1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Definitions for the 'struct sk_buff' memory handlers. 4 * 5 * Authors: 6 * Alan Cox, <gw4pts@gw4pts.ampr.org> 7 * Florian La Roche, <rzsfl@rz.uni-sb.de> 8 */ 9 10 #ifndef _LINUX_SKBUFF_H 11 #define _LINUX_SKBUFF_H 12 13 #include <linux/kernel.h> 14 #include <linux/compiler.h> 15 #include <linux/time.h> 16 #include <linux/bug.h> 17 #include <linux/bvec.h> 18 #include <linux/cache.h> 19 #include <linux/rbtree.h> 20 #include <linux/socket.h> 21 #include <linux/refcount.h> 22 23 #include <linux/atomic.h> 24 #include <asm/types.h> 25 #include <linux/spinlock.h> 26 #include <linux/net.h> 27 #include <linux/textsearch.h> 28 #include <net/checksum.h> 29 #include <linux/rcupdate.h> 30 #include <linux/hrtimer.h> 31 #include <linux/dma-mapping.h> 32 #include <linux/netdev_features.h> 33 #include <linux/sched.h> 34 #include <linux/sched/clock.h> 35 #include <net/flow_dissector.h> 36 #include <linux/splice.h> 37 #include <linux/in6.h> 38 #include <linux/if_packet.h> 39 #include <net/flow.h> 40 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 41 #include <linux/netfilter/nf_conntrack_common.h> 42 #endif 43 44 /* The interface for checksum offload between the stack and networking drivers 45 * is as follows... 46 * 47 * A. IP checksum related features 48 * 49 * Drivers advertise checksum offload capabilities in the features of a device. 50 * From the stack's point of view these are capabilities offered by the driver. 51 * A driver typically only advertises features that it is capable of offloading 52 * to its device. 53 * 54 * The checksum related features are: 55 * 56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one 57 * IP (one's complement) checksum for any combination 58 * of protocols or protocol layering. The checksum is 59 * computed and set in a packet per the CHECKSUM_PARTIAL 60 * interface (see below). 61 * 62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain 63 * TCP or UDP packets over IPv4. These are specifically 64 * unencapsulated packets of the form IPv4|TCP or 65 * IPv4|UDP where the Protocol field in the IPv4 header 66 * is TCP or UDP. The IPv4 header may contain IP options. 67 * This feature cannot be set in features for a device 68 * with NETIF_F_HW_CSUM also set. This feature is being 69 * DEPRECATED (see below). 70 * 71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain 72 * TCP or UDP packets over IPv6. These are specifically 73 * unencapsulated packets of the form IPv6|TCP or 74 * IPv6|UDP where the Next Header field in the IPv6 75 * header is either TCP or UDP. IPv6 extension headers 76 * are not supported with this feature. This feature 77 * cannot be set in features for a device with 78 * NETIF_F_HW_CSUM also set. This feature is being 79 * DEPRECATED (see below). 80 * 81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. 82 * This flag is only used to disable the RX checksum 83 * feature for a device. The stack will accept receive 84 * checksum indication in packets received on a device 85 * regardless of whether NETIF_F_RXCSUM is set. 86 * 87 * B. Checksumming of received packets by device. Indication of checksum 88 * verification is set in skb->ip_summed. Possible values are: 89 * 90 * CHECKSUM_NONE: 91 * 92 * Device did not checksum this packet e.g. due to lack of capabilities. 93 * The packet contains full (though not verified) checksum in packet but 94 * not in skb->csum. Thus, skb->csum is undefined in this case. 95 * 96 * CHECKSUM_UNNECESSARY: 97 * 98 * The hardware you're dealing with doesn't calculate the full checksum 99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 101 * if their checksums are okay. skb->csum is still undefined in this case 102 * though. A driver or device must never modify the checksum field in the 103 * packet even if checksum is verified. 104 * 105 * CHECKSUM_UNNECESSARY is applicable to following protocols: 106 * TCP: IPv6 and IPv4. 107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 108 * zero UDP checksum for either IPv4 or IPv6, the networking stack 109 * may perform further validation in this case. 110 * GRE: only if the checksum is present in the header. 111 * SCTP: indicates the CRC in SCTP header has been validated. 112 * FCOE: indicates the CRC in FC frame has been validated. 113 * 114 * skb->csum_level indicates the number of consecutive checksums found in 115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 117 * and a device is able to verify the checksums for UDP (possibly zero), 118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to 119 * two. If the device were only able to verify the UDP checksum and not 120 * GRE, either because it doesn't support GRE checksum or because GRE 121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 122 * not considered in this case). 123 * 124 * CHECKSUM_COMPLETE: 125 * 126 * This is the most generic way. The device supplied checksum of the _whole_ 127 * packet as seen by netif_rx() and fills in skb->csum. This means the 128 * hardware doesn't need to parse L3/L4 headers to implement this. 129 * 130 * Notes: 131 * - Even if device supports only some protocols, but is able to produce 132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 134 * 135 * CHECKSUM_PARTIAL: 136 * 137 * A checksum is set up to be offloaded to a device as described in the 138 * output description for CHECKSUM_PARTIAL. This may occur on a packet 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel 140 * on the same host, or it may be set in the input path in GRO or remote 141 * checksum offload. For the purposes of checksum verification, the checksum 142 * referred to by skb->csum_start + skb->csum_offset and any preceding 143 * checksums in the packet are considered verified. Any checksums in the 144 * packet that are after the checksum being offloaded are not considered to 145 * be verified. 146 * 147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload 148 * in the skb->ip_summed for a packet. Values are: 149 * 150 * CHECKSUM_PARTIAL: 151 * 152 * The driver is required to checksum the packet as seen by hard_start_xmit() 153 * from skb->csum_start up to the end, and to record/write the checksum at 154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the 155 * csum_start and csum_offset values are valid values given the length and 156 * offset of the packet, but it should not attempt to validate that the 157 * checksum refers to a legitimate transport layer checksum -- it is the 158 * purview of the stack to validate that csum_start and csum_offset are set 159 * correctly. 160 * 161 * When the stack requests checksum offload for a packet, the driver MUST 162 * ensure that the checksum is set correctly. A driver can either offload the 163 * checksum calculation to the device, or call skb_checksum_help (in the case 164 * that the device does not support offload for a particular checksum). 165 * 166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of 167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate 168 * checksum offload capability. 169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based 170 * on network device checksumming capabilities: if a packet does not match 171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of 172 * csum_not_inet, see item D.) is called to resolve the checksum. 173 * 174 * CHECKSUM_NONE: 175 * 176 * The skb was already checksummed by the protocol, or a checksum is not 177 * required. 178 * 179 * CHECKSUM_UNNECESSARY: 180 * 181 * This has the same meaning as CHECKSUM_NONE for checksum offload on 182 * output. 183 * 184 * CHECKSUM_COMPLETE: 185 * Not used in checksum output. If a driver observes a packet with this value 186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set. 187 * 188 * D. Non-IP checksum (CRC) offloads 189 * 190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of 191 * offloading the SCTP CRC in a packet. To perform this offload the stack 192 * will set csum_start and csum_offset accordingly, set ip_summed to 193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in 194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c. 195 * A driver that supports both IP checksum offload and SCTP CRC32c offload 196 * must verify which offload is configured for a packet by testing the 197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve 198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 199 * 200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of 201 * offloading the FCOE CRC in a packet. To perform this offload the stack 202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset 203 * accordingly. Note that there is no indication in the skbuff that the 204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 205 * both IP checksum offload and FCOE CRC offload must verify which offload 206 * is configured for a packet, presumably by inspecting packet headers. 207 * 208 * E. Checksumming on output with GSO. 209 * 210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload 211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as 213 * part of the GSO operation is implied. If a checksum is being offloaded 214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and 215 * csum_offset are set to refer to the outermost checksum being offloaded 216 * (two offloaded checksums are possible with UDP encapsulation). 217 */ 218 219 /* Don't change this without changing skb_csum_unnecessary! */ 220 #define CHECKSUM_NONE 0 221 #define CHECKSUM_UNNECESSARY 1 222 #define CHECKSUM_COMPLETE 2 223 #define CHECKSUM_PARTIAL 3 224 225 /* Maximum value in skb->csum_level */ 226 #define SKB_MAX_CSUM_LEVEL 3 227 228 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 229 #define SKB_WITH_OVERHEAD(X) \ 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 231 #define SKB_MAX_ORDER(X, ORDER) \ 232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 233 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 234 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 235 236 /* return minimum truesize of one skb containing X bytes of data */ 237 #define SKB_TRUESIZE(X) ((X) + \ 238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 240 241 struct ahash_request; 242 struct net_device; 243 struct scatterlist; 244 struct pipe_inode_info; 245 struct iov_iter; 246 struct napi_struct; 247 struct bpf_prog; 248 union bpf_attr; 249 struct skb_ext; 250 251 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 252 struct nf_bridge_info { 253 enum { 254 BRNF_PROTO_UNCHANGED, 255 BRNF_PROTO_8021Q, 256 BRNF_PROTO_PPPOE 257 } orig_proto:8; 258 u8 pkt_otherhost:1; 259 u8 in_prerouting:1; 260 u8 bridged_dnat:1; 261 __u16 frag_max_size; 262 struct net_device *physindev; 263 264 /* always valid & non-NULL from FORWARD on, for physdev match */ 265 struct net_device *physoutdev; 266 union { 267 /* prerouting: detect dnat in orig/reply direction */ 268 __be32 ipv4_daddr; 269 struct in6_addr ipv6_daddr; 270 271 /* after prerouting + nat detected: store original source 272 * mac since neigh resolution overwrites it, only used while 273 * skb is out in neigh layer. 274 */ 275 char neigh_header[8]; 276 }; 277 }; 278 #endif 279 280 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 281 /* Chain in tc_skb_ext will be used to share the tc chain with 282 * ovs recirc_id. It will be set to the current chain by tc 283 * and read by ovs to recirc_id. 284 */ 285 struct tc_skb_ext { 286 __u32 chain; 287 __u16 mru; 288 bool post_ct; 289 }; 290 #endif 291 292 struct sk_buff_head { 293 /* These two members must be first. */ 294 struct sk_buff *next; 295 struct sk_buff *prev; 296 297 __u32 qlen; 298 spinlock_t lock; 299 }; 300 301 struct sk_buff; 302 303 /* To allow 64K frame to be packed as single skb without frag_list we 304 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 305 * buffers which do not start on a page boundary. 306 * 307 * Since GRO uses frags we allocate at least 16 regardless of page 308 * size. 309 */ 310 #if (65536/PAGE_SIZE + 1) < 16 311 #define MAX_SKB_FRAGS 16UL 312 #else 313 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 314 #endif 315 extern int sysctl_max_skb_frags; 316 317 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 318 * segment using its current segmentation instead. 319 */ 320 #define GSO_BY_FRAGS 0xFFFF 321 322 typedef struct bio_vec skb_frag_t; 323 324 /** 325 * skb_frag_size() - Returns the size of a skb fragment 326 * @frag: skb fragment 327 */ 328 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 329 { 330 return frag->bv_len; 331 } 332 333 /** 334 * skb_frag_size_set() - Sets the size of a skb fragment 335 * @frag: skb fragment 336 * @size: size of fragment 337 */ 338 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 339 { 340 frag->bv_len = size; 341 } 342 343 /** 344 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 345 * @frag: skb fragment 346 * @delta: value to add 347 */ 348 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 349 { 350 frag->bv_len += delta; 351 } 352 353 /** 354 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 355 * @frag: skb fragment 356 * @delta: value to subtract 357 */ 358 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 359 { 360 frag->bv_len -= delta; 361 } 362 363 /** 364 * skb_frag_must_loop - Test if %p is a high memory page 365 * @p: fragment's page 366 */ 367 static inline bool skb_frag_must_loop(struct page *p) 368 { 369 #if defined(CONFIG_HIGHMEM) 370 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 371 return true; 372 #endif 373 return false; 374 } 375 376 /** 377 * skb_frag_foreach_page - loop over pages in a fragment 378 * 379 * @f: skb frag to operate on 380 * @f_off: offset from start of f->bv_page 381 * @f_len: length from f_off to loop over 382 * @p: (temp var) current page 383 * @p_off: (temp var) offset from start of current page, 384 * non-zero only on first page. 385 * @p_len: (temp var) length in current page, 386 * < PAGE_SIZE only on first and last page. 387 * @copied: (temp var) length so far, excluding current p_len. 388 * 389 * A fragment can hold a compound page, in which case per-page 390 * operations, notably kmap_atomic, must be called for each 391 * regular page. 392 */ 393 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 394 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 395 p_off = (f_off) & (PAGE_SIZE - 1), \ 396 p_len = skb_frag_must_loop(p) ? \ 397 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 398 copied = 0; \ 399 copied < f_len; \ 400 copied += p_len, p++, p_off = 0, \ 401 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 402 403 #define HAVE_HW_TIME_STAMP 404 405 /** 406 * struct skb_shared_hwtstamps - hardware time stamps 407 * @hwtstamp: hardware time stamp transformed into duration 408 * since arbitrary point in time 409 * 410 * Software time stamps generated by ktime_get_real() are stored in 411 * skb->tstamp. 412 * 413 * hwtstamps can only be compared against other hwtstamps from 414 * the same device. 415 * 416 * This structure is attached to packets as part of the 417 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 418 */ 419 struct skb_shared_hwtstamps { 420 ktime_t hwtstamp; 421 }; 422 423 /* Definitions for tx_flags in struct skb_shared_info */ 424 enum { 425 /* generate hardware time stamp */ 426 SKBTX_HW_TSTAMP = 1 << 0, 427 428 /* generate software time stamp when queueing packet to NIC */ 429 SKBTX_SW_TSTAMP = 1 << 1, 430 431 /* device driver is going to provide hardware time stamp */ 432 SKBTX_IN_PROGRESS = 1 << 2, 433 434 /* generate wifi status information (where possible) */ 435 SKBTX_WIFI_STATUS = 1 << 4, 436 437 /* generate software time stamp when entering packet scheduling */ 438 SKBTX_SCHED_TSTAMP = 1 << 6, 439 }; 440 441 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 442 SKBTX_SCHED_TSTAMP) 443 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 444 445 /* Definitions for flags in struct skb_shared_info */ 446 enum { 447 /* use zcopy routines */ 448 SKBFL_ZEROCOPY_ENABLE = BIT(0), 449 450 /* This indicates at least one fragment might be overwritten 451 * (as in vmsplice(), sendfile() ...) 452 * If we need to compute a TX checksum, we'll need to copy 453 * all frags to avoid possible bad checksum 454 */ 455 SKBFL_SHARED_FRAG = BIT(1), 456 }; 457 458 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 459 460 /* 461 * The callback notifies userspace to release buffers when skb DMA is done in 462 * lower device, the skb last reference should be 0 when calling this. 463 * The zerocopy_success argument is true if zero copy transmit occurred, 464 * false on data copy or out of memory error caused by data copy attempt. 465 * The ctx field is used to track device context. 466 * The desc field is used to track userspace buffer index. 467 */ 468 struct ubuf_info { 469 void (*callback)(struct sk_buff *, struct ubuf_info *, 470 bool zerocopy_success); 471 union { 472 struct { 473 unsigned long desc; 474 void *ctx; 475 }; 476 struct { 477 u32 id; 478 u16 len; 479 u16 zerocopy:1; 480 u32 bytelen; 481 }; 482 }; 483 refcount_t refcnt; 484 u8 flags; 485 486 struct mmpin { 487 struct user_struct *user; 488 unsigned int num_pg; 489 } mmp; 490 }; 491 492 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 493 494 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 495 void mm_unaccount_pinned_pages(struct mmpin *mmp); 496 497 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size); 498 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 499 struct ubuf_info *uarg); 500 501 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 502 503 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 504 bool success); 505 506 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len); 507 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 508 struct msghdr *msg, int len, 509 struct ubuf_info *uarg); 510 511 /* This data is invariant across clones and lives at 512 * the end of the header data, ie. at skb->end. 513 */ 514 struct skb_shared_info { 515 __u8 flags; 516 __u8 meta_len; 517 __u8 nr_frags; 518 __u8 tx_flags; 519 unsigned short gso_size; 520 /* Warning: this field is not always filled in (UFO)! */ 521 unsigned short gso_segs; 522 struct sk_buff *frag_list; 523 struct skb_shared_hwtstamps hwtstamps; 524 unsigned int gso_type; 525 u32 tskey; 526 527 /* 528 * Warning : all fields before dataref are cleared in __alloc_skb() 529 */ 530 atomic_t dataref; 531 532 /* Intermediate layers must ensure that destructor_arg 533 * remains valid until skb destructor */ 534 void * destructor_arg; 535 536 /* must be last field, see pskb_expand_head() */ 537 skb_frag_t frags[MAX_SKB_FRAGS]; 538 }; 539 540 /* We divide dataref into two halves. The higher 16 bits hold references 541 * to the payload part of skb->data. The lower 16 bits hold references to 542 * the entire skb->data. A clone of a headerless skb holds the length of 543 * the header in skb->hdr_len. 544 * 545 * All users must obey the rule that the skb->data reference count must be 546 * greater than or equal to the payload reference count. 