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 <net/checksum.h> 27 #include <linux/rcupdate.h> 28 #include <linux/dma-mapping.h> 29 #include <linux/netdev_features.h> 30 #include <net/flow_dissector.h> 31 #include <linux/in6.h> 32 #include <linux/if_packet.h> 33 #include <linux/llist.h> 34 #include <net/flow.h> 35 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 36 #include <linux/netfilter/nf_conntrack_common.h> 37 #endif 38 #include <net/net_debug.h> 39 #include <net/dropreason-core.h> 40 #include <net/netmem.h> 41 42 /** 43 * DOC: skb checksums 44 * 45 * The interface for checksum offload between the stack and networking drivers 46 * is as follows... 47 * 48 * IP checksum related features 49 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 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 * .. flat-table:: Checksum related device features 57 * :widths: 1 10 58 * 59 * * - %NETIF_F_HW_CSUM 60 * - The driver (or its device) is able to compute one 61 * IP (one's complement) checksum for any combination 62 * of protocols or protocol layering. The checksum is 63 * computed and set in a packet per the CHECKSUM_PARTIAL 64 * interface (see below). 65 * 66 * * - %NETIF_F_IP_CSUM 67 * - Driver (device) is only able to checksum plain 68 * TCP or UDP packets over IPv4. These are specifically 69 * unencapsulated packets of the form IPv4|TCP or 70 * IPv4|UDP where the Protocol field in the IPv4 header 71 * is TCP or UDP. The IPv4 header may contain IP options. 72 * This feature cannot be set in features for a device 73 * with NETIF_F_HW_CSUM also set. This feature is being 74 * DEPRECATED (see below). 75 * 76 * * - %NETIF_F_IPV6_CSUM 77 * - Driver (device) is only able to checksum plain 78 * TCP or UDP packets over IPv6. These are specifically 79 * unencapsulated packets of the form IPv6|TCP or 80 * IPv6|UDP where the Next Header field in the IPv6 81 * header is either TCP or UDP. IPv6 extension headers 82 * are not supported with this feature. This feature 83 * cannot be set in features for a device with 84 * NETIF_F_HW_CSUM also set. This feature is being 85 * DEPRECATED (see below). 86 * 87 * * - %NETIF_F_RXCSUM 88 * - Driver (device) performs receive checksum offload. 89 * This flag is only used to disable the RX checksum 90 * feature for a device. The stack will accept receive 91 * checksum indication in packets received on a device 92 * regardless of whether NETIF_F_RXCSUM is set. 93 * 94 * Checksumming of received packets by device 95 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 96 * 97 * Indication of checksum verification is set in &sk_buff.ip_summed. 98 * Possible values are: 99 * 100 * - %CHECKSUM_NONE 101 * 102 * Device did not checksum this packet e.g. due to lack of capabilities. 103 * The packet contains full (though not verified) checksum in packet but 104 * not in skb->csum. Thus, skb->csum is undefined in this case. 105 * 106 * - %CHECKSUM_UNNECESSARY 107 * 108 * The hardware you're dealing with doesn't calculate the full checksum 109 * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums 110 * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY 111 * if their checksums are okay. &sk_buff.csum is still undefined in this case 112 * though. A driver or device must never modify the checksum field in the 113 * packet even if checksum is verified. 114 * 115 * %CHECKSUM_UNNECESSARY is applicable to following protocols: 116 * 117 * - TCP: IPv6 and IPv4. 118 * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 119 * zero UDP checksum for either IPv4 or IPv6, the networking stack 120 * may perform further validation in this case. 121 * - GRE: only if the checksum is present in the header. 122 * - SCTP: indicates the CRC in SCTP header has been validated. 123 * - FCOE: indicates the CRC in FC frame has been validated. 124 * 125 * &sk_buff.csum_level indicates the number of consecutive checksums found in 126 * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY. 127 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 128 * and a device is able to verify the checksums for UDP (possibly zero), 129 * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to 130 * two. If the device were only able to verify the UDP checksum and not 131 * GRE, either because it doesn't support GRE checksum or because GRE 132 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 133 * not considered in this case). 134 * 135 * - %CHECKSUM_COMPLETE 136 * 137 * This is the most generic way. The device supplied checksum of the _whole_ 138 * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the 139 * hardware doesn't need to parse L3/L4 headers to implement this. 140 * 141 * Notes: 142 * 143 * - Even if device supports only some protocols, but is able to produce 144 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 145 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. 146 * 147 * - %CHECKSUM_PARTIAL 148 * 149 * A checksum is set up to be offloaded to a device as described in the 150 * output description for CHECKSUM_PARTIAL. This may occur on a packet 151 * received directly from another Linux OS, e.g., a virtualized Linux kernel 152 * on the same host, or it may be set in the input path in GRO or remote 153 * checksum offload. For the purposes of checksum verification, the checksum 154 * referred to by skb->csum_start + skb->csum_offset and any preceding 155 * checksums in the packet are considered verified. Any checksums in the 156 * packet that are after the checksum being offloaded are not considered to 157 * be verified. 158 * 159 * Checksumming on transmit for non-GSO 160 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 161 * 162 * The stack requests checksum offload in the &sk_buff.ip_summed for a packet. 163 * Values are: 164 * 165 * - %CHECKSUM_PARTIAL 166 * 167 * The driver is required to checksum the packet as seen by hard_start_xmit() 168 * from &sk_buff.csum_start up to the end, and to record/write the checksum at 169 * offset &sk_buff.csum_start + &sk_buff.csum_offset. 170 * A driver may verify that the 171 * csum_start and csum_offset values are valid values given the length and 172 * offset of the packet, but it should not attempt to validate that the 173 * checksum refers to a legitimate transport layer checksum -- it is the 174 * purview of the stack to validate that csum_start and csum_offset are set 175 * correctly. 176 * 177 * When the stack requests checksum offload for a packet, the driver MUST 178 * ensure that the checksum is set correctly. A driver can either offload the 179 * checksum calculation to the device, or call skb_checksum_help (in the case 180 * that the device does not support offload for a particular checksum). 181 * 182 * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of 183 * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate 184 * checksum offload capability. 185 * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based 186 * on network device checksumming capabilities: if a packet does not match 187 * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of 188 * &sk_buff.csum_not_inet, see :ref:`crc`) 189 * is called to resolve the checksum. 190 * 191 * - %CHECKSUM_NONE 192 * 193 * The skb was already checksummed by the protocol, or a checksum is not 194 * required. 195 * 196 * - %CHECKSUM_UNNECESSARY 197 * 198 * This has the same meaning as CHECKSUM_NONE for checksum offload on 199 * output. 200 * 201 * - %CHECKSUM_COMPLETE 202 * 203 * Not used in checksum output. If a driver observes a packet with this value 204 * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set. 205 * 206 * .. _crc: 207 * 208 * Non-IP checksum (CRC) offloads 209 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 210 * 211 * .. flat-table:: 212 * :widths: 1 10 213 * 214 * * - %NETIF_F_SCTP_CRC 215 * - This feature indicates that a device is capable of 216 * offloading the SCTP CRC in a packet. To perform this offload the stack 217 * will set csum_start and csum_offset accordingly, set ip_summed to 218 * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication 219 * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c. 220 * A driver that supports both IP checksum offload and SCTP CRC32c offload 221 * must verify which offload is configured for a packet by testing the 222 * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to 223 * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. 224 * 225 * * - %NETIF_F_FCOE_CRC 226 * - This feature indicates that a device is capable of offloading the FCOE 227 * CRC in a packet. To perform this offload the stack will set ip_summed 228 * to %CHECKSUM_PARTIAL and set csum_start and csum_offset 229 * accordingly. Note that there is no indication in the skbuff that the 230 * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports 231 * both IP checksum offload and FCOE CRC offload must verify which offload 232 * is configured for a packet, presumably by inspecting packet headers. 233 * 234 * Checksumming on output with GSO 235 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 236 * 237 * In the case of a GSO packet (skb_is_gso() is true), checksum offload 238 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the 239 * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as 240 * part of the GSO operation is implied. If a checksum is being offloaded 241 * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and 242 * csum_offset are set to refer to the outermost checksum being offloaded 243 * (two offloaded checksums are possible with UDP encapsulation). 244 */ 245 246 /* Don't change this without changing skb_csum_unnecessary! */ 247 #define CHECKSUM_NONE 0 248 #define CHECKSUM_UNNECESSARY 1 249 #define CHECKSUM_COMPLETE 2 250 #define CHECKSUM_PARTIAL 3 251 252 /* Maximum value in skb->csum_level */ 253 #define SKB_MAX_CSUM_LEVEL 3 254 255 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 256 #define SKB_WITH_OVERHEAD(X) \ 257 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 258 259 /* For X bytes available in skb->head, what is the minimal 260 * allocation needed, knowing struct skb_shared_info needs 261 * to be aligned. 262 */ 263 #define SKB_HEAD_ALIGN(X) (SKB_DATA_ALIGN(X) + \ 264 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 265 266 #define SKB_MAX_ORDER(X, ORDER) \ 267 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 268 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 269 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 270 271 /* return minimum truesize of one skb containing X bytes of data */ 272 #define SKB_TRUESIZE(X) ((X) + \ 273 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 274 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 275 276 struct ahash_request; 277 struct net_device; 278 struct scatterlist; 279 struct pipe_inode_info; 280 struct iov_iter; 281 struct napi_struct; 282 struct bpf_prog; 283 union bpf_attr; 284 struct skb_ext; 285 struct ts_config; 286 287 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 288 struct nf_bridge_info { 289 enum { 290 BRNF_PROTO_UNCHANGED, 291 BRNF_PROTO_8021Q, 292 BRNF_PROTO_PPPOE 293 } orig_proto:8; 294 u8 pkt_otherhost:1; 295 u8 in_prerouting:1; 296 u8 bridged_dnat:1; 297 u8 sabotage_in_done:1; 298 __u16 frag_max_size; 299 int physinif; 300 301 /* always valid & non-NULL from FORWARD on, for physdev match */ 302 struct net_device *physoutdev; 303 union { 304 /* prerouting: detect dnat in orig/reply direction */ 305 __be32 ipv4_daddr; 306 struct in6_addr ipv6_daddr; 307 308 /* after prerouting + nat detected: store original source 309 * mac since neigh resolution overwrites it, only used while 310 * skb is out in neigh layer. 