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 #include <net/page_pool.h> 41 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 42 #include <linux/netfilter/nf_conntrack_common.h> 43 #endif 44 45 /* The interface for checksum offload between the stack and networking drivers 46 * is as follows... 47 * 48 * A. IP checksum related features 49 * 50 * Drivers advertise checksum offload capabilities in the features of a device. 51 * From the stack's point of view these are capabilities offered by the driver. 52 * A driver typically only advertises features that it is capable of offloading 53 * to its device. 54 * 55 * The checksum related features are: 56 * 57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one 58 * IP (one's complement) checksum for any combination 59 * of protocols or protocol layering. The checksum is 60 * computed and set in a packet per the CHECKSUM_PARTIAL 61 * interface (see below). 62 * 63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain 64 * TCP or UDP packets over IPv4. These are specifically 65 * unencapsulated packets of the form IPv4|TCP or 66 * IPv4|UDP where the Protocol field in the IPv4 header 67 * is TCP or UDP. The IPv4 header may contain IP options. 68 * This feature cannot be set in features for a device 69 * with NETIF_F_HW_CSUM also set. This feature is being 70 * DEPRECATED (see below). 71 * 72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain 73 * TCP or UDP packets over IPv6. These are specifically 74 * unencapsulated packets of the form IPv6|TCP or 75 * IPv6|UDP where the Next Header field in the IPv6 76 * header is either TCP or UDP. IPv6 extension headers 77 * are not supported with this feature. This feature 78 * cannot be set in features for a device with 79 * NETIF_F_HW_CSUM also set. This feature is being 80 * DEPRECATED (see below). 81 * 82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. 83 * This flag is only used to disable the RX checksum 84 * feature for a device. The stack will accept receive 85 * checksum indication in packets received on a device 86 * regardless of whether NETIF_F_RXCSUM is set. 87 * 88 * B. Checksumming of received packets by device. Indication of checksum 89 * verification is set in skb->ip_summed. Possible values are: 90 * 91 * CHECKSUM_NONE: 92 * 93 * Device did not checksum this packet e.g. due to lack of capabilities. 94 * The packet contains full (though not verified) checksum in packet but 95 * not in skb->csum. Thus, skb->csum is undefined in this case. 96 * 97 * CHECKSUM_UNNECESSARY: 98 * 99 * The hardware you're dealing with doesn't calculate the full checksum 100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 102 * if their checksums are okay. skb->csum is still undefined in this case 103 * though. A driver or device must never modify the checksum field in the 104 * packet even if checksum is verified. 105 * 106 * CHECKSUM_UNNECESSARY is applicable to following protocols: 107 * TCP: IPv6 and IPv4. 108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 109 * zero UDP checksum for either IPv4 or IPv6, the networking stack 110 * may perform further validation in this case. 111 * GRE: only if the checksum is present in the header. 112 * SCTP: indicates the CRC in SCTP header has been validated. 113 * FCOE: indicates the CRC in FC frame has been validated. 114 * 115 * skb->csum_level indicates the number of consecutive checksums found in 116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 118 * and a device is able to verify the checksums for UDP (possibly zero), 119 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to 120 * two. If the device were only able to verify the UDP checksum and not 121 * GRE, either because it doesn't support GRE checksum or because GRE 122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 123 * not considered in this case). 124 * 125 * CHECKSUM_COMPLETE: 126 * 127 * This is the most generic way. The device supplied checksum of the _whole_ 128 * packet as seen by netif_rx() and fills in skb->csum. This means the 129 * hardware doesn't need to parse L3/L4 headers to implement this. 130 * 131 * Notes: 132 * - Even if device supports only some protocols, but is able to produce 133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 135 * 136 * CHECKSUM_PARTIAL: 137 * 138 * A checksum is set up to be offloaded to a device as described in the 139 * output description for CHECKSUM_PARTIAL. This may occur on a packet 140 * received directly from another Linux OS, e.g., a virtualized Linux kernel 141 * on the same host, or it may be set in the input path in GRO or remote 142 * checksum offload. For the purposes of checksum verification, the checksum 143 * referred to by skb->csum_start + skb->csum_offset and any preceding 144 * checksums in the packet are considered verified. Any checksums in the 145 * packet that are after the checksum being offloaded are not considered to 146 * be verified. 147 * 148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload 149 * in the skb->ip_summed for a packet. Values are: 150 * 151 * CHECKSUM_PARTIAL: 152 * 153 * The driver is required to checksum the packet as seen by hard_start_xmit() 154 * from skb->csum_start up to the end, and to record/write the checksum at 155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the 156 * csum_start and csum_offset values are valid values given the length and 157 * offset of the packet, but it should not attempt to validate that the 158 * checksum refers to a legitimate transport layer checksum -- it is the 159 * purview of the stack to validate that csum_start and csum_offset are set 160 * correctly. 161 * 162 * When the stack requests checksum offload for a packet, the driver MUST 163 * ensure that the checksum is set correctly. A driver can either offload the 164 * checksum calculation to the device, or call skb_checksum_help (in the case 165 * that the device does not support offload for a particular checksum). 166 * 167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of 168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate 169 * checksum offload capability. 170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based 171 * on network device checksumming capabilities: if a packet does not match 172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of 173 * csum_not_inet, see item D.) is called to resolve the checksum. 174 * 175 * CHECKSUM_NONE: 176 * 177 * The skb was already checksummed by the protocol, or a checksum is not 178 * required. 179 * 180 * CHECKSUM_UNNECESSARY: 181 * 182 * This has the same meaning as CHECKSUM_NONE for checksum offload on 183 * output. 184 * 185 * CHECKSUM_COMPLETE: 186 * Not used in checksum output. If a driver observes a packet with this value 187 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set. 188 * 189 * D. Non-IP checksum (CRC) offloads 190 * 191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of 192 * offloading the SCTP CRC in a packet. To perform this offload the stack 193 * will set csum_start and csum_offset accordingly, set ip_summed to 194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in 195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c. 196 * A driver that supports both IP checksum offload and SCTP CRC32c offload 197 * must verify which offload is configured for a packet by testing the 198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve 199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 200 * 201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of 202 * offloading the FCOE CRC in a packet. To perform this offload the stack 203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset 204 * accordingly. Note that there is no indication in the skbuff that the 205 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 206 * both IP checksum offload and FCOE CRC offload must verify which offload 207 * is configured for a packet, presumably by inspecting packet headers. 208 * 209 * E. Checksumming on output with GSO. 210 * 211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload 212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as 214 * part of the GSO operation is implied. If a checksum is being offloaded 215 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and 216 * csum_offset are set to refer to the outermost checksum being offloaded 217 * (two offloaded checksums are possible with UDP encapsulation). 218 */ 219 220 /* Don't change this without changing skb_csum_unnecessary! */ 221 #define CHECKSUM_NONE 0 222 #define CHECKSUM_UNNECESSARY 1 223 #define CHECKSUM_COMPLETE 2 224 #define CHECKSUM_PARTIAL 3 225 226 /* Maximum value in skb->csum_level */ 227 #define SKB_MAX_CSUM_LEVEL 3 228 229 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 230 #define SKB_WITH_OVERHEAD(X) \ 231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 232 #define SKB_MAX_ORDER(X, ORDER) \ 233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 234 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 235 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 236 237 /* return minimum truesize of one skb containing X bytes of data */ 238 #define SKB_TRUESIZE(X) ((X) + \ 239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 241 242 struct ahash_request; 243 struct net_device; 244 struct scatterlist; 245 struct pipe_inode_info; 246 struct iov_iter; 247 struct napi_struct; 248 struct bpf_prog; 249 union bpf_attr; 250 struct skb_ext; 251 252 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 253 struct nf_bridge_info { 254 enum { 255 BRNF_PROTO_UNCHANGED, 256 BRNF_PROTO_8021Q, 257 BRNF_PROTO_PPPOE 258 } orig_proto:8; 259 u8 pkt_otherhost:1; 260 u8 in_prerouting:1; 261 u8 bridged_dnat:1; 262 __u16 frag_max_size; 263 struct net_device *physindev; 264 265 /* always valid & non-NULL from FORWARD on, for physdev match */ 266 struct net_device *physoutdev; 267 union { 268 /* prerouting: detect dnat in orig/reply direction */ 269 __be32 ipv4_daddr; 270 struct in6_addr ipv6_daddr; 271 272 /* after prerouting + nat detected: store original source 273 * mac since neigh resolution overwrites it, only used while 274 * skb is out in neigh layer. 275 */ 276 char neigh_header[8]; 277 }; 278 }; 279 #endif 280 281 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 282 /* Chain in tc_skb_ext will be used to share the tc chain with 283 * ovs recirc_id. It will be set to the current chain by tc 284 * and read by ovs to recirc_id. 285 */ 286 struct tc_skb_ext { 287 __u32 chain; 288 __u16 mru; 289 bool post_ct; 290 }; 291 #endif 292 293 struct sk_buff_head { 294 /* These two members must be first. */ 295 struct sk_buff *next; 296 struct sk_buff *prev; 297 298 __u32 qlen; 299 spinlock_t lock; 300 }; 301 302 struct sk_buff; 303 304 /* To allow 64K frame to be packed as single skb without frag_list we 305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 306 * buffers which do not start on a page boundary. 307 * 308 * Since GRO uses frags we allocate at least 16 regardless of page 309 * size. 310 */ 311 #if (65536/PAGE_SIZE + 1) < 16 312 #define MAX_SKB_FRAGS 16UL 313 #else 314 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 315 #endif 316 extern int sysctl_max_skb_frags; 317 318 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 319 * segment using its current segmentation instead. 320 */ 321 #define GSO_BY_FRAGS 0xFFFF 322 323 typedef struct bio_vec skb_frag_t; 324 325 /** 326 * skb_frag_size() - Returns the size of a skb fragment 327 * @frag: skb fragment 328 */ 329 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 330 { 331 return frag->bv_len; 332 } 333 334 /** 335 * skb_frag_size_set() - Sets the size of a skb fragment 336 * @frag: skb fragment 337 * @size: size of fragment 338 */ 339 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 340 { 341 frag->bv_len = size; 342 } 343 344 /** 345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 346 * @frag: skb fragment 347 * @delta: value to add 348 */ 349 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 350 { 351 frag->bv_len += delta; 352 } 353 354 /** 355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 356 * @frag: skb fragment 357 * @delta: value to subtract 358 */ 359 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 360 { 361 frag->bv_len -= delta; 362 } 363 364 /** 365 * skb_frag_must_loop - Test if %p is a high memory page 366 * @p: fragment's page 367 */ 368 static inline bool skb_frag_must_loop(struct page *p) 369 { 370 #if defined(CONFIG_HIGHMEM) 371 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 372 return true; 373 #endif 374 return false; 375 } 376 377 /** 378 * skb_frag_foreach_page - loop over pages in a fragment 379 * 380 * @f: skb frag to operate on 381 * @f_off: offset from start of f->bv_page 382 * @f_len: length from f_off to loop over 383 * @p: (temp var) current page 384 * @p_off: (temp var) offset from start of current page, 385 * non-zero only on first page. 386 * @p_len: (temp var) length in current page, 387 * < PAGE_SIZE only on first and last page. 388 * @copied: (temp var) length so far, excluding current p_len. 389 * 390 * A fragment can hold a compound page, in which case per-page 391 * operations, notably kmap_atomic, must be called for each 392 * regular page. 393 */ 394 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 396 p_off = (f_off) & (PAGE_SIZE - 1), \ 397 p_len = skb_frag_must_loop(p) ? \ 398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 399 copied = 0; \ 400 copied < f_len; \ 401 copied += p_len, p++, p_off = 0, \ 402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 403 404 #define HAVE_HW_TIME_STAMP 405 406 /** 407 * struct skb_shared_hwtstamps - hardware time stamps 408 * @hwtstamp: hardware time stamp transformed into duration 409 * since arbitrary point in time 410 * 411 * Software time stamps generated by ktime_get_real() are stored in 412 * skb->tstamp. 413 * 414 * hwtstamps can only be compared against other hwtstamps from 415 * the same device. 416 * 417 * This structure is attached to packets as part of the 418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 419 */ 420 struct skb_shared_hwtstamps { 421 ktime_t hwtstamp; 422 }; 423 424 /* Definitions for tx_flags in struct skb_shared_info */ 425 enum { 426 /* generate hardware time stamp */ 427 SKBTX_HW_TSTAMP = 1 << 0, 428 429 /* generate software time stamp when queueing packet to NIC */ 430 SKBTX_SW_TSTAMP = 1 << 1, 431 432 /* device driver is going to provide hardware time stamp */ 433 SKBTX_IN_PROGRESS = 1 << 2, 434 435 /* generate wifi status information (where possible) */ 436 SKBTX_WIFI_STATUS = 1 << 4, 437 438 /* generate software time stamp when entering packet scheduling */ 439 SKBTX_SCHED_TSTAMP = 1 << 6, 440 }; 441 442 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 443 SKBTX_SCHED_TSTAMP) 444 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 445 446 /* Definitions for flags in struct skb_shared_info */ 447 enum { 448 /* use zcopy routines */ 449 SKBFL_ZEROCOPY_ENABLE = BIT(0), 450 451 /* This indicates at least one fragment might be overwritten 452 * (as in vmsplice(), sendfile() ...) 453 * If we need to compute a TX checksum, we'll need to copy 454 * all frags to avoid possible bad checksum 455 */ 456 SKBFL_SHARED_FRAG = BIT(1), 457 }; 458 459 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 460 461 /* 462 * The callback notifies userspace to release buffers when skb DMA is done in 463 * lower device, the skb last reference should be 0 when calling this. 464 * The zerocopy_success argument is true if zero copy transmit occurred, 465 * false on data copy or out of memory error caused by data copy attempt. 466 * The ctx field is used to track device context. 467 * The desc field is used to track userspace buffer index. 468 */ 469 struct ubuf_info { 470 void (*callback)(struct sk_buff *, struct ubuf_info *, 471 bool zerocopy_success); 472 union { 473 struct { 474 unsigned long desc; 475 void *ctx; 476 }; 477 struct { 478 u32 id; 479 u16 len; 480 u16 zerocopy:1; 481 u32 bytelen; 482 }; 483 }; 484 refcount_t refcnt; 485 u8 flags; 486 487 struct mmpin { 488 struct user_struct *user; 489 unsigned int num_pg; 490 } mmp; 491 }; 492 493 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 494 495 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 496 void mm_unaccount_pinned_pages(struct mmpin *mmp); 497 498 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size); 499 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 500 struct ubuf_info *uarg); 501 502 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 503 504 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 505 bool success); 506 507 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len); 508 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 509 struct msghdr *msg, int len, 510 struct ubuf_info *uarg); 511 512 /* This data is invariant across clones and lives at 513 * the end of the header data, ie. at skb->end. 514 */ 515 struct skb_shared_info { 516 __u8 flags; 517 __u8 meta_len; 518 __u8 nr_frags; 519 __u8 tx_flags; 520 unsigned short gso_size; 521 /* Warning: this field is not always filled in (UFO)! */ 522 unsigned short gso_segs; 523 struct sk_buff *frag_list; 524 struct skb_shared_hwtstamps hwtstamps; 525 unsigned int gso_type; 526 u32 tskey; 527 528 /* 529 * Warning : all fields before dataref are cleared in __alloc_skb() 530 */ 531 atomic_t dataref; 532 533 /* Intermediate layers must ensure that destructor_arg 534 * remains valid until skb destructor */ 535 void * destructor_arg; 536 537 /* must be last field, see pskb_expand_head() */ 538 skb_frag_t frags[MAX_SKB_FRAGS]; 539 }; 540 541 /* We divide dataref into two halves. The higher 16 bits hold references 542 * to the payload part of skb->data. The lower 16 bits hold references to 543 * the entire skb->data. A clone of a headerless skb holds the length of 544 * the header in skb->hdr_len. 545 * 546 * All users must obey the rule that the skb->data reference count must be 547 * greater than or equal to the payload reference count. 