optional hardware time stamping information (if non NULL) or 3924 * generates a software time stamp (otherwise), then queues the clone 3925 * to the error queue of the socket. Errors are silently ignored. 3926 */ 3927 void skb_tstamp_tx(struct sk_buff *orig_skb, 3928 struct skb_shared_hwtstamps *hwtstamps); 3929 3930 /** 3931 * skb_tx_timestamp() - Driver hook for transmit timestamping 3932 * 3933 * Ethernet MAC Drivers should call this function in their hard_xmit() 3934 * function immediately before giving the sk_buff to the MAC hardware. 3935 * 3936 * Specifically, one should make absolutely sure that this function is 3937 * called before TX completion of this packet can trigger. Otherwise 3938 * the packet could potentially already be freed. 3939 * 3940 * @skb: A socket buffer. 3941 */ 3942 static inline void skb_tx_timestamp(struct sk_buff *skb) 3943 { 3944 skb_clone_tx_timestamp(skb); 3945 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3946 skb_tstamp_tx(skb, NULL); 3947 } 3948 3949 /** 3950 * skb_complete_wifi_ack - deliver skb with wifi status 3951 * 3952 * @skb: the original outgoing packet 3953 * @acked: ack status 3954 * 3955 */ 3956 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3957 3958 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3959 __sum16 __skb_checksum_complete(struct sk_buff *skb); 3960 3961 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3962 { 3963 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3964 skb->csum_valid || 3965 (skb->ip_summed == CHECKSUM_PARTIAL && 3966 skb_checksum_start_offset(skb) >= 0)); 3967 } 3968 3969 /** 3970 * skb_checksum_complete - Calculate checksum of an entire packet 3971 * @skb: packet to process 3972 * 3973 * This function calculates the checksum over the entire packet plus 3974 * the value of skb->csum. The latter can be used to supply the 3975 * checksum of a pseudo header as used by TCP/UDP. It returns the 3976 * checksum. 3977 * 3978 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 3979 * this function can be used to verify that checksum on received 3980 * packets. In that case the function should return zero if the 3981 * checksum is correct. In particular, this function will return zero 3982 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 3983 * hardware has already verified the correctness of the checksum. 3984 */ 3985 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 3986 { 3987 return skb_csum_unnecessary(skb) ? 3988 0 : __skb_checksum_complete(skb); 3989 } 3990 3991 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 3992 { 3993 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3994 if (skb->csum_level == 0) 3995 skb->ip_summed = CHECKSUM_NONE; 3996 else 3997 skb->csum_level--; 3998 } 3999 } 4000 4001 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4002 { 4003 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4004 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4005 skb->csum_level++; 4006 } else if (skb->ip_summed == CHECKSUM_NONE) { 4007 skb->ip_summed = CHECKSUM_UNNECESSARY; 4008 skb->csum_level = 0; 4009 } 4010 } 4011 4012 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4013 { 4014 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4015 skb->ip_summed = CHECKSUM_NONE; 4016 skb->csum_level = 0; 4017 } 4018 } 4019 4020 /* Check if we need to perform checksum complete validation. 4021 * 4022 * Returns true if checksum complete is needed, false otherwise 4023 * (either checksum is unnecessary or zero checksum is allowed). 4024 */ 4025 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4026 bool zero_okay, 4027 __sum16 check) 4028 { 4029 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4030 skb->csum_valid = 1; 4031 __skb_decr_checksum_unnecessary(skb); 4032 return false; 4033 } 4034 4035 return true; 4036 } 4037 4038 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4039 * in checksum_init. 4040 */ 4041 #define CHECKSUM_BREAK 76 4042 4043 /* Unset checksum-complete 4044 * 4045 * Unset checksum complete can be done when packet is being modified 4046 * (uncompressed for instance) and checksum-complete value is 4047 * invalidated. 4048 */ 4049 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4050 { 4051 if (skb->ip_summed == CHECKSUM_COMPLETE) 4052 skb->ip_summed = CHECKSUM_NONE; 4053 } 4054 4055 /* Validate (init) checksum based on checksum complete. 