311 */ 312 char neigh_header[8]; 313 }; 314 }; 315 #endif 316 317 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 318 /* Chain in tc_skb_ext will be used to share the tc chain with 319 * ovs recirc_id. It will be set to the current chain by tc 320 * and read by ovs to recirc_id. 321 */ 322 struct tc_skb_ext { 323 union { 324 u64 act_miss_cookie; 325 __u32 chain; 326 }; 327 __u16 mru; 328 __u16 zone; 329 u8 post_ct:1; 330 u8 post_ct_snat:1; 331 u8 post_ct_dnat:1; 332 u8 act_miss:1; /* Set if act_miss_cookie is used */ 333 u8 l2_miss:1; /* Set by bridge upon FDB or MDB miss */ 334 }; 335 #endif 336 337 struct sk_buff_head { 338 /* These two members must be first to match sk_buff. */ 339 struct_group_tagged(sk_buff_list, list, 340 struct sk_buff *next; 341 struct sk_buff *prev; 342 ); 343 344 __u32 qlen; 345 spinlock_t lock; 346 }; 347 348 struct sk_buff; 349 350 #ifndef CONFIG_MAX_SKB_FRAGS 351 # define CONFIG_MAX_SKB_FRAGS 17 352 #endif 353 354 #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS 355 356 extern int sysctl_max_skb_frags; 357 358 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 359 * segment using its current segmentation instead. 360 */ 361 #define GSO_BY_FRAGS 0xFFFF 362 363 typedef struct skb_frag { 364 netmem_ref netmem; 365 unsigned int len; 366 unsigned int offset; 367 } skb_frag_t; 368 369 /** 370 * skb_frag_size() - Returns the size of a skb fragment 371 * @frag: skb fragment 372 */ 373 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 374 { 375 return frag->len; 376 } 377 378 /** 379 * skb_frag_size_set() - Sets the size of a skb fragment 380 * @frag: skb fragment 381 * @size: size of fragment 382 */ 383 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 384 { 385 frag->len = size; 386 } 387 388 /** 389 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 390 * @frag: skb fragment 391 * @delta: value to add 392 */ 393 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 394 { 395 frag->len += delta; 396 } 397 398 /** 399 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 400 * @frag: skb fragment 401 * @delta: value to subtract 402 */ 403 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 404 { 405 frag->len -= delta; 406 } 407 408 /** 409 * skb_frag_must_loop - Test if %p is a high memory page 410 * @p: fragment's page 411 */ 412 static inline bool skb_frag_must_loop(struct page *p) 413 { 414 #if defined(CONFIG_HIGHMEM) 415 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 416 return true; 417 #endif 418 return false; 419 } 420 421 /** 422 * skb_frag_foreach_page - loop over pages in a fragment 423 * 424 * @f: skb frag to operate on 425 * @f_off: offset from start of f->netmem 426 * @f_len: length from f_off to loop over 427 * @p: (temp var) current page 428 * @p_off: (temp var) offset from start of current page, 429 * non-zero only on first page. 430 * @p_len: (temp var) length in current page, 431 * < PAGE_SIZE only on first and last page. 432 * @copied: (temp var) length so far, excluding current p_len. 433 * 434 * A fragment can hold a compound page, in which case per-page 435 * operations, notably kmap_atomic, must be called for each 436 * regular page. 437 */ 438 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 439 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 440 p_off = (f_off) & (PAGE_SIZE - 1), \ 441 p_len = skb_frag_must_loop(p) ? \ 442 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 443 copied = 0; \ 444 copied < f_len; \ 445 copied += p_len, p++, p_off = 0, \ 446 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 447 448 /** 449 * struct skb_shared_hwtstamps - hardware time stamps 450 * @hwtstamp: hardware time stamp transformed into duration 451 * since arbitrary point in time 452 * @netdev_data: address/cookie of network device driver used as 453 * reference to actual hardware time stamp 454 * 455 * Software time stamps generated by ktime_get_real() are stored in 456 * skb->tstamp. 457 * 458 * hwtstamps can only be compared against other hwtstamps from 459 * the same device. 460 * 461 * This structure is attached to packets as part of the 462 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 463 */ 464 struct skb_shared_hwtstamps { 465 union { 466 ktime_t hwtstamp; 467 void *netdev_data; 468 }; 469 }; 470 471 /* Definitions for tx_flags in struct skb_shared_info */ 472 enum { 473 /* generate hardware time stamp */ 474 SKBTX_HW_TSTAMP = 1 << 0, 475 476 /* generate software time stamp when queueing packet to NIC */ 477 SKBTX_SW_TSTAMP = 1 << 1, 478 479 /* device driver is going to provide hardware time stamp */ 480 SKBTX_IN_PROGRESS = 1 << 2, 481 482 /* generate hardware time stamp based on cycles if supported */ 483 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3, 484 485 /* generate wifi status information (where possible) */ 486 SKBTX_WIFI_STATUS = 1 << 4, 487 488 /* determine hardware time stamp based on time or cycles */ 489 SKBTX_HW_TSTAMP_NETDEV = 1 << 5, 490 491 /* generate software time stamp when entering packet scheduling */ 492 SKBTX_SCHED_TSTAMP = 1 << 6, 493 }; 494 495 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 496 SKBTX_SCHED_TSTAMP) 497 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \ 498 SKBTX_HW_TSTAMP_USE_CYCLES | \ 499 SKBTX_ANY_SW_TSTAMP) 500 501 /* Definitions for flags in struct skb_shared_info */ 502 enum { 503 /* use zcopy routines */ 504 SKBFL_ZEROCOPY_ENABLE = BIT(0), 505 506 /* This indicates at least one fragment might be overwritten 507 * (as in vmsplice(), sendfile() ...) 508 * If we need to compute a TX checksum, we'll need to copy 509 * all frags to avoid possible bad checksum 510 */ 511 SKBFL_SHARED_FRAG = BIT(1), 512 513 /* segment contains only zerocopy data and should not be 514 * charged to the kernel memory. 515 */ 516 SKBFL_PURE_ZEROCOPY = BIT(2), 517 518 SKBFL_DONT_ORPHAN = BIT(3), 519 520 /* page references are managed by the ubuf_info, so it's safe to 521 * use frags only up until ubuf_info is released 522 */ 523 SKBFL_MANAGED_FRAG_REFS = BIT(4), 524 }; 525 526 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 527 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \ 528 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS) 529 530 /* 531 * The callback notifies userspace to release buffers when skb DMA is done in 532 * lower device, the skb last reference should be 0 when calling this. 533 * The zerocopy_success argument is true if zero copy transmit occurred, 534 * false on data copy or out of memory error caused by data copy attempt. 535 * The ctx field is used to track device context. 536 * The desc field is used to track userspace buffer index. 537 */ 538 struct ubuf_info { 539 void (*callback)(struct sk_buff *, struct ubuf_info *, 540 bool zerocopy_success); 541 refcount_t refcnt; 542 u8 flags; 543 }; 544 545 struct ubuf_info_msgzc { 546 struct ubuf_info ubuf; 547 548 union { 549 struct { 550 unsigned long desc; 551 void *ctx; 552 }; 553 struct { 554 u32 id; 555 u16 len; 556 u16 zerocopy:1; 557 u32 bytelen; 558 }; 559 }; 560 561 struct mmpin { 562 struct user_struct *user; 563 unsigned int num_pg; 564 } mmp; 565 }; 566 567 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 568 #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \ 569 ubuf) 570 571 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 572 void mm_unaccount_pinned_pages(struct mmpin *mmp); 573 574 /* Preserve some data across TX submission and completion. 575 * 576 * Note, this state is stored in the driver. Extending the layout 577 * might need some special care. 578 */ 579 struct xsk_tx_metadata_compl { 580 __u64 *tx_timestamp; 581 }; 582 583 /* This data is invariant across clones and lives at 584 * the end of the header data, ie. at skb->end. 585 */ 586 struct skb_shared_info { 587 __u8 flags; 588 __u8 meta_len; 589 __u8 nr_frags; 590 __u8 tx_flags; 591 unsigned short gso_size; 592 /* Warning: this field is not always filled in (UFO)! */ 593 unsigned short gso_segs; 594 struct sk_buff *frag_list; 595 union { 596 struct skb_shared_hwtstamps hwtstamps; 597 struct xsk_tx_metadata_compl xsk_meta; 598 }; 599 unsigned int gso_type; 600 u32 tskey; 601 602 /* 603 * Warning : all fields before dataref are cleared in __alloc_skb() 604 */ 605 atomic_t dataref; 606 unsigned int xdp_frags_size; 607 608 /* Intermediate layers must ensure that destructor_arg 609 * remains valid until skb destructor */ 610 void * destructor_arg; 611 612 /* must be last field, see pskb_expand_head() */ 613 skb_frag_t frags[MAX_SKB_FRAGS]; 614 }; 615 616 /** 617 * DOC: dataref and headerless skbs 618 * 619 * Transport layers send out clones of payload skbs they hold for 620 * retransmissions. To allow lower layers of the stack to prepend their headers 621 * we split &skb_shared_info.dataref into two halves. 622 * The lower 16 bits count the overall number of references. 623 * The higher 16 bits indicate how many of the references are payload-only. 624 * skb_header_cloned() checks if skb is allowed to add / write the headers. 625 * 626 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr 627 * (via __skb_header_release()). Any clone created from marked skb will get 628 * &sk_buff.hdr_len populated with the available headroom. 629 * If there's the only clone in existence it's able to modify the headroom 630 * at will. The sequence of calls inside the transport layer is:: 631 * 632 * <alloc skb> 633 * skb_reserve() 634 * __skb_header_release() 635 * skb_clone() 636 * // send the clone down the stack 637 * 638 * This is not a very generic construct and it depends on the transport layers 639 * doing the right thing. In practice there's usually only one payload-only skb. 640 * Having multiple payload-only skbs with different lengths of hdr_len is not 641 * possible. The payload-only skbs should never leave their owner. 642 */ 643 #define SKB_DATAREF_SHIFT 16 644 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 645 646 647 enum { 648 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 649 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 650 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 651 }; 652 653 enum { 654 SKB_GSO_TCPV4 = 1 << 0, 655 656 /* This indicates the skb is from an untrusted source. */ 657 SKB_GSO_DODGY = 1 << 1, 658 659 /* This indicates the tcp segment has CWR set. */ 660 SKB_GSO_TCP_ECN = 1 << 2, 661 662 SKB_GSO_TCP_FIXEDID = 1 << 3, 663 664 SKB_GSO_TCPV6 = 1 << 4, 665 666 SKB_GSO_FCOE = 1 << 5, 667 668 SKB_GSO_GRE = 1 << 6, 669 670 SKB_GSO_GRE_CSUM = 1 << 7, 671 672 SKB_GSO_IPXIP4 = 1 << 8, 673 674 SKB_GSO_IPXIP6 = 1 << 9, 675 676 SKB_GSO_UDP_TUNNEL = 1 << 10, 677 678 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 679 680 SKB_GSO_PARTIAL = 1 << 12, 681 682 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 683 684 SKB_GSO_SCTP = 1 << 14, 685 686 SKB_GSO_ESP = 1 << 15, 687 688 SKB_GSO_UDP = 1 << 16, 689 690 SKB_GSO_UDP_L4 = 1 << 17, 691 692 SKB_GSO_FRAGLIST = 1 << 18, 693 }; 694 695 #if BITS_PER_LONG > 32 696 #define NET_SKBUFF_DATA_USES_OFFSET 1 697 #endif 698 699 #ifdef NET_SKBUFF_DATA_USES_OFFSET 700 typedef unsigned int sk_buff_data_t; 701 #else 702 typedef unsigned char *sk_buff_data_t; 703 #endif 704 705 /** 706 * DOC: Basic sk_buff geometry 707 * 708 * struct sk_buff itself is a metadata structure and does not hold any packet 709 * data. All the data is held in associated buffers. 