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, <gw4pts@gw4pts.ampr.org> 7 * Florian La Roche, <rzsfl@rz.uni-sb.de> 8 */ 9 10 #ifndef _LINUX_SKBUFF_H 11 #define _LINUX_SKBUFF_H 12 13 #include <linux/kernel.h> 14 #include <linux/compiler.h> 15 #include <linux/time.h> 16 #include <linux/bug.h> 17 #include <linux/bvec.h> 18 #include <linux/cache.h> 19 #include <linux/rbtree.h> 20 #include <linux/socket.h> 21 #include <linux/refcount.h> 22 23 #include <linux/atomic.h> 24 #include <asm/types.h> 25 #include <linux/spinlock.h> 26 #include <linux/net.h> 27 #include <linux/textsearch.h> 28 #include <net/checksum.h> 29 #include <linux/rcupdate.h> 30 #include <linux/hrtimer.h> 31 #include <linux/dma-mapping.h> 32 #include <linux/netdev_features.h> 33 #include <linux/sched.h> 34 #include <linux/sched/clock.h> 35 #include <net/flow_dissector.h> 36 #include <linux/splice.h> 37 #include <linux/in6.h> 38 #include <linux/if_packet.h> 39 #include <net/flow.h> 40 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 41 #include <linux/netfilter/nf_conntrack_common.h> 42 #endif 43 44 /* The interface for checksum offload between the stack and networking drivers 45 * is as follows... 46 * 47 * A. IP checksum related features 48 * 49 * Drivers advertise checksum offload capabilities in the features of a device. 50 * From the stack's point of view these are capabilities offered by the driver. 51 * A driver typically only advertises features that it is capable of offloading 52 * to its device. 53 * 54 * The checksum related features are: 55 * 56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one 57 * IP (one's complement) checksum for any combination 58 * of protocols or protocol layering. The checksum is 59 * computed and set in a packet per the CHECKSUM_PARTIAL 60 * interface (see below). 61 * 62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain 63 * TCP or UDP packets over IPv4. These are specifically 64 * unencapsulated packets of the form IPv4|TCP or 65 * IPv4|UDP where the Protocol field in the IPv4 header 66 * is TCP or UDP. The IPv4 header may contain IP options. 67 * This feature cannot be set in features for a device 68 * with NETIF_F_HW_CSUM also set. This feature is being 69 * DEPRECATED (see below). 70 * 71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain 72 * TCP or UDP packets over IPv6. These are specifically 73 * unencapsulated packets of the form IPv6|TCP or 74 * IPv6|UDP where the Next Header field in the IPv6 75 * header is either TCP or UDP. IPv6 extension headers 76 * are not supported with this feature. This feature 77 * cannot be set in features for a device with 78 * NETIF_F_HW_CSUM also set. This feature is being 79 * DEPRECATED (see below). 80 * 81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. 82 * This flag is only used to disable the RX checksum 83 * feature for a device. The stack will accept receive 84 * checksum indication in packets received on a device 85 * regardless of whether NETIF_F_RXCSUM is set. 86 * 87 * B. Checksumming of received packets by device. Indication of checksum 88 * verification is set in skb->ip_summed. Possible values are: 89 * 90 * CHECKSUM_NONE: 91 * 92 * Device did not checksum this packet e.g. due to lack of capabilities. 93 * The packet contains full (though not verified) checksum in packet but 94 * not in skb->csum. Thus, skb->csum is undefined in this case. 95 * 96 * CHECKSUM_UNNECESSARY: 97 * 98 * The hardware you're dealing with doesn't calculate the full checksum 99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 101 * if their checksums are okay. skb->csum is still undefined in this case 102 * though. A driver or device must never modify the checksum field in the 103 * packet even if checksum is verified. 104 * 105 * CHECKSUM_UNNECESSARY is applicable to following protocols: 106 * TCP: IPv6 and IPv4. 107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 108 * zero UDP checksum for either IPv4 or IPv6, the networking stack 109 * may perform further validation in this case. 110 * GRE: only if the checksum is present in the header. 111 * SCTP: indicates the CRC in SCTP header has been validated. 112 * FCOE: indicates the CRC in FC frame has been validated. 113 * 114 * skb->csum_level indicates the number of consecutive checksums found in 115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 117 * and a device is able to verify the checksums for UDP (possibly zero), 118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to 119 * two. If the device were only able to verify the UDP checksum and not 120 * GRE, either because it doesn't support GRE checksum or because GRE 121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 122 * not considered in this case). 123 * 124 * CHECKSUM_COMPLETE: 125 * 126 * This is the most generic way. The device supplied checksum of the _whole_ 127 * packet as seen by netif_rx() and fills in skb->csum. This means the 128 * hardware doesn't need to parse L3/L4 headers to implement this. 129 * 130 * Notes: 131 * - Even if device supports only some protocols, but is able to produce 132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 134 * 135 * CHECKSUM_PARTIAL: 136 * 137 * A checksum is set up to be offloaded to a device as described in the 138 * output description for CHECKSUM_PARTIAL. This may occur on a packet 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel 140 * on the same host, or it may be set in the input path in GRO or remote 141 * checksum offload. For the purposes of checksum verification, the checksum 142 * referred to by skb->csum_start + skb->csum_offset and any preceding 143 * checksums in the packet are considered verified. Any checksums in the 144 * packet that are after the checksum being offloaded are not considered to 145 * be verified. 146 * 147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload 148 * in the skb->ip_summed for a packet. Values are: 149 * 150 * CHECKSUM_PARTIAL: 151 * 152 * The driver is required to checksum the packet as seen by hard_start_xmit() 153 * from skb->csum_start up to the end, and to record/write the checksum at 154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the 155 * csum_start and csum_offset values are valid values given the length and 156 * offset of the packet, but it should not attempt to validate that the 157 * checksum refers to a legitimate transport layer checksum -- it is the 158 * purview of the stack to validate that csum_start and csum_offset are set 159 * correctly. 160 * 161 * When the stack requests checksum offload for a packet, the driver MUST 162 * ensure that the checksum is set correctly. A driver can either offload the 163 * checksum calculation to the device, or call skb_checksum_help (in the case 164 * that the device does not support offload for a particular checksum). 165 * 166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of 167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate 168 * checksum offload capability. 169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based 170 * on network device checksumming capabilities: if a packet does not match 171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of 172 * csum_not_inet, see item D.) is called to resolve the checksum. 173 * 174 * CHECKSUM_NONE: 175 * 176 * The skb was already checksummed by the protocol, or a checksum is not 177 * required. 178 * 179 * CHECKSUM_UNNECESSARY: 180 * 181 * This has the same meaning as CHECKSUM_NONE for checksum offload on 182 * output. 183 * 184 * CHECKSUM_COMPLETE: 185 * Not used in checksum output. If a driver observes a packet with this value 186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set. 187 * 188 * D. Non-IP checksum (CRC) offloads 189 * 190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of 191 * offloading the SCTP CRC in a packet. To perform this offload the stack 192 * will set csum_start and csum_offset accordingly, set ip_summed to 193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in 194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c. 195 * A driver that supports both IP checksum offload and SCTP CRC32c offload 196 * must verify which offload is configured for a packet by testing the 197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve 198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 199 * 200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of 201 * offloading the FCOE CRC in a packet. To perform this offload the stack 202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset 203 * accordingly. Note that there is no indication in the skbuff that the 204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 205 * both IP checksum offload and FCOE CRC offload must verify which offload 206 * is configured for a packet, presumably by inspecting packet headers. 207 * 208 * E. Checksumming on output with GSO. 209 * 210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload 211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as 213 * part of the GSO operation is implied. If a checksum is being offloaded 214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and 215 * csum_offset are set to refer to the outermost checksum being offloaded 216 * (two offloaded checksums are possible with UDP encapsulation). 217 */ 218 219 /* Don't change this without changing skb_csum_unnecessary! */ 220 #define CHECKSUM_NONE 0 221 #define CHECKSUM_UNNECESSARY 1 222 #define CHECKSUM_COMPLETE 2 223 #define CHECKSUM_PARTIAL 3 224 225 /* Maximum value in skb->csum_level */ 226 #define SKB_MAX_CSUM_LEVEL 3 227 228 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 229 #define SKB_WITH_OVERHEAD(X) \ 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 231 #define SKB_MAX_ORDER(X, ORDER) \ 232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 233 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 234 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 235 236 /* return minimum truesize of one skb containing X bytes of data */ 237 #define SKB_TRUESIZE(X) ((X) + \ 238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 240 241 struct ahash_request; 242 struct net_device; 243 struct scatterlist; 244 struct pipe_inode_info; 245 struct iov_iter; 246 struct napi_struct; 247 struct bpf_prog; 248 union bpf_attr; 249 struct skb_ext; 250 251 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 252 struct nf_bridge_info { 253 enum { 254 BRNF_PROTO_UNCHANGED, 255 BRNF_PROTO_8021Q, 256 BRNF_PROTO_PPPOE 257 } orig_proto:8; 258 u8 pkt_otherhost:1; 259 u8 in_prerouting:1; 260 u8 bridged_dnat:1; 261 __u16 frag_max_size; 262 struct net_device *physindev; 263 264 /* always valid & non-NULL from FORWARD on, for physdev match */ 265 struct net_device *physoutdev; 266 union { 267 /* prerouting: detect dnat in orig/reply direction */ 268 __be32 ipv4_daddr; 269 struct in6_addr ipv6_daddr; 270 271 /* after prerouting + nat detected: store original source 272 * mac since neigh resolution overwrites it, only used while 273 * skb is out in neigh layer. 274 */ 275 char neigh_header[8]; 276 }; 277 }; 278 #endif 279 280 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 281 /* Chain in tc_skb_ext will be used to share the tc chain with 282 * ovs recirc_id. It will be set to the current chain by tc 283 * and read by ovs to recirc_id. 284 */ 285 struct tc_skb_ext { 286 __u32 chain; 287 __u16 mru; 288 bool post_ct; 289 }; 290 #endif 291 292 struct sk_buff_head { 293 /* These two members must be first. */ 294 struct sk_buff *next; 295 struct sk_buff *prev; 296 297 __u32 qlen; 298 spinlock_t lock; 299 }; 300 301 struct sk_buff; 302 303 /* To allow 64K frame to be packed as single skb without frag_list we 304 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 305 * buffers which do not start on a page boundary. 306 * 307 * Since GRO uses frags we allocate at least 16 regardless of page 308 * size. 309 */ 310 #if (65536/PAGE_SIZE + 1) < 16 311 #define MAX_SKB_FRAGS 16UL 312 #else 313 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 314 #endif 315 extern int sysctl_max_skb_frags; 316 317 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 318 * segment using its current segmentation instead. 319 */ 320 #define GSO_BY_FRAGS 0xFFFF 321 322 typedef struct bio_vec skb_frag_t; 323 324 /** 325 * skb_frag_size() - Returns the size of a skb fragment 326 * @frag: skb fragment 327 */ 328 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 329 { 330 return frag->bv_len; 331 } 332 333 /** 334 * skb_frag_size_set() - Sets the size of a skb fragment 335 * @frag: skb fragment 336 * @size: size of fragment 337 */ 338 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 339 { 340 frag->bv_len = size; 341 } 342 343 /** 344 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 345 * @frag: skb fragment 346 * @delta: value to add 347 */ 348 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 349 { 350 frag->bv_len += delta; 351 } 352 353 /** 354 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 355 * @frag: skb fragment 356 * @delta: value to subtract 357 */ 358 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 359 { 360 frag->bv_len -= delta; 361 } 362 363 /** 364 * skb_frag_must_loop - Test if %p is a high memory page 365 * @p: fragment's page 366 */ 367 static inline bool skb_frag_must_loop(struct page *p) 368 { 369 #if defined(CONFIG_HIGHMEM) 370 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 371 return true; 372 #endif 373 return false; 374 } 375 376 /** 377 * skb_frag_foreach_page - loop over pages in a fragment 378 * 379 * @f: skb frag to operate on 380 * @f_off: offset from start of f->bv_page 381 * @f_len: length from f_off to loop over 382 * @p: (temp var) current page 383 * @p_off: (temp var) offset from start of current page, 384 * non-zero only on first page. 385 * @p_len: (temp var) length in current page, 386 * < PAGE_SIZE only on first and last page. 387 * @copied: (temp var) length so far, excluding current p_len. 388 * 389 * A fragment can hold a compound page, in which case per-page 390 * operations, notably kmap_atomic, must be called for each 391 * regular page. 392 */ 393 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 394 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 395 p_off = (f_off) & (PAGE_SIZE - 1), \ 396 p_len = skb_frag_must_loop(p) ? \ 397 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 398 copied = 0; \ 399 copied < f_len; \ 400 copied += p_len, p++, p_off = 0, \ 401 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 402 403 #define HAVE_HW_TIME_STAMP 404 405 /** 406 * struct skb_shared_hwtstamps - hardware time stamps 407 * @hwtstamp: hardware time stamp transformed into duration 408 * since arbitrary point in time 409 * 410 * Software time stamps generated by ktime_get_real() are stored in 411 * skb->tstamp. 412 * 413 * hwtstamps can only be compared against other hwtstamps from 414 * the same device. 415 * 416 * This structure is attached to packets as part of the 417 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 418 */ 419 struct skb_shared_hwtstamps { 420 ktime_t hwtstamp; 421 }; 422 423 /* Definitions for tx_flags in struct skb_shared_info */ 424 enum { 425 /* generate hardware time stamp */ 426 SKBTX_HW_TSTAMP = 1 << 0, 427 428 /* generate software time stamp when queueing packet to NIC */ 429 SKBTX_SW_TSTAMP = 1 << 1, 430 431 /* device driver is going to provide hardware time stamp */ 432 SKBTX_IN_PROGRESS = 1 << 2, 433 434 /* generate wifi status information (where possible) */ 435 SKBTX_WIFI_STATUS = 1 << 4, 436 437 /* generate software time stamp when entering packet scheduling */ 438 SKBTX_SCHED_TSTAMP = 1 << 6, 439 }; 440 441 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 442 SKBTX_SCHED_TSTAMP) 443 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 444 445 /* Definitions for flags in struct skb_shared_info */ 446 enum { 447 /* use zcopy routines */ 448 SKBFL_ZEROCOPY_ENABLE = BIT(0), 449 450 /* This indicates at least one fragment might be overwritten 451 * (as in vmsplice(), sendfile() ...) 452 * If we need to compute a TX checksum, we'll need to copy 453 * all frags to avoid possible bad checksum 454 */ 455 SKBFL_SHARED_FRAG = BIT(1), 456 }; 457 458 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 459 460 /* 461 * The callback notifies userspace to release buffers when skb DMA is done in 462 * lower device, the skb last reference should be 0 when calling this. 463 * The zerocopy_success argument is true if zero copy transmit occurred, 464 * false on data copy or out of memory error caused by data copy attempt. 465 * The ctx field is used to track device context. 466 * The desc field is used to track userspace buffer index. 467 */ 468 struct ubuf_info { 469 void (*callback)(struct sk_buff *, struct ubuf_info *, 470 bool zerocopy_success); 471 union { 472 struct { 473 unsigned long desc; 474 void *ctx; 475 }; 476 struct { 477 u32 id; 478 u16 len; 479 u16 zerocopy:1; 480 u32 bytelen; 481 }; 482 }; 483 refcount_t refcnt; 484 u8 flags; 485 486 struct mmpin { 487 struct user_struct *user; 488 unsigned int num_pg; 489 } mmp; 490 }; 491 492 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 493 494 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 495 void mm_unaccount_pinned_pages(struct mmpin *mmp); 496 497 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size); 498 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 499 struct ubuf_info *uarg); 500 501 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 502 503 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 504 bool success); 505 506 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len); 507 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 508 struct msghdr *msg, int len, 509 struct ubuf_info *uarg); 510 511 /* This data is invariant across clones and lives at 512 * the end of the header data, ie. at skb->end. 513 */ 514 struct skb_shared_info { 515 __u8 flags; 516 __u8 meta_len; 517 __u8 nr_frags; 518 __u8 tx_flags; 519 unsigned short gso_size; 520 /* Warning: this field is not always filled in (UFO)! */ 521 unsigned short gso_segs; 522 struct sk_buff *frag_list; 523 struct skb_shared_hwtstamps hwtstamps; 524 unsigned int gso_type; 525 u32 tskey; 526 527 /* 528 * Warning : all fields before dataref are cleared in __alloc_skb() 529 */ 530 atomic_t dataref; 531 532 /* Intermediate layers must ensure that destructor_arg 533 * remains valid until skb destructor */ 534 void * destructor_arg; 535 536 /* must be last field, see pskb_expand_head() */ 537 skb_frag_t frags[MAX_SKB_FRAGS]; 538 }; 539 540 /* We divide dataref into two halves. The higher 16 bits hold references 541 * to the payload part of skb->data. The lower 16 bits hold references to 542 * the entire skb->data. A clone of a headerless skb holds the length of 543 * the header in skb->hdr_len. 544 * 545 * All users must obey the rule that the skb->data reference count must be 546 * greater than or equal to the payload reference count. 