4056 * 4057 * Return values: 4058 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4059 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4060 * checksum is stored in skb->csum for use in __skb_checksum_complete 4061 * non-zero: value of invalid checksum 4062 * 4063 */ 4064 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4065 bool complete, 4066 __wsum psum) 4067 { 4068 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4069 if (!csum_fold(csum_add(psum, skb->csum))) { 4070 skb->csum_valid = 1; 4071 return 0; 4072 } 4073 } 4074 4075 skb->csum = psum; 4076 4077 if (complete || skb->len <= CHECKSUM_BREAK) { 4078 __sum16 csum; 4079 4080 csum = __skb_checksum_complete(skb); 4081 skb->csum_valid = !csum; 4082 return csum; 4083 } 4084 4085 return 0; 4086 } 4087 4088 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4089 { 4090 return 0; 4091 } 4092 4093 /* Perform checksum validate (init). Note that this is a macro since we only 4094 * want to calculate the pseudo header which is an input function if necessary. 4095 * First we try to validate without any computation (checksum unnecessary) and 4096 * then calculate based on checksum complete calling the function to compute 4097 * pseudo header. 4098 * 4099 * Return values: 4100 * 0: checksum is validated or try to in skb_checksum_complete 4101 * non-zero: value of invalid checksum 4102 */ 4103 #define __skb_checksum_validate(skb, proto, complete, \ 4104 zero_okay, check, compute_pseudo) \ 4105 ({ \ 4106 __sum16 __ret = 0; \ 4107 skb->csum_valid = 0; \ 4108 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4109 __ret = __skb_checksum_validate_complete(skb, \ 4110 complete, compute_pseudo(skb, proto)); \ 4111 __ret; \ 4112 }) 4113 4114 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4115 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4116 4117 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4118 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4119 4120 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4121 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4122 4123 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4124 compute_pseudo) \ 4125 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4126 4127 #define skb_checksum_simple_validate(skb) \ 4128 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4129 4130 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4131 { 4132 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4133 } 4134 4135 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4136 { 4137 skb->csum = ~pseudo; 4138 skb->ip_summed = CHECKSUM_COMPLETE; 4139 } 4140 4141 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4142 do { \ 4143 if (__skb_checksum_convert_check(skb)) \ 4144 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4145 } while (0) 4146 4147 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4148 u16 start, u16 offset) 4149 { 4150 skb->ip_summed = CHECKSUM_PARTIAL; 4151 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4152 skb->csum_offset = offset - start; 4153 } 4154 4155 /* Update skbuf and packet to reflect the remote checksum offload operation. 4156 * When called, ptr indicates the starting point for skb->csum when 4157 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4158 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4159 */ 4160 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4161 int start, int offset, bool nopartial) 4162 { 4163 __wsum delta; 4164 4165 if (!nopartial) { 4166 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4167 return; 4168 } 4169 4170 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4171 __skb_checksum_complete(skb); 4172 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4173 } 4174 4175 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4176 4177 /* Adjust skb->csum since we changed the packet */ 4178 skb->csum = csum_add(skb->csum, delta); 4179 } 4180 4181 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4182 { 4183 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4184 return (void *)(skb->_nfct & NFCT_PTRMASK); 4185 #else 4186 return NULL; 4187 #endif 4188 } 4189 4190 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4191 { 4192 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4193 return skb->_nfct; 4194 #else 4195 return 0UL; 4196 #endif 4197 } 4198 4199 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4200 { 4201 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4202 skb->_nfct = nfct; 4203 #endif 4204 } 4205 4206 #ifdef CONFIG_SKB_EXTENSIONS 4207 enum skb_ext_id { 4208 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4209 SKB_EXT_BRIDGE_NF, 4210 #endif 4211 #ifdef CONFIG_XFRM 4212 SKB_EXT_SEC_PATH, 4213 #endif 4214 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4215 TC_SKB_EXT, 