710 * 711 * &sk_buff.head points to the main "head" buffer. The head buffer is divided 712 * into two parts: 713 * 714 * - data buffer, containing headers and sometimes payload; 715 * this is the part of the skb operated on by the common helpers 716 * such as skb_put() or skb_pull(); 717 * - shared info (struct skb_shared_info) which holds an array of pointers 718 * to read-only data in the (page, offset, length) format. 719 * 720 * Optionally &skb_shared_info.frag_list may point to another skb. 721 * 722 * Basic diagram may look like this:: 723 * 724 * --------------- 725 * | sk_buff | 726 * --------------- 727 * ,--------------------------- + head 728 * / ,----------------- + data 729 * / / ,----------- + tail 730 * | | | , + end 731 * | | | | 732 * v v v v 733 * ----------------------------------------------- 734 * | headroom | data | tailroom | skb_shared_info | 735 * ----------------------------------------------- 736 * + [page frag] 737 * + [page frag] 738 * + [page frag] 739 * + [page frag] --------- 740 * + frag_list --> | sk_buff | 741 * --------- 742 * 743 */ 744 745 /** 746 * struct sk_buff - socket buffer 747 * @next: Next buffer in list 748 * @prev: Previous buffer in list 749 * @tstamp: Time we arrived/left 750 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 751 * for retransmit timer 752 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 753 * @list: queue head 754 * @ll_node: anchor in an llist (eg socket defer_list) 755 * @sk: Socket we are owned by 756 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 757 * fragmentation management 758 * @dev: Device we arrived on/are leaving by 759 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 760 * @cb: Control buffer. Free for use by every layer. Put private vars here 761 * @_skb_refdst: destination entry (with norefcount bit) 762 * @len: Length of actual data 763 * @data_len: Data length 764 * @mac_len: Length of link layer header 765 * @hdr_len: writable header length of cloned skb 766 * @csum: Checksum (must include start/offset pair) 767 * @csum_start: Offset from skb->head where checksumming should start 768 * @csum_offset: Offset from csum_start where checksum should be stored 769 * @priority: Packet queueing priority 770 * @ignore_df: allow local fragmentation 771 * @cloned: Head may be cloned (check refcnt to be sure) 772 * @ip_summed: Driver fed us an IP checksum 773 * @nohdr: Payload reference only, must not modify header 774 * @pkt_type: Packet class 775 * @fclone: skbuff clone status 776 * @ipvs_property: skbuff is owned by ipvs 777 * @inner_protocol_type: whether the inner protocol is 778 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 779 * @remcsum_offload: remote checksum offload is enabled 780 * @offload_fwd_mark: Packet was L2-forwarded in hardware 781 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 782 * @tc_skip_classify: do not classify packet. set by IFB device 783 * @tc_at_ingress: used within tc_classify to distinguish in/egress 784 * @redirected: packet was redirected by packet classifier 785 * @from_ingress: packet was redirected from the ingress path 786 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h 787 * @peeked: this packet has been seen already, so stats have been 788 * done for it, don't do them again 789 * @nf_trace: netfilter packet trace flag 790 * @protocol: Packet protocol from driver 791 * @destructor: Destruct function 792 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 793 * @_sk_redir: socket redirection information for skmsg 794 * @_nfct: Associated connection, if any (with nfctinfo bits) 795 * @skb_iif: ifindex of device we arrived on 796 * @tc_index: Traffic control index 797 * @hash: the packet hash 798 * @queue_mapping: Queue mapping for multiqueue devices 799 * @head_frag: skb was allocated from page fragments, 800 * not allocated by kmalloc() or vmalloc(). 801 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 802 * @pp_recycle: mark the packet for recycling instead of freeing (implies 803 * page_pool support on driver) 804 * @active_extensions: active extensions (skb_ext_id types) 805 * @ndisc_nodetype: router type (from link layer) 806 * @ooo_okay: allow the mapping of a socket to a queue to be changed 807 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 808 * ports. 809 * @sw_hash: indicates hash was computed in software stack 810 * @wifi_acked_valid: wifi_acked was set 811 * @wifi_acked: whether frame was acked on wifi or not 812 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 813 * @encapsulation: indicates the inner headers in the skbuff are valid 814 * @encap_hdr_csum: software checksum is needed 815 * @csum_valid: checksum is already valid 816 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 817 * @csum_complete_sw: checksum was completed by software 818 * @csum_level: indicates the number of consecutive checksums found in 819 * the packet minus one that have been verified as 820 * CHECKSUM_UNNECESSARY (max 3) 821 * @dst_pending_confirm: need to confirm neighbour 822 * @decrypted: Decrypted SKB 823 * @slow_gro: state present at GRO time, slower prepare step required 824 * @mono_delivery_time: When set, skb->tstamp has the 825 * delivery_time in mono clock base (i.e. EDT). Otherwise, the 826 * skb->tstamp has the (rcv) timestamp at ingress and 827 * delivery_time at egress. 828 * @napi_id: id of the NAPI struct this skb came from 829 * @sender_cpu: (aka @napi_id) source CPU in XPS 830 * @alloc_cpu: CPU which did the skb allocation. 831 * @secmark: security marking 832 * @mark: Generic packet mark 833 * @reserved_tailroom: (aka @mark) number of bytes of free space available 834 * at the tail of an sk_buff 835 * @vlan_all: vlan fields (proto & tci) 836 * @vlan_proto: vlan encapsulation protocol 837 * @vlan_tci: vlan tag control information 838 * @inner_protocol: Protocol (encapsulation) 839 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 840 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 841 * @inner_transport_header: Inner transport layer header (encapsulation) 842 * @inner_network_header: Network layer header (encapsulation) 843 * @inner_mac_header: Link layer header (encapsulation) 844 * @transport_header: Transport layer header 845 * @network_header: Network layer header 846 * @mac_header: Link layer header 847 * @kcov_handle: KCOV remote handle for remote coverage collection 848 * @tail: Tail pointer 849 * @end: End pointer 850 * @head: Head of buffer 851 * @data: Data head pointer 852 * @truesize: Buffer size 853 * @users: User count - see {datagram,tcp}.c 854 * @extensions: allocated extensions, valid if active_extensions is nonzero 855 */ 856 857 struct sk_buff { 858 union { 859 struct { 860 /* These two members must be first to match sk_buff_head. */ 861 struct sk_buff *next; 862 struct sk_buff *prev; 863 864 union { 865 struct net_device *dev; 866 /* Some protocols might use this space to store information, 867 * while device pointer would be NULL. 868 * UDP receive path is one user. 869 */ 870 unsigned long dev_scratch; 871 }; 872 }; 873 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 874 struct list_head list; 875 struct llist_node ll_node; 876 }; 877 878 union { 879 struct sock *sk; 880 int ip_defrag_offset; 881 }; 882 883 union { 884 ktime_t tstamp; 885 u64 skb_mstamp_ns; /* earliest departure time */ 886 }; 887 /* 888 * This is the control buffer. It is free to use for every 889 * layer. Please put your private variables there. If you 890 * want to keep them across layers you have to do a skb_clone() 891 * first. This is owned by whoever has the skb queued ATM. 892 */ 893 char cb[48] __aligned(8); 894 895 union { 896 struct { 897 unsigned long _skb_refdst; 898 void (*destructor)(struct sk_buff *skb); 899 }; 900 struct list_head tcp_tsorted_anchor; 901 #ifdef CONFIG_NET_SOCK_MSG 902 unsigned long _sk_redir; 903 #endif 904 }; 905 906 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 907 unsigned long _nfct; 908 #endif 909 unsigned int len, 910 data_len; 911 __u16 mac_len, 912 hdr_len; 913 914 /* Following fields are _not_ copied in __copy_skb_header() 915 * Note that queue_mapping is here mostly to fill a hole. 916 */ 917 __u16 queue_mapping; 918 919 /* if you move cloned around you also must adapt those constants */ 920 #ifdef __BIG_ENDIAN_BITFIELD 921 #define CLONED_MASK (1 << 7) 922 #else 923 #define CLONED_MASK 1 924 #endif 925 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset) 926 927 /* private: */ 928 __u8 __cloned_offset[0]; 929 /* public: */ 930 __u8 cloned:1, 931 nohdr:1, 932 fclone:2, 933 peeked:1, 934 head_frag:1, 935 pfmemalloc:1, 936 pp_recycle:1; /* page_pool recycle indicator */ 937 #ifdef CONFIG_SKB_EXTENSIONS 938 __u8 active_extensions; 939 #endif 940 941 /* Fields enclosed in headers group are copied 942 * using a single memcpy() in __copy_skb_header() 943 */ 944 struct_group(headers, 945 946 /* private: */ 947 __u8 __pkt_type_offset[0]; 948 /* public: */ 949 __u8 pkt_type:3; /* see PKT_TYPE_MAX */ 950 __u8 ignore_df:1; 951 __u8 dst_pending_confirm:1; 952 __u8 ip_summed:2; 953 __u8 ooo_okay:1; 954 955 /* private: */ 956 __u8 __mono_tc_offset[0]; 957 /* public: */ 958 __u8 mono_delivery_time:1; /* See SKB_MONO_DELIVERY_TIME_MASK */ 959 #ifdef CONFIG_NET_XGRESS 960 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */ 961 __u8 tc_skip_classify:1; 962 #endif 963 __u8 remcsum_offload:1; 964 __u8 csum_complete_sw:1; 965 __u8 csum_level:2; 966 __u8 inner_protocol_type:1; 967 968 __u8 l4_hash:1; 969 __u8 sw_hash:1; 970 #ifdef CONFIG_WIRELESS 971 __u8 wifi_acked_valid:1; 972 __u8 wifi_acked:1; 973 #endif 974 __u8 no_fcs:1; 975 /* Indicates the inner headers are valid in the skbuff. */ 976 __u8 encapsulation:1; 977 __u8 encap_hdr_csum:1; 978 __u8 csum_valid:1; 979 #ifdef CONFIG_IPV6_NDISC_NODETYPE 980 __u8 ndisc_nodetype:2; 981 #endif 982 983 #if IS_ENABLED(CONFIG_IP_VS) 984 __u8 ipvs_property:1; 985 #endif 986 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) 987 __u8 nf_trace:1; 988 #endif 989 #ifdef CONFIG_NET_SWITCHDEV 990 __u8 offload_fwd_mark:1; 991 __u8 offload_l3_fwd_mark:1; 992 #endif 993 __u8 redirected:1; 994 #ifdef CONFIG_NET_REDIRECT 995 __u8 from_ingress:1; 996 #endif 997 #ifdef CONFIG_NETFILTER_SKIP_EGRESS 998 __u8 nf_skip_egress:1; 999 #endif 1000 #ifdef CONFIG_TLS_DEVICE 1001 __u8 decrypted:1; 1002 #endif 1003 __u8 slow_gro:1; 1004 #if IS_ENABLED(CONFIG_IP_SCTP) 1005 __u8 csum_not_inet:1; 1006 #endif 1007 1008 #if defined(CONFIG_NET_SCHED) || defined(CONFIG_NET_XGRESS) 1009 __u16 tc_index; /* traffic control index */ 1010 #endif 1011 1012 u16 alloc_cpu; 1013 1014 union { 1015 __wsum csum; 1016 struct { 1017 __u16 csum_start; 1018 __u16 csum_offset; 1019 }; 1020 }; 1021 __u32 priority; 1022 int skb_iif; 1023 __u32 hash; 1024 union { 1025 u32 vlan_all; 1026 struct { 1027 __be16 vlan_proto; 1028 __u16 vlan_tci; 1029 }; 1030 }; 1031 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 1032 union { 1033 unsigned int napi_id; 1034 unsigned int sender_cpu; 1035 }; 1036 #endif 1037 #ifdef CONFIG_NETWORK_SECMARK 1038 __u32 secmark; 1039 #endif 1040 1041 union { 1042 __u32 mark; 1043 __u32 reserved_tailroom; 1044 }; 1045 1046 union { 1047 __be16 inner_protocol; 1048 __u8 inner_ipproto; 1049 }; 1050 1051 __u16 inner_transport_header; 1052 __u16 inner_network_header; 1053 __u16 inner_mac_header; 1054 1055 __be16 protocol; 1056 __u16 transport_header; 1057 __u16 network_header; 1058 __u16 mac_header; 1059 1060 #ifdef CONFIG_KCOV 1061 u64 kcov_handle; 1062 #endif 1063 1064 ); /* end headers group */ 1065 1066 /* These elements must be at the end, see alloc_skb() for details. */ 1067 sk_buff_data_t tail; 1068 sk_buff_data_t end; 1069 unsigned char *head, 1070 *data; 1071 unsigned int truesize; 1072 refcount_t users; 1073 1074 #ifdef CONFIG_SKB_EXTENSIONS 1075 /* only usable after checking ->active_extensions != 0 */ 1076 struct skb_ext *extensions; 1077 #endif 1078 }; 1079 1080 /* if you move pkt_type around you also must adapt those constants */ 1081 #ifdef __BIG_ENDIAN_BITFIELD 1082 #define PKT_TYPE_MAX (7 << 5) 1083 #else 1084 #define PKT_TYPE_MAX 7 1085 #endif 1086 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset) 1087 1088 /* if you move tc_at_ingress or mono_delivery_time 1089 * around, you also must adapt these constants. 