547 * 548 * Holding a reference to the payload part means that the user does not 549 * care about modifications to the header part of skb->data. 550 */ 551 #define SKB_DATAREF_SHIFT 16 552 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 553 554 555 enum { 556 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 557 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 558 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 559 }; 560 561 enum { 562 SKB_GSO_TCPV4 = 1 << 0, 563 564 /* This indicates the skb is from an untrusted source. */ 565 SKB_GSO_DODGY = 1 << 1, 566 567 /* This indicates the tcp segment has CWR set. */ 568 SKB_GSO_TCP_ECN = 1 << 2, 569 570 SKB_GSO_TCP_FIXEDID = 1 << 3, 571 572 SKB_GSO_TCPV6 = 1 << 4, 573 574 SKB_GSO_FCOE = 1 << 5, 575 576 SKB_GSO_GRE = 1 << 6, 577 578 SKB_GSO_GRE_CSUM = 1 << 7, 579 580 SKB_GSO_IPXIP4 = 1 << 8, 581 582 SKB_GSO_IPXIP6 = 1 << 9, 583 584 SKB_GSO_UDP_TUNNEL = 1 << 10, 585 586 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 587 588 SKB_GSO_PARTIAL = 1 << 12, 589 590 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 591 592 SKB_GSO_SCTP = 1 << 14, 593 594 SKB_GSO_ESP = 1 << 15, 595 596 SKB_GSO_UDP = 1 << 16, 597 598 SKB_GSO_UDP_L4 = 1 << 17, 599 600 SKB_GSO_FRAGLIST = 1 << 18, 601 }; 602 603 #if BITS_PER_LONG > 32 604 #define NET_SKBUFF_DATA_USES_OFFSET 1 605 #endif 606 607 #ifdef NET_SKBUFF_DATA_USES_OFFSET 608 typedef unsigned int sk_buff_data_t; 609 #else 610 typedef unsigned char *sk_buff_data_t; 611 #endif 612 613 /** 614 * struct sk_buff - socket buffer 615 * @next: Next buffer in list 616 * @prev: Previous buffer in list 617 * @tstamp: Time we arrived/left 618 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 619 * for retransmit timer 620 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 621 * @list: queue head 622 * @sk: Socket we are owned by 623 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 624 * fragmentation management 625 * @dev: Device we arrived on/are leaving by 626 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 627 * @cb: Control buffer. Free for use by every layer. Put private vars here 628 * @_skb_refdst: destination entry (with norefcount bit) 629 * @sp: the security path, used for xfrm 630 * @len: Length of actual data 631 * @data_len: Data length 632 * @mac_len: Length of link layer header 633 * @hdr_len: writable header length of cloned skb 634 * @csum: Checksum (must include start/offset pair) 635 * @csum_start: Offset from skb->head where checksumming should start 636 * @csum_offset: Offset from csum_start where checksum should be stored 637 * @priority: Packet queueing priority 638 * @ignore_df: allow local fragmentation 639 * @cloned: Head may be cloned (check refcnt to be sure) 640 * @ip_summed: Driver fed us an IP checksum 641 * @nohdr: Payload reference only, must not modify header 642 * @pkt_type: Packet class 643 * @fclone: skbuff clone status 644 * @ipvs_property: skbuff is owned by ipvs 645 * @inner_protocol_type: whether the inner protocol is 646 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 647 * @remcsum_offload: remote checksum offload is enabled 648 * @offload_fwd_mark: Packet was L2-forwarded in hardware 649 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 650 * @tc_skip_classify: do not classify packet. set by IFB device 651 * @tc_at_ingress: used within tc_classify to distinguish in/egress 652 * @redirected: packet was redirected by packet classifier 653 * @from_ingress: packet was redirected from the ingress path 654 * @peeked: this packet has been seen already, so stats have been 655 * done for it, don't do them again 656 * @nf_trace: netfilter packet trace flag 657 * @protocol: Packet protocol from driver 658 * @destructor: Destruct function 659 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 660 * @_nfct: Associated connection, if any (with nfctinfo bits) 661 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 662 * @skb_iif: ifindex of device we arrived on 663 * @tc_index: Traffic control index 664 * @hash: the packet hash 665 * @queue_mapping: Queue mapping for multiqueue devices 666 * @head_frag: skb was allocated from page fragments, 667 * not allocated by kmalloc() or vmalloc(). 668 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 669 * @active_extensions: active extensions (skb_ext_id types) 670 * @ndisc_nodetype: router type (from link layer) 671 * @ooo_okay: allow the mapping of a socket to a queue to be changed 672 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 673 * ports. 674 * @sw_hash: indicates hash was computed in software stack 675 * @wifi_acked_valid: wifi_acked was set 676 * @wifi_acked: whether frame was acked on wifi or not 677 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 678 * @encapsulation: indicates the inner headers in the skbuff are valid 679 * @encap_hdr_csum: software checksum is needed 680 * @csum_valid: checksum is already valid 681 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 682 * @csum_complete_sw: checksum was completed by software 683 * @csum_level: indicates the number of consecutive checksums found in 684 * the packet minus one that have been verified as 685 * CHECKSUM_UNNECESSARY (max 3) 686 * @dst_pending_confirm: need to confirm neighbour 687 * @decrypted: Decrypted SKB 688 * @napi_id: id of the NAPI struct this skb came from 689 * @sender_cpu: (aka @napi_id) source CPU in XPS 690 * @secmark: security marking 691 * @mark: Generic packet mark 692 * @reserved_tailroom: (aka @mark) number of bytes of free space available 693 * at the tail of an sk_buff 694 * @vlan_present: VLAN tag is present 695 * @vlan_proto: vlan encapsulation protocol 696 * @vlan_tci: vlan tag control information 697 * @inner_protocol: Protocol (encapsulation) 698 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 699 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 700 * @inner_transport_header: Inner transport layer header (encapsulation) 701 * @inner_network_header: Network layer header (encapsulation) 702 * @inner_mac_header: Link layer header (encapsulation) 703 * @transport_header: Transport layer header 704 * @network_header: Network layer header 705 * @mac_header: Link layer header 706 * @kcov_handle: KCOV remote handle for remote coverage collection 707 * @tail: Tail pointer 708 * @end: End pointer 709 * @head: Head of buffer 710 * @data: Data head pointer 711 * @truesize: Buffer size 712 * @users: User count - see {datagram,tcp}.c 713 * @extensions: allocated extensions, valid if active_extensions is nonzero 714 */ 715 716 struct sk_buff { 717 union { 718 struct { 719 /* These two members must be first. */ 720 struct sk_buff *next; 721 struct sk_buff *prev; 722 723 union { 724 struct net_device *dev; 725 /* Some protocols might use this space to store information, 726 * while device pointer would be NULL. 727 * UDP receive path is one user. 728 */ 729 unsigned long dev_scratch; 730 }; 731 }; 732 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 733 struct list_head list; 734 }; 735 736 union { 737 struct sock *sk; 738 int ip_defrag_offset; 739 }; 740 741 union { 742 ktime_t tstamp; 743 u64 skb_mstamp_ns; /* earliest departure time */ 744 }; 745 /* 746 * This is the control buffer. It is free to use for every 747 * layer. Please put your private variables there. If you 748 * want to keep them across layers you have to do a skb_clone() 749 * first. This is owned by whoever has the skb queued ATM. 750 */ 751 char cb[48] __aligned(8); 752 753 union { 754 struct { 755 unsigned long _skb_refdst; 756 void (*destructor)(struct sk_buff *skb); 757 }; 758 struct list_head tcp_tsorted_anchor; 759 }; 760 761 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 762 unsigned long _nfct; 763 #endif 764 unsigned int len, 765 data_len; 766 __u16 mac_len, 767 hdr_len; 768 769 /* Following fields are _not_ copied in __copy_skb_header() 770 * Note that queue_mapping is here mostly to fill a hole. 771 */ 772 __u16 queue_mapping; 773 774 /* if you move cloned around you also must adapt those constants */ 775 #ifdef __BIG_ENDIAN_BITFIELD 776 #define CLONED_MASK (1 << 7) 777 #else 778 #define CLONED_MASK 1 779 #endif 780 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset) 781 782 /* private: */ 783 __u8 __cloned_offset[0]; 784 /* public: */ 785 __u8 cloned:1, 786 nohdr:1, 787 fclone:2, 788 peeked:1, 789 head_frag:1, 790 pfmemalloc:1; 791 #ifdef CONFIG_SKB_EXTENSIONS 792 __u8 active_extensions; 793 #endif 794 /* fields enclosed in headers_start/headers_end are copied 795 * using a single memcpy() in __copy_skb_header() 796 */ 797 /* private: */ 798 __u32 headers_start[0]; 799 /* public: */ 800 801 /* if you move pkt_type around you also must adapt those constants */ 802 #ifdef __BIG_ENDIAN_BITFIELD 803 #define PKT_TYPE_MAX (7 << 5) 804 #else 805 #define PKT_TYPE_MAX 7 806 #endif 807 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 808 809 /* private: */ 810 __u8 __pkt_type_offset[0]; 811 /* public: */ 812 __u8 pkt_type:3; 813 __u8 ignore_df:1; 814 __u8 nf_trace:1; 815 __u8 ip_summed:2; 816 __u8 ooo_okay:1; 817 818 __u8 l4_hash:1; 819 __u8 sw_hash:1; 820 __u8 wifi_acked_valid:1; 821 __u8 wifi_acked:1; 822 __u8 no_fcs:1; 823 /* Indicates the inner headers are valid in the skbuff. */ 824 __u8 encapsulation:1; 825 __u8 encap_hdr_csum:1; 826 __u8 csum_valid:1; 827 828 #ifdef __BIG_ENDIAN_BITFIELD 829 #define PKT_VLAN_PRESENT_BIT 7 830 #else 831 #define PKT_VLAN_PRESENT_BIT 0 832 #endif 833 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset) 834 /* private: */ 835 __u8 __pkt_vlan_present_offset[0]; 836 /* public: */ 837 __u8 vlan_present:1; 838 __u8 csum_complete_sw:1; 839 __u8 csum_level:2; 840 __u8 csum_not_inet:1; 841 __u8 dst_pending_confirm:1; 842 #ifdef CONFIG_IPV6_NDISC_NODETYPE 843 __u8 ndisc_nodetype:2; 844 #endif 845 846 __u8 ipvs_property:1; 847 __u8 inner_protocol_type:1; 848 __u8 remcsum_offload:1; 849 #ifdef CONFIG_NET_SWITCHDEV 850 __u8 offload_fwd_mark:1; 851 __u8 offload_l3_fwd_mark:1; 852 #endif 853 #ifdef CONFIG_NET_CLS_ACT 854 __u8 tc_skip_classify:1; 855 __u8 tc_at_ingress:1; 856 #endif 857 #ifdef CONFIG_NET_REDIRECT 858 __u8 redirected:1; 859 __u8 from_ingress:1; 860 #endif 861 #ifdef CONFIG_TLS_DEVICE 862 __u8 decrypted:1; 863 #endif 864 865 #ifdef CONFIG_NET_SCHED 866 __u16 tc_index; /* traffic control index */ 867 #endif 868 869 union { 870 __wsum csum; 871 struct { 872 __u16 csum_start; 873 __u16 csum_offset; 874 }; 875 }; 876 __u32 priority; 877 int skb_iif; 878 __u32 hash; 879 __be16 vlan_proto; 880 __u16 vlan_tci; 881 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 882 union { 883 unsigned int napi_id; 884 unsigned int sender_cpu; 885 }; 886 #endif 887 #ifdef CONFIG_NETWORK_SECMARK 888 __u32 secmark; 889 #endif 890 891 union { 892 __u32 mark; 893 __u32 reserved_tailroom; 894 }; 895 896 union { 897 __be16 inner_protocol; 898 __u8 inner_ipproto; 899 }; 900 901 __u16 inner_transport_header; 902 __u16 inner_network_header; 903 __u16 inner_mac_header; 904 905 __be16 protocol; 906 __u16 transport_header; 907 __u16 network_header; 908 __u16 mac_header; 909 910 #ifdef CONFIG_KCOV 911 u64 kcov_handle; 912 #endif 913 914 /* private: */ 915 __u32 headers_end[0]; 916 /* public: */ 917 918 /* These elements must be at the end, see alloc_skb() for details. */ 919 sk_buff_data_t tail; 920 sk_buff_data_t end; 921 unsigned char *head, 922 *data; 923 unsigned int truesize; 924 refcount_t users; 925 926 #ifdef CONFIG_SKB_EXTENSIONS 927 /* only useable after checking ->active_extensions != 0 */ 928 struct skb_ext *extensions; 929 #endif 930 }; 931 932 #ifdef __KERNEL__ 933 /* 934 * Handling routines are only of interest to the kernel 935 */ 936 937 #define SKB_ALLOC_FCLONE 0x01 938 #define SKB_ALLOC_RX 0x02 939 #define SKB_ALLOC_NAPI 0x04 940 941 /** 942 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 943 * @skb: buffer 944 */ 945 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 946 { 947 return unlikely(skb->pfmemalloc); 948 } 949 950 /* 951 * skb might have a dst pointer attached, refcounted or not. 952 * _skb_refdst low order bit is set if refcount was _not_ taken 953 */ 954 #define SKB_DST_NOREF 1UL 955 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 956 957 /** 958 * skb_dst - returns skb dst_entry 959 * @skb: buffer 960 * 961 * Returns skb dst_entry, regardless of reference taken or not. 962 */ 963 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 964 { 965 /* If refdst was not refcounted, check we still are in a 966 * rcu_read_lock section 967 */ 968 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 969 !rcu_read_lock_held() && 970 !rcu_read_lock_bh_held()); 971 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 972 } 973 974 /** 975 * skb_dst_set - sets skb dst 976 * @skb: buffer 977 * @dst: dst entry 978 * 979 * Sets skb dst, assuming a reference was taken on dst and should 980 * be released by skb_dst_drop() 981 */ 982 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 983 { 984 skb->_skb_refdst = (unsigned long)dst; 985 } 986 987 /** 988 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 989 * @skb: buffer 990 * @dst: dst entry 991 * 992 * Sets skb dst, assuming a reference was not taken on dst. 993 * If dst entry is cached, we do not take reference and dst_release 994 * will be avoided by refdst_drop. If dst entry is not cached, we take 995 * reference, so that last dst_release can destroy the dst immediately. 996 */ 997 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 998 { 999 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1000 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1001 } 1002 1003 /** 1004 * skb_dst_is_noref - Test if skb dst isn't refcounted 1005 * @skb: buffer 1006 */ 1007 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1008 { 1009 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1010 } 1011 1012 /** 1013 * skb_rtable - Returns the skb &rtable 1014 * @skb: buffer 1015 */ 1016 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1017 { 1018 return (struct rtable *)skb_dst(skb); 1019 } 1020 1021 /* For mangling skb->pkt_type from user space side from applications 1022 * such as nft, tc, etc, we only allow a conservative subset of 1023 * possible pkt_types to be set. 1024 */ 1025 static inline bool skb_pkt_type_ok(u32 ptype) 1026 { 1027 return ptype <= PACKET_OTHERHOST; 1028 } 1029 1030 /** 1031 * skb_napi_id - Returns the skb's NAPI id 1032 * @skb: buffer 1033 */ 1034 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1035 { 1036 #ifdef CONFIG_NET_RX_BUSY_POLL 1037 return skb->napi_id; 1038 #else 1039 return 0; 1040 #endif 1041 } 1042 1043 /** 1044 * skb_unref - decrement the skb's reference count 1045 * @skb: buffer 1046 * 1047 * Returns true if we can free the skb. 1048 */ 1049 static inline bool skb_unref(struct sk_buff *skb) 1050 { 1051 if (unlikely(!skb)) 1052 return false; 1053 if (likely(refcount_read(&skb->users) == 1)) 1054 smp_rmb(); 1055 else if (likely(!refcount_dec_and_test(&skb->users))) 1056 return false; 1057 1058 return true; 1059 } 1060 1061 void skb_release_head_state(struct sk_buff *skb); 1062 void kfree_skb(struct sk_buff *skb); 1063 void kfree_skb_list(struct sk_buff *segs); 1064 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1065 void skb_tx_error(struct sk_buff *skb); 1066 1067 #ifdef CONFIG_TRACEPOINTS 1068 void consume_skb(struct sk_buff *skb); 1069 #else 1070 static inline void consume_skb(struct sk_buff *skb) 1071 { 1072 return kfree_skb(skb); 1073 } 1074 #endif 1075 1076 void __consume_stateless_skb(struct sk_buff *skb); 1077 void __kfree_skb(struct sk_buff *skb); 1078 extern struct kmem_cache *skbuff_head_cache; 1079 1080 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1081 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1082 bool *fragstolen, int *delta_truesize); 1083 1084 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1085 int node); 1086 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1087 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1088 struct sk_buff *build_skb_around(struct sk_buff *skb, 1089 void *data, unsigned int frag_size); 1090 1091 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1092 1093 /** 1094 * alloc_skb - allocate a network buffer 1095 * @size: size to allocate 1096 * @priority: allocation mask 1097 * 1098 * This function is a convenient wrapper around __alloc_skb(). 1099 */ 1100 static inline struct sk_buff *alloc_skb(unsigned int size, 1101 gfp_t priority) 1102 { 1103 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1104 } 1105 1106 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1107 unsigned long data_len, 1108 int max_page_order, 1109 int *errcode, 1110 gfp_t gfp_mask); 1111 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1112 1113 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1114 struct sk_buff_fclones { 1115 struct sk_buff skb1; 1116 1117 struct sk_buff skb2; 1118 1119 refcount_t fclone_ref; 1120 }; 1121 1122 /** 1123 * skb_fclone_busy - check if fclone is busy 1124 * @sk: socket 1125 * @skb: buffer 1126 * 1127 * Returns true if skb is a fast clone, and its clone is not freed. 