548 * 549 * Holding a reference to the payload part means that the user does not 550 * care about modifications to the header part of skb->data. 551 */ 552 #define SKB_DATAREF_SHIFT 16 553 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 554 555 556 enum { 557 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 558 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 559 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 560 }; 561 562 enum { 563 SKB_GSO_TCPV4 = 1 << 0, 564 565 /* This indicates the skb is from an untrusted source. */ 566 SKB_GSO_DODGY = 1 << 1, 567 568 /* This indicates the tcp segment has CWR set. */ 569 SKB_GSO_TCP_ECN = 1 << 2, 570 571 SKB_GSO_TCP_FIXEDID = 1 << 3, 572 573 SKB_GSO_TCPV6 = 1 << 4, 574 575 SKB_GSO_FCOE = 1 << 5, 576 577 SKB_GSO_GRE = 1 << 6, 578 579 SKB_GSO_GRE_CSUM = 1 << 7, 580 581 SKB_GSO_IPXIP4 = 1 << 8, 582 583 SKB_GSO_IPXIP6 = 1 << 9, 584 585 SKB_GSO_UDP_TUNNEL = 1 << 10, 586 587 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 588 589 SKB_GSO_PARTIAL = 1 << 12, 590 591 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 592 593 SKB_GSO_SCTP = 1 << 14, 594 595 SKB_GSO_ESP = 1 << 15, 596 597 SKB_GSO_UDP = 1 << 16, 598 599 SKB_GSO_UDP_L4 = 1 << 17, 600 601 SKB_GSO_FRAGLIST = 1 << 18, 602 }; 603 604 #if BITS_PER_LONG > 32 605 #define NET_SKBUFF_DATA_USES_OFFSET 1 606 #endif 607 608 #ifdef NET_SKBUFF_DATA_USES_OFFSET 609 typedef unsigned int sk_buff_data_t; 610 #else 611 typedef unsigned char *sk_buff_data_t; 612 #endif 613 614 /** 615 * struct sk_buff - socket buffer 616 * @next: Next buffer in list 617 * @prev: Previous buffer in list 618 * @tstamp: Time we arrived/left 619 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 620 * for retransmit timer 621 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 622 * @list: queue head 623 * @sk: Socket we are owned by 624 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 625 * fragmentation management 626 * @dev: Device we arrived on/are leaving by 627 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 628 * @cb: Control buffer. Free for use by every layer. Put private vars here 629 * @_skb_refdst: destination entry (with norefcount bit) 630 * @sp: the security path, used for xfrm 631 * @len: Length of actual data 632 * @data_len: Data length 633 * @mac_len: Length of link layer header 634 * @hdr_len: writable header length of cloned skb 635 * @csum: Checksum (must include start/offset pair) 636 * @csum_start: Offset from skb->head where checksumming should start 637 * @csum_offset: Offset from csum_start where checksum should be stored 638 * @priority: Packet queueing priority 639 * @ignore_df: allow local fragmentation 640 * @cloned: Head may be cloned (check refcnt to be sure) 641 * @ip_summed: Driver fed us an IP checksum 642 * @nohdr: Payload reference only, must not modify header 643 * @pkt_type: Packet class 644 * @fclone: skbuff clone status 645 * @ipvs_property: skbuff is owned by ipvs 646 * @inner_protocol_type: whether the inner protocol is 647 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 648 * @remcsum_offload: remote checksum offload is enabled 649 * @offload_fwd_mark: Packet was L2-forwarded in hardware 650 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 651 * @tc_skip_classify: do not classify packet. set by IFB device 652 * @tc_at_ingress: used within tc_classify to distinguish in/egress 653 * @redirected: packet was redirected by packet classifier 654 * @from_ingress: packet was redirected from the ingress path 655 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h 656 * @peeked: this packet has been seen already, so stats have been 657 * done for it, don't do them again 658 * @nf_trace: netfilter packet trace flag 659 * @protocol: Packet protocol from driver 660 * @destructor: Destruct function 661 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 662 * @_sk_redir: socket redirection information for skmsg 663 * @_nfct: Associated connection, if any (with nfctinfo bits) 664 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 665 * @skb_iif: ifindex of device we arrived on 666 * @tc_index: Traffic control index 667 * @hash: the packet hash 668 * @queue_mapping: Queue mapping for multiqueue devices 669 * @head_frag: skb was allocated from page fragments, 670 * not allocated by kmalloc() or vmalloc(). 671 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 672 * @pp_recycle: mark the packet for recycling instead of freeing (implies 673 * page_pool support on driver) 674 * @active_extensions: active extensions (skb_ext_id types) 675 * @ndisc_nodetype: router type (from link layer) 676 * @ooo_okay: allow the mapping of a socket to a queue to be changed 677 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 678 * ports. 679 * @sw_hash: indicates hash was computed in software stack 680 * @wifi_acked_valid: wifi_acked was set 681 * @wifi_acked: whether frame was acked on wifi or not 682 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 683 * @encapsulation: indicates the inner headers in the skbuff are valid 684 * @encap_hdr_csum: software checksum is needed 685 * @csum_valid: checksum is already valid 686 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 687 * @csum_complete_sw: checksum was completed by software 688 * @csum_level: indicates the number of consecutive checksums found in 689 * the packet minus one that have been verified as 690 * CHECKSUM_UNNECESSARY (max 3) 691 * @dst_pending_confirm: need to confirm neighbour 692 * @decrypted: Decrypted SKB 693 * @slow_gro: state present at GRO time, slower prepare step required 694 * @napi_id: id of the NAPI struct this skb came from 695 * @sender_cpu: (aka @napi_id) source CPU in XPS 696 * @secmark: security marking 697 * @mark: Generic packet mark 698 * @reserved_tailroom: (aka @mark) number of bytes of free space available 699 * at the tail of an sk_buff 700 * @vlan_present: VLAN tag is present 701 * @vlan_proto: vlan encapsulation protocol 702 * @vlan_tci: vlan tag control information 703 * @inner_protocol: Protocol (encapsulation) 704 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 705 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 706 * @inner_transport_header: Inner transport layer header (encapsulation) 707 * @inner_network_header: Network layer header (encapsulation) 708 * @inner_mac_header: Link layer header (encapsulation) 709 * @transport_header: Transport layer header 710 * @network_header: Network layer header 711 * @mac_header: Link layer header 712 * @kcov_handle: KCOV remote handle for remote coverage collection 713 * @tail: Tail pointer 714 * @end: End pointer 715 * @head: Head of buffer 716 * @data: Data head pointer 717 * @truesize: Buffer size 718 * @users: User count - see {datagram,tcp}.c 719 * @extensions: allocated extensions, valid if active_extensions is nonzero 720 */ 721 722 struct sk_buff { 723 union { 724 struct { 725 /* These two members must be first. */ 726 struct sk_buff *next; 727 struct sk_buff *prev; 728 729 union { 730 struct net_device *dev; 731 /* Some protocols might use this space to store information, 732 * while device pointer would be NULL. 733 * UDP receive path is one user. 734 */ 735 unsigned long dev_scratch; 736 }; 737 }; 738 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 739 struct list_head list; 740 }; 741 742 union { 743 struct sock *sk; 744 int ip_defrag_offset; 745 }; 746 747 union { 748 ktime_t tstamp; 749 u64 skb_mstamp_ns; /* earliest departure time */ 750 }; 751 /* 752 * This is the control buffer. It is free to use for every 753 * layer. Please put your private variables there. If you 754 * want to keep them across layers you have to do a skb_clone() 755 * first. This is owned by whoever has the skb queued ATM. 756 */ 757 char cb[48] __aligned(8); 758 759 union { 760 struct { 761 unsigned long _skb_refdst; 762 void (*destructor)(struct sk_buff *skb); 763 }; 764 struct list_head tcp_tsorted_anchor; 765 #ifdef CONFIG_NET_SOCK_MSG 766 unsigned long _sk_redir; 767 #endif 768 }; 769 770 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 771 unsigned long _nfct; 772 #endif 773 unsigned int len, 774 data_len; 775 __u16 mac_len, 776 hdr_len; 777 778 /* Following fields are _not_ copied in __copy_skb_header() 779 * Note that queue_mapping is here mostly to fill a hole. 780 */ 781 __u16 queue_mapping; 782 783 /* if you move cloned around you also must adapt those constants */ 784 #ifdef __BIG_ENDIAN_BITFIELD 785 #define CLONED_MASK (1 << 7) 786 #else 787 #define CLONED_MASK 1 788 #endif 789 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset) 790 791 /* private: */ 792 __u8 __cloned_offset[0]; 793 /* public: */ 794 __u8 cloned:1, 795 nohdr:1, 796 fclone:2, 797 peeked:1, 798 head_frag:1, 799 pfmemalloc:1, 800 pp_recycle:1; /* page_pool recycle indicator */ 801 #ifdef CONFIG_SKB_EXTENSIONS 802 __u8 active_extensions; 803 #endif 804 805 /* fields enclosed in headers_start/headers_end are copied 806 * using a single memcpy() in __copy_skb_header() 807 */ 808 /* private: */ 809 __u32 headers_start[0]; 810 /* public: */ 811 812 /* if you move pkt_type around you also must adapt those constants */ 813 #ifdef __BIG_ENDIAN_BITFIELD 814 #define PKT_TYPE_MAX (7 << 5) 815 #else 816 #define PKT_TYPE_MAX 7 817 #endif 818 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 819 820 /* private: */ 821 __u8 __pkt_type_offset[0]; 822 /* public: */ 823 __u8 pkt_type:3; 824 __u8 ignore_df:1; 825 __u8 nf_trace:1; 826 __u8 ip_summed:2; 827 __u8 ooo_okay:1; 828 829 __u8 l4_hash:1; 830 __u8 sw_hash:1; 831 __u8 wifi_acked_valid:1; 832 __u8 wifi_acked:1; 833 __u8 no_fcs:1; 834 /* Indicates the inner headers are valid in the skbuff. */ 835 __u8 encapsulation:1; 836 __u8 encap_hdr_csum:1; 837 __u8 csum_valid:1; 838 839 #ifdef __BIG_ENDIAN_BITFIELD 840 #define PKT_VLAN_PRESENT_BIT 7 841 #else 842 #define PKT_VLAN_PRESENT_BIT 0 843 #endif 844 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset) 845 /* private: */ 846 __u8 __pkt_vlan_present_offset[0]; 847 /* public: */ 848 __u8 vlan_present:1; 849 __u8 csum_complete_sw:1; 850 __u8 csum_level:2; 851 __u8 csum_not_inet:1; 852 __u8 dst_pending_confirm:1; 853 #ifdef CONFIG_IPV6_NDISC_NODETYPE 854 __u8 ndisc_nodetype:2; 855 #endif 856 857 __u8 ipvs_property:1; 858 __u8 inner_protocol_type:1; 859 __u8 remcsum_offload:1; 860 #ifdef CONFIG_NET_SWITCHDEV 861 __u8 offload_fwd_mark:1; 862 __u8 offload_l3_fwd_mark:1; 863 #endif 864 #ifdef CONFIG_NET_CLS_ACT 865 __u8 tc_skip_classify:1; 866 __u8 tc_at_ingress:1; 867 #endif 868 __u8 redirected:1; 869 #ifdef CONFIG_NET_REDIRECT 870 __u8 from_ingress:1; 871 #endif 872 #ifdef CONFIG_NETFILTER_SKIP_EGRESS 873 __u8 nf_skip_egress:1; 874 #endif 875 #ifdef CONFIG_TLS_DEVICE 876 __u8 decrypted:1; 877 #endif 878 __u8 slow_gro:1; 879 880 #ifdef CONFIG_NET_SCHED 881 __u16 tc_index; /* traffic control index */ 882 #endif 883 884 union { 885 __wsum csum; 886 struct { 887 __u16 csum_start; 888 __u16 csum_offset; 889 }; 890 }; 891 __u32 priority; 892 int skb_iif; 893 __u32 hash; 894 __be16 vlan_proto; 895 __u16 vlan_tci; 896 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 897 union { 898 unsigned int napi_id; 899 unsigned int sender_cpu; 900 }; 901 #endif 902 #ifdef CONFIG_NETWORK_SECMARK 903 __u32 secmark; 904 #endif 905 906 union { 907 __u32 mark; 908 __u32 reserved_tailroom; 909 }; 910 911 union { 912 __be16 inner_protocol; 913 __u8 inner_ipproto; 914 }; 915 916 __u16 inner_transport_header; 917 __u16 inner_network_header; 918 __u16 inner_mac_header; 919 920 __be16 protocol; 921 __u16 transport_header; 922 __u16 network_header; 923 __u16 mac_header; 924 925 #ifdef CONFIG_KCOV 926 u64 kcov_handle; 927 #endif 928 929 /* private: */ 930 __u32 headers_end[0]; 931 /* public: */ 932 933 /* These elements must be at the end, see alloc_skb() for details. */ 934 sk_buff_data_t tail; 935 sk_buff_data_t end; 936 unsigned char *head, 937 *data; 938 unsigned int truesize; 939 refcount_t users; 940 941 #ifdef CONFIG_SKB_EXTENSIONS 942 /* only useable after checking ->active_extensions != 0 */ 943 struct skb_ext *extensions; 944 #endif 945 }; 946 947 #ifdef __KERNEL__ 948 /* 949 * Handling routines are only of interest to the kernel 950 */ 951 952 #define SKB_ALLOC_FCLONE 0x01 953 #define SKB_ALLOC_RX 0x02 954 #define SKB_ALLOC_NAPI 0x04 955 956 /** 957 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 958 * @skb: buffer 959 */ 960 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 961 { 962 return unlikely(skb->pfmemalloc); 963 } 964 965 /* 966 * skb might have a dst pointer attached, refcounted or not. 967 * _skb_refdst low order bit is set if refcount was _not_ taken 968 */ 969 #define SKB_DST_NOREF 1UL 970 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 971 972 /** 973 * skb_dst - returns skb dst_entry 974 * @skb: buffer 975 * 976 * Returns skb dst_entry, regardless of reference taken or not. 977 */ 978 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 979 { 980 /* If refdst was not refcounted, check we still are in a 981 * rcu_read_lock section 982 */ 983 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 984 !rcu_read_lock_held() && 985 !rcu_read_lock_bh_held()); 986 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 987 } 988 989 /** 990 * skb_dst_set - sets skb dst 991 * @skb: buffer 992 * @dst: dst entry 993 * 994 * Sets skb dst, assuming a reference was taken on dst and should 995 * be released by skb_dst_drop() 996 */ 997 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 998 { 999 skb->slow_gro |= !!dst; 1000 skb->_skb_refdst = (unsigned long)dst; 1001 } 1002 1003 /** 1004 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 1005 * @skb: buffer 1006 * @dst: dst entry 1007 * 1008 * Sets skb dst, assuming a reference was not taken on dst. 1009 * If dst entry is cached, we do not take reference and dst_release 1010 * will be avoided by refdst_drop. If dst entry is not cached, we take 1011 * reference, so that last dst_release can destroy the dst immediately. 1012 */ 1013 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 1014 { 1015 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1016 skb->slow_gro |= !!dst; 1017 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1018 } 1019 1020 /** 1021 * skb_dst_is_noref - Test if skb dst isn't refcounted 1022 * @skb: buffer 1023 */ 1024 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1025 { 1026 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1027 } 1028 1029 /** 1030 * skb_rtable - Returns the skb &rtable 1031 * @skb: buffer 1032 */ 1033 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1034 { 1035 return (struct rtable *)skb_dst(skb); 1036 } 1037 1038 /* For mangling skb->pkt_type from user space side from applications 1039 * such as nft, tc, etc, we only allow a conservative subset of 1040 * possible pkt_types to be set. 1041 */ 1042 static inline bool skb_pkt_type_ok(u32 ptype) 1043 { 1044 return ptype <= PACKET_OTHERHOST; 1045 } 1046 1047 /** 1048 * skb_napi_id - Returns the skb's NAPI id 1049 * @skb: buffer 1050 */ 1051 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1052 { 1053 #ifdef CONFIG_NET_RX_BUSY_POLL 1054 return skb->napi_id; 1055 #else 1056 return 0; 1057 #endif 1058 } 1059 1060 /** 1061 * skb_unref - decrement the skb's reference count 1062 * @skb: buffer 1063 * 1064 * Returns true if we can free the skb. 1065 */ 1066 static inline bool skb_unref(struct sk_buff *skb) 1067 { 1068 if (unlikely(!skb)) 1069 return false; 1070 if (likely(refcount_read(&skb->users) == 1)) 1071 smp_rmb(); 1072 else if (likely(!refcount_dec_and_test(&skb->users))) 1073 return false; 1074 1075 return true; 1076 } 1077 1078 void skb_release_head_state(struct sk_buff *skb); 1079 void kfree_skb(struct sk_buff *skb); 1080 void kfree_skb_list(struct sk_buff *segs); 1081 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1082 void skb_tx_error(struct sk_buff *skb); 1083 1084 #ifdef CONFIG_TRACEPOINTS 1085 void consume_skb(struct sk_buff *skb); 1086 #else 1087 static inline void consume_skb(struct sk_buff *skb) 1088 { 1089 return kfree_skb(skb); 1090 } 1091 #endif 1092 1093 void __consume_stateless_skb(struct sk_buff *skb); 1094 void __kfree_skb(struct sk_buff *skb); 1095 extern struct kmem_cache *skbuff_head_cache; 1096 1097 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1098 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1099 bool *fragstolen, int *delta_truesize); 1100 1101 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1102 int node); 1103 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1104 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1105 struct sk_buff *build_skb_around(struct sk_buff *skb, 1106 void *data, unsigned int frag_size); 1107 1108 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1109 1110 /** 1111 * alloc_skb - allocate a network buffer 1112 * @size: size to allocate 1113 * @priority: allocation mask 1114 * 1115 * This function is a convenient wrapper around __alloc_skb(). 1116 */ 1117 static inline struct sk_buff *alloc_skb(unsigned int size, 1118 gfp_t priority) 1119 { 1120 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1121 } 1122 1123 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1124 unsigned long data_len, 1125 int max_page_order, 1126 int *errcode, 1127 gfp_t gfp_mask); 1128 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1129 1130 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1131 struct sk_buff_fclones { 1132 struct sk_buff skb1; 1133 1134 struct sk_buff skb2; 1135 1136 refcount_t fclone_ref; 1137 }; 1138 1139 /** 1140 * skb_fclone_busy - check if fclone is busy 1141 * @sk: socket 1142 * @skb: buffer 1143 * 1144 * Returns true if skb is a fast clone, and its clone is not freed. 1145 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1146 * so we also check that this didnt happen. 