1090 */ 1091 #ifdef __BIG_ENDIAN_BITFIELD 1092 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 7) 1093 #define TC_AT_INGRESS_MASK (1 << 6) 1094 #else 1095 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 0) 1096 #define TC_AT_INGRESS_MASK (1 << 1) 1097 #endif 1098 #define SKB_BF_MONO_TC_OFFSET offsetof(struct sk_buff, __mono_tc_offset) 1099 1100 #ifdef __KERNEL__ 1101 /* 1102 * Handling routines are only of interest to the kernel 1103 */ 1104 1105 #define SKB_ALLOC_FCLONE 0x01 1106 #define SKB_ALLOC_RX 0x02 1107 #define SKB_ALLOC_NAPI 0x04 1108 1109 /** 1110 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 1111 * @skb: buffer 1112 */ 1113 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 1114 { 1115 return unlikely(skb->pfmemalloc); 1116 } 1117 1118 /* 1119 * skb might have a dst pointer attached, refcounted or not. 1120 * _skb_refdst low order bit is set if refcount was _not_ taken 1121 */ 1122 #define SKB_DST_NOREF 1UL 1123 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 1124 1125 /** 1126 * skb_dst - returns skb dst_entry 1127 * @skb: buffer 1128 * 1129 * Returns skb dst_entry, regardless of reference taken or not. 1130 */ 1131 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 1132 { 1133 /* If refdst was not refcounted, check we still are in a 1134 * rcu_read_lock section 1135 */ 1136 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 1137 !rcu_read_lock_held() && 1138 !rcu_read_lock_bh_held()); 1139 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 1140 } 1141 1142 /** 1143 * skb_dst_set - sets skb dst 1144 * @skb: buffer 1145 * @dst: dst entry 1146 * 1147 * Sets skb dst, assuming a reference was taken on dst and should 1148 * be released by skb_dst_drop() 1149 */ 1150 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 1151 { 1152 skb->slow_gro |= !!dst; 1153 skb->_skb_refdst = (unsigned long)dst; 1154 } 1155 1156 /** 1157 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 1158 * @skb: buffer 1159 * @dst: dst entry 1160 * 1161 * Sets skb dst, assuming a reference was not taken on dst. 1162 * If dst entry is cached, we do not take reference and dst_release 1163 * will be avoided by refdst_drop. If dst entry is not cached, we take 1164 * reference, so that last dst_release can destroy the dst immediately. 1165 */ 1166 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 1167 { 1168 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1169 skb->slow_gro |= !!dst; 1170 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1171 } 1172 1173 /** 1174 * skb_dst_is_noref - Test if skb dst isn't refcounted 1175 * @skb: buffer 1176 */ 1177 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1178 { 1179 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1180 } 1181 1182 /** 1183 * skb_rtable - Returns the skb &rtable 1184 * @skb: buffer 1185 */ 1186 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1187 { 1188 return (struct rtable *)skb_dst(skb); 1189 } 1190 1191 /* For mangling skb->pkt_type from user space side from applications 1192 * such as nft, tc, etc, we only allow a conservative subset of 1193 * possible pkt_types to be set. 1194 */ 1195 static inline bool skb_pkt_type_ok(u32 ptype) 1196 { 1197 return ptype <= PACKET_OTHERHOST; 1198 } 1199 1200 /** 1201 * skb_napi_id - Returns the skb's NAPI id 1202 * @skb: buffer 1203 */ 1204 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1205 { 1206 #ifdef CONFIG_NET_RX_BUSY_POLL 1207 return skb->napi_id; 1208 #else 1209 return 0; 1210 #endif 1211 } 1212 1213 static inline bool skb_wifi_acked_valid(const struct sk_buff *skb) 1214 { 1215 #ifdef CONFIG_WIRELESS 1216 return skb->wifi_acked_valid; 1217 #else 1218 return 0; 1219 #endif 1220 } 1221 1222 /** 1223 * skb_unref - decrement the skb's reference count 1224 * @skb: buffer 1225 * 1226 * Returns true if we can free the skb. 1227 */ 1228 static inline bool skb_unref(struct sk_buff *skb) 1229 { 1230 if (unlikely(!skb)) 1231 return false; 1232 if (likely(refcount_read(&skb->users) == 1)) 1233 smp_rmb(); 1234 else if (likely(!refcount_dec_and_test(&skb->users))) 1235 return false; 1236 1237 return true; 1238 } 1239 1240 static inline bool skb_data_unref(const struct sk_buff *skb, 1241 struct skb_shared_info *shinfo) 1242 { 1243 int bias; 1244 1245 if (!skb->cloned) 1246 return true; 1247 1248 bias = skb->nohdr ? (1 << SKB_DATAREF_SHIFT) + 1 : 1; 1249 1250 if (atomic_read(&shinfo->dataref) == bias) 1251 smp_rmb(); 1252 else if (atomic_sub_return(bias, &shinfo->dataref)) 1253 return false; 1254 1255 return true; 1256 } 1257 1258 void __fix_address 1259 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason); 1260 1261 /** 1262 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason 1263 * @skb: buffer to free 1264 */ 1265 static inline void kfree_skb(struct sk_buff *skb) 1266 { 1267 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 1268 } 1269 1270 void skb_release_head_state(struct sk_buff *skb); 1271 void kfree_skb_list_reason(struct sk_buff *segs, 1272 enum skb_drop_reason reason); 1273 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1274 void skb_tx_error(struct sk_buff *skb); 1275 1276 static inline void kfree_skb_list(struct sk_buff *segs) 1277 { 1278 kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED); 1279 } 1280 1281 #ifdef CONFIG_TRACEPOINTS 1282 void consume_skb(struct sk_buff *skb); 1283 #else 1284 static inline void consume_skb(struct sk_buff *skb) 1285 { 1286 return kfree_skb(skb); 1287 } 1288 #endif 1289 1290 void __consume_stateless_skb(struct sk_buff *skb); 1291 void __kfree_skb(struct sk_buff *skb); 1292 1293 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1294 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1295 bool *fragstolen, int *delta_truesize); 1296 1297 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1298 int node); 1299 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1300 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1301 struct sk_buff *build_skb_around(struct sk_buff *skb, 1302 void *data, unsigned int frag_size); 1303 void skb_attempt_defer_free(struct sk_buff *skb); 1304 1305 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1306 struct sk_buff *slab_build_skb(void *data); 1307 1308 /** 1309 * alloc_skb - allocate a network buffer 1310 * @size: size to allocate 1311 * @priority: allocation mask 1312 * 1313 * This function is a convenient wrapper around __alloc_skb(). 1314 */ 1315 static inline struct sk_buff *alloc_skb(unsigned int size, 1316 gfp_t priority) 1317 { 1318 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1319 } 1320 1321 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1322 unsigned long data_len, 1323 int max_page_order, 1324 int *errcode, 1325 gfp_t gfp_mask); 1326 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1327 1328 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1329 struct sk_buff_fclones { 1330 struct sk_buff skb1; 1331 1332 struct sk_buff skb2; 1333 1334 refcount_t fclone_ref; 1335 }; 1336 1337 /** 1338 * skb_fclone_busy - check if fclone is busy 1339 * @sk: socket 1340 * @skb: buffer 1341 * 1342 * Returns true if skb is a fast clone, and its clone is not freed. 1343 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1344 * so we also check that didn't happen. 1345 */ 1346 static inline bool skb_fclone_busy(const struct sock *sk, 1347 const struct sk_buff *skb) 1348 { 1349 const struct sk_buff_fclones *fclones; 1350 1351 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1352 1353 return skb->fclone == SKB_FCLONE_ORIG && 1354 refcount_read(&fclones->fclone_ref) > 1 && 1355 READ_ONCE(fclones->skb2.sk) == sk; 1356 } 1357 1358 /** 1359 * alloc_skb_fclone - allocate a network buffer from fclone cache 1360 * @size: size to allocate 1361 * @priority: allocation mask 1362 * 1363 * This function is a convenient wrapper around __alloc_skb(). 1364 */ 1365 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1366 gfp_t priority) 1367 { 1368 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1369 } 1370 1371 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1372 void skb_headers_offset_update(struct sk_buff *skb, int off); 1373 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1374 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1375 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1376 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1377 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1378 gfp_t gfp_mask, bool fclone); 1379 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1380 gfp_t gfp_mask) 1381 { 1382 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1383 } 1384 1385 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1386 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1387 unsigned int headroom); 1388 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom); 1389 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1390 int newtailroom, gfp_t priority); 1391 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1392 int offset, int len); 1393 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1394 int offset, int len); 1395 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1396 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1397 1398 /** 1399 * skb_pad - zero pad the tail of an skb 1400 * @skb: buffer to pad 1401 * @pad: space to pad 1402 * 1403 * Ensure that a buffer is followed by a padding area that is zero 1404 * filled. Used by network drivers which may DMA or transfer data 1405 * beyond the buffer end onto the wire. 1406 * 1407 * May return error in out of memory cases. The skb is freed on error. 