547 * 548 * Holding a reference to the payload part means that the user does not 549 * care about modifications to the header part of skb->data. 550 */ 551 #define SKB_DATAREF_SHIFT 16 552 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 553 554 555 enum { 556 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 557 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 558 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 559 }; 560 561 enum { 562 SKB_GSO_TCPV4 = 1 << 0, 563 564 /* This indicates the skb is from an untrusted source. */ 565 SKB_GSO_DODGY = 1 << 1, 566 567 /* This indicates the tcp segment has CWR set. */ 568 SKB_GSO_TCP_ECN = 1 << 2, 569 570 SKB_GSO_TCP_FIXEDID = 1 << 3, 571 572 SKB_GSO_TCPV6 = 1 << 4, 573 574 SKB_GSO_FCOE = 1 << 5, 575 576 SKB_GSO_GRE = 1 << 6, 577 578 SKB_GSO_GRE_CSUM = 1 << 7, 579 580 SKB_GSO_IPXIP4 = 1 << 8, 581 582 SKB_GSO_IPXIP6 = 1 << 9, 583 584 SKB_GSO_UDP_TUNNEL = 1 << 10, 585 586 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 587 588 SKB_GSO_PARTIAL = 1 << 12, 589 590 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 591 592 SKB_GSO_SCTP = 1 << 14, 593 594 SKB_GSO_ESP = 1 << 15, 595 596 SKB_GSO_UDP = 1 << 16, 597 598 SKB_GSO_UDP_L4 = 1 << 17, 599 600 SKB_GSO_FRAGLIST = 1 << 18, 601 }; 602 603 #if BITS_PER_LONG > 32 604 #define NET_SKBUFF_DATA_USES_OFFSET 1 605 #endif 606 607 #ifdef NET_SKBUFF_DATA_USES_OFFSET 608 typedef unsigned int sk_buff_data_t; 609 #else 610 typedef unsigned char *sk_buff_data_t; 611 #endif 612 613 /** 614 * struct sk_buff - socket buffer 615 * @next: Next buffer in list 616 * @prev: Previous buffer in list 617 * @tstamp: Time we arrived/left 618 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 619 * for retransmit timer 620 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 621 * @list: queue head 622 * @sk: Socket we are owned by 623 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 624 * fragmentation management 625 * @dev: Device we arrived on/are leaving by 626 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 627 * @cb: Control buffer. Free for use by every layer. Put private vars here 628 * @_skb_refdst: destination entry (with norefcount bit) 629 * @sp: the security path, used for xfrm 630 * @len: Length of actual data 631 * @data_len: Data length 632 * @mac_len: Length of link layer header 633 * @hdr_len: writable header length of cloned skb 634 * @csum: Checksum (must include start/offset pair) 635 * @csum_start: Offset from skb->head where checksumming should start 636 * @csum_offset: Offset from csum_start where checksum should be stored 637 * @priority: Packet queueing priority 638 * @ignore_df: allow local fragmentation 639 * @cloned: Head may be cloned (check refcnt to be sure) 640 * @ip_summed: Driver fed us an IP checksum 641 * @nohdr: Payload reference only, must not modify header 642 * @pkt_type: Packet class 643 * @fclone: skbuff clone status 644 * @ipvs_property: skbuff is owned by ipvs 645 * @inner_protocol_type: whether the inner protocol is 646 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 647 * @remcsum_offload: remote checksum offload is enabled 648 * @offload_fwd_mark: Packet was L2-forwarded in hardware 649 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 650 * @tc_skip_classify: do not classify packet. set by IFB device 651 * @tc_at_ingress: used within tc_classify to distinguish in/egress 652 * @redirected: packet was redirected by packet classifier 653 * @from_ingress: packet was redirected from the ingress path 654 * @peeked: this packet has been seen already, so stats have been 655 * done for it, don't do them again 656 * @nf_trace: netfilter packet trace flag 657 * @protocol: Packet protocol from driver 658 * @destructor: Destruct function 659 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 660 * @_sk_redir: socket redirection information for skmsg 661 * @_nfct: Associated connection, if any (with nfctinfo bits) 662 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 663 * @skb_iif: ifindex of device we arrived on 664 * @tc_index: Traffic control index 665 * @hash: the packet hash 666 * @queue_mapping: Queue mapping for multiqueue devices 667 * @head_frag: skb was allocated from page fragments, 668 * not allocated by kmalloc() or vmalloc(). 669 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 670 * @active_extensions: active extensions (skb_ext_id types) 671 * @ndisc_nodetype: router type (from link layer) 672 * @ooo_okay: allow the mapping of a socket to a queue to be changed 673 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 674 * ports. 675 * @sw_hash: indicates hash was computed in software stack 676 * @wifi_acked_valid: wifi_acked was set 677 * @wifi_acked: whether frame was acked on wifi or not 678 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 679 * @encapsulation: indicates the inner headers in the skbuff are valid 680 * @encap_hdr_csum: software checksum is needed 681 * @csum_valid: checksum is already valid 682 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 683 * @csum_complete_sw: checksum was completed by software 684 * @csum_level: indicates the number of consecutive checksums found in 685 * the packet minus one that have been verified as 686 * CHECKSUM_UNNECESSARY (max 3) 687 * @dst_pending_confirm: need to confirm neighbour 688 * @decrypted: Decrypted SKB 689 * @napi_id: id of the NAPI struct this skb came from 690 * @sender_cpu: (aka @napi_id) source CPU in XPS 691 * @secmark: security marking 692 * @mark: Generic packet mark 693 * @reserved_tailroom: (aka @mark) number of bytes of free space available 694 * at the tail of an sk_buff 695 * @vlan_present: VLAN tag is present 696 * @vlan_proto: vlan encapsulation protocol 697 * @vlan_tci: vlan tag control information 698 * @inner_protocol: Protocol (encapsulation) 699 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 700 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 701 * @inner_transport_header: Inner transport layer header (encapsulation) 702 * @inner_network_header: Network layer header (encapsulation) 703 * @inner_mac_header: Link layer header (encapsulation) 704 * @transport_header: Transport layer header 705 * @network_header: Network layer header 706 * @mac_header: Link layer header 707 * @kcov_handle: KCOV remote handle for remote coverage collection 708 * @tail: Tail pointer 709 * @end: End pointer 710 * @head: Head of buffer 711 * @data: Data head pointer 712 * @truesize: Buffer size 713 * @users: User count - see {datagram,tcp}.c 714 * @extensions: allocated extensions, valid if active_extensions is nonzero 715 */ 716 717 struct sk_buff { 718 union { 719 struct { 720 /* These two members must be first. */ 721 struct sk_buff *next; 722 struct sk_buff *prev; 723 724 union { 725 struct net_device *dev; 726 /* Some protocols might use this space to store information, 727 * while device pointer would be NULL. 728 * UDP receive path is one user. 729 */ 730 unsigned long dev_scratch; 731 }; 732 }; 733 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 734 struct list_head list; 735 }; 736 737 union { 738 struct sock *sk; 739 int ip_defrag_offset; 740 }; 741 742 union { 743 ktime_t tstamp; 744 u64 skb_mstamp_ns; /* earliest departure time */ 745 }; 746 /* 747 * This is the control buffer. It is free to use for every 748 * layer. Please put your private variables there. If you 749 * want to keep them across layers you have to do a skb_clone() 750 * first. This is owned by whoever has the skb queued ATM. 751 */ 752 char cb[48] __aligned(8); 753 754 union { 755 struct { 756 unsigned long _skb_refdst; 757 void (*destructor)(struct sk_buff *skb); 758 }; 759 struct list_head tcp_tsorted_anchor; 760 #ifdef CONFIG_NET_SOCK_MSG 761 unsigned long _sk_redir; 762 #endif 763 }; 764 765 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 766 unsigned long _nfct; 767 #endif 768 unsigned int len, 769 data_len; 770 __u16 mac_len, 771 hdr_len; 772 773 /* Following fields are _not_ copied in __copy_skb_header() 774 * Note that queue_mapping is here mostly to fill a hole. 775 */ 776 __u16 queue_mapping; 777 778 /* if you move cloned around you also must adapt those constants */ 779 #ifdef __BIG_ENDIAN_BITFIELD 780 #define CLONED_MASK (1 << 7) 781 #else 782 #define CLONED_MASK 1 783 #endif 784 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset) 785 786 /* private: */ 787 __u8 __cloned_offset[0]; 788 /* public: */ 789 __u8 cloned:1, 790 nohdr:1, 791 fclone:2, 792 peeked:1, 793 head_frag:1, 794 pfmemalloc:1; 795 #ifdef CONFIG_SKB_EXTENSIONS 796 __u8 active_extensions; 797 #endif 798 /* fields enclosed in headers_start/headers_end are copied 799 * using a single memcpy() in __copy_skb_header() 800 */ 801 /* private: */ 802 __u32 headers_start[0]; 803 /* public: */ 804 805 /* if you move pkt_type around you also must adapt those constants */ 806 #ifdef __BIG_ENDIAN_BITFIELD 807 #define PKT_TYPE_MAX (7 << 5) 808 #else 809 #define PKT_TYPE_MAX 7 810 #endif 811 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 812 813 /* private: */ 814 __u8 __pkt_type_offset[0]; 815 /* public: */ 816 __u8 pkt_type:3; 817 __u8 ignore_df:1; 818 __u8 nf_trace:1; 819 __u8 ip_summed:2; 820 __u8 ooo_okay:1; 821 822 __u8 l4_hash:1; 823 __u8 sw_hash:1; 824 __u8 wifi_acked_valid:1; 825 __u8 wifi_acked:1; 826 __u8 no_fcs:1; 827 /* Indicates the inner headers are valid in the skbuff. */ 828 __u8 encapsulation:1; 829 __u8 encap_hdr_csum:1; 830 __u8 csum_valid:1; 831 832 #ifdef __BIG_ENDIAN_BITFIELD 833 #define PKT_VLAN_PRESENT_BIT 7 834 #else 835 #define PKT_VLAN_PRESENT_BIT 0 836 #endif 837 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset) 838 /* private: */ 839 __u8 __pkt_vlan_present_offset[0]; 840 /* public: */ 841 __u8 vlan_present:1; 842 __u8 csum_complete_sw:1; 843 __u8 csum_level:2; 844 __u8 csum_not_inet:1; 845 __u8 dst_pending_confirm:1; 846 #ifdef CONFIG_IPV6_NDISC_NODETYPE 847 __u8 ndisc_nodetype:2; 848 #endif 849 850 __u8 ipvs_property:1; 851 __u8 inner_protocol_type:1; 852 __u8 remcsum_offload:1; 853 #ifdef CONFIG_NET_SWITCHDEV 854 __u8 offload_fwd_mark:1; 855 __u8 offload_l3_fwd_mark:1; 856 #endif 857 #ifdef CONFIG_NET_CLS_ACT 858 __u8 tc_skip_classify:1; 859 __u8 tc_at_ingress:1; 860 #endif 861 #ifdef CONFIG_NET_REDIRECT 862 __u8 redirected:1; 863 __u8 from_ingress:1; 864 #endif 865 #ifdef CONFIG_TLS_DEVICE 866 __u8 decrypted:1; 867 #endif 868 869 #ifdef CONFIG_NET_SCHED 870 __u16 tc_index; /* traffic control index */ 871 #endif 872 873 union { 874 __wsum csum; 875 struct { 876 __u16 csum_start; 877 __u16 csum_offset; 878 }; 879 }; 880 __u32 priority; 881 int skb_iif; 882 __u32 hash; 883 __be16 vlan_proto; 884 __u16 vlan_tci; 885 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 886 union { 887 unsigned int napi_id; 888 unsigned int sender_cpu; 889 }; 890 #endif 891 #ifdef CONFIG_NETWORK_SECMARK 892 __u32 secmark; 893 #endif 894 895 union { 896 __u32 mark; 897 __u32 reserved_tailroom; 898 }; 899 900 union { 901 __be16 inner_protocol; 902 __u8 inner_ipproto; 903 }; 904 905 __u16 inner_transport_header; 906 __u16 inner_network_header; 907 __u16 inner_mac_header; 908 909 __be16 protocol; 910 __u16 transport_header; 911 __u16 network_header; 912 __u16 mac_header; 913 914 #ifdef CONFIG_KCOV 915 u64 kcov_handle; 916 #endif 917 918 /* private: */ 919 __u32 headers_end[0]; 920 /* public: */ 921 922 /* These elements must be at the end, see alloc_skb() for details. */ 923 sk_buff_data_t tail; 924 sk_buff_data_t end; 925 unsigned char *head, 926 *data; 927 unsigned int truesize; 928 refcount_t users; 929 930 #ifdef CONFIG_SKB_EXTENSIONS 931 /* only useable after checking ->active_extensions != 0 */ 932 struct skb_ext *extensions; 933 #endif 934 }; 935 936 #ifdef __KERNEL__ 937 /* 938 * Handling routines are only of interest to the kernel 939 */ 940 941 #define SKB_ALLOC_FCLONE 0x01 942 #define SKB_ALLOC_RX 0x02 943 #define SKB_ALLOC_NAPI 0x04 944 945 /** 946 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 947 * @skb: buffer 948 */ 949 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 950 { 951 return unlikely(skb->pfmemalloc); 952 } 953 954 /* 955 * skb might have a dst pointer attached, refcounted or not. 956 * _skb_refdst low order bit is set if refcount was _not_ taken 957 */ 958 #define SKB_DST_NOREF 1UL 959 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 960 961 /** 962 * skb_dst - returns skb dst_entry 963 * @skb: buffer 964 * 965 * Returns skb dst_entry, regardless of reference taken or not. 966 */ 967 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 968 { 969 /* If refdst was not refcounted, check we still are in a 970 * rcu_read_lock section 971 */ 972 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 973 !rcu_read_lock_held() && 974 !rcu_read_lock_bh_held()); 975 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 976 } 977 978 /** 979 * skb_dst_set - sets skb dst 980 * @skb: buffer 981 * @dst: dst entry 982 * 983 * Sets skb dst, assuming a reference was taken on dst and should 984 * be released by skb_dst_drop() 985 */ 986 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 987 { 988 skb->_skb_refdst = (unsigned long)dst; 989 } 990 991 /** 992 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 993 * @skb: buffer 994 * @dst: dst entry 995 * 996 * Sets skb dst, assuming a reference was not taken on dst. 997 * If dst entry is cached, we do not take reference and dst_release 998 * will be avoided by refdst_drop. If dst entry is not cached, we take 999 * reference, so that last dst_release can destroy the dst immediately. 1000 */ 1001 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 1002 { 1003 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1004 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1005 } 1006 1007 /** 1008 * skb_dst_is_noref - Test if skb dst isn't refcounted 1009 * @skb: buffer 1010 */ 1011 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1012 { 1013 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1014 } 1015 1016 /** 1017 * skb_rtable - Returns the skb &rtable 1018 * @skb: buffer 1019 */ 1020 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1021 { 1022 return (struct rtable *)skb_dst(skb); 1023 } 1024 1025 /* For mangling skb->pkt_type from user space side from applications 1026 * such as nft, tc, etc, we only allow a conservative subset of 1027 * possible pkt_types to be set. 1028 */ 1029 static inline bool skb_pkt_type_ok(u32 ptype) 1030 { 1031 return ptype <= PACKET_OTHERHOST; 1032 } 1033 1034 /** 1035 * skb_napi_id - Returns the skb's NAPI id 1036 * @skb: buffer 1037 */ 1038 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1039 { 1040 #ifdef CONFIG_NET_RX_BUSY_POLL 1041 return skb->napi_id; 1042 #else 1043 return 0; 1044 #endif 1045 } 1046 1047 /** 1048 * skb_unref - decrement the skb's reference count 1049 * @skb: buffer 1050 * 1051 * Returns true if we can free the skb. 1052 */ 1053 static inline bool skb_unref(struct sk_buff *skb) 1054 { 1055 if (unlikely(!skb)) 1056 return false; 1057 if (likely(refcount_read(&skb->users) == 1)) 1058 smp_rmb(); 1059 else if (likely(!refcount_dec_and_test(&skb->users))) 1060 return false; 1061 1062 return true; 1063 } 1064 1065 void skb_release_head_state(struct sk_buff *skb); 1066 void kfree_skb(struct sk_buff *skb); 1067 void kfree_skb_list(struct sk_buff *segs); 1068 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1069 void skb_tx_error(struct sk_buff *skb); 1070 1071 #ifdef CONFIG_TRACEPOINTS 1072 void consume_skb(struct sk_buff *skb); 1073 #else 1074 static inline void consume_skb(struct sk_buff *skb) 1075 { 1076 return kfree_skb(skb); 1077 } 1078 #endif 1079 1080 void __consume_stateless_skb(struct sk_buff *skb); 1081 void __kfree_skb(struct sk_buff *skb); 1082 extern struct kmem_cache *skbuff_head_cache; 1083 1084 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1085 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1086 bool *fragstolen, int *delta_truesize); 1087 1088 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1089 int node); 1090 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1091 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1092 struct sk_buff *build_skb_around(struct sk_buff *skb, 1093 void *data, unsigned int frag_size); 1094 1095 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1096 1097 /** 1098 * alloc_skb - allocate a network buffer 1099 * @size: size to allocate 1100 * @priority: allocation mask 1101 * 1102 * This function is a convenient wrapper around __alloc_skb(). 1103 */ 1104 static inline struct sk_buff *alloc_skb(unsigned int size, 1105 gfp_t priority) 1106 { 1107 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1108 } 1109 1110 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1111 unsigned long data_len, 1112 int max_page_order, 1113 int *errcode, 1114 gfp_t gfp_mask); 1115 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1116 1117 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1118 struct sk_buff_fclones { 1119 struct sk_buff skb1; 1120 1121 struct sk_buff skb2; 1122 1123 refcount_t fclone_ref; 1124 }; 1125 1126 /** 1127 * skb_fclone_busy - check if fclone is busy 1128 * @sk: socket 1129 * @skb: buffer 1130 * 1131 * Returns true if skb is a fast clone, and its clone is not freed. 