1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Definitions for the 'struct sk_buff' memory handlers. 4 * 5 * Authors: 6 * Alan Cox, <[email protected]> 7 * Florian La Roche, <[email protected]> 8 */ 9 10 #ifndef _LINUX_SKBUFF_H 11 #define _LINUX_SKBUFF_H 12 13 #include <linux/kernel.h> 14 #include <linux/compiler.h> 15 #include <linux/time.h> 16 #include <linux/bug.h> 17 #include <linux/bvec.h> 18 #include <linux/cache.h> 19 #include <linux/rbtree.h> 20 #include <linux/socket.h> 21 #include <linux/refcount.h> 22 23 #include <linux/atomic.h> 24 #include <asm/types.h> 25 #include <linux/spinlock.h> 26 #include <linux/net.h> 27 #include <linux/textsearch.h> 28 #include <net/checksum.h> 29 #include <linux/rcupdate.h> 30 #include <linux/hrtimer.h> 31 #include <linux/dma-mapping.h> 32 #include <linux/netdev_features.h> 33 #include <linux/sched.h> 34 #include <linux/sched/clock.h> 35 #include <net/flow_dissector.h> 36 #include <linux/splice.h> 37 #include <linux/in6.h> 38 #include <linux/if_packet.h> 39 #include <linux/llist.h> 40 #include <net/flow.h> 41 #include <net/page_pool.h> 42 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 43 #include <linux/netfilter/nf_conntrack_common.h> 44 #endif 45 46 /* The interface for checksum offload between the stack and networking drivers 47 * is as follows... 48 * 49 * A. IP checksum related features 50 * 51 * Drivers advertise checksum offload capabilities in the features of a device. 52 * From the stack's point of view these are capabilities offered by the driver. 53 * A driver typically only advertises features that it is capable of offloading 54 * to its device. 55 * 56 * The checksum related features are: 57 * 58 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one 59 * IP (one's complement) checksum for any combination 60 * of protocols or protocol layering. The checksum is 61 * computed and set in a packet per the CHECKSUM_PARTIAL 62 * interface (see below). 63 * 64 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain 65 * TCP or UDP packets over IPv4. These are specifically 66 * unencapsulated packets of the form IPv4|TCP or 67 * IPv4|UDP where the Protocol field in the IPv4 header 68 * is TCP or UDP. The IPv4 header may contain IP options. 69 * This feature cannot be set in features for a device 70 * with NETIF_F_HW_CSUM also set. This feature is being 71 * DEPRECATED (see below). 72 * 73 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain 74 * TCP or UDP packets over IPv6. These are specifically 75 * unencapsulated packets of the form IPv6|TCP or 76 * IPv6|UDP where the Next Header field in the IPv6 77 * header is either TCP or UDP. IPv6 extension headers 78 * are not supported with this feature. This feature 79 * cannot be set in features for a device with 80 * NETIF_F_HW_CSUM also set. This feature is being 81 * DEPRECATED (see below). 82 * 83 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. 84 * This flag is only used to disable the RX checksum 85 * feature for a device. The stack will accept receive 86 * checksum indication in packets received on a device 87 * regardless of whether NETIF_F_RXCSUM is set. 88 * 89 * B. Checksumming of received packets by device. Indication of checksum 90 * verification is set in skb->ip_summed. Possible values are: 91 * 92 * CHECKSUM_NONE: 93 * 94 * Device did not checksum this packet e.g. due to lack of capabilities. 95 * The packet contains full (though not verified) checksum in packet but 96 * not in skb->csum. Thus, skb->csum is undefined in this case. 97 * 98 * CHECKSUM_UNNECESSARY: 99 * 100 * The hardware you're dealing with doesn't calculate the full checksum 101 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 102 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 103 * if their checksums are okay. skb->csum is still undefined in this case 104 * though. A driver or device must never modify the checksum field in the 105 * packet even if checksum is verified. 106 * 107 * CHECKSUM_UNNECESSARY is applicable to following protocols: 108 * TCP: IPv6 and IPv4. 109 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 110 * zero UDP checksum for either IPv4 or IPv6, the networking stack 111 * may perform further validation in this case. 112 * GRE: only if the checksum is present in the header. 113 * SCTP: indicates the CRC in SCTP header has been validated. 114 * FCOE: indicates the CRC in FC frame has been validated. 115 * 116 * skb->csum_level indicates the number of consecutive checksums found in 117 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 118 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 119 * and a device is able to verify the checksums for UDP (possibly zero), 120 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to 121 * two. If the device were only able to verify the UDP checksum and not 122 * GRE, either because it doesn't support GRE checksum or because GRE 123 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 124 * not considered in this case). 125 * 126 * CHECKSUM_COMPLETE: 127 * 128 * This is the most generic way. The device supplied checksum of the _whole_ 129 * packet as seen by netif_rx() and fills in skb->csum. This means the 130 * hardware doesn't need to parse L3/L4 headers to implement this. 131 * 132 * Notes: 133 * - Even if device supports only some protocols, but is able to produce 134 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 135 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 136 * 137 * CHECKSUM_PARTIAL: 138 * 139 * A checksum is set up to be offloaded to a device as described in the 140 * output description for CHECKSUM_PARTIAL. This may occur on a packet 141 * received directly from another Linux OS, e.g., a virtualized Linux kernel 142 * on the same host, or it may be set in the input path in GRO or remote 143 * checksum offload. For the purposes of checksum verification, the checksum 144 * referred to by skb->csum_start + skb->csum_offset and any preceding 145 * checksums in the packet are considered verified. Any checksums in the 146 * packet that are after the checksum being offloaded are not considered to 147 * be verified. 148 * 149 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload 150 * in the skb->ip_summed for a packet. Values are: 151 * 152 * CHECKSUM_PARTIAL: 153 * 154 * The driver is required to checksum the packet as seen by hard_start_xmit() 155 * from skb->csum_start up to the end, and to record/write the checksum at 156 * offset skb->csum_start + skb->csum_offset. A driver may verify that the 157 * csum_start and csum_offset values are valid values given the length and 158 * offset of the packet, but it should not attempt to validate that the 159 * checksum refers to a legitimate transport layer checksum -- it is the 160 * purview of the stack to validate that csum_start and csum_offset are set 161 * correctly. 162 * 163 * When the stack requests checksum offload for a packet, the driver MUST 164 * ensure that the checksum is set correctly. A driver can either offload the 165 * checksum calculation to the device, or call skb_checksum_help (in the case 166 * that the device does not support offload for a particular checksum). 167 * 168 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of 169 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate 170 * checksum offload capability. 171 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based 172 * on network device checksumming capabilities: if a packet does not match 173 * them, skb_checksum_help or skb_crc32c_help (depending on the value of 174 * csum_not_inet, see item D.) is called to resolve the checksum. 175 * 176 * CHECKSUM_NONE: 177 * 178 * The skb was already checksummed by the protocol, or a checksum is not 179 * required. 180 * 181 * CHECKSUM_UNNECESSARY: 182 * 183 * This has the same meaning as CHECKSUM_NONE for checksum offload on 184 * output. 185 * 186 * CHECKSUM_COMPLETE: 187 * Not used in checksum output. If a driver observes a packet with this value 188 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set. 189 * 190 * D. Non-IP checksum (CRC) offloads 191 * 192 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of 193 * offloading the SCTP CRC in a packet. To perform this offload the stack 194 * will set csum_start and csum_offset accordingly, set ip_summed to 195 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in 196 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c. 197 * A driver that supports both IP checksum offload and SCTP CRC32c offload 198 * must verify which offload is configured for a packet by testing the 199 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve 200 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 201 * 202 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of 203 * offloading the FCOE CRC in a packet. To perform this offload the stack 204 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset 205 * accordingly. Note that there is no indication in the skbuff that the 206 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 207 * both IP checksum offload and FCOE CRC offload must verify which offload 208 * is configured for a packet, presumably by inspecting packet headers. 209 * 210 * E. Checksumming on output with GSO. 211 * 212 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload 213 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 214 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as 215 * part of the GSO operation is implied. If a checksum is being offloaded 216 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and 217 * csum_offset are set to refer to the outermost checksum being offloaded 218 * (two offloaded checksums are possible with UDP encapsulation). 219 */ 220 221 /* Don't change this without changing skb_csum_unnecessary! */ 222 #define CHECKSUM_NONE 0 223 #define CHECKSUM_UNNECESSARY 1 224 #define CHECKSUM_COMPLETE 2 225 #define CHECKSUM_PARTIAL 3 226 227 /* Maximum value in skb->csum_level */ 228 #define SKB_MAX_CSUM_LEVEL 3 229 230 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 231 #define SKB_WITH_OVERHEAD(X) \ 232 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 233 #define SKB_MAX_ORDER(X, ORDER) \ 234 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 235 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 236 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 237 238 /* return minimum truesize of one skb containing X bytes of data */ 239 #define SKB_TRUESIZE(X) ((X) + \ 240 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 241 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 242 243 struct ahash_request; 244 struct net_device; 245 struct scatterlist; 246 struct pipe_inode_info; 247 struct iov_iter; 248 struct napi_struct; 249 struct bpf_prog; 250 union bpf_attr; 251 struct skb_ext; 252 253 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 254 struct nf_bridge_info { 255 enum { 256 BRNF_PROTO_UNCHANGED, 257 BRNF_PROTO_8021Q, 258 BRNF_PROTO_PPPOE 259 } orig_proto:8; 260 u8 pkt_otherhost:1; 261 u8 in_prerouting:1; 262 u8 bridged_dnat:1; 263 __u16 frag_max_size; 264 struct net_device *physindev; 265 266 /* always valid & non-NULL from FORWARD on, for physdev match */ 267 struct net_device *physoutdev; 268 union { 269 /* prerouting: detect dnat in orig/reply direction */ 270 __be32 ipv4_daddr; 271 struct in6_addr ipv6_daddr; 272 273 /* after prerouting + nat detected: store original source 274 * mac since neigh resolution overwrites it, only used while 275 * skb is out in neigh layer. 276 */ 277 char neigh_header[8]; 278 }; 279 }; 280 #endif 281 282 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 283 /* Chain in tc_skb_ext will be used to share the tc chain with 284 * ovs recirc_id. It will be set to the current chain by tc 285 * and read by ovs to recirc_id. 286 */ 287 struct tc_skb_ext { 288 __u32 chain; 289 __u16 mru; 290 __u16 zone; 291 u8 post_ct:1; 292 u8 post_ct_snat:1; 293 u8 post_ct_dnat:1; 294 }; 295 #endif 296 297 struct sk_buff_head { 298 /* These two members must be first to match sk_buff. */ 299 struct_group_tagged(sk_buff_list, list, 300 struct sk_buff *next; 301 struct sk_buff *prev; 302 ); 303 304 __u32 qlen; 305 spinlock_t lock; 306 }; 307 308 struct sk_buff; 309 310 /* The reason of skb drop, which is used in kfree_skb_reason(). 311 * en...maybe they should be splited by group? 312 * 313 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is 314 * used to translate the reason to string. 315 */ 316 enum skb_drop_reason { 317 SKB_DROP_REASON_NOT_SPECIFIED, 318 SKB_DROP_REASON_NO_SOCKET, 319 SKB_DROP_REASON_PKT_TOO_SMALL, 320 SKB_DROP_REASON_TCP_CSUM, 321 SKB_DROP_REASON_TCP_FILTER, 322 SKB_DROP_REASON_UDP_CSUM, 323 SKB_DROP_REASON_MAX, 324 }; 325 326 /* To allow 64K frame to be packed as single skb without frag_list we 327 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 328 * buffers which do not start on a page boundary. 329 * 330 * Since GRO uses frags we allocate at least 16 regardless of page 331 * size. 332 */ 333 #if (65536/PAGE_SIZE + 1) < 16 334 #define MAX_SKB_FRAGS 16UL 335 #else 336 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 337 #endif 338 extern int sysctl_max_skb_frags; 339 340 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 341 * segment using its current segmentation instead. 342 */ 343 #define GSO_BY_FRAGS 0xFFFF 344 345 typedef struct bio_vec skb_frag_t; 346 347 /** 348 * skb_frag_size() - Returns the size of a skb fragment 349 * @frag: skb fragment 350 */ 351 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 352 { 353 return frag->bv_len; 354 } 355 356 /** 357 * skb_frag_size_set() - Sets the size of a skb fragment 358 * @frag: skb fragment 359 * @size: size of fragment 360 */ 361 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 362 { 363 frag->bv_len = size; 364 } 365 366 /** 367 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 368 * @frag: skb fragment 369 * @delta: value to add 370 */ 371 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 372 { 373 frag->bv_len += delta; 374 } 375 376 /** 377 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 378 * @frag: skb fragment 379 * @delta: value to subtract 380 */ 381 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 382 { 383 frag->bv_len -= delta; 384 } 385 386 /** 387 * skb_frag_must_loop - Test if %p is a high memory page 388 * @p: fragment's page 389 */ 390 static inline bool skb_frag_must_loop(struct page *p) 391 { 392 #if defined(CONFIG_HIGHMEM) 393 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 394 return true; 395 #endif 396 return false; 397 } 398 399 /** 400 * skb_frag_foreach_page - loop over pages in a fragment 401 * 402 * @f: skb frag to operate on 403 * @f_off: offset from start of f->bv_page 404 * @f_len: length from f_off to loop over 405 * @p: (temp var) current page 406 * @p_off: (temp var) offset from start of current page, 407 * non-zero only on first page. 408 * @p_len: (temp var) length in current page, 409 * < PAGE_SIZE only on first and last page. 410 * @copied: (temp var) length so far, excluding current p_len. 411 * 412 * A fragment can hold a compound page, in which case per-page 413 * operations, notably kmap_atomic, must be called for each 414 * regular page. 415 */ 416 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 417 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 418 p_off = (f_off) & (PAGE_SIZE - 1), \ 419 p_len = skb_frag_must_loop(p) ? \ 420 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 421 copied = 0; \ 422 copied < f_len; \ 423 copied += p_len, p++, p_off = 0, \ 424 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 425 426 #define HAVE_HW_TIME_STAMP 427 428 /** 429 * struct skb_shared_hwtstamps - hardware time stamps 430 * @hwtstamp: hardware time stamp transformed into duration 431 * since arbitrary point in time 432 * 433 * Software time stamps generated by ktime_get_real() are stored in 434 * skb->tstamp. 435 * 436 * hwtstamps can only be compared against other hwtstamps from 437 * the same device. 438 * 439 * This structure is attached to packets as part of the 440 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 441 */ 442 struct skb_shared_hwtstamps { 443 ktime_t hwtstamp; 444 }; 445 446 /* Definitions for tx_flags in struct skb_shared_info */ 447 enum { 448 /* generate hardware time stamp */ 449 SKBTX_HW_TSTAMP = 1 << 0, 450 451 /* generate software time stamp when queueing packet to NIC */ 452 SKBTX_SW_TSTAMP = 1 << 1, 453 454 /* device driver is going to provide hardware time stamp */ 455 SKBTX_IN_PROGRESS = 1 << 2, 456 457 /* generate wifi status information (where possible) */ 458 SKBTX_WIFI_STATUS = 1 << 4, 459 460 /* generate software time stamp when entering packet scheduling */ 461 SKBTX_SCHED_TSTAMP = 1 << 6, 462 }; 463 464 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 465 SKBTX_SCHED_TSTAMP) 466 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 467 468 /* Definitions for flags in struct skb_shared_info */ 469 enum { 470 /* use zcopy routines */ 471 SKBFL_ZEROCOPY_ENABLE = BIT(0), 472 473 /* This indicates at least one fragment might be overwritten 474 * (as in vmsplice(), sendfile() ...) 475 * If we need to compute a TX checksum, we'll need to copy 476 * all frags to avoid possible bad checksum 477 */ 478 SKBFL_SHARED_FRAG = BIT(1), 479 480 /* segment contains only zerocopy data and should not be 481 * charged to the kernel memory. 482 */ 483 SKBFL_PURE_ZEROCOPY = BIT(2), 484 }; 485 486 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 487 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY) 488 489 /* 490 * The callback notifies userspace to release buffers when skb DMA is done in 491 * lower device, the skb last reference should be 0 when calling this. 492 * The zerocopy_success argument is true if zero copy transmit occurred, 493 * false on data copy or out of memory error caused by data copy attempt. 494 * The ctx field is used to track device context. 495 * The desc field is used to track userspace buffer index. 