1408 */ 1409 static inline int skb_pad(struct sk_buff *skb, int pad) 1410 { 1411 return __skb_pad(skb, pad, true); 1412 } 1413 #define dev_kfree_skb(a) consume_skb(a) 1414 1415 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1416 int offset, size_t size, size_t max_frags); 1417 1418 struct skb_seq_state { 1419 __u32 lower_offset; 1420 __u32 upper_offset; 1421 __u32 frag_idx; 1422 __u32 stepped_offset; 1423 struct sk_buff *root_skb; 1424 struct sk_buff *cur_skb; 1425 __u8 *frag_data; 1426 __u32 frag_off; 1427 }; 1428 1429 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1430 unsigned int to, struct skb_seq_state *st); 1431 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1432 struct skb_seq_state *st); 1433 void skb_abort_seq_read(struct skb_seq_state *st); 1434 1435 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1436 unsigned int to, struct ts_config *config); 1437 1438 /* 1439 * Packet hash types specify the type of hash in skb_set_hash. 1440 * 1441 * Hash types refer to the protocol layer addresses which are used to 1442 * construct a packet's hash. The hashes are used to differentiate or identify 1443 * flows of the protocol layer for the hash type. Hash types are either 1444 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1445 * 1446 * Properties of hashes: 1447 * 1448 * 1) Two packets in different flows have different hash values 1449 * 2) Two packets in the same flow should have the same hash value 1450 * 1451 * A hash at a higher layer is considered to be more specific. A driver should 1452 * set the most specific hash possible. 1453 * 1454 * A driver cannot indicate a more specific hash than the layer at which a hash 1455 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1456 * 1457 * A driver may indicate a hash level which is less specific than the 1458 * actual layer the hash was computed on. For instance, a hash computed 1459 * at L4 may be considered an L3 hash. This should only be done if the 1460 * driver can't unambiguously determine that the HW computed the hash at 1461 * the higher layer. Note that the "should" in the second property above 1462 * permits this. 1463 */ 1464 enum pkt_hash_types { 1465 PKT_HASH_TYPE_NONE, /* Undefined type */ 1466 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1467 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1468 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1469 }; 1470 1471 static inline void skb_clear_hash(struct sk_buff *skb) 1472 { 1473 skb->hash = 0; 1474 skb->sw_hash = 0; 1475 skb->l4_hash = 0; 1476 } 1477 1478 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1479 { 1480 if (!skb->l4_hash) 1481 skb_clear_hash(skb); 1482 } 1483 1484 static inline void 1485 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1486 { 1487 skb->l4_hash = is_l4; 1488 skb->sw_hash = is_sw; 1489 skb->hash = hash; 1490 } 1491 1492 static inline void 1493 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1494 { 1495 /* Used by drivers to set hash from HW */ 1496 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1497 } 1498 1499 static inline void 1500 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1501 { 1502 __skb_set_hash(skb, hash, true, is_l4); 1503 } 1504 1505 void __skb_get_hash(struct sk_buff *skb); 1506 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1507 u32 skb_get_poff(const struct sk_buff *skb); 1508 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1509 const struct flow_keys_basic *keys, int hlen); 1510 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1511 const void *data, int hlen_proto); 1512 1513 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1514 int thoff, u8 ip_proto) 1515 { 1516 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1517 } 1518 1519 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1520 const struct flow_dissector_key *key, 1521 unsigned int key_count); 1522 1523 struct bpf_flow_dissector; 1524 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1525 __be16 proto, int nhoff, int hlen, unsigned int flags); 1526 1527 bool __skb_flow_dissect(const struct net *net, 1528 const struct sk_buff *skb, 1529 struct flow_dissector *flow_dissector, 1530 void *target_container, const void *data, 1531 __be16 proto, int nhoff, int hlen, unsigned int flags); 1532 1533 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1534 struct flow_dissector *flow_dissector, 1535 void *target_container, unsigned int flags) 1536 { 1537 return __skb_flow_dissect(NULL, skb, flow_dissector, 1538 target_container, NULL, 0, 0, 0, flags); 1539 } 1540 1541 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1542 struct flow_keys *flow, 1543 unsigned int flags) 1544 { 1545 memset(flow, 0, sizeof(*flow)); 1546 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1547 flow, NULL, 0, 0, 0, flags); 1548 } 1549 1550 static inline bool 1551 skb_flow_dissect_flow_keys_basic(const struct net *net, 1552 const struct sk_buff *skb, 1553 struct flow_keys_basic *flow, 1554 const void *data, __be16 proto, 1555 int nhoff, int hlen, unsigned int flags) 1556 { 1557 memset(flow, 0, sizeof(*flow)); 1558 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1559 data, proto, nhoff, hlen, flags); 1560 } 1561 1562 void skb_flow_dissect_meta(const struct sk_buff *skb, 1563 struct flow_dissector *flow_dissector, 1564 void *target_container); 1565 1566 /* Gets a skb connection tracking info, ctinfo map should be a 1567 * map of mapsize to translate enum ip_conntrack_info states 1568 * to user states. 1569 */ 1570 void 1571 skb_flow_dissect_ct(const struct sk_buff *skb, 1572 struct flow_dissector *flow_dissector, 1573 void *target_container, 1574 u16 *ctinfo_map, size_t mapsize, 1575 bool post_ct, u16 zone); 1576 void 1577 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1578 struct flow_dissector *flow_dissector, 1579 void *target_container); 1580 1581 void skb_flow_dissect_hash(const struct sk_buff *skb, 1582 struct flow_dissector *flow_dissector, 1583 void *target_container); 1584 1585 static inline __u32 skb_get_hash(struct sk_buff *skb) 1586 { 1587 if (!skb->l4_hash && !skb->sw_hash) 1588 __skb_get_hash(skb); 1589 1590 return skb->hash; 1591 } 1592 1593 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1594 { 1595 if (!skb->l4_hash && !skb->sw_hash) { 1596 struct flow_keys keys; 1597 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1598 1599 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1600 } 1601 1602 return skb->hash; 1603 } 1604 1605 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1606 const siphash_key_t *perturb); 1607 1608 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1609 { 1610 return skb->hash; 1611 } 1612 1613 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1614 { 1615 to->hash = from->hash; 1616 to->sw_hash = from->sw_hash; 1617 to->l4_hash = from->l4_hash; 1618 }; 1619 1620 static inline int skb_cmp_decrypted(const struct sk_buff *skb1, 1621 const struct sk_buff *skb2) 1622 { 1623 #ifdef CONFIG_TLS_DEVICE 1624 return skb2->decrypted - skb1->decrypted; 1625 #else 1626 return 0; 1627 #endif 1628 } 1629 1630 static inline void skb_copy_decrypted(struct sk_buff *to, 1631 const struct sk_buff *from) 1632 { 1633 #ifdef CONFIG_TLS_DEVICE 1634 to->decrypted = from->decrypted; 1635 #endif 1636 } 1637 1638 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1639 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1640 { 1641 return skb->head + skb->end; 1642 } 1643 1644 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1645 { 1646 return skb->end; 1647 } 1648 1649 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) 1650 { 1651 skb->end = offset; 1652 } 1653 #else 1654 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1655 { 1656 return skb->end; 1657 } 1658 1659 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1660 { 1661 return skb->end - skb->head; 1662 } 1663 1664 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) 1665 { 1666 skb->end = skb->head + offset; 1667 } 1668 #endif 1669 1670 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 1671 struct ubuf_info *uarg); 1672 1673 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 1674 1675 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 1676 bool success); 1677 1678 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk, 1679 struct sk_buff *skb, struct iov_iter *from, 1680 size_t length); 1681 1682 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb, 1683 struct msghdr *msg, int len) 1684 { 1685 return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len); 1686 } 1687 1688 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 1689 struct msghdr *msg, int len, 1690 struct ubuf_info *uarg); 1691 1692 /* Internal */ 1693 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1694 1695 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1696 { 1697 return &skb_shinfo(skb)->hwtstamps; 1698 } 1699 1700 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1701 { 1702 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1703 1704 return is_zcopy ? skb_uarg(skb) : NULL; 1705 } 1706 1707 static inline bool skb_zcopy_pure(const struct sk_buff *skb) 1708 { 1709 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY; 1710 } 1711 1712 static inline bool skb_zcopy_managed(const struct sk_buff *skb) 1713 { 1714 return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS; 1715 } 1716 1717 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1, 1718 const struct sk_buff *skb2) 1719 { 1720 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2); 1721 } 1722 1723 static inline void net_zcopy_get(struct ubuf_info *uarg) 1724 { 1725 refcount_inc(&uarg->refcnt); 1726 } 1727 1728 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1729 { 1730 skb_shinfo(skb)->destructor_arg = uarg; 1731 skb_shinfo(skb)->flags |= uarg->flags; 1732 } 1733 1734 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1735 bool *have_ref) 1736 { 1737 if (skb && uarg && !skb_zcopy(skb)) { 1738 if (unlikely(have_ref && *have_ref)) 1739 *have_ref = false; 1740 else 1741 net_zcopy_get(uarg); 1742 skb_zcopy_init(skb, uarg); 1743 } 1744 } 1745 1746 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1747 { 1748 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1749 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1750 } 1751 1752 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1753 { 1754 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1755 } 1756 1757 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1758 { 1759 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1760 } 1761 1762 static inline void net_zcopy_put(struct ubuf_info *uarg) 1763 { 1764 if (uarg) 1765 uarg->callback(NULL, uarg, true); 1766 } 1767 1768 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1769 { 1770 if (uarg) { 1771 if (uarg->callback == msg_zerocopy_callback) 1772 msg_zerocopy_put_abort(uarg, have_uref); 1773 else if (have_uref) 1774 net_zcopy_put(uarg); 1775 } 1776 } 1777 1778 /* Release a reference on a zerocopy structure */ 1779 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1780 { 1781 struct ubuf_info *uarg = skb_zcopy(skb); 1782 1783 if (uarg) { 1784 if (!skb_zcopy_is_nouarg(skb)) 1785 uarg->callback(skb, uarg, zerocopy_success); 1786 1787 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY; 1788 } 1789 } 1790 1791 void __skb_zcopy_downgrade_managed(struct sk_buff *skb); 1792 1793 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb) 1794 { 1795 if (unlikely(skb_zcopy_managed(skb))) 1796 __skb_zcopy_downgrade_managed(skb); 1797 } 1798 1799 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1800 { 1801 skb->next = NULL; 1802 } 1803 1804 static inline void skb_poison_list(struct sk_buff *skb) 1805 { 1806 #ifdef CONFIG_DEBUG_NET 1807 skb->next = SKB_LIST_POISON_NEXT; 1808 #endif 1809 } 1810 1811 /* Iterate through singly-linked GSO fragments of an skb. */ 1812 #define skb_list_walk_safe(first, skb, next_skb) \ 1813 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1814 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1815 1816 static inline void skb_list_del_init(struct sk_buff *skb) 1817 { 1818 __list_del_entry(&skb->list); 1819 skb_mark_not_on_list(skb); 1820 } 1821 1822 /** 1823 * skb_queue_empty - check if a queue is empty 1824 * @list: queue head 1825 * 1826 * Returns true if the queue is empty, false otherwise. 