1128 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1129 * so we also check that this didnt happen. 1130 */ 1131 static inline bool skb_fclone_busy(const struct sock *sk, 1132 const struct sk_buff *skb) 1133 { 1134 const struct sk_buff_fclones *fclones; 1135 1136 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1137 1138 return skb->fclone == SKB_FCLONE_ORIG && 1139 refcount_read(&fclones->fclone_ref) > 1 && 1140 fclones->skb2.sk == sk; 1141 } 1142 1143 /** 1144 * alloc_skb_fclone - allocate a network buffer from fclone cache 1145 * @size: size to allocate 1146 * @priority: allocation mask 1147 * 1148 * This function is a convenient wrapper around __alloc_skb(). 1149 */ 1150 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1151 gfp_t priority) 1152 { 1153 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1154 } 1155 1156 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1157 void skb_headers_offset_update(struct sk_buff *skb, int off); 1158 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1159 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1160 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1161 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1162 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1163 gfp_t gfp_mask, bool fclone); 1164 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1165 gfp_t gfp_mask) 1166 { 1167 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1168 } 1169 1170 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1171 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1172 unsigned int headroom); 1173 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1174 int newtailroom, gfp_t priority); 1175 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1176 int offset, int len); 1177 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1178 int offset, int len); 1179 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1180 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1181 1182 /** 1183 * skb_pad - zero pad the tail of an skb 1184 * @skb: buffer to pad 1185 * @pad: space to pad 1186 * 1187 * Ensure that a buffer is followed by a padding area that is zero 1188 * filled. Used by network drivers which may DMA or transfer data 1189 * beyond the buffer end onto the wire. 1190 * 1191 * May return error in out of memory cases. The skb is freed on error. 1192 */ 1193 static inline int skb_pad(struct sk_buff *skb, int pad) 1194 { 1195 return __skb_pad(skb, pad, true); 1196 } 1197 #define dev_kfree_skb(a) consume_skb(a) 1198 1199 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1200 int offset, size_t size); 1201 1202 struct skb_seq_state { 1203 __u32 lower_offset; 1204 __u32 upper_offset; 1205 __u32 frag_idx; 1206 __u32 stepped_offset; 1207 struct sk_buff *root_skb; 1208 struct sk_buff *cur_skb; 1209 __u8 *frag_data; 1210 __u32 frag_off; 1211 }; 1212 1213 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1214 unsigned int to, struct skb_seq_state *st); 1215 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1216 struct skb_seq_state *st); 1217 void skb_abort_seq_read(struct skb_seq_state *st); 1218 1219 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1220 unsigned int to, struct ts_config *config); 1221 1222 /* 1223 * Packet hash types specify the type of hash in skb_set_hash. 1224 * 1225 * Hash types refer to the protocol layer addresses which are used to 1226 * construct a packet's hash. The hashes are used to differentiate or identify 1227 * flows of the protocol layer for the hash type. Hash types are either 1228 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1229 * 1230 * Properties of hashes: 1231 * 1232 * 1) Two packets in different flows have different hash values 1233 * 2) Two packets in the same flow should have the same hash value 1234 * 1235 * A hash at a higher layer is considered to be more specific. A driver should 1236 * set the most specific hash possible. 1237 * 1238 * A driver cannot indicate a more specific hash than the layer at which a hash 1239 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1240 * 1241 * A driver may indicate a hash level which is less specific than the 1242 * actual layer the hash was computed on. For instance, a hash computed 1243 * at L4 may be considered an L3 hash. This should only be done if the 1244 * driver can't unambiguously determine that the HW computed the hash at 1245 * the higher layer. Note that the "should" in the second property above 1246 * permits this. 1247 */ 1248 enum pkt_hash_types { 1249 PKT_HASH_TYPE_NONE, /* Undefined type */ 1250 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1251 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1252 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1253 }; 1254 1255 static inline void skb_clear_hash(struct sk_buff *skb) 1256 { 1257 skb->hash = 0; 1258 skb->sw_hash = 0; 1259 skb->l4_hash = 0; 1260 } 1261 1262 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1263 { 1264 if (!skb->l4_hash) 1265 skb_clear_hash(skb); 1266 } 1267 1268 static inline void 1269 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1270 { 1271 skb->l4_hash = is_l4; 1272 skb->sw_hash = is_sw; 1273 skb->hash = hash; 1274 } 1275 1276 static inline void 1277 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1278 { 1279 /* Used by drivers to set hash from HW */ 1280 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1281 } 1282 1283 static inline void 1284 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1285 { 1286 __skb_set_hash(skb, hash, true, is_l4); 1287 } 1288 1289 void __skb_get_hash(struct sk_buff *skb); 1290 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1291 u32 skb_get_poff(const struct sk_buff *skb); 1292 u32 __skb_get_poff(const struct sk_buff *skb, void *data, 1293 const struct flow_keys_basic *keys, int hlen); 1294 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1295 void *data, int hlen_proto); 1296 1297 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1298 int thoff, u8 ip_proto) 1299 { 1300 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1301 } 1302 1303 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1304 const struct flow_dissector_key *key, 1305 unsigned int key_count); 1306 1307 struct bpf_flow_dissector; 1308 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1309 __be16 proto, int nhoff, int hlen, unsigned int flags); 1310 1311 bool __skb_flow_dissect(const struct net *net, 1312 const struct sk_buff *skb, 1313 struct flow_dissector *flow_dissector, 1314 void *target_container, 1315 void *data, __be16 proto, int nhoff, int hlen, 1316 unsigned int flags); 1317 1318 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1319 struct flow_dissector *flow_dissector, 1320 void *target_container, unsigned int flags) 1321 { 1322 return __skb_flow_dissect(NULL, skb, flow_dissector, 1323 target_container, NULL, 0, 0, 0, flags); 1324 } 1325 1326 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1327 struct flow_keys *flow, 1328 unsigned int flags) 1329 { 1330 memset(flow, 0, sizeof(*flow)); 1331 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1332 flow, NULL, 0, 0, 0, flags); 1333 } 1334 1335 static inline bool 1336 skb_flow_dissect_flow_keys_basic(const struct net *net, 1337 const struct sk_buff *skb, 1338 struct flow_keys_basic *flow, void *data, 1339 __be16 proto, int nhoff, int hlen, 1340 unsigned int flags) 1341 { 1342 memset(flow, 0, sizeof(*flow)); 1343 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1344 data, proto, nhoff, hlen, flags); 1345 } 1346 1347 void skb_flow_dissect_meta(const struct sk_buff *skb, 1348 struct flow_dissector *flow_dissector, 1349 void *target_container); 1350 1351 /* Gets a skb connection tracking info, ctinfo map should be a 1352 * map of mapsize to translate enum ip_conntrack_info states 1353 * to user states. 1354 */ 1355 void 1356 skb_flow_dissect_ct(const struct sk_buff *skb, 1357 struct flow_dissector *flow_dissector, 1358 void *target_container, 1359 u16 *ctinfo_map, size_t mapsize, 1360 bool post_ct); 1361 void 1362 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1363 struct flow_dissector *flow_dissector, 1364 void *target_container); 1365 1366 void skb_flow_dissect_hash(const struct sk_buff *skb, 1367 struct flow_dissector *flow_dissector, 1368 void *target_container); 1369 1370 static inline __u32 skb_get_hash(struct sk_buff *skb) 1371 { 1372 if (!skb->l4_hash && !skb->sw_hash) 1373 __skb_get_hash(skb); 1374 1375 return skb->hash; 1376 } 1377 1378 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1379 { 1380 if (!skb->l4_hash && !skb->sw_hash) { 1381 struct flow_keys keys; 1382 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1383 1384 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1385 } 1386 1387 return skb->hash; 1388 } 1389 1390 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1391 const siphash_key_t *perturb); 1392 1393 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1394 { 1395 return skb->hash; 1396 } 1397 1398 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1399 { 1400 to->hash = from->hash; 1401 to->sw_hash = from->sw_hash; 1402 to->l4_hash = from->l4_hash; 1403 }; 1404 1405 static inline void skb_copy_decrypted(struct sk_buff *to, 1406 const struct sk_buff *from) 1407 { 1408 #ifdef CONFIG_TLS_DEVICE 1409 to->decrypted = from->decrypted; 1410 #endif 1411 } 1412 1413 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1414 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1415 { 1416 return skb->head + skb->end; 1417 } 1418 1419 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1420 { 1421 return skb->end; 1422 } 1423 #else 1424 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1425 { 1426 return skb->end; 1427 } 1428 1429 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1430 { 1431 return skb->end - skb->head; 1432 } 1433 #endif 1434 1435 /* Internal */ 1436 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1437 1438 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1439 { 1440 return &skb_shinfo(skb)->hwtstamps; 1441 } 1442 1443 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1444 { 1445 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1446 1447 return is_zcopy ? skb_uarg(skb) : NULL; 1448 } 1449 1450 static inline void net_zcopy_get(struct ubuf_info *uarg) 1451 { 1452 refcount_inc(&uarg->refcnt); 1453 } 1454 1455 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1456 { 1457 skb_shinfo(skb)->destructor_arg = uarg; 1458 skb_shinfo(skb)->flags |= uarg->flags; 1459 } 1460 1461 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1462 bool *have_ref) 1463 { 1464 if (skb && uarg && !skb_zcopy(skb)) { 1465 if (unlikely(have_ref && *have_ref)) 1466 *have_ref = false; 1467 else 1468 net_zcopy_get(uarg); 1469 skb_zcopy_init(skb, uarg); 1470 } 1471 } 1472 1473 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1474 { 1475 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1476 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1477 } 1478 1479 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1480 { 1481 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1482 } 1483 1484 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1485 { 1486 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1487 } 1488 1489 static inline void net_zcopy_put(struct ubuf_info *uarg) 1490 { 1491 if (uarg) 1492 uarg->callback(NULL, uarg, true); 1493 } 1494 1495 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1496 { 1497 if (uarg) { 1498 if (uarg->callback == msg_zerocopy_callback) 1499 msg_zerocopy_put_abort(uarg, have_uref); 1500 else if (have_uref) 1501 net_zcopy_put(uarg); 1502 } 1503 } 1504 1505 /* Release a reference on a zerocopy structure */ 1506 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1507 { 1508 struct ubuf_info *uarg = skb_zcopy(skb); 1509 1510 if (uarg) { 1511 if (!skb_zcopy_is_nouarg(skb)) 1512 uarg->callback(skb, uarg, zerocopy_success); 1513 1514 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG; 1515 } 1516 } 1517 1518 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1519 { 1520 skb->next = NULL; 1521 } 1522 1523 /* Iterate through singly-linked GSO fragments of an skb. */ 1524 #define skb_list_walk_safe(first, skb, next_skb) \ 1525 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1526 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1527 1528 static inline void skb_list_del_init(struct sk_buff *skb) 1529 { 1530 __list_del_entry(&skb->list); 1531 skb_mark_not_on_list(skb); 1532 } 1533 1534 /** 1535 * skb_queue_empty - check if a queue is empty 1536 * @list: queue head 1537 * 1538 * Returns true if the queue is empty, false otherwise. 1539 */ 1540 static inline int skb_queue_empty(const struct sk_buff_head *list) 1541 { 1542 return list->next == (const struct sk_buff *) list; 1543 } 1544 1545 /** 1546 * skb_queue_empty_lockless - check if a queue is empty 1547 * @list: queue head 1548 * 1549 * Returns true if the queue is empty, false otherwise. 1550 * This variant can be used in lockless contexts. 1551 */ 1552 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1553 { 1554 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1555 } 1556 1557 1558 /** 1559 * skb_queue_is_last - check if skb is the last entry in the queue 1560 * @list: queue head 1561 * @skb: buffer 1562 * 1563 * Returns true if @skb is the last buffer on the list. 1564 */ 1565 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1566 const struct sk_buff *skb) 1567 { 1568 return skb->next == (const struct sk_buff *) list; 1569 } 1570 1571 /** 1572 * skb_queue_is_first - check if skb is the first entry in the queue 1573 * @list: queue head 1574 * @skb: buffer 1575 * 1576 * Returns true if @skb is the first buffer on the list. 1577 */ 1578 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1579 const struct sk_buff *skb) 1580 { 1581 return skb->prev == (const struct sk_buff *) list; 1582 } 1583 1584 /** 1585 * skb_queue_next - return the next packet in the queue 1586 * @list: queue head 1587 * @skb: current buffer 1588 * 1589 * Return the next packet in @list after @skb. It is only valid to 1590 * call this if skb_queue_is_last() evaluates to false. 1591 */ 1592 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1593 const struct sk_buff *skb) 1594 { 1595 /* This BUG_ON may seem severe, but if we just return then we 1596 * are going to dereference garbage. 1597 */ 1598 BUG_ON(skb_queue_is_last(list, skb)); 1599 return skb->next; 1600 } 1601 1602 /** 1603 * skb_queue_prev - return the prev packet in the queue 1604 * @list: queue head 1605 * @skb: current buffer 1606 * 1607 * Return the prev packet in @list before @skb. It is only valid to 1608 * call this if skb_queue_is_first() evaluates to false. 1609 */ 1610 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1611 const struct sk_buff *skb) 1612 { 1613 /* This BUG_ON may seem severe, but if we just return then we 1614 * are going to dereference garbage. 1615 */ 1616 BUG_ON(skb_queue_is_first(list, skb)); 1617 return skb->prev; 1618 } 1619 1620 /** 1621 * skb_get - reference buffer 1622 * @skb: buffer to reference 1623 * 1624 * Makes another reference to a socket buffer and returns a pointer 1625 * to the buffer. 1626 */ 1627 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1628 { 1629 refcount_inc(&skb->users); 1630 return skb; 1631 } 1632 1633 /* 1634 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1635 */ 1636 1637 /** 1638 * skb_cloned - is the buffer a clone 1639 * @skb: buffer to check 1640 * 1641 * Returns true if the buffer was generated with skb_clone() and is 1642 * one of multiple shared copies of the buffer. Cloned buffers are 1643 * shared data so must not be written to under normal circumstances. 1644 */ 1645 static inline int skb_cloned(const struct sk_buff *skb) 1646 { 1647 return skb->cloned && 1648 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1649 } 1650 1651 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1652 { 1653 might_sleep_if(gfpflags_allow_blocking(pri)); 1654 1655 if (skb_cloned(skb)) 1656 return pskb_expand_head(skb, 0, 0, pri); 1657 1658 return 0; 1659 } 1660 1661 /** 1662 * skb_header_cloned - is the header a clone 1663 * @skb: buffer to check 1664 * 1665 * Returns true if modifying the header part of the buffer requires 1666 * the data to be copied. 1667 */ 1668 static inline int skb_header_cloned(const struct sk_buff *skb) 1669 { 1670 int dataref; 1671 1672 if (!skb->cloned) 1673 return 0; 1674 1675 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1676 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1677 return dataref != 1; 1678 } 1679 1680 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1681 { 1682 might_sleep_if(gfpflags_allow_blocking(pri)); 1683 1684 if (skb_header_cloned(skb)) 1685 return pskb_expand_head(skb, 0, 0, pri); 1686 1687 return 0; 1688 } 1689 1690 /** 1691 * __skb_header_release - release reference to header 1692 * @skb: buffer to operate on 1693 */ 1694 static inline void __skb_header_release(struct sk_buff *skb) 1695 { 1696 skb->nohdr = 1; 1697 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1698 } 1699 1700 1701 /** 1702 * skb_shared - is the buffer shared 1703 * @skb: buffer to check 1704 * 1705 * Returns true if more than one person has a reference to this 1706 * buffer. 1707 */ 1708 static inline int skb_shared(const struct sk_buff *skb) 1709 { 1710 return refcount_read(&skb->users) != 1; 1711 } 1712 1713 /** 1714 * skb_share_check - check if buffer is shared and if so clone it 1715 * @skb: buffer to check 1716 * @pri: priority for memory allocation 1717 * 1718 * If the buffer is shared the buffer is cloned and the old copy 1719 * drops a reference. A new clone with a single reference is returned. 