4216 #endif 4217 #if IS_ENABLED(CONFIG_MPTCP) 4218 SKB_EXT_MPTCP, 4219 #endif 4220 SKB_EXT_NUM, /* must be last */ 4221 }; 4222 4223 /** 4224 * struct skb_ext - sk_buff extensions 4225 * @refcnt: 1 on allocation, deallocated on 0 4226 * @offset: offset to add to @data to obtain extension address 4227 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4228 * @data: start of extension data, variable sized 4229 * 4230 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4231 * to use 'u8' types while allowing up to 2kb worth of extension data. 4232 */ 4233 struct skb_ext { 4234 refcount_t refcnt; 4235 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4236 u8 chunks; /* same */ 4237 char data[] __aligned(8); 4238 }; 4239 4240 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4241 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4242 struct skb_ext *ext); 4243 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4244 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4245 void __skb_ext_put(struct skb_ext *ext); 4246 4247 static inline void skb_ext_put(struct sk_buff *skb) 4248 { 4249 if (skb->active_extensions) 4250 __skb_ext_put(skb->extensions); 4251 } 4252 4253 static inline void __skb_ext_copy(struct sk_buff *dst, 4254 const struct sk_buff *src) 4255 { 4256 dst->active_extensions = src->active_extensions; 4257 4258 if (src->active_extensions) { 4259 struct skb_ext *ext = src->extensions; 4260 4261 refcount_inc(&ext->refcnt); 4262 dst->extensions = ext; 4263 } 4264 } 4265 4266 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4267 { 4268 skb_ext_put(dst); 4269 __skb_ext_copy(dst, src); 4270 } 4271 4272 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4273 { 4274 return !!ext->offset[i]; 4275 } 4276 4277 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4278 { 4279 return skb->active_extensions & (1 << id); 4280 } 4281 4282 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4283 { 4284 if (skb_ext_exist(skb, id)) 4285 __skb_ext_del(skb, id); 4286 } 4287 4288 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4289 { 4290 if (skb_ext_exist(skb, id)) { 4291 struct skb_ext *ext = skb->extensions; 4292 4293 return (void *)ext + (ext->offset[id] << 3); 4294 } 4295 4296 return NULL; 4297 } 4298 4299 static inline void skb_ext_reset(struct sk_buff *skb) 4300 { 4301 if (unlikely(skb->active_extensions)) { 4302 __skb_ext_put(skb->extensions); 4303 skb->active_extensions = 0; 4304 } 4305 } 4306 4307 static inline bool skb_has_extensions(struct sk_buff *skb) 4308 { 4309 return unlikely(skb->active_extensions); 4310 } 4311 #else 4312 static inline void skb_ext_put(struct sk_buff *skb) {} 4313 static inline void skb_ext_reset(struct sk_buff *skb) {} 4314 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4315 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4316 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4317 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4318 #endif /* CONFIG_SKB_EXTENSIONS */ 4319 4320 static inline void nf_reset_ct(struct sk_buff *skb) 4321 { 4322 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4323 nf_conntrack_put(skb_nfct(skb)); 4324 skb->_nfct = 0; 4325 #endif 4326 } 4327 4328 static inline void nf_reset_trace(struct sk_buff *skb) 4329 { 4330 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4331 skb->nf_trace = 0; 4332 #endif 4333 } 4334 4335 static inline void ipvs_reset(struct sk_buff *skb) 4336 { 4337 #if IS_ENABLED(CONFIG_IP_VS) 4338 skb->ipvs_property = 0; 4339 #endif 4340 } 4341 4342 /* Note: This doesn't put any conntrack info in dst. */ 4343 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4344 bool copy) 4345 { 4346 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4347 dst->_nfct = src->_nfct; 4348 nf_conntrack_get(skb_nfct(src)); 4349 #endif 4350 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4351 if (copy) 4352 dst->nf_trace = src->nf_trace; 4353 #endif 4354 } 4355 4356 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4357 { 4358 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4359 nf_conntrack_put(skb_nfct(dst)); 4360 #endif 4361 __nf_copy(dst, src, true); 4362 } 4363 4364 #ifdef CONFIG_NETWORK_SECMARK 4365 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4366 { 4367 to->secmark = from->secmark; 4368 } 4369 4370 static inline void skb_init_secmark(struct sk_buff *skb) 4371 { 4372 skb->secmark = 0; 4373 } 4374 #else 4375 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4376 { } 4377 4378 static inline void skb_init_secmark(struct sk_buff *skb) 4379 { } 4380 #endif 4381 4382 static inline