1827 */ 1828 static inline int skb_queue_empty(const struct sk_buff_head *list) 1829 { 1830 return list->next == (const struct sk_buff *) list; 1831 } 1832 1833 /** 1834 * skb_queue_empty_lockless - check if a queue is empty 1835 * @list: queue head 1836 * 1837 * Returns true if the queue is empty, false otherwise. 1838 * This variant can be used in lockless contexts. 1839 */ 1840 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1841 { 1842 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1843 } 1844 1845 1846 /** 1847 * skb_queue_is_last - check if skb is the last entry in the queue 1848 * @list: queue head 1849 * @skb: buffer 1850 * 1851 * Returns true if @skb is the last buffer on the list. 1852 */ 1853 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1854 const struct sk_buff *skb) 1855 { 1856 return skb->next == (const struct sk_buff *) list; 1857 } 1858 1859 /** 1860 * skb_queue_is_first - check if skb is the first entry in the queue 1861 * @list: queue head 1862 * @skb: buffer 1863 * 1864 * Returns true if @skb is the first buffer on the list. 1865 */ 1866 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1867 const struct sk_buff *skb) 1868 { 1869 return skb->prev == (const struct sk_buff *) list; 1870 } 1871 1872 /** 1873 * skb_queue_next - return the next packet in the queue 1874 * @list: queue head 1875 * @skb: current buffer 1876 * 1877 * Return the next packet in @list after @skb. It is only valid to 1878 * call this if skb_queue_is_last() evaluates to false. 1879 */ 1880 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1881 const struct sk_buff *skb) 1882 { 1883 /* This BUG_ON may seem severe, but if we just return then we 1884 * are going to dereference garbage. 1885 */ 1886 BUG_ON(skb_queue_is_last(list, skb)); 1887 return skb->next; 1888 } 1889 1890 /** 1891 * skb_queue_prev - return the prev packet in the queue 1892 * @list: queue head 1893 * @skb: current buffer 1894 * 1895 * Return the prev packet in @list before @skb. It is only valid to 1896 * call this if skb_queue_is_first() evaluates to false. 1897 */ 1898 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1899 const struct sk_buff *skb) 1900 { 1901 /* This BUG_ON may seem severe, but if we just return then we 1902 * are going to dereference garbage. 1903 */ 1904 BUG_ON(skb_queue_is_first(list, skb)); 1905 return skb->prev; 1906 } 1907 1908 /** 1909 * skb_get - reference buffer 1910 * @skb: buffer to reference 1911 * 1912 * Makes another reference to a socket buffer and returns a pointer 1913 * to the buffer. 1914 */ 1915 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1916 { 1917 refcount_inc(&skb->users); 1918 return skb; 1919 } 1920 1921 /* 1922 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1923 */ 1924 1925 /** 1926 * skb_cloned - is the buffer a clone 1927 * @skb: buffer to check 1928 * 1929 * Returns true if the buffer was generated with skb_clone() and is 1930 * one of multiple shared copies of the buffer. Cloned buffers are 1931 * shared data so must not be written to under normal circumstances. 1932 */ 1933 static inline int skb_cloned(const struct sk_buff *skb) 1934 { 1935 return skb->cloned && 1936 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1937 } 1938 1939 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1940 { 1941 might_sleep_if(gfpflags_allow_blocking(pri)); 1942 1943 if (skb_cloned(skb)) 1944 return pskb_expand_head(skb, 0, 0, pri); 1945 1946 return 0; 1947 } 1948 1949 /* This variant of skb_unclone() makes sure skb->truesize 1950 * and skb_end_offset() are not changed, whenever a new skb->head is needed. 1951 * 1952 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X)) 1953 * when various debugging features are in place. 1954 */ 1955 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri); 1956 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri) 1957 { 1958 might_sleep_if(gfpflags_allow_blocking(pri)); 1959 1960 if (skb_cloned(skb)) 1961 return __skb_unclone_keeptruesize(skb, pri); 1962 return 0; 1963 } 1964 1965 /** 1966 * skb_header_cloned - is the header a clone 1967 * @skb: buffer to check 1968 * 1969 * Returns true if modifying the header part of the buffer requires 1970 * the data to be copied. 1971 */ 1972 static inline int skb_header_cloned(const struct sk_buff *skb) 1973 { 1974 int dataref; 1975 1976 if (!skb->cloned) 1977 return 0; 1978 1979 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1980 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1981 return dataref != 1; 1982 } 1983 1984 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1985 { 1986 might_sleep_if(gfpflags_allow_blocking(pri)); 1987 1988 if (skb_header_cloned(skb)) 1989 return pskb_expand_head(skb, 0, 0, pri); 1990 1991 return 0; 1992 } 1993 1994 /** 1995 * __skb_header_release() - allow clones to use the headroom 1996 * @skb: buffer to operate on 1997 * 1998 * See "DOC: dataref and headerless skbs". 1999 */ 2000 static inline void __skb_header_release(struct sk_buff *skb) 2001 { 2002 skb->nohdr = 1; 2003 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 2004 } 2005 2006 2007 /** 2008 * skb_shared - is the buffer shared 2009 * @skb: buffer to check 2010 * 2011 * Returns true if more than one person has a reference to this 2012 * buffer. 2013 */ 2014 static inline int skb_shared(const struct sk_buff *skb) 2015 { 2016 return refcount_read(&skb->users) != 1; 2017 } 2018 2019 /** 2020 * skb_share_check - check if buffer is shared and if so clone it 2021 * @skb: buffer to check 2022 * @pri: priority for memory allocation 2023 * 2024 * If the buffer is shared the buffer is cloned and the old copy 2025 * drops a reference. A new clone with a single reference is returned. 2026 * If the buffer is not shared the original buffer is returned. When 2027 * being called from interrupt status or with spinlocks held pri must 2028 * be GFP_ATOMIC. 2029 * 2030 * NULL is returned on a memory allocation failure. 2031 */ 2032 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 2033 { 2034 might_sleep_if(gfpflags_allow_blocking(pri)); 2035 if (skb_shared(skb)) { 2036 struct sk_buff *nskb = skb_clone(skb, pri); 2037 2038 if (likely(nskb)) 2039 consume_skb(skb); 2040 else 2041 kfree_skb(skb); 2042 skb = nskb; 2043 } 2044 return skb; 2045 } 2046 2047 /* 2048 * Copy shared buffers into a new sk_buff. We effectively do COW on 2049 * packets to handle cases where we have a local reader and forward 2050 * and a couple of other messy ones. The normal one is tcpdumping 2051 * a packet that's being forwarded. 2052 */ 2053 2054 /** 2055 * skb_unshare - make a copy of a shared buffer 2056 * @skb: buffer to check 2057 * @pri: priority for memory allocation 2058 * 2059 * If the socket buffer is a clone then this function creates a new 2060 * copy of the data, drops a reference count on the old copy and returns 2061 * the new copy with the reference count at 1. If the buffer is not a clone 2062 * the original buffer is returned. When called with a spinlock held or 2063 * from interrupt state @pri must be %GFP_ATOMIC 2064 * 2065 * %NULL is returned on a memory allocation failure. 2066 */ 2067 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 2068 gfp_t pri) 2069 { 2070 might_sleep_if(gfpflags_allow_blocking(pri)); 2071 if (skb_cloned(skb)) { 2072 struct sk_buff *nskb = skb_copy(skb, pri); 2073 2074 /* Free our shared copy */ 2075 if (likely(nskb)) 2076 consume_skb(skb); 2077 else 2078 kfree_skb(skb); 2079 skb = nskb; 2080 } 2081 return skb; 2082 } 2083 2084 /** 2085 * skb_peek - peek at the head of an &sk_buff_head 2086 * @list_: list to peek at 2087 * 2088 * Peek an &sk_buff. Unlike most other operations you _MUST_ 2089 * be careful with this one. A peek leaves the buffer on the 2090 * list and someone else may run off with it. You must hold 2091 * the appropriate locks or have a private queue to do this. 2092 * 2093 * Returns %NULL for an empty list or a pointer to the head element. 2094 * The reference count is not incremented and the reference is therefore 2095 * volatile. Use with caution. 2096 */ 2097 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 2098 { 2099 struct sk_buff *skb = list_->next; 2100 2101 if (skb == (struct sk_buff *)list_) 2102 skb = NULL; 2103 return skb; 2104 } 2105 2106 /** 2107 * __skb_peek - peek at the head of a non-empty &sk_buff_head 2108 * @list_: list to peek at 2109 * 2110 * Like skb_peek(), but the caller knows that the list is not empty. 2111 */ 2112 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 2113 { 2114 return list_->next; 2115 } 2116 2117 /** 2118 * skb_peek_next - peek skb following the given one from a queue 2119 * @skb: skb to start from 2120 * @list_: list to peek at 2121 * 2122 * Returns %NULL when the end of the list is met or a pointer to the 2123 * next element. The reference count is not incremented and the 2124 * reference is therefore volatile. Use with caution. 2125 */ 2126 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 2127 const struct sk_buff_head *list_) 2128 { 2129 struct sk_buff *next = skb->next; 2130 2131 if (next == (struct sk_buff *)list_) 2132 next = NULL; 2133 return next; 2134 } 2135 2136 /** 2137 * skb_peek_tail - peek at the tail of an &sk_buff_head 2138 * @list_: list to peek at 2139 * 2140 * Peek an &sk_buff. Unlike most other operations you _MUST_ 2141 * be careful with this one. A peek leaves the buffer on the 2142 * list and someone else may run off with it. You must hold 2143 * the appropriate locks or have a private queue to do this. 2144 * 2145 * Returns %NULL for an empty list or a pointer to the tail element. 2146 * The reference count is not incremented and the reference is therefore 2147 * volatile. Use with caution. 2148 */ 2149 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 2150 { 2151 struct sk_buff *skb = READ_ONCE(list_->prev); 2152 2153 if (skb == (struct sk_buff *)list_) 2154 skb = NULL; 2155 return skb; 2156 2157 } 2158 2159 /** 2160 * skb_queue_len - get queue length 2161 * @list_: list to measure 2162 * 2163 * Return the length of an &sk_buff queue. 2164 */ 2165 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 2166 { 2167 return list_->qlen; 2168 } 2169 2170 /** 2171 * skb_queue_len_lockless - get queue length 2172 * @list_: list to measure 2173 * 2174 * Return the length of an &sk_buff queue. 2175 * This variant can be used in lockless contexts. 2176 */ 2177 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 2178 { 2179 return READ_ONCE(list_->qlen); 2180 } 2181 2182 /** 2183 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 2184 * @list: queue to initialize 2185 * 2186 * This initializes only the list and queue length aspects of 2187 * an sk_buff_head object. This allows to initialize the list 2188 * aspects of an sk_buff_head without reinitializing things like 2189 * the spinlock. It can also be used for on-stack sk_buff_head 2190 * objects where the spinlock is known to not be used. 2191 */ 2192 static inline void __skb_queue_head_init(struct sk_buff_head *list) 2193 { 2194 list->prev = list->next = (struct sk_buff *)list; 2195 list->qlen = 0; 2196 } 2197 2198 /* 2199 * This function creates a split out lock class for each invocation; 2200 * this is needed for now since a whole lot of users of the skb-queue 2201 * infrastructure in drivers have different locking usage (in hardirq) 2202 * than the networking core (in softirq only). In the long run either the 2203 * network layer or drivers should need annotation to consolidate the 2204 * main types of usage into 3 classes. 