1720 * If the buffer is not shared the original buffer is returned. When 1721 * being called from interrupt status or with spinlocks held pri must 1722 * be GFP_ATOMIC. 1723 * 1724 * NULL is returned on a memory allocation failure. 1725 */ 1726 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1727 { 1728 might_sleep_if(gfpflags_allow_blocking(pri)); 1729 if (skb_shared(skb)) { 1730 struct sk_buff *nskb = skb_clone(skb, pri); 1731 1732 if (likely(nskb)) 1733 consume_skb(skb); 1734 else 1735 kfree_skb(skb); 1736 skb = nskb; 1737 } 1738 return skb; 1739 } 1740 1741 /* 1742 * Copy shared buffers into a new sk_buff. We effectively do COW on 1743 * packets to handle cases where we have a local reader and forward 1744 * and a couple of other messy ones. The normal one is tcpdumping 1745 * a packet thats being forwarded. 1746 */ 1747 1748 /** 1749 * skb_unshare - make a copy of a shared buffer 1750 * @skb: buffer to check 1751 * @pri: priority for memory allocation 1752 * 1753 * If the socket buffer is a clone then this function creates a new 1754 * copy of the data, drops a reference count on the old copy and returns 1755 * the new copy with the reference count at 1. If the buffer is not a clone 1756 * the original buffer is returned. When called with a spinlock held or 1757 * from interrupt state @pri must be %GFP_ATOMIC 1758 * 1759 * %NULL is returned on a memory allocation failure. 1760 */ 1761 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1762 gfp_t pri) 1763 { 1764 might_sleep_if(gfpflags_allow_blocking(pri)); 1765 if (skb_cloned(skb)) { 1766 struct sk_buff *nskb = skb_copy(skb, pri); 1767 1768 /* Free our shared copy */ 1769 if (likely(nskb)) 1770 consume_skb(skb); 1771 else 1772 kfree_skb(skb); 1773 skb = nskb; 1774 } 1775 return skb; 1776 } 1777 1778 /** 1779 * skb_peek - peek at the head of an &sk_buff_head 1780 * @list_: list to peek at 1781 * 1782 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1783 * be careful with this one. A peek leaves the buffer on the 1784 * list and someone else may run off with it. You must hold 1785 * the appropriate locks or have a private queue to do this. 1786 * 1787 * Returns %NULL for an empty list or a pointer to the head element. 1788 * The reference count is not incremented and the reference is therefore 1789 * volatile. Use with caution. 1790 */ 1791 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1792 { 1793 struct sk_buff *skb = list_->next; 1794 1795 if (skb == (struct sk_buff *)list_) 1796 skb = NULL; 1797 return skb; 1798 } 1799 1800 /** 1801 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1802 * @list_: list to peek at 1803 * 1804 * Like skb_peek(), but the caller knows that the list is not empty. 1805 */ 1806 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1807 { 1808 return list_->next; 1809 } 1810 1811 /** 1812 * skb_peek_next - peek skb following the given one from a queue 1813 * @skb: skb to start from 1814 * @list_: list to peek at 1815 * 1816 * Returns %NULL when the end of the list is met or a pointer to the 1817 * next element. The reference count is not incremented and the 1818 * reference is therefore volatile. Use with caution. 1819 */ 1820 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1821 const struct sk_buff_head *list_) 1822 { 1823 struct sk_buff *next = skb->next; 1824 1825 if (next == (struct sk_buff *)list_) 1826 next = NULL; 1827 return next; 1828 } 1829 1830 /** 1831 * skb_peek_tail - peek at the tail of an &sk_buff_head 1832 * @list_: list to peek at 1833 * 1834 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1835 * be careful with this one. A peek leaves the buffer on the 1836 * list and someone else may run off with it. You must hold 1837 * the appropriate locks or have a private queue to do this. 1838 * 1839 * Returns %NULL for an empty list or a pointer to the tail element. 1840 * The reference count is not incremented and the reference is therefore 1841 * volatile. Use with caution. 1842 */ 1843 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1844 { 1845 struct sk_buff *skb = READ_ONCE(list_->prev); 1846 1847 if (skb == (struct sk_buff *)list_) 1848 skb = NULL; 1849 return skb; 1850 1851 } 1852 1853 /** 1854 * skb_queue_len - get queue length 1855 * @list_: list to measure 1856 * 1857 * Return the length of an &sk_buff queue. 1858 */ 1859 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1860 { 1861 return list_->qlen; 1862 } 1863 1864 /** 1865 * skb_queue_len_lockless - get queue length 1866 * @list_: list to measure 1867 * 1868 * Return the length of an &sk_buff queue. 1869 * This variant can be used in lockless contexts. 1870 */ 1871 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1872 { 1873 return READ_ONCE(list_->qlen); 1874 } 1875 1876 /** 1877 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1878 * @list: queue to initialize 1879 * 1880 * This initializes only the list and queue length aspects of 1881 * an sk_buff_head object. This allows to initialize the list 1882 * aspects of an sk_buff_head without reinitializing things like 1883 * the spinlock. It can also be used for on-stack sk_buff_head 1884 * objects where the spinlock is known to not be used. 1885 */ 1886 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1887 { 1888 list->prev = list->next = (struct sk_buff *)list; 1889 list->qlen = 0; 1890 } 1891 1892 /* 1893 * This function creates a split out lock class for each invocation; 1894 * this is needed for now since a whole lot of users of the skb-queue 1895 * infrastructure in drivers have different locking usage (in hardirq) 1896 * than the networking core (in softirq only). In the long run either the 1897 * network layer or drivers should need annotation to consolidate the 1898 * main types of usage into 3 classes. 1899 */ 1900 static inline void skb_queue_head_init(struct sk_buff_head *list) 1901 { 1902 spin_lock_init(&list->lock); 1903 __skb_queue_head_init(list); 1904 } 1905 1906 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1907 struct lock_class_key *class) 1908 { 1909 skb_queue_head_init(list); 1910 lockdep_set_class(&list->lock, class); 1911 } 1912 1913 /* 1914 * Insert an sk_buff on a list. 1915 * 1916 * The "__skb_xxxx()" functions are the non-atomic ones that 1917 * can only be called with interrupts disabled. 1918 */ 1919 static inline void __skb_insert(struct sk_buff *newsk, 1920 struct sk_buff *prev, struct sk_buff *next, 1921 struct sk_buff_head *list) 1922 { 1923 /* See skb_queue_empty_lockless() and skb_peek_tail() 1924 * for the opposite READ_ONCE() 1925 */ 1926 WRITE_ONCE(newsk->next, next); 1927 WRITE_ONCE(newsk->prev, prev); 1928 WRITE_ONCE(next->prev, newsk); 1929 WRITE_ONCE(prev->next, newsk); 1930 list->qlen++; 1931 } 1932 1933 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1934 struct sk_buff *prev, 1935 struct sk_buff *next) 1936 { 1937 struct sk_buff *first = list->next; 1938 struct sk_buff *last = list->prev; 1939 1940 WRITE_ONCE(first->prev, prev); 1941 WRITE_ONCE(prev->next, first); 1942 1943 WRITE_ONCE(last->next, next); 1944 WRITE_ONCE(next->prev, last); 1945 } 1946 1947 /** 1948 * skb_queue_splice - join two skb lists, this is designed for stacks 1949 * @list: the new list to add 1950 * @head: the place to add it in the first list 1951 */ 1952 static inline void skb_queue_splice(const struct sk_buff_head *list, 1953 struct sk_buff_head *head) 1954 { 1955 if (!skb_queue_empty(list)) { 1956 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1957 head->qlen += list->qlen; 1958 } 1959 } 1960 1961 /** 1962 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1963 * @list: the new list to add 1964 * @head: the place to add it in the first list 1965 * 1966 * The list at @list is reinitialised 1967 */ 1968 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1969 struct sk_buff_head *head) 1970 { 1971 if (!skb_queue_empty(list)) { 1972 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1973 head->qlen += list->qlen; 1974 __skb_queue_head_init(list); 1975 } 1976 } 1977 1978 /** 1979 * skb_queue_splice_tail - join two skb lists, each list being a queue 1980 * @list: the new list to add 1981 * @head: the place to add it in the first list 1982 */ 1983 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1984 struct sk_buff_head *head) 1985 { 1986 if (!skb_queue_empty(list)) { 1987 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1988 head->qlen += list->qlen; 1989 } 1990 } 1991 1992 /** 1993 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1994 * @list: the new list to add 1995 * @head: the place to add it in the first list 1996 * 1997 * Each of the lists is a queue. 1998 * The list at @list is reinitialised 1999 */ 2000 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2001 struct sk_buff_head *head) 2002 { 2003 if (!skb_queue_empty(list)) { 2004 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2005 head->qlen += list->qlen; 2006 __skb_queue_head_init(list); 2007 } 2008 } 2009 2010 /** 2011 * __skb_queue_after - queue a buffer at the list head 2012 * @list: list to use 2013 * @prev: place after this buffer 2014 * @newsk: buffer to queue 2015 * 2016 * Queue a buffer int the middle of a list. This function takes no locks 2017 * and you must therefore hold required locks before calling it. 2018 * 2019 * A buffer cannot be placed on two lists at the same time. 2020 */ 2021 static inline void __skb_queue_after(struct sk_buff_head *list, 2022 struct sk_buff *prev, 2023 struct sk_buff *newsk) 2024 { 2025 __skb_insert(newsk, prev, prev->next, list); 2026 } 2027 2028 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2029 struct sk_buff_head *list); 2030 2031 static inline void __skb_queue_before(struct sk_buff_head *list, 2032 struct sk_buff *next, 2033 struct sk_buff *newsk) 2034 { 2035 __skb_insert(newsk, next->prev, next, list); 2036 } 2037 2038 /** 2039 * __skb_queue_head - queue a buffer at the list head 2040 * @list: list to use 2041 * @newsk: buffer to queue 2042 * 2043 * Queue a buffer at the start of a list. This function takes no locks 2044 * and you must therefore hold required locks before calling it. 2045 * 2046 * A buffer cannot be placed on two lists at the same time. 2047 */ 2048 static inline void __skb_queue_head(struct sk_buff_head *list, 2049 struct sk_buff *newsk) 2050 { 2051 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2052 } 2053 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2054 2055 /** 2056 * __skb_queue_tail - queue a buffer at the list tail 2057 * @list: list to use 2058 * @newsk: buffer to queue 2059 * 2060 * Queue a buffer at the end of a list. This function takes no locks 2061 * and you must therefore hold required locks before calling it. 2062 * 2063 * A buffer cannot be placed on two lists at the same time. 2064 */ 2065 static inline void __skb_queue_tail(struct sk_buff_head *list, 2066 struct sk_buff *newsk) 2067 { 2068 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2069 } 2070 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2071 2072 /* 2073 * remove sk_buff from list. _Must_ be called atomically, and with 2074 * the list known.. 2075 */ 2076 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2077 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2078 { 2079 struct sk_buff *next, *prev; 2080 2081 WRITE_ONCE(list->qlen, list->qlen - 1); 2082 next = skb->next; 2083 prev = skb->prev; 2084 skb->next = skb->prev = NULL; 2085 WRITE_ONCE(next->prev, prev); 2086 WRITE_ONCE(prev->next, next); 2087 } 2088 2089 /** 2090 * __skb_dequeue - remove from the head of the queue 2091 * @list: list to dequeue from 2092 * 2093 * Remove the head of the list. This function does not take any locks 2094 * so must be used with appropriate locks held only. The head item is 2095 * returned or %NULL if the list is empty. 2096 */ 2097 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2098 { 2099 struct sk_buff *skb = skb_peek(list); 2100 if (skb) 2101 __skb_unlink(skb, list); 2102 return skb; 2103 } 2104 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2105 2106 /** 2107 * __skb_dequeue_tail - remove from the tail of the queue 2108 * @list: list to dequeue from 2109 * 2110 * Remove the tail of the list. This function does not take any locks 2111 * so must be used with appropriate locks held only. The tail item is 2112 * returned or %NULL if the list is empty. 2113 */ 2114 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2115 { 2116 struct sk_buff *skb = skb_peek_tail(list); 2117 if (skb) 2118 __skb_unlink(skb, list); 2119 return skb; 2120 } 2121 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2122 2123 2124 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2125 { 2126 return skb->data_len; 2127 } 2128 2129 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2130 { 2131 return skb->len - skb->data_len; 2132 } 2133 2134 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2135 { 2136 unsigned int i, len = 0; 2137 2138 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2139 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2140 return len; 2141 } 2142 2143 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2144 { 2145 return skb_headlen(skb) + __skb_pagelen(skb); 2146 } 2147 2148 /** 2149 * __skb_fill_page_desc - initialise a paged fragment in an skb 2150 * @skb: buffer containing fragment to be initialised 2151 * @i: paged fragment index to initialise 2152 * @page: the page to use for this fragment 2153 * @off: the offset to the data with @page 2154 * @size: the length of the data 2155 * 2156 * Initialises the @i'th fragment of @skb to point to &size bytes at 2157 * offset @off within @page. 2158 * 2159 * Does not take any additional reference on the fragment. 2160 */ 2161 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2162 struct page *page, int off, int size) 2163 { 2164 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2165 2166 /* 2167 * Propagate page pfmemalloc to the skb if we can. The problem is 2168 * that not all callers have unique ownership of the page but rely 2169 * on page_is_pfmemalloc doing the right thing(tm). 2170 */ 2171 frag->bv_page = page; 2172 frag->bv_offset = off; 2173 skb_frag_size_set(frag, size); 2174 2175 page = compound_head(page); 2176 if (page_is_pfmemalloc(page)) 2177 skb->pfmemalloc = true; 2178 } 2179 2180 /** 2181 * skb_fill_page_desc - initialise a paged fragment in an skb 2182 * @skb: buffer containing fragment to be initialised 2183 * @i: paged fragment index to initialise 2184 * @page: the page to use for this fragment 2185 * @off: the offset to the data with @page 2186 * @size: the length of the data 2187 * 2188 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2189 * @skb to point to @size bytes at offset @off within @page. In 2190 * addition updates @skb such that @i is the last fragment. 2191 * 2192 * Does not take any additional reference on the fragment. 2193 */ 2194 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2195 struct page *page, int off, int size) 2196 { 2197 __skb_fill_page_desc(skb, i, page, off, size); 2198 skb_shinfo(skb)->nr_frags = i + 1; 2199 } 2200 2201 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2202 int size, unsigned int truesize); 2203 2204 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2205 unsigned int truesize); 2206 2207 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2208 2209 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2210 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2211 { 2212 return skb->head + skb->tail; 2213 } 2214 2215 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2216 { 2217 skb->tail = skb->data - skb->head; 2218 } 2219 2220 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2221 { 2222 skb_reset_tail_pointer(skb); 2223 skb->tail += offset; 2224 } 2225 2226 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2227 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2228 { 2229 return skb->tail; 2230 } 2231 2232 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2233 { 2234 skb->tail = skb->data; 2235 } 2236 2237 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2238 { 2239 skb->tail = skb->data + offset; 2240 } 2241 2242 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2243 2244 /* 2245 * Add data to an sk_buff 2246 */ 2247 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2248 void *skb_put(struct sk_buff *skb, unsigned int len); 2249 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2250 { 2251 void *tmp = skb_tail_pointer(skb); 2252 SKB_LINEAR_ASSERT(skb); 2253 skb->tail += len; 2254 skb->len += len; 2255 return tmp; 2256 } 2257 2258 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2259 { 2260 void *tmp = __skb_put(skb, len); 2261 2262 memset(tmp, 0, len); 2263 return tmp; 2264 } 2265 2266 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2267 unsigned int len) 2268 { 2269 void *tmp = __skb_put(skb, len); 2270 2271 memcpy(tmp, data, len); 2272 return tmp; 2273 } 2274 2275 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2276 { 2277 *(u8 *)__skb_put(skb, 1) = val; 2278 } 2279 2280 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2281 { 2282 void *tmp = skb_put(skb, len); 2283 2284 memset(tmp, 0, len); 2285 2286 return tmp; 2287 } 2288 2289 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2290 unsigned int len) 2291 { 2292 void *tmp = skb_put(skb, len); 2293 2294 memcpy(tmp, data, len); 2295 2296 return tmp; 2297 } 2298 2299 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2300 { 2301 *(u8 *)skb_put(skb, 1) = val; 2302 } 2303 2304 void *skb_push(struct sk_buff *skb, unsigned int len); 2305 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2306 { 2307 skb->data -= len; 2308 skb->len += len; 2309 return skb->data; 2310 } 2311 2312 void *skb_pull(struct sk_buff *skb, unsigned int len); 2313 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2314 { 2315 skb->len -= len; 2316 BUG_ON(skb->len < skb->data_len); 2317 return skb->data += len; 2318 } 2319 2320 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2321 { 2322 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2323 } 2324 2325 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2326 2327 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2328 { 2329 if (len > skb_headlen(skb) && 2330 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2331 return NULL; 2332 skb->len -= len; 2333 return skb->data += len; 2334 } 2335 2336 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2337 { 2338 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2339 } 2340 2341 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2342 { 2343 if (likely(len <= skb_headlen(skb))) 2344 return true; 2345 if (unlikely(len > skb->len)) 2346 return false; 2347 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2348 } 2349 2350 void skb_condense(struct sk_buff *skb); 2351 2352 /** 2353 * skb_headroom - bytes at buffer head 2354 * @skb: buffer to check 2355 * 2356 * Return the number of bytes of free space at the head of an &sk_buff. 2357 */ 2358 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2359 { 2360 return skb->data - skb->head; 2361 } 2362 2363 /** 2364 * skb_tailroom - bytes at buffer end 2365 * @skb: buffer to check 2366 * 2367 * Return the number of bytes of free space at the tail of an sk_buff 2368 */ 2369 static inline int skb_tailroom(const struct sk_buff *skb) 2370 { 2371 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2372 } 2373 2374 /** 2375 * skb_availroom - bytes at buffer end 2376 * @skb: buffer to check 2377 * 2378 * Return the number of bytes of free space at the tail of an sk_buff 2379 * allocated by sk_stream_alloc() 2380 */ 2381 static inline int skb_availroom(const struct sk_buff *skb) 2382 { 2383 if (skb_is_nonlinear(skb)) 2384 return 0; 2385 2386 return skb->end - skb->tail - skb->reserved_tailroom; 2387 } 2388 2389 /** 2390 * skb_reserve - adjust headroom 2391 * @skb: buffer to alter 2392 * @len: bytes to move 2393 * 2394 * Increase the headroom of an empty &sk_buff by reducing the tail 2395 * room. This is only allowed for an empty buffer. 