1132 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1133 * so we also check that this didnt happen. 1134 */ 1135 static inline bool skb_fclone_busy(const struct sock *sk, 1136 const struct sk_buff *skb) 1137 { 1138 const struct sk_buff_fclones *fclones; 1139 1140 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1141 1142 return skb->fclone == SKB_FCLONE_ORIG && 1143 refcount_read(&fclones->fclone_ref) > 1 && 1144 READ_ONCE(fclones->skb2.sk) == sk; 1145 } 1146 1147 /** 1148 * alloc_skb_fclone - allocate a network buffer from fclone cache 1149 * @size: size to allocate 1150 * @priority: allocation mask 1151 * 1152 * This function is a convenient wrapper around __alloc_skb(). 1153 */ 1154 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1155 gfp_t priority) 1156 { 1157 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1158 } 1159 1160 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1161 void skb_headers_offset_update(struct sk_buff *skb, int off); 1162 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1163 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1164 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1165 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1166 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1167 gfp_t gfp_mask, bool fclone); 1168 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1169 gfp_t gfp_mask) 1170 { 1171 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1172 } 1173 1174 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1175 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1176 unsigned int headroom); 1177 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1178 int newtailroom, gfp_t priority); 1179 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1180 int offset, int len); 1181 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1182 int offset, int len); 1183 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1184 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1185 1186 /** 1187 * skb_pad - zero pad the tail of an skb 1188 * @skb: buffer to pad 1189 * @pad: space to pad 1190 * 1191 * Ensure that a buffer is followed by a padding area that is zero 1192 * filled. Used by network drivers which may DMA or transfer data 1193 * beyond the buffer end onto the wire. 1194 * 1195 * May return error in out of memory cases. The skb is freed on error. 1196 */ 1197 static inline int skb_pad(struct sk_buff *skb, int pad) 1198 { 1199 return __skb_pad(skb, pad, true); 1200 } 1201 #define dev_kfree_skb(a) consume_skb(a) 1202 1203 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1204 int offset, size_t size); 1205 1206 struct skb_seq_state { 1207 __u32 lower_offset; 1208 __u32 upper_offset; 1209 __u32 frag_idx; 1210 __u32 stepped_offset; 1211 struct sk_buff *root_skb; 1212 struct sk_buff *cur_skb; 1213 __u8 *frag_data; 1214 __u32 frag_off; 1215 }; 1216 1217 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1218 unsigned int to, struct skb_seq_state *st); 1219 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1220 struct skb_seq_state *st); 1221 void skb_abort_seq_read(struct skb_seq_state *st); 1222 1223 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1224 unsigned int to, struct ts_config *config); 1225 1226 /* 1227 * Packet hash types specify the type of hash in skb_set_hash. 1228 * 1229 * Hash types refer to the protocol layer addresses which are used to 1230 * construct a packet's hash. The hashes are used to differentiate or identify 1231 * flows of the protocol layer for the hash type. Hash types are either 1232 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1233 * 1234 * Properties of hashes: 1235 * 1236 * 1) Two packets in different flows have different hash values 1237 * 2) Two packets in the same flow should have the same hash value 1238 * 1239 * A hash at a higher layer is considered to be more specific. A driver should 1240 * set the most specific hash possible. 1241 * 1242 * A driver cannot indicate a more specific hash than the layer at which a hash 1243 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1244 * 1245 * A driver may indicate a hash level which is less specific than the 1246 * actual layer the hash was computed on. For instance, a hash computed 1247 * at L4 may be considered an L3 hash. This should only be done if the 1248 * driver can't unambiguously determine that the HW computed the hash at 1249 * the higher layer. Note that the "should" in the second property above 1250 * permits this. 1251 */ 1252 enum pkt_hash_types { 1253 PKT_HASH_TYPE_NONE, /* Undefined type */ 1254 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1255 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1256 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1257 }; 1258 1259 static inline void skb_clear_hash(struct sk_buff *skb) 1260 { 1261 skb->hash = 0; 1262 skb->sw_hash = 0; 1263 skb->l4_hash = 0; 1264 } 1265 1266 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1267 { 1268 if (!skb->l4_hash) 1269 skb_clear_hash(skb); 1270 } 1271 1272 static inline void 1273 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1274 { 1275 skb->l4_hash = is_l4; 1276 skb->sw_hash = is_sw; 1277 skb->hash = hash; 1278 } 1279 1280 static inline void 1281 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1282 { 1283 /* Used by drivers to set hash from HW */ 1284 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1285 } 1286 1287 static inline void 1288 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1289 { 1290 __skb_set_hash(skb, hash, true, is_l4); 1291 } 1292 1293 void __skb_get_hash(struct sk_buff *skb); 1294 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1295 u32 skb_get_poff(const struct sk_buff *skb); 1296 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1297 const struct flow_keys_basic *keys, int hlen); 1298 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1299 const void *data, int hlen_proto); 1300 1301 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1302 int thoff, u8 ip_proto) 1303 { 1304 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1305 } 1306 1307 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1308 const struct flow_dissector_key *key, 1309 unsigned int key_count); 1310 1311 struct bpf_flow_dissector; 1312 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1313 __be16 proto, int nhoff, int hlen, unsigned int flags); 1314 1315 bool __skb_flow_dissect(const struct net *net, 1316 const struct sk_buff *skb, 1317 struct flow_dissector *flow_dissector, 1318 void *target_container, const void *data, 1319 __be16 proto, int nhoff, int hlen, unsigned int flags); 1320 1321 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1322 struct flow_dissector *flow_dissector, 1323 void *target_container, unsigned int flags) 1324 { 1325 return __skb_flow_dissect(NULL, skb, flow_dissector, 1326 target_container, NULL, 0, 0, 0, flags); 1327 } 1328 1329 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1330 struct flow_keys *flow, 1331 unsigned int flags) 1332 { 1333 memset(flow, 0, sizeof(*flow)); 1334 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1335 flow, NULL, 0, 0, 0, flags); 1336 } 1337 1338 static inline bool 1339 skb_flow_dissect_flow_keys_basic(const struct net *net, 1340 const struct sk_buff *skb, 1341 struct flow_keys_basic *flow, 1342 const void *data, __be16 proto, 1343 int nhoff, int hlen, unsigned int flags) 1344 { 1345 memset(flow, 0, sizeof(*flow)); 1346 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1347 data, proto, nhoff, hlen, flags); 1348 } 1349 1350 void skb_flow_dissect_meta(const struct sk_buff *skb, 1351 struct flow_dissector *flow_dissector, 1352 void *target_container); 1353 1354 /* Gets a skb connection tracking info, ctinfo map should be a 1355 * map of mapsize to translate enum ip_conntrack_info states 1356 * to user states. 1357 */ 1358 void 1359 skb_flow_dissect_ct(const struct sk_buff *skb, 1360 struct flow_dissector *flow_dissector, 1361 void *target_container, 1362 u16 *ctinfo_map, size_t mapsize, 1363 bool post_ct); 1364 void 1365 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1366 struct flow_dissector *flow_dissector, 1367 void *target_container); 1368 1369 void skb_flow_dissect_hash(const struct sk_buff *skb, 1370 struct flow_dissector *flow_dissector, 1371 void *target_container); 1372 1373 static inline __u32 skb_get_hash(struct sk_buff *skb) 1374 { 1375 if (!skb->l4_hash && !skb->sw_hash) 1376 __skb_get_hash(skb); 1377 1378 return skb->hash; 1379 } 1380 1381 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1382 { 1383 if (!skb->l4_hash && !skb->sw_hash) { 1384 struct flow_keys keys; 1385 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1386 1387 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1388 } 1389 1390 return skb->hash; 1391 } 1392 1393 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1394 const siphash_key_t *perturb); 1395 1396 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1397 { 1398 return skb->hash; 1399 } 1400 1401 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1402 { 1403 to->hash = from->hash; 1404 to->sw_hash = from->sw_hash; 1405 to->l4_hash = from->l4_hash; 1406 }; 1407 1408 static inline void skb_copy_decrypted(struct sk_buff *to, 1409 const struct sk_buff *from) 1410 { 1411 #ifdef CONFIG_TLS_DEVICE 1412 to->decrypted = from->decrypted; 1413 #endif 1414 } 1415 1416 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1417 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1418 { 1419 return skb->head + skb->end; 1420 } 1421 1422 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1423 { 1424 return skb->end; 1425 } 1426 #else 1427 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1428 { 1429 return skb->end; 1430 } 1431 1432 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1433 { 1434 return skb->end - skb->head; 1435 } 1436 #endif 1437 1438 /* Internal */ 1439 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1440 1441 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1442 { 1443 return &skb_shinfo(skb)->hwtstamps; 1444 } 1445 1446 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1447 { 1448 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1449 1450 return is_zcopy ? skb_uarg(skb) : NULL; 1451 } 1452 1453 static inline void net_zcopy_get(struct ubuf_info *uarg) 1454 { 1455 refcount_inc(&uarg->refcnt); 1456 } 1457 1458 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1459 { 1460 skb_shinfo(skb)->destructor_arg = uarg; 1461 skb_shinfo(skb)->flags |= uarg->flags; 1462 } 1463 1464 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1465 bool *have_ref) 1466 { 1467 if (skb && uarg && !skb_zcopy(skb)) { 1468 if (unlikely(have_ref && *have_ref)) 1469 *have_ref = false; 1470 else 1471 net_zcopy_get(uarg); 1472 skb_zcopy_init(skb, uarg); 1473 } 1474 } 1475 1476 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1477 { 1478 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1479 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1480 } 1481 1482 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1483 { 1484 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1485 } 1486 1487 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1488 { 1489 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1490 } 1491 1492 static inline void net_zcopy_put(struct ubuf_info *uarg) 1493 { 1494 if (uarg) 1495 uarg->callback(NULL, uarg, true); 1496 } 1497 1498 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1499 { 1500 if (uarg) { 1501 if (uarg->callback == msg_zerocopy_callback) 1502 msg_zerocopy_put_abort(uarg, have_uref); 1503 else if (have_uref) 1504 net_zcopy_put(uarg); 1505 } 1506 } 1507 1508 /* Release a reference on a zerocopy structure */ 1509 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1510 { 1511 struct ubuf_info *uarg = skb_zcopy(skb); 1512 1513 if (uarg) { 1514 if (!skb_zcopy_is_nouarg(skb)) 1515 uarg->callback(skb, uarg, zerocopy_success); 1516 1517 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG; 1518 } 1519 } 1520 1521 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1522 { 1523 skb->next = NULL; 1524 } 1525 1526 /* Iterate through singly-linked GSO fragments of an skb. */ 1527 #define skb_list_walk_safe(first, skb, next_skb) \ 1528 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1529 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1530 1531 static inline void skb_list_del_init(struct sk_buff *skb) 1532 { 1533 __list_del_entry(&skb->list); 1534 skb_mark_not_on_list(skb); 1535 } 1536 1537 /** 1538 * skb_queue_empty - check if a queue is empty 1539 * @list: queue head 1540 * 1541 * Returns true if the queue is empty, false otherwise. 1542 */ 1543 static inline int skb_queue_empty(const struct sk_buff_head *list) 1544 { 1545 return list->next == (const struct sk_buff *) list; 1546 } 1547 1548 /** 1549 * skb_queue_empty_lockless - check if a queue is empty 1550 * @list: queue head 1551 * 1552 * Returns true if the queue is empty, false otherwise. 1553 * This variant can be used in lockless contexts. 1554 */ 1555 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1556 { 1557 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1558 } 1559 1560 1561 /** 1562 * skb_queue_is_last - check if skb is the last entry in the queue 1563 * @list: queue head 1564 * @skb: buffer 1565 * 1566 * Returns true if @skb is the last buffer on the list. 1567 */ 1568 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1569 const struct sk_buff *skb) 1570 { 1571 return skb->next == (const struct sk_buff *) list; 1572 } 1573 1574 /** 1575 * skb_queue_is_first - check if skb is the first entry in the queue 1576 * @list: queue head 1577 * @skb: buffer 1578 * 1579 * Returns true if @skb is the first buffer on the list. 1580 */ 1581 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1582 const struct sk_buff *skb) 1583 { 1584 return skb->prev == (const struct sk_buff *) list; 1585 } 1586 1587 /** 1588 * skb_queue_next - return the next packet in the queue 1589 * @list: queue head 1590 * @skb: current buffer 1591 * 1592 * Return the next packet in @list after @skb. It is only valid to 1593 * call this if skb_queue_is_last() evaluates to false. 1594 */ 1595 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1596 const struct sk_buff *skb) 1597 { 1598 /* This BUG_ON may seem severe, but if we just return then we 1599 * are going to dereference garbage. 1600 */ 1601 BUG_ON(skb_queue_is_last(list, skb)); 1602 return skb->next; 1603 } 1604 1605 /** 1606 * skb_queue_prev - return the prev packet in the queue 1607 * @list: queue head 1608 * @skb: current buffer 1609 * 1610 * Return the prev packet in @list before @skb. It is only valid to 1611 * call this if skb_queue_is_first() evaluates to false. 1612 */ 1613 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1614 const struct sk_buff *skb) 1615 { 1616 /* This BUG_ON may seem severe, but if we just return then we 1617 * are going to dereference garbage. 1618 */ 1619 BUG_ON(skb_queue_is_first(list, skb)); 1620 return skb->prev; 1621 } 1622 1623 /** 1624 * skb_get - reference buffer 1625 * @skb: buffer to reference 1626 * 1627 * Makes another reference to a socket buffer and returns a pointer 1628 * to the buffer. 1629 */ 1630 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1631 { 1632 refcount_inc(&skb->users); 1633 return skb; 1634 } 1635 1636 /* 1637 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1638 */ 1639 1640 /** 1641 * skb_cloned - is the buffer a clone 1642 * @skb: buffer to check 1643 * 1644 * Returns true if the buffer was generated with skb_clone() and is 1645 * one of multiple shared copies of the buffer. Cloned buffers are 1646 * shared data so must not be written to under normal circumstances. 1647 */ 1648 static inline int skb_cloned(const struct sk_buff *skb) 1649 { 1650 return skb->cloned && 1651 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1652 } 1653 1654 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1655 { 1656 might_sleep_if(gfpflags_allow_blocking(pri)); 1657 1658 if (skb_cloned(skb)) 1659 return pskb_expand_head(skb, 0, 0, pri); 1660 1661 return 0; 1662 } 1663 1664 /** 1665 * skb_header_cloned - is the header a clone 1666 * @skb: buffer to check 1667 * 1668 * Returns true if modifying the header part of the buffer requires 1669 * the data to be copied. 1670 */ 1671 static inline int skb_header_cloned(const struct sk_buff *skb) 1672 { 1673 int dataref; 1674 1675 if (!skb->cloned) 1676 return 0; 1677 1678 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1679 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1680 return dataref != 1; 1681 } 1682 1683 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1684 { 1685 might_sleep_if(gfpflags_allow_blocking(pri)); 1686 1687 if (skb_header_cloned(skb)) 1688 return pskb_expand_head(skb, 0, 0, pri); 1689 1690 return 0; 1691 } 1692 1693 /** 1694 * __skb_header_release - release reference to header 1695 * @skb: buffer to operate on 1696 */ 1697 static inline void __skb_header_release(struct sk_buff *skb) 1698 { 1699 skb->nohdr = 1; 1700 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1701 } 1702 1703 1704 /** 1705 * skb_shared - is the buffer shared 1706 * @skb: buffer to check 1707 * 1708 * Returns true if more than one person has a reference to this 1709 * buffer. 1710 */ 1711 static inline int skb_shared(const struct sk_buff *skb) 1712 { 1713 return refcount_read(&skb->users) != 1; 1714 } 1715 1716 /** 1717 * skb_share_check - check if buffer is shared and if so clone it 1718 * @skb: buffer to check 1719 * @pri: priority for memory allocation 1720 * 1721 * If the buffer is shared the buffer is cloned and the old copy 1722 * drops a reference. A new clone with a single reference is returned. 