2205 */ 2206 static inline void skb_queue_head_init(struct sk_buff_head *list) 2207 { 2208 spin_lock_init(&list->lock); 2209 __skb_queue_head_init(list); 2210 } 2211 2212 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 2213 struct lock_class_key *class) 2214 { 2215 skb_queue_head_init(list); 2216 lockdep_set_class(&list->lock, class); 2217 } 2218 2219 /* 2220 * Insert an sk_buff on a list. 2221 * 2222 * The "__skb_xxxx()" functions are the non-atomic ones that 2223 * can only be called with interrupts disabled. 2224 */ 2225 static inline void __skb_insert(struct sk_buff *newsk, 2226 struct sk_buff *prev, struct sk_buff *next, 2227 struct sk_buff_head *list) 2228 { 2229 /* See skb_queue_empty_lockless() and skb_peek_tail() 2230 * for the opposite READ_ONCE() 2231 */ 2232 WRITE_ONCE(newsk->next, next); 2233 WRITE_ONCE(newsk->prev, prev); 2234 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk); 2235 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk); 2236 WRITE_ONCE(list->qlen, list->qlen + 1); 2237 } 2238 2239 static inline void __skb_queue_splice(const struct sk_buff_head *list, 2240 struct sk_buff *prev, 2241 struct sk_buff *next) 2242 { 2243 struct sk_buff *first = list->next; 2244 struct sk_buff *last = list->prev; 2245 2246 WRITE_ONCE(first->prev, prev); 2247 WRITE_ONCE(prev->next, first); 2248 2249 WRITE_ONCE(last->next, next); 2250 WRITE_ONCE(next->prev, last); 2251 } 2252 2253 /** 2254 * skb_queue_splice - join two skb lists, this is designed for stacks 2255 * @list: the new list to add 2256 * @head: the place to add it in the first list 2257 */ 2258 static inline void skb_queue_splice(const struct sk_buff_head *list, 2259 struct sk_buff_head *head) 2260 { 2261 if (!skb_queue_empty(list)) { 2262 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 2263 head->qlen += list->qlen; 2264 } 2265 } 2266 2267 /** 2268 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 2269 * @list: the new list to add 2270 * @head: the place to add it in the first list 2271 * 2272 * The list at @list is reinitialised 2273 */ 2274 static inline void skb_queue_splice_init(struct sk_buff_head *list, 2275 struct sk_buff_head *head) 2276 { 2277 if (!skb_queue_empty(list)) { 2278 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 2279 head->qlen += list->qlen; 2280 __skb_queue_head_init(list); 2281 } 2282 } 2283 2284 /** 2285 * skb_queue_splice_tail - join two skb lists, each list being a queue 2286 * @list: the new list to add 2287 * @head: the place to add it in the first list 2288 */ 2289 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 2290 struct sk_buff_head *head) 2291 { 2292 if (!skb_queue_empty(list)) { 2293 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2294 head->qlen += list->qlen; 2295 } 2296 } 2297 2298 /** 2299 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 2300 * @list: the new list to add 2301 * @head: the place to add it in the first list 2302 * 2303 * Each of the lists is a queue. 2304 * The list at @list is reinitialised 2305 */ 2306 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2307 struct sk_buff_head *head) 2308 { 2309 if (!skb_queue_empty(list)) { 2310 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2311 head->qlen += list->qlen; 2312 __skb_queue_head_init(list); 2313 } 2314 } 2315 2316 /** 2317 * __skb_queue_after - queue a buffer at the list head 2318 * @list: list to use 2319 * @prev: place after this buffer 2320 * @newsk: buffer to queue 2321 * 2322 * Queue a buffer int the middle of a list. This function takes no locks 2323 * and you must therefore hold required locks before calling it. 2324 * 2325 * A buffer cannot be placed on two lists at the same time. 2326 */ 2327 static inline void __skb_queue_after(struct sk_buff_head *list, 2328 struct sk_buff *prev, 2329 struct sk_buff *newsk) 2330 { 2331 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list); 2332 } 2333 2334 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2335 struct sk_buff_head *list); 2336 2337 static inline void __skb_queue_before(struct sk_buff_head *list, 2338 struct sk_buff *next, 2339 struct sk_buff *newsk) 2340 { 2341 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list); 2342 } 2343 2344 /** 2345 * __skb_queue_head - queue a buffer at the list head 2346 * @list: list to use 2347 * @newsk: buffer to queue 2348 * 2349 * Queue a buffer at the start of a list. This function takes no locks 2350 * and you must therefore hold required locks before calling it. 2351 * 2352 * A buffer cannot be placed on two lists at the same time. 2353 */ 2354 static inline void __skb_queue_head(struct sk_buff_head *list, 2355 struct sk_buff *newsk) 2356 { 2357 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2358 } 2359 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2360 2361 /** 2362 * __skb_queue_tail - queue a buffer at the list tail 2363 * @list: list to use 2364 * @newsk: buffer to queue 2365 * 2366 * Queue a buffer at the end of a list. This function takes no locks 2367 * and you must therefore hold required locks before calling it. 2368 * 2369 * A buffer cannot be placed on two lists at the same time. 2370 */ 2371 static inline void __skb_queue_tail(struct sk_buff_head *list, 2372 struct sk_buff *newsk) 2373 { 2374 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2375 } 2376 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2377 2378 /* 2379 * remove sk_buff from list. _Must_ be called atomically, and with 2380 * the list known.. 2381 */ 2382 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2383 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2384 { 2385 struct sk_buff *next, *prev; 2386 2387 WRITE_ONCE(list->qlen, list->qlen - 1); 2388 next = skb->next; 2389 prev = skb->prev; 2390 skb->next = skb->prev = NULL; 2391 WRITE_ONCE(next->prev, prev); 2392 WRITE_ONCE(prev->next, next); 2393 } 2394 2395 /** 2396 * __skb_dequeue - remove from the head of the queue 2397 * @list: list to dequeue from 2398 * 2399 * Remove the head of the list. This function does not take any locks 2400 * so must be used with appropriate locks held only. The head item is 2401 * returned or %NULL if the list is empty. 2402 */ 2403 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2404 { 2405 struct sk_buff *skb = skb_peek(list); 2406 if (skb) 2407 __skb_unlink(skb, list); 2408 return skb; 2409 } 2410 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2411 2412 /** 2413 * __skb_dequeue_tail - remove from the tail of the queue 2414 * @list: list to dequeue from 2415 * 2416 * Remove the tail of the list. This function does not take any locks 2417 * so must be used with appropriate locks held only. The tail item is 2418 * returned or %NULL if the list is empty. 2419 */ 2420 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2421 { 2422 struct sk_buff *skb = skb_peek_tail(list); 2423 if (skb) 2424 __skb_unlink(skb, list); 2425 return skb; 2426 } 2427 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2428 2429 2430 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2431 { 2432 return skb->data_len; 2433 } 2434 2435 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2436 { 2437 return skb->len - skb->data_len; 2438 } 2439 2440 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2441 { 2442 unsigned int i, len = 0; 2443 2444 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2445 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2446 return len; 2447 } 2448 2449 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2450 { 2451 return skb_headlen(skb) + __skb_pagelen(skb); 2452 } 2453 2454 static inline void skb_frag_fill_netmem_desc(skb_frag_t *frag, 2455 netmem_ref netmem, int off, 2456 int size) 2457 { 2458 frag->netmem = netmem; 2459 frag->offset = off; 2460 skb_frag_size_set(frag, size); 2461 } 2462 2463 static inline void skb_frag_fill_page_desc(skb_frag_t *frag, 2464 struct page *page, 2465 int off, int size) 2466 { 2467 skb_frag_fill_netmem_desc(frag, page_to_netmem(page), off, size); 2468 } 2469 2470 static inline void __skb_fill_netmem_desc_noacc(struct skb_shared_info *shinfo, 2471 int i, netmem_ref netmem, 2472 int off, int size) 2473 { 2474 skb_frag_t *frag = &shinfo->frags[i]; 2475 2476 skb_frag_fill_netmem_desc(frag, netmem, off, size); 2477 } 2478 2479 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo, 2480 int i, struct page *page, 2481 int off, int size) 2482 { 2483 __skb_fill_netmem_desc_noacc(shinfo, i, page_to_netmem(page), off, 2484 size); 2485 } 2486 2487 /** 2488 * skb_len_add - adds a number to len fields of skb 2489 * @skb: buffer to add len to 2490 * @delta: number of bytes to add 2491 */ 2492 static inline void skb_len_add(struct sk_buff *skb, int delta) 2493 { 2494 skb->len += delta; 2495 skb->data_len += delta; 2496 skb->truesize += delta; 2497 } 2498 2499 /** 2500 * __skb_fill_netmem_desc - initialise a fragment in an skb 2501 * @skb: buffer containing fragment to be initialised 2502 * @i: fragment index to initialise 2503 * @netmem: the netmem to use for this fragment 2504 * @off: the offset to the data with @page 2505 * @size: the length of the data 2506 * 2507 * Initialises the @i'th fragment of @skb to point to &size bytes at 2508 * offset @off within @page. 2509 * 2510 * Does not take any additional reference on the fragment. 2511 */ 2512 static inline void __skb_fill_netmem_desc(struct sk_buff *skb, int i, 2513 netmem_ref netmem, int off, int size) 2514 { 2515 struct page *page = netmem_to_page(netmem); 2516 2517 __skb_fill_netmem_desc_noacc(skb_shinfo(skb), i, netmem, off, size); 2518 2519 /* Propagate page pfmemalloc to the skb if we can. The problem is 2520 * that not all callers have unique ownership of the page but rely 2521 * on page_is_pfmemalloc doing the right thing(tm). 2522 */ 2523 page = compound_head(page); 2524 if (page_is_pfmemalloc(page)) 2525 skb->pfmemalloc = true; 2526 } 2527 2528 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2529 struct page *page, int off, int size) 2530 { 2531 __skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size); 2532 } 2533 2534 static inline void skb_fill_netmem_desc(struct sk_buff *skb, int i, 2535 netmem_ref netmem, int off, int size) 2536 { 2537 __skb_fill_netmem_desc(skb, i, netmem, off, size); 2538 skb_shinfo(skb)->nr_frags = i + 1; 2539 } 2540 2541 /** 2542 * skb_fill_page_desc - initialise a paged fragment in an skb 2543 * @skb: buffer containing fragment to be initialised 2544 * @i: paged fragment index to initialise 2545 * @page: the page to use for this fragment 2546 * @off: the offset to the data with @page 2547 * @size: the length of the data 2548 * 2549 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2550 * @skb to point to @size bytes at offset @off within @page. In 2551 * addition updates @skb such that @i is the last fragment. 2552 * 2553 * Does not take any additional reference on the fragment. 