496 */ 497 struct ubuf_info { 498 void (*callback)(struct sk_buff *, struct ubuf_info *, 499 bool zerocopy_success); 500 union { 501 struct { 502 unsigned long desc; 503 void *ctx; 504 }; 505 struct { 506 u32 id; 507 u16 len; 508 u16 zerocopy:1; 509 u32 bytelen; 510 }; 511 }; 512 refcount_t refcnt; 513 u8 flags; 514 515 struct mmpin { 516 struct user_struct *user; 517 unsigned int num_pg; 518 } mmp; 519 }; 520 521 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 522 523 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 524 void mm_unaccount_pinned_pages(struct mmpin *mmp); 525 526 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size); 527 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 528 struct ubuf_info *uarg); 529 530 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 531 532 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 533 bool success); 534 535 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len); 536 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 537 struct msghdr *msg, int len, 538 struct ubuf_info *uarg); 539 540 /* This data is invariant across clones and lives at 541 * the end of the header data, ie. at skb->end. 542 */ 543 struct skb_shared_info { 544 __u8 flags; 545 __u8 meta_len; 546 __u8 nr_frags; 547 __u8 tx_flags; 548 unsigned short gso_size; 549 /* Warning: this field is not always filled in (UFO)! */ 550 unsigned short gso_segs; 551 struct sk_buff *frag_list; 552 struct skb_shared_hwtstamps hwtstamps; 553 unsigned int gso_type; 554 u32 tskey; 555 556 /* 557 * Warning : all fields before dataref are cleared in __alloc_skb() 558 */ 559 atomic_t dataref; 560 unsigned int xdp_frags_size; 561 562 /* Intermediate layers must ensure that destructor_arg 563 * remains valid until skb destructor */ 564 void * destructor_arg; 565 566 /* must be last field, see pskb_expand_head() */ 567 skb_frag_t frags[MAX_SKB_FRAGS]; 568 }; 569 570 /* We divide dataref into two halves. The higher 16 bits hold references 571 * to the payload part of skb->data. The lower 16 bits hold references to 572 * the entire skb->data. A clone of a headerless skb holds the length of 573 * the header in skb->hdr_len. 574 * 575 * All users must obey the rule that the skb->data reference count must be 576 * greater than or equal to the payload reference count. 577 * 578 * Holding a reference to the payload part means that the user does not 579 * care about modifications to the header part of skb->data. 580 */ 581 #define SKB_DATAREF_SHIFT 16 582 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 583 584 585 enum { 586 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 587 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 588 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 589 }; 590 591 enum { 592 SKB_GSO_TCPV4 = 1 << 0, 593 594 /* This indicates the skb is from an untrusted source. */ 595 SKB_GSO_DODGY = 1 << 1, 596 597 /* This indicates the tcp segment has CWR set. */ 598 SKB_GSO_TCP_ECN = 1 << 2, 599 600 SKB_GSO_TCP_FIXEDID = 1 << 3, 601 602 SKB_GSO_TCPV6 = 1 << 4, 603 604 SKB_GSO_FCOE = 1 << 5, 605 606 SKB_GSO_GRE = 1 << 6, 607 608 SKB_GSO_GRE_CSUM = 1 << 7, 609 610 SKB_GSO_IPXIP4 = 1 << 8, 611 612 SKB_GSO_IPXIP6 = 1 << 9, 613 614 SKB_GSO_UDP_TUNNEL = 1 << 10, 615 616 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 617 618 SKB_GSO_PARTIAL = 1 << 12, 619 620 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 621 622 SKB_GSO_SCTP = 1 << 14, 623 624 SKB_GSO_ESP = 1 << 15, 625 626 SKB_GSO_UDP = 1 << 16, 627 628 SKB_GSO_UDP_L4 = 1 << 17, 629 630 SKB_GSO_FRAGLIST = 1 << 18, 631 }; 632 633 #if BITS_PER_LONG > 32 634 #define NET_SKBUFF_DATA_USES_OFFSET 1 635 #endif 636 637 #ifdef NET_SKBUFF_DATA_USES_OFFSET 638 typedef unsigned int sk_buff_data_t; 639 #else 640 typedef unsigned char *sk_buff_data_t; 641 #endif 642 643 /** 644 * struct sk_buff - socket buffer 645 * @next: Next buffer in list 646 * @prev: Previous buffer in list 647 * @tstamp: Time we arrived/left 648 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 649 * for retransmit timer 650 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 651 * @list: queue head 652 * @ll_node: anchor in an llist (eg socket defer_list) 653 * @sk: Socket we are owned by 654 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 655 * fragmentation management 656 * @dev: Device we arrived on/are leaving by 657 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 658 * @cb: Control buffer. Free for use by every layer. Put private vars here 659 * @_skb_refdst: destination entry (with norefcount bit) 660 * @sp: the security path, used for xfrm 661 * @len: Length of actual data 662 * @data_len: Data length 663 * @mac_len: Length of link layer header 664 * @hdr_len: writable header length of cloned skb 665 * @csum: Checksum (must include start/offset pair) 666 * @csum_start: Offset from skb->head where checksumming should start 667 * @csum_offset: Offset from csum_start where checksum should be stored 668 * @priority: Packet queueing priority 669 * @ignore_df: allow local fragmentation 670 * @cloned: Head may be cloned (check refcnt to be sure) 671 * @ip_summed: Driver fed us an IP checksum 672 * @nohdr: Payload reference only, must not modify header 673 * @pkt_type: Packet class 674 * @fclone: skbuff clone status 675 * @ipvs_property: skbuff is owned by ipvs 676 * @inner_protocol_type: whether the inner protocol is 677 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 678 * @remcsum_offload: remote checksum offload is enabled 679 * @offload_fwd_mark: Packet was L2-forwarded in hardware 680 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 681 * @tc_skip_classify: do not classify packet. set by IFB device 682 * @tc_at_ingress: used within tc_classify to distinguish in/egress 683 * @redirected: packet was redirected by packet classifier 684 * @from_ingress: packet was redirected from the ingress path 685 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h 686 * @peeked: this packet has been seen already, so stats have been 687 * done for it, don't do them again 688 * @nf_trace: netfilter packet trace flag 689 * @protocol: Packet protocol from driver 690 * @destructor: Destruct function 691 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 692 * @_sk_redir: socket redirection information for skmsg 693 * @_nfct: Associated connection, if any (with nfctinfo bits) 694 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 695 * @skb_iif: ifindex of device we arrived on 696 * @tc_index: Traffic control index 697 * @hash: the packet hash 698 * @queue_mapping: Queue mapping for multiqueue devices 699 * @head_frag: skb was allocated from page fragments, 700 * not allocated by kmalloc() or vmalloc(). 701 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 702 * @pp_recycle: mark the packet for recycling instead of freeing (implies 703 * page_pool support on driver) 704 * @active_extensions: active extensions (skb_ext_id types) 705 * @ndisc_nodetype: router type (from link layer) 706 * @ooo_okay: allow the mapping of a socket to a queue to be changed 707 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 708 * ports. 709 * @sw_hash: indicates hash was computed in software stack 710 * @wifi_acked_valid: wifi_acked was set 711 * @wifi_acked: whether frame was acked on wifi or not 712 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 713 * @encapsulation: indicates the inner headers in the skbuff are valid 714 * @encap_hdr_csum: software checksum is needed 715 * @csum_valid: checksum is already valid 716 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 717 * @csum_complete_sw: checksum was completed by software 718 * @csum_level: indicates the number of consecutive checksums found in 719 * the packet minus one that have been verified as 720 * CHECKSUM_UNNECESSARY (max 3) 721 * @dst_pending_confirm: need to confirm neighbour 722 * @decrypted: Decrypted SKB 723 * @slow_gro: state present at GRO time, slower prepare step required 724 * @napi_id: id of the NAPI struct this skb came from 725 * @sender_cpu: (aka @napi_id) source CPU in XPS 726 * @secmark: security marking 727 * @mark: Generic packet mark 728 * @reserved_tailroom: (aka @mark) number of bytes of free space available 729 * at the tail of an sk_buff 730 * @vlan_present: VLAN tag is present 731 * @vlan_proto: vlan encapsulation protocol 732 * @vlan_tci: vlan tag control information 733 * @inner_protocol: Protocol (encapsulation) 734 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 735 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 736 * @inner_transport_header: Inner transport layer header (encapsulation) 737 * @inner_network_header: Network layer header (encapsulation) 738 * @inner_mac_header: Link layer header (encapsulation) 739 * @transport_header: Transport layer header 740 * @network_header: Network layer header 741 * @mac_header: Link layer header 742 * @kcov_handle: KCOV remote handle for remote coverage collection 743 * @tail: Tail pointer 744 * @end: End pointer 745 * @head: Head of buffer 746 * @data: Data head pointer 747 * @truesize: Buffer size 748 * @users: User count - see {datagram,tcp}.c 749 * @extensions: allocated extensions, valid if active_extensions is nonzero 750 */ 751 752 struct sk_buff { 753 union { 754 struct { 755 /* These two members must be first to match sk_buff_head. */ 756 struct sk_buff *next; 757 struct sk_buff *prev; 758 759 union { 760 struct net_device *dev; 761 /* Some protocols might use this space to store information, 762 * while device pointer would be NULL. 763 * UDP receive path is one user. 764 */ 765 unsigned long dev_scratch; 766 }; 767 }; 768 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 769 struct list_head list; 770 struct llist_node ll_node; 771 }; 772 773 union { 774 struct sock *sk; 775 int ip_defrag_offset; 776 }; 777 778 union { 779 ktime_t tstamp; 780 u64 skb_mstamp_ns; /* earliest departure time */ 781 }; 782 /* 783 * This is the control buffer. It is free to use for every 784 * layer. Please put your private variables there. If you 785 * want to keep them across layers you have to do a skb_clone() 786 * first. This is owned by whoever has the skb queued ATM. 787 */ 788 char cb[48] __aligned(8); 789 790 union { 791 struct { 792 unsigned long _skb_refdst; 793 void (*destructor)(struct sk_buff *skb); 794 }; 795 struct list_head tcp_tsorted_anchor; 796 #ifdef CONFIG_NET_SOCK_MSG 797 unsigned long _sk_redir; 798 #endif 799 }; 800 801 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 802 unsigned long _nfct; 803 #endif 804 unsigned int len, 805 data_len; 806 __u16 mac_len, 807 hdr_len; 808 809 /* Following fields are _not_ copied in __copy_skb_header() 810 * Note that queue_mapping is here mostly to fill a hole. 811 */ 812 __u16 queue_mapping; 813 814 /* if you move cloned around you also must adapt those constants */ 815 #ifdef __BIG_ENDIAN_BITFIELD 816 #define CLONED_MASK (1 << 7) 817 #else 818 #define CLONED_MASK 1 819 #endif 820 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset) 821 822 /* private: */ 823 __u8 __cloned_offset[0]; 824 /* public: */ 825 __u8 cloned:1, 826 nohdr:1, 827 fclone:2, 828 peeked:1, 829 head_frag:1, 830 pfmemalloc:1, 831 pp_recycle:1; /* page_pool recycle indicator */ 832 #ifdef CONFIG_SKB_EXTENSIONS 833 __u8 active_extensions; 834 #endif 835 836 /* Fields enclosed in headers group are copied 837 * using a single memcpy() in __copy_skb_header() 838 */ 839 struct_group(headers, 840 841 /* private: */ 842 __u8 __pkt_type_offset[0]; 843 /* public: */ 844 __u8 pkt_type:3; /* see PKT_TYPE_MAX */ 845 __u8 ignore_df:1; 846 __u8 nf_trace:1; 847 __u8 ip_summed:2; 848 __u8 ooo_okay:1; 849 850 __u8 l4_hash:1; 851 __u8 sw_hash:1; 852 __u8 wifi_acked_valid:1; 853 __u8 wifi_acked:1; 854 __u8 no_fcs:1; 855 /* Indicates the inner headers are valid in the skbuff. */ 856 __u8 encapsulation:1; 857 __u8 encap_hdr_csum:1; 858 __u8 csum_valid:1; 859 860 /* private: */ 861 __u8 __pkt_vlan_present_offset[0]; 862 /* public: */ 863 __u8 vlan_present:1; /* See PKT_VLAN_PRESENT_BIT */ 864 __u8 csum_complete_sw:1; 865 __u8 csum_level:2; 866 __u8 csum_not_inet:1; 867 __u8 dst_pending_confirm:1; 868 #ifdef CONFIG_IPV6_NDISC_NODETYPE 869 __u8 ndisc_nodetype:2; 870 #endif 871 872 __u8 ipvs_property:1; 873 __u8 inner_protocol_type:1; 874 __u8 remcsum_offload:1; 875 #ifdef CONFIG_NET_SWITCHDEV 876 __u8 offload_fwd_mark:1; 877 __u8 offload_l3_fwd_mark:1; 878 #endif 879 #ifdef CONFIG_NET_CLS_ACT 880 __u8 tc_skip_classify:1; 881 __u8 tc_at_ingress:1; 882 #endif 883 __u8 redirected:1; 884 #ifdef CONFIG_NET_REDIRECT 885 __u8 from_ingress:1; 886 #endif 887 #ifdef CONFIG_NETFILTER_SKIP_EGRESS 888 __u8 nf_skip_egress:1; 889 #endif 890 #ifdef CONFIG_TLS_DEVICE 891 __u8 decrypted:1; 892 #endif 893 __u8 slow_gro:1; 894 895 #ifdef CONFIG_NET_SCHED 896 __u16 tc_index; /* traffic control index */ 897 #endif 898 899 union { 900 __wsum csum; 901 struct { 902 __u16 csum_start; 903 __u16 csum_offset; 904 }; 905 }; 906 __u32 priority; 907 int skb_iif; 908 __u32 hash; 909 __be16 vlan_proto; 910 __u16 vlan_tci; 911 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 912 union { 913 unsigned int napi_id; 914 unsigned int sender_cpu; 915 }; 916 #endif 917 #ifdef CONFIG_NETWORK_SECMARK 918 __u32 secmark; 919 #endif 920 921 union { 922 __u32 mark; 923 __u32 reserved_tailroom; 924 }; 925 926 union { 927 __be16 inner_protocol; 928 __u8 inner_ipproto; 929 }; 930 931 __u16 inner_transport_header; 932 __u16 inner_network_header; 933 __u16 inner_mac_header; 934 935 __be16 protocol; 936 __u16 transport_header; 937 __u16 network_header; 938 __u16 mac_header; 939 940 #ifdef CONFIG_KCOV 941 u64 kcov_handle; 942 #endif 943 944 ); /* end headers group */ 945 946 /* These elements must be at the end, see alloc_skb() for details. */ 947 sk_buff_data_t tail; 948 sk_buff_data_t end; 949 unsigned char *head, 950 *data; 951 unsigned int truesize; 952 refcount_t users; 953 954 #ifdef CONFIG_SKB_EXTENSIONS 955 /* only useable after checking ->active_extensions != 0 */ 956 struct skb_ext *extensions; 957 #endif 958 }; 959 960 /* if you move pkt_type around you also must adapt those constants */ 961 #ifdef __BIG_ENDIAN_BITFIELD 962 #define PKT_TYPE_MAX (7 << 5) 963 #else 964 #define PKT_TYPE_MAX 7 965 #endif 966 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset) 967 968 /* if you move pkt_vlan_present around you also must adapt these constants */ 969 #ifdef __BIG_ENDIAN_BITFIELD 970 #define PKT_VLAN_PRESENT_BIT 7 971 #else 972 #define PKT_VLAN_PRESENT_BIT 0 973 #endif 974 #define PKT_VLAN_PRESENT_OFFSET offsetof(struct sk_buff, __pkt_vlan_present_offset) 975 976 #ifdef __KERNEL__ 977 /* 978 * Handling routines are only of interest to the kernel 979 */ 980 981 #define SKB_ALLOC_FCLONE 0x01 982 #define SKB_ALLOC_RX 0x02 983 #define SKB_ALLOC_NAPI 0x04 984 985 /** 986 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 987 * @skb: buffer 988 */ 989 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 990 { 991 return unlikely(skb->pfmemalloc); 992 } 993 994 /* 995 * skb might have a dst pointer attached, refcounted or not. 996 * _skb_refdst low order bit is set if refcount was _not_ taken 997 */ 998 #define SKB_DST_NOREF 1UL 999 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 1000 1001 /** 1002 * skb_dst - returns skb dst_entry 1003 * @skb: buffer 1004 * 1005 * Returns skb dst_entry, regardless of reference taken or not. 1006 */ 1007 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 1008 { 1009 /* If refdst was not refcounted, check we still are in a 1010 * rcu_read_lock section 1011 */ 1012 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 1013 !rcu_read_lock_held() && 1014 !rcu_read_lock_bh_held()); 1015 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 1016 } 1017 1018 /** 1019 * skb_dst_set - sets skb dst 1020 * @skb: buffer 1021 * @dst: dst entry 1022 * 1023 * Sets skb dst, assuming a reference was taken on dst and should 1024 * be released by skb_dst_drop() 1025 */ 1026 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 1027 { 1028 skb->slow_gro |= !!dst; 1029 skb->_skb_refdst = (unsigned long)dst; 1030 } 1031 1032 /** 1033 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 1034 * @skb: buffer 1035 * @dst: dst entry 1036 * 1037 * Sets skb dst, assuming a reference was not taken on dst. 1038 * If dst entry is cached, we do not take reference and dst_release 1039 * will be avoided by refdst_drop. If dst entry is not cached, we take 1040 * reference, so that last dst_release can destroy the dst immediately. 1041 */ 1042 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 1043 { 1044 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1045 skb->slow_gro |= !!dst; 1046 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1047 } 1048 1049 /** 1050 * skb_dst_is_noref - Test if skb dst isn't refcounted 1051 * @skb: buffer 1052 */ 1053 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1054 { 1055 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1056 } 1057 1058 /** 1059 * skb_rtable - Returns the skb &rtable 1060 * @skb: buffer 1061 */ 1062 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1063 { 1064 return (struct rtable *)skb_dst(skb); 1065 } 1066 1067 /* For mangling skb->pkt_type from user space side from applications 1068 * such as nft, tc, etc, we only allow a conservative subset of 1069 * possible pkt_types to be set. 1070 */ 1071 static inline bool skb_pkt_type_ok(u32 ptype) 1072 { 1073 return ptype <= PACKET_OTHERHOST; 1074 } 1075 1076 /** 1077 * skb_napi_id - Returns the skb's NAPI id 1078 * @skb: buffer 1079 */ 1080 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1081 { 1082 #ifdef CONFIG_NET_RX_BUSY_POLL 1083 return skb->napi_id; 1084 #else 1085 return 0; 1086 #endif 1087 } 1088 1089 /** 1090 * skb_unref - decrement the skb's reference count 1091 * @skb: buffer 1092 * 1093 * Returns true if we can free the skb. 1094 */ 1095 static inline bool skb_unref(struct sk_buff *skb) 1096 { 1097 if (unlikely(!skb)) 1098 return false; 1099 if (likely(refcount_read(&skb->users) == 1)) 1100 smp_rmb(); 1101 else if (likely(!refcount_dec_and_test(&skb->users))) 1102 return false; 1103 1104 return true; 1105 } 1106 1107 void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason); 1108 1109 /** 1110 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason 1111 * @skb: buffer to free 1112 */ 1113 static inline void kfree_skb(struct sk_buff *skb) 1114 { 1115 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 1116 } 1117 1118 void skb_release_head_state(struct sk_buff *skb); 1119 void kfree_skb_list(struct sk_buff *segs); 1120 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1121 void skb_tx_error(struct sk_buff *skb); 1122 1123 #ifdef CONFIG_TRACEPOINTS 1124 void consume_skb(struct sk_buff *skb); 1125 #else 1126 static inline void consume_skb(struct sk_buff *skb) 1127 { 1128 return kfree_skb(skb); 1129 } 1130 #endif 1131 1132 void __consume_stateless_skb(struct sk_buff *skb); 1133 void __kfree_skb(struct sk_buff *skb); 1134 extern struct kmem_cache *skbuff_head_cache; 1135 1136 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1137 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1138 bool *fragstolen, int *delta_truesize); 1139 1140 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1141 int node); 1142 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1143 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1144 struct sk_buff *build_skb_around(struct sk_buff *skb, 1145 void *data, unsigned int frag_size); 1146 1147 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1148 1149 /** 1150 * alloc_skb - allocate a network buffer 1151 * @size: size to allocate 1152 * @priority: allocation mask 1153 * 1154 * This function is a convenient wrapper around __alloc_skb(). 