1723 * If the buffer is not shared the original buffer is returned. When 1724 * being called from interrupt status or with spinlocks held pri must 1725 * be GFP_ATOMIC. 1726 * 1727 * NULL is returned on a memory allocation failure. 1728 */ 1729 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1730 { 1731 might_sleep_if(gfpflags_allow_blocking(pri)); 1732 if (skb_shared(skb)) { 1733 struct sk_buff *nskb = skb_clone(skb, pri); 1734 1735 if (likely(nskb)) 1736 consume_skb(skb); 1737 else 1738 kfree_skb(skb); 1739 skb = nskb; 1740 } 1741 return skb; 1742 } 1743 1744 /* 1745 * Copy shared buffers into a new sk_buff. We effectively do COW on 1746 * packets to handle cases where we have a local reader and forward 1747 * and a couple of other messy ones. The normal one is tcpdumping 1748 * a packet thats being forwarded. 1749 */ 1750 1751 /** 1752 * skb_unshare - make a copy of a shared buffer 1753 * @skb: buffer to check 1754 * @pri: priority for memory allocation 1755 * 1756 * If the socket buffer is a clone then this function creates a new 1757 * copy of the data, drops a reference count on the old copy and returns 1758 * the new copy with the reference count at 1. If the buffer is not a clone 1759 * the original buffer is returned. When called with a spinlock held or 1760 * from interrupt state @pri must be %GFP_ATOMIC 1761 * 1762 * %NULL is returned on a memory allocation failure. 1763 */ 1764 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1765 gfp_t pri) 1766 { 1767 might_sleep_if(gfpflags_allow_blocking(pri)); 1768 if (skb_cloned(skb)) { 1769 struct sk_buff *nskb = skb_copy(skb, pri); 1770 1771 /* Free our shared copy */ 1772 if (likely(nskb)) 1773 consume_skb(skb); 1774 else 1775 kfree_skb(skb); 1776 skb = nskb; 1777 } 1778 return skb; 1779 } 1780 1781 /** 1782 * skb_peek - peek at the head of an &sk_buff_head 1783 * @list_: list to peek at 1784 * 1785 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1786 * be careful with this one. A peek leaves the buffer on the 1787 * list and someone else may run off with it. You must hold 1788 * the appropriate locks or have a private queue to do this. 1789 * 1790 * Returns %NULL for an empty list or a pointer to the head element. 1791 * The reference count is not incremented and the reference is therefore 1792 * volatile. Use with caution. 1793 */ 1794 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1795 { 1796 struct sk_buff *skb = list_->next; 1797 1798 if (skb == (struct sk_buff *)list_) 1799 skb = NULL; 1800 return skb; 1801 } 1802 1803 /** 1804 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1805 * @list_: list to peek at 1806 * 1807 * Like skb_peek(), but the caller knows that the list is not empty. 1808 */ 1809 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1810 { 1811 return list_->next; 1812 } 1813 1814 /** 1815 * skb_peek_next - peek skb following the given one from a queue 1816 * @skb: skb to start from 1817 * @list_: list to peek at 1818 * 1819 * Returns %NULL when the end of the list is met or a pointer to the 1820 * next element. The reference count is not incremented and the 1821 * reference is therefore volatile. Use with caution. 1822 */ 1823 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1824 const struct sk_buff_head *list_) 1825 { 1826 struct sk_buff *next = skb->next; 1827 1828 if (next == (struct sk_buff *)list_) 1829 next = NULL; 1830 return next; 1831 } 1832 1833 /** 1834 * skb_peek_tail - peek at the tail of an &sk_buff_head 1835 * @list_: list to peek at 1836 * 1837 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1838 * be careful with this one. A peek leaves the buffer on the 1839 * list and someone else may run off with it. You must hold 1840 * the appropriate locks or have a private queue to do this. 1841 * 1842 * Returns %NULL for an empty list or a pointer to the tail element. 1843 * The reference count is not incremented and the reference is therefore 1844 * volatile. Use with caution. 1845 */ 1846 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1847 { 1848 struct sk_buff *skb = READ_ONCE(list_->prev); 1849 1850 if (skb == (struct sk_buff *)list_) 1851 skb = NULL; 1852 return skb; 1853 1854 } 1855 1856 /** 1857 * skb_queue_len - get queue length 1858 * @list_: list to measure 1859 * 1860 * Return the length of an &sk_buff queue. 1861 */ 1862 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1863 { 1864 return list_->qlen; 1865 } 1866 1867 /** 1868 * skb_queue_len_lockless - get queue length 1869 * @list_: list to measure 1870 * 1871 * Return the length of an &sk_buff queue. 1872 * This variant can be used in lockless contexts. 1873 */ 1874 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1875 { 1876 return READ_ONCE(list_->qlen); 1877 } 1878 1879 /** 1880 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1881 * @list: queue to initialize 1882 * 1883 * This initializes only the list and queue length aspects of 1884 * an sk_buff_head object. This allows to initialize the list 1885 * aspects of an sk_buff_head without reinitializing things like 1886 * the spinlock. It can also be used for on-stack sk_buff_head 1887 * objects where the spinlock is known to not be used. 1888 */ 1889 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1890 { 1891 list->prev = list->next = (struct sk_buff *)list; 1892 list->qlen = 0; 1893 } 1894 1895 /* 1896 * This function creates a split out lock class for each invocation; 1897 * this is needed for now since a whole lot of users of the skb-queue 1898 * infrastructure in drivers have different locking usage (in hardirq) 1899 * than the networking core (in softirq only). In the long run either the 1900 * network layer or drivers should need annotation to consolidate the 1901 * main types of usage into 3 classes. 1902 */ 1903 static inline void skb_queue_head_init(struct sk_buff_head *list) 1904 { 1905 spin_lock_init(&list->lock); 1906 __skb_queue_head_init(list); 1907 } 1908 1909 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1910 struct lock_class_key *class) 1911 { 1912 skb_queue_head_init(list); 1913 lockdep_set_class(&list->lock, class); 1914 } 1915 1916 /* 1917 * Insert an sk_buff on a list. 1918 * 1919 * The "__skb_xxxx()" functions are the non-atomic ones that 1920 * can only be called with interrupts disabled. 1921 */ 1922 static inline void __skb_insert(struct sk_buff *newsk, 1923 struct sk_buff *prev, struct sk_buff *next, 1924 struct sk_buff_head *list) 1925 { 1926 /* See skb_queue_empty_lockless() and skb_peek_tail() 1927 * for the opposite READ_ONCE() 1928 */ 1929 WRITE_ONCE(newsk->next, next); 1930 WRITE_ONCE(newsk->prev, prev); 1931 WRITE_ONCE(next->prev, newsk); 1932 WRITE_ONCE(prev->next, newsk); 1933 list->qlen++; 1934 } 1935 1936 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1937 struct sk_buff *prev, 1938 struct sk_buff *next) 1939 { 1940 struct sk_buff *first = list->next; 1941 struct sk_buff *last = list->prev; 1942 1943 WRITE_ONCE(first->prev, prev); 1944 WRITE_ONCE(prev->next, first); 1945 1946 WRITE_ONCE(last->next, next); 1947 WRITE_ONCE(next->prev, last); 1948 } 1949 1950 /** 1951 * skb_queue_splice - join two skb lists, this is designed for stacks 1952 * @list: the new list to add 1953 * @head: the place to add it in the first list 1954 */ 1955 static inline void skb_queue_splice(const struct sk_buff_head *list, 1956 struct sk_buff_head *head) 1957 { 1958 if (!skb_queue_empty(list)) { 1959 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1960 head->qlen += list->qlen; 1961 } 1962 } 1963 1964 /** 1965 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1966 * @list: the new list to add 1967 * @head: the place to add it in the first list 1968 * 1969 * The list at @list is reinitialised 1970 */ 1971 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1972 struct sk_buff_head *head) 1973 { 1974 if (!skb_queue_empty(list)) { 1975 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1976 head->qlen += list->qlen; 1977 __skb_queue_head_init(list); 1978 } 1979 } 1980 1981 /** 1982 * skb_queue_splice_tail - join two skb lists, each list being a queue 1983 * @list: the new list to add 1984 * @head: the place to add it in the first list 1985 */ 1986 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1987 struct sk_buff_head *head) 1988 { 1989 if (!skb_queue_empty(list)) { 1990 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1991 head->qlen += list->qlen; 1992 } 1993 } 1994 1995 /** 1996 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1997 * @list: the new list to add 1998 * @head: the place to add it in the first list 1999 * 2000 * Each of the lists is a queue. 2001 * The list at @list is reinitialised 2002 */ 2003 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2004 struct sk_buff_head *head) 2005 { 2006 if (!skb_queue_empty(list)) { 2007 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2008 head->qlen += list->qlen; 2009 __skb_queue_head_init(list); 2010 } 2011 } 2012 2013 /** 2014 * __skb_queue_after - queue a buffer at the list head 2015 * @list: list to use 2016 * @prev: place after this buffer 2017 * @newsk: buffer to queue 2018 * 2019 * Queue a buffer int the middle of a list. This function takes no locks 2020 * and you must therefore hold required locks before calling it. 2021 * 2022 * A buffer cannot be placed on two lists at the same time. 2023 */ 2024 static inline void __skb_queue_after(struct sk_buff_head *list, 2025 struct sk_buff *prev, 2026 struct sk_buff *newsk) 2027 { 2028 __skb_insert(newsk, prev, prev->next, list); 2029 } 2030 2031 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2032 struct sk_buff_head *list); 2033 2034 static inline void __skb_queue_before(struct sk_buff_head *list, 2035 struct sk_buff *next, 2036 struct sk_buff *newsk) 2037 { 2038 __skb_insert(newsk, next->prev, next, list); 2039 } 2040 2041 /** 2042 * __skb_queue_head - queue a buffer at the list head 2043 * @list: list to use 2044 * @newsk: buffer to queue 2045 * 2046 * Queue a buffer at the start of a list. This function takes no locks 2047 * and you must therefore hold required locks before calling it. 2048 * 2049 * A buffer cannot be placed on two lists at the same time. 2050 */ 2051 static inline void __skb_queue_head(struct sk_buff_head *list, 2052 struct sk_buff *newsk) 2053 { 2054 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2055 } 2056 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2057 2058 /** 2059 * __skb_queue_tail - queue a buffer at the list tail 2060 * @list: list to use 2061 * @newsk: buffer to queue 2062 * 2063 * Queue a buffer at the end of a list. This function takes no locks 2064 * and you must therefore hold required locks before calling it. 2065 * 2066 * A buffer cannot be placed on two lists at the same time. 2067 */ 2068 static inline void __skb_queue_tail(struct sk_buff_head *list, 2069 struct sk_buff *newsk) 2070 { 2071 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2072 } 2073 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2074 2075 /* 2076 * remove sk_buff from list. _Must_ be called atomically, and with 2077 * the list known.. 2078 */ 2079 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2080 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2081 { 2082 struct sk_buff *next, *prev; 2083 2084 WRITE_ONCE(list->qlen, list->qlen - 1); 2085 next = skb->next; 2086 prev = skb->prev; 2087 skb->next = skb->prev = NULL; 2088 WRITE_ONCE(next->prev, prev); 2089 WRITE_ONCE(prev->next, next); 2090 } 2091 2092 /** 2093 * __skb_dequeue - remove from the head of the queue 2094 * @list: list to dequeue from 2095 * 2096 * Remove the head of the list. This function does not take any locks 2097 * so must be used with appropriate locks held only. The head item is 2098 * returned or %NULL if the list is empty. 2099 */ 2100 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2101 { 2102 struct sk_buff *skb = skb_peek(list); 2103 if (skb) 2104 __skb_unlink(skb, list); 2105 return skb; 2106 } 2107 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2108 2109 /** 2110 * __skb_dequeue_tail - remove from the tail of the queue 2111 * @list: list to dequeue from 2112 * 2113 * Remove the tail of the list. This function does not take any locks 2114 * so must be used with appropriate locks held only. The tail item is 2115 * returned or %NULL if the list is empty. 2116 */ 2117 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2118 { 2119 struct sk_buff *skb = skb_peek_tail(list); 2120 if (skb) 2121 __skb_unlink(skb, list); 2122 return skb; 2123 } 2124 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2125 2126 2127 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2128 { 2129 return skb->data_len; 2130 } 2131 2132 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2133 { 2134 return skb->len - skb->data_len; 2135 } 2136 2137 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2138 { 2139 unsigned int i, len = 0; 2140 2141 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2142 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2143 return len; 2144 } 2145 2146 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2147 { 2148 return skb_headlen(skb) + __skb_pagelen(skb); 2149 } 2150 2151 /** 2152 * __skb_fill_page_desc - initialise a paged fragment in an skb 2153 * @skb: buffer containing fragment to be initialised 2154 * @i: paged fragment index to initialise 2155 * @page: the page to use for this fragment 2156 * @off: the offset to the data with @page 2157 * @size: the length of the data 2158 * 2159 * Initialises the @i'th fragment of @skb to point to &size bytes at 2160 * offset @off within @page. 2161 * 2162 * Does not take any additional reference on the fragment. 2163 */ 2164 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2165 struct page *page, int off, int size) 2166 { 2167 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2168 2169 /* 2170 * Propagate page pfmemalloc to the skb if we can. The problem is 2171 * that not all callers have unique ownership of the page but rely 2172 * on page_is_pfmemalloc doing the right thing(tm). 2173 */ 2174 frag->bv_page = page; 2175 frag->bv_offset = off; 2176 skb_frag_size_set(frag, size); 2177 2178 page = compound_head(page); 2179 if (page_is_pfmemalloc(page)) 2180 skb->pfmemalloc = true; 2181 } 2182 2183 /** 2184 * skb_fill_page_desc - initialise a paged fragment in an skb 2185 * @skb: buffer containing fragment to be initialised 2186 * @i: paged fragment index to initialise 2187 * @page: the page to use for this fragment 2188 * @off: the offset to the data with @page 2189 * @size: the length of the data 2190 * 2191 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2192 * @skb to point to @size bytes at offset @off within @page. In 2193 * addition updates @skb such that @i is the last fragment. 2194 * 2195 * Does not take any additional reference on the fragment. 2196 */ 2197 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2198 struct page *page, int off, int size) 2199 { 2200 __skb_fill_page_desc(skb, i, page, off, size); 2201 skb_shinfo(skb)->nr_frags = i + 1; 2202 } 2203 2204 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2205 int size, unsigned int truesize); 2206 2207 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2208 unsigned int truesize); 2209 2210 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2211 2212 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2213 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2214 { 2215 return skb->head + skb->tail; 2216 } 2217 2218 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2219 { 2220 skb->tail = skb->data - skb->head; 2221 } 2222 2223 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2224 { 2225 skb_reset_tail_pointer(skb); 2226 skb->tail += offset; 2227 } 2228 2229 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2230 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2231 { 2232 return skb->tail; 2233 } 2234 2235 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2236 { 2237 skb->tail = skb->data; 2238 } 2239 2240 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2241 { 2242 skb->tail = skb->data + offset; 2243 } 2244 2245 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2246 2247 /* 2248 * Add data to an sk_buff 2249 */ 2250 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2251 void *skb_put(struct sk_buff *skb, unsigned int len); 2252 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2253 { 2254 void *tmp = skb_tail_pointer(skb); 2255 SKB_LINEAR_ASSERT(skb); 2256 skb->tail += len; 2257 skb->len += len; 2258 return tmp; 2259 } 2260 2261 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2262 { 2263 void *tmp = __skb_put(skb, len); 2264 2265 memset(tmp, 0, len); 2266 return tmp; 2267 } 2268 2269 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2270 unsigned int len) 2271 { 2272 void *tmp = __skb_put(skb, len); 2273 2274 memcpy(tmp, data, len); 2275 return tmp; 2276 } 2277 2278 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2279 { 2280 *(u8 *)__skb_put(skb, 1) = val; 2281 } 2282 2283 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2284 { 2285 void *tmp = skb_put(skb, len); 2286 2287 memset(tmp, 0, len); 2288 2289 return tmp; 2290 } 2291 2292 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2293 unsigned int len) 2294 { 2295 void *tmp = skb_put(skb, len); 2296 2297 memcpy(tmp, data, len); 2298 2299 return tmp; 2300 } 2301 2302 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2303 { 2304 *(u8 *)skb_put(skb, 1) = val; 2305 } 2306 2307 void *skb_push(struct sk_buff *skb, unsigned int len); 2308 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2309 { 2310 skb->data -= len; 2311 skb->len += len; 2312 return skb->data; 2313 } 2314 2315 void *skb_pull(struct sk_buff *skb, unsigned int len); 2316 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2317 { 2318 skb->len -= len; 2319 BUG_ON(skb->len < skb->data_len); 2320 return skb->data += len; 2321 } 2322 2323 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2324 { 2325 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2326 } 2327 2328 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2329 2330 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2331 { 2332 if (len > skb_headlen(skb) && 2333 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2334 return NULL; 2335 skb->len -= len; 2336 return skb->data += len; 2337 } 2338 2339 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2340 { 2341 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2342 } 2343 2344 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2345 { 2346 if (likely(len <= skb_headlen(skb))) 2347 return true; 2348 if (unlikely(len > skb->len)) 2349 return false; 2350 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2351 } 2352 2353 void skb_condense(struct sk_buff *skb); 2354 2355 /** 2356 * skb_headroom - bytes at buffer head 2357 * @skb: buffer to check 2358 * 2359 * Return the number of bytes of free space at the head of an &sk_buff. 2360 */ 2361 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2362 { 2363 return skb->data - skb->head; 2364 } 2365 2366 /** 2367 * skb_tailroom - bytes at buffer end 2368 * @skb: buffer to check 2369 * 2370 * Return the number of bytes of free space at the tail of an sk_buff 2371 */ 2372 static inline int skb_tailroom(const struct sk_buff *skb) 2373 { 2374 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2375 } 2376 2377 /** 2378 * skb_availroom - bytes at buffer end 2379 * @skb: buffer to check 2380 * 2381 * Return the number of bytes of free space at the tail of an sk_buff 2382 * allocated by sk_stream_alloc() 2383 */ 2384 static inline int skb_availroom(const struct sk_buff *skb) 2385 { 2386 if (skb_is_nonlinear(skb)) 2387 return 0; 2388 2389 return skb->end - skb->tail - skb->reserved_tailroom; 2390 } 2391 2392 /** 2393 * skb_reserve - adjust headroom 2394 * @skb: buffer to alter 2395 * @len: bytes to move 2396 * 2397 * Increase the headroom of an empty &sk_buff by reducing the tail 2398 * room. This is only allowed for an empty buffer. 