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