1147 */ 1148 static inline bool skb_fclone_busy(const struct sock *sk, 1149 const struct sk_buff *skb) 1150 { 1151 const struct sk_buff_fclones *fclones; 1152 1153 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1154 1155 return skb->fclone == SKB_FCLONE_ORIG && 1156 refcount_read(&fclones->fclone_ref) > 1 && 1157 READ_ONCE(fclones->skb2.sk) == sk; 1158 } 1159 1160 /** 1161 * alloc_skb_fclone - allocate a network buffer from fclone cache 1162 * @size: size to allocate 1163 * @priority: allocation mask 1164 * 1165 * This function is a convenient wrapper around __alloc_skb(). 1166 */ 1167 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1168 gfp_t priority) 1169 { 1170 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1171 } 1172 1173 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1174 void skb_headers_offset_update(struct sk_buff *skb, int off); 1175 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1176 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1177 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1178 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1179 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1180 gfp_t gfp_mask, bool fclone); 1181 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1182 gfp_t gfp_mask) 1183 { 1184 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1185 } 1186 1187 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1188 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1189 unsigned int headroom); 1190 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom); 1191 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1192 int newtailroom, gfp_t priority); 1193 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1194 int offset, int len); 1195 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1196 int offset, int len); 1197 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1198 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1199 1200 /** 1201 * skb_pad - zero pad the tail of an skb 1202 * @skb: buffer to pad 1203 * @pad: space to pad 1204 * 1205 * Ensure that a buffer is followed by a padding area that is zero 1206 * filled. Used by network drivers which may DMA or transfer data 1207 * beyond the buffer end onto the wire. 1208 * 1209 * May return error in out of memory cases. The skb is freed on error. 1210 */ 1211 static inline int skb_pad(struct sk_buff *skb, int pad) 1212 { 1213 return __skb_pad(skb, pad, true); 1214 } 1215 #define dev_kfree_skb(a) consume_skb(a) 1216 1217 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1218 int offset, size_t size); 1219 1220 struct skb_seq_state { 1221 __u32 lower_offset; 1222 __u32 upper_offset; 1223 __u32 frag_idx; 1224 __u32 stepped_offset; 1225 struct sk_buff *root_skb; 1226 struct sk_buff *cur_skb; 1227 __u8 *frag_data; 1228 __u32 frag_off; 1229 }; 1230 1231 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1232 unsigned int to, struct skb_seq_state *st); 1233 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1234 struct skb_seq_state *st); 1235 void skb_abort_seq_read(struct skb_seq_state *st); 1236 1237 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1238 unsigned int to, struct ts_config *config); 1239 1240 /* 1241 * Packet hash types specify the type of hash in skb_set_hash. 1242 * 1243 * Hash types refer to the protocol layer addresses which are used to 1244 * construct a packet's hash. The hashes are used to differentiate or identify 1245 * flows of the protocol layer for the hash type. Hash types are either 1246 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1247 * 1248 * Properties of hashes: 1249 * 1250 * 1) Two packets in different flows have different hash values 1251 * 2) Two packets in the same flow should have the same hash value 1252 * 1253 * A hash at a higher layer is considered to be more specific. A driver should 1254 * set the most specific hash possible. 1255 * 1256 * A driver cannot indicate a more specific hash than the layer at which a hash 1257 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1258 * 1259 * A driver may indicate a hash level which is less specific than the 1260 * actual layer the hash was computed on. For instance, a hash computed 1261 * at L4 may be considered an L3 hash. This should only be done if the 1262 * driver can't unambiguously determine that the HW computed the hash at 1263 * the higher layer. Note that the "should" in the second property above 1264 * permits this. 1265 */ 1266 enum pkt_hash_types { 1267 PKT_HASH_TYPE_NONE, /* Undefined type */ 1268 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1269 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1270 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1271 }; 1272 1273 static inline void skb_clear_hash(struct sk_buff *skb) 1274 { 1275 skb->hash = 0; 1276 skb->sw_hash = 0; 1277 skb->l4_hash = 0; 1278 } 1279 1280 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1281 { 1282 if (!skb->l4_hash) 1283 skb_clear_hash(skb); 1284 } 1285 1286 static inline void 1287 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1288 { 1289 skb->l4_hash = is_l4; 1290 skb->sw_hash = is_sw; 1291 skb->hash = hash; 1292 } 1293 1294 static inline void 1295 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1296 { 1297 /* Used by drivers to set hash from HW */ 1298 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1299 } 1300 1301 static inline void 1302 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1303 { 1304 __skb_set_hash(skb, hash, true, is_l4); 1305 } 1306 1307 void __skb_get_hash(struct sk_buff *skb); 1308 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1309 u32 skb_get_poff(const struct sk_buff *skb); 1310 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1311 const struct flow_keys_basic *keys, int hlen); 1312 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1313 const void *data, int hlen_proto); 1314 1315 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1316 int thoff, u8 ip_proto) 1317 { 1318 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1319 } 1320 1321 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1322 const struct flow_dissector_key *key, 1323 unsigned int key_count); 1324 1325 struct bpf_flow_dissector; 1326 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1327 __be16 proto, int nhoff, int hlen, unsigned int flags); 1328 1329 bool __skb_flow_dissect(const struct net *net, 1330 const struct sk_buff *skb, 1331 struct flow_dissector *flow_dissector, 1332 void *target_container, const void *data, 1333 __be16 proto, int nhoff, int hlen, unsigned int flags); 1334 1335 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1336 struct flow_dissector *flow_dissector, 1337 void *target_container, unsigned int flags) 1338 { 1339 return __skb_flow_dissect(NULL, skb, flow_dissector, 1340 target_container, NULL, 0, 0, 0, flags); 1341 } 1342 1343 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1344 struct flow_keys *flow, 1345 unsigned int flags) 1346 { 1347 memset(flow, 0, sizeof(*flow)); 1348 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1349 flow, NULL, 0, 0, 0, flags); 1350 } 1351 1352 static inline bool 1353 skb_flow_dissect_flow_keys_basic(const struct net *net, 1354 const struct sk_buff *skb, 1355 struct flow_keys_basic *flow, 1356 const void *data, __be16 proto, 1357 int nhoff, int hlen, unsigned int flags) 1358 { 1359 memset(flow, 0, sizeof(*flow)); 1360 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1361 data, proto, nhoff, hlen, flags); 1362 } 1363 1364 void skb_flow_dissect_meta(const struct sk_buff *skb, 1365 struct flow_dissector *flow_dissector, 1366 void *target_container); 1367 1368 /* Gets a skb connection tracking info, ctinfo map should be a 1369 * map of mapsize to translate enum ip_conntrack_info states 1370 * to user states. 1371 */ 1372 void 1373 skb_flow_dissect_ct(const struct sk_buff *skb, 1374 struct flow_dissector *flow_dissector, 1375 void *target_container, 1376 u16 *ctinfo_map, size_t mapsize, 1377 bool post_ct); 1378 void 1379 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1380 struct flow_dissector *flow_dissector, 1381 void *target_container); 1382 1383 void skb_flow_dissect_hash(const struct sk_buff *skb, 1384 struct flow_dissector *flow_dissector, 1385 void *target_container); 1386 1387 static inline __u32 skb_get_hash(struct sk_buff *skb) 1388 { 1389 if (!skb->l4_hash && !skb->sw_hash) 1390 __skb_get_hash(skb); 1391 1392 return skb->hash; 1393 } 1394 1395 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1396 { 1397 if (!skb->l4_hash && !skb->sw_hash) { 1398 struct flow_keys keys; 1399 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1400 1401 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1402 } 1403 1404 return skb->hash; 1405 } 1406 1407 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1408 const siphash_key_t *perturb); 1409 1410 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1411 { 1412 return skb->hash; 1413 } 1414 1415 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1416 { 1417 to->hash = from->hash; 1418 to->sw_hash = from->sw_hash; 1419 to->l4_hash = from->l4_hash; 1420 }; 1421 1422 static inline void skb_copy_decrypted(struct sk_buff *to, 1423 const struct sk_buff *from) 1424 { 1425 #ifdef CONFIG_TLS_DEVICE 1426 to->decrypted = from->decrypted; 1427 #endif 1428 } 1429 1430 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1431 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1432 { 1433 return skb->head + skb->end; 1434 } 1435 1436 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1437 { 1438 return skb->end; 1439 } 1440 #else 1441 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1442 { 1443 return skb->end; 1444 } 1445 1446 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1447 { 1448 return skb->end - skb->head; 1449 } 1450 #endif 1451 1452 /* Internal */ 1453 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1454 1455 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1456 { 1457 return &skb_shinfo(skb)->hwtstamps; 1458 } 1459 1460 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1461 { 1462 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1463 1464 return is_zcopy ? skb_uarg(skb) : NULL; 1465 } 1466 1467 static inline void net_zcopy_get(struct ubuf_info *uarg) 1468 { 1469 refcount_inc(&uarg->refcnt); 1470 } 1471 1472 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1473 { 1474 skb_shinfo(skb)->destructor_arg = uarg; 1475 skb_shinfo(skb)->flags |= uarg->flags; 1476 } 1477 1478 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1479 bool *have_ref) 1480 { 1481 if (skb && uarg && !skb_zcopy(skb)) { 1482 if (unlikely(have_ref && *have_ref)) 1483 *have_ref = false; 1484 else 1485 net_zcopy_get(uarg); 1486 skb_zcopy_init(skb, uarg); 1487 } 1488 } 1489 1490 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1491 { 1492 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1493 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1494 } 1495 1496 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1497 { 1498 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1499 } 1500 1501 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1502 { 1503 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1504 } 1505 1506 static inline void net_zcopy_put(struct ubuf_info *uarg) 1507 { 1508 if (uarg) 1509 uarg->callback(NULL, uarg, true); 1510 } 1511 1512 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1513 { 1514 if (uarg) { 1515 if (uarg->callback == msg_zerocopy_callback) 1516 msg_zerocopy_put_abort(uarg, have_uref); 1517 else if (have_uref) 1518 net_zcopy_put(uarg); 1519 } 1520 } 1521 1522 /* Release a reference on a zerocopy structure */ 1523 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1524 { 1525 struct ubuf_info *uarg = skb_zcopy(skb); 1526 1527 if (uarg) { 1528 if (!skb_zcopy_is_nouarg(skb)) 1529 uarg->callback(skb, uarg, zerocopy_success); 1530 1531 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG; 1532 } 1533 } 1534 1535 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1536 { 1537 skb->next = NULL; 1538 } 1539 1540 /* Iterate through singly-linked GSO fragments of an skb. */ 1541 #define skb_list_walk_safe(first, skb, next_skb) \ 1542 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1543 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1544 1545 static inline void skb_list_del_init(struct sk_buff *skb) 1546 { 1547 __list_del_entry(&skb->list); 1548 skb_mark_not_on_list(skb); 1549 } 1550 1551 /** 1552 * skb_queue_empty - check if a queue is empty 1553 * @list: queue head 1554 * 1555 * Returns true if the queue is empty, false otherwise. 1556 */ 1557 static inline int skb_queue_empty(const struct sk_buff_head *list) 1558 { 1559 return list->next == (const struct sk_buff *) list; 1560 } 1561 1562 /** 1563 * skb_queue_empty_lockless - check if a queue is empty 1564 * @list: queue head 1565 * 1566 * Returns true if the queue is empty, false otherwise. 1567 * This variant can be used in lockless contexts. 1568 */ 1569 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1570 { 1571 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1572 } 1573 1574 1575 /** 1576 * skb_queue_is_last - check if skb is the last entry in the queue 1577 * @list: queue head 1578 * @skb: buffer 1579 * 1580 * Returns true if @skb is the last buffer on the list. 1581 */ 1582 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1583 const struct sk_buff *skb) 1584 { 1585 return skb->next == (const struct sk_buff *) list; 1586 } 1587 1588 /** 1589 * skb_queue_is_first - check if skb is the first entry in the queue 1590 * @list: queue head 1591 * @skb: buffer 1592 * 1593 * Returns true if @skb is the first buffer on the list. 1594 */ 1595 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1596 const struct sk_buff *skb) 1597 { 1598 return skb->prev == (const struct sk_buff *) list; 1599 } 1600 1601 /** 1602 * skb_queue_next - return the next packet in the queue 1603 * @list: queue head 1604 * @skb: current buffer 1605 * 1606 * Return the next packet in @list after @skb. It is only valid to 1607 * call this if skb_queue_is_last() evaluates to false. 1608 */ 1609 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1610 const struct sk_buff *skb) 1611 { 1612 /* This BUG_ON may seem severe, but if we just return then we 1613 * are going to dereference garbage. 1614 */ 1615 BUG_ON(skb_queue_is_last(list, skb)); 1616 return skb->next; 1617 } 1618 1619 /** 1620 * skb_queue_prev - return the prev packet in the queue 1621 * @list: queue head 1622 * @skb: current buffer 1623 * 1624 * Return the prev packet in @list before @skb. It is only valid to 1625 * call this if skb_queue_is_first() evaluates to false. 1626 */ 1627 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1628 const struct sk_buff *skb) 1629 { 1630 /* This BUG_ON may seem severe, but if we just return then we 1631 * are going to dereference garbage. 1632 */ 1633 BUG_ON(skb_queue_is_first(list, skb)); 1634 return skb->prev; 1635 } 1636 1637 /** 1638 * skb_get - reference buffer 1639 * @skb: buffer to reference 1640 * 1641 * Makes another reference to a socket buffer and returns a pointer 1642 * to the buffer. 1643 */ 1644 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1645 { 1646 refcount_inc(&skb->users); 1647 return skb; 1648 } 1649 1650 /* 1651 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1652 */ 1653 1654 /** 1655 * skb_cloned - is the buffer a clone 1656 * @skb: buffer to check 1657 * 1658 * Returns true if the buffer was generated with skb_clone() and is 1659 * one of multiple shared copies of the buffer. Cloned buffers are 1660 * shared data so must not be written to under normal circumstances. 1661 */ 1662 static inline int skb_cloned(const struct sk_buff *skb) 1663 { 1664 return skb->cloned && 1665 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1666 } 1667 1668 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1669 { 1670 might_sleep_if(gfpflags_allow_blocking(pri)); 1671 1672 if (skb_cloned(skb)) 1673 return pskb_expand_head(skb, 0, 0, pri); 1674 1675 return 0; 1676 } 1677 1678 /** 1679 * skb_header_cloned - is the header a clone 1680 * @skb: buffer to check 1681 * 1682 * Returns true if modifying the header part of the buffer requires 1683 * the data to be copied. 1684 */ 1685 static inline int skb_header_cloned(const struct sk_buff *skb) 1686 { 1687 int dataref; 1688 1689 if (!skb->cloned) 1690 return 0; 1691 1692 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1693 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1694 return dataref != 1; 1695 } 1696 1697 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1698 { 1699 might_sleep_if(gfpflags_allow_blocking(pri)); 1700 1701 if (skb_header_cloned(skb)) 1702 return pskb_expand_head(skb, 0, 0, pri); 1703 1704 return 0; 1705 } 1706 1707 /** 1708 * __skb_header_release - release reference to header 1709 * @skb: buffer to operate on 1710 */ 1711 static inline void __skb_header_release(struct sk_buff *skb) 1712 { 1713 skb->nohdr = 1; 1714 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1715 } 1716 1717 1718 /** 1719 * skb_shared - is the buffer shared 1720 * @skb: buffer to check 1721 * 1722 * Returns true if more than one person has a reference to this 1723 * buffer. 1724 */ 1725 static inline int skb_shared(const struct sk_buff *skb) 1726 { 1727 return refcount_read(&skb->users) != 1; 1728 } 1729 1730 /** 1731 * skb_share_check - check if buffer is shared and if so clone it 1732 * @skb: buffer to check 1733 * @pri: priority for memory allocation 1734 * 1735 * If the buffer is shared the buffer is cloned and the old copy 1736 * drops a reference. A new clone with a single reference is returned. 1737 * If the buffer is not shared the original buffer is returned. When 1738 * being called from interrupt status or with spinlocks held pri must 1739 * be GFP_ATOMIC. 1740 * 1741 * NULL is returned on a memory allocation failure. 