2399 */ 2400 static inline void skb_reserve(struct sk_buff *skb, int len) 2401 { 2402 skb->data += len; 2403 skb->tail += len; 2404 } 2405 2406 /** 2407 * skb_tailroom_reserve - adjust reserved_tailroom 2408 * @skb: buffer to alter 2409 * @mtu: maximum amount of headlen permitted 2410 * @needed_tailroom: minimum amount of reserved_tailroom 2411 * 2412 * Set reserved_tailroom so that headlen can be as large as possible but 2413 * not larger than mtu and tailroom cannot be smaller than 2414 * needed_tailroom. 2415 * The required headroom should already have been reserved before using 2416 * this function. 2417 */ 2418 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2419 unsigned int needed_tailroom) 2420 { 2421 SKB_LINEAR_ASSERT(skb); 2422 if (mtu < skb_tailroom(skb) - needed_tailroom) 2423 /* use at most mtu */ 2424 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2425 else 2426 /* use up to all available space */ 2427 skb->reserved_tailroom = needed_tailroom; 2428 } 2429 2430 #define ENCAP_TYPE_ETHER 0 2431 #define ENCAP_TYPE_IPPROTO 1 2432 2433 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2434 __be16 protocol) 2435 { 2436 skb->inner_protocol = protocol; 2437 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2438 } 2439 2440 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2441 __u8 ipproto) 2442 { 2443 skb->inner_ipproto = ipproto; 2444 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2445 } 2446 2447 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2448 { 2449 skb->inner_mac_header = skb->mac_header; 2450 skb->inner_network_header = skb->network_header; 2451 skb->inner_transport_header = skb->transport_header; 2452 } 2453 2454 static inline void skb_reset_mac_len(struct sk_buff *skb) 2455 { 2456 skb->mac_len = skb->network_header - skb->mac_header; 2457 } 2458 2459 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2460 *skb) 2461 { 2462 return skb->head + skb->inner_transport_header; 2463 } 2464 2465 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2466 { 2467 return skb_inner_transport_header(skb) - skb->data; 2468 } 2469 2470 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2471 { 2472 skb->inner_transport_header = skb->data - skb->head; 2473 } 2474 2475 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2476 const int offset) 2477 { 2478 skb_reset_inner_transport_header(skb); 2479 skb->inner_transport_header += offset; 2480 } 2481 2482 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2483 { 2484 return skb->head + skb->inner_network_header; 2485 } 2486 2487 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2488 { 2489 skb->inner_network_header = skb->data - skb->head; 2490 } 2491 2492 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2493 const int offset) 2494 { 2495 skb_reset_inner_network_header(skb); 2496 skb->inner_network_header += offset; 2497 } 2498 2499 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2500 { 2501 return skb->head + skb->inner_mac_header; 2502 } 2503 2504 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2505 { 2506 skb->inner_mac_header = skb->data - skb->head; 2507 } 2508 2509 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2510 const int offset) 2511 { 2512 skb_reset_inner_mac_header(skb); 2513 skb->inner_mac_header += offset; 2514 } 2515 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2516 { 2517 return skb->transport_header != (typeof(skb->transport_header))~0U; 2518 } 2519 2520 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2521 { 2522 return skb->head + skb->transport_header; 2523 } 2524 2525 static inline void skb_reset_transport_header(struct sk_buff *skb) 2526 { 2527 skb->transport_header = skb->data - skb->head; 2528 } 2529 2530 static inline void skb_set_transport_header(struct sk_buff *skb, 2531 const int offset) 2532 { 2533 skb_reset_transport_header(skb); 2534 skb->transport_header += offset; 2535 } 2536 2537 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2538 { 2539 return skb->head + skb->network_header; 2540 } 2541 2542 static inline void skb_reset_network_header(struct sk_buff *skb) 2543 { 2544 skb->network_header = skb->data - skb->head; 2545 } 2546 2547 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2548 { 2549 skb_reset_network_header(skb); 2550 skb->network_header += offset; 2551 } 2552 2553 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2554 { 2555 return skb->head + skb->mac_header; 2556 } 2557 2558 static inline int skb_mac_offset(const struct sk_buff *skb) 2559 { 2560 return skb_mac_header(skb) - skb->data; 2561 } 2562 2563 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2564 { 2565 return skb->network_header - skb->mac_header; 2566 } 2567 2568 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2569 { 2570 return skb->mac_header != (typeof(skb->mac_header))~0U; 2571 } 2572 2573 static inline void skb_unset_mac_header(struct sk_buff *skb) 2574 { 2575 skb->mac_header = (typeof(skb->mac_header))~0U; 2576 } 2577 2578 static inline void skb_reset_mac_header(struct sk_buff *skb) 2579 { 2580 skb->mac_header = skb->data - skb->head; 2581 } 2582 2583 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2584 { 2585 skb_reset_mac_header(skb); 2586 skb->mac_header += offset; 2587 } 2588 2589 static inline void skb_pop_mac_header(struct sk_buff *skb) 2590 { 2591 skb->mac_header = skb->network_header; 2592 } 2593 2594 static inline void skb_probe_transport_header(struct sk_buff *skb) 2595 { 2596 struct flow_keys_basic keys; 2597 2598 if (skb_transport_header_was_set(skb)) 2599 return; 2600 2601 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2602 NULL, 0, 0, 0, 0)) 2603 skb_set_transport_header(skb, keys.control.thoff); 2604 } 2605 2606 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2607 { 2608 if (skb_mac_header_was_set(skb)) { 2609 const unsigned char *old_mac = skb_mac_header(skb); 2610 2611 skb_set_mac_header(skb, -skb->mac_len); 2612 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2613 } 2614 } 2615 2616 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2617 { 2618 return skb->csum_start - skb_headroom(skb); 2619 } 2620 2621 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2622 { 2623 return skb->head + skb->csum_start; 2624 } 2625 2626 static inline int skb_transport_offset(const struct sk_buff *skb) 2627 { 2628 return skb_transport_header(skb) - skb->data; 2629 } 2630 2631 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2632 { 2633 return skb->transport_header - skb->network_header; 2634 } 2635 2636 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2637 { 2638 return skb->inner_transport_header - skb->inner_network_header; 2639 } 2640 2641 static inline int skb_network_offset(const struct sk_buff *skb) 2642 { 2643 return skb_network_header(skb) - skb->data; 2644 } 2645 2646 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2647 { 2648 return skb_inner_network_header(skb) - skb->data; 2649 } 2650 2651 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2652 { 2653 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2654 } 2655 2656 /* 2657 * CPUs often take a performance hit when accessing unaligned memory 2658 * locations. The actual performance hit varies, it can be small if the 2659 * hardware handles it or large if we have to take an exception and fix it 2660 * in software. 2661 * 2662 * Since an ethernet header is 14 bytes network drivers often end up with 2663 * the IP header at an unaligned offset. The IP header can be aligned by 2664 * shifting the start of the packet by 2 bytes. Drivers should do this 2665 * with: 2666 * 2667 * skb_reserve(skb, NET_IP_ALIGN); 2668 * 2669 * The downside to this alignment of the IP header is that the DMA is now 2670 * unaligned. On some architectures the cost of an unaligned DMA is high 2671 * and this cost outweighs the gains made by aligning the IP header. 2672 * 2673 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2674 * to be overridden. 2675 */ 2676 #ifndef NET_IP_ALIGN 2677 #define NET_IP_ALIGN 2 2678 #endif 2679 2680 /* 2681 * The networking layer reserves some headroom in skb data (via 2682 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2683 * the header has to grow. In the default case, if the header has to grow 2684 * 32 bytes or less we avoid the reallocation. 2685 * 2686 * Unfortunately this headroom changes the DMA alignment of the resulting 2687 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2688 * on some architectures. An architecture can override this value, 2689 * perhaps setting it to a cacheline in size (since that will maintain 2690 * cacheline alignment of the DMA). It must be a power of 2. 2691 * 2692 * Various parts of the networking layer expect at least 32 bytes of 2693 * headroom, you should not reduce this. 2694 * 2695 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2696 * to reduce average number of cache lines per packet. 2697 * get_rps_cpu() for example only access one 64 bytes aligned block : 2698 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2699 */ 2700 #ifndef NET_SKB_PAD 2701 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2702 #endif 2703 2704 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2705 2706 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2707 { 2708 if (WARN_ON(skb_is_nonlinear(skb))) 2709 return; 2710 skb->len = len; 2711 skb_set_tail_pointer(skb, len); 2712 } 2713 2714 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2715 { 2716 __skb_set_length(skb, len); 2717 } 2718 2719 void skb_trim(struct sk_buff *skb, unsigned int len); 2720 2721 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2722 { 2723 if (skb->data_len) 2724 return ___pskb_trim(skb, len); 2725 __skb_trim(skb, len); 2726 return 0; 2727 } 2728 2729 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2730 { 2731 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2732 } 2733 2734 /** 2735 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2736 * @skb: buffer to alter 2737 * @len: new length 2738 * 2739 * This is identical to pskb_trim except that the caller knows that 2740 * the skb is not cloned so we should never get an error due to out- 2741 * of-memory. 2742 */ 2743 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2744 { 2745 int err = pskb_trim(skb, len); 2746 BUG_ON(err); 2747 } 2748 2749 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2750 { 2751 unsigned int diff = len - skb->len; 2752 2753 if (skb_tailroom(skb) < diff) { 2754 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2755 GFP_ATOMIC); 2756 if (ret) 2757 return ret; 2758 } 2759 __skb_set_length(skb, len); 2760 return 0; 2761 } 2762 2763 /** 2764 * skb_orphan - orphan a buffer 2765 * @skb: buffer to orphan 2766 * 2767 * If a buffer currently has an owner then we call the owner's 2768 * destructor function and make the @skb unowned. The buffer continues 2769 * to exist but is no longer charged to its former owner. 2770 */ 2771 static inline void skb_orphan(struct sk_buff *skb) 2772 { 2773 if (skb->destructor) { 2774 skb->destructor(skb); 2775 skb->destructor = NULL; 2776 skb->sk = NULL; 2777 } else { 2778 BUG_ON(skb->sk); 2779 } 2780 } 2781 2782 /** 2783 * skb_orphan_frags - orphan the frags contained in a buffer 2784 * @skb: buffer to orphan frags from 2785 * @gfp_mask: allocation mask for replacement pages 2786 * 2787 * For each frag in the SKB which needs a destructor (i.e. has an 2788 * owner) create a copy of that frag and release the original 2789 * page by calling the destructor. 2790 */ 2791 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2792 { 2793 if (likely(!skb_zcopy(skb))) 2794 return 0; 2795 if (!skb_zcopy_is_nouarg(skb) && 2796 skb_uarg(skb)->callback == msg_zerocopy_callback) 2797 return 0; 2798 return skb_copy_ubufs(skb, gfp_mask); 2799 } 2800 2801 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2802 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2803 { 2804 if (likely(!skb_zcopy(skb))) 2805 return 0; 2806 return skb_copy_ubufs(skb, gfp_mask); 2807 } 2808 2809 /** 2810 * __skb_queue_purge - empty a list 2811 * @list: list to empty 2812 * 2813 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2814 * the list and one reference dropped. This function does not take the 2815 * list lock and the caller must hold the relevant locks to use it. 2816 */ 2817 static inline void __skb_queue_purge(struct sk_buff_head *list) 2818 { 2819 struct sk_buff *skb; 2820 while ((skb = __skb_dequeue(list)) != NULL) 2821 kfree_skb(skb); 2822 } 2823 void skb_queue_purge(struct sk_buff_head *list); 2824 2825 unsigned int skb_rbtree_purge(struct rb_root *root); 2826 2827 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2828 2829 /** 2830 * netdev_alloc_frag - allocate a page fragment 2831 * @fragsz: fragment size 2832 * 2833 * Allocates a frag from a page for receive buffer. 2834 * Uses GFP_ATOMIC allocations. 2835 */ 2836 static inline void *netdev_alloc_frag(unsigned int fragsz) 2837 { 2838 return __netdev_alloc_frag_align(fragsz, ~0u); 2839 } 2840 2841 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 2842 unsigned int align) 2843 { 2844 WARN_ON_ONCE(!is_power_of_2(align)); 2845 return __netdev_alloc_frag_align(fragsz, -align); 2846 } 2847 2848 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2849 gfp_t gfp_mask); 2850 2851 /** 2852 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2853 * @dev: network device to receive on 2854 * @length: length to allocate 2855 * 2856 * Allocate a new &sk_buff and assign it a usage count of one. The 2857 * buffer has unspecified headroom built in. Users should allocate 2858 * the headroom they think they need without accounting for the 2859 * built in space. The built in space is used for optimisations. 2860 * 2861 * %NULL is returned if there is no free memory. Although this function 2862 * allocates memory it can be called from an interrupt. 2863 */ 2864 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2865 unsigned int length) 2866 { 2867 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2868 } 2869 2870 /* legacy helper around __netdev_alloc_skb() */ 2871 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2872 gfp_t gfp_mask) 2873 { 2874 return __netdev_alloc_skb(NULL, length, gfp_mask); 2875 } 2876 2877 /* legacy helper around netdev_alloc_skb() */ 2878 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2879 { 2880 return netdev_alloc_skb(NULL, length); 2881 } 2882 2883 2884 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2885 unsigned int length, gfp_t gfp) 2886 { 2887 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2888 2889 if (NET_IP_ALIGN && skb) 2890 skb_reserve(skb, NET_IP_ALIGN); 2891 return skb; 2892 } 2893 2894 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2895 unsigned int length) 2896 { 2897 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2898 } 2899 2900 static inline void skb_free_frag(void *addr) 2901 { 2902 page_frag_free(addr); 2903 } 2904 2905 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2906 2907 static inline void *napi_alloc_frag(unsigned int fragsz) 2908 { 2909 return __napi_alloc_frag_align(fragsz, ~0u); 2910 } 2911 2912 static inline void *napi_alloc_frag_align(unsigned int fragsz, 2913 unsigned int align) 2914 { 2915 WARN_ON_ONCE(!is_power_of_2(align)); 2916 return __napi_alloc_frag_align(fragsz, -align); 2917 } 2918 2919 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2920 unsigned int length, gfp_t gfp_mask); 2921 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2922 unsigned int length) 2923 { 2924 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2925 } 2926 void napi_consume_skb(struct sk_buff *skb, int budget); 2927 2928 void napi_skb_free_stolen_head(struct sk_buff *skb); 2929 void __kfree_skb_defer(struct sk_buff *skb); 2930 2931 /** 2932 * __dev_alloc_pages - allocate page for network Rx 2933 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2934 * @order: size of the allocation 2935 * 2936 * Allocate a new page. 2937 * 2938 * %NULL is returned if there is no free memory. 