2396 */ 2397 static inline void skb_reserve(struct sk_buff *skb, int len) 2398 { 2399 skb->data += len; 2400 skb->tail += len; 2401 } 2402 2403 /** 2404 * skb_tailroom_reserve - adjust reserved_tailroom 2405 * @skb: buffer to alter 2406 * @mtu: maximum amount of headlen permitted 2407 * @needed_tailroom: minimum amount of reserved_tailroom 2408 * 2409 * Set reserved_tailroom so that headlen can be as large as possible but 2410 * not larger than mtu and tailroom cannot be smaller than 2411 * needed_tailroom. 2412 * The required headroom should already have been reserved before using 2413 * this function. 2414 */ 2415 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2416 unsigned int needed_tailroom) 2417 { 2418 SKB_LINEAR_ASSERT(skb); 2419 if (mtu < skb_tailroom(skb) - needed_tailroom) 2420 /* use at most mtu */ 2421 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2422 else 2423 /* use up to all available space */ 2424 skb->reserved_tailroom = needed_tailroom; 2425 } 2426 2427 #define ENCAP_TYPE_ETHER 0 2428 #define ENCAP_TYPE_IPPROTO 1 2429 2430 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2431 __be16 protocol) 2432 { 2433 skb->inner_protocol = protocol; 2434 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2435 } 2436 2437 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2438 __u8 ipproto) 2439 { 2440 skb->inner_ipproto = ipproto; 2441 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2442 } 2443 2444 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2445 { 2446 skb->inner_mac_header = skb->mac_header; 2447 skb->inner_network_header = skb->network_header; 2448 skb->inner_transport_header = skb->transport_header; 2449 } 2450 2451 static inline void skb_reset_mac_len(struct sk_buff *skb) 2452 { 2453 skb->mac_len = skb->network_header - skb->mac_header; 2454 } 2455 2456 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2457 *skb) 2458 { 2459 return skb->head + skb->inner_transport_header; 2460 } 2461 2462 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2463 { 2464 return skb_inner_transport_header(skb) - skb->data; 2465 } 2466 2467 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2468 { 2469 skb->inner_transport_header = skb->data - skb->head; 2470 } 2471 2472 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2473 const int offset) 2474 { 2475 skb_reset_inner_transport_header(skb); 2476 skb->inner_transport_header += offset; 2477 } 2478 2479 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2480 { 2481 return skb->head + skb->inner_network_header; 2482 } 2483 2484 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2485 { 2486 skb->inner_network_header = skb->data - skb->head; 2487 } 2488 2489 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2490 const int offset) 2491 { 2492 skb_reset_inner_network_header(skb); 2493 skb->inner_network_header += offset; 2494 } 2495 2496 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2497 { 2498 return skb->head + skb->inner_mac_header; 2499 } 2500 2501 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2502 { 2503 skb->inner_mac_header = skb->data - skb->head; 2504 } 2505 2506 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2507 const int offset) 2508 { 2509 skb_reset_inner_mac_header(skb); 2510 skb->inner_mac_header += offset; 2511 } 2512 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2513 { 2514 return skb->transport_header != (typeof(skb->transport_header))~0U; 2515 } 2516 2517 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2518 { 2519 return skb->head + skb->transport_header; 2520 } 2521 2522 static inline void skb_reset_transport_header(struct sk_buff *skb) 2523 { 2524 skb->transport_header = skb->data - skb->head; 2525 } 2526 2527 static inline void skb_set_transport_header(struct sk_buff *skb, 2528 const int offset) 2529 { 2530 skb_reset_transport_header(skb); 2531 skb->transport_header += offset; 2532 } 2533 2534 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2535 { 2536 return skb->head + skb->network_header; 2537 } 2538 2539 static inline void skb_reset_network_header(struct sk_buff *skb) 2540 { 2541 skb->network_header = skb->data - skb->head; 2542 } 2543 2544 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2545 { 2546 skb_reset_network_header(skb); 2547 skb->network_header += offset; 2548 } 2549 2550 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2551 { 2552 return skb->head + skb->mac_header; 2553 } 2554 2555 static inline int skb_mac_offset(const struct sk_buff *skb) 2556 { 2557 return skb_mac_header(skb) - skb->data; 2558 } 2559 2560 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2561 { 2562 return skb->network_header - skb->mac_header; 2563 } 2564 2565 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2566 { 2567 return skb->mac_header != (typeof(skb->mac_header))~0U; 2568 } 2569 2570 static inline void skb_unset_mac_header(struct sk_buff *skb) 2571 { 2572 skb->mac_header = (typeof(skb->mac_header))~0U; 2573 } 2574 2575 static inline void skb_reset_mac_header(struct sk_buff *skb) 2576 { 2577 skb->mac_header = skb->data - skb->head; 2578 } 2579 2580 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2581 { 2582 skb_reset_mac_header(skb); 2583 skb->mac_header += offset; 2584 } 2585 2586 static inline void skb_pop_mac_header(struct sk_buff *skb) 2587 { 2588 skb->mac_header = skb->network_header; 2589 } 2590 2591 static inline void skb_probe_transport_header(struct sk_buff *skb) 2592 { 2593 struct flow_keys_basic keys; 2594 2595 if (skb_transport_header_was_set(skb)) 2596 return; 2597 2598 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2599 NULL, 0, 0, 0, 0)) 2600 skb_set_transport_header(skb, keys.control.thoff); 2601 } 2602 2603 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2604 { 2605 if (skb_mac_header_was_set(skb)) { 2606 const unsigned char *old_mac = skb_mac_header(skb); 2607 2608 skb_set_mac_header(skb, -skb->mac_len); 2609 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2610 } 2611 } 2612 2613 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2614 { 2615 return skb->csum_start - skb_headroom(skb); 2616 } 2617 2618 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2619 { 2620 return skb->head + skb->csum_start; 2621 } 2622 2623 static inline int skb_transport_offset(const struct sk_buff *skb) 2624 { 2625 return skb_transport_header(skb) - skb->data; 2626 } 2627 2628 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2629 { 2630 return skb->transport_header - skb->network_header; 2631 } 2632 2633 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2634 { 2635 return skb->inner_transport_header - skb->inner_network_header; 2636 } 2637 2638 static inline int skb_network_offset(const struct sk_buff *skb) 2639 { 2640 return skb_network_header(skb) - skb->data; 2641 } 2642 2643 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2644 { 2645 return skb_inner_network_header(skb) - skb->data; 2646 } 2647 2648 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2649 { 2650 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2651 } 2652 2653 /* 2654 * CPUs often take a performance hit when accessing unaligned memory 2655 * locations. The actual performance hit varies, it can be small if the 2656 * hardware handles it or large if we have to take an exception and fix it 2657 * in software. 2658 * 2659 * Since an ethernet header is 14 bytes network drivers often end up with 2660 * the IP header at an unaligned offset. The IP header can be aligned by 2661 * shifting the start of the packet by 2 bytes. Drivers should do this 2662 * with: 2663 * 2664 * skb_reserve(skb, NET_IP_ALIGN); 2665 * 2666 * The downside to this alignment of the IP header is that the DMA is now 2667 * unaligned. On some architectures the cost of an unaligned DMA is high 2668 * and this cost outweighs the gains made by aligning the IP header. 2669 * 2670 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2671 * to be overridden. 2672 */ 2673 #ifndef NET_IP_ALIGN 2674 #define NET_IP_ALIGN 2 2675 #endif 2676 2677 /* 2678 * The networking layer reserves some headroom in skb data (via 2679 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2680 * the header has to grow. In the default case, if the header has to grow 2681 * 32 bytes or less we avoid the reallocation. 2682 * 2683 * Unfortunately this headroom changes the DMA alignment of the resulting 2684 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2685 * on some architectures. An architecture can override this value, 2686 * perhaps setting it to a cacheline in size (since that will maintain 2687 * cacheline alignment of the DMA). It must be a power of 2. 2688 * 2689 * Various parts of the networking layer expect at least 32 bytes of 2690 * headroom, you should not reduce this. 2691 * 2692 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2693 * to reduce average number of cache lines per packet. 2694 * get_rps_cpu() for example only access one 64 bytes aligned block : 2695 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2696 */ 2697 #ifndef NET_SKB_PAD 2698 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2699 #endif 2700 2701 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2702 2703 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2704 { 2705 if (WARN_ON(skb_is_nonlinear(skb))) 2706 return; 2707 skb->len = len; 2708 skb_set_tail_pointer(skb, len); 2709 } 2710 2711 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2712 { 2713 __skb_set_length(skb, len); 2714 } 2715 2716 void skb_trim(struct sk_buff *skb, unsigned int len); 2717 2718 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2719 { 2720 if (skb->data_len) 2721 return ___pskb_trim(skb, len); 2722 __skb_trim(skb, len); 2723 return 0; 2724 } 2725 2726 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2727 { 2728 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2729 } 2730 2731 /** 2732 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2733 * @skb: buffer to alter 2734 * @len: new length 2735 * 2736 * This is identical to pskb_trim except that the caller knows that 2737 * the skb is not cloned so we should never get an error due to out- 2738 * of-memory. 2739 */ 2740 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2741 { 2742 int err = pskb_trim(skb, len); 2743 BUG_ON(err); 2744 } 2745 2746 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2747 { 2748 unsigned int diff = len - skb->len; 2749 2750 if (skb_tailroom(skb) < diff) { 2751 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2752 GFP_ATOMIC); 2753 if (ret) 2754 return ret; 2755 } 2756 __skb_set_length(skb, len); 2757 return 0; 2758 } 2759 2760 /** 2761 * skb_orphan - orphan a buffer 2762 * @skb: buffer to orphan 2763 * 2764 * If a buffer currently has an owner then we call the owner's 2765 * destructor function and make the @skb unowned. The buffer continues 2766 * to exist but is no longer charged to its former owner. 2767 */ 2768 static inline void skb_orphan(struct sk_buff *skb) 2769 { 2770 if (skb->destructor) { 2771 skb->destructor(skb); 2772 skb->destructor = NULL; 2773 skb->sk = NULL; 2774 } else { 2775 BUG_ON(skb->sk); 2776 } 2777 } 2778 2779 /** 2780 * skb_orphan_frags - orphan the frags contained in a buffer 2781 * @skb: buffer to orphan frags from 2782 * @gfp_mask: allocation mask for replacement pages 2783 * 2784 * For each frag in the SKB which needs a destructor (i.e. has an 2785 * owner) create a copy of that frag and release the original 2786 * page by calling the destructor. 2787 */ 2788 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2789 { 2790 if (likely(!skb_zcopy(skb))) 2791 return 0; 2792 if (!skb_zcopy_is_nouarg(skb) && 2793 skb_uarg(skb)->callback == msg_zerocopy_callback) 2794 return 0; 2795 return skb_copy_ubufs(skb, gfp_mask); 2796 } 2797 2798 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2799 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2800 { 2801 if (likely(!skb_zcopy(skb))) 2802 return 0; 2803 return skb_copy_ubufs(skb, gfp_mask); 2804 } 2805 2806 /** 2807 * __skb_queue_purge - empty a list 2808 * @list: list to empty 2809 * 2810 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2811 * the list and one reference dropped. This function does not take the 2812 * list lock and the caller must hold the relevant locks to use it. 2813 */ 2814 static inline void __skb_queue_purge(struct sk_buff_head *list) 2815 { 2816 struct sk_buff *skb; 2817 while ((skb = __skb_dequeue(list)) != NULL) 2818 kfree_skb(skb); 2819 } 2820 void skb_queue_purge(struct sk_buff_head *list); 2821 2822 unsigned int skb_rbtree_purge(struct rb_root *root); 2823 2824 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2825 2826 /** 2827 * netdev_alloc_frag - allocate a page fragment 2828 * @fragsz: fragment size 2829 * 2830 * Allocates a frag from a page for receive buffer. 2831 * Uses GFP_ATOMIC allocations. 2832 */ 2833 static inline void *netdev_alloc_frag(unsigned int fragsz) 2834 { 2835 return __netdev_alloc_frag_align(fragsz, ~0u); 2836 } 2837 2838 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 2839 unsigned int align) 2840 { 2841 WARN_ON_ONCE(!is_power_of_2(align)); 2842 return __netdev_alloc_frag_align(fragsz, -align); 2843 } 2844 2845 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2846 gfp_t gfp_mask); 2847 2848 /** 2849 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2850 * @dev: network device to receive on 2851 * @length: length to allocate 2852 * 2853 * Allocate a new &sk_buff and assign it a usage count of one. The 2854 * buffer has unspecified headroom built in. Users should allocate 2855 * the headroom they think they need without accounting for the 2856 * built in space. The built in space is used for optimisations. 2857 * 2858 * %NULL is returned if there is no free memory. Although this function 2859 * allocates memory it can be called from an interrupt. 2860 */ 2861 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2862 unsigned int length) 2863 { 2864 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2865 } 2866 2867 /* legacy helper around __netdev_alloc_skb() */ 2868 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2869 gfp_t gfp_mask) 2870 { 2871 return __netdev_alloc_skb(NULL, length, gfp_mask); 2872 } 2873 2874 /* legacy helper around netdev_alloc_skb() */ 2875 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2876 { 2877 return netdev_alloc_skb(NULL, length); 2878 } 2879 2880 2881 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2882 unsigned int length, gfp_t gfp) 2883 { 2884 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2885 2886 if (NET_IP_ALIGN && skb) 2887 skb_reserve(skb, NET_IP_ALIGN); 2888 return skb; 2889 } 2890 2891 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2892 unsigned int length) 2893 { 2894 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2895 } 2896 2897 static inline void skb_free_frag(void *addr) 2898 { 2899 page_frag_free(addr); 2900 } 2901 2902 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2903 2904 static inline void *napi_alloc_frag(unsigned int fragsz) 2905 { 2906 return __napi_alloc_frag_align(fragsz, ~0u); 2907 } 2908 2909 static inline void *napi_alloc_frag_align(unsigned int fragsz, 2910 unsigned int align) 2911 { 2912 WARN_ON_ONCE(!is_power_of_2(align)); 2913 return __napi_alloc_frag_align(fragsz, -align); 2914 } 2915 2916 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2917 unsigned int length, gfp_t gfp_mask); 2918 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2919 unsigned int length) 2920 { 2921 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2922 } 2923 void napi_consume_skb(struct sk_buff *skb, int budget); 2924 2925 void napi_skb_free_stolen_head(struct sk_buff *skb); 2926 void __kfree_skb_defer(struct sk_buff *skb); 2927 2928 /** 2929 * __dev_alloc_pages - allocate page for network Rx 2930 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2931 * @order: size of the allocation 2932 * 2933 * Allocate a new page. 2934 * 2935 * %NULL is returned if there is no free memory. 