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