1155 */ 1156 static inline struct sk_buff *alloc_skb(unsigned int size, 1157 gfp_t priority) 1158 { 1159 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1160 } 1161 1162 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1163 unsigned long data_len, 1164 int max_page_order, 1165 int *errcode, 1166 gfp_t gfp_mask); 1167 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1168 1169 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1170 struct sk_buff_fclones { 1171 struct sk_buff skb1; 1172 1173 struct sk_buff skb2; 1174 1175 refcount_t fclone_ref; 1176 }; 1177 1178 /** 1179 * skb_fclone_busy - check if fclone is busy 1180 * @sk: socket 1181 * @skb: buffer 1182 * 1183 * Returns true if skb is a fast clone, and its clone is not freed. 1184 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1185 * so we also check that this didnt happen. 1186 */ 1187 static inline bool skb_fclone_busy(const struct sock *sk, 1188 const struct sk_buff *skb) 1189 { 1190 const struct sk_buff_fclones *fclones; 1191 1192 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1193 1194 return skb->fclone == SKB_FCLONE_ORIG && 1195 refcount_read(&fclones->fclone_ref) > 1 && 1196 READ_ONCE(fclones->skb2.sk) == sk; 1197 } 1198 1199 /** 1200 * alloc_skb_fclone - allocate a network buffer from fclone cache 1201 * @size: size to allocate 1202 * @priority: allocation mask 1203 * 1204 * This function is a convenient wrapper around __alloc_skb(). 1205 */ 1206 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1207 gfp_t priority) 1208 { 1209 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1210 } 1211 1212 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1213 void skb_headers_offset_update(struct sk_buff *skb, int off); 1214 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1215 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1216 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1217 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1218 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1219 gfp_t gfp_mask, bool fclone); 1220 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1221 gfp_t gfp_mask) 1222 { 1223 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1224 } 1225 1226 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1227 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1228 unsigned int headroom); 1229 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom); 1230 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1231 int newtailroom, gfp_t priority); 1232 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1233 int offset, int len); 1234 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1235 int offset, int len); 1236 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1237 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1238 1239 /** 1240 * skb_pad - zero pad the tail of an skb 1241 * @skb: buffer to pad 1242 * @pad: space to pad 1243 * 1244 * Ensure that a buffer is followed by a padding area that is zero 1245 * filled. Used by network drivers which may DMA or transfer data 1246 * beyond the buffer end onto the wire. 1247 * 1248 * May return error in out of memory cases. The skb is freed on error. 1249 */ 1250 static inline int skb_pad(struct sk_buff *skb, int pad) 1251 { 1252 return __skb_pad(skb, pad, true); 1253 } 1254 #define dev_kfree_skb(a) consume_skb(a) 1255 1256 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1257 int offset, size_t size); 1258 1259 struct skb_seq_state { 1260 __u32 lower_offset; 1261 __u32 upper_offset; 1262 __u32 frag_idx; 1263 __u32 stepped_offset; 1264 struct sk_buff *root_skb; 1265 struct sk_buff *cur_skb; 1266 __u8 *frag_data; 1267 __u32 frag_off; 1268 }; 1269 1270 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1271 unsigned int to, struct skb_seq_state *st); 1272 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1273 struct skb_seq_state *st); 1274 void skb_abort_seq_read(struct skb_seq_state *st); 1275 1276 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1277 unsigned int to, struct ts_config *config); 1278 1279 /* 1280 * Packet hash types specify the type of hash in skb_set_hash. 1281 * 1282 * Hash types refer to the protocol layer addresses which are used to 1283 * construct a packet's hash. The hashes are used to differentiate or identify 1284 * flows of the protocol layer for the hash type. Hash types are either 1285 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1286 * 1287 * Properties of hashes: 1288 * 1289 * 1) Two packets in different flows have different hash values 1290 * 2) Two packets in the same flow should have the same hash value 1291 * 1292 * A hash at a higher layer is considered to be more specific. A driver should 1293 * set the most specific hash possible. 1294 * 1295 * A driver cannot indicate a more specific hash than the layer at which a hash 1296 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1297 * 1298 * A driver may indicate a hash level which is less specific than the 1299 * actual layer the hash was computed on. For instance, a hash computed 1300 * at L4 may be considered an L3 hash. This should only be done if the 1301 * driver can't unambiguously determine that the HW computed the hash at 1302 * the higher layer. Note that the "should" in the second property above 1303 * permits this. 1304 */ 1305 enum pkt_hash_types { 1306 PKT_HASH_TYPE_NONE, /* Undefined type */ 1307 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1308 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1309 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1310 }; 1311 1312 static inline void skb_clear_hash(struct sk_buff *skb) 1313 { 1314 skb->hash = 0; 1315 skb->sw_hash = 0; 1316 skb->l4_hash = 0; 1317 } 1318 1319 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1320 { 1321 if (!skb->l4_hash) 1322 skb_clear_hash(skb); 1323 } 1324 1325 static inline void 1326 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1327 { 1328 skb->l4_hash = is_l4; 1329 skb->sw_hash = is_sw; 1330 skb->hash = hash; 1331 } 1332 1333 static inline void 1334 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1335 { 1336 /* Used by drivers to set hash from HW */ 1337 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1338 } 1339 1340 static inline void 1341 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1342 { 1343 __skb_set_hash(skb, hash, true, is_l4); 1344 } 1345 1346 void __skb_get_hash(struct sk_buff *skb); 1347 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1348 u32 skb_get_poff(const struct sk_buff *skb); 1349 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1350 const struct flow_keys_basic *keys, int hlen); 1351 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1352 const void *data, int hlen_proto); 1353 1354 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1355 int thoff, u8 ip_proto) 1356 { 1357 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1358 } 1359 1360 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1361 const struct flow_dissector_key *key, 1362 unsigned int key_count); 1363 1364 struct bpf_flow_dissector; 1365 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1366 __be16 proto, int nhoff, int hlen, unsigned int flags); 1367 1368 bool __skb_flow_dissect(const struct net *net, 1369 const struct sk_buff *skb, 1370 struct flow_dissector *flow_dissector, 1371 void *target_container, const void *data, 1372 __be16 proto, int nhoff, int hlen, unsigned int flags); 1373 1374 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1375 struct flow_dissector *flow_dissector, 1376 void *target_container, unsigned int flags) 1377 { 1378 return __skb_flow_dissect(NULL, skb, flow_dissector, 1379 target_container, NULL, 0, 0, 0, flags); 1380 } 1381 1382 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1383 struct flow_keys *flow, 1384 unsigned int flags) 1385 { 1386 memset(flow, 0, sizeof(*flow)); 1387 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1388 flow, NULL, 0, 0, 0, flags); 1389 } 1390 1391 static inline bool 1392 skb_flow_dissect_flow_keys_basic(const struct net *net, 1393 const struct sk_buff *skb, 1394 struct flow_keys_basic *flow, 1395 const void *data, __be16 proto, 1396 int nhoff, int hlen, unsigned int flags) 1397 { 1398 memset(flow, 0, sizeof(*flow)); 1399 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1400 data, proto, nhoff, hlen, flags); 1401 } 1402 1403 void skb_flow_dissect_meta(const struct sk_buff *skb, 1404 struct flow_dissector *flow_dissector, 1405 void *target_container); 1406 1407 /* Gets a skb connection tracking info, ctinfo map should be a 1408 * map of mapsize to translate enum ip_conntrack_info states 1409 * to user states. 1410 */ 1411 void 1412 skb_flow_dissect_ct(const struct sk_buff *skb, 1413 struct flow_dissector *flow_dissector, 1414 void *target_container, 1415 u16 *ctinfo_map, size_t mapsize, 1416 bool post_ct, u16 zone); 1417 void 1418 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1419 struct flow_dissector *flow_dissector, 1420 void *target_container); 1421 1422 void skb_flow_dissect_hash(const struct sk_buff *skb, 1423 struct flow_dissector *flow_dissector, 1424 void *target_container); 1425 1426 static inline __u32 skb_get_hash(struct sk_buff *skb) 1427 { 1428 if (!skb->l4_hash && !skb->sw_hash) 1429 __skb_get_hash(skb); 1430 1431 return skb->hash; 1432 } 1433 1434 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1435 { 1436 if (!skb->l4_hash && !skb->sw_hash) { 1437 struct flow_keys keys; 1438 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1439 1440 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1441 } 1442 1443 return skb->hash; 1444 } 1445 1446 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1447 const siphash_key_t *perturb); 1448 1449 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1450 { 1451 return skb->hash; 1452 } 1453 1454 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1455 { 1456 to->hash = from->hash; 1457 to->sw_hash = from->sw_hash; 1458 to->l4_hash = from->l4_hash; 1459 }; 1460 1461 static inline void skb_copy_decrypted(struct sk_buff *to, 1462 const struct sk_buff *from) 1463 { 1464 #ifdef CONFIG_TLS_DEVICE 1465 to->decrypted = from->decrypted; 1466 #endif 1467 } 1468 1469 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1470 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1471 { 1472 return skb->head + skb->end; 1473 } 1474 1475 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1476 { 1477 return skb->end; 1478 } 1479 #else 1480 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1481 { 1482 return skb->end; 1483 } 1484 1485 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1486 { 1487 return skb->end - skb->head; 1488 } 1489 #endif 1490 1491 /* Internal */ 1492 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1493 1494 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1495 { 1496 return &skb_shinfo(skb)->hwtstamps; 1497 } 1498 1499 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1500 { 1501 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1502 1503 return is_zcopy ? skb_uarg(skb) : NULL; 1504 } 1505 1506 static inline bool skb_zcopy_pure(const struct sk_buff *skb) 1507 { 1508 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY; 1509 } 1510 1511 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1, 1512 const struct sk_buff *skb2) 1513 { 1514 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2); 1515 } 1516 1517 static inline void net_zcopy_get(struct ubuf_info *uarg) 1518 { 1519 refcount_inc(&uarg->refcnt); 1520 } 1521 1522 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1523 { 1524 skb_shinfo(skb)->destructor_arg = uarg; 1525 skb_shinfo(skb)->flags |= uarg->flags; 1526 } 1527 1528 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1529 bool *have_ref) 1530 { 1531 if (skb && uarg && !skb_zcopy(skb)) { 1532 if (unlikely(have_ref && *have_ref)) 1533 *have_ref = false; 1534 else 1535 net_zcopy_get(uarg); 1536 skb_zcopy_init(skb, uarg); 1537 } 1538 } 1539 1540 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1541 { 1542 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1543 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1544 } 1545 1546 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1547 { 1548 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1549 } 1550 1551 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1552 { 1553 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1554 } 1555 1556 static inline void net_zcopy_put(struct ubuf_info *uarg) 1557 { 1558 if (uarg) 1559 uarg->callback(NULL, uarg, true); 1560 } 1561 1562 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1563 { 1564 if (uarg) { 1565 if (uarg->callback == msg_zerocopy_callback) 1566 msg_zerocopy_put_abort(uarg, have_uref); 1567 else if (have_uref) 1568 net_zcopy_put(uarg); 1569 } 1570 } 1571 1572 /* Release a reference on a zerocopy structure */ 1573 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1574 { 1575 struct ubuf_info *uarg = skb_zcopy(skb); 1576 1577 if (uarg) { 1578 if (!skb_zcopy_is_nouarg(skb)) 1579 uarg->callback(skb, uarg, zerocopy_success); 1580 1581 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY; 1582 } 1583 } 1584 1585 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1586 { 1587 skb->next = NULL; 1588 } 1589 1590 /* Iterate through singly-linked GSO fragments of an skb. */ 1591 #define skb_list_walk_safe(first, skb, next_skb) \ 1592 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1593 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1594 1595 static inline void skb_list_del_init(struct sk_buff *skb) 1596 { 1597 __list_del_entry(&skb->list); 1598 skb_mark_not_on_list(skb); 1599 } 1600 1601 /** 1602 * skb_queue_empty - check if a queue is empty 1603 * @list: queue head 1604 * 1605 * Returns true if the queue is empty, false otherwise. 1606 */ 1607 static inline int skb_queue_empty(const struct sk_buff_head *list) 1608 { 1609 return list->next == (const struct sk_buff *) list; 1610 } 1611 1612 /** 1613 * skb_queue_empty_lockless - check if a queue is empty 1614 * @list: queue head 1615 * 1616 * Returns true if the queue is empty, false otherwise. 1617 * This variant can be used in lockless contexts. 1618 */ 1619 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1620 { 1621 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1622 } 1623 1624 1625 /** 1626 * skb_queue_is_last - check if skb is the last entry in the queue 1627 * @list: queue head 1628 * @skb: buffer 1629 * 1630 * Returns true if @skb is the last buffer on the list. 1631 */ 1632 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1633 const struct sk_buff *skb) 1634 { 1635 return skb->next == (const struct sk_buff *) list; 1636 } 1637 1638 /** 1639 * skb_queue_is_first - check if skb is the first entry in the queue 1640 * @list: queue head 1641 * @skb: buffer 1642 * 1643 * Returns true if @skb is the first buffer on the list. 1644 */ 1645 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1646 const struct sk_buff *skb) 1647 { 1648 return skb->prev == (const struct sk_buff *) list; 1649 } 1650 1651 /** 1652 * skb_queue_next - return the next packet in the queue 1653 * @list: queue head 1654 * @skb: current buffer 1655 * 1656 * Return the next packet in @list after @skb. It is only valid to 1657 * call this if skb_queue_is_last() evaluates to false. 1658 */ 1659 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1660 const struct sk_buff *skb) 1661 { 1662 /* This BUG_ON may seem severe, but if we just return then we 1663 * are going to dereference garbage. 1664 */ 1665 BUG_ON(skb_queue_is_last(list, skb)); 1666 return skb->next; 1667 } 1668 1669 /** 1670 * skb_queue_prev - return the prev packet in the queue 1671 * @list: queue head 1672 * @skb: current buffer 1673 * 1674 * Return the prev packet in @list before @skb. It is only valid to 1675 * call this if skb_queue_is_first() evaluates to false. 1676 */ 1677 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1678 const struct sk_buff *skb) 1679 { 1680 /* This BUG_ON may seem severe, but if we just return then we 1681 * are going to dereference garbage. 1682 */ 1683 BUG_ON(skb_queue_is_first(list, skb)); 1684 return skb->prev; 1685 } 1686 1687 /** 1688 * skb_get - reference buffer 1689 * @skb: buffer to reference 1690 * 1691 * Makes another reference to a socket buffer and returns a pointer 1692 * to the buffer. 1693 */ 1694 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1695 { 1696 refcount_inc(&skb->users); 1697 return skb; 1698 } 1699 1700 /* 1701 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1702 */ 1703 1704 /** 1705 * skb_cloned - is the buffer a clone 1706 * @skb: buffer to check 1707 * 1708 * Returns true if the buffer was generated with skb_clone() and is 1709 * one of multiple shared copies of the buffer. Cloned buffers are 1710 * shared data so must not be written to under normal circumstances. 1711 */ 1712 static inline int skb_cloned(const struct sk_buff *skb) 1713 { 1714 return skb->cloned && 1715 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1716 } 1717 1718 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1719 { 1720 might_sleep_if(gfpflags_allow_blocking(pri)); 1721 1722 if (skb_cloned(skb)) 1723 return pskb_expand_head(skb, 0, 0, pri); 1724 1725 return 0; 1726 } 1727 1728 /* This variant of skb_unclone() makes sure skb->truesize is not changed */ 1729 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri) 1730 { 1731 might_sleep_if(gfpflags_allow_blocking(pri)); 1732 1733 if (skb_cloned(skb)) { 1734 unsigned int save = skb->truesize; 1735 int res; 1736 1737 res = pskb_expand_head(skb, 0, 0, pri); 1738 skb->truesize = save; 1739 return res; 1740 } 1741 return 0; 1742 } 1743 1744 /** 1745 * skb_header_cloned - is the header a clone 1746 * @skb: buffer to check 1747 * 1748 * Returns true if modifying the header part of the buffer requires 1749 * the data to be copied. 