2939 */ 2940 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2941 unsigned int order) 2942 { 2943 /* This piece of code contains several assumptions. 2944 * 1. This is for device Rx, therefor a cold page is preferred. 2945 * 2. The expectation is the user wants a compound page. 2946 * 3. If requesting a order 0 page it will not be compound 2947 * due to the check to see if order has a value in prep_new_page 2948 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2949 * code in gfp_to_alloc_flags that should be enforcing this. 2950 */ 2951 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 2952 2953 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2954 } 2955 2956 static inline struct page *dev_alloc_pages(unsigned int order) 2957 { 2958 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 2959 } 2960 2961 /** 2962 * __dev_alloc_page - allocate a page for network Rx 2963 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2964 * 2965 * Allocate a new page. 2966 * 2967 * %NULL is returned if there is no free memory. 2968 */ 2969 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2970 { 2971 return __dev_alloc_pages(gfp_mask, 0); 2972 } 2973 2974 static inline struct page *dev_alloc_page(void) 2975 { 2976 return dev_alloc_pages(0); 2977 } 2978 2979 /** 2980 * dev_page_is_reusable - check whether a page can be reused for network Rx 2981 * @page: the page to test 2982 * 2983 * A page shouldn't be considered for reusing/recycling if it was allocated 2984 * under memory pressure or at a distant memory node. 2985 * 2986 * Returns false if this page should be returned to page allocator, true 2987 * otherwise. 2988 */ 2989 static inline bool dev_page_is_reusable(const struct page *page) 2990 { 2991 return likely(page_to_nid(page) == numa_mem_id() && 2992 !page_is_pfmemalloc(page)); 2993 } 2994 2995 /** 2996 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2997 * @page: The page that was allocated from skb_alloc_page 2998 * @skb: The skb that may need pfmemalloc set 2999 */ 3000 static inline void skb_propagate_pfmemalloc(const struct page *page, 3001 struct sk_buff *skb) 3002 { 3003 if (page_is_pfmemalloc(page)) 3004 skb->pfmemalloc = true; 3005 } 3006 3007 /** 3008 * skb_frag_off() - Returns the offset of a skb fragment 3009 * @frag: the paged fragment 3010 */ 3011 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3012 { 3013 return frag->bv_offset; 3014 } 3015 3016 /** 3017 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3018 * @frag: skb fragment 3019 * @delta: value to add 3020 */ 3021 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3022 { 3023 frag->bv_offset += delta; 3024 } 3025 3026 /** 3027 * skb_frag_off_set() - Sets the offset of a skb fragment 3028 * @frag: skb fragment 3029 * @offset: offset of fragment 3030 */ 3031 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3032 { 3033 frag->bv_offset = offset; 3034 } 3035 3036 /** 3037 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3038 * @fragto: skb fragment where offset is set 3039 * @fragfrom: skb fragment offset is copied from 3040 */ 3041 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3042 const skb_frag_t *fragfrom) 3043 { 3044 fragto->bv_offset = fragfrom->bv_offset; 3045 } 3046 3047 /** 3048 * skb_frag_page - retrieve the page referred to by a paged fragment 3049 * @frag: the paged fragment 3050 * 3051 * Returns the &struct page associated with @frag. 3052 */ 3053 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3054 { 3055 return frag->bv_page; 3056 } 3057 3058 /** 3059 * __skb_frag_ref - take an addition reference on a paged fragment. 3060 * @frag: the paged fragment 3061 * 3062 * Takes an additional reference on the paged fragment @frag. 3063 */ 3064 static inline void __skb_frag_ref(skb_frag_t *frag) 3065 { 3066 get_page(skb_frag_page(frag)); 3067 } 3068 3069 /** 3070 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3071 * @skb: the buffer 3072 * @f: the fragment offset. 3073 * 3074 * Takes an additional reference on the @f'th paged fragment of @skb. 3075 */ 3076 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3077 { 3078 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3079 } 3080 3081 /** 3082 * __skb_frag_unref - release a reference on a paged fragment. 3083 * @frag: the paged fragment 3084 * 3085 * Releases a reference on the paged fragment @frag. 3086 */ 3087 static inline void __skb_frag_unref(skb_frag_t *frag) 3088 { 3089 put_page(skb_frag_page(frag)); 3090 } 3091 3092 /** 3093 * skb_frag_unref - release a reference on a paged fragment of an skb. 3094 * @skb: the buffer 3095 * @f: the fragment offset 3096 * 3097 * Releases a reference on the @f'th paged fragment of @skb. 3098 */ 3099 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3100 { 3101 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 3102 } 3103 3104 /** 3105 * skb_frag_address - gets the address of the data contained in a paged fragment 3106 * @frag: the paged fragment buffer 3107 * 3108 * Returns the address of the data within @frag. The page must already 3109 * be mapped. 3110 */ 3111 static inline void *skb_frag_address(const skb_frag_t *frag) 3112 { 3113 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3114 } 3115 3116 /** 3117 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3118 * @frag: the paged fragment buffer 3119 * 3120 * Returns the address of the data within @frag. Checks that the page 3121 * is mapped and returns %NULL otherwise. 3122 */ 3123 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3124 { 3125 void *ptr = page_address(skb_frag_page(frag)); 3126 if (unlikely(!ptr)) 3127 return NULL; 3128 3129 return ptr + skb_frag_off(frag); 3130 } 3131 3132 /** 3133 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3134 * @fragto: skb fragment where page is set 3135 * @fragfrom: skb fragment page is copied from 3136 */ 3137 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3138 const skb_frag_t *fragfrom) 3139 { 3140 fragto->bv_page = fragfrom->bv_page; 3141 } 3142 3143 /** 3144 * __skb_frag_set_page - sets the page contained in a paged fragment 3145 * @frag: the paged fragment 3146 * @page: the page to set 3147 * 3148 * Sets the fragment @frag to contain @page. 3149 */ 3150 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3151 { 3152 frag->bv_page = page; 3153 } 3154 3155 /** 3156 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3157 * @skb: the buffer 3158 * @f: the fragment offset 3159 * @page: the page to set 3160 * 3161 * Sets the @f'th fragment of @skb to contain @page. 3162 */ 3163 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3164 struct page *page) 3165 { 3166 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3167 } 3168 3169 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3170 3171 /** 3172 * skb_frag_dma_map - maps a paged fragment via the DMA API 3173 * @dev: the device to map the fragment to 3174 * @frag: the paged fragment to map 3175 * @offset: the offset within the fragment (starting at the 3176 * fragment's own offset) 3177 * @size: the number of bytes to map 3178 * @dir: the direction of the mapping (``PCI_DMA_*``) 3179 * 3180 * Maps the page associated with @frag to @device. 3181 */ 3182 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3183 const skb_frag_t *frag, 3184 size_t offset, size_t size, 3185 enum dma_data_direction dir) 3186 { 3187 return dma_map_page(dev, skb_frag_page(frag), 3188 skb_frag_off(frag) + offset, size, dir); 3189 } 3190 3191 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3192 gfp_t gfp_mask) 3193 { 3194 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3195 } 3196 3197 3198 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3199 gfp_t gfp_mask) 3200 { 3201 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3202 } 3203 3204 3205 /** 3206 * skb_clone_writable - is the header of a clone writable 3207 * @skb: buffer to check 3208 * @len: length up to which to write 3209 * 3210 * Returns true if modifying the header part of the cloned buffer 3211 * does not requires the data to be copied. 3212 */ 3213 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3214 { 3215 return !skb_header_cloned(skb) && 3216 skb_headroom(skb) + len <= skb->hdr_len; 3217 } 3218 3219 static inline int skb_try_make_writable(struct sk_buff *skb, 3220 unsigned int write_len) 3221 { 3222 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3223 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3224 } 3225 3226 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3227 int cloned) 3228 { 3229 int delta = 0; 3230 3231 if (headroom > skb_headroom(skb)) 3232 delta = headroom - skb_headroom(skb); 3233 3234 if (delta || cloned) 3235 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3236 GFP_ATOMIC); 3237 return 0; 3238 } 3239 3240 /** 3241 * skb_cow - copy header of skb when it is required 3242 * @skb: buffer to cow 3243 * @headroom: needed headroom 3244 * 3245 * If the skb passed lacks sufficient headroom or its data part 3246 * is shared, data is reallocated. If reallocation fails, an error 3247 * is returned and original skb is not changed. 3248 * 3249 * The result is skb with writable area skb->head...skb->tail 3250 * and at least @headroom of space at head. 3251 */ 3252 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3253 { 3254 return __skb_cow(skb, headroom, skb_cloned(skb)); 3255 } 3256 3257 /** 3258 * skb_cow_head - skb_cow but only making the head writable 3259 * @skb: buffer to cow 3260 * @headroom: needed headroom 3261 * 3262 * This function is identical to skb_cow except that we replace the 3263 * skb_cloned check by skb_header_cloned. It should be used when 3264 * you only need to push on some header and do not need to modify 3265 * the data. 3266 */ 3267 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3268 { 3269 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3270 } 3271 3272 /** 3273 * skb_padto - pad an skbuff up to a minimal size 3274 * @skb: buffer to pad 3275 * @len: minimal length 3276 * 3277 * Pads up a buffer to ensure the trailing bytes exist and are 3278 * blanked. If the buffer already contains sufficient data it 3279 * is untouched. Otherwise it is extended. Returns zero on 3280 * success. The skb is freed on error. 3281 */ 3282 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3283 { 3284 unsigned int size = skb->len; 3285 if (likely(size >= len)) 3286 return 0; 3287 return skb_pad(skb, len - size); 3288 } 3289 3290 /** 3291 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3292 * @skb: buffer to pad 3293 * @len: minimal length 3294 * @free_on_error: free buffer on error 3295 * 3296 * Pads up a buffer to ensure the trailing bytes exist and are 3297 * blanked. If the buffer already contains sufficient data it 3298 * is untouched. Otherwise it is extended. Returns zero on 3299 * success. The skb is freed on error if @free_on_error is true. 3300 */ 3301 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3302 unsigned int len, 3303 bool free_on_error) 3304 { 3305 unsigned int size = skb->len; 3306 3307 if (unlikely(size < len)) { 3308 len -= size; 3309 if (__skb_pad(skb, len, free_on_error)) 3310 return -ENOMEM; 3311 __skb_put(skb, len); 3312 } 3313 return 0; 3314 } 3315 3316 /** 3317 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3318 * @skb: buffer to pad 3319 * @len: minimal length 3320 * 3321 * Pads up a buffer to ensure the trailing bytes exist and are 3322 * blanked. If the buffer already contains sufficient data it 3323 * is untouched. Otherwise it is extended. Returns zero on 3324 * success. The skb is freed on error. 3325 */ 3326 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3327 { 3328 return __skb_put_padto(skb, len, true); 3329 } 3330 3331 static inline int skb_add_data(struct sk_buff *skb, 3332 struct iov_iter *from, int copy) 3333 { 3334 const int off = skb->len; 3335 3336 if (skb->ip_summed == CHECKSUM_NONE) { 3337 __wsum csum = 0; 3338 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3339 &csum, from)) { 3340 skb->csum = csum_block_add(skb->csum, csum, off); 3341 return 0; 3342 } 3343 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3344 return 0; 3345 3346 __skb_trim(skb, off); 3347 return -EFAULT; 3348 } 3349 3350 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3351 const struct page *page, int off) 3352 { 3353 if (skb_zcopy(skb)) 3354 return false; 3355 if (i) { 3356 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3357 3358 return page == skb_frag_page(frag) && 3359 off == skb_frag_off(frag) + skb_frag_size(frag); 3360 } 3361 return false; 3362 } 3363 3364 static inline int __skb_linearize(struct sk_buff *skb) 3365 { 3366 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3367 } 3368 3369 /** 3370 * skb_linearize - convert paged skb to linear one 3371 * @skb: buffer to linarize 3372 * 3373 * If there is no free memory -ENOMEM is returned, otherwise zero 3374 * is returned and the old skb data released. 3375 */ 3376 static inline int skb_linearize(struct sk_buff *skb) 3377 { 3378 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3379 } 3380 3381 /** 3382 * skb_has_shared_frag - can any frag be overwritten 3383 * @skb: buffer to test 3384 * 3385 * Return true if the skb has at least one frag that might be modified 3386 * by an external entity (as in vmsplice()/sendfile()) 3387 */ 3388 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3389 { 3390 return skb_is_nonlinear(skb) && 3391 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3392 } 3393 3394 /** 3395 * skb_linearize_cow - make sure skb is linear and writable 3396 * @skb: buffer to process 3397 * 3398 * If there is no free memory -ENOMEM is returned, otherwise zero 3399 * is returned and the old skb data released. 3400 */ 3401 static inline int skb_linearize_cow(struct sk_buff *skb) 3402 { 3403 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3404 __skb_linearize(skb) : 0; 3405 } 3406 3407 static __always_inline void 3408 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3409 unsigned int off) 3410 { 3411 if (skb->ip_summed == CHECKSUM_COMPLETE) 3412 skb->csum = csum_block_sub(skb->csum, 3413 csum_partial(start, len, 0), off); 3414 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3415 skb_checksum_start_offset(skb) < 0) 3416 skb->ip_summed = CHECKSUM_NONE; 3417 } 3418 3419 /** 3420 * skb_postpull_rcsum - update checksum for received skb after pull 3421 * @skb: buffer to update 3422 * @start: start of data before pull 3423 * @len: length of data pulled 3424 * 3425 * After doing a pull on a received packet, you need to call this to 3426 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3427 * CHECKSUM_NONE so that it can be recomputed from scratch. 3428 */ 3429 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3430 const void *start, unsigned int len) 3431 { 3432 __skb_postpull_rcsum(skb, start, len, 0); 3433 } 3434 3435 static __always_inline void 3436 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3437 unsigned int off) 3438 { 3439 if (skb->ip_summed == CHECKSUM_COMPLETE) 3440 skb->csum = csum_block_add(skb->csum, 3441 csum_partial(start, len, 0), off); 3442 } 3443 3444 /** 3445 * skb_postpush_rcsum - update checksum for received skb after push 3446 * @skb: buffer to update 3447 * @start: start of data after push 3448 * @len: length of data pushed 3449 * 3450 * After doing a push on a received packet, you need to call this to 3451 * update the CHECKSUM_COMPLETE checksum. 3452 */ 3453 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3454 const void *start, unsigned int len) 3455 { 3456 __skb_postpush_rcsum(skb, start, len, 0); 3457 } 3458 3459 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3460 3461 /** 3462 * skb_push_rcsum - push skb and update receive checksum 3463 * @skb: buffer to update 3464 * @len: length of data pulled 3465 * 3466 * This function performs an skb_push on the packet and updates 3467 * the CHECKSUM_COMPLETE checksum. It should be used on 3468 * receive path processing instead of skb_push unless you know 3469 * that the checksum difference is zero (e.g., a valid IP header) 3470 * or you are setting ip_summed to CHECKSUM_NONE. 3471 */ 3472 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3473 { 3474 skb_push(skb, len); 3475 skb_postpush_rcsum(skb, skb->data, len); 3476 return skb->data; 3477 } 3478 3479 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3480 /** 3481 * pskb_trim_rcsum - trim received skb and update checksum 3482 * @skb: buffer to trim 3483 * @len: new length 3484 * 3485 * This is exactly the same as pskb_trim except that it ensures the 3486 * checksum of received packets are still valid after the operation. 3487 * It can change skb pointers. 