1742 */ 1743 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1744 { 1745 might_sleep_if(gfpflags_allow_blocking(pri)); 1746 if (skb_shared(skb)) { 1747 struct sk_buff *nskb = skb_clone(skb, pri); 1748 1749 if (likely(nskb)) 1750 consume_skb(skb); 1751 else 1752 kfree_skb(skb); 1753 skb = nskb; 1754 } 1755 return skb; 1756 } 1757 1758 /* 1759 * Copy shared buffers into a new sk_buff. We effectively do COW on 1760 * packets to handle cases where we have a local reader and forward 1761 * and a couple of other messy ones. The normal one is tcpdumping 1762 * a packet thats being forwarded. 1763 */ 1764 1765 /** 1766 * skb_unshare - make a copy of a shared buffer 1767 * @skb: buffer to check 1768 * @pri: priority for memory allocation 1769 * 1770 * If the socket buffer is a clone then this function creates a new 1771 * copy of the data, drops a reference count on the old copy and returns 1772 * the new copy with the reference count at 1. If the buffer is not a clone 1773 * the original buffer is returned. When called with a spinlock held or 1774 * from interrupt state @pri must be %GFP_ATOMIC 1775 * 1776 * %NULL is returned on a memory allocation failure. 1777 */ 1778 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1779 gfp_t pri) 1780 { 1781 might_sleep_if(gfpflags_allow_blocking(pri)); 1782 if (skb_cloned(skb)) { 1783 struct sk_buff *nskb = skb_copy(skb, pri); 1784 1785 /* Free our shared copy */ 1786 if (likely(nskb)) 1787 consume_skb(skb); 1788 else 1789 kfree_skb(skb); 1790 skb = nskb; 1791 } 1792 return skb; 1793 } 1794 1795 /** 1796 * skb_peek - peek at the head of an &sk_buff_head 1797 * @list_: list to peek at 1798 * 1799 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1800 * be careful with this one. A peek leaves the buffer on the 1801 * list and someone else may run off with it. You must hold 1802 * the appropriate locks or have a private queue to do this. 1803 * 1804 * Returns %NULL for an empty list or a pointer to the head element. 1805 * The reference count is not incremented and the reference is therefore 1806 * volatile. Use with caution. 1807 */ 1808 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1809 { 1810 struct sk_buff *skb = list_->next; 1811 1812 if (skb == (struct sk_buff *)list_) 1813 skb = NULL; 1814 return skb; 1815 } 1816 1817 /** 1818 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1819 * @list_: list to peek at 1820 * 1821 * Like skb_peek(), but the caller knows that the list is not empty. 1822 */ 1823 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1824 { 1825 return list_->next; 1826 } 1827 1828 /** 1829 * skb_peek_next - peek skb following the given one from a queue 1830 * @skb: skb to start from 1831 * @list_: list to peek at 1832 * 1833 * Returns %NULL when the end of the list is met or a pointer to the 1834 * next element. The reference count is not incremented and the 1835 * reference is therefore volatile. Use with caution. 1836 */ 1837 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1838 const struct sk_buff_head *list_) 1839 { 1840 struct sk_buff *next = skb->next; 1841 1842 if (next == (struct sk_buff *)list_) 1843 next = NULL; 1844 return next; 1845 } 1846 1847 /** 1848 * skb_peek_tail - peek at the tail of an &sk_buff_head 1849 * @list_: list to peek at 1850 * 1851 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1852 * be careful with this one. A peek leaves the buffer on the 1853 * list and someone else may run off with it. You must hold 1854 * the appropriate locks or have a private queue to do this. 1855 * 1856 * Returns %NULL for an empty list or a pointer to the tail element. 1857 * The reference count is not incremented and the reference is therefore 1858 * volatile. Use with caution. 1859 */ 1860 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1861 { 1862 struct sk_buff *skb = READ_ONCE(list_->prev); 1863 1864 if (skb == (struct sk_buff *)list_) 1865 skb = NULL; 1866 return skb; 1867 1868 } 1869 1870 /** 1871 * skb_queue_len - get queue length 1872 * @list_: list to measure 1873 * 1874 * Return the length of an &sk_buff queue. 1875 */ 1876 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1877 { 1878 return list_->qlen; 1879 } 1880 1881 /** 1882 * skb_queue_len_lockless - get queue length 1883 * @list_: list to measure 1884 * 1885 * Return the length of an &sk_buff queue. 1886 * This variant can be used in lockless contexts. 1887 */ 1888 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1889 { 1890 return READ_ONCE(list_->qlen); 1891 } 1892 1893 /** 1894 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1895 * @list: queue to initialize 1896 * 1897 * This initializes only the list and queue length aspects of 1898 * an sk_buff_head object. This allows to initialize the list 1899 * aspects of an sk_buff_head without reinitializing things like 1900 * the spinlock. It can also be used for on-stack sk_buff_head 1901 * objects where the spinlock is known to not be used. 1902 */ 1903 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1904 { 1905 list->prev = list->next = (struct sk_buff *)list; 1906 list->qlen = 0; 1907 } 1908 1909 /* 1910 * This function creates a split out lock class for each invocation; 1911 * this is needed for now since a whole lot of users of the skb-queue 1912 * infrastructure in drivers have different locking usage (in hardirq) 1913 * than the networking core (in softirq only). In the long run either the 1914 * network layer or drivers should need annotation to consolidate the 1915 * main types of usage into 3 classes. 1916 */ 1917 static inline void skb_queue_head_init(struct sk_buff_head *list) 1918 { 1919 spin_lock_init(&list->lock); 1920 __skb_queue_head_init(list); 1921 } 1922 1923 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1924 struct lock_class_key *class) 1925 { 1926 skb_queue_head_init(list); 1927 lockdep_set_class(&list->lock, class); 1928 } 1929 1930 /* 1931 * Insert an sk_buff on a list. 1932 * 1933 * The "__skb_xxxx()" functions are the non-atomic ones that 1934 * can only be called with interrupts disabled. 1935 */ 1936 static inline void __skb_insert(struct sk_buff *newsk, 1937 struct sk_buff *prev, struct sk_buff *next, 1938 struct sk_buff_head *list) 1939 { 1940 /* See skb_queue_empty_lockless() and skb_peek_tail() 1941 * for the opposite READ_ONCE() 1942 */ 1943 WRITE_ONCE(newsk->next, next); 1944 WRITE_ONCE(newsk->prev, prev); 1945 WRITE_ONCE(next->prev, newsk); 1946 WRITE_ONCE(prev->next, newsk); 1947 WRITE_ONCE(list->qlen, list->qlen + 1); 1948 } 1949 1950 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1951 struct sk_buff *prev, 1952 struct sk_buff *next) 1953 { 1954 struct sk_buff *first = list->next; 1955 struct sk_buff *last = list->prev; 1956 1957 WRITE_ONCE(first->prev, prev); 1958 WRITE_ONCE(prev->next, first); 1959 1960 WRITE_ONCE(last->next, next); 1961 WRITE_ONCE(next->prev, last); 1962 } 1963 1964 /** 1965 * skb_queue_splice - join two skb lists, this is designed for stacks 1966 * @list: the new list to add 1967 * @head: the place to add it in the first list 1968 */ 1969 static inline void skb_queue_splice(const struct sk_buff_head *list, 1970 struct sk_buff_head *head) 1971 { 1972 if (!skb_queue_empty(list)) { 1973 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1974 head->qlen += list->qlen; 1975 } 1976 } 1977 1978 /** 1979 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1980 * @list: the new list to add 1981 * @head: the place to add it in the first list 1982 * 1983 * The list at @list is reinitialised 1984 */ 1985 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1986 struct sk_buff_head *head) 1987 { 1988 if (!skb_queue_empty(list)) { 1989 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1990 head->qlen += list->qlen; 1991 __skb_queue_head_init(list); 1992 } 1993 } 1994 1995 /** 1996 * skb_queue_splice_tail - join two skb lists, each list being a queue 1997 * @list: the new list to add 1998 * @head: the place to add it in the first list 1999 */ 2000 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 2001 struct sk_buff_head *head) 2002 { 2003 if (!skb_queue_empty(list)) { 2004 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2005 head->qlen += list->qlen; 2006 } 2007 } 2008 2009 /** 2010 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 2011 * @list: the new list to add 2012 * @head: the place to add it in the first list 2013 * 2014 * Each of the lists is a queue. 2015 * The list at @list is reinitialised 2016 */ 2017 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2018 struct sk_buff_head *head) 2019 { 2020 if (!skb_queue_empty(list)) { 2021 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2022 head->qlen += list->qlen; 2023 __skb_queue_head_init(list); 2024 } 2025 } 2026 2027 /** 2028 * __skb_queue_after - queue a buffer at the list head 2029 * @list: list to use 2030 * @prev: place after this buffer 2031 * @newsk: buffer to queue 2032 * 2033 * Queue a buffer int the middle of a list. This function takes no locks 2034 * and you must therefore hold required locks before calling it. 2035 * 2036 * A buffer cannot be placed on two lists at the same time. 2037 */ 2038 static inline void __skb_queue_after(struct sk_buff_head *list, 2039 struct sk_buff *prev, 2040 struct sk_buff *newsk) 2041 { 2042 __skb_insert(newsk, prev, prev->next, list); 2043 } 2044 2045 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2046 struct sk_buff_head *list); 2047 2048 static inline void __skb_queue_before(struct sk_buff_head *list, 2049 struct sk_buff *next, 2050 struct sk_buff *newsk) 2051 { 2052 __skb_insert(newsk, next->prev, next, list); 2053 } 2054 2055 /** 2056 * __skb_queue_head - queue a buffer at the list head 2057 * @list: list to use 2058 * @newsk: buffer to queue 2059 * 2060 * Queue a buffer at the start of a list. This function takes no locks 2061 * and you must therefore hold required locks before calling it. 2062 * 2063 * A buffer cannot be placed on two lists at the same time. 2064 */ 2065 static inline void __skb_queue_head(struct sk_buff_head *list, 2066 struct sk_buff *newsk) 2067 { 2068 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2069 } 2070 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2071 2072 /** 2073 * __skb_queue_tail - queue a buffer at the list tail 2074 * @list: list to use 2075 * @newsk: buffer to queue 2076 * 2077 * Queue a buffer at the end of a list. This function takes no locks 2078 * and you must therefore hold required locks before calling it. 2079 * 2080 * A buffer cannot be placed on two lists at the same time. 2081 */ 2082 static inline void __skb_queue_tail(struct sk_buff_head *list, 2083 struct sk_buff *newsk) 2084 { 2085 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2086 } 2087 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2088 2089 /* 2090 * remove sk_buff from list. _Must_ be called atomically, and with 2091 * the list known.. 2092 */ 2093 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2094 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2095 { 2096 struct sk_buff *next, *prev; 2097 2098 WRITE_ONCE(list->qlen, list->qlen - 1); 2099 next = skb->next; 2100 prev = skb->prev; 2101 skb->next = skb->prev = NULL; 2102 WRITE_ONCE(next->prev, prev); 2103 WRITE_ONCE(prev->next, next); 2104 } 2105 2106 /** 2107 * __skb_dequeue - remove from the head of the queue 2108 * @list: list to dequeue from 2109 * 2110 * Remove the head of the list. This function does not take any locks 2111 * so must be used with appropriate locks held only. The head item is 2112 * returned or %NULL if the list is empty. 2113 */ 2114 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2115 { 2116 struct sk_buff *skb = skb_peek(list); 2117 if (skb) 2118 __skb_unlink(skb, list); 2119 return skb; 2120 } 2121 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2122 2123 /** 2124 * __skb_dequeue_tail - remove from the tail of the queue 2125 * @list: list to dequeue from 2126 * 2127 * Remove the tail of the list. This function does not take any locks 2128 * so must be used with appropriate locks held only. The tail item is 2129 * returned or %NULL if the list is empty. 2130 */ 2131 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2132 { 2133 struct sk_buff *skb = skb_peek_tail(list); 2134 if (skb) 2135 __skb_unlink(skb, list); 2136 return skb; 2137 } 2138 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2139 2140 2141 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2142 { 2143 return skb->data_len; 2144 } 2145 2146 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2147 { 2148 return skb->len - skb->data_len; 2149 } 2150 2151 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2152 { 2153 unsigned int i, len = 0; 2154 2155 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2156 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2157 return len; 2158 } 2159 2160 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2161 { 2162 return skb_headlen(skb) + __skb_pagelen(skb); 2163 } 2164 2165 /** 2166 * __skb_fill_page_desc - initialise a paged fragment in an skb 2167 * @skb: buffer containing fragment to be initialised 2168 * @i: paged fragment index to initialise 2169 * @page: the page to use for this fragment 2170 * @off: the offset to the data with @page 2171 * @size: the length of the data 2172 * 2173 * Initialises the @i'th fragment of @skb to point to &size bytes at 2174 * offset @off within @page. 2175 * 2176 * Does not take any additional reference on the fragment. 2177 */ 2178 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2179 struct page *page, int off, int size) 2180 { 2181 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2182 2183 /* 2184 * Propagate page pfmemalloc to the skb if we can. The problem is 2185 * that not all callers have unique ownership of the page but rely 2186 * on page_is_pfmemalloc doing the right thing(tm). 2187 */ 2188 frag->bv_page = page; 2189 frag->bv_offset = off; 2190 skb_frag_size_set(frag, size); 2191 2192 page = compound_head(page); 2193 if (page_is_pfmemalloc(page)) 2194 skb->pfmemalloc = true; 2195 } 2196 2197 /** 2198 * skb_fill_page_desc - initialise a paged fragment in an skb 2199 * @skb: buffer containing fragment to be initialised 2200 * @i: paged fragment index to initialise 2201 * @page: the page to use for this fragment 2202 * @off: the offset to the data with @page 2203 * @size: the length of the data 2204 * 2205 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2206 * @skb to point to @size bytes at offset @off within @page. In 2207 * addition updates @skb such that @i is the last fragment. 2208 * 2209 * Does not take any additional reference on the fragment. 2210 */ 2211 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2212 struct page *page, int off, int size) 2213 { 2214 __skb_fill_page_desc(skb, i, page, off, size); 2215 skb_shinfo(skb)->nr_frags = i + 1; 2216 } 2217 2218 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2219 int size, unsigned int truesize); 2220 2221 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2222 unsigned int truesize); 2223 2224 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2225 2226 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2227 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2228 { 2229 return skb->head + skb->tail; 2230 } 2231 2232 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2233 { 2234 skb->tail = skb->data - skb->head; 2235 } 2236 2237 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2238 { 2239 skb_reset_tail_pointer(skb); 2240 skb->tail += offset; 2241 } 2242 2243 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2244 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2245 { 2246 return skb->tail; 2247 } 2248 2249 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2250 { 2251 skb->tail = skb->data; 2252 } 2253 2254 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2255 { 2256 skb->tail = skb->data + offset; 2257 } 2258 2259 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2260 2261 /* 2262 * Add data to an sk_buff 2263 */ 2264 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2265 void *skb_put(struct sk_buff *skb, unsigned int len); 2266 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2267 { 2268 void *tmp = skb_tail_pointer(skb); 2269 SKB_LINEAR_ASSERT(skb); 2270 skb->tail += len; 2271 skb->len += len; 2272 return tmp; 2273 } 2274 2275 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2276 { 2277 void *tmp = __skb_put(skb, len); 2278 2279 memset(tmp, 0, len); 2280 return tmp; 2281 } 2282 2283 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2284 unsigned int len) 2285 { 2286 void *tmp = __skb_put(skb, len); 2287 2288 memcpy(tmp, data, len); 2289 return tmp; 2290 } 2291 2292 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2293 { 2294 *(u8 *)__skb_put(skb, 1) = val; 2295 } 2296 2297 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2298 { 2299 void *tmp = skb_put(skb, len); 2300 2301 memset(tmp, 0, len); 2302 2303 return tmp; 2304 } 2305 2306 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2307 unsigned int len) 2308 { 2309 void *tmp = skb_put(skb, len); 2310 2311 memcpy(tmp, data, len); 2312 2313 return tmp; 2314 } 2315 2316 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2317 { 2318 *(u8 *)skb_put(skb, 1) = val; 2319 } 2320 2321 void *skb_push(struct sk_buff *skb, unsigned int len); 2322 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2323 { 2324 skb->data -= len; 2325 skb->len += len; 2326 return skb->data; 2327 } 2328 2329 void *skb_pull(struct sk_buff *skb, unsigned int len); 2330 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2331 { 2332 skb->len -= len; 2333 BUG_ON(skb->len < skb->data_len); 2334 return skb->data += len; 2335 } 2336 2337 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2338 { 2339 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2340 } 2341 2342 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2343 2344 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2345 { 2346 if (len > skb_headlen(skb) && 2347 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2348 return NULL; 2349 skb->len -= len; 2350 return skb->data += len; 2351 } 2352 2353 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2354 { 2355 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2356 } 2357 2358 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2359 { 2360 if (likely(len <= skb_headlen(skb))) 2361 return true; 2362 if (unlikely(len > skb->len)) 2363 return false; 2364 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2365 } 2366 2367 void skb_condense(struct sk_buff *skb); 2368 2369 /** 2370 * skb_headroom - bytes at buffer head 2371 * @skb: buffer to check 2372 * 2373 * Return the number of bytes of free space at the head of an &sk_buff. 2374 */ 2375 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2376 { 2377 return skb->data - skb->head; 2378 } 2379 2380 /** 2381 * skb_tailroom - bytes at buffer end 2382 * @skb: buffer to check 2383 * 2384 * Return the number of bytes of free space at the tail of an sk_buff 2385 */ 2386 static inline int skb_tailroom(const struct sk_buff *skb) 2387 { 2388 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2389 } 2390 2391 /** 2392 * skb_availroom - bytes at buffer end 2393 * @skb: buffer to check 2394 * 2395 * Return the number of bytes of free space at the tail of an sk_buff 2396 * allocated by sk_stream_alloc() 2397 */ 2398 static inline int skb_availroom(const struct sk_buff *skb) 2399 { 2400 if (skb_is_nonlinear(skb)) 2401 return 0; 2402 2403 return skb->end - skb->tail - skb->reserved_tailroom; 2404 } 2405 2406 /** 2407 * skb_reserve - adjust headroom 2408 * @skb: buffer to alter 2409 * @len: bytes to move 2410 * 2411 * Increase the headroom of an empty &sk_buff by reducing the tail 2412 * room. This is only allowed for an empty buffer. 