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