3488 */ 3489 3490 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3491 { 3492 if (likely(len >= skb->len)) 3493 return 0; 3494 return pskb_trim_rcsum_slow(skb, len); 3495 } 3496 3497 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3498 { 3499 if (skb->ip_summed == CHECKSUM_COMPLETE) 3500 skb->ip_summed = CHECKSUM_NONE; 3501 __skb_trim(skb, len); 3502 return 0; 3503 } 3504 3505 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3506 { 3507 if (skb->ip_summed == CHECKSUM_COMPLETE) 3508 skb->ip_summed = CHECKSUM_NONE; 3509 return __skb_grow(skb, len); 3510 } 3511 3512 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3513 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3514 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3515 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3516 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3517 3518 #define skb_queue_walk(queue, skb) \ 3519 for (skb = (queue)->next; \ 3520 skb != (struct sk_buff *)(queue); \ 3521 skb = skb->next) 3522 3523 #define skb_queue_walk_safe(queue, skb, tmp) \ 3524 for (skb = (queue)->next, tmp = skb->next; \ 3525 skb != (struct sk_buff *)(queue); \ 3526 skb = tmp, tmp = skb->next) 3527 3528 #define skb_queue_walk_from(queue, skb) \ 3529 for (; skb != (struct sk_buff *)(queue); \ 3530 skb = skb->next) 3531 3532 #define skb_rbtree_walk(skb, root) \ 3533 for (skb = skb_rb_first(root); skb != NULL; \ 3534 skb = skb_rb_next(skb)) 3535 3536 #define skb_rbtree_walk_from(skb) \ 3537 for (; skb != NULL; \ 3538 skb = skb_rb_next(skb)) 3539 3540 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3541 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3542 skb = tmp) 3543 3544 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3545 for (tmp = skb->next; \ 3546 skb != (struct sk_buff *)(queue); \ 3547 skb = tmp, tmp = skb->next) 3548 3549 #define skb_queue_reverse_walk(queue, skb) \ 3550 for (skb = (queue)->prev; \ 3551 skb != (struct sk_buff *)(queue); \ 3552 skb = skb->prev) 3553 3554 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3555 for (skb = (queue)->prev, tmp = skb->prev; \ 3556 skb != (struct sk_buff *)(queue); \ 3557 skb = tmp, tmp = skb->prev) 3558 3559 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3560 for (tmp = skb->prev; \ 3561 skb != (struct sk_buff *)(queue); \ 3562 skb = tmp, tmp = skb->prev) 3563 3564 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3565 { 3566 return skb_shinfo(skb)->frag_list != NULL; 3567 } 3568 3569 static inline void skb_frag_list_init(struct sk_buff *skb) 3570 { 3571 skb_shinfo(skb)->frag_list = NULL; 3572 } 3573 3574 #define skb_walk_frags(skb, iter) \ 3575 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3576 3577 3578 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3579 int *err, long *timeo_p, 3580 const struct sk_buff *skb); 3581 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3582 struct sk_buff_head *queue, 3583 unsigned int flags, 3584 int *off, int *err, 3585 struct sk_buff **last); 3586 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3587 struct sk_buff_head *queue, 3588 unsigned int flags, int *off, int *err, 3589 struct sk_buff **last); 3590 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3591 struct sk_buff_head *sk_queue, 3592 unsigned int flags, int *off, int *err); 3593 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3594 int *err); 3595 __poll_t datagram_poll(struct file *file, struct socket *sock, 3596 struct poll_table_struct *wait); 3597 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3598 struct iov_iter *to, int size); 3599 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3600 struct msghdr *msg, int size) 3601 { 3602 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3603 } 3604 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3605 struct msghdr *msg); 3606 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3607 struct iov_iter *to, int len, 3608 struct ahash_request *hash); 3609 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3610 struct iov_iter *from, int len); 3611 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3612 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3613 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3614 static inline void skb_free_datagram_locked(struct sock *sk, 3615 struct sk_buff *skb) 3616 { 3617 __skb_free_datagram_locked(sk, skb, 0); 3618 } 3619 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3620 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3621 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3622 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3623 int len); 3624 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3625 struct pipe_inode_info *pipe, unsigned int len, 3626 unsigned int flags); 3627 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3628 int len); 3629 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len); 3630 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3631 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3632 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3633 int len, int hlen); 3634 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3635 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3636 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3637 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3638 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3639 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3640 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3641 unsigned int offset); 3642 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3643 int skb_ensure_writable(struct sk_buff *skb, int write_len); 3644 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3645 int skb_vlan_pop(struct sk_buff *skb); 3646 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3647 int skb_eth_pop(struct sk_buff *skb); 3648 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3649 const unsigned char *src); 3650 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3651 int mac_len, bool ethernet); 3652 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3653 bool ethernet); 3654 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3655 int skb_mpls_dec_ttl(struct sk_buff *skb); 3656 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3657 gfp_t gfp); 3658 3659 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3660 { 3661 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3662 } 3663 3664 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3665 { 3666 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3667 } 3668 3669 struct skb_checksum_ops { 3670 __wsum (*update)(const void *mem, int len, __wsum wsum); 3671 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3672 }; 3673 3674 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3675 3676 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3677 __wsum csum, const struct skb_checksum_ops *ops); 3678 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3679 __wsum csum); 3680 3681 static inline void * __must_check 3682 __skb_header_pointer(const struct sk_buff *skb, int offset, int len, 3683 const void *data, int hlen, void *buffer) 3684 { 3685 if (likely(hlen - offset >= len)) 3686 return (void *)data + offset; 3687 3688 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) 3689 return NULL; 3690 3691 return buffer; 3692 } 3693 3694 static inline void * __must_check 3695 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3696 { 3697 return __skb_header_pointer(skb, offset, len, skb->data, 3698 skb_headlen(skb), buffer); 3699 } 3700 3701 /** 3702 * skb_needs_linearize - check if we need to linearize a given skb 3703 * depending on the given device features. 3704 * @skb: socket buffer to check 3705 * @features: net device features 3706 * 3707 * Returns true if either: 3708 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3709 * 2. skb is fragmented and the device does not support SG. 3710 */ 3711 static inline bool skb_needs_linearize(struct sk_buff *skb, 3712 netdev_features_t features) 3713 { 3714 return skb_is_nonlinear(skb) && 3715 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3716 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3717 } 3718 3719 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3720 void *to, 3721 const unsigned int len) 3722 { 3723 memcpy(to, skb->data, len); 3724 } 3725 3726 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3727 const int offset, void *to, 3728 const unsigned int len) 3729 { 3730 memcpy(to, skb->data + offset, len); 3731 } 3732 3733 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3734 const void *from, 3735 const unsigned int len) 3736 { 3737 memcpy(skb->data, from, len); 3738 } 3739 3740 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3741 const int offset, 3742 const void *from, 3743 const unsigned int len) 3744 { 3745 memcpy(skb->data + offset, from, len); 3746 } 3747 3748 void skb_init(void); 3749 3750 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3751 { 3752 return skb->tstamp; 3753 } 3754 3755 /** 3756 * skb_get_timestamp - get timestamp from a skb 3757 * @skb: skb to get stamp from 3758 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3759 * 3760 * Timestamps are stored in the skb as offsets to a base timestamp. 3761 * This function converts the offset back to a struct timeval and stores 3762 * it in stamp. 3763 */ 3764 static inline void skb_get_timestamp(const struct sk_buff *skb, 3765 struct __kernel_old_timeval *stamp) 3766 { 3767 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3768 } 3769 3770 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3771 struct __kernel_sock_timeval *stamp) 3772 { 3773 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3774 3775 stamp->tv_sec = ts.tv_sec; 3776 stamp->tv_usec = ts.tv_nsec / 1000; 3777 } 3778 3779 static inline void skb_get_timestampns(const struct sk_buff *skb, 3780 struct __kernel_old_timespec *stamp) 3781 { 3782 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3783 3784 stamp->tv_sec = ts.tv_sec; 3785 stamp->tv_nsec = ts.tv_nsec; 3786 } 3787 3788 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3789 struct __kernel_timespec *stamp) 3790 { 3791 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3792 3793 stamp->tv_sec = ts.tv_sec; 3794 stamp->tv_nsec = ts.tv_nsec; 3795 } 3796 3797 static inline void __net_timestamp(struct sk_buff *skb) 3798 { 3799 skb->tstamp = ktime_get_real(); 3800 } 3801 3802 static inline ktime_t net_timedelta(ktime_t t) 3803 { 3804 return ktime_sub(ktime_get_real(), t); 3805 } 3806 3807 static inline ktime_t net_invalid_timestamp(void) 3808 { 3809 return 0; 3810 } 3811 3812 static inline u8 skb_metadata_len(const struct sk_buff *skb) 3813 { 3814 return skb_shinfo(skb)->meta_len; 3815 } 3816 3817 static inline void *skb_metadata_end(const struct sk_buff *skb) 3818 { 3819 return skb_mac_header(skb); 3820 } 3821 3822 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3823 const struct sk_buff *skb_b, 3824 u8 meta_len) 3825 { 3826 const void *a = skb_metadata_end(skb_a); 3827 const void *b = skb_metadata_end(skb_b); 3828 /* Using more efficient varaiant than plain call to memcmp(). */ 3829 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3830 u64 diffs = 0; 3831 3832 switch (meta_len) { 3833 #define __it(x, op) (x -= sizeof(u##op)) 3834 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3835 case 32: diffs |= __it_diff(a, b, 64); 3836 fallthrough; 3837 case 24: diffs |= __it_diff(a, b, 64); 3838 fallthrough; 3839 case 16: diffs |= __it_diff(a, b, 64); 3840 fallthrough; 3841 case 8: diffs |= __it_diff(a, b, 64); 3842 break; 3843 case 28: diffs |= __it_diff(a, b, 64); 3844 fallthrough; 3845 case 20: diffs |= __it_diff(a, b, 64); 3846 fallthrough; 3847 case 12: diffs |= __it_diff(a, b, 64); 3848 fallthrough; 3849 case 4: diffs |= __it_diff(a, b, 32); 3850 break; 3851 } 3852 return diffs; 3853 #else 3854 return memcmp(a - meta_len, b - meta_len, meta_len); 3855 #endif 3856 } 3857 3858 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3859 const struct sk_buff *skb_b) 3860 { 3861 u8 len_a = skb_metadata_len(skb_a); 3862 u8 len_b = skb_metadata_len(skb_b); 3863 3864 if (!(len_a | len_b)) 3865 return false; 3866 3867 return len_a != len_b ? 3868 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3869 } 3870 3871 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3872 { 3873 skb_shinfo(skb)->meta_len = meta_len; 3874 } 3875 3876 static inline void skb_metadata_clear(struct sk_buff *skb) 3877 { 3878 skb_metadata_set(skb, 0); 3879 } 3880 3881 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3882 3883 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3884 3885 void skb_clone_tx_timestamp(struct sk_buff *skb); 3886 bool skb_defer_rx_timestamp(struct sk_buff *skb); 3887 3888 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3889 3890 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3891 { 3892 } 3893 3894 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3895 { 3896 return false; 3897 } 3898 3899 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3900 3901 /** 3902 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3903 * 3904 * PHY drivers may accept clones of transmitted packets for 3905 * timestamping via their phy_driver.txtstamp method. These drivers 3906 * must call this function to return the skb back to the stack with a 3907 * timestamp. 3908 * 3909 * @skb: clone of the original outgoing packet 3910 * @hwtstamps: hardware time stamps 3911 * 3912 */ 3913 void skb_complete_tx_timestamp(struct sk_buff *skb, 3914 struct skb_shared_hwtstamps *hwtstamps); 3915 3916 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 3917 struct skb_shared_hwtstamps *hwtstamps, 3918 struct sock *sk, int tstype); 3919 3920 /** 3921 * skb_tstamp_tx - queue clone of skb with send time stamps 3922 * @orig_skb: the original outgoing packet 3923 * @hwtstamps: hardware time stamps, may be NULL if not available 3924 * 3925 * If the skb has a socket associated, then this function clones the 3926 * skb (thus sharing the actual data and optional structures), stores 3927 * the optional hardware time stamping information (if non NULL) or 3928 * generates a software time stamp (otherwise), then queues the clone 3929 * to the error queue of the socket. Errors are silently ignored. 3930 */ 3931 void skb_tstamp_tx(struct sk_buff *orig_skb, 3932 struct skb_shared_hwtstamps *hwtstamps); 3933 3934 /** 3935 * skb_tx_timestamp() - Driver hook for transmit timestamping 3936 * 3937 * Ethernet MAC Drivers should call this function in their hard_xmit() 3938 * function immediately before giving the sk_buff to the MAC hardware. 3939 * 3940 * Specifically, one should make absolutely sure that this function is 3941 * called before TX completion of this packet can trigger. Otherwise 3942 * the packet could potentially already be freed. 3943 * 3944 * @skb: A socket buffer. 3945 */ 3946 static inline void skb_tx_timestamp(struct sk_buff *skb) 3947 { 3948 skb_clone_tx_timestamp(skb); 3949 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 3950 skb_tstamp_tx(skb, NULL); 3951 } 3952 3953 /** 3954 * skb_complete_wifi_ack - deliver skb with wifi status 3955 * 3956 * @skb: the original outgoing packet 3957 * @acked: ack status 3958 * 3959 */ 3960 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 3961 3962 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 3963 __sum16 __skb_checksum_complete(struct sk_buff *skb); 3964 3965 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 3966 { 3967 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 3968 skb->csum_valid || 3969 (skb->ip_summed == CHECKSUM_PARTIAL && 3970 skb_checksum_start_offset(skb) >= 0)); 3971 } 3972 3973 /** 3974 * skb_checksum_complete - Calculate checksum of an entire packet 3975 * @skb: packet to process 3976 * 3977 * This function calculates the checksum over the entire packet plus 3978 * the value of skb->csum. The latter can be used to supply the 3979 * checksum of a pseudo header as used by TCP/UDP. It returns the 3980 * checksum. 3981 * 3982 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 3983 * this function can be used to verify that checksum on received 3984 * packets. In that case the function should return zero if the 3985 * checksum is correct. In particular, this function will return zero 3986 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 3987 * hardware has already verified the correctness of the checksum. 3988 */ 3989 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 3990 { 3991 return skb_csum_unnecessary(skb) ? 3992 0 : __skb_checksum_complete(skb); 3993 } 3994 3995 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 3996 { 3997 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 3998 if (skb->csum_level == 0) 3999 skb->ip_summed = CHECKSUM_NONE; 4000 else 4001 skb->csum_level--; 4002 } 4003 } 4004 4005 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4006 { 4007 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4008 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4009 skb->csum_level++; 4010 } else if (skb->ip_summed == CHECKSUM_NONE) { 4011 skb->ip_summed = CHECKSUM_UNNECESSARY; 4012 skb->csum_level = 0; 4013 } 4014 } 4015 4016 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4017 { 4018 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4019 skb->ip_summed = CHECKSUM_NONE; 4020 skb->csum_level = 0; 4021 } 4022 } 4023 4024 /* Check if we need to perform checksum complete validation. 4025 * 4026 * Returns true if checksum complete is needed, false otherwise 4027 * (either checksum is unnecessary or zero checksum is allowed). 4028 */ 4029 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4030 bool zero_okay, 4031 __sum16 check) 4032 { 4033 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4034 skb->csum_valid = 1; 4035 __skb_decr_checksum_unnecessary(skb); 4036 return false; 4037 } 4038 4039 return true; 4040 } 4041 4042 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4043 * in checksum_init. 4044 */ 4045 #define CHECKSUM_BREAK 76 4046 4047 /* Unset checksum-complete 4048 * 4049 * Unset checksum complete can be done when packet is being modified 4050 * (uncompressed for instance) and checksum-complete value is 4051 * invalidated. 4052 */ 4053 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4054 { 4055 if (skb->ip_summed == CHECKSUM_COMPLETE) 4056 skb->ip_summed = CHECKSUM_NONE; 4057 } 4058 4059 /* Validate (init) checksum based on checksum complete. 4060 * 4061 * Return values: 4062 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4063 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4064 * checksum is stored in skb->csum for use in __skb_checksum_complete 4065 * non-zero: value of invalid checksum 4066 * 4067 */ 4068 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4069 bool complete, 4070 __wsum psum) 4071 { 4072 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4073 if (!csum_fold(csum_add(psum, skb->csum))) { 4074 skb->csum_valid = 1; 4075 return 0; 4076 } 4077 } 4078 4079 skb->csum = psum; 4080 4081 if (complete || skb->len <= CHECKSUM_BREAK) { 4082 __sum16 csum; 4083 4084 csum = __skb_checksum_complete(skb); 4085 skb->csum_valid = !csum; 4086 return csum; 4087 } 4088 4089 return 0; 4090 } 4091 4092 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4093 { 4094 return 0; 4095 } 4096 4097 /* Perform checksum validate (init). Note that this is a macro since we only 4098 * want to calculate the pseudo header which is an input function if necessary. 