2413 */ 2414 static inline void skb_reserve(struct sk_buff *skb, int len) 2415 { 2416 skb->data += len; 2417 skb->tail += len; 2418 } 2419 2420 /** 2421 * skb_tailroom_reserve - adjust reserved_tailroom 2422 * @skb: buffer to alter 2423 * @mtu: maximum amount of headlen permitted 2424 * @needed_tailroom: minimum amount of reserved_tailroom 2425 * 2426 * Set reserved_tailroom so that headlen can be as large as possible but 2427 * not larger than mtu and tailroom cannot be smaller than 2428 * needed_tailroom. 2429 * The required headroom should already have been reserved before using 2430 * this function. 2431 */ 2432 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2433 unsigned int needed_tailroom) 2434 { 2435 SKB_LINEAR_ASSERT(skb); 2436 if (mtu < skb_tailroom(skb) - needed_tailroom) 2437 /* use at most mtu */ 2438 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2439 else 2440 /* use up to all available space */ 2441 skb->reserved_tailroom = needed_tailroom; 2442 } 2443 2444 #define ENCAP_TYPE_ETHER 0 2445 #define ENCAP_TYPE_IPPROTO 1 2446 2447 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2448 __be16 protocol) 2449 { 2450 skb->inner_protocol = protocol; 2451 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2452 } 2453 2454 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2455 __u8 ipproto) 2456 { 2457 skb->inner_ipproto = ipproto; 2458 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2459 } 2460 2461 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2462 { 2463 skb->inner_mac_header = skb->mac_header; 2464 skb->inner_network_header = skb->network_header; 2465 skb->inner_transport_header = skb->transport_header; 2466 } 2467 2468 static inline void skb_reset_mac_len(struct sk_buff *skb) 2469 { 2470 skb->mac_len = skb->network_header - skb->mac_header; 2471 } 2472 2473 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2474 *skb) 2475 { 2476 return skb->head + skb->inner_transport_header; 2477 } 2478 2479 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2480 { 2481 return skb_inner_transport_header(skb) - skb->data; 2482 } 2483 2484 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2485 { 2486 skb->inner_transport_header = skb->data - skb->head; 2487 } 2488 2489 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2490 const int offset) 2491 { 2492 skb_reset_inner_transport_header(skb); 2493 skb->inner_transport_header += offset; 2494 } 2495 2496 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2497 { 2498 return skb->head + skb->inner_network_header; 2499 } 2500 2501 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2502 { 2503 skb->inner_network_header = skb->data - skb->head; 2504 } 2505 2506 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2507 const int offset) 2508 { 2509 skb_reset_inner_network_header(skb); 2510 skb->inner_network_header += offset; 2511 } 2512 2513 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2514 { 2515 return skb->head + skb->inner_mac_header; 2516 } 2517 2518 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2519 { 2520 skb->inner_mac_header = skb->data - skb->head; 2521 } 2522 2523 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2524 const int offset) 2525 { 2526 skb_reset_inner_mac_header(skb); 2527 skb->inner_mac_header += offset; 2528 } 2529 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2530 { 2531 return skb->transport_header != (typeof(skb->transport_header))~0U; 2532 } 2533 2534 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2535 { 2536 return skb->head + skb->transport_header; 2537 } 2538 2539 static inline void skb_reset_transport_header(struct sk_buff *skb) 2540 { 2541 skb->transport_header = skb->data - skb->head; 2542 } 2543 2544 static inline void skb_set_transport_header(struct sk_buff *skb, 2545 const int offset) 2546 { 2547 skb_reset_transport_header(skb); 2548 skb->transport_header += offset; 2549 } 2550 2551 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2552 { 2553 return skb->head + skb->network_header; 2554 } 2555 2556 static inline void skb_reset_network_header(struct sk_buff *skb) 2557 { 2558 skb->network_header = skb->data - skb->head; 2559 } 2560 2561 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2562 { 2563 skb_reset_network_header(skb); 2564 skb->network_header += offset; 2565 } 2566 2567 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2568 { 2569 return skb->head + skb->mac_header; 2570 } 2571 2572 static inline int skb_mac_offset(const struct sk_buff *skb) 2573 { 2574 return skb_mac_header(skb) - skb->data; 2575 } 2576 2577 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2578 { 2579 return skb->network_header - skb->mac_header; 2580 } 2581 2582 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2583 { 2584 return skb->mac_header != (typeof(skb->mac_header))~0U; 2585 } 2586 2587 static inline void skb_unset_mac_header(struct sk_buff *skb) 2588 { 2589 skb->mac_header = (typeof(skb->mac_header))~0U; 2590 } 2591 2592 static inline void skb_reset_mac_header(struct sk_buff *skb) 2593 { 2594 skb->mac_header = skb->data - skb->head; 2595 } 2596 2597 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2598 { 2599 skb_reset_mac_header(skb); 2600 skb->mac_header += offset; 2601 } 2602 2603 static inline void skb_pop_mac_header(struct sk_buff *skb) 2604 { 2605 skb->mac_header = skb->network_header; 2606 } 2607 2608 static inline void skb_probe_transport_header(struct sk_buff *skb) 2609 { 2610 struct flow_keys_basic keys; 2611 2612 if (skb_transport_header_was_set(skb)) 2613 return; 2614 2615 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2616 NULL, 0, 0, 0, 0)) 2617 skb_set_transport_header(skb, keys.control.thoff); 2618 } 2619 2620 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2621 { 2622 if (skb_mac_header_was_set(skb)) { 2623 const unsigned char *old_mac = skb_mac_header(skb); 2624 2625 skb_set_mac_header(skb, -skb->mac_len); 2626 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2627 } 2628 } 2629 2630 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2631 { 2632 return skb->csum_start - skb_headroom(skb); 2633 } 2634 2635 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2636 { 2637 return skb->head + skb->csum_start; 2638 } 2639 2640 static inline int skb_transport_offset(const struct sk_buff *skb) 2641 { 2642 return skb_transport_header(skb) - skb->data; 2643 } 2644 2645 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2646 { 2647 return skb->transport_header - skb->network_header; 2648 } 2649 2650 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2651 { 2652 return skb->inner_transport_header - skb->inner_network_header; 2653 } 2654 2655 static inline int skb_network_offset(const struct sk_buff *skb) 2656 { 2657 return skb_network_header(skb) - skb->data; 2658 } 2659 2660 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2661 { 2662 return skb_inner_network_header(skb) - skb->data; 2663 } 2664 2665 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2666 { 2667 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2668 } 2669 2670 /* 2671 * CPUs often take a performance hit when accessing unaligned memory 2672 * locations. The actual performance hit varies, it can be small if the 2673 * hardware handles it or large if we have to take an exception and fix it 2674 * in software. 2675 * 2676 * Since an ethernet header is 14 bytes network drivers often end up with 2677 * the IP header at an unaligned offset. The IP header can be aligned by 2678 * shifting the start of the packet by 2 bytes. Drivers should do this 2679 * with: 2680 * 2681 * skb_reserve(skb, NET_IP_ALIGN); 2682 * 2683 * The downside to this alignment of the IP header is that the DMA is now 2684 * unaligned. On some architectures the cost of an unaligned DMA is high 2685 * and this cost outweighs the gains made by aligning the IP header. 2686 * 2687 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2688 * to be overridden. 2689 */ 2690 #ifndef NET_IP_ALIGN 2691 #define NET_IP_ALIGN 2 2692 #endif 2693 2694 /* 2695 * The networking layer reserves some headroom in skb data (via 2696 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2697 * the header has to grow. In the default case, if the header has to grow 2698 * 32 bytes or less we avoid the reallocation. 2699 * 2700 * Unfortunately this headroom changes the DMA alignment of the resulting 2701 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2702 * on some architectures. An architecture can override this value, 2703 * perhaps setting it to a cacheline in size (since that will maintain 2704 * cacheline alignment of the DMA). It must be a power of 2. 2705 * 2706 * Various parts of the networking layer expect at least 32 bytes of 2707 * headroom, you should not reduce this. 2708 * 2709 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2710 * to reduce average number of cache lines per packet. 2711 * get_rps_cpu() for example only access one 64 bytes aligned block : 2712 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2713 */ 2714 #ifndef NET_SKB_PAD 2715 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2716 #endif 2717 2718 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2719 2720 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2721 { 2722 if (WARN_ON(skb_is_nonlinear(skb))) 2723 return; 2724 skb->len = len; 2725 skb_set_tail_pointer(skb, len); 2726 } 2727 2728 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2729 { 2730 __skb_set_length(skb, len); 2731 } 2732 2733 void skb_trim(struct sk_buff *skb, unsigned int len); 2734 2735 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2736 { 2737 if (skb->data_len) 2738 return ___pskb_trim(skb, len); 2739 __skb_trim(skb, len); 2740 return 0; 2741 } 2742 2743 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2744 { 2745 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2746 } 2747 2748 /** 2749 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2750 * @skb: buffer to alter 2751 * @len: new length 2752 * 2753 * This is identical to pskb_trim except that the caller knows that 2754 * the skb is not cloned so we should never get an error due to out- 2755 * of-memory. 2756 */ 2757 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2758 { 2759 int err = pskb_trim(skb, len); 2760 BUG_ON(err); 2761 } 2762 2763 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2764 { 2765 unsigned int diff = len - skb->len; 2766 2767 if (skb_tailroom(skb) < diff) { 2768 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2769 GFP_ATOMIC); 2770 if (ret) 2771 return ret; 2772 } 2773 __skb_set_length(skb, len); 2774 return 0; 2775 } 2776 2777 /** 2778 * skb_orphan - orphan a buffer 2779 * @skb: buffer to orphan 2780 * 2781 * If a buffer currently has an owner then we call the owner's 2782 * destructor function and make the @skb unowned. The buffer continues 2783 * to exist but is no longer charged to its former owner. 2784 */ 2785 static inline void skb_orphan(struct sk_buff *skb) 2786 { 2787 if (skb->destructor) { 2788 skb->destructor(skb); 2789 skb->destructor = NULL; 2790 skb->sk = NULL; 2791 } else { 2792 BUG_ON(skb->sk); 2793 } 2794 } 2795 2796 /** 2797 * skb_orphan_frags - orphan the frags contained in a buffer 2798 * @skb: buffer to orphan frags from 2799 * @gfp_mask: allocation mask for replacement pages 2800 * 2801 * For each frag in the SKB which needs a destructor (i.e. has an 2802 * owner) create a copy of that frag and release the original 2803 * page by calling the destructor. 2804 */ 2805 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2806 { 2807 if (likely(!skb_zcopy(skb))) 2808 return 0; 2809 if (!skb_zcopy_is_nouarg(skb) && 2810 skb_uarg(skb)->callback == msg_zerocopy_callback) 2811 return 0; 2812 return skb_copy_ubufs(skb, gfp_mask); 2813 } 2814 2815 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2816 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2817 { 2818 if (likely(!skb_zcopy(skb))) 2819 return 0; 2820 return skb_copy_ubufs(skb, gfp_mask); 2821 } 2822 2823 /** 2824 * __skb_queue_purge - empty a list 2825 * @list: list to empty 2826 * 2827 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2828 * the list and one reference dropped. This function does not take the 2829 * list lock and the caller must hold the relevant locks to use it. 2830 */ 2831 static inline void __skb_queue_purge(struct sk_buff_head *list) 2832 { 2833 struct sk_buff *skb; 2834 while ((skb = __skb_dequeue(list)) != NULL) 2835 kfree_skb(skb); 2836 } 2837 void skb_queue_purge(struct sk_buff_head *list); 2838 2839 unsigned int skb_rbtree_purge(struct rb_root *root); 2840 2841 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2842 2843 /** 2844 * netdev_alloc_frag - allocate a page fragment 2845 * @fragsz: fragment size 2846 * 2847 * Allocates a frag from a page for receive buffer. 2848 * Uses GFP_ATOMIC allocations. 2849 */ 2850 static inline void *netdev_alloc_frag(unsigned int fragsz) 2851 { 2852 return __netdev_alloc_frag_align(fragsz, ~0u); 2853 } 2854 2855 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 2856 unsigned int align) 2857 { 2858 WARN_ON_ONCE(!is_power_of_2(align)); 2859 return __netdev_alloc_frag_align(fragsz, -align); 2860 } 2861 2862 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2863 gfp_t gfp_mask); 2864 2865 /** 2866 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2867 * @dev: network device to receive on 2868 * @length: length to allocate 2869 * 2870 * Allocate a new &sk_buff and assign it a usage count of one. The 2871 * buffer has unspecified headroom built in. Users should allocate 2872 * the headroom they think they need without accounting for the 2873 * built in space. The built in space is used for optimisations. 2874 * 2875 * %NULL is returned if there is no free memory. Although this function 2876 * allocates memory it can be called from an interrupt. 2877 */ 2878 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2879 unsigned int length) 2880 { 2881 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2882 } 2883 2884 /* legacy helper around __netdev_alloc_skb() */ 2885 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2886 gfp_t gfp_mask) 2887 { 2888 return __netdev_alloc_skb(NULL, length, gfp_mask); 2889 } 2890 2891 /* legacy helper around netdev_alloc_skb() */ 2892 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2893 { 2894 return netdev_alloc_skb(NULL, length); 2895 } 2896 2897 2898 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2899 unsigned int length, gfp_t gfp) 2900 { 2901 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2902 2903 if (NET_IP_ALIGN && skb) 2904 skb_reserve(skb, NET_IP_ALIGN); 2905 return skb; 2906 } 2907 2908 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2909 unsigned int length) 2910 { 2911 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2912 } 2913 2914 static inline void skb_free_frag(void *addr) 2915 { 2916 page_frag_free(addr); 2917 } 2918 2919 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2920 2921 static inline void *napi_alloc_frag(unsigned int fragsz) 2922 { 2923 return __napi_alloc_frag_align(fragsz, ~0u); 2924 } 2925 2926 static inline void *napi_alloc_frag_align(unsigned int fragsz, 2927 unsigned int align) 2928 { 2929 WARN_ON_ONCE(!is_power_of_2(align)); 2930 return __napi_alloc_frag_align(fragsz, -align); 2931 } 2932 2933 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2934 unsigned int length, gfp_t gfp_mask); 2935 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2936 unsigned int length) 2937 { 2938 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2939 } 2940 void napi_consume_skb(struct sk_buff *skb, int budget); 2941 2942 void napi_skb_free_stolen_head(struct sk_buff *skb); 2943 void __kfree_skb_defer(struct sk_buff *skb); 2944 2945 /** 2946 * __dev_alloc_pages - allocate page for network Rx 2947 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2948 * @order: size of the allocation 2949 * 2950 * Allocate a new page. 2951 * 2952 * %NULL is returned if there is no free memory. 2953 */ 2954 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2955 unsigned int order) 2956 { 2957 /* This piece of code contains several assumptions. 2958 * 1. This is for device Rx, therefor a cold page is preferred. 