1750 */ 1751 static inline int skb_header_cloned(const struct sk_buff *skb) 1752 { 1753 int dataref; 1754 1755 if (!skb->cloned) 1756 return 0; 1757 1758 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1759 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1760 return dataref != 1; 1761 } 1762 1763 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1764 { 1765 might_sleep_if(gfpflags_allow_blocking(pri)); 1766 1767 if (skb_header_cloned(skb)) 1768 return pskb_expand_head(skb, 0, 0, pri); 1769 1770 return 0; 1771 } 1772 1773 /** 1774 * __skb_header_release - release reference to header 1775 * @skb: buffer to operate on 1776 */ 1777 static inline void __skb_header_release(struct sk_buff *skb) 1778 { 1779 skb->nohdr = 1; 1780 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1781 } 1782 1783 1784 /** 1785 * skb_shared - is the buffer shared 1786 * @skb: buffer to check 1787 * 1788 * Returns true if more than one person has a reference to this 1789 * buffer. 1790 */ 1791 static inline int skb_shared(const struct sk_buff *skb) 1792 { 1793 return refcount_read(&skb->users) != 1; 1794 } 1795 1796 /** 1797 * skb_share_check - check if buffer is shared and if so clone it 1798 * @skb: buffer to check 1799 * @pri: priority for memory allocation 1800 * 1801 * If the buffer is shared the buffer is cloned and the old copy 1802 * drops a reference. A new clone with a single reference is returned. 1803 * If the buffer is not shared the original buffer is returned. When 1804 * being called from interrupt status or with spinlocks held pri must 1805 * be GFP_ATOMIC. 1806 * 1807 * NULL is returned on a memory allocation failure. 1808 */ 1809 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1810 { 1811 might_sleep_if(gfpflags_allow_blocking(pri)); 1812 if (skb_shared(skb)) { 1813 struct sk_buff *nskb = skb_clone(skb, pri); 1814 1815 if (likely(nskb)) 1816 consume_skb(skb); 1817 else 1818 kfree_skb(skb); 1819 skb = nskb; 1820 } 1821 return skb; 1822 } 1823 1824 /* 1825 * Copy shared buffers into a new sk_buff. We effectively do COW on 1826 * packets to handle cases where we have a local reader and forward 1827 * and a couple of other messy ones. The normal one is tcpdumping 1828 * a packet thats being forwarded. 1829 */ 1830 1831 /** 1832 * skb_unshare - make a copy of a shared buffer 1833 * @skb: buffer to check 1834 * @pri: priority for memory allocation 1835 * 1836 * If the socket buffer is a clone then this function creates a new 1837 * copy of the data, drops a reference count on the old copy and returns 1838 * the new copy with the reference count at 1. If the buffer is not a clone 1839 * the original buffer is returned. When called with a spinlock held or 1840 * from interrupt state @pri must be %GFP_ATOMIC 1841 * 1842 * %NULL is returned on a memory allocation failure. 1843 */ 1844 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1845 gfp_t pri) 1846 { 1847 might_sleep_if(gfpflags_allow_blocking(pri)); 1848 if (skb_cloned(skb)) { 1849 struct sk_buff *nskb = skb_copy(skb, pri); 1850 1851 /* Free our shared copy */ 1852 if (likely(nskb)) 1853 consume_skb(skb); 1854 else 1855 kfree_skb(skb); 1856 skb = nskb; 1857 } 1858 return skb; 1859 } 1860 1861 /** 1862 * skb_peek - peek at the head of an &sk_buff_head 1863 * @list_: list to peek at 1864 * 1865 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1866 * be careful with this one. A peek leaves the buffer on the 1867 * list and someone else may run off with it. You must hold 1868 * the appropriate locks or have a private queue to do this. 1869 * 1870 * Returns %NULL for an empty list or a pointer to the head element. 1871 * The reference count is not incremented and the reference is therefore 1872 * volatile. Use with caution. 1873 */ 1874 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1875 { 1876 struct sk_buff *skb = list_->next; 1877 1878 if (skb == (struct sk_buff *)list_) 1879 skb = NULL; 1880 return skb; 1881 } 1882 1883 /** 1884 * __skb_peek - peek at the head of a non-empty &sk_buff_head 1885 * @list_: list to peek at 1886 * 1887 * Like skb_peek(), but the caller knows that the list is not empty. 1888 */ 1889 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 1890 { 1891 return list_->next; 1892 } 1893 1894 /** 1895 * skb_peek_next - peek skb following the given one from a queue 1896 * @skb: skb to start from 1897 * @list_: list to peek at 1898 * 1899 * Returns %NULL when the end of the list is met or a pointer to the 1900 * next element. The reference count is not incremented and the 1901 * reference is therefore volatile. Use with caution. 1902 */ 1903 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1904 const struct sk_buff_head *list_) 1905 { 1906 struct sk_buff *next = skb->next; 1907 1908 if (next == (struct sk_buff *)list_) 1909 next = NULL; 1910 return next; 1911 } 1912 1913 /** 1914 * skb_peek_tail - peek at the tail of an &sk_buff_head 1915 * @list_: list to peek at 1916 * 1917 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1918 * be careful with this one. A peek leaves the buffer on the 1919 * list and someone else may run off with it. You must hold 1920 * the appropriate locks or have a private queue to do this. 1921 * 1922 * Returns %NULL for an empty list or a pointer to the tail element. 1923 * The reference count is not incremented and the reference is therefore 1924 * volatile. Use with caution. 1925 */ 1926 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1927 { 1928 struct sk_buff *skb = READ_ONCE(list_->prev); 1929 1930 if (skb == (struct sk_buff *)list_) 1931 skb = NULL; 1932 return skb; 1933 1934 } 1935 1936 /** 1937 * skb_queue_len - get queue length 1938 * @list_: list to measure 1939 * 1940 * Return the length of an &sk_buff queue. 1941 */ 1942 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1943 { 1944 return list_->qlen; 1945 } 1946 1947 /** 1948 * skb_queue_len_lockless - get queue length 1949 * @list_: list to measure 1950 * 1951 * Return the length of an &sk_buff queue. 1952 * This variant can be used in lockless contexts. 1953 */ 1954 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 1955 { 1956 return READ_ONCE(list_->qlen); 1957 } 1958 1959 /** 1960 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1961 * @list: queue to initialize 1962 * 1963 * This initializes only the list and queue length aspects of 1964 * an sk_buff_head object. This allows to initialize the list 1965 * aspects of an sk_buff_head without reinitializing things like 1966 * the spinlock. It can also be used for on-stack sk_buff_head 1967 * objects where the spinlock is known to not be used. 1968 */ 1969 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1970 { 1971 list->prev = list->next = (struct sk_buff *)list; 1972 list->qlen = 0; 1973 } 1974 1975 /* 1976 * This function creates a split out lock class for each invocation; 1977 * this is needed for now since a whole lot of users of the skb-queue 1978 * infrastructure in drivers have different locking usage (in hardirq) 1979 * than the networking core (in softirq only). In the long run either the 1980 * network layer or drivers should need annotation to consolidate the 1981 * main types of usage into 3 classes. 1982 */ 1983 static inline void skb_queue_head_init(struct sk_buff_head *list) 1984 { 1985 spin_lock_init(&list->lock); 1986 __skb_queue_head_init(list); 1987 } 1988 1989 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1990 struct lock_class_key *class) 1991 { 1992 skb_queue_head_init(list); 1993 lockdep_set_class(&list->lock, class); 1994 } 1995 1996 /* 1997 * Insert an sk_buff on a list. 1998 * 1999 * The "__skb_xxxx()" functions are the non-atomic ones that 2000 * can only be called with interrupts disabled. 2001 */ 2002 static inline void __skb_insert(struct sk_buff *newsk, 2003 struct sk_buff *prev, struct sk_buff *next, 2004 struct sk_buff_head *list) 2005 { 2006 /* See skb_queue_empty_lockless() and skb_peek_tail() 2007 * for the opposite READ_ONCE() 2008 */ 2009 WRITE_ONCE(newsk->next, next); 2010 WRITE_ONCE(newsk->prev, prev); 2011 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk); 2012 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk); 2013 WRITE_ONCE(list->qlen, list->qlen + 1); 2014 } 2015 2016 static inline void __skb_queue_splice(const struct sk_buff_head *list, 2017 struct sk_buff *prev, 2018 struct sk_buff *next) 2019 { 2020 struct sk_buff *first = list->next; 2021 struct sk_buff *last = list->prev; 2022 2023 WRITE_ONCE(first->prev, prev); 2024 WRITE_ONCE(prev->next, first); 2025 2026 WRITE_ONCE(last->next, next); 2027 WRITE_ONCE(next->prev, last); 2028 } 2029 2030 /** 2031 * skb_queue_splice - join two skb lists, this is designed for stacks 2032 * @list: the new list to add 2033 * @head: the place to add it in the first list 2034 */ 2035 static inline void skb_queue_splice(const struct sk_buff_head *list, 2036 struct sk_buff_head *head) 2037 { 2038 if (!skb_queue_empty(list)) { 2039 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 2040 head->qlen += list->qlen; 2041 } 2042 } 2043 2044 /** 2045 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 2046 * @list: the new list to add 2047 * @head: the place to add it in the first list 2048 * 2049 * The list at @list is reinitialised 2050 */ 2051 static inline void skb_queue_splice_init(struct sk_buff_head *list, 2052 struct sk_buff_head *head) 2053 { 2054 if (!skb_queue_empty(list)) { 2055 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 2056 head->qlen += list->qlen; 2057 __skb_queue_head_init(list); 2058 } 2059 } 2060 2061 /** 2062 * skb_queue_splice_tail - join two skb lists, each list being a queue 2063 * @list: the new list to add 2064 * @head: the place to add it in the first list 2065 */ 2066 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 2067 struct sk_buff_head *head) 2068 { 2069 if (!skb_queue_empty(list)) { 2070 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2071 head->qlen += list->qlen; 2072 } 2073 } 2074 2075 /** 2076 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 2077 * @list: the new list to add 2078 * @head: the place to add it in the first list 2079 * 2080 * Each of the lists is a queue. 2081 * The list at @list is reinitialised 2082 */ 2083 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2084 struct sk_buff_head *head) 2085 { 2086 if (!skb_queue_empty(list)) { 2087 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2088 head->qlen += list->qlen; 2089 __skb_queue_head_init(list); 2090 } 2091 } 2092 2093 /** 2094 * __skb_queue_after - queue a buffer at the list head 2095 * @list: list to use 2096 * @prev: place after this buffer 2097 * @newsk: buffer to queue 2098 * 2099 * Queue a buffer int the middle of a list. This function takes no locks 2100 * and you must therefore hold required locks before calling it. 2101 * 2102 * A buffer cannot be placed on two lists at the same time. 2103 */ 2104 static inline void __skb_queue_after(struct sk_buff_head *list, 2105 struct sk_buff *prev, 2106 struct sk_buff *newsk) 2107 { 2108 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list); 2109 } 2110 2111 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2112 struct sk_buff_head *list); 2113 2114 static inline void __skb_queue_before(struct sk_buff_head *list, 2115 struct sk_buff *next, 2116 struct sk_buff *newsk) 2117 { 2118 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list); 2119 } 2120 2121 /** 2122 * __skb_queue_head - queue a buffer at the list head 2123 * @list: list to use 2124 * @newsk: buffer to queue 2125 * 2126 * Queue a buffer at the start of a list. This function takes no locks 2127 * and you must therefore hold required locks before calling it. 2128 * 2129 * A buffer cannot be placed on two lists at the same time. 2130 */ 2131 static inline void __skb_queue_head(struct sk_buff_head *list, 2132 struct sk_buff *newsk) 2133 { 2134 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2135 } 2136 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2137 2138 /** 2139 * __skb_queue_tail - queue a buffer at the list tail 2140 * @list: list to use 2141 * @newsk: buffer to queue 2142 * 2143 * Queue a buffer at the end of a list. This function takes no locks 2144 * and you must therefore hold required locks before calling it. 2145 * 2146 * A buffer cannot be placed on two lists at the same time. 2147 */ 2148 static inline void __skb_queue_tail(struct sk_buff_head *list, 2149 struct sk_buff *newsk) 2150 { 2151 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2152 } 2153 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2154 2155 /* 2156 * remove sk_buff from list. _Must_ be called atomically, and with 2157 * the list known.. 2158 */ 2159 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2160 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2161 { 2162 struct sk_buff *next, *prev; 2163 2164 WRITE_ONCE(list->qlen, list->qlen - 1); 2165 next = skb->next; 2166 prev = skb->prev; 2167 skb->next = skb->prev = NULL; 2168 WRITE_ONCE(next->prev, prev); 2169 WRITE_ONCE(prev->next, next); 2170 } 2171 2172 /** 2173 * __skb_dequeue - remove from the head of the queue 2174 * @list: list to dequeue from 2175 * 2176 * Remove the head of the list. This function does not take any locks 2177 * so must be used with appropriate locks held only. The head item is 2178 * returned or %NULL if the list is empty. 2179 */ 2180 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2181 { 2182 struct sk_buff *skb = skb_peek(list); 2183 if (skb) 2184 __skb_unlink(skb, list); 2185 return skb; 2186 } 2187 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2188 2189 /** 2190 * __skb_dequeue_tail - remove from the tail of the queue 2191 * @list: list to dequeue from 2192 * 2193 * Remove the tail of the list. This function does not take any locks 2194 * so must be used with appropriate locks held only. The tail item is 2195 * returned or %NULL if the list is empty. 2196 */ 2197 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2198 { 2199 struct sk_buff *skb = skb_peek_tail(list); 2200 if (skb) 2201 __skb_unlink(skb, list); 2202 return skb; 2203 } 2204 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2205 2206 2207 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2208 { 2209 return skb->data_len; 2210 } 2211 2212 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2213 { 2214 return skb->len - skb->data_len; 2215 } 2216 2217 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2218 { 2219 unsigned int i, len = 0; 2220 2221 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2222 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2223 return len; 2224 } 2225 2226 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2227 { 2228 return skb_headlen(skb) + __skb_pagelen(skb); 2229 } 2230 2231 /** 2232 * __skb_fill_page_desc - initialise a paged fragment in an skb 2233 * @skb: buffer containing fragment to be initialised 2234 * @i: paged fragment index to initialise 2235 * @page: the page to use for this fragment 2236 * @off: the offset to the data with @page 2237 * @size: the length of the data 2238 * 2239 * Initialises the @i'th fragment of @skb to point to &size bytes at 2240 * offset @off within @page. 2241 * 2242 * Does not take any additional reference on the fragment. 2243 */ 2244 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2245 struct page *page, int off, int size) 2246 { 2247 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2248 2249 /* 2250 * Propagate page pfmemalloc to the skb if we can. The problem is 2251 * that not all callers have unique ownership of the page but rely 2252 * on page_is_pfmemalloc doing the right thing(tm). 2253 */ 2254 frag->bv_page = page; 2255 frag->bv_offset = off; 2256 skb_frag_size_set(frag, size); 2257 2258 page = compound_head(page); 2259 if (page_is_pfmemalloc(page)) 2260 skb->pfmemalloc = true; 2261 } 2262 2263 /** 2264 * skb_fill_page_desc - initialise a paged fragment in an skb 2265 * @skb: buffer containing fragment to be initialised 2266 * @i: paged fragment index to initialise 2267 * @page: the page to use for this fragment 2268 * @off: the offset to the data with @page 2269 * @size: the length of the data 2270 * 2271 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2272 * @skb to point to @size bytes at offset @off within @page. In 2273 * addition updates @skb such that @i is the last fragment. 2274 * 2275 * Does not take any additional reference on the fragment. 