int secpath_exists(const struct sk_buff *skb) 4383 { 4384 #ifdef CONFIG_XFRM 4385 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4386 #else 4387 return 0; 4388 #endif 4389 } 4390 4391 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4392 { 4393 return !skb->destructor && 4394 !secpath_exists(skb) && 4395 !skb_nfct(skb) && 4396 !skb->_skb_refdst && 4397 !skb_has_frag_list(skb); 4398 } 4399 4400 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4401 { 4402 skb->queue_mapping = queue_mapping; 4403 } 4404 4405 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4406 { 4407 return skb->queue_mapping; 4408 } 4409 4410 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4411 { 4412 to->queue_mapping = from->queue_mapping; 4413 } 4414 4415 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4416 { 4417 skb->queue_mapping = rx_queue + 1; 4418 } 4419 4420 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4421 { 4422 return skb->queue_mapping - 1; 4423 } 4424 4425 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4426 { 4427 return skb->queue_mapping != 0; 4428 } 4429 4430 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4431 { 4432 skb->dst_pending_confirm = val; 4433 } 4434 4435 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4436 { 4437 return skb->dst_pending_confirm != 0; 4438 } 4439 4440 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4441 { 4442 #ifdef CONFIG_XFRM 4443 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4444 #else 4445 return NULL; 4446 #endif 4447 } 4448 4449 /* Keeps track of mac header offset relative to skb->head. 4450 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4451 * For non-tunnel skb it points to skb_mac_header() and for 4452 * tunnel skb it points to outer mac header. 4453 * Keeps track of level of encapsulation of network headers. 4454 */ 4455 struct skb_gso_cb { 4456 union { 4457 int mac_offset; 4458 int data_offset; 4459 }; 4460 int encap_level; 4461 __wsum csum; 4462 __u16 csum_start; 4463 }; 4464 #define SKB_GSO_CB_OFFSET 32 4465 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4466 4467 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4468 { 4469 return (skb_mac_header(inner_skb) - inner_skb->head) - 4470 SKB_GSO_CB(inner_skb)->mac_offset; 4471 } 4472 4473 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4474 { 4475 int new_headroom, headroom; 4476 int ret; 4477 4478 headroom = skb_headroom(skb); 4479 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4480 if (ret) 4481 return ret; 4482 4483 new_headroom = skb_headroom(skb); 4484 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4485 return 0; 4486 } 4487 4488 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4489 { 4490 /* Do not update partial checksums if remote checksum is enabled. */ 4491 if (skb->remcsum_offload) 4492 return; 4493 4494 SKB_GSO_CB(skb)->csum = res; 4495 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4496 } 4497 4498 /* Compute the checksum for a gso segment. First compute the checksum value 4499 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4500 * then add in skb->csum (checksum from csum_start to end of packet). 4501 * skb->csum and csum_start are then updated to reflect the checksum of the 4502 * resultant packet starting from the transport header-- the resultant checksum 4503 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4504 * header. 4505 */ 4506 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4507 { 4508 unsigned char *csum_start = skb_transport_header(skb); 4509 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4510 __wsum partial = SKB_GSO_CB(skb)->csum; 4511 4512 SKB_GSO_CB(skb)->csum = res; 4513 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4514 4515 return csum_fold(csum_partial(csum_start, plen, partial)); 4516 } 4517 4518 static inline bool skb_is_gso(const struct sk_buff *skb) 4519 { 4520 return skb_shinfo(skb)->gso_size; 4521 } 4522 4523 /* Note: Should be called only if skb_is_gso(skb) is true */ 4524 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4525 { 4526 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4527 } 4528 4529 /* Note: Should be called only if skb_is_gso(skb) is true */ 4530 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4531 { 4532 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4533 } 4534 4535 /* Note: Should be called only if skb_is_gso(skb) is true */ 4536 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4537 { 4538 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4539 } 4540 4541 static inline void skb_gso_reset(struct sk_buff *skb) 