2959 * 2. The expectation is the user wants a compound page. 2960 * 3. If requesting a order 0 page it will not be compound 2961 * due to the check to see if order has a value in prep_new_page 2962 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2963 * code in gfp_to_alloc_flags that should be enforcing this. 2964 */ 2965 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2966 2967 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2968 } 2969 2970 static inline struct page *dev_alloc_pages(unsigned int order) 2971 { 2972 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2973 } 2974 2975 /** 2976 * __dev_alloc_page - allocate a page for network Rx 2977 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2978 * 2979 * Allocate a new page. 2980 * 2981 * %NULL is returned if there is no free memory. 2982 */ 2983 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2984 { 2985 return __dev_alloc_pages(gfp_mask, 0); 2986 } 2987 2988 static inline struct page *dev_alloc_page(void) 2989 { 2990 return dev_alloc_pages(0); 2991 } 2992 2993 /** 2994 * dev_page_is_reusable - check whether a page can be reused for network Rx 2995 * @page: the page to test 2996 * 2997 * A page shouldn't be considered for reusing/recycling if it was allocated 2998 * under memory pressure or at a distant memory node. 2999 * 3000 * Returns false if this page should be returned to page allocator, true 3001 * otherwise. 3002 */ 3003 static inline bool dev_page_is_reusable(const struct page *page) 3004 { 3005 return likely(page_to_nid(page) == numa_mem_id() && 3006 !page_is_pfmemalloc(page)); 3007 } 3008 3009 /** 3010 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 3011 * @page: The page that was allocated from skb_alloc_page 3012 * @skb: The skb that may need pfmemalloc set 3013 */ 3014 static inline void skb_propagate_pfmemalloc(const struct page *page, 3015 struct sk_buff *skb) 3016 { 3017 if (page_is_pfmemalloc(page)) 3018 skb->pfmemalloc = true; 3019 } 3020 3021 /** 3022 * skb_frag_off() - Returns the offset of a skb fragment 3023 * @frag: the paged fragment 3024 */ 3025 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3026 { 3027 return frag->bv_offset; 3028 } 3029 3030 /** 3031 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3032 * @frag: skb fragment 3033 * @delta: value to add 3034 */ 3035 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3036 { 3037 frag->bv_offset += delta; 3038 } 3039 3040 /** 3041 * skb_frag_off_set() - Sets the offset of a skb fragment 3042 * @frag: skb fragment 3043 * @offset: offset of fragment 3044 */ 3045 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3046 { 3047 frag->bv_offset = offset; 3048 } 3049 3050 /** 3051 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3052 * @fragto: skb fragment where offset is set 3053 * @fragfrom: skb fragment offset is copied from 3054 */ 3055 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3056 const skb_frag_t *fragfrom) 3057 { 3058 fragto->bv_offset = fragfrom->bv_offset; 3059 } 3060 3061 /** 3062 * skb_frag_page - retrieve the page referred to by a paged fragment 3063 * @frag: the paged fragment 3064 * 3065 * Returns the &struct page associated with @frag. 3066 */ 3067 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3068 { 3069 return frag->bv_page; 3070 } 3071 3072 /** 3073 * __skb_frag_ref - take an addition reference on a paged fragment. 3074 * @frag: the paged fragment 3075 * 3076 * Takes an additional reference on the paged fragment @frag. 3077 */ 3078 static inline void __skb_frag_ref(skb_frag_t *frag) 3079 { 3080 get_page(skb_frag_page(frag)); 3081 } 3082 3083 /** 3084 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3085 * @skb: the buffer 3086 * @f: the fragment offset. 3087 * 3088 * Takes an additional reference on the @f'th paged fragment of @skb. 3089 */ 3090 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3091 { 3092 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3093 } 3094 3095 /** 3096 * __skb_frag_unref - release a reference on a paged fragment. 3097 * @frag: the paged fragment 3098 * @recycle: recycle the page if allocated via page_pool 3099 * 3100 * Releases a reference on the paged fragment @frag 3101 * or recycles the page via the page_pool API. 3102 */ 3103 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) 3104 { 3105 struct page *page = skb_frag_page(frag); 3106 3107 #ifdef CONFIG_PAGE_POOL 3108 if (recycle && page_pool_return_skb_page(page)) 3109 return; 3110 #endif 3111 put_page(page); 3112 } 3113 3114 /** 3115 * skb_frag_unref - release a reference on a paged fragment of an skb. 3116 * @skb: the buffer 3117 * @f: the fragment offset 3118 * 3119 * Releases a reference on the @f'th paged fragment of @skb. 3120 */ 3121 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3122 { 3123 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle); 3124 } 3125 3126 /** 3127 * skb_frag_address - gets the address of the data contained in a paged fragment 3128 * @frag: the paged fragment buffer 3129 * 3130 * Returns the address of the data within @frag. The page must already 3131 * be mapped. 3132 */ 3133 static inline void *skb_frag_address(const skb_frag_t *frag) 3134 { 3135 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3136 } 3137 3138 /** 3139 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3140 * @frag: the paged fragment buffer 3141 * 3142 * Returns the address of the data within @frag. Checks that the page 3143 * is mapped and returns %NULL otherwise. 3144 */ 3145 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3146 { 3147 void *ptr = page_address(skb_frag_page(frag)); 3148 if (unlikely(!ptr)) 3149 return NULL; 3150 3151 return ptr + skb_frag_off(frag); 3152 } 3153 3154 /** 3155 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3156 * @fragto: skb fragment where page is set 3157 * @fragfrom: skb fragment page is copied from 3158 */ 3159 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3160 const skb_frag_t *fragfrom) 3161 { 3162 fragto->bv_page = fragfrom->bv_page; 3163 } 3164 3165 /** 3166 * __skb_frag_set_page - sets the page contained in a paged fragment 3167 * @frag: the paged fragment 3168 * @page: the page to set 3169 * 3170 * Sets the fragment @frag to contain @page. 3171 */ 3172 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3173 { 3174 frag->bv_page = page; 3175 } 3176 3177 /** 3178 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3179 * @skb: the buffer 3180 * @f: the fragment offset 3181 * @page: the page to set 3182 * 3183 * Sets the @f'th fragment of @skb to contain @page. 3184 */ 3185 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3186 struct page *page) 3187 { 3188 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3189 } 3190 3191 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3192 3193 /** 3194 * skb_frag_dma_map - maps a paged fragment via the DMA API 3195 * @dev: the device to map the fragment to 3196 * @frag: the paged fragment to map 3197 * @offset: the offset within the fragment (starting at the 3198 * fragment's own offset) 3199 * @size: the number of bytes to map 3200 * @dir: the direction of the mapping (``PCI_DMA_*``) 3201 * 3202 * Maps the page associated with @frag to @device. 3203 */ 3204 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3205 const skb_frag_t *frag, 3206 size_t offset, size_t size, 3207 enum dma_data_direction dir) 3208 { 3209 return dma_map_page(dev, skb_frag_page(frag), 3210 skb_frag_off(frag) + offset, size, dir); 3211 } 3212 3213 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3214 gfp_t gfp_mask) 3215 { 3216 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3217 } 3218 3219 3220 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3221 gfp_t gfp_mask) 3222 { 3223 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3224 } 3225 3226 3227 /** 3228 * skb_clone_writable - is the header of a clone writable 3229 * @skb: buffer to check 3230 * @len: length up to which to write 3231 * 3232 * Returns true if modifying the header part of the cloned buffer 3233 * does not requires the data to be copied. 3234 */ 3235 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3236 { 3237 return !skb_header_cloned(skb) && 3238 skb_headroom(skb) + len <= skb->hdr_len; 3239 } 3240 3241 static inline int skb_try_make_writable(struct sk_buff *skb, 3242 unsigned int write_len) 3243 { 3244 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3245 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3246 } 3247 3248 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3249 int cloned) 3250 { 3251 int delta = 0; 3252 3253 if (headroom > skb_headroom(skb)) 3254 delta = headroom - skb_headroom(skb); 3255 3256 if (delta || cloned) 3257 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3258 GFP_ATOMIC); 3259 return 0; 3260 } 3261 3262 /** 3263 * skb_cow - copy header of skb when it is required 3264 * @skb: buffer to cow 3265 * @headroom: needed headroom 3266 * 3267 * If the skb passed lacks sufficient headroom or its data part 3268 * is shared, data is reallocated. If reallocation fails, an error 3269 * is returned and original skb is not changed. 3270 * 3271 * The result is skb with writable area skb->head...skb->tail 3272 * and at least @headroom of space at head. 3273 */ 3274 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3275 { 3276 return __skb_cow(skb, headroom, skb_cloned(skb)); 3277 } 3278 3279 /** 3280 * skb_cow_head - skb_cow but only making the head writable 3281 * @skb: buffer to cow 3282 * @headroom: needed headroom 3283 * 3284 * This function is identical to skb_cow except that we replace the 3285 * skb_cloned check by skb_header_cloned. It should be used when 3286 * you only need to push on some header and do not need to modify 3287 * the data. 3288 */ 3289 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3290 { 3291 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3292 } 3293 3294 /** 3295 * skb_padto - pad an skbuff up to a minimal size 3296 * @skb: buffer to pad 3297 * @len: minimal length 3298 * 3299 * Pads up a buffer to ensure the trailing bytes exist and are 3300 * blanked. If the buffer already contains sufficient data it 3301 * is untouched. Otherwise it is extended. Returns zero on 3302 * success. The skb is freed on error. 3303 */ 3304 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3305 { 3306 unsigned int size = skb->len; 3307 if (likely(size >= len)) 3308 return 0; 3309 return skb_pad(skb, len - size); 3310 } 3311 3312 /** 3313 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3314 * @skb: buffer to pad 3315 * @len: minimal length 3316 * @free_on_error: free buffer on error 3317 * 3318 * Pads up a buffer to ensure the trailing bytes exist and are 3319 * blanked. If the buffer already contains sufficient data it 3320 * is untouched. Otherwise it is extended. Returns zero on 3321 * success. The skb is freed on error if @free_on_error is true. 3322 */ 3323 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3324 unsigned int len, 3325 bool free_on_error) 3326 { 3327 unsigned int size = skb->len; 3328 3329 if (unlikely(size < len)) { 3330 len -= size; 3331 if (__skb_pad(skb, len, free_on_error)) 3332 return -ENOMEM; 3333 __skb_put(skb, len); 3334 } 3335 return 0; 3336 } 3337 3338 /** 3339 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3340 * @skb: buffer to pad 3341 * @len: minimal length 3342 * 3343 * Pads up a buffer to ensure the trailing bytes exist and are 3344 * blanked. If the buffer already contains sufficient data it 3345 * is untouched. Otherwise it is extended. Returns zero on 3346 * success. The skb is freed on error. 3347 */ 3348 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3349 { 3350 return __skb_put_padto(skb, len, true); 3351 } 3352 3353 static inline int skb_add_data(struct sk_buff *skb, 3354 struct iov_iter *from, int copy) 3355 { 3356 const int off = skb->len; 3357 3358 if (skb->ip_summed == CHECKSUM_NONE) { 3359 __wsum csum = 0; 3360 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3361 &csum, from)) { 3362 skb->csum = csum_block_add(skb->csum, csum, off); 3363 return 0; 3364 } 3365 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3366 return 0; 3367 3368 __skb_trim(skb, off); 3369 return -EFAULT; 3370 } 3371 3372 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3373 const struct page *page, int off) 3374 { 3375 if (skb_zcopy(skb)) 3376 return false; 3377 if (i) { 3378 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3379 3380 return page == skb_frag_page(frag) && 3381 off == skb_frag_off(frag) + skb_frag_size(frag); 3382 } 3383 return false; 3384 } 3385 3386 static inline int __skb_linearize(struct sk_buff *skb) 3387 { 3388 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3389 } 3390 3391 /** 3392 * skb_linearize - convert paged skb to linear one 3393 * @skb: buffer to linarize 3394 * 3395 * If there is no free memory -ENOMEM is returned, otherwise zero 3396 * is returned and the old skb data released. 3397 */ 3398 static inline int skb_linearize(struct sk_buff *skb) 3399 { 3400 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3401 } 3402 3403 /** 3404 * skb_has_shared_frag - can any frag be overwritten 3405 * @skb: buffer to test 3406 * 3407 * Return true if the skb has at least one frag that might be modified 3408 * by an external entity (as in vmsplice()/sendfile()) 3409 */ 3410 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3411 { 3412 return skb_is_nonlinear(skb) && 3413 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3414 } 3415 3416 /** 3417 * skb_linearize_cow - make sure skb is linear and writable 3418 * @skb: buffer to process 3419 * 3420 * If there is no free memory -ENOMEM is returned, otherwise zero 3421 * is returned and the old skb data released. 3422 */ 3423 static inline int skb_linearize_cow(struct sk_buff *skb) 3424 { 3425 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3426 __skb_linearize(skb) : 0; 3427 } 3428 3429 static __always_inline void 3430 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3431 unsigned int off) 3432 { 3433 if (skb->ip_summed == CHECKSUM_COMPLETE) 3434 skb->csum = csum_block_sub(skb->csum, 3435 csum_partial(start, len, 0), off); 3436 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3437 skb_checksum_start_offset(skb) < 0) 3438 skb->ip_summed = CHECKSUM_NONE; 3439 } 3440 3441 /** 3442 * skb_postpull_rcsum - update checksum for received skb after pull 3443 * @skb: buffer to update 3444 * @start: start of data before pull 3445 * @len: length of data pulled 3446 * 3447 * After doing a pull on a received packet, you need to call this to 3448 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3449 * CHECKSUM_NONE so that it can be recomputed from scratch. 3450 */ 3451 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3452 const void *start, unsigned int len) 3453 { 3454 __skb_postpull_rcsum(skb, start, len, 0); 3455 } 3456 3457 static __always_inline void 3458 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3459 unsigned int off) 3460 { 3461 if (skb->ip_summed == CHECKSUM_COMPLETE) 3462 skb->csum = csum_block_add(skb->csum, 3463 csum_partial(start, len, 0), off); 3464 } 3465 3466 /** 3467 * skb_postpush_rcsum - update checksum for received skb after push 3468 * @skb: buffer to update 3469 * @start: start of data after push 3470 * @len: length of data pushed 3471 * 3472 * After doing a push on a received packet, you need to call this to 3473 * update the CHECKSUM_COMPLETE checksum. 3474 */ 3475 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3476 const void *start, unsigned int len) 3477 { 3478 __skb_postpush_rcsum(skb, start, len, 0); 3479 } 3480 3481 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3482 3483 /** 3484 * skb_push_rcsum - push skb and update receive checksum 3485 * @skb: buffer to update 3486 * @len: length of data pulled 3487 * 3488 * This function performs an skb_push on the packet and updates 3489 * the CHECKSUM_COMPLETE checksum. It should be used on 3490 * receive path processing instead of skb_push unless you know 3491 * that the checksum difference is zero (e.g., a valid IP header) 3492 * or you are setting ip_summed to CHECKSUM_NONE. 3493 */ 3494 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3495 { 3496 skb_push(skb, len); 3497 skb_postpush_rcsum(skb, skb->data, len); 3498 return skb->data; 3499 } 3500 3501 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3502 /** 3503 * pskb_trim_rcsum - trim received skb and update checksum 3504 * @skb: buffer to trim 3505 * @len: new length 3506 * 3507 * This is exactly the same as pskb_trim except that it ensures the 3508 * checksum of received packets are still valid after the operation. 3509 * It can change skb pointers. 