2276 */ 2277 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2278 struct page *page, int off, int size) 2279 { 2280 __skb_fill_page_desc(skb, i, page, off, size); 2281 skb_shinfo(skb)->nr_frags = i + 1; 2282 } 2283 2284 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2285 int size, unsigned int truesize); 2286 2287 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2288 unsigned int truesize); 2289 2290 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2291 2292 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2293 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2294 { 2295 return skb->head + skb->tail; 2296 } 2297 2298 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2299 { 2300 skb->tail = skb->data - skb->head; 2301 } 2302 2303 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2304 { 2305 skb_reset_tail_pointer(skb); 2306 skb->tail += offset; 2307 } 2308 2309 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2310 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2311 { 2312 return skb->tail; 2313 } 2314 2315 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2316 { 2317 skb->tail = skb->data; 2318 } 2319 2320 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2321 { 2322 skb->tail = skb->data + offset; 2323 } 2324 2325 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2326 2327 /* 2328 * Add data to an sk_buff 2329 */ 2330 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2331 void *skb_put(struct sk_buff *skb, unsigned int len); 2332 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2333 { 2334 void *tmp = skb_tail_pointer(skb); 2335 SKB_LINEAR_ASSERT(skb); 2336 skb->tail += len; 2337 skb->len += len; 2338 return tmp; 2339 } 2340 2341 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2342 { 2343 void *tmp = __skb_put(skb, len); 2344 2345 memset(tmp, 0, len); 2346 return tmp; 2347 } 2348 2349 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2350 unsigned int len) 2351 { 2352 void *tmp = __skb_put(skb, len); 2353 2354 memcpy(tmp, data, len); 2355 return tmp; 2356 } 2357 2358 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2359 { 2360 *(u8 *)__skb_put(skb, 1) = val; 2361 } 2362 2363 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2364 { 2365 void *tmp = skb_put(skb, len); 2366 2367 memset(tmp, 0, len); 2368 2369 return tmp; 2370 } 2371 2372 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2373 unsigned int len) 2374 { 2375 void *tmp = skb_put(skb, len); 2376 2377 memcpy(tmp, data, len); 2378 2379 return tmp; 2380 } 2381 2382 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2383 { 2384 *(u8 *)skb_put(skb, 1) = val; 2385 } 2386 2387 void *skb_push(struct sk_buff *skb, unsigned int len); 2388 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2389 { 2390 skb->data -= len; 2391 skb->len += len; 2392 return skb->data; 2393 } 2394 2395 void *skb_pull(struct sk_buff *skb, unsigned int len); 2396 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2397 { 2398 skb->len -= len; 2399 BUG_ON(skb->len < skb->data_len); 2400 return skb->data += len; 2401 } 2402 2403 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2404 { 2405 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2406 } 2407 2408 void *skb_pull_data(struct sk_buff *skb, size_t len); 2409 2410 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2411 2412 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len) 2413 { 2414 if (len > skb_headlen(skb) && 2415 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 2416 return NULL; 2417 skb->len -= len; 2418 return skb->data += len; 2419 } 2420 2421 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2422 { 2423 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 2424 } 2425 2426 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2427 { 2428 if (likely(len <= skb_headlen(skb))) 2429 return true; 2430 if (unlikely(len > skb->len)) 2431 return false; 2432 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 2433 } 2434 2435 void skb_condense(struct sk_buff *skb); 2436 2437 /** 2438 * skb_headroom - bytes at buffer head 2439 * @skb: buffer to check 2440 * 2441 * Return the number of bytes of free space at the head of an &sk_buff. 2442 */ 2443 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2444 { 2445 return skb->data - skb->head; 2446 } 2447 2448 /** 2449 * skb_tailroom - bytes at buffer end 2450 * @skb: buffer to check 2451 * 2452 * Return the number of bytes of free space at the tail of an sk_buff 2453 */ 2454 static inline int skb_tailroom(const struct sk_buff *skb) 2455 { 2456 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2457 } 2458 2459 /** 2460 * skb_availroom - bytes at buffer end 2461 * @skb: buffer to check 2462 * 2463 * Return the number of bytes of free space at the tail of an sk_buff 2464 * allocated by sk_stream_alloc() 2465 */ 2466 static inline int skb_availroom(const struct sk_buff *skb) 2467 { 2468 if (skb_is_nonlinear(skb)) 2469 return 0; 2470 2471 return skb->end - skb->tail - skb->reserved_tailroom; 2472 } 2473 2474 /** 2475 * skb_reserve - adjust headroom 2476 * @skb: buffer to alter 2477 * @len: bytes to move 2478 * 2479 * Increase the headroom of an empty &sk_buff by reducing the tail 2480 * room. This is only allowed for an empty buffer. 2481 */ 2482 static inline void skb_reserve(struct sk_buff *skb, int len) 2483 { 2484 skb->data += len; 2485 skb->tail += len; 2486 } 2487 2488 /** 2489 * skb_tailroom_reserve - adjust reserved_tailroom 2490 * @skb: buffer to alter 2491 * @mtu: maximum amount of headlen permitted 2492 * @needed_tailroom: minimum amount of reserved_tailroom 2493 * 2494 * Set reserved_tailroom so that headlen can be as large as possible but 2495 * not larger than mtu and tailroom cannot be smaller than 2496 * needed_tailroom. 2497 * The required headroom should already have been reserved before using 2498 * this function. 2499 */ 2500 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2501 unsigned int needed_tailroom) 2502 { 2503 SKB_LINEAR_ASSERT(skb); 2504 if (mtu < skb_tailroom(skb) - needed_tailroom) 2505 /* use at most mtu */ 2506 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2507 else 2508 /* use up to all available space */ 2509 skb->reserved_tailroom = needed_tailroom; 2510 } 2511 2512 #define ENCAP_TYPE_ETHER 0 2513 #define ENCAP_TYPE_IPPROTO 1 2514 2515 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2516 __be16 protocol) 2517 { 2518 skb->inner_protocol = protocol; 2519 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2520 } 2521 2522 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2523 __u8 ipproto) 2524 { 2525 skb->inner_ipproto = ipproto; 2526 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2527 } 2528 2529 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2530 { 2531 skb->inner_mac_header = skb->mac_header; 2532 skb->inner_network_header = skb->network_header; 2533 skb->inner_transport_header = skb->transport_header; 2534 } 2535 2536 static inline void skb_reset_mac_len(struct sk_buff *skb) 2537 { 2538 skb->mac_len = skb->network_header - skb->mac_header; 2539 } 2540 2541 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2542 *skb) 2543 { 2544 return skb->head + skb->inner_transport_header; 2545 } 2546 2547 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2548 { 2549 return skb_inner_transport_header(skb) - skb->data; 2550 } 2551 2552 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2553 { 2554 skb->inner_transport_header = skb->data - skb->head; 2555 } 2556 2557 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2558 const int offset) 2559 { 2560 skb_reset_inner_transport_header(skb); 2561 skb->inner_transport_header += offset; 2562 } 2563 2564 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2565 { 2566 return skb->head + skb->inner_network_header; 2567 } 2568 2569 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2570 { 2571 skb->inner_network_header = skb->data - skb->head; 2572 } 2573 2574 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2575 const int offset) 2576 { 2577 skb_reset_inner_network_header(skb); 2578 skb->inner_network_header += offset; 2579 } 2580 2581 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2582 { 2583 return skb->head + skb->inner_mac_header; 2584 } 2585 2586 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2587 { 2588 skb->inner_mac_header = skb->data - skb->head; 2589 } 2590 2591 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2592 const int offset) 2593 { 2594 skb_reset_inner_mac_header(skb); 2595 skb->inner_mac_header += offset; 2596 } 2597 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2598 { 2599 return skb->transport_header != (typeof(skb->transport_header))~0U; 2600 } 2601 2602 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2603 { 2604 return skb->head + skb->transport_header; 2605 } 2606 2607 static inline void skb_reset_transport_header(struct sk_buff *skb) 2608 { 2609 skb->transport_header = skb->data - skb->head; 2610 } 2611 2612 static inline void skb_set_transport_header(struct sk_buff *skb, 2613 const int offset) 2614 { 2615 skb_reset_transport_header(skb); 2616 skb->transport_header += offset; 2617 } 2618 2619 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2620 { 2621 return skb->head + skb->network_header; 2622 } 2623 2624 static inline void skb_reset_network_header(struct sk_buff *skb) 2625 { 2626 skb->network_header = skb->data - skb->head; 2627 } 2628 2629 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2630 { 2631 skb_reset_network_header(skb); 2632 skb->network_header += offset; 2633 } 2634 2635 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2636 { 2637 return skb->head + skb->mac_header; 2638 } 2639 2640 static inline int skb_mac_offset(const struct sk_buff *skb) 2641 { 2642 return skb_mac_header(skb) - skb->data; 2643 } 2644 2645 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2646 { 2647 return skb->network_header - skb->mac_header; 2648 } 2649 2650 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2651 { 2652 return skb->mac_header != (typeof(skb->mac_header))~0U; 2653 } 2654 2655 static inline void skb_unset_mac_header(struct sk_buff *skb) 2656 { 2657 skb->mac_header = (typeof(skb->mac_header))~0U; 2658 } 2659 2660 static inline void skb_reset_mac_header(struct sk_buff *skb) 2661 { 2662 skb->mac_header = skb->data - skb->head; 2663 } 2664 2665 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2666 { 2667 skb_reset_mac_header(skb); 2668 skb->mac_header += offset; 2669 } 2670 2671 static inline void skb_pop_mac_header(struct sk_buff *skb) 2672 { 2673 skb->mac_header = skb->network_header; 2674 } 2675 2676 static inline void skb_probe_transport_header(struct sk_buff *skb) 2677 { 2678 struct flow_keys_basic keys; 2679 2680 if (skb_transport_header_was_set(skb)) 2681 return; 2682 2683 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2684 NULL, 0, 0, 0, 0)) 2685 skb_set_transport_header(skb, keys.control.thoff); 2686 } 2687 2688 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2689 { 2690 if (skb_mac_header_was_set(skb)) { 2691 const unsigned char *old_mac = skb_mac_header(skb); 2692 2693 skb_set_mac_header(skb, -skb->mac_len); 2694 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2695 } 2696 } 2697 2698 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2699 { 2700 return skb->csum_start - skb_headroom(skb); 2701 } 2702 2703 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2704 { 2705 return skb->head + skb->csum_start; 2706 } 2707 2708 static inline int skb_transport_offset(const struct sk_buff *skb) 2709 { 2710 return skb_transport_header(skb) - skb->data; 2711 } 2712 2713 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2714 { 2715 return skb->transport_header - skb->network_header; 2716 } 2717 2718 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2719 { 2720 return skb->inner_transport_header - skb->inner_network_header; 2721 } 2722 2723 static inline int skb_network_offset(const struct sk_buff *skb) 2724 { 2725 return skb_network_header(skb) - skb->data; 2726 } 2727 2728 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2729 { 2730 return skb_inner_network_header(skb) - skb->data; 2731 } 2732 2733 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2734 { 2735 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2736 } 2737 2738 /* 2739 * CPUs often take a performance hit when accessing unaligned memory 2740 * locations. The actual performance hit varies, it can be small if the 2741 * hardware handles it or large if we have to take an exception and fix it 2742 * in software. 2743 * 2744 * Since an ethernet header is 14 bytes network drivers often end up with 2745 * the IP header at an unaligned offset. The IP header can be aligned by 2746 * shifting the start of the packet by 2 bytes. Drivers should do this 2747 * with: 2748 * 2749 * skb_reserve(skb, NET_IP_ALIGN); 2750 * 2751 * The downside to this alignment of the IP header is that the DMA is now 2752 * unaligned. On some architectures the cost of an unaligned DMA is high 2753 * and this cost outweighs the gains made by aligning the IP header. 2754 * 2755 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2756 * to be overridden. 2757 */ 2758 #ifndef NET_IP_ALIGN 2759 #define NET_IP_ALIGN 2 2760 #endif 2761 2762 /* 2763 * The networking layer reserves some headroom in skb data (via 2764 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2765 * the header has to grow. In the default case, if the header has to grow 2766 * 32 bytes or less we avoid the reallocation. 2767 * 2768 * Unfortunately this headroom changes the DMA alignment of the resulting 2769 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2770 * on some architectures. An architecture can override this value, 2771 * perhaps setting it to a cacheline in size (since that will maintain 2772 * cacheline alignment of the DMA). It must be a power of 2. 2773 * 2774 * Various parts of the networking layer expect at least 32 bytes of 2775 * headroom, you should not reduce this. 2776 * 2777 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2778 * to reduce average number of cache lines per packet. 2779 * get_rps_cpu() for example only access one 64 bytes aligned block : 2780 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2781 */ 2782 #ifndef NET_SKB_PAD 2783 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2784 #endif 2785 2786 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2787 2788 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 2789 { 2790 if (WARN_ON(skb_is_nonlinear(skb))) 2791 return; 2792 skb->len = len; 2793 skb_set_tail_pointer(skb, len); 2794 } 2795 2796 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2797 { 2798 __skb_set_length(skb, len); 2799 } 2800 2801 void skb_trim(struct sk_buff *skb, unsigned int len); 2802 2803 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2804 { 2805 if (skb->data_len) 2806 return ___pskb_trim(skb, len); 2807 __skb_trim(skb, len); 2808 return 0; 2809 } 2810 2811 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2812 { 2813 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2814 } 2815 2816 /** 2817 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2818 * @skb: buffer to alter 2819 * @len: new length 2820 * 2821 * This is identical to pskb_trim except that the caller knows that 2822 * the skb is not cloned so we should never get an error due to out- 2823 * of-memory. 2824 */ 2825 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2826 { 2827 int err = pskb_trim(skb, len); 2828 BUG_ON(err); 2829 } 2830 2831 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 2832 { 2833 unsigned int diff = len - skb->len; 2834 2835 if (skb_tailroom(skb) < diff) { 2836 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 2837 GFP_ATOMIC); 2838 if (ret) 2839 return ret; 2840 } 2841 __skb_set_length(skb, len); 2842 return 0; 2843 } 2844 2845 /** 2846 * skb_orphan - orphan a buffer 2847 * @skb: buffer to orphan 2848 * 2849 * If a buffer currently has an owner then we call the owner's 2850 * destructor function and make the @skb unowned. The buffer continues 2851 * to exist but is no longer charged to its former owner. 2852 */ 2853 static inline void skb_orphan(struct sk_buff *skb) 2854 { 2855 if (skb->destructor) { 2856 skb->destructor(skb); 2857 skb->destructor = NULL; 2858 skb->sk = NULL; 2859 } else { 2860 BUG_ON(skb->sk); 2861 } 2862 } 2863 2864 /** 2865 * skb_orphan_frags - orphan the frags contained in a buffer 2866 * @skb: buffer to orphan frags from 2867 * @gfp_mask: allocation mask for replacement pages 2868 * 2869 * For each frag in the SKB which needs a destructor (i.e. has an 2870 * owner) create a copy of that frag and release the original 2871 * page by calling the destructor. 2872 */ 2873 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2874 { 2875 if (likely(!skb_zcopy(skb))) 2876 return 0; 2877 if (!skb_zcopy_is_nouarg(skb) && 2878 skb_uarg(skb)->callback == msg_zerocopy_callback) 2879 return 0; 2880 return skb_copy_ubufs(skb, gfp_mask); 2881 } 2882 2883 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 2884 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 2885 { 2886 if (likely(!skb_zcopy(skb))) 2887 return 0; 2888 return skb_copy_ubufs(skb, gfp_mask); 2889 } 2890 2891 /** 2892 * __skb_queue_purge - empty a list 2893 * @list: list to empty 2894 * 2895 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2896 * the list and one reference dropped. This function does not take the 2897 * list lock and the caller must hold the relevant locks to use it. 2898 */ 2899 static inline void __skb_queue_purge(struct sk_buff_head *list) 2900 { 2901 struct sk_buff *skb; 2902 while ((skb = __skb_dequeue(list)) != NULL) 2903 kfree_skb(skb); 2904 } 2905 void skb_queue_purge(struct sk_buff_head *list); 2906 2907 unsigned int skb_rbtree_purge(struct rb_root *root); 2908 2909 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2910 2911 /** 2912 * netdev_alloc_frag - allocate a page fragment 2913 * @fragsz: fragment size 2914 * 2915 * Allocates a frag from a page for receive buffer. 2916 * Uses GFP_ATOMIC allocations. 2917 */ 2918 static inline void *netdev_alloc_frag(unsigned int fragsz) 2919 { 2920 return __netdev_alloc_frag_align(fragsz, ~0u); 2921 } 2922 2923 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 2924 unsigned int align) 2925 { 2926 WARN_ON_ONCE(!is_power_of_2(align)); 2927 return __netdev_alloc_frag_align(fragsz, -align); 2928 } 2929 2930 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2931 gfp_t gfp_mask); 2932 2933 /** 2934 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2935 * @dev: network device to receive on 2936 * @length: length to allocate 2937 * 2938 * Allocate a new &sk_buff and assign it a usage count of one. The 2939 * buffer has unspecified headroom built in. Users should allocate 2940 * the headroom they think they need without accounting for the 2941 * built in space. The built in space is used for optimisations. 2942 * 2943 * %NULL is returned if there is no free memory. Although this function 2944 * allocates memory it can be called from an interrupt. 2945 */ 2946 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2947 unsigned int length) 2948 { 2949 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2950 } 2951 2952 /* legacy helper around __netdev_alloc_skb() */ 2953 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2954 gfp_t gfp_mask) 2955 { 2956 return __netdev_alloc_skb(NULL, length, gfp_mask); 2957 } 2958 2959 /* legacy helper around netdev_alloc_skb() */ 2960 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2961 { 2962 return netdev_alloc_skb(NULL, length); 2963 } 2964 2965 2966 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2967 unsigned int length, gfp_t gfp) 2968 { 2969 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2970 2971 if (NET_IP_ALIGN && skb) 2972 skb_reserve(skb, NET_IP_ALIGN); 2973 return skb; 2974 } 2975 2976 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2977 unsigned int length) 2978 { 2979 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2980 } 2981 2982 static inline void skb_free_frag(void *addr) 2983 { 2984 page_frag_free(addr); 2985 } 2986 2987 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 2988 2989 static inline void *napi_alloc_frag(unsigned int fragsz) 2990 { 2991 return __napi_alloc_frag_align(fragsz, ~0u); 2992 } 2993 2994 static inline void *napi_alloc_frag_align(unsigned int fragsz, 2995 unsigned int align) 2996 { 2997 WARN_ON_ONCE(!is_power_of_2(align)); 2998 return __napi_alloc_frag_align(fragsz, -align); 2999 } 3000 3001 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 3002 unsigned int length, gfp_t gfp_mask); 3003 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 3004 unsigned int length) 3005 { 3006 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 3007 } 3008 void napi_consume_skb(struct sk_buff *skb, int budget); 3009 3010 void napi_skb_free_stolen_head(struct sk_buff *skb); 3011 void __kfree_skb_defer(struct sk_buff *skb); 3012 3013 /** 3014 * __dev_alloc_pages - allocate page for network Rx 3015 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 3016 * @order: size of the allocation 3017 * 3018 * Allocate a new page. 3019 * 3020 * %NULL is returned if there is no free memory. 