4099 * First we try to validate without any computation (checksum unnecessary) and 4100 * then calculate based on checksum complete calling the function to compute 4101 * pseudo header. 4102 * 4103 * Return values: 4104 * 0: checksum is validated or try to in skb_checksum_complete 4105 * non-zero: value of invalid checksum 4106 */ 4107 #define __skb_checksum_validate(skb, proto, complete, \ 4108 zero_okay, check, compute_pseudo) \ 4109 ({ \ 4110 __sum16 __ret = 0; \ 4111 skb->csum_valid = 0; \ 4112 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4113 __ret = __skb_checksum_validate_complete(skb, \ 4114 complete, compute_pseudo(skb, proto)); \ 4115 __ret; \ 4116 }) 4117 4118 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4119 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4120 4121 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4122 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4123 4124 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4125 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4126 4127 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4128 compute_pseudo) \ 4129 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4130 4131 #define skb_checksum_simple_validate(skb) \ 4132 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4133 4134 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4135 { 4136 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4137 } 4138 4139 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4140 { 4141 skb->csum = ~pseudo; 4142 skb->ip_summed = CHECKSUM_COMPLETE; 4143 } 4144 4145 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4146 do { \ 4147 if (__skb_checksum_convert_check(skb)) \ 4148 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4149 } while (0) 4150 4151 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4152 u16 start, u16 offset) 4153 { 4154 skb->ip_summed = CHECKSUM_PARTIAL; 4155 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4156 skb->csum_offset = offset - start; 4157 } 4158 4159 /* Update skbuf and packet to reflect the remote checksum offload operation. 4160 * When called, ptr indicates the starting point for skb->csum when 4161 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4162 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4163 */ 4164 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4165 int start, int offset, bool nopartial) 4166 { 4167 __wsum delta; 4168 4169 if (!nopartial) { 4170 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4171 return; 4172 } 4173 4174 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4175 __skb_checksum_complete(skb); 4176 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4177 } 4178 4179 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4180 4181 /* Adjust skb->csum since we changed the packet */ 4182 skb->csum = csum_add(skb->csum, delta); 4183 } 4184 4185 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4186 { 4187 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4188 return (void *)(skb->_nfct & NFCT_PTRMASK); 4189 #else 4190 return NULL; 4191 #endif 4192 } 4193 4194 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4195 { 4196 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4197 return skb->_nfct; 4198 #else 4199 return 0UL; 4200 #endif 4201 } 4202 4203 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4204 { 4205 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4206 skb->_nfct = nfct; 4207 #endif 4208 } 4209 4210 #ifdef CONFIG_SKB_EXTENSIONS 4211 enum skb_ext_id { 4212 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4213 SKB_EXT_BRIDGE_NF, 4214 #endif 4215 #ifdef CONFIG_XFRM 4216 SKB_EXT_SEC_PATH, 4217 #endif 4218 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4219 TC_SKB_EXT, 4220 #endif 4221 #if IS_ENABLED(CONFIG_MPTCP) 4222 SKB_EXT_MPTCP, 4223 #endif 4224 SKB_EXT_NUM, /* must be last */ 4225 }; 4226 4227 /** 4228 * struct skb_ext - sk_buff extensions 4229 * @refcnt: 1 on allocation, deallocated on 0 4230 * @offset: offset to add to @data to obtain extension address 4231 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4232 * @data: start of extension data, variable sized 4233 * 4234 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4235 * to use 'u8' types while allowing up to 2kb worth of extension data. 4236 */ 4237 struct skb_ext { 4238 refcount_t refcnt; 4239 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4240 u8 chunks; /* same */ 4241 char data[] __aligned(8); 4242 }; 4243 4244 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4245 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4246 struct skb_ext *ext); 4247 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4248 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4249 void __skb_ext_put(struct skb_ext *ext); 4250 4251 static inline void skb_ext_put(struct sk_buff *skb) 4252 { 4253 if (skb->active_extensions) 4254 __skb_ext_put(skb->extensions); 4255 } 4256 4257 static inline void __skb_ext_copy(struct sk_buff *dst, 4258 const struct sk_buff *src) 4259 { 4260 dst->active_extensions = src->active_extensions; 4261 4262 if (src->active_extensions) { 4263 struct skb_ext *ext = src->extensions; 4264 4265 refcount_inc(&ext->refcnt); 4266 dst->extensions = ext; 4267 } 4268 } 4269 4270 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4271 { 4272 skb_ext_put(dst); 4273 __skb_ext_copy(dst, src); 4274 } 4275 4276 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4277 { 4278 return !!ext->offset[i]; 4279 } 4280 4281 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4282 { 4283 return skb->active_extensions & (1 << id); 4284 } 4285 4286 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4287 { 4288 if (skb_ext_exist(skb, id)) 4289 __skb_ext_del(skb, id); 4290 } 4291 4292 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4293 { 4294 if (skb_ext_exist(skb, id)) { 4295 struct skb_ext *ext = skb->extensions; 4296 4297 return (void *)ext + (ext->offset[id] << 3); 4298 } 4299 4300 return NULL; 4301 } 4302 4303 static inline void skb_ext_reset(struct sk_buff *skb) 4304 { 4305 if (unlikely(skb->active_extensions)) { 4306 __skb_ext_put(skb->extensions); 4307 skb->active_extensions = 0; 4308 } 4309 } 4310 4311 static inline bool skb_has_extensions(struct sk_buff *skb) 4312 { 4313 return unlikely(skb->active_extensions); 4314 } 4315 #else 4316 static inline void skb_ext_put(struct sk_buff *skb) {} 4317 static inline void skb_ext_reset(struct sk_buff *skb) {} 4318 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4319 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4320 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4321 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4322 #endif /* CONFIG_SKB_EXTENSIONS */ 4323 4324 static inline void nf_reset_ct(struct sk_buff *skb) 4325 { 4326 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4327 nf_conntrack_put(skb_nfct(skb)); 4328 skb->_nfct = 0; 4329 #endif 4330 } 4331 4332 static inline void nf_reset_trace(struct sk_buff *skb) 4333 { 4334 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4335 skb->nf_trace = 0; 4336 #endif 4337 } 4338 4339 static inline void ipvs_reset(struct sk_buff *skb) 4340 { 4341 #if IS_ENABLED(CONFIG_IP_VS) 4342 skb->ipvs_property = 0; 4343 #endif 4344 } 4345 4346 /* Note: This doesn't put any conntrack info in dst. */ 4347 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4348 bool copy) 4349 { 4350 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4351 dst->_nfct = src->_nfct; 4352 nf_conntrack_get(skb_nfct(src)); 4353 #endif 4354 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4355 if (copy) 4356 dst->nf_trace = src->nf_trace; 4357 #endif 4358 } 4359 4360 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4361 { 4362 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4363 nf_conntrack_put(skb_nfct(dst)); 4364 #endif 4365 __nf_copy(dst, src, true); 4366 } 4367 4368 #ifdef CONFIG_NETWORK_SECMARK 4369 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4370 { 4371 to->secmark = from->secmark; 4372 } 4373 4374 static inline void skb_init_secmark(struct sk_buff *skb) 4375 { 4376 skb->secmark = 0; 4377 } 4378 #else 4379 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4380 { } 4381 4382 static inline void skb_init_secmark(struct sk_buff *skb) 4383 { } 4384 #endif 4385 4386 static inline int secpath_exists(const struct sk_buff *skb) 4387 { 4388 #ifdef CONFIG_XFRM 4389 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4390 #else 4391 return 0; 4392 #endif 4393 } 4394 4395 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4396 { 4397 return !skb->destructor && 4398 !secpath_exists(skb) && 4399 !skb_nfct(skb) && 4400 !skb->_skb_refdst && 4401 !skb_has_frag_list(skb); 4402 } 4403 4404 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4405 { 4406 skb->queue_mapping = queue_mapping; 4407 } 4408 4409 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4410 { 4411 return skb->queue_mapping; 4412 } 4413 4414 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4415 { 4416 to->queue_mapping = from->queue_mapping; 4417 } 4418 4419 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4420 { 4421 skb->queue_mapping = rx_queue + 1; 4422 } 4423 4424 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4425 { 4426 return skb->queue_mapping - 1; 4427 } 4428 4429 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4430 { 4431 return skb->queue_mapping != 0; 4432 } 4433 4434 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4435 { 4436 skb->dst_pending_confirm = val; 4437 } 4438 4439 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4440 { 4441 return skb->dst_pending_confirm != 0; 4442 } 4443 4444 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4445 { 4446 #ifdef CONFIG_XFRM 4447 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4448 #else 4449 return NULL; 4450 #endif 4451 } 4452 4453 /* Keeps track of mac header offset relative to skb->head. 4454 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4455 * For non-tunnel skb it points to skb_mac_header() and for 4456 * tunnel skb it points to outer mac header. 4457 * Keeps track of level of encapsulation of network headers. 4458 */ 4459 struct skb_gso_cb { 4460 union { 4461 int mac_offset; 4462 int data_offset; 4463 }; 4464 int encap_level; 4465 __wsum csum; 4466 __u16 csum_start; 4467 }; 4468 #define SKB_GSO_CB_OFFSET 32 4469 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4470 4471 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4472 { 4473 return (skb_mac_header(inner_skb) - inner_skb->head) - 4474 SKB_GSO_CB(inner_skb)->mac_offset; 4475 } 4476 4477 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4478 { 4479 int new_headroom, headroom; 4480 int ret; 4481 4482 headroom = skb_headroom(skb); 4483 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4484 if (ret) 4485 return ret; 4486 4487 new_headroom = skb_headroom(skb); 4488 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4489 return 0; 4490 } 4491 4492 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4493 { 4494 /* Do not update partial checksums if remote checksum is enabled. */ 4495 if (skb->remcsum_offload) 4496 return; 4497 4498 SKB_GSO_CB(skb)->csum = res; 4499 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4500 } 4501 4502 /* Compute the checksum for a gso segment. First compute the checksum value 4503 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4504 * then add in skb->csum (checksum from csum_start to end of packet). 4505 * skb->csum and csum_start are then updated to reflect the checksum of the 4506 * resultant packet starting from the transport header-- the resultant checksum 4507 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4508 * header. 4509 */ 4510 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4511 { 4512 unsigned char *csum_start = skb_transport_header(skb); 4513 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4514 __wsum partial = SKB_GSO_CB(skb)->csum; 4515 4516 SKB_GSO_CB(skb)->csum = res; 4517 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4518 4519 return csum_fold(csum_partial(csum_start, plen, partial)); 4520 } 4521 4522 static inline bool skb_is_gso(const struct sk_buff *skb) 4523 { 4524 return skb_shinfo(skb)->gso_size; 4525 } 4526 4527 /* Note: Should be called only if skb_is_gso(skb) is true */ 4528 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4529 { 4530 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4531 } 4532 4533 /* Note: Should be called only if skb_is_gso(skb) is true */ 4534 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4535 { 4536 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4537 } 4538 4539 /* Note: Should be called only if skb_is_gso(skb) is true */ 4540 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4541 { 4542 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4543 } 4544 4545 static inline void skb_gso_reset(struct sk_buff *skb) 4546 { 4547 skb_shinfo(skb)->gso_size = 0; 4548 skb_shinfo(skb)->gso_segs = 0; 4549 skb_shinfo(skb)->gso_type = 0; 4550 } 4551 4552 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4553 u16 increment) 4554 { 4555 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4556 return; 4557 shinfo->gso_size += increment; 4558 } 4559 4560 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4561 u16 decrement) 4562 { 4563 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4564 return; 4565 shinfo->gso_size -= decrement; 4566 } 4567 4568 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4569 4570 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4571 { 4572 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4573 * wanted then gso_type will be set. */ 4574 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4575 4576 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4577 unlikely(shinfo->gso_type == 0)) { 4578 __skb_warn_lro_forwarding(skb); 4579 return true; 4580 } 4581 return false; 4582 } 4583 4584 static inline void skb_forward_csum(struct sk_buff *skb) 4585 { 4586 /* Unfortunately we don't support this one. Any brave souls? */ 4587 if (skb->ip_summed == CHECKSUM_COMPLETE) 4588 skb->ip_summed = CHECKSUM_NONE; 4589 } 4590 4591 /** 4592 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4593 * @skb: skb to check 4594 * 4595 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4596 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4597 * use this helper, to document places where we make this assertion. 4598 */ 4599 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4600 { 4601 #ifdef DEBUG 4602 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4603 #endif 4604 } 4605 4606 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4607 4608 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4609 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4610 unsigned int transport_len, 4611 __sum16(*skb_chkf)(struct sk_buff *skb)); 4612 4613 /** 4614 * skb_head_is_locked - Determine if the skb->head is locked down 4615 * @skb: skb to check 4616 * 4617 * The head on skbs build around a head frag can be removed if they are 4618 * not cloned. This function returns true if the skb head is locked down 4619 * due to either being allocated via kmalloc, or by being a clone with 4620 * multiple references to the head. 4621 */ 4622 static inline bool skb_head_is_locked(const struct sk_buff *skb) 4623 { 4624 return !skb->head_frag || skb_cloned(skb); 4625 } 4626 4627 /* Local Checksum Offload. 4628 * Compute outer checksum based on the assumption that the 4629 * inner checksum will be offloaded later. 4630 * See Documentation/networking/checksum-offloads.rst for 4631 * explanation of how this works. 4632 * Fill in outer checksum adjustment (e.g. with sum of outer 4633 * pseudo-header) before calling. 4634 * Also ensure that inner checksum is in linear data area. 4635 */ 4636 static inline __wsum lco_csum(struct sk_buff *skb) 4637 { 4638 unsigned char *csum_start = skb_checksum_start(skb); 4639 unsigned char *l4_hdr = skb_transport_header(skb); 4640 __wsum partial; 4641 4642 /* Start with complement of inner checksum adjustment */ 4643 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4644 skb->csum_offset)); 4645 4646 /* Add in checksum of our headers (incl. outer checksum 4647 * adjustment filled in by caller) and return result. 4648 */ 4649 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4650 } 4651 4652 static inline bool skb_is_redirected(const struct sk_buff *skb) 4653 { 4654 #ifdef CONFIG_NET_REDIRECT 4655 return skb->redirected; 4656 #else 4657 return false; 4658 #endif 4659 } 4660 4661 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4662 { 4663 #ifdef CONFIG_NET_REDIRECT 4664 skb->redirected = 1; 4665 skb->from_ingress = from_ingress; 4666 if (skb->from_ingress) 4667 skb->tstamp = 0; 4668 #endif 4669 } 4670 4671 static inline void skb_reset_redirect(struct sk_buff *skb) 4672 { 4673 #ifdef CONFIG_NET_REDIRECT 4674 skb->redirected = 0; 4675 #endif 4676 } 4677 4678 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 4679 { 4680 return skb->csum_not_inet; 4681 } 4682 4683 static inline void skb_set_kcov_handle(struct sk_buff *skb, 4684 const u64 kcov_handle) 4685 { 4686 #ifdef CONFIG_KCOV 4687 skb->kcov_handle = kcov_handle; 4688 #endif 4689 } 4690 4691 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4692 { 4693 #ifdef CONFIG_KCOV 4694 return skb->kcov_handle; 4695 #else 4696 return 0; 4697 #endif 4698 } 4699 4700 #endif /* __KERNEL__ */ 4701 #endif /* _LINUX_SKBUFF_H */ 4702
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.