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