3510 */ 3511 3512 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3513 { 3514 if (likely(len >= skb->len)) 3515 return 0; 3516 return pskb_trim_rcsum_slow(skb, len); 3517 } 3518 3519 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3520 { 3521 if (skb->ip_summed == CHECKSUM_COMPLETE) 3522 skb->ip_summed = CHECKSUM_NONE; 3523 __skb_trim(skb, len); 3524 return 0; 3525 } 3526 3527 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3528 { 3529 if (skb->ip_summed == CHECKSUM_COMPLETE) 3530 skb->ip_summed = CHECKSUM_NONE; 3531 return __skb_grow(skb, len); 3532 } 3533 3534 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3535 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3536 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3537 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3538 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3539 3540 #define skb_queue_walk(queue, skb) \ 3541 for (skb = (queue)->next; \ 3542 skb != (struct sk_buff *)(queue); \ 3543 skb = skb->next) 3544 3545 #define skb_queue_walk_safe(queue, skb, tmp) \ 3546 for (skb = (queue)->next, tmp = skb->next; \ 3547 skb != (struct sk_buff *)(queue); \ 3548 skb = tmp, tmp = skb->next) 3549 3550 #define skb_queue_walk_from(queue, skb) \ 3551 for (; skb != (struct sk_buff *)(queue); \ 3552 skb = skb->next) 3553 3554 #define skb_rbtree_walk(skb, root) \ 3555 for (skb = skb_rb_first(root); skb != NULL; \ 3556 skb = skb_rb_next(skb)) 3557 3558 #define skb_rbtree_walk_from(skb) \ 3559 for (; skb != NULL; \ 3560 skb = skb_rb_next(skb)) 3561 3562 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3563 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3564 skb = tmp) 3565 3566 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3567 for (tmp = skb->next; \ 3568 skb != (struct sk_buff *)(queue); \ 3569 skb = tmp, tmp = skb->next) 3570 3571 #define skb_queue_reverse_walk(queue, skb) \ 3572 for (skb = (queue)->prev; \ 3573 skb != (struct sk_buff *)(queue); \ 3574 skb = skb->prev) 3575 3576 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3577 for (skb = (queue)->prev, tmp = skb->prev; \ 3578 skb != (struct sk_buff *)(queue); \ 3579 skb = tmp, tmp = skb->prev) 3580 3581 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3582 for (tmp = skb->prev; \ 3583 skb != (struct sk_buff *)(queue); \ 3584 skb = tmp, tmp = skb->prev) 3585 3586 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3587 { 3588 return skb_shinfo(skb)->frag_list != NULL; 3589 } 3590 3591 static inline void skb_frag_list_init(struct sk_buff *skb) 3592 { 3593 skb_shinfo(skb)->frag_list = NULL; 3594 } 3595 3596 #define skb_walk_frags(skb, iter) \ 3597 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3598 3599 3600 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3601 int *err, long *timeo_p, 3602 const struct sk_buff *skb); 3603 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3604 struct sk_buff_head *queue, 3605 unsigned int flags, 3606 int *off, int *err, 3607 struct sk_buff **last); 3608 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3609 struct sk_buff_head *queue, 3610 unsigned int flags, int *off, int *err, 3611 struct sk_buff **last); 3612 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3613 struct sk_buff_head *sk_queue, 3614 unsigned int flags, int *off, int *err); 3615 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3616 int *err); 3617 __poll_t datagram_poll(struct file *file, struct socket *sock, 3618 struct poll_table_struct *wait); 3619 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3620 struct iov_iter *to, int size); 3621 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3622 struct msghdr *msg, int size) 3623 { 3624 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3625 } 3626 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3627 struct msghdr *msg); 3628 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3629 struct iov_iter *to, int len, 3630 struct ahash_request *hash); 3631 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3632 struct iov_iter *from, int len); 3633 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3634 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3635 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3636 static inline void skb_free_datagram_locked(struct sock *sk, 3637 struct sk_buff *skb) 3638 { 3639 __skb_free_datagram_locked(sk, skb, 0); 3640 } 3641 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3642 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3643 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3644 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3645 int len); 3646 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3647 struct pipe_inode_info *pipe, unsigned int len, 3648 unsigned int flags); 3649 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3650 int len); 3651 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len); 3652 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3653 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3654 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3655 int len, int hlen); 3656 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3657 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3658 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3659 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3660 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3661 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3662 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3663 unsigned int offset); 3664 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3665 int skb_ensure_writable(struct sk_buff *skb, int write_len); 3666 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3667 int skb_vlan_pop(struct sk_buff *skb); 3668 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3669 int skb_eth_pop(struct sk_buff *skb); 3670 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3671 const unsigned char *src); 3672 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3673 int mac_len, bool ethernet); 3674 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3675 bool ethernet); 3676 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3677 int skb_mpls_dec_ttl(struct sk_buff *skb); 3678 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3679 gfp_t gfp); 3680 3681 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3682 { 3683 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3684 } 3685 3686 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3687 { 3688 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3689 } 3690 3691 struct skb_checksum_ops { 3692 __wsum (*update)(const void *mem, int len, __wsum wsum); 3693 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3694 }; 3695 3696 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3697 3698 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3699 __wsum csum, const struct skb_checksum_ops *ops); 3700 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3701 __wsum csum); 3702 3703 static inline void * __must_check 3704 __skb_header_pointer(const struct sk_buff *skb, int offset, int len, 3705 const void *data, int hlen, void *buffer) 3706 { 3707 if (likely(hlen - offset >= len)) 3708 return (void *)data + offset; 3709 3710 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) 3711 return NULL; 3712 3713 return buffer; 3714 } 3715 3716 static inline void * __must_check 3717 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3718 { 3719 return __skb_header_pointer(skb, offset, len, skb->data, 3720 skb_headlen(skb), buffer); 3721 } 3722 3723 /** 3724 * skb_needs_linearize - check if we need to linearize a given skb 3725 * depending on the given device features. 3726 * @skb: socket buffer to check 3727 * @features: net device features 3728 * 3729 * Returns true if either: 3730 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3731 * 2. skb is fragmented and the device does not support SG. 3732 */ 3733 static inline bool skb_needs_linearize(struct sk_buff *skb, 3734 netdev_features_t features) 3735 { 3736 return skb_is_nonlinear(skb) && 3737 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3738 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3739 } 3740 3741 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3742 void *to, 3743 const unsigned int len) 3744 { 3745 memcpy(to, skb->data, len); 3746 } 3747 3748 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3749 const int offset, void *to, 3750 const unsigned int len) 3751 { 3752 memcpy(to, skb->data + offset, len); 3753 } 3754 3755 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3756 const void *from, 3757 const unsigned int len) 3758 { 3759 memcpy(skb->data, from, len); 3760 } 3761 3762 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3763 const int offset, 3764 const void *from, 3765 const unsigned int len) 3766 { 3767 memcpy(skb->data + offset, from, len); 3768 } 3769 3770 void skb_init(void); 3771 3772 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3773 { 3774 return skb->tstamp; 3775 } 3776 3777 /** 3778 * skb_get_timestamp - get timestamp from a skb 3779 * @skb: skb to get stamp from 3780 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3781 * 3782 * Timestamps are stored in the skb as offsets to a base timestamp. 3783 * This function converts the offset back to a struct timeval and stores 3784 * it in stamp. 3785 */ 3786 static inline void skb_get_timestamp(const struct sk_buff *skb, 3787 struct __kernel_old_timeval *stamp) 3788 { 3789 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3790 } 3791 3792 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3793 struct __kernel_sock_timeval *stamp) 3794 { 3795 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3796 3797 stamp->tv_sec = ts.tv_sec; 3798 stamp->tv_usec = ts.tv_nsec / 1000; 3799 } 3800 3801 static inline void skb_get_timestampns(const struct sk_buff *skb, 3802 struct __kernel_old_timespec *stamp) 3803 { 3804 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3805 3806 stamp->tv_sec = ts.tv_sec; 3807 stamp->tv_nsec = ts.tv_nsec; 3808 } 3809 3810 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3811 struct __kernel_timespec *stamp) 3812 { 3813 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3814 3815 stamp->tv_sec = ts.tv_sec; 3816 stamp->tv_nsec = ts.tv_nsec; 3817 } 3818 3819 static inline void __net_timestamp(struct sk_buff *skb) 3820 { 3821 skb->tstamp = ktime_get_real(); 3822 } 3823 3824 static inline ktime_t net_timedelta(ktime_t t) 3825 { 3826 return ktime_sub(ktime_get_real(), t); 3827 } 3828 3829 static inline ktime_t net_invalid_timestamp(void) 3830 { 3831 return 0; 3832 } 3833 3834 static inline u8 skb_metadata_len(const struct sk_buff *skb) 3835 { 3836 return skb_shinfo(skb)->meta_len; 3837 } 3838 3839 static inline void *skb_metadata_end(const struct sk_buff *skb) 3840 { 3841 return skb_mac_header(skb); 3842 } 3843 3844 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3845 const struct sk_buff *skb_b, 3846 u8 meta_len) 3847 { 3848 const void *a = skb_metadata_end(skb_a); 3849 const void *b = skb_metadata_end(skb_b); 3850 /* Using more efficient varaiant than plain call to memcmp(). */ 3851 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3852 u64 diffs = 0; 3853 3854 switch (meta_len) { 3855 #define __it(x, op) (x -= sizeof(u##op)) 3856 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3857 case 32: diffs |= __it_diff(a, b, 64); 3858 fallthrough; 3859 case 24: diffs |= __it_diff(a, b, 64); 3860 fallthrough; 3861 case 16: diffs |= __it_diff(a, b, 64); 3862 fallthrough; 3863 case 8: diffs |= __it_diff(a, b, 64); 3864 break; 3865 case 28: diffs |= __it_diff(a, b, 64); 3866 fallthrough; 3867 case 20: diffs |= __it_diff(a, b, 64); 3868 fallthrough; 3869 case 12: diffs |= __it_diff(a, b, 64); 3870 fallthrough; 3871 case 4: diffs |= __it_diff(a, b, 32); 3872 break; 3873 } 3874 return diffs; 3875 #else 3876 return memcmp(a - meta_len, b - meta_len, meta_len); 3877 #endif 3878 } 3879 3880 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3881 const struct sk_buff *skb_b) 3882 { 3883 u8 len_a = skb_metadata_len(skb_a); 3884 u8 len_b = skb_metadata_len(skb_b); 3885 3886 if (!(len_a | len_b)) 3887 return false; 3888 3889 return len_a != len_b ? 3890 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3891 } 3892 3893 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3894 { 3895 skb_shinfo(skb)->meta_len = meta_len; 3896 } 3897 3898 static inline void skb_metadata_clear(struct sk_buff *skb) 3899 { 3900 skb_metadata_set(skb, 0); 3901 } 3902 3903 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3904 3905 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3906 3907 void skb_clone_tx_timestamp(struct sk_buff *skb); 3908 bool skb_defer_rx_timestamp(struct sk_buff *skb); 3909 3910 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3911 3912 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3913 { 3914 } 3915 3916 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3917 { 3918 return false; 3919 } 3920 3921 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3922 3923 /** 3924 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3925 * 3926 * PHY drivers may accept clones of transmitted packets for 3927 * timestamping via their phy_driver.txtstamp method. These drivers 3928 * must call this function to return the skb back to the stack with a 3929 * timestamp. 3930 * 3931 * @skb: clone of the original outgoing packet 3932 * @hwtstamps: hardware time stamps 3933 * 3934 */ 3935 void skb_complete_tx_timestamp(struct sk_buff *skb, 3936 struct skb_shared_hwtstamps *hwtstamps); 3937 3938 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 3939 struct skb_shared_hwtstamps *hwtstamps, 3940 struct sock *sk, int tstype); 3941 3942 /** 3943 * skb_tstamp_tx - queue clone of skb with send time stamps 3944 * @orig_skb: the original outgoing packet 3945 * @hwtstamps: hardware time stamps, may be NULL if not available 3946 * 3947 * If the skb has a socket associated, then this function clones the 3948 * skb (thus sharing the actual data and optional structures), stores 3949 * the optional hardware time stamping information (if non NULL) or 3950 * generates a software time stamp (otherwise), then queues the clone 3951 * to the error queue of the socket. Errors are silently ignored. 3952 */ 3953 void skb_tstamp_tx(struct sk_buff *orig_skb, 3954 struct skb_shared_hwtstamps *hwtstamps); 3955 3956 /** 3957 * skb_tx_timestamp() - Driver hook for transmit timestamping 3958 * 3959 * Ethernet MAC Drivers should call this function in their hard_xmit() 3960 * function immediately before giving the sk_buff to the MAC hardware. 3961 * 3962 * Specifically, one should make absolutely sure that this function is 3963 * called before TX completion of this packet can trigger. Otherwise 3964 * the packet could potentially already be freed. 3965 * 3966 * @skb: A socket buffer. 3967 */ 3968 static inline void skb_tx_timestamp(struct sk_buff *skb) 3969 { 3970 skb_clone_tx_timestamp(skb); 3971 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3972 skb_tstamp_tx(skb, NULL); 3973 } 3974 3975 /** 3976 * skb_complete_wifi_ack - deliver skb with wifi status 3977 * 3978 * @skb: the original outgoing packet 3979 * @acked: ack status 3980 * 3981 */ 3982 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3983 3984 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3985 __sum16 __skb_checksum_complete(struct sk_buff *skb); 3986 3987 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3988 { 3989 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3990 skb->csum_valid || 3991 (skb->ip_summed == CHECKSUM_PARTIAL && 3992 skb_checksum_start_offset(skb) >= 0)); 3993 } 3994 3995 /** 3996 * skb_checksum_complete - Calculate checksum of an entire packet 3997 * @skb: packet to process 3998 * 3999 * This function calculates the checksum over the entire packet plus 4000 * the value of skb->csum. The latter can be used to supply the 4001 * checksum of a pseudo header as used by TCP/UDP. It returns the 4002 * checksum. 4003 * 4004 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 4005 * this function can be used to verify that checksum on received 4006 * packets. In that case the function should return zero if the 4007 * checksum is correct. In particular, this function will return zero 4008 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 4009 * hardware has already verified the correctness of the checksum. 4010 */ 4011 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 4012 { 4013 return skb_csum_unnecessary(skb) ? 4014 0 : __skb_checksum_complete(skb); 4015 } 4016 4017 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 4018 { 4019 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4020 if (skb->csum_level == 0) 4021 skb->ip_summed = CHECKSUM_NONE; 4022 else 4023 skb->csum_level--; 4024 } 4025 } 4026 4027 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4028 { 4029 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4030 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4031 skb->csum_level++; 4032 } else if (skb->ip_summed == CHECKSUM_NONE) { 4033 skb->ip_summed = CHECKSUM_UNNECESSARY; 4034 skb->csum_level = 0; 4035 } 4036 } 4037 4038 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4039 { 4040 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4041 skb->ip_summed = CHECKSUM_NONE; 4042 skb->csum_level = 0; 4043 } 4044 } 4045 4046 /* Check if we need to perform checksum complete validation. 4047 * 4048 * Returns true if checksum complete is needed, false otherwise 4049 * (either checksum is unnecessary or zero checksum is allowed). 4050 */ 4051 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4052 bool zero_okay, 4053 __sum16 check) 4054 { 4055 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4056 skb->csum_valid = 1; 4057 __skb_decr_checksum_unnecessary(skb); 4058 return false; 4059 } 4060 4061 return true; 4062 } 4063 4064 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4065 * in checksum_init. 4066 */ 4067 #define CHECKSUM_BREAK 76 4068 4069 /* Unset checksum-complete 4070 * 4071 * Unset checksum complete can be done when packet is being modified 4072 * (uncompressed for instance) and checksum-complete value is 4073 * invalidated. 4074 */ 4075 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4076 { 4077 if (skb->ip_summed == CHECKSUM_COMPLETE) 4078 skb->ip_summed = CHECKSUM_NONE; 4079 } 4080 4081 /* Validate (init) checksum based on checksum complete. 4082 * 4083 * Return values: 4084 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4085 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4086 * checksum is stored in skb->csum for use in __skb_checksum_complete 4087 * non-zero: value of invalid checksum 4088 * 4089 */ 4090 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4091 bool complete, 4092 __wsum psum) 4093 { 4094 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4095 if (!csum_fold(csum_add(psum, skb->csum))) { 4096 skb->csum_valid = 1; 4097 return 0; 4098 } 4099 } 4100 4101 skb->csum = psum; 4102 4103 if (complete || skb->len <= CHECKSUM_BREAK) { 4104 __sum16 csum; 4105 4106 csum = __skb_checksum_complete(skb); 4107 skb->csum_valid = !csum; 4108 return csum; 4109 } 4110 4111 return 0; 4112 } 4113 4114 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4115 { 4116 return 0; 4117 } 4118 4119 /* Perform checksum validate (init). Note that this is a macro since we only 4120 * want to calculate the pseudo header which is an input function if necessary. 