3021 */ 3022 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 3023 unsigned int order) 3024 { 3025 /* This piece of code contains several assumptions. 3026 * 1. This is for device Rx, therefor a cold page is preferred. 3027 * 2. The expectation is the user wants a compound page. 3028 * 3. If requesting a order 0 page it will not be compound 3029 * due to the check to see if order has a value in prep_new_page 3030 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 3031 * code in gfp_to_alloc_flags that should be enforcing this. 3032 */ 3033 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 3034 3035 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 3036 } 3037 3038 static inline struct page *dev_alloc_pages(unsigned int order) 3039 { 3040 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 3041 } 3042 3043 /** 3044 * __dev_alloc_page - allocate a page for network Rx 3045 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 3046 * 3047 * Allocate a new page. 3048 * 3049 * %NULL is returned if there is no free memory. 3050 */ 3051 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 3052 { 3053 return __dev_alloc_pages(gfp_mask, 0); 3054 } 3055 3056 static inline struct page *dev_alloc_page(void) 3057 { 3058 return dev_alloc_pages(0); 3059 } 3060 3061 /** 3062 * dev_page_is_reusable - check whether a page can be reused for network Rx 3063 * @page: the page to test 3064 * 3065 * A page shouldn't be considered for reusing/recycling if it was allocated 3066 * under memory pressure or at a distant memory node. 3067 * 3068 * Returns false if this page should be returned to page allocator, true 3069 * otherwise. 3070 */ 3071 static inline bool dev_page_is_reusable(const struct page *page) 3072 { 3073 return likely(page_to_nid(page) == numa_mem_id() && 3074 !page_is_pfmemalloc(page)); 3075 } 3076 3077 /** 3078 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 3079 * @page: The page that was allocated from skb_alloc_page 3080 * @skb: The skb that may need pfmemalloc set 3081 */ 3082 static inline void skb_propagate_pfmemalloc(const struct page *page, 3083 struct sk_buff *skb) 3084 { 3085 if (page_is_pfmemalloc(page)) 3086 skb->pfmemalloc = true; 3087 } 3088 3089 /** 3090 * skb_frag_off() - Returns the offset of a skb fragment 3091 * @frag: the paged fragment 3092 */ 3093 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3094 { 3095 return frag->bv_offset; 3096 } 3097 3098 /** 3099 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3100 * @frag: skb fragment 3101 * @delta: value to add 3102 */ 3103 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3104 { 3105 frag->bv_offset += delta; 3106 } 3107 3108 /** 3109 * skb_frag_off_set() - Sets the offset of a skb fragment 3110 * @frag: skb fragment 3111 * @offset: offset of fragment 3112 */ 3113 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3114 { 3115 frag->bv_offset = offset; 3116 } 3117 3118 /** 3119 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3120 * @fragto: skb fragment where offset is set 3121 * @fragfrom: skb fragment offset is copied from 3122 */ 3123 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3124 const skb_frag_t *fragfrom) 3125 { 3126 fragto->bv_offset = fragfrom->bv_offset; 3127 } 3128 3129 /** 3130 * skb_frag_page - retrieve the page referred to by a paged fragment 3131 * @frag: the paged fragment 3132 * 3133 * Returns the &struct page associated with @frag. 3134 */ 3135 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3136 { 3137 return frag->bv_page; 3138 } 3139 3140 /** 3141 * __skb_frag_ref - take an addition reference on a paged fragment. 3142 * @frag: the paged fragment 3143 * 3144 * Takes an additional reference on the paged fragment @frag. 3145 */ 3146 static inline void __skb_frag_ref(skb_frag_t *frag) 3147 { 3148 get_page(skb_frag_page(frag)); 3149 } 3150 3151 /** 3152 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3153 * @skb: the buffer 3154 * @f: the fragment offset. 3155 * 3156 * Takes an additional reference on the @f'th paged fragment of @skb. 3157 */ 3158 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3159 { 3160 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3161 } 3162 3163 /** 3164 * __skb_frag_unref - release a reference on a paged fragment. 3165 * @frag: the paged fragment 3166 * @recycle: recycle the page if allocated via page_pool 3167 * 3168 * Releases a reference on the paged fragment @frag 3169 * or recycles the page via the page_pool API. 3170 */ 3171 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) 3172 { 3173 struct page *page = skb_frag_page(frag); 3174 3175 #ifdef CONFIG_PAGE_POOL 3176 if (recycle && page_pool_return_skb_page(page)) 3177 return; 3178 #endif 3179 put_page(page); 3180 } 3181 3182 /** 3183 * skb_frag_unref - release a reference on a paged fragment of an skb. 3184 * @skb: the buffer 3185 * @f: the fragment offset 3186 * 3187 * Releases a reference on the @f'th paged fragment of @skb. 3188 */ 3189 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3190 { 3191 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle); 3192 } 3193 3194 /** 3195 * skb_frag_address - gets the address of the data contained in a paged fragment 3196 * @frag: the paged fragment buffer 3197 * 3198 * Returns the address of the data within @frag. The page must already 3199 * be mapped. 3200 */ 3201 static inline void *skb_frag_address(const skb_frag_t *frag) 3202 { 3203 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3204 } 3205 3206 /** 3207 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3208 * @frag: the paged fragment buffer 3209 * 3210 * Returns the address of the data within @frag. Checks that the page 3211 * is mapped and returns %NULL otherwise. 3212 */ 3213 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3214 { 3215 void *ptr = page_address(skb_frag_page(frag)); 3216 if (unlikely(!ptr)) 3217 return NULL; 3218 3219 return ptr + skb_frag_off(frag); 3220 } 3221 3222 /** 3223 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3224 * @fragto: skb fragment where page is set 3225 * @fragfrom: skb fragment page is copied from 3226 */ 3227 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3228 const skb_frag_t *fragfrom) 3229 { 3230 fragto->bv_page = fragfrom->bv_page; 3231 } 3232 3233 /** 3234 * __skb_frag_set_page - sets the page contained in a paged fragment 3235 * @frag: the paged fragment 3236 * @page: the page to set 3237 * 3238 * Sets the fragment @frag to contain @page. 3239 */ 3240 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3241 { 3242 frag->bv_page = page; 3243 } 3244 3245 /** 3246 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3247 * @skb: the buffer 3248 * @f: the fragment offset 3249 * @page: the page to set 3250 * 3251 * Sets the @f'th fragment of @skb to contain @page. 3252 */ 3253 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3254 struct page *page) 3255 { 3256 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3257 } 3258 3259 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3260 3261 /** 3262 * skb_frag_dma_map - maps a paged fragment via the DMA API 3263 * @dev: the device to map the fragment to 3264 * @frag: the paged fragment to map 3265 * @offset: the offset within the fragment (starting at the 3266 * fragment's own offset) 3267 * @size: the number of bytes to map 3268 * @dir: the direction of the mapping (``PCI_DMA_*``) 3269 * 3270 * Maps the page associated with @frag to @device. 3271 */ 3272 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3273 const skb_frag_t *frag, 3274 size_t offset, size_t size, 3275 enum dma_data_direction dir) 3276 { 3277 return dma_map_page(dev, skb_frag_page(frag), 3278 skb_frag_off(frag) + offset, size, dir); 3279 } 3280 3281 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3282 gfp_t gfp_mask) 3283 { 3284 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3285 } 3286 3287 3288 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3289 gfp_t gfp_mask) 3290 { 3291 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3292 } 3293 3294 3295 /** 3296 * skb_clone_writable - is the header of a clone writable 3297 * @skb: buffer to check 3298 * @len: length up to which to write 3299 * 3300 * Returns true if modifying the header part of the cloned buffer 3301 * does not requires the data to be copied. 3302 */ 3303 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3304 { 3305 return !skb_header_cloned(skb) && 3306 skb_headroom(skb) + len <= skb->hdr_len; 3307 } 3308 3309 static inline int skb_try_make_writable(struct sk_buff *skb, 3310 unsigned int write_len) 3311 { 3312 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3313 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3314 } 3315 3316 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3317 int cloned) 3318 { 3319 int delta = 0; 3320 3321 if (headroom > skb_headroom(skb)) 3322 delta = headroom - skb_headroom(skb); 3323 3324 if (delta || cloned) 3325 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3326 GFP_ATOMIC); 3327 return 0; 3328 } 3329 3330 /** 3331 * skb_cow - copy header of skb when it is required 3332 * @skb: buffer to cow 3333 * @headroom: needed headroom 3334 * 3335 * If the skb passed lacks sufficient headroom or its data part 3336 * is shared, data is reallocated. If reallocation fails, an error 3337 * is returned and original skb is not changed. 3338 * 3339 * The result is skb with writable area skb->head...skb->tail 3340 * and at least @headroom of space at head. 3341 */ 3342 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3343 { 3344 return __skb_cow(skb, headroom, skb_cloned(skb)); 3345 } 3346 3347 /** 3348 * skb_cow_head - skb_cow but only making the head writable 3349 * @skb: buffer to cow 3350 * @headroom: needed headroom 3351 * 3352 * This function is identical to skb_cow except that we replace the 3353 * skb_cloned check by skb_header_cloned. It should be used when 3354 * you only need to push on some header and do not need to modify 3355 * the data. 3356 */ 3357 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3358 { 3359 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3360 } 3361 3362 /** 3363 * skb_padto - pad an skbuff up to a minimal size 3364 * @skb: buffer to pad 3365 * @len: minimal length 3366 * 3367 * Pads up a buffer to ensure the trailing bytes exist and are 3368 * blanked. If the buffer already contains sufficient data it 3369 * is untouched. Otherwise it is extended. Returns zero on 3370 * success. The skb is freed on error. 3371 */ 3372 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3373 { 3374 unsigned int size = skb->len; 3375 if (likely(size >= len)) 3376 return 0; 3377 return skb_pad(skb, len - size); 3378 } 3379 3380 /** 3381 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3382 * @skb: buffer to pad 3383 * @len: minimal length 3384 * @free_on_error: free buffer on error 3385 * 3386 * Pads up a buffer to ensure the trailing bytes exist and are 3387 * blanked. If the buffer already contains sufficient data it 3388 * is untouched. Otherwise it is extended. Returns zero on 3389 * success. The skb is freed on error if @free_on_error is true. 3390 */ 3391 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3392 unsigned int len, 3393 bool free_on_error) 3394 { 3395 unsigned int size = skb->len; 3396 3397 if (unlikely(size < len)) { 3398 len -= size; 3399 if (__skb_pad(skb, len, free_on_error)) 3400 return -ENOMEM; 3401 __skb_put(skb, len); 3402 } 3403 return 0; 3404 } 3405 3406 /** 3407 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3408 * @skb: buffer to pad 3409 * @len: minimal length 3410 * 3411 * Pads up a buffer to ensure the trailing bytes exist and are 3412 * blanked. If the buffer already contains sufficient data it 3413 * is untouched. Otherwise it is extended. Returns zero on 3414 * success. The skb is freed on error. 3415 */ 3416 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3417 { 3418 return __skb_put_padto(skb, len, true); 3419 } 3420 3421 static inline int skb_add_data(struct sk_buff *skb, 3422 struct iov_iter *from, int copy) 3423 { 3424 const int off = skb->len; 3425 3426 if (skb->ip_summed == CHECKSUM_NONE) { 3427 __wsum csum = 0; 3428 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3429 &csum, from)) { 3430 skb->csum = csum_block_add(skb->csum, csum, off); 3431 return 0; 3432 } 3433 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3434 return 0; 3435 3436 __skb_trim(skb, off); 3437 return -EFAULT; 3438 } 3439 3440 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3441 const struct page *page, int off) 3442 { 3443 if (skb_zcopy(skb)) 3444 return false; 3445 if (i) { 3446 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3447 3448 return page == skb_frag_page(frag) && 3449 off == skb_frag_off(frag) + skb_frag_size(frag); 3450 } 3451 return false; 3452 } 3453 3454 static inline int __skb_linearize(struct sk_buff *skb) 3455 { 3456 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3457 } 3458 3459 /** 3460 * skb_linearize - convert paged skb to linear one 3461 * @skb: buffer to linarize 3462 * 3463 * If there is no free memory -ENOMEM is returned, otherwise zero 3464 * is returned and the old skb data released. 3465 */ 3466 static inline int skb_linearize(struct sk_buff *skb) 3467 { 3468 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3469 } 3470 3471 /** 3472 * skb_has_shared_frag - can any frag be overwritten 3473 * @skb: buffer to test 3474 * 3475 * Return true if the skb has at least one frag that might be modified 3476 * by an external entity (as in vmsplice()/sendfile()) 3477 */ 3478 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3479 { 3480 return skb_is_nonlinear(skb) && 3481 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3482 } 3483 3484 /** 3485 * skb_linearize_cow - make sure skb is linear and writable 3486 * @skb: buffer to process 3487 * 3488 * If there is no free memory -ENOMEM is returned, otherwise zero 3489 * is returned and the old skb data released. 3490 */ 3491 static inline int skb_linearize_cow(struct sk_buff *skb) 3492 { 3493 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3494 __skb_linearize(skb) : 0; 3495 } 3496 3497 static __always_inline void 3498 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3499 unsigned int off) 3500 { 3501 if (skb->ip_summed == CHECKSUM_COMPLETE) 3502 skb->csum = csum_block_sub(skb->csum, 3503 csum_partial(start, len, 0), off); 3504 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3505 skb_checksum_start_offset(skb) < 0) 3506 skb->ip_summed = CHECKSUM_NONE; 3507 } 3508 3509 /** 3510 * skb_postpull_rcsum - update checksum for received skb after pull 3511 * @skb: buffer to update 3512 * @start: start of data before pull 3513 * @len: length of data pulled 3514 * 3515 * After doing a pull on a received packet, you need to call this to 3516 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3517 * CHECKSUM_NONE so that it can be recomputed from scratch. 3518 */ 3519 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3520 const void *start, unsigned int len) 3521 { 3522 if (skb->ip_summed == CHECKSUM_COMPLETE) 3523 skb->csum = wsum_negate(csum_partial(start, len, 3524 wsum_negate(skb->csum))); 3525 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3526 skb_checksum_start_offset(skb) < 0) 3527 skb->ip_summed = CHECKSUM_NONE; 3528 } 3529 3530 static __always_inline void 3531 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3532 unsigned int off) 3533 { 3534 if (skb->ip_summed == CHECKSUM_COMPLETE) 3535 skb->csum = csum_block_add(skb->csum, 3536 csum_partial(start, len, 0), off); 3537 } 3538 3539 /** 3540 * skb_postpush_rcsum - update checksum for received skb after push 3541 * @skb: buffer to update 3542 * @start: start of data after push 3543 * @len: length of data pushed 3544 * 3545 * After doing a push on a received packet, you need to call this to 3546 * update the CHECKSUM_COMPLETE checksum. 3547 */ 3548 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3549 const void *start, unsigned int len) 3550 { 3551 __skb_postpush_rcsum(skb, start, len, 0); 3552 } 3553 3554 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3555 3556 /** 3557 * skb_push_rcsum - push skb and update receive checksum 3558 * @skb: buffer to update 3559 * @len: length of data pulled 3560 * 3561 * This function performs an skb_push on the packet and updates 3562 * the CHECKSUM_COMPLETE checksum. It should be used on 3563 * receive path processing instead of skb_push unless you know 3564 * that the checksum difference is zero (e.g., a valid IP header) 3565 * or you are setting ip_summed to CHECKSUM_NONE. 3566 */ 3567 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3568 { 3569 skb_push(skb, len); 3570 skb_postpush_rcsum(skb, skb->data, len); 3571 return skb->data; 3572 } 3573 3574 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3575 /** 3576 * pskb_trim_rcsum - trim received skb and update checksum 3577 * @skb: buffer to trim 3578 * @len: new length 3579 * 3580 * This is exactly the same as pskb_trim except that it ensures the 3581 * checksum of received packets are still valid after the operation. 3582 * It can change skb pointers. 