2936 */ 2937 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2938 unsigned int order) 2939 { 2940 /* This piece of code contains several assumptions. 2941 * 1. This is for device Rx, therefor a cold page is preferred. 2942 * 2. The expectation is the user wants a compound page. 2943 * 3. If requesting a order 0 page it will not be compound 2944 * due to the check to see if order has a value in prep_new_page 2945 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2946 * code in gfp_to_alloc_flags that should be enforcing this. 2947 */ 2948 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2949 2950 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2951 } 2952 2953 static inline struct page *dev_alloc_pages(unsigned int order) 2954 { 2955 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2956 } 2957 2958 /** 2959 * __dev_alloc_page - allocate a page for network Rx 2960 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2961 * 2962 * Allocate a new page. 2963 * 2964 * %NULL is returned if there is no free memory. 2965 */ 2966 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2967 { 2968 return __dev_alloc_pages(gfp_mask, 0); 2969 } 2970 2971 static inline struct page *dev_alloc_page(void) 2972 { 2973 return dev_alloc_pages(0); 2974 } 2975 2976 /** 2977 * dev_page_is_reusable - check whether a page can be reused for network Rx 2978 * @page: the page to test 2979 * 2980 * A page shouldn't be considered for reusing/recycling if it was allocated 2981 * under memory pressure or at a distant memory node. 2982 * 2983 * Returns false if this page should be returned to page allocator, true 2984 * otherwise. 2985 */ 2986 static inline bool dev_page_is_reusable(const struct page *page) 2987 { 2988 return likely(page_to_nid(page) == numa_mem_id() && 2989 !page_is_pfmemalloc(page)); 2990 } 2991 2992 /** 2993 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2994 * @page: The page that was allocated from skb_alloc_page 2995 * @skb: The skb that may need pfmemalloc set 2996 */ 2997 static inline void skb_propagate_pfmemalloc(const struct page *page, 2998 struct sk_buff *skb) 2999 { 3000 if (page_is_pfmemalloc(page)) 3001 skb->pfmemalloc = true; 3002 } 3003 3004 /** 3005 * skb_frag_off() - Returns the offset of a skb fragment 3006 * @frag: the paged fragment 3007 */ 3008 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3009 { 3010 return frag->bv_offset; 3011 } 3012 3013 /** 3014 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3015 * @frag: skb fragment 3016 * @delta: value to add 3017 */ 3018 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3019 { 3020 frag->bv_offset += delta; 3021 } 3022 3023 /** 3024 * skb_frag_off_set() - Sets the offset of a skb fragment 3025 * @frag: skb fragment 3026 * @offset: offset of fragment 3027 */ 3028 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3029 { 3030 frag->bv_offset = offset; 3031 } 3032 3033 /** 3034 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3035 * @fragto: skb fragment where offset is set 3036 * @fragfrom: skb fragment offset is copied from 3037 */ 3038 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3039 const skb_frag_t *fragfrom) 3040 { 3041 fragto->bv_offset = fragfrom->bv_offset; 3042 } 3043 3044 /** 3045 * skb_frag_page - retrieve the page referred to by a paged fragment 3046 * @frag: the paged fragment 3047 * 3048 * Returns the &struct page associated with @frag. 3049 */ 3050 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3051 { 3052 return frag->bv_page; 3053 } 3054 3055 /** 3056 * __skb_frag_ref - take an addition reference on a paged fragment. 3057 * @frag: the paged fragment 3058 * 3059 * Takes an additional reference on the paged fragment @frag. 3060 */ 3061 static inline void __skb_frag_ref(skb_frag_t *frag) 3062 { 3063 get_page(skb_frag_page(frag)); 3064 } 3065 3066 /** 3067 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3068 * @skb: the buffer 3069 * @f: the fragment offset. 3070 * 3071 * Takes an additional reference on the @f'th paged fragment of @skb. 3072 */ 3073 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3074 { 3075 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3076 } 3077 3078 /** 3079 * __skb_frag_unref - release a reference on a paged fragment. 3080 * @frag: the paged fragment 3081 * 3082 * Releases a reference on the paged fragment @frag. 3083 */ 3084 static inline void __skb_frag_unref(skb_frag_t *frag) 3085 { 3086 put_page(skb_frag_page(frag)); 3087 } 3088 3089 /** 3090 * skb_frag_unref - release a reference on a paged fragment of an skb. 3091 * @skb: the buffer 3092 * @f: the fragment offset 3093 * 3094 * Releases a reference on the @f'th paged fragment of @skb. 3095 */ 3096 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3097 { 3098 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 3099 } 3100 3101 /** 3102 * skb_frag_address - gets the address of the data contained in a paged fragment 3103 * @frag: the paged fragment buffer 3104 * 3105 * Returns the address of the data within @frag. The page must already 3106 * be mapped. 3107 */ 3108 static inline void *skb_frag_address(const skb_frag_t *frag) 3109 { 3110 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3111 } 3112 3113 /** 3114 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3115 * @frag: the paged fragment buffer 3116 * 3117 * Returns the address of the data within @frag. Checks that the page 3118 * is mapped and returns %NULL otherwise. 3119 */ 3120 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3121 { 3122 void *ptr = page_address(skb_frag_page(frag)); 3123 if (unlikely(!ptr)) 3124 return NULL; 3125 3126 return ptr + skb_frag_off(frag); 3127 } 3128 3129 /** 3130 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3131 * @fragto: skb fragment where page is set 3132 * @fragfrom: skb fragment page is copied from 3133 */ 3134 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3135 const skb_frag_t *fragfrom) 3136 { 3137 fragto->bv_page = fragfrom->bv_page; 3138 } 3139 3140 /** 3141 * __skb_frag_set_page - sets the page contained in a paged fragment 3142 * @frag: the paged fragment 3143 * @page: the page to set 3144 * 3145 * Sets the fragment @frag to contain @page. 3146 */ 3147 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3148 { 3149 frag->bv_page = page; 3150 } 3151 3152 /** 3153 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3154 * @skb: the buffer 3155 * @f: the fragment offset 3156 * @page: the page to set 3157 * 3158 * Sets the @f'th fragment of @skb to contain @page. 3159 */ 3160 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3161 struct page *page) 3162 { 3163 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3164 } 3165 3166 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3167 3168 /** 3169 * skb_frag_dma_map - maps a paged fragment via the DMA API 3170 * @dev: the device to map the fragment to 3171 * @frag: the paged fragment to map 3172 * @offset: the offset within the fragment (starting at the 3173 * fragment's own offset) 3174 * @size: the number of bytes to map 3175 * @dir: the direction of the mapping (``PCI_DMA_*``) 3176 * 3177 * Maps the page associated with @frag to @device. 3178 */ 3179 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3180 const skb_frag_t *frag, 3181 size_t offset, size_t size, 3182 enum dma_data_direction dir) 3183 { 3184 return dma_map_page(dev, skb_frag_page(frag), 3185 skb_frag_off(frag) + offset, size, dir); 3186 } 3187 3188 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3189 gfp_t gfp_mask) 3190 { 3191 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3192 } 3193 3194 3195 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3196 gfp_t gfp_mask) 3197 { 3198 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3199 } 3200 3201 3202 /** 3203 * skb_clone_writable - is the header of a clone writable 3204 * @skb: buffer to check 3205 * @len: length up to which to write 3206 * 3207 * Returns true if modifying the header part of the cloned buffer 3208 * does not requires the data to be copied. 3209 */ 3210 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3211 { 3212 return !skb_header_cloned(skb) && 3213 skb_headroom(skb) + len <= skb->hdr_len; 3214 } 3215 3216 static inline int skb_try_make_writable(struct sk_buff *skb, 3217 unsigned int write_len) 3218 { 3219 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3220 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3221 } 3222 3223 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3224 int cloned) 3225 { 3226 int delta = 0; 3227 3228 if (headroom > skb_headroom(skb)) 3229 delta = headroom - skb_headroom(skb); 3230 3231 if (delta || cloned) 3232 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3233 GFP_ATOMIC); 3234 return 0; 3235 } 3236 3237 /** 3238 * skb_cow - copy header of skb when it is required 3239 * @skb: buffer to cow 3240 * @headroom: needed headroom 3241 * 3242 * If the skb passed lacks sufficient headroom or its data part 3243 * is shared, data is reallocated. If reallocation fails, an error 3244 * is returned and original skb is not changed. 3245 * 3246 * The result is skb with writable area skb->head...skb->tail 3247 * and at least @headroom of space at head. 3248 */ 3249 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3250 { 3251 return __skb_cow(skb, headroom, skb_cloned(skb)); 3252 } 3253 3254 /** 3255 * skb_cow_head - skb_cow but only making the head writable 3256 * @skb: buffer to cow 3257 * @headroom: needed headroom 3258 * 3259 * This function is identical to skb_cow except that we replace the 3260 * skb_cloned check by skb_header_cloned. It should be used when 3261 * you only need to push on some header and do not need to modify 3262 * the data. 3263 */ 3264 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3265 { 3266 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3267 } 3268 3269 /** 3270 * skb_padto - pad an skbuff up to a minimal size 3271 * @skb: buffer to pad 3272 * @len: minimal length 3273 * 3274 * Pads up a buffer to ensure the trailing bytes exist and are 3275 * blanked. If the buffer already contains sufficient data it 3276 * is untouched. Otherwise it is extended. Returns zero on 3277 * success. The skb is freed on error. 3278 */ 3279 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3280 { 3281 unsigned int size = skb->len; 3282 if (likely(size >= len)) 3283 return 0; 3284 return skb_pad(skb, len - size); 3285 } 3286 3287 /** 3288 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3289 * @skb: buffer to pad 3290 * @len: minimal length 3291 * @free_on_error: free buffer on error 3292 * 3293 * Pads up a buffer to ensure the trailing bytes exist and are 3294 * blanked. If the buffer already contains sufficient data it 3295 * is untouched. Otherwise it is extended. Returns zero on 3296 * success. The skb is freed on error if @free_on_error is true. 3297 */ 3298 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3299 unsigned int len, 3300 bool free_on_error) 3301 { 3302 unsigned int size = skb->len; 3303 3304 if (unlikely(size < len)) { 3305 len -= size; 3306 if (__skb_pad(skb, len, free_on_error)) 3307 return -ENOMEM; 3308 __skb_put(skb, len); 3309 } 3310 return 0; 3311 } 3312 3313 /** 3314 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3315 * @skb: buffer to pad 3316 * @len: minimal length 3317 * 3318 * Pads up a buffer to ensure the trailing bytes exist and are 3319 * blanked. If the buffer already contains sufficient data it 3320 * is untouched. Otherwise it is extended. Returns zero on 3321 * success. The skb is freed on error. 3322 */ 3323 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3324 { 3325 return __skb_put_padto(skb, len, true); 3326 } 3327 3328 static inline int skb_add_data(struct sk_buff *skb, 3329 struct iov_iter *from, int copy) 3330 { 3331 const int off = skb->len; 3332 3333 if (skb->ip_summed == CHECKSUM_NONE) { 3334 __wsum csum = 0; 3335 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3336 &csum, from)) { 3337 skb->csum = csum_block_add(skb->csum, csum, off); 3338 return 0; 3339 } 3340 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3341 return 0; 3342 3343 __skb_trim(skb, off); 3344 return -EFAULT; 3345 } 3346 3347 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3348 const struct page *page, int off) 3349 { 3350 if (skb_zcopy(skb)) 3351 return false; 3352 if (i) { 3353 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3354 3355 return page == skb_frag_page(frag) && 3356 off == skb_frag_off(frag) + skb_frag_size(frag); 3357 } 3358 return false; 3359 } 3360 3361 static inline int __skb_linearize(struct sk_buff *skb) 3362 { 3363 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3364 } 3365 3366 /** 3367 * skb_linearize - convert paged skb to linear one 3368 * @skb: buffer to linarize 3369 * 3370 * If there is no free memory -ENOMEM is returned, otherwise zero 3371 * is returned and the old skb data released. 3372 */ 3373 static inline int skb_linearize(struct sk_buff *skb) 3374 { 3375 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3376 } 3377 3378 /** 3379 * skb_has_shared_frag - can any frag be overwritten 3380 * @skb: buffer to test 3381 * 3382 * Return true if the skb has at least one frag that might be modified 3383 * by an external entity (as in vmsplice()/sendfile()) 3384 */ 3385 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3386 { 3387 return skb_is_nonlinear(skb) && 3388 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3389 } 3390 3391 /** 3392 * skb_linearize_cow - make sure skb is linear and writable 3393 * @skb: buffer to process 3394 * 3395 * If there is no free memory -ENOMEM is returned, otherwise zero 3396 * is returned and the old skb data released. 3397 */ 3398 static inline int skb_linearize_cow(struct sk_buff *skb) 3399 { 3400 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3401 __skb_linearize(skb) : 0; 3402 } 3403 3404 static __always_inline void 3405 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3406 unsigned int off) 3407 { 3408 if (skb->ip_summed == CHECKSUM_COMPLETE) 3409 skb->csum = csum_block_sub(skb->csum, 3410 csum_partial(start, len, 0), off); 3411 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3412 skb_checksum_start_offset(skb) < 0) 3413 skb->ip_summed = CHECKSUM_NONE; 3414 } 3415 3416 /** 3417 * skb_postpull_rcsum - update checksum for received skb after pull 3418 * @skb: buffer to update 3419 * @start: start of data before pull 3420 * @len: length of data pulled 3421 * 3422 * After doing a pull on a received packet, you need to call this to 3423 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3424 * CHECKSUM_NONE so that it can be recomputed from scratch. 3425 */ 3426 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3427 const void *start, unsigned int len) 3428 { 3429 __skb_postpull_rcsum(skb, start, len, 0); 3430 } 3431 3432 static __always_inline void 3433 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3434 unsigned int off) 3435 { 3436 if (skb->ip_summed == CHECKSUM_COMPLETE) 3437 skb->csum = csum_block_add(skb->csum, 3438 csum_partial(start, len, 0), off); 3439 } 3440 3441 /** 3442 * skb_postpush_rcsum - update checksum for received skb after push 3443 * @skb: buffer to update 3444 * @start: start of data after push 3445 * @len: length of data pushed 3446 * 3447 * After doing a push on a received packet, you need to call this to 3448 * update the CHECKSUM_COMPLETE checksum. 3449 */ 3450 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3451 const void *start, unsigned int len) 3452 { 3453 __skb_postpush_rcsum(skb, start, len, 0); 3454 } 3455 3456 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3457 3458 /** 3459 * skb_push_rcsum - push skb and update receive checksum 3460 * @skb: buffer to update 3461 * @len: length of data pulled 3462 * 3463 * This function performs an skb_push on the packet and updates 3464 * the CHECKSUM_COMPLETE checksum. It should be used on 3465 * receive path processing instead of skb_push unless you know 3466 * that the checksum difference is zero (e.g., a valid IP header) 3467 * or you are setting ip_summed to CHECKSUM_NONE. 3468 */ 3469 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3470 { 3471 skb_push(skb, len); 3472 skb_postpush_rcsum(skb, skb->data, len); 3473 return skb->data; 3474 } 3475 3476 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3477 /** 3478 * pskb_trim_rcsum - trim received skb and update checksum 3479 * @skb: buffer to trim 3480 * @len: new length 3481 * 3482 * This is exactly the same as pskb_trim except that it ensures the 3483 * checksum of received packets are still valid after the operation. 3484 * It can change skb pointers. 