4542 { 4543 skb_shinfo(skb)->gso_size = 0; 4544 skb_shinfo(skb)->gso_segs = 0; 4545 skb_shinfo(skb)->gso_type = 0; 4546 } 4547 4548 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4549 u16 increment) 4550 { 4551 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4552 return; 4553 shinfo->gso_size += increment; 4554 } 4555 4556 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4557 u16 decrement) 4558 { 4559 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4560 return; 4561 shinfo->gso_size -= decrement; 4562 } 4563 4564 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4565 4566 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4567 { 4568 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4569 * wanted then gso_type will be set. */ 4570 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4571 4572 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4573 unlikely(shinfo->gso_type == 0)) { 4574 __skb_warn_lro_forwarding(skb); 4575 return true; 4576 } 4577 return false; 4578 } 4579 4580 static inline void skb_forward_csum(struct sk_buff *skb) 4581 { 4582 /* Unfortunately we don't support this one. Any brave souls? */ 4583 if (skb->ip_summed == CHECKSUM_COMPLETE) 4584 skb->ip_summed = CHECKSUM_NONE; 4585 } 4586 4587 /** 4588 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4589 * @skb: skb to check 4590 * 4591 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4592 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4593 * use this helper, to document places where we make this assertion. 4594 */ 4595 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4596 { 4597 #ifdef DEBUG 4598 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4599 #endif 4600 } 4601 4602 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4603 4604 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4605 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4606 unsigned int transport_len, 4607 __sum16(*skb_chkf)(struct sk_buff *skb)); 4608 4609 /** 4610 * skb_head_is_locked - Determine if the skb->head is locked down 4611 * @skb: skb to check 4612 * 4613 * The head on skbs build around a head frag can be removed if they are 4614 * not cloned. This function returns true if the skb head is locked down 4615 * due to either being allocated via kmalloc, or by being a clone with 4616 * multiple references to the head. 4617 */ 4618 static inline bool skb_head_is_locked(const struct sk_buff *skb) 4619 { 4620 return !skb->head_frag || skb_cloned(skb); 4621 } 4622 4623 /* Local Checksum Offload. 4624 * Compute outer checksum based on the assumption that the 4625 * inner checksum will be offloaded later. 4626 * See Documentation/networking/checksum-offloads.rst for 4627 * explanation of how this works. 4628 * Fill in outer checksum adjustment (e.g. with sum of outer 4629 * pseudo-header) before calling. 4630 * Also ensure that inner checksum is in linear data area. 4631 */ 4632 static inline __wsum lco_csum(struct sk_buff *skb) 4633 { 4634 unsigned char *csum_start = skb_checksum_start(skb); 4635 unsigned char *l4_hdr = skb_transport_header(skb); 4636 __wsum partial; 4637 4638 /* Start with complement of inner checksum adjustment */ 4639 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4640 skb->csum_offset)); 4641 4642 /* Add in checksum of our headers (incl. outer checksum 4643 * adjustment filled in by caller) and return result. 4644 */ 4645 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4646 } 4647 4648 static inline bool skb_is_redirected(const struct sk_buff *skb) 4649 { 4650 #ifdef CONFIG_NET_REDIRECT 4651 return skb->redirected; 4652 #else 4653 return false; 4654 #endif 4655 } 4656 4657 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4658 { 4659 #ifdef CONFIG_NET_REDIRECT 4660 skb->redirected = 1; 4661 skb->from_ingress = from_ingress; 4662 if (skb->from_ingress) 4663 skb->tstamp = 0; 4664 #endif 4665 } 4666 4667 static inline void skb_reset_redirect(struct sk_buff *skb) 4668 { 4669 #ifdef CONFIG_NET_REDIRECT 4670 skb->redirected = 0; 4671 #endif 4672 } 4673 4674 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 4675 { 4676 return skb->csum_not_inet; 4677 } 4678 4679 static inline void skb_set_kcov_handle(struct sk_buff *skb, 4680 const u64 kcov_handle) 4681 { 4682 #ifdef CONFIG_KCOV 4683 skb->kcov_handle = kcov_handle; 4684 #endif 4685 } 4686 4687 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4688 { 4689 #ifdef CONFIG_KCOV 4690 return skb->kcov_handle; 4691 #else 4692 return 0; 4693 #endif 4694 } 4695 4696 #endif /* __KERNEL__ */ 4697 #endif /* _LINUX_SKBUFF_H */ 4698