4121 * First we try to validate without any computation (checksum unnecessary) and 4122 * then calculate based on checksum complete calling the function to compute 4123 * pseudo header. 4124 * 4125 * Return values: 4126 * 0: checksum is validated or try to in skb_checksum_complete 4127 * non-zero: value of invalid checksum 4128 */ 4129 #define __skb_checksum_validate(skb, proto, complete, \ 4130 zero_okay, check, compute_pseudo) \ 4131 ({ \ 4132 __sum16 __ret = 0; \ 4133 skb->csum_valid = 0; \ 4134 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4135 __ret = __skb_checksum_validate_complete(skb, \ 4136 complete, compute_pseudo(skb, proto)); \ 4137 __ret; \ 4138 }) 4139 4140 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4141 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4142 4143 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4144 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4145 4146 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4147 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4148 4149 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4150 compute_pseudo) \ 4151 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4152 4153 #define skb_checksum_simple_validate(skb) \ 4154 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4155 4156 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4157 { 4158 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4159 } 4160 4161 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4162 { 4163 skb->csum = ~pseudo; 4164 skb->ip_summed = CHECKSUM_COMPLETE; 4165 } 4166 4167 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4168 do { \ 4169 if (__skb_checksum_convert_check(skb)) \ 4170 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4171 } while (0) 4172 4173 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4174 u16 start, u16 offset) 4175 { 4176 skb->ip_summed = CHECKSUM_PARTIAL; 4177 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4178 skb->csum_offset = offset - start; 4179 } 4180 4181 /* Update skbuf and packet to reflect the remote checksum offload operation. 4182 * When called, ptr indicates the starting point for skb->csum when 4183 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4184 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4185 */ 4186 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4187 int start, int offset, bool nopartial) 4188 { 4189 __wsum delta; 4190 4191 if (!nopartial) { 4192 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4193 return; 4194 } 4195 4196 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4197 __skb_checksum_complete(skb); 4198 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4199 } 4200 4201 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4202 4203 /* Adjust skb->csum since we changed the packet */ 4204 skb->csum = csum_add(skb->csum, delta); 4205 } 4206 4207 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4208 { 4209 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4210 return (void *)(skb->_nfct & NFCT_PTRMASK); 4211 #else 4212 return NULL; 4213 #endif 4214 } 4215 4216 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4217 { 4218 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4219 return skb->_nfct; 4220 #else 4221 return 0UL; 4222 #endif 4223 } 4224 4225 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4226 { 4227 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4228 skb->slow_gro |= !!nfct; 4229 skb->_nfct = nfct; 4230 #endif 4231 } 4232 4233 #ifdef CONFIG_SKB_EXTENSIONS 4234 enum skb_ext_id { 4235 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4236 SKB_EXT_BRIDGE_NF, 4237 #endif 4238 #ifdef CONFIG_XFRM 4239 SKB_EXT_SEC_PATH, 4240 #endif 4241 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4242 TC_SKB_EXT, 4243 #endif 4244 #if IS_ENABLED(CONFIG_MPTCP) 4245 SKB_EXT_MPTCP, 4246 #endif 4247 #if IS_ENABLED(CONFIG_MCTP_FLOWS) 4248 SKB_EXT_MCTP, 4249 #endif 4250 SKB_EXT_NUM, /* must be last */ 4251 }; 4252 4253 /** 4254 * struct skb_ext - sk_buff extensions 4255 * @refcnt: 1 on allocation, deallocated on 0 4256 * @offset: offset to add to @data to obtain extension address 4257 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4258 * @data: start of extension data, variable sized 4259 * 4260 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4261 * to use 'u8' types while allowing up to 2kb worth of extension data. 4262 */ 4263 struct skb_ext { 4264 refcount_t refcnt; 4265 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4266 u8 chunks; /* same */ 4267 char data[] __aligned(8); 4268 }; 4269 4270 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4271 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4272 struct skb_ext *ext); 4273 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4274 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4275 void __skb_ext_put(struct skb_ext *ext); 4276 4277 static inline void skb_ext_put(struct sk_buff *skb) 4278 { 4279 if (skb->active_extensions) 4280 __skb_ext_put(skb->extensions); 4281 } 4282 4283 static inline void __skb_ext_copy(struct sk_buff *dst, 4284 const struct sk_buff *src) 4285 { 4286 dst->active_extensions = src->active_extensions; 4287 4288 if (src->active_extensions) { 4289 struct skb_ext *ext = src->extensions; 4290 4291 refcount_inc(&ext->refcnt); 4292 dst->extensions = ext; 4293 } 4294 } 4295 4296 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4297 { 4298 skb_ext_put(dst); 4299 __skb_ext_copy(dst, src); 4300 } 4301 4302 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4303 { 4304 return !!ext->offset[i]; 4305 } 4306 4307 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4308 { 4309 return skb->active_extensions & (1 << id); 4310 } 4311 4312 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4313 { 4314 if (skb_ext_exist(skb, id)) 4315 __skb_ext_del(skb, id); 4316 } 4317 4318 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4319 { 4320 if (skb_ext_exist(skb, id)) { 4321 struct skb_ext *ext = skb->extensions; 4322 4323 return (void *)ext + (ext->offset[id] << 3); 4324 } 4325 4326 return NULL; 4327 } 4328 4329 static inline void skb_ext_reset(struct sk_buff *skb) 4330 { 4331 if (unlikely(skb->active_extensions)) { 4332 __skb_ext_put(skb->extensions); 4333 skb->active_extensions = 0; 4334 } 4335 } 4336 4337 static inline bool skb_has_extensions(struct sk_buff *skb) 4338 { 4339 return unlikely(skb->active_extensions); 4340 } 4341 #else 4342 static inline void skb_ext_put(struct sk_buff *skb) {} 4343 static inline void skb_ext_reset(struct sk_buff *skb) {} 4344 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4345 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4346 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4347 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4348 #endif /* CONFIG_SKB_EXTENSIONS */ 4349 4350 static inline void nf_reset_ct(struct sk_buff *skb) 4351 { 4352 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4353 nf_conntrack_put(skb_nfct(skb)); 4354 skb->_nfct = 0; 4355 #endif 4356 } 4357 4358 static inline void nf_reset_trace(struct sk_buff *skb) 4359 { 4360 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4361 skb->nf_trace = 0; 4362 #endif 4363 } 4364 4365 static inline void ipvs_reset(struct sk_buff *skb) 4366 { 4367 #if IS_ENABLED(CONFIG_IP_VS) 4368 skb->ipvs_property = 0; 4369 #endif 4370 } 4371 4372 /* Note: This doesn't put any conntrack info in dst. */ 4373 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4374 bool copy) 4375 { 4376 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4377 dst->_nfct = src->_nfct; 4378 nf_conntrack_get(skb_nfct(src)); 4379 #endif 4380 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4381 if (copy) 4382 dst->nf_trace = src->nf_trace; 4383 #endif 4384 } 4385 4386 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4387 { 4388 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4389 nf_conntrack_put(skb_nfct(dst)); 4390 #endif 4391 dst->slow_gro = src->slow_gro; 4392 __nf_copy(dst, src, true); 4393 } 4394 4395 #ifdef CONFIG_NETWORK_SECMARK 4396 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4397 { 4398 to->secmark = from->secmark; 4399 } 4400 4401 static inline void skb_init_secmark(struct sk_buff *skb) 4402 { 4403 skb->secmark = 0; 4404 } 4405 #else 4406 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4407 { } 4408 4409 static inline void skb_init_secmark(struct sk_buff *skb) 4410 { } 4411 #endif 4412 4413 static inline int secpath_exists(const struct sk_buff *skb) 4414 { 4415 #ifdef CONFIG_XFRM 4416 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4417 #else 4418 return 0; 4419 #endif 4420 } 4421 4422 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4423 { 4424 return !skb->destructor && 4425 !secpath_exists(skb) && 4426 !skb_nfct(skb) && 4427 !skb->_skb_refdst && 4428 !skb_has_frag_list(skb); 4429 } 4430 4431 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4432 { 4433 skb->queue_mapping = queue_mapping; 4434 } 4435 4436 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4437 { 4438 return skb->queue_mapping; 4439 } 4440 4441 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4442 { 4443 to->queue_mapping = from->queue_mapping; 4444 } 4445 4446 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4447 { 4448 skb->queue_mapping = rx_queue + 1; 4449 } 4450 4451 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4452 { 4453 return skb->queue_mapping - 1; 4454 } 4455 4456 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4457 { 4458 return skb->queue_mapping != 0; 4459 } 4460 4461 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4462 { 4463 skb->dst_pending_confirm = val; 4464 } 4465 4466 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4467 { 4468 return skb->dst_pending_confirm != 0; 4469 } 4470 4471 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4472 { 4473 #ifdef CONFIG_XFRM 4474 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4475 #else 4476 return NULL; 4477 #endif 4478 } 4479 4480 /* Keeps track of mac header offset relative to skb->head. 4481 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4482 * For non-tunnel skb it points to skb_mac_header() and for 4483 * tunnel skb it points to outer mac header. 4484 * Keeps track of level of encapsulation of network headers. 4485 */ 4486 struct skb_gso_cb { 4487 union { 4488 int mac_offset; 4489 int data_offset; 4490 }; 4491 int encap_level; 4492 __wsum csum; 4493 __u16 csum_start; 4494 }; 4495 #define SKB_GSO_CB_OFFSET 32 4496 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4497 4498 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4499 { 4500 return (skb_mac_header(inner_skb) - inner_skb->head) - 4501 SKB_GSO_CB(inner_skb)->mac_offset; 4502 } 4503 4504 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4505 { 4506 int new_headroom, headroom; 4507 int ret; 4508 4509 headroom = skb_headroom(skb); 4510 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4511 if (ret) 4512 return ret; 4513 4514 new_headroom = skb_headroom(skb); 4515 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4516 return 0; 4517 } 4518 4519 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4520 { 4521 /* Do not update partial checksums if remote checksum is enabled. */ 4522 if (skb->remcsum_offload) 4523 return; 4524 4525 SKB_GSO_CB(skb)->csum = res; 4526 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4527 } 4528 4529 /* Compute the checksum for a gso segment. First compute the checksum value 4530 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4531 * then add in skb->csum (checksum from csum_start to end of packet). 4532 * skb->csum and csum_start are then updated to reflect the checksum of the 4533 * resultant packet starting from the transport header-- the resultant checksum 4534 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4535 * header. 4536 */ 4537 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4538 { 4539 unsigned char *csum_start = skb_transport_header(skb); 4540 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4541 __wsum partial = SKB_GSO_CB(skb)->csum; 4542 4543 SKB_GSO_CB(skb)->csum = res; 4544 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4545 4546 return csum_fold(csum_partial(csum_start, plen, partial)); 4547 } 4548 4549 static inline bool skb_is_gso(const struct sk_buff *skb) 4550 { 4551 return skb_shinfo(skb)->gso_size; 4552 } 4553 4554 /* Note: Should be called only if skb_is_gso(skb) is true */ 4555 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4556 { 4557 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4558 } 4559 4560 /* Note: Should be called only if skb_is_gso(skb) is true */ 4561 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4562 { 4563 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4564 } 4565 4566 /* Note: Should be called only if skb_is_gso(skb) is true */ 4567 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4568 { 4569 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4570 } 4571 4572 static inline void skb_gso_reset(struct sk_buff *skb) 4573 { 4574 skb_shinfo(skb)->gso_size = 0; 4575 skb_shinfo(skb)->gso_segs = 0; 4576 skb_shinfo(skb)->gso_type = 0; 4577 } 4578 4579 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4580 u16 increment) 4581 { 4582 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4583 return; 4584 shinfo->gso_size += increment; 4585 } 4586 4587 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4588 u16 decrement) 4589 { 4590 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4591 return; 4592 shinfo->gso_size -= decrement; 4593 } 4594 4595 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4596 4597 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4598 { 4599 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4600 * wanted then gso_type will be set. */ 4601 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4602 4603 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4604 unlikely(shinfo->gso_type == 0)) { 4605 __skb_warn_lro_forwarding(skb); 4606 return true; 4607 } 4608 return false; 4609 } 4610 4611 static inline void skb_forward_csum(struct sk_buff *skb) 4612 { 4613 /* Unfortunately we don't support this one. Any brave souls? */ 4614 if (skb->ip_summed == CHECKSUM_COMPLETE) 4615 skb->ip_summed = CHECKSUM_NONE; 4616 } 4617 4618 /** 4619 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4620 * @skb: skb to check 4621 * 4622 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4623 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4624 * use this helper, to document places where we make this assertion. 4625 */ 4626 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4627 { 4628 #ifdef DEBUG 4629 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4630 #endif 4631 } 4632 4633 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4634 4635 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4636 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4637 unsigned int transport_len, 4638 __sum16(*skb_chkf)(struct sk_buff *skb)); 4639 4640 /** 4641 * skb_head_is_locked - Determine if the skb->head is locked down 4642 * @skb: skb to check 4643 * 4644 * The head on skbs build around a head frag can be removed if they are 4645 * not cloned. This function returns true if the skb head is locked down 4646 * due to either being allocated via kmalloc, or by being a clone with 4647 * multiple references to the head. 4648 */ 4649 static inline bool skb_head_is_locked(const struct sk_buff *skb) 4650 { 4651 return !skb->head_frag || skb_cloned(skb); 4652 } 4653 4654 /* Local Checksum Offload. 4655 * Compute outer checksum based on the assumption that the 4656 * inner checksum will be offloaded later. 4657 * See Documentation/networking/checksum-offloads.rst for 4658 * explanation of how this works. 4659 * Fill in outer checksum adjustment (e.g. with sum of outer 4660 * pseudo-header) before calling. 4661 * Also ensure that inner checksum is in linear data area. 4662 */ 4663 static inline __wsum lco_csum(struct sk_buff *skb) 4664 { 4665 unsigned char *csum_start = skb_checksum_start(skb); 4666 unsigned char *l4_hdr = skb_transport_header(skb); 4667 __wsum partial; 4668 4669 /* Start with complement of inner checksum adjustment */ 4670 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4671 skb->csum_offset)); 4672 4673 /* Add in checksum of our headers (incl. outer checksum 4674 * adjustment filled in by caller) and return result. 4675 */ 4676 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4677 } 4678 4679 static inline bool skb_is_redirected(const struct sk_buff *skb) 4680 { 4681 return skb->redirected; 4682 } 4683 4684 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4685 { 4686 skb->redirected = 1; 4687 #ifdef CONFIG_NET_REDIRECT 4688 skb->from_ingress = from_ingress; 4689 if (skb->from_ingress) 4690 skb->tstamp = 0; 4691 #endif 4692 } 4693 4694 static inline void skb_reset_redirect(struct sk_buff *skb) 4695 { 4696 skb->redirected = 0; 4697 } 4698 4699 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 4700 { 4701 return skb->csum_not_inet; 4702 } 4703 4704 static inline void skb_set_kcov_handle(struct sk_buff *skb, 4705 const u64 kcov_handle) 4706 { 4707 #ifdef CONFIG_KCOV 4708 skb->kcov_handle = kcov_handle; 4709 #endif 4710 } 4711 4712 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4713 { 4714 #ifdef CONFIG_KCOV 4715 return skb->kcov_handle; 4716 #else 4717 return 0; 4718 #endif 4719 } 4720 4721 #ifdef CONFIG_PAGE_POOL 4722 static inline void skb_mark_for_recycle(struct sk_buff *skb) 4723 { 4724 skb->pp_recycle = 1; 4725 } 4726 #endif 4727 4728 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data) 4729 { 4730 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle) 4731 return false; 4732 return page_pool_return_skb_page(virt_to_page(data)); 4733 } 4734 4735 #endif /* __KERNEL__ */ 4736 #endif /* _LINUX_SKBUFF_H */ 4737