3485 */ 3486 3487 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3488 { 3489 if (likely(len >= skb->len)) 3490 return 0; 3491 return pskb_trim_rcsum_slow(skb, len); 3492 } 3493 3494 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3495 { 3496 if (skb->ip_summed == CHECKSUM_COMPLETE) 3497 skb->ip_summed = CHECKSUM_NONE; 3498 __skb_trim(skb, len); 3499 return 0; 3500 } 3501 3502 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3503 { 3504 if (skb->ip_summed == CHECKSUM_COMPLETE) 3505 skb->ip_summed = CHECKSUM_NONE; 3506 return __skb_grow(skb, len); 3507 } 3508 3509 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3510 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3511 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3512 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3513 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3514 3515 #define skb_queue_walk(queue, skb) \ 3516 for (skb = (queue)->next; \ 3517 skb != (struct sk_buff *)(queue); \ 3518 skb = skb->next) 3519 3520 #define skb_queue_walk_safe(queue, skb, tmp) \ 3521 for (skb = (queue)->next, tmp = skb->next; \ 3522 skb != (struct sk_buff *)(queue); \ 3523 skb = tmp, tmp = skb->next) 3524 3525 #define skb_queue_walk_from(queue, skb) \ 3526 for (; skb != (struct sk_buff *)(queue); \ 3527 skb = skb->next) 3528 3529 #define skb_rbtree_walk(skb, root) \ 3530 for (skb = skb_rb_first(root); skb != NULL; \ 3531 skb = skb_rb_next(skb)) 3532 3533 #define skb_rbtree_walk_from(skb) \ 3534 for (; skb != NULL; \ 3535 skb = skb_rb_next(skb)) 3536 3537 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3538 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3539 skb = tmp) 3540 3541 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3542 for (tmp = skb->next; \ 3543 skb != (struct sk_buff *)(queue); \ 3544 skb = tmp, tmp = skb->next) 3545 3546 #define skb_queue_reverse_walk(queue, skb) \ 3547 for (skb = (queue)->prev; \ 3548 skb != (struct sk_buff *)(queue); \ 3549 skb = skb->prev) 3550 3551 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3552 for (skb = (queue)->prev, tmp = skb->prev; \ 3553 skb != (struct sk_buff *)(queue); \ 3554 skb = tmp, tmp = skb->prev) 3555 3556 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3557 for (tmp = skb->prev; \ 3558 skb != (struct sk_buff *)(queue); \ 3559 skb = tmp, tmp = skb->prev) 3560 3561 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3562 { 3563 return skb_shinfo(skb)->frag_list != NULL; 3564 } 3565 3566 static inline void skb_frag_list_init(struct sk_buff *skb) 3567 { 3568 skb_shinfo(skb)->frag_list = NULL; 3569 } 3570 3571 #define skb_walk_frags(skb, iter) \ 3572 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3573 3574 3575 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3576 int *err, long *timeo_p, 3577 const struct sk_buff *skb); 3578 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3579 struct sk_buff_head *queue, 3580 unsigned int flags, 3581 int *off, int *err, 3582 struct sk_buff **last); 3583 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3584 struct sk_buff_head *queue, 3585 unsigned int flags, int *off, int *err, 3586 struct sk_buff **last); 3587 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3588 struct sk_buff_head *sk_queue, 3589 unsigned int flags, int *off, int *err); 3590 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3591 int *err); 3592 __poll_t datagram_poll(struct file *file, struct socket *sock, 3593 struct poll_table_struct *wait); 3594 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3595 struct iov_iter *to, int size); 3596 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3597 struct msghdr *msg, int size) 3598 { 3599 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3600 } 3601 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3602 struct msghdr *msg); 3603 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3604 struct iov_iter *to, int len, 3605 struct ahash_request *hash); 3606 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3607 struct iov_iter *from, int len); 3608 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3609 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3610 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3611 static inline void skb_free_datagram_locked(struct sock *sk, 3612 struct sk_buff *skb) 3613 { 3614 __skb_free_datagram_locked(sk, skb, 0); 3615 } 3616 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3617 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3618 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3619 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3620 int len); 3621 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3622 struct pipe_inode_info *pipe, unsigned int len, 3623 unsigned int flags); 3624 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3625 int len); 3626 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3627 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3628 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3629 int len, int hlen); 3630 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3631 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3632 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3633 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3634 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3635 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3636 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3637 unsigned int offset); 3638 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3639 int skb_ensure_writable(struct sk_buff *skb, int write_len); 3640 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3641 int skb_vlan_pop(struct sk_buff *skb); 3642 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3643 int skb_eth_pop(struct sk_buff *skb); 3644 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3645 const unsigned char *src); 3646 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3647 int mac_len, bool ethernet); 3648 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3649 bool ethernet); 3650 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3651 int skb_mpls_dec_ttl(struct sk_buff *skb); 3652 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3653 gfp_t gfp); 3654 3655 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3656 { 3657 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3658 } 3659 3660 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3661 { 3662 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3663 } 3664 3665 struct skb_checksum_ops { 3666 __wsum (*update)(const void *mem, int len, __wsum wsum); 3667 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3668 }; 3669 3670 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3671 3672 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3673 __wsum csum, const struct skb_checksum_ops *ops); 3674 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3675 __wsum csum); 3676 3677 static inline void * __must_check 3678 __skb_header_pointer(const struct sk_buff *skb, int offset, 3679 int len, void *data, int hlen, void *buffer) 3680 { 3681 if (hlen - offset >= len) 3682 return data + offset; 3683 3684 if (!skb || 3685 skb_copy_bits(skb, offset, buffer, len) < 0) 3686 return NULL; 3687 3688 return buffer; 3689 } 3690 3691 static inline void * __must_check 3692 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3693 { 3694 return __skb_header_pointer(skb, offset, len, skb->data, 3695 skb_headlen(skb), buffer); 3696 } 3697 3698 /** 3699 * skb_needs_linearize - check if we need to linearize a given skb 3700 * depending on the given device features. 3701 * @skb: socket buffer to check 3702 * @features: net device features 3703 * 3704 * Returns true if either: 3705 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3706 * 2. skb is fragmented and the device does not support SG. 3707 */ 3708 static inline bool skb_needs_linearize(struct sk_buff *skb, 3709 netdev_features_t features) 3710 { 3711 return skb_is_nonlinear(skb) && 3712 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3713 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3714 } 3715 3716 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3717 void *to, 3718 const unsigned int len) 3719 { 3720 memcpy(to, skb->data, len); 3721 } 3722 3723 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3724 const int offset, void *to, 3725 const unsigned int len) 3726 { 3727 memcpy(to, skb->data + offset, len); 3728 } 3729 3730 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3731 const void *from, 3732 const unsigned int len) 3733 { 3734 memcpy(skb->data, from, len); 3735 } 3736 3737 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3738 const int offset, 3739 const void *from, 3740 const unsigned int len) 3741 { 3742 memcpy(skb->data + offset, from, len); 3743 } 3744 3745 void skb_init(void); 3746 3747 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3748 { 3749 return skb->tstamp; 3750 } 3751 3752 /** 3753 * skb_get_timestamp - get timestamp from a skb 3754 * @skb: skb to get stamp from 3755 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3756 * 3757 * Timestamps are stored in the skb as offsets to a base timestamp. 3758 * This function converts the offset back to a struct timeval and stores 3759 * it in stamp. 3760 */ 3761 static inline void skb_get_timestamp(const struct sk_buff *skb, 3762 struct __kernel_old_timeval *stamp) 3763 { 3764 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3765 } 3766 3767 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3768 struct __kernel_sock_timeval *stamp) 3769 { 3770 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3771 3772 stamp->tv_sec = ts.tv_sec; 3773 stamp->tv_usec = ts.tv_nsec / 1000; 3774 } 3775 3776 static inline void skb_get_timestampns(const struct sk_buff *skb, 3777 struct __kernel_old_timespec *stamp) 3778 { 3779 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3780 3781 stamp->tv_sec = ts.tv_sec; 3782 stamp->tv_nsec = ts.tv_nsec; 3783 } 3784 3785 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3786 struct __kernel_timespec *stamp) 3787 { 3788 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3789 3790 stamp->tv_sec = ts.tv_sec; 3791 stamp->tv_nsec = ts.tv_nsec; 3792 } 3793 3794 static inline void __net_timestamp(struct sk_buff *skb) 3795 { 3796 skb->tstamp = ktime_get_real(); 3797 } 3798 3799 static inline ktime_t net_timedelta(ktime_t t) 3800 { 3801 return ktime_sub(ktime_get_real(), t); 3802 } 3803 3804 static inline ktime_t net_invalid_timestamp(void) 3805 { 3806 return 0; 3807 } 3808 3809 static inline u8 skb_metadata_len(const struct sk_buff *skb) 3810 { 3811 return skb_shinfo(skb)->meta_len; 3812 } 3813 3814 static inline void *skb_metadata_end(const struct sk_buff *skb) 3815 { 3816 return skb_mac_header(skb); 3817 } 3818 3819 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3820 const struct sk_buff *skb_b, 3821 u8 meta_len) 3822 { 3823 const void *a = skb_metadata_end(skb_a); 3824 const void *b = skb_metadata_end(skb_b); 3825 /* Using more efficient varaiant than plain call to memcmp(). */ 3826 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3827 u64 diffs = 0; 3828 3829 switch (meta_len) { 3830 #define __it(x, op) (x -= sizeof(u##op)) 3831 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3832 case 32: diffs |= __it_diff(a, b, 64); 3833 fallthrough; 3834 case 24: diffs |= __it_diff(a, b, 64); 3835 fallthrough; 3836 case 16: diffs |= __it_diff(a, b, 64); 3837 fallthrough; 3838 case 8: diffs |= __it_diff(a, b, 64); 3839 break; 3840 case 28: diffs |= __it_diff(a, b, 64); 3841 fallthrough; 3842 case 20: diffs |= __it_diff(a, b, 64); 3843 fallthrough; 3844 case 12: diffs |= __it_diff(a, b, 64); 3845 fallthrough; 3846 case 4: diffs |= __it_diff(a, b, 32); 3847 break; 3848 } 3849 return diffs; 3850 #else 3851 return memcmp(a - meta_len, b - meta_len, meta_len); 3852 #endif 3853 } 3854 3855 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3856 const struct sk_buff *skb_b) 3857 { 3858 u8 len_a = skb_metadata_len(skb_a); 3859 u8 len_b = skb_metadata_len(skb_b); 3860 3861 if (!(len_a | len_b)) 3862 return false; 3863 3864 return len_a != len_b ? 3865 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3866 } 3867 3868 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3869 { 3870 skb_shinfo(skb)->meta_len = meta_len; 3871 } 3872 3873 static inline void skb_metadata_clear(struct sk_buff *skb) 3874 { 3875 skb_metadata_set(skb, 0); 3876 } 3877 3878 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3879 3880 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3881 3882 void skb_clone_tx_timestamp(struct sk_buff *skb); 3883 bool skb_defer_rx_timestamp(struct sk_buff *skb); 3884 3885 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3886 3887 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3888 { 3889 } 3890 3891 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3892 { 3893 return false; 3894 } 3895 3896 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3897 3898 /** 3899 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3900 * 3901 * PHY drivers may accept clones of transmitted packets for 3902 * timestamping via their phy_driver.txtstamp method. These drivers 3903 * must call this function to return the skb back to the stack with a 3904 * timestamp. 3905 * 3906 * @skb: clone of the original outgoing packet 3907 * @hwtstamps: hardware time stamps 3908 * 3909 */ 3910 void skb_complete_tx_timestamp(struct sk_buff *skb, 3911 struct skb_shared_hwtstamps *hwtstamps); 3912 3913 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 3914 struct skb_shared_hwtstamps *hwtstamps, 3915 struct sock *sk, int tstype); 3916 3917 /** 3918 * skb_tstamp_tx - queue clone of skb with send time stamps 3919 * @orig_skb: the original outgoing packet 3920 * @hwtstamps: hardware time stamps, may be NULL if not available 3921 * 3922 * If the skb has a socket associated, then this function clones the 3923 * skb (thus sharing the actual data and optional structures), stores 3924 * the optional hardware time stamping information (if non NULL) or 3925 * generates a software time stamp (otherwise), then queues the clone 3926 * to the error queue of the socket. Errors are silently ignored. 3927 */ 3928 void skb_tstamp_tx(struct sk_buff *orig_skb, 3929 struct skb_shared_hwtstamps *hwtstamps); 3930 3931 /** 3932 * skb_tx_timestamp() - Driver hook for transmit timestamping 3933 * 3934 * Ethernet MAC Drivers should call this function in their hard_xmit() 3935 * function immediately before giving the sk_buff to the MAC hardware. 3936 * 3937 * Specifically, one should make absolutely sure that this function is 3938 * called before TX completion of this packet can trigger. Otherwise 3939 * the packet could potentially already be freed. 3940 * 3941 * @skb: A socket buffer. 3942 */ 3943 static inline void skb_tx_timestamp(struct sk_buff *skb) 3944 { 3945 skb_clone_tx_timestamp(skb); 3946 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3947 skb_tstamp_tx(skb, NULL); 3948 } 3949 3950 /** 3951 * skb_complete_wifi_ack - deliver skb with wifi status 3952 * 3953 * @skb: the original outgoing packet 3954 * @acked: ack status 3955 * 3956 */ 3957 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3958 3959 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3960 __sum16 __skb_checksum_complete(struct sk_buff *skb); 3961 3962 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3963 { 3964 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3965 skb->csum_valid || 3966 (skb->ip_summed == CHECKSUM_PARTIAL && 3967 skb_checksum_start_offset(skb) >= 0)); 3968 } 3969 3970 /** 3971 * skb_checksum_complete - Calculate checksum of an entire packet 3972 * @skb: packet to process 3973 * 3974 * This function calculates the checksum over the entire packet plus 3975 * the value of skb->csum. The latter can be used to supply the 3976 * checksum of a pseudo header as used by TCP/UDP. It returns the 3977 * checksum. 3978 * 3979 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 3980 * this function can be used to verify that checksum on received 3981 * packets. In that case the function should return zero if the 3982 * checksum is correct. In particular, this function will return zero 3983 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 3984 * hardware has already verified the correctness of the checksum. 3985 */ 3986 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 3987 { 3988 return skb_csum_unnecessary(skb) ? 3989 0 : __skb_checksum_complete(skb); 3990 } 3991 3992 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 3993 { 3994 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3995 if (skb->csum_level == 0) 3996 skb->ip_summed = CHECKSUM_NONE; 3997 else 3998 skb->csum_level--; 3999 } 4000 } 4001 4002 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4003 { 4004 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4005 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4006 skb->csum_level++; 4007 } else if (skb->ip_summed == CHECKSUM_NONE) { 4008 skb->ip_summed = CHECKSUM_UNNECESSARY; 4009 skb->csum_level = 0; 4010 } 4011 } 4012 4013 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4014 { 4015 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4016 skb->ip_summed = CHECKSUM_NONE; 4017 skb->csum_level = 0; 4018 } 4019 } 4020 4021 /* Check if we need to perform checksum complete validation. 4022 * 4023 * Returns true if checksum complete is needed, false otherwise 4024 * (either checksum is unnecessary or zero checksum is allowed). 4025 */ 4026 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4027 bool zero_okay, 4028 __sum16 check) 4029 { 4030 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4031 skb->csum_valid = 1; 4032 __skb_decr_checksum_unnecessary(skb); 4033 return false; 4034 } 4035 4036 return true; 4037 } 4038 4039 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4040 * in checksum_init. 4041 */ 4042 #define CHECKSUM_BREAK 76 4043 4044 /* Unset checksum-complete 4045 * 4046 * Unset checksum complete can be done when packet is being modified 4047 * (uncompressed for instance) and checksum-complete value is 4048 * invalidated. 4049 */ 4050 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4051 { 4052 if (skb->ip_summed == CHECKSUM_COMPLETE) 4053 skb->ip_summed = CHECKSUM_NONE; 4054 } 4055 4056 /* Validate (init) checksum based on checksum complete. 4057 * 4058 * Return values: 4059 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4060 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4061 * checksum is stored in skb->csum for use in __skb_checksum_complete 4062 * non-zero: value of invalid checksum 4063 * 4064 */ 4065 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4066 bool complete, 4067 __wsum psum) 4068 { 4069 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4070 if (!csum_fold(csum_add(psum, skb->csum))) { 4071 skb->csum_valid = 1; 4072 return 0; 4073 } 4074 } 4075 4076 skb->csum = psum; 4077 4078 if (complete || skb->len <= CHECKSUM_BREAK) { 4079 __sum16 csum; 4080 4081 csum = __skb_checksum_complete(skb); 4082 skb->csum_valid = !csum; 4083 return csum; 4084 } 4085 4086 return 0; 4087 } 4088 4089 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4090 { 4091 return 0; 4092 } 4093 4094 /* Perform checksum validate (init). Note that this is a macro since we only 4095 * want to calculate the pseudo header which is an input function if necessary. 