3583 */ 3584 3585 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3586 { 3587 if (likely(len >= skb->len)) 3588 return 0; 3589 return pskb_trim_rcsum_slow(skb, len); 3590 } 3591 3592 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3593 { 3594 if (skb->ip_summed == CHECKSUM_COMPLETE) 3595 skb->ip_summed = CHECKSUM_NONE; 3596 __skb_trim(skb, len); 3597 return 0; 3598 } 3599 3600 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3601 { 3602 if (skb->ip_summed == CHECKSUM_COMPLETE) 3603 skb->ip_summed = CHECKSUM_NONE; 3604 return __skb_grow(skb, len); 3605 } 3606 3607 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3608 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3609 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3610 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3611 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3612 3613 #define skb_queue_walk(queue, skb) \ 3614 for (skb = (queue)->next; \ 3615 skb != (struct sk_buff *)(queue); \ 3616 skb = skb->next) 3617 3618 #define skb_queue_walk_safe(queue, skb, tmp) \ 3619 for (skb = (queue)->next, tmp = skb->next; \ 3620 skb != (struct sk_buff *)(queue); \ 3621 skb = tmp, tmp = skb->next) 3622 3623 #define skb_queue_walk_from(queue, skb) \ 3624 for (; skb != (struct sk_buff *)(queue); \ 3625 skb = skb->next) 3626 3627 #define skb_rbtree_walk(skb, root) \ 3628 for (skb = skb_rb_first(root); skb != NULL; \ 3629 skb = skb_rb_next(skb)) 3630 3631 #define skb_rbtree_walk_from(skb) \ 3632 for (; skb != NULL; \ 3633 skb = skb_rb_next(skb)) 3634 3635 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3636 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3637 skb = tmp) 3638 3639 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3640 for (tmp = skb->next; \ 3641 skb != (struct sk_buff *)(queue); \ 3642 skb = tmp, tmp = skb->next) 3643 3644 #define skb_queue_reverse_walk(queue, skb) \ 3645 for (skb = (queue)->prev; \ 3646 skb != (struct sk_buff *)(queue); \ 3647 skb = skb->prev) 3648 3649 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3650 for (skb = (queue)->prev, tmp = skb->prev; \ 3651 skb != (struct sk_buff *)(queue); \ 3652 skb = tmp, tmp = skb->prev) 3653 3654 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3655 for (tmp = skb->prev; \ 3656 skb != (struct sk_buff *)(queue); \ 3657 skb = tmp, tmp = skb->prev) 3658 3659 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3660 { 3661 return skb_shinfo(skb)->frag_list != NULL; 3662 } 3663 3664 static inline void skb_frag_list_init(struct sk_buff *skb) 3665 { 3666 skb_shinfo(skb)->frag_list = NULL; 3667 } 3668 3669 #define skb_walk_frags(skb, iter) \ 3670 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3671 3672 3673 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3674 int *err, long *timeo_p, 3675 const struct sk_buff *skb); 3676 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3677 struct sk_buff_head *queue, 3678 unsigned int flags, 3679 int *off, int *err, 3680 struct sk_buff **last); 3681 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3682 struct sk_buff_head *queue, 3683 unsigned int flags, int *off, int *err, 3684 struct sk_buff **last); 3685 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3686 struct sk_buff_head *sk_queue, 3687 unsigned int flags, int *off, int *err); 3688 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 3689 int *err); 3690 __poll_t datagram_poll(struct file *file, struct socket *sock, 3691 struct poll_table_struct *wait); 3692 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3693 struct iov_iter *to, int size); 3694 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3695 struct msghdr *msg, int size) 3696 { 3697 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3698 } 3699 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3700 struct msghdr *msg); 3701 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3702 struct iov_iter *to, int len, 3703 struct ahash_request *hash); 3704 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3705 struct iov_iter *from, int len); 3706 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3707 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3708 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3709 static inline void skb_free_datagram_locked(struct sock *sk, 3710 struct sk_buff *skb) 3711 { 3712 __skb_free_datagram_locked(sk, skb, 0); 3713 } 3714 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3715 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3716 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3717 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3718 int len); 3719 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3720 struct pipe_inode_info *pipe, unsigned int len, 3721 unsigned int flags); 3722 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3723 int len); 3724 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len); 3725 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3726 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3727 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3728 int len, int hlen); 3729 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3730 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3731 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3732 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3733 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3734 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3735 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3736 unsigned int offset); 3737 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 3738 int skb_ensure_writable(struct sk_buff *skb, int write_len); 3739 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 3740 int skb_vlan_pop(struct sk_buff *skb); 3741 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 3742 int skb_eth_pop(struct sk_buff *skb); 3743 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 3744 const unsigned char *src); 3745 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 3746 int mac_len, bool ethernet); 3747 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 3748 bool ethernet); 3749 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 3750 int skb_mpls_dec_ttl(struct sk_buff *skb); 3751 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 3752 gfp_t gfp); 3753 3754 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 3755 { 3756 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 3757 } 3758 3759 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 3760 { 3761 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 3762 } 3763 3764 struct skb_checksum_ops { 3765 __wsum (*update)(const void *mem, int len, __wsum wsum); 3766 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 3767 }; 3768 3769 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 3770 3771 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 3772 __wsum csum, const struct skb_checksum_ops *ops); 3773 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 3774 __wsum csum); 3775 3776 static inline void * __must_check 3777 __skb_header_pointer(const struct sk_buff *skb, int offset, int len, 3778 const void *data, int hlen, void *buffer) 3779 { 3780 if (likely(hlen - offset >= len)) 3781 return (void *)data + offset; 3782 3783 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) 3784 return NULL; 3785 3786 return buffer; 3787 } 3788 3789 static inline void * __must_check 3790 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 3791 { 3792 return __skb_header_pointer(skb, offset, len, skb->data, 3793 skb_headlen(skb), buffer); 3794 } 3795 3796 /** 3797 * skb_needs_linearize - check if we need to linearize a given skb 3798 * depending on the given device features. 3799 * @skb: socket buffer to check 3800 * @features: net device features 3801 * 3802 * Returns true if either: 3803 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 3804 * 2. skb is fragmented and the device does not support SG. 3805 */ 3806 static inline bool skb_needs_linearize(struct sk_buff *skb, 3807 netdev_features_t features) 3808 { 3809 return skb_is_nonlinear(skb) && 3810 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 3811 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 3812 } 3813 3814 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 3815 void *to, 3816 const unsigned int len) 3817 { 3818 memcpy(to, skb->data, len); 3819 } 3820 3821 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 3822 const int offset, void *to, 3823 const unsigned int len) 3824 { 3825 memcpy(to, skb->data + offset, len); 3826 } 3827 3828 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 3829 const void *from, 3830 const unsigned int len) 3831 { 3832 memcpy(skb->data, from, len); 3833 } 3834 3835 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 3836 const int offset, 3837 const void *from, 3838 const unsigned int len) 3839 { 3840 memcpy(skb->data + offset, from, len); 3841 } 3842 3843 void skb_init(void); 3844 3845 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 3846 { 3847 return skb->tstamp; 3848 } 3849 3850 /** 3851 * skb_get_timestamp - get timestamp from a skb 3852 * @skb: skb to get stamp from 3853 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 3854 * 3855 * Timestamps are stored in the skb as offsets to a base timestamp. 3856 * This function converts the offset back to a struct timeval and stores 3857 * it in stamp. 3858 */ 3859 static inline void skb_get_timestamp(const struct sk_buff *skb, 3860 struct __kernel_old_timeval *stamp) 3861 { 3862 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 3863 } 3864 3865 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 3866 struct __kernel_sock_timeval *stamp) 3867 { 3868 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3869 3870 stamp->tv_sec = ts.tv_sec; 3871 stamp->tv_usec = ts.tv_nsec / 1000; 3872 } 3873 3874 static inline void skb_get_timestampns(const struct sk_buff *skb, 3875 struct __kernel_old_timespec *stamp) 3876 { 3877 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3878 3879 stamp->tv_sec = ts.tv_sec; 3880 stamp->tv_nsec = ts.tv_nsec; 3881 } 3882 3883 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 3884 struct __kernel_timespec *stamp) 3885 { 3886 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 3887 3888 stamp->tv_sec = ts.tv_sec; 3889 stamp->tv_nsec = ts.tv_nsec; 3890 } 3891 3892 static inline void __net_timestamp(struct sk_buff *skb) 3893 { 3894 skb->tstamp = ktime_get_real(); 3895 } 3896 3897 static inline ktime_t net_timedelta(ktime_t t) 3898 { 3899 return ktime_sub(ktime_get_real(), t); 3900 } 3901 3902 static inline ktime_t net_invalid_timestamp(void) 3903 { 3904 return 0; 3905 } 3906 3907 static inline u8 skb_metadata_len(const struct sk_buff *skb) 3908 { 3909 return skb_shinfo(skb)->meta_len; 3910 } 3911 3912 static inline void *skb_metadata_end(const struct sk_buff *skb) 3913 { 3914 return skb_mac_header(skb); 3915 } 3916 3917 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 3918 const struct sk_buff *skb_b, 3919 u8 meta_len) 3920 { 3921 const void *a = skb_metadata_end(skb_a); 3922 const void *b = skb_metadata_end(skb_b); 3923 /* Using more efficient varaiant than plain call to memcmp(). */ 3924 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 3925 u64 diffs = 0; 3926 3927 switch (meta_len) { 3928 #define __it(x, op) (x -= sizeof(u##op)) 3929 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 3930 case 32: diffs |= __it_diff(a, b, 64); 3931 fallthrough; 3932 case 24: diffs |= __it_diff(a, b, 64); 3933 fallthrough; 3934 case 16: diffs |= __it_diff(a, b, 64); 3935 fallthrough; 3936 case 8: diffs |= __it_diff(a, b, 64); 3937 break; 3938 case 28: diffs |= __it_diff(a, b, 64); 3939 fallthrough; 3940 case 20: diffs |= __it_diff(a, b, 64); 3941 fallthrough; 3942 case 12: diffs |= __it_diff(a, b, 64); 3943 fallthrough; 3944 case 4: diffs |= __it_diff(a, b, 32); 3945 break; 3946 } 3947 return diffs; 3948 #else 3949 return memcmp(a - meta_len, b - meta_len, meta_len); 3950 #endif 3951 } 3952 3953 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 3954 const struct sk_buff *skb_b) 3955 { 3956 u8 len_a = skb_metadata_len(skb_a); 3957 u8 len_b = skb_metadata_len(skb_b); 3958 3959 if (!(len_a | len_b)) 3960 return false; 3961 3962 return len_a != len_b ? 3963 true : __skb_metadata_differs(skb_a, skb_b, len_a); 3964 } 3965 3966 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 3967 { 3968 skb_shinfo(skb)->meta_len = meta_len; 3969 } 3970 3971 static inline void skb_metadata_clear(struct sk_buff *skb) 3972 { 3973 skb_metadata_set(skb, 0); 3974 } 3975 3976 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 3977 3978 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 3979 3980 void skb_clone_tx_timestamp(struct sk_buff *skb); 3981 bool skb_defer_rx_timestamp(struct sk_buff *skb); 3982 3983 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 3984 3985 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 3986 { 3987 } 3988 3989 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 3990 { 3991 return false; 3992 } 3993 3994 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 3995 3996 /** 3997 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 3998 * 3999 * PHY drivers may accept clones of transmitted packets for 4000 * timestamping via their phy_driver.txtstamp method. These drivers 4001 * must call this function to return the skb back to the stack with a 4002 * timestamp. 4003 * 4004 * @skb: clone of the original outgoing packet 4005 * @hwtstamps: hardware time stamps 4006 * 4007 */ 4008 void skb_complete_tx_timestamp(struct sk_buff *skb, 4009 struct skb_shared_hwtstamps *hwtstamps); 4010 4011 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 4012 struct skb_shared_hwtstamps *hwtstamps, 4013 struct sock *sk, int tstype); 4014 4015 /** 4016 * skb_tstamp_tx - queue clone of skb with send time stamps 4017 * @orig_skb: the original outgoing packet 4018 * @hwtstamps: hardware time stamps, may be NULL if not available 4019 * 4020 * If the skb has a socket associated, then this function clones the 4021 * skb (thus sharing the actual data and optional structures), stores 4022 * the optional hardware time stamping information (if non NULL) or 4023 * generates a software time stamp (otherwise), then queues the clone 4024 * to the error queue of the socket. Errors are silently ignored. 4025 */ 4026 void skb_tstamp_tx(struct sk_buff *orig_skb, 4027 struct skb_shared_hwtstamps *hwtstamps); 4028 4029 /** 4030 * skb_tx_timestamp() - Driver hook for transmit timestamping 4031 * 4032 * Ethernet MAC Drivers should call this function in their hard_xmit() 4033 * function immediately before giving the sk_buff to the MAC hardware. 4034 * 4035 * Specifically, one should make absolutely sure that this function is 4036 * called before TX completion of this packet can trigger. Otherwise 4037 * the packet could potentially already be freed. 4038 * 4039 * @skb: A socket buffer. 4040 */ 4041 static inline void skb_tx_timestamp(struct sk_buff *skb) 4042 { 4043 skb_clone_tx_timestamp(skb); 4044 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 4045 skb_tstamp_tx(skb, NULL); 4046 } 4047 4048 /** 4049 * skb_complete_wifi_ack - deliver skb with wifi status 4050 * 4051 * @skb: the original outgoing packet 4052 * @acked: ack status 4053 * 4054 */ 4055 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 4056 4057 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 4058 __sum16 __skb_checksum_complete(struct sk_buff *skb); 4059 4060 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 4061 { 4062 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 4063 skb->csum_valid || 4064 (skb->ip_summed == CHECKSUM_PARTIAL && 4065 skb_checksum_start_offset(skb) >= 0)); 4066 } 4067 4068 /** 4069 * skb_checksum_complete - Calculate checksum of an entire packet 4070 * @skb: packet to process 4071 * 4072 * This function calculates the checksum over the entire packet plus 4073 * the value of skb->csum. The latter can be used to supply the 4074 * checksum of a pseudo header as used by TCP/UDP. It returns the 4075 * checksum. 4076 * 4077 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 4078 * this function can be used to verify that checksum on received 4079 * packets. In that case the function should return zero if the 4080 * checksum is correct. In particular, this function will return zero 4081 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 4082 * hardware has already verified the correctness of the checksum. 4083 */ 4084 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 4085 { 4086 return skb_csum_unnecessary(skb) ? 4087 0 : __skb_checksum_complete(skb); 4088 } 4089 4090 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 4091 { 4092 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4093 if (skb->csum_level == 0) 4094 skb->ip_summed = CHECKSUM_NONE; 4095 else 4096 skb->csum_level--; 4097 } 4098 } 4099 4100 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4101 { 4102 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4103 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4104 skb->csum_level++; 4105 } else if (skb->ip_summed == CHECKSUM_NONE) { 4106 skb->ip_summed = CHECKSUM_UNNECESSARY; 4107 skb->csum_level = 0; 4108 } 4109 } 4110 4111 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4112 { 4113 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4114 skb->ip_summed = CHECKSUM_NONE; 4115 skb->csum_level = 0; 4116 } 4117 } 4118 4119 /* Check if we need to perform checksum complete validation. 4120 * 4121 * Returns true if checksum complete is needed, false otherwise 4122 * (either checksum is unnecessary or zero checksum is allowed). 4123 */ 4124 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4125 bool zero_okay, 4126 __sum16 check) 4127 { 4128 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4129 skb->csum_valid = 1; 4130 __skb_decr_checksum_unnecessary(skb); 4131 return false; 4132 } 4133 4134 return true; 4135 } 4136 4137 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4138 * in checksum_init. 4139 */ 4140 #define CHECKSUM_BREAK 76 4141 4142 /* Unset checksum-complete 4143 * 4144 * Unset checksum complete can be done when packet is being modified 4145 * (uncompressed for instance) and checksum-complete value is 4146 * invalidated. 4147 */ 4148 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4149 { 4150 if (skb->ip_summed == CHECKSUM_COMPLETE) 4151 skb->ip_summed = CHECKSUM_NONE; 4152 } 4153 4154 /* Validate (init) checksum based on checksum complete. 