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