4096 * First we try to validate without any computation (checksum unnecessary) and 4097 * then calculate based on checksum complete calling the function to compute 4098 * pseudo header. 4099 * 4100 * Return values: 4101 * 0: checksum is validated or try to in skb_checksum_complete 4102 * non-zero: value of invalid checksum 4103 */ 4104 #define __skb_checksum_validate(skb, proto, complete, \ 4105 zero_okay, check, compute_pseudo) \ 4106 ({ \ 4107 __sum16 __ret = 0; \ 4108 skb->csum_valid = 0; \ 4109 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4110 __ret = __skb_checksum_validate_complete(skb, \ 4111 complete, compute_pseudo(skb, proto)); \ 4112 __ret; \ 4113 }) 4114 4115 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4116 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4117 4118 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4119 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4120 4121 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4122 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4123 4124 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4125 compute_pseudo) \ 4126 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4127 4128 #define skb_checksum_simple_validate(skb) \ 4129 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4130 4131 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4132 { 4133 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4134 } 4135 4136 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4137 { 4138 skb->csum = ~pseudo; 4139 skb->ip_summed = CHECKSUM_COMPLETE; 4140 } 4141 4142 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4143 do { \ 4144 if (__skb_checksum_convert_check(skb)) \ 4145 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4146 } while (0) 4147 4148 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4149 u16 start, u16 offset) 4150 { 4151 skb->ip_summed = CHECKSUM_PARTIAL; 4152 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4153 skb->csum_offset = offset - start; 4154 } 4155 4156 /* Update skbuf and packet to reflect the remote checksum offload operation. 4157 * When called, ptr indicates the starting point for skb->csum when 4158 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4159 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4160 */ 4161 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4162 int start, int offset, bool nopartial) 4163 { 4164 __wsum delta; 4165 4166 if (!nopartial) { 4167 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4168 return; 4169 } 4170 4171 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4172 __skb_checksum_complete(skb); 4173 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4174 } 4175 4176 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4177 4178 /* Adjust skb->csum since we changed the packet */ 4179 skb->csum = csum_add(skb->csum, delta); 4180 } 4181 4182 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4183 { 4184 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4185 return (void *)(skb->_nfct & NFCT_PTRMASK); 4186 #else 4187 return NULL; 4188 #endif 4189 } 4190 4191 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4192 { 4193 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4194 return skb->_nfct; 4195 #else 4196 return 0UL; 4197 #endif 4198 } 4199 4200 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4201 { 4202 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4203 skb->_nfct = nfct; 4204 #endif 4205 } 4206 4207 #ifdef CONFIG_SKB_EXTENSIONS 4208 enum skb_ext_id { 4209 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4210 SKB_EXT_BRIDGE_NF, 4211 #endif 4212 #ifdef CONFIG_XFRM 4213 SKB_EXT_SEC_PATH, 4214 #endif 4215 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4216 TC_SKB_EXT, 4217 #endif 4218 #if IS_ENABLED(CONFIG_MPTCP) 4219 SKB_EXT_MPTCP, 4220 #endif 4221 SKB_EXT_NUM, /* must be last */ 4222 }; 4223 4224 /** 4225 * struct skb_ext - sk_buff extensions 4226 * @refcnt: 1 on allocation, deallocated on 0 4227 * @offset: offset to add to @data to obtain extension address 4228 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4229 * @data: start of extension data, variable sized 4230 * 4231 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4232 * to use 'u8' types while allowing up to 2kb worth of extension data. 4233 */ 4234 struct skb_ext { 4235 refcount_t refcnt; 4236 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4237 u8 chunks; /* same */ 4238 char data[] __aligned(8); 4239 }; 4240 4241 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4242 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4243 struct skb_ext *ext); 4244 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4245 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4246 void __skb_ext_put(struct skb_ext *ext); 4247 4248 static inline void skb_ext_put(struct sk_buff *skb) 4249 { 4250 if (skb->active_extensions) 4251 __skb_ext_put(skb->extensions); 4252 } 4253 4254 static inline void __skb_ext_copy(struct sk_buff *dst, 4255 const struct sk_buff *src) 4256 { 4257 dst->active_extensions = src->active_extensions; 4258 4259 if (src->active_extensions) { 4260 struct skb_ext *ext = src->extensions; 4261 4262 refcount_inc(&ext->refcnt); 4263 dst->extensions = ext; 4264 } 4265 } 4266 4267 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4268 { 4269 skb_ext_put(dst); 4270 __skb_ext_copy(dst, src); 4271 } 4272 4273 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4274 { 4275 return !!ext->offset[i]; 4276 } 4277 4278 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4279 { 4280 return skb->active_extensions & (1 << id); 4281 } 4282 4283 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4284 { 4285 if (skb_ext_exist(skb, id)) 4286 __skb_ext_del(skb, id); 4287 } 4288 4289 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4290 { 4291 if (skb_ext_exist(skb, id)) { 4292 struct skb_ext *ext = skb->extensions; 4293 4294 return (void *)ext + (ext->offset[id] << 3); 4295 } 4296 4297 return NULL; 4298 } 4299 4300 static inline void skb_ext_reset(struct sk_buff *skb) 4301 { 4302 if (unlikely(skb->active_extensions)) { 4303 __skb_ext_put(skb->extensions); 4304 skb->active_extensions = 0; 4305 } 4306 } 4307 4308 static inline bool skb_has_extensions(struct sk_buff *skb) 4309 { 4310 return unlikely(skb->active_extensions); 4311 } 4312 #else 4313 static inline void skb_ext_put(struct sk_buff *skb) {} 4314 static inline void skb_ext_reset(struct sk_buff *skb) {} 4315 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4316 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4317 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4318 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4319 #endif /* CONFIG_SKB_EXTENSIONS */ 4320 4321 static inline void nf_reset_ct(struct sk_buff *skb) 4322 { 4323 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4324 nf_conntrack_put(skb_nfct(skb)); 4325 skb->_nfct = 0; 4326 #endif 4327 } 4328 4329 static inline void nf_reset_trace(struct sk_buff *skb) 4330 { 4331 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4332 skb->nf_trace = 0; 4333 #endif 4334 } 4335 4336 static inline void ipvs_reset(struct sk_buff *skb) 4337 { 4338 #if IS_ENABLED(CONFIG_IP_VS) 4339 skb->ipvs_property = 0; 4340 #endif 4341 } 4342 4343 /* Note: This doesn't put any conntrack info in dst. */ 4344 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4345 bool copy) 4346 { 4347 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4348 dst->_nfct = src->_nfct; 4349 nf_conntrack_get(skb_nfct(src)); 4350 #endif 4351 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4352 if (copy) 4353 dst->nf_trace = src->nf_trace; 4354 #endif 4355 } 4356 4357 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4358 { 4359 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4360 nf_conntrack_put(skb_nfct(dst)); 4361 #endif 4362 __nf_copy(dst, src, true); 4363 } 4364 4365 #ifdef CONFIG_NETWORK_SECMARK 4366 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4367 { 4368 to->secmark = from->secmark; 4369 } 4370 4371 static inline void skb_init_secmark(struct sk_buff *skb) 4372 { 4373 skb->secmark = 0; 4374 } 4375 #else 4376 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4377 { } 4378 4379 static inline void skb_init_secmark(struct sk_buff *skb) 4380 { } 4381 #endif 4382 4383 static inline int secpath_exists(const struct sk_buff *skb) 4384 { 4385 #ifdef CONFIG_XFRM 4386 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4387 #else 4388 return 0; 4389 #endif 4390 } 4391 4392 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4393 { 4394 return !skb->destructor && 4395 !secpath_exists(skb) && 4396 !skb_nfct(skb) && 4397 !skb->_skb_refdst && 4398 !skb_has_frag_list(skb); 4399 } 4400 4401 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4402 { 4403 skb->queue_mapping = queue_mapping; 4404 } 4405 4406 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4407 { 4408 return skb->queue_mapping; 4409 } 4410 4411 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4412 { 4413 to->queue_mapping = from->queue_mapping; 4414 } 4415 4416 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4417 { 4418 skb->queue_mapping = rx_queue + 1; 4419 } 4420 4421 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4422 { 4423 return skb->queue_mapping - 1; 4424 } 4425 4426 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4427 { 4428 return skb->queue_mapping != 0; 4429 } 4430 4431 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4432 { 4433 skb->dst_pending_confirm = val; 4434 } 4435 4436 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4437 { 4438 return skb->dst_pending_confirm != 0; 4439 } 4440 4441 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4442 { 4443 #ifdef CONFIG_XFRM 4444 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4445 #else 4446 return NULL; 4447 #endif 4448 } 4449 4450 /* Keeps track of mac header offset relative to skb->head. 4451 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4452 * For non-tunnel skb it points to skb_mac_header() and for 4453 * tunnel skb it points to outer mac header. 4454 * Keeps track of level of encapsulation of network headers. 4455 */ 4456 struct skb_gso_cb { 4457 union { 4458 int mac_offset; 4459 int data_offset; 4460 }; 4461 int encap_level; 4462 __wsum csum; 4463 __u16 csum_start; 4464 }; 4465 #define SKB_GSO_CB_OFFSET 32 4466 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4467 4468 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4469 { 4470 return (skb_mac_header(inner_skb) - inner_skb->head) - 4471 SKB_GSO_CB(inner_skb)->mac_offset; 4472 } 4473 4474 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4475 { 4476 int new_headroom, headroom; 4477 int ret; 4478 4479 headroom = skb_headroom(skb); 4480 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4481 if (ret) 4482 return ret; 4483 4484 new_headroom = skb_headroom(skb); 4485 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4486 return 0; 4487 } 4488 4489 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4490 { 4491 /* Do not update partial checksums if remote checksum is enabled. */ 4492 if (skb->remcsum_offload) 4493 return; 4494 4495 SKB_GSO_CB(skb)->csum = res; 4496 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4497 } 4498 4499 /* Compute the checksum for a gso segment. First compute the checksum value 4500 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4501 * then add in skb->csum (checksum from csum_start to end of packet). 4502 * skb->csum and csum_start are then updated to reflect the checksum of the 4503 * resultant packet starting from the transport header-- the resultant checksum 4504 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4505 * header. 4506 */ 4507 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4508 { 4509 unsigned char *csum_start = skb_transport_header(skb); 4510 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4511 __wsum partial = SKB_GSO_CB(skb)->csum; 4512 4513 SKB_GSO_CB(skb)->csum = res; 4514 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4515 4516 return csum_fold(csum_partial(csum_start, plen, partial)); 4517 } 4518 4519 static inline bool skb_is_gso(const struct sk_buff *skb) 4520 { 4521 return skb_shinfo(skb)->gso_size; 4522 } 4523 4524 /* Note: Should be called only if skb_is_gso(skb) is true */ 4525 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4526 { 4527 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4528 } 4529 4530 /* Note: Should be called only if skb_is_gso(skb) is true */ 4531 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4532 { 4533 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4534 } 4535 4536 /* Note: Should be called only if skb_is_gso(skb) is true */ 4537 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4538 { 4539 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4540 } 4541 4542 static inline void skb_gso_reset(struct sk_buff *skb) 4543 { 4544 skb_shinfo(skb)->gso_size = 0; 4545 skb_shinfo(skb)->gso_segs = 0; 4546 skb_shinfo(skb)->gso_type = 0; 4547 } 4548 4549 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4550 u16 increment) 4551 { 4552 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4553 return; 4554 shinfo->gso_size += increment; 4555 } 4556 4557 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4558 u16 decrement) 4559 { 4560 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4561 return; 4562 shinfo->gso_size -= decrement; 4563 } 4564 4565 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4566 4567 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4568 { 4569 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4570 * wanted then gso_type will be set. */ 4571 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4572 4573 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4574 unlikely(shinfo->gso_type == 0)) { 4575 __skb_warn_lro_forwarding(skb); 4576 return true; 4577 } 4578 return false; 4579 } 4580 4581 static inline void skb_forward_csum(struct sk_buff *skb) 4582 { 4583 /* Unfortunately we don't support this one. Any brave souls? */ 4584 if (skb->ip_summed == CHECKSUM_COMPLETE) 4585 skb->ip_summed = CHECKSUM_NONE; 4586 } 4587 4588 /** 4589 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4590 * @skb: skb to check 4591 * 4592 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4593 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4594 * use this helper, to document places where we make this assertion. 4595 */ 4596 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4597 { 4598 #ifdef DEBUG 4599 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4600 #endif 4601 } 4602 4603 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4604 4605 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4606 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4607 unsigned int transport_len, 4608 __sum16(*skb_chkf)(struct sk_buff *skb)); 4609 4610 /** 4611 * skb_head_is_locked - Determine if the skb->head is locked down 4612 * @skb: skb to check 4613 * 4614 * The head on skbs build around a head frag can be removed if they are 4615 * not cloned. This function returns true if the skb head is locked down 4616 * due to either being allocated via kmalloc, or by being a clone with 4617 * multiple references to the head. 4618 */ 4619 static inline bool skb_head_is_locked(const struct sk_buff *skb) 4620 { 4621 return !skb->head_frag || skb_cloned(skb); 4622 } 4623 4624 /* Local Checksum Offload. 4625 * Compute outer checksum based on the assumption that the 4626 * inner checksum will be offloaded later. 4627 * See Documentation/networking/checksum-offloads.rst for 4628 * explanation of how this works. 4629 * Fill in outer checksum adjustment (e.g. with sum of outer 4630 * pseudo-header) before calling. 4631 * Also ensure that inner checksum is in linear data area. 4632 */ 4633 static inline __wsum lco_csum(struct sk_buff *skb) 4634 { 4635 unsigned char *csum_start = skb_checksum_start(skb); 4636 unsigned char *l4_hdr = skb_transport_header(skb); 4637 __wsum partial; 4638 4639 /* Start with complement of inner checksum adjustment */ 4640 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4641 skb->csum_offset)); 4642 4643 /* Add in checksum of our headers (incl. outer checksum 4644 * adjustment filled in by caller) and return result. 4645 */ 4646 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4647 } 4648 4649 static inline bool skb_is_redirected(const struct sk_buff *skb) 4650 { 4651 #ifdef CONFIG_NET_REDIRECT 4652 return skb->redirected; 4653 #else 4654 return false; 4655 #endif 4656 } 4657 4658 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4659 { 4660 #ifdef CONFIG_NET_REDIRECT 4661 skb->redirected = 1; 4662 skb->from_ingress = from_ingress; 4663 if (skb->from_ingress) 4664 skb->tstamp = 0; 4665 #endif 4666 } 4667 4668 static inline void skb_reset_redirect(struct sk_buff *skb) 4669 { 4670 #ifdef CONFIG_NET_REDIRECT 4671 skb->redirected = 0; 4672 #endif 4673 } 4674 4675 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 4676 { 4677 return skb->csum_not_inet; 4678 } 4679 4680 static inline void skb_set_kcov_handle(struct sk_buff *skb, 4681 const u64 kcov_handle) 4682 { 4683 #ifdef CONFIG_KCOV 4684 skb->kcov_handle = kcov_handle; 4685 #endif 4686 } 4687 4688 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4689 { 4690 #ifdef CONFIG_KCOV 4691 return skb->kcov_handle; 4692 #else 4693 return 0; 4694 #endif 4695 } 4696 4697 #endif /* __KERNEL__ */ 4698 #endif /* _LINUX_SKBUFF_H */ 4699
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.