4155 * 4156 * Return values: 4157 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4158 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4159 * checksum is stored in skb->csum for use in __skb_checksum_complete 4160 * non-zero: value of invalid checksum 4161 * 4162 */ 4163 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4164 bool complete, 4165 __wsum psum) 4166 { 4167 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4168 if (!csum_fold(csum_add(psum, skb->csum))) { 4169 skb->csum_valid = 1; 4170 return 0; 4171 } 4172 } 4173 4174 skb->csum = psum; 4175 4176 if (complete || skb->len <= CHECKSUM_BREAK) { 4177 __sum16 csum; 4178 4179 csum = __skb_checksum_complete(skb); 4180 skb->csum_valid = !csum; 4181 return csum; 4182 } 4183 4184 return 0; 4185 } 4186 4187 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4188 { 4189 return 0; 4190 } 4191 4192 /* Perform checksum validate (init). Note that this is a macro since we only 4193 * want to calculate the pseudo header which is an input function if necessary. 4194 * First we try to validate without any computation (checksum unnecessary) and 4195 * then calculate based on checksum complete calling the function to compute 4196 * pseudo header. 4197 * 4198 * Return values: 4199 * 0: checksum is validated or try to in skb_checksum_complete 4200 * non-zero: value of invalid checksum 4201 */ 4202 #define __skb_checksum_validate(skb, proto, complete, \ 4203 zero_okay, check, compute_pseudo) \ 4204 ({ \ 4205 __sum16 __ret = 0; \ 4206 skb->csum_valid = 0; \ 4207 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4208 __ret = __skb_checksum_validate_complete(skb, \ 4209 complete, compute_pseudo(skb, proto)); \ 4210 __ret; \ 4211 }) 4212 4213 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4214 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4215 4216 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4217 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4218 4219 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4220 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4221 4222 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4223 compute_pseudo) \ 4224 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4225 4226 #define skb_checksum_simple_validate(skb) \ 4227 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4228 4229 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4230 { 4231 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4232 } 4233 4234 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4235 { 4236 skb->csum = ~pseudo; 4237 skb->ip_summed = CHECKSUM_COMPLETE; 4238 } 4239 4240 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4241 do { \ 4242 if (__skb_checksum_convert_check(skb)) \ 4243 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4244 } while (0) 4245 4246 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4247 u16 start, u16 offset) 4248 { 4249 skb->ip_summed = CHECKSUM_PARTIAL; 4250 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4251 skb->csum_offset = offset - start; 4252 } 4253 4254 /* Update skbuf and packet to reflect the remote checksum offload operation. 4255 * When called, ptr indicates the starting point for skb->csum when 4256 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4257 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4258 */ 4259 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4260 int start, int offset, bool nopartial) 4261 { 4262 __wsum delta; 4263 4264 if (!nopartial) { 4265 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4266 return; 4267 } 4268 4269 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4270 __skb_checksum_complete(skb); 4271 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4272 } 4273 4274 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4275 4276 /* Adjust skb->csum since we changed the packet */ 4277 skb->csum = csum_add(skb->csum, delta); 4278 } 4279 4280 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4281 { 4282 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4283 return (void *)(skb->_nfct & NFCT_PTRMASK); 4284 #else 4285 return NULL; 4286 #endif 4287 } 4288 4289 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4290 { 4291 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4292 return skb->_nfct; 4293 #else 4294 return 0UL; 4295 #endif 4296 } 4297 4298 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4299 { 4300 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4301 skb->slow_gro |= !!nfct; 4302 skb->_nfct = nfct; 4303 #endif 4304 } 4305 4306 #ifdef CONFIG_SKB_EXTENSIONS 4307 enum skb_ext_id { 4308 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4309 SKB_EXT_BRIDGE_NF, 4310 #endif 4311 #ifdef CONFIG_XFRM 4312 SKB_EXT_SEC_PATH, 4313 #endif 4314 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4315 TC_SKB_EXT, 4316 #endif 4317 #if IS_ENABLED(CONFIG_MPTCP) 4318 SKB_EXT_MPTCP, 4319 #endif 4320 #if IS_ENABLED(CONFIG_MCTP_FLOWS) 4321 SKB_EXT_MCTP, 4322 #endif 4323 SKB_EXT_NUM, /* must be last */ 4324 }; 4325 4326 /** 4327 * struct skb_ext - sk_buff extensions 4328 * @refcnt: 1 on allocation, deallocated on 0 4329 * @offset: offset to add to @data to obtain extension address 4330 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4331 * @data: start of extension data, variable sized 4332 * 4333 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4334 * to use 'u8' types while allowing up to 2kb worth of extension data. 4335 */ 4336 struct skb_ext { 4337 refcount_t refcnt; 4338 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4339 u8 chunks; /* same */ 4340 char data[] __aligned(8); 4341 }; 4342 4343 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4344 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4345 struct skb_ext *ext); 4346 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4347 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4348 void __skb_ext_put(struct skb_ext *ext); 4349 4350 static inline void skb_ext_put(struct sk_buff *skb) 4351 { 4352 if (skb->active_extensions) 4353 __skb_ext_put(skb->extensions); 4354 } 4355 4356 static inline void __skb_ext_copy(struct sk_buff *dst, 4357 const struct sk_buff *src) 4358 { 4359 dst->active_extensions = src->active_extensions; 4360 4361 if (src->active_extensions) { 4362 struct skb_ext *ext = src->extensions; 4363 4364 refcount_inc(&ext->refcnt); 4365 dst->extensions = ext; 4366 } 4367 } 4368 4369 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4370 { 4371 skb_ext_put(dst); 4372 __skb_ext_copy(dst, src); 4373 } 4374 4375 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4376 { 4377 return !!ext->offset[i]; 4378 } 4379 4380 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4381 { 4382 return skb->active_extensions & (1 << id); 4383 } 4384 4385 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4386 { 4387 if (skb_ext_exist(skb, id)) 4388 __skb_ext_del(skb, id); 4389 } 4390 4391 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4392 { 4393 if (skb_ext_exist(skb, id)) { 4394 struct skb_ext *ext = skb->extensions; 4395 4396 return (void *)ext + (ext->offset[id] << 3); 4397 } 4398 4399 return NULL; 4400 } 4401 4402 static inline void skb_ext_reset(struct sk_buff *skb) 4403 { 4404 if (unlikely(skb->active_extensions)) { 4405 __skb_ext_put(skb->extensions); 4406 skb->active_extensions = 0; 4407 } 4408 } 4409 4410 static inline bool skb_has_extensions(struct sk_buff *skb) 4411 { 4412 return unlikely(skb->active_extensions); 4413 } 4414 #else 4415 static inline void skb_ext_put(struct sk_buff *skb) {} 4416 static inline void skb_ext_reset(struct sk_buff *skb) {} 4417 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4418 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4419 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4420 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4421 #endif /* CONFIG_SKB_EXTENSIONS */ 4422 4423 static inline void nf_reset_ct(struct sk_buff *skb) 4424 { 4425 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4426 nf_conntrack_put(skb_nfct(skb)); 4427 skb->_nfct = 0; 4428 #endif 4429 } 4430 4431 static inline void nf_reset_trace(struct sk_buff *skb) 4432 { 4433 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4434 skb->nf_trace = 0; 4435 #endif 4436 } 4437 4438 static inline void ipvs_reset(struct sk_buff *skb) 4439 { 4440 #if IS_ENABLED(CONFIG_IP_VS) 4441 skb->ipvs_property = 0; 4442 #endif 4443 } 4444 4445 /* Note: This doesn't put any conntrack info in dst. */ 4446 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4447 bool copy) 4448 { 4449 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4450 dst->_nfct = src->_nfct; 4451 nf_conntrack_get(skb_nfct(src)); 4452 #endif 4453 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 4454 if (copy) 4455 dst->nf_trace = src->nf_trace; 4456 #endif 4457 } 4458 4459 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4460 { 4461 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4462 nf_conntrack_put(skb_nfct(dst)); 4463 #endif 4464 dst->slow_gro = src->slow_gro; 4465 __nf_copy(dst, src, true); 4466 } 4467 4468 #ifdef CONFIG_NETWORK_SECMARK 4469 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4470 { 4471 to->secmark = from->secmark; 4472 } 4473 4474 static inline void skb_init_secmark(struct sk_buff *skb) 4475 { 4476 skb->secmark = 0; 4477 } 4478 #else 4479 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4480 { } 4481 4482 static inline void skb_init_secmark(struct sk_buff *skb) 4483 { } 4484 #endif 4485 4486 static inline int secpath_exists(const struct sk_buff *skb) 4487 { 4488 #ifdef CONFIG_XFRM 4489 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4490 #else 4491 return 0; 4492 #endif 4493 } 4494 4495 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4496 { 4497 return !skb->destructor && 4498 !secpath_exists(skb) && 4499 !skb_nfct(skb) && 4500 !skb->_skb_refdst && 4501 !skb_has_frag_list(skb); 4502 } 4503 4504 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4505 { 4506 skb->queue_mapping = queue_mapping; 4507 } 4508 4509 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4510 { 4511 return skb->queue_mapping; 4512 } 4513 4514 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4515 { 4516 to->queue_mapping = from->queue_mapping; 4517 } 4518 4519 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4520 { 4521 skb->queue_mapping = rx_queue + 1; 4522 } 4523 4524 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4525 { 4526 return skb->queue_mapping - 1; 4527 } 4528 4529 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4530 { 4531 return skb->queue_mapping != 0; 4532 } 4533 4534 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4535 { 4536 skb->dst_pending_confirm = val; 4537 } 4538 4539 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4540 { 4541 return skb->dst_pending_confirm != 0; 4542 } 4543 4544 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4545 { 4546 #ifdef CONFIG_XFRM 4547 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4548 #else 4549 return NULL; 4550 #endif 4551 } 4552 4553 /* Keeps track of mac header offset relative to skb->head. 4554 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4555 * For non-tunnel skb it points to skb_mac_header() and for 4556 * tunnel skb it points to outer mac header. 4557 * Keeps track of level of encapsulation of network headers. 4558 */ 4559 struct skb_gso_cb { 4560 union { 4561 int mac_offset; 4562 int data_offset; 4563 }; 4564 int encap_level; 4565 __wsum csum; 4566 __u16 csum_start; 4567 }; 4568 #define SKB_GSO_CB_OFFSET 32 4569 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4570 4571 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4572 { 4573 return (skb_mac_header(inner_skb) - inner_skb->head) - 4574 SKB_GSO_CB(inner_skb)->mac_offset; 4575 } 4576 4577 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4578 { 4579 int new_headroom, headroom; 4580 int ret; 4581 4582 headroom = skb_headroom(skb); 4583 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4584 if (ret) 4585 return ret; 4586 4587 new_headroom = skb_headroom(skb); 4588 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4589 return 0; 4590 } 4591 4592 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4593 { 4594 /* Do not update partial checksums if remote checksum is enabled. */ 4595 if (skb->remcsum_offload) 4596 return; 4597 4598 SKB_GSO_CB(skb)->csum = res; 4599 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4600 } 4601 4602 /* Compute the checksum for a gso segment. First compute the checksum value 4603 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4604 * then add in skb->csum (checksum from csum_start to end of packet). 4605 * skb->csum and csum_start are then updated to reflect the checksum of the 4606 * resultant packet starting from the transport header-- the resultant checksum 4607 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4608 * header. 4609 */ 4610 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4611 { 4612 unsigned char *csum_start = skb_transport_header(skb); 4613 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4614 __wsum partial = SKB_GSO_CB(skb)->csum; 4615 4616 SKB_GSO_CB(skb)->csum = res; 4617 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4618 4619 return csum_fold(csum_partial(csum_start, plen, partial)); 4620 } 4621 4622 static inline bool skb_is_gso(const struct sk_buff *skb) 4623 { 4624 return skb_shinfo(skb)->gso_size; 4625 } 4626 4627 /* Note: Should be called only if skb_is_gso(skb) is true */ 4628 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4629 { 4630 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4631 } 4632 4633 /* Note: Should be called only if skb_is_gso(skb) is true */ 4634 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4635 { 4636 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4637 } 4638 4639 /* Note: Should be called only if skb_is_gso(skb) is true */ 4640 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4641 { 4642 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4643 } 4644 4645 static inline void skb_gso_reset(struct sk_buff *skb) 4646 { 4647 skb_shinfo(skb)->gso_size = 0; 4648 skb_shinfo(skb)->gso_segs = 0; 4649 skb_shinfo(skb)->gso_type = 0; 4650 } 4651 4652 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4653 u16 increment) 4654 { 4655 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4656 return; 4657 shinfo->gso_size += increment; 4658 } 4659 4660 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4661 u16 decrement) 4662 { 4663 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4664 return; 4665 shinfo->gso_size -= decrement; 4666 } 4667 4668 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4669 4670 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4671 { 4672 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4673 * wanted then gso_type will be set. */ 4674 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4675 4676 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4677 unlikely(shinfo->gso_type == 0)) { 4678 __skb_warn_lro_forwarding(skb); 4679 return true; 4680 } 4681 return false; 4682 } 4683 4684 static inline void skb_forward_csum(struct sk_buff *skb) 4685 { 4686 /* Unfortunately we don't support this one. Any brave souls? */ 4687 if (skb->ip_summed == CHECKSUM_COMPLETE) 4688 skb->ip_summed = CHECKSUM_NONE; 4689 } 4690 4691 /** 4692 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 4693 * @skb: skb to check 4694 * 4695 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 4696 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 4697 * use this helper, to document places where we make this assertion. 4698 */ 4699 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 4700 { 4701 #ifdef DEBUG 4702 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 4703 #endif 4704 } 4705 4706 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 4707 4708 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 4709 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 4710 unsigned int transport_len, 4711 __sum16(*skb_chkf)(struct sk_buff *skb)); 4712 4713 /** 4714 * skb_head_is_locked - Determine if the skb->head is locked down 4715 * @skb: skb to check 4716 * 4717 * The head on skbs build around a head frag can be removed if they are 4718 * not cloned. This function returns true if the skb head is locked down 4719 * due to either being allocated via kmalloc, or by being a clone with 4720 * multiple references to the head. 4721 */ 4722 static inline bool skb_head_is_locked(const struct sk_buff *skb) 4723 { 4724 return !skb->head_frag || skb_cloned(skb); 4725 } 4726 4727 /* Local Checksum Offload. 4728 * Compute outer checksum based on the assumption that the 4729 * inner checksum will be offloaded later. 4730 * See Documentation/networking/checksum-offloads.rst for 4731 * explanation of how this works. 4732 * Fill in outer checksum adjustment (e.g. with sum of outer 4733 * pseudo-header) before calling. 4734 * Also ensure that inner checksum is in linear data area. 4735 */ 4736 static inline __wsum lco_csum(struct sk_buff *skb) 4737 { 4738 unsigned char *csum_start = skb_checksum_start(skb); 4739 unsigned char *l4_hdr = skb_transport_header(skb); 4740 __wsum partial; 4741 4742 /* Start with complement of inner checksum adjustment */ 4743 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 4744 skb->csum_offset)); 4745 4746 /* Add in checksum of our headers (incl. outer checksum 4747 * adjustment filled in by caller) and return result. 4748 */ 4749 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 4750 } 4751 4752 static inline bool skb_is_redirected(const struct sk_buff *skb) 4753 { 4754 return skb->redirected; 4755 } 4756 4757 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 4758 { 4759 skb->redirected = 1; 4760 #ifdef CONFIG_NET_REDIRECT 4761 skb->from_ingress = from_ingress; 4762 if (skb->from_ingress) 4763 skb->tstamp = 0; 4764 #endif 4765 } 4766 4767 static inline void skb_reset_redirect(struct sk_buff *skb) 4768 { 4769 skb->redirected = 0; 4770 } 4771 4772 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 4773 { 4774 return skb->csum_not_inet; 4775 } 4776 4777 static inline void skb_set_kcov_handle(struct sk_buff *skb, 4778 const u64 kcov_handle) 4779 { 4780 #ifdef CONFIG_KCOV 4781 skb->kcov_handle = kcov_handle; 4782 #endif 4783 } 4784 4785 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 4786 { 4787 #ifdef CONFIG_KCOV 4788 return skb->kcov_handle; 4789 #else 4790 return 0; 4791 #endif 4792 } 4793 4794 #ifdef CONFIG_PAGE_POOL 4795 static inline void skb_mark_for_recycle(struct sk_buff *skb) 4796 { 4797 skb->pp_recycle = 1; 4798 } 4799 #endif 4800 4801 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data) 4802 { 4803 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle) 4804 return false; 4805 return page_pool_return_skb_page(virt_to_page(data)); 4806 } 4807 4808 #endif /* __KERNEL__ */ 4809 #endif /* _LINUX_SKBUFF_H */ 4810