2554 */ 2555 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2556 struct page *page, int off, int size) 2557 { 2558 skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size); 2559 } 2560 2561 /** 2562 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb 2563 * @skb: buffer containing fragment to be initialised 2564 * @i: paged fragment index to initialise 2565 * @page: the page to use for this fragment 2566 * @off: the offset to the data with @page 2567 * @size: the length of the data 2568 * 2569 * Variant of skb_fill_page_desc() which does not deal with 2570 * pfmemalloc, if page is not owned by us. 2571 */ 2572 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i, 2573 struct page *page, int off, 2574 int size) 2575 { 2576 struct skb_shared_info *shinfo = skb_shinfo(skb); 2577 2578 __skb_fill_page_desc_noacc(shinfo, i, page, off, size); 2579 shinfo->nr_frags = i + 1; 2580 } 2581 2582 void skb_add_rx_frag_netmem(struct sk_buff *skb, int i, netmem_ref netmem, 2583 int off, int size, unsigned int truesize); 2584 2585 static inline void skb_add_rx_frag(struct sk_buff *skb, int i, 2586 struct page *page, int off, int size, 2587 unsigned int truesize) 2588 { 2589 skb_add_rx_frag_netmem(skb, i, page_to_netmem(page), off, size, 2590 truesize); 2591 } 2592 2593 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2594 unsigned int truesize); 2595 2596 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2597 2598 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2599 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2600 { 2601 return skb->head + skb->tail; 2602 } 2603 2604 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2605 { 2606 skb->tail = skb->data - skb->head; 2607 } 2608 2609 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2610 { 2611 skb_reset_tail_pointer(skb); 2612 skb->tail += offset; 2613 } 2614 2615 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2616 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2617 { 2618 return skb->tail; 2619 } 2620 2621 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2622 { 2623 skb->tail = skb->data; 2624 } 2625 2626 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2627 { 2628 skb->tail = skb->data + offset; 2629 } 2630 2631 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2632 2633 static inline void skb_assert_len(struct sk_buff *skb) 2634 { 2635 #ifdef CONFIG_DEBUG_NET 2636 if (WARN_ONCE(!skb->len, "%s\n", __func__)) 2637 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false); 2638 #endif /* CONFIG_DEBUG_NET */ 2639 } 2640 2641 /* 2642 * Add data to an sk_buff 2643 */ 2644 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2645 void *skb_put(struct sk_buff *skb, unsigned int len); 2646 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2647 { 2648 void *tmp = skb_tail_pointer(skb); 2649 SKB_LINEAR_ASSERT(skb); 2650 skb->tail += len; 2651 skb->len += len; 2652 return tmp; 2653 } 2654 2655 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2656 { 2657 void *tmp = __skb_put(skb, len); 2658 2659 memset(tmp, 0, len); 2660 return tmp; 2661 } 2662 2663 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2664 unsigned int len) 2665 { 2666 void *tmp = __skb_put(skb, len); 2667 2668 memcpy(tmp, data, len); 2669 return tmp; 2670 } 2671 2672 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2673 { 2674 *(u8 *)__skb_put(skb, 1) = val; 2675 } 2676 2677 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2678 { 2679 void *tmp = skb_put(skb, len); 2680 2681 memset(tmp, 0, len); 2682 2683 return tmp; 2684 } 2685 2686 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2687 unsigned int len) 2688 { 2689 void *tmp = skb_put(skb, len); 2690 2691 memcpy(tmp, data, len); 2692 2693 return tmp; 2694 } 2695 2696 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2697 { 2698 *(u8 *)skb_put(skb, 1) = val; 2699 } 2700 2701 void *skb_push(struct sk_buff *skb, unsigned int len); 2702 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2703 { 2704 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX); 2705 2706 skb->data -= len; 2707 skb->len += len; 2708 return skb->data; 2709 } 2710 2711 void *skb_pull(struct sk_buff *skb, unsigned int len); 2712 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2713 { 2714 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX); 2715 2716 skb->len -= len; 2717 if (unlikely(skb->len < skb->data_len)) { 2718 #if defined(CONFIG_DEBUG_NET) 2719 skb->len += len; 2720 pr_err("__skb_pull(len=%u)\n", len); 2721 skb_dump(KERN_ERR, skb, false); 2722 #endif 2723 BUG(); 2724 } 2725 return skb->data += len; 2726 } 2727 2728 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2729 { 2730 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2731 } 2732 2733 void *skb_pull_data(struct sk_buff *skb, size_t len); 2734 2735 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2736 2737 static inline enum skb_drop_reason 2738 pskb_may_pull_reason(struct sk_buff *skb, unsigned int len) 2739 { 2740 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX); 2741 2742 if (likely(len <= skb_headlen(skb))) 2743 return SKB_NOT_DROPPED_YET; 2744 2745 if (unlikely(len > skb->len)) 2746 return SKB_DROP_REASON_PKT_TOO_SMALL; 2747 2748 if (unlikely(!__pskb_pull_tail(skb, len - skb_headlen(skb)))) 2749 return SKB_DROP_REASON_NOMEM; 2750 2751 return SKB_NOT_DROPPED_YET; 2752 } 2753 2754 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2755 { 2756 return pskb_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET; 2757 } 2758 2759 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2760 { 2761 if (!pskb_may_pull(skb, len)) 2762 return NULL; 2763 2764 skb->len -= len; 2765 return skb->data += len; 2766 } 2767 2768 void skb_condense(struct sk_buff *skb); 2769 2770 /** 2771 * skb_headroom - bytes at buffer head 2772 * @skb: buffer to check 2773 * 2774 * Return the number of bytes of free space at the head of an &sk_buff. 2775 */ 2776 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2777 { 2778 return skb->data - skb->head; 2779 } 2780 2781 /** 2782 * skb_tailroom - bytes at buffer end 2783 * @skb: buffer to check 2784 * 2785 * Return the number of bytes of free space at the tail of an sk_buff 2786 */ 2787 static inline int skb_tailroom(const struct sk_buff *skb) 2788 { 2789 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2790 } 2791 2792 /** 2793 * skb_availroom - bytes at buffer end 2794 * @skb: buffer to check 2795 * 2796 * Return the number of bytes of free space at the tail of an sk_buff 2797 * allocated by sk_stream_alloc() 2798 */ 2799 static inline int skb_availroom(const struct sk_buff *skb) 2800 { 2801 if (skb_is_nonlinear(skb)) 2802 return 0; 2803 2804 return skb->end - skb->tail - skb->reserved_tailroom; 2805 } 2806 2807 /** 2808 * skb_reserve - adjust headroom 2809 * @skb: buffer to alter 2810 * @len: bytes to move 2811 * 2812 * Increase the headroom of an empty &sk_buff by reducing the tail 2813 * room. This is only allowed for an empty buffer. 2814 */ 2815 static inline void skb_reserve(struct sk_buff *skb, int len) 2816 { 2817 skb->data += len; 2818 skb->tail += len; 2819 } 2820 2821 /** 2822 * skb_tailroom_reserve - adjust reserved_tailroom 2823 * @skb: buffer to alter 2824 * @mtu: maximum amount of headlen permitted 2825 * @needed_tailroom: minimum amount of reserved_tailroom 2826 * 2827 * Set reserved_tailroom so that headlen can be as large as possible but 2828 * not larger than mtu and tailroom cannot be smaller than 2829 * needed_tailroom. 2830 * The required headroom should already have been reserved before using 2831 * this function. 2832 */ 2833 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2834 unsigned int needed_tailroom) 2835 { 2836 SKB_LINEAR_ASSERT(skb); 2837 if (mtu < skb_tailroom(skb) - needed_tailroom) 2838 /* use at most mtu */ 2839 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2840 else 2841 /* use up to all available space */ 2842 skb->reserved_tailroom = needed_tailroom; 2843 } 2844 2845 #define ENCAP_TYPE_ETHER 0 2846 #define ENCAP_TYPE_IPPROTO 1 2847 2848 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2849 __be16 protocol) 2850 { 2851 skb->inner_protocol = protocol; 2852 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2853 } 2854 2855 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2856 __u8 ipproto) 2857 { 2858 skb->inner_ipproto = ipproto; 2859 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2860 } 2861 2862 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2863 { 2864 skb->inner_mac_header = skb->mac_header; 2865 skb->inner_network_header = skb->network_header; 2866 skb->inner_transport_header = skb->transport_header; 2867 } 2868 2869 static inline void skb_reset_mac_len(struct sk_buff *skb) 2870 { 2871 skb->mac_len = skb->network_header - skb->mac_header; 2872 } 2873 2874 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2875 *skb) 2876 { 2877 return skb->head + skb->inner_transport_header; 2878 } 2879 2880 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2881 { 2882 return skb_inner_transport_header(skb) - skb->data; 2883 } 2884 2885 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2886 { 2887 skb->inner_transport_header = skb->data - skb->head; 2888 } 2889 2890 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2891 const int offset) 2892 { 2893 skb_reset_inner_transport_header(skb); 2894 skb->inner_transport_header += offset; 2895 } 2896 2897 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2898 { 2899 return skb->head + skb->inner_network_header; 2900 } 2901 2902 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2903 { 2904 skb->inner_network_header = skb->data - skb->head; 2905 } 2906 2907 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2908 const int offset) 2909 { 2910 skb_reset_inner_network_header(skb); 2911 skb->inner_network_header += offset; 2912 } 2913 2914 static inline bool skb_inner_network_header_was_set(const struct sk_buff *skb) 2915 { 2916 return skb->inner_network_header > 0; 2917 } 2918 2919 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2920 { 2921 return skb->head + skb->inner_mac_header; 2922 } 2923 2924 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2925 { 2926 skb->inner_mac_header = skb->data - skb->head; 2927 } 2928 2929 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2930 const int offset) 2931 { 2932 skb_reset_inner_mac_header(skb); 2933 skb->inner_mac_header += offset; 2934 } 2935 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2936 { 2937 return skb->transport_header != (typeof(skb->transport_header))~0U; 2938 } 2939 2940 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2941 { 2942 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb)); 2943 return skb->head + skb->transport_header; 2944 } 2945 2946 static inline void skb_reset_transport_header(struct sk_buff *skb) 2947 { 2948 skb->transport_header = skb->data - skb->head; 2949 } 2950 2951 static inline void skb_set_transport_header(struct sk_buff *skb, 2952 const int offset) 2953 { 2954 skb_reset_transport_header(skb); 2955 skb->transport_header += offset; 2956 } 2957 2958 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2959 { 2960 return skb->head + skb->network_header; 2961 } 2962 2963 static inline void skb_reset_network_header(struct sk_buff *skb) 2964 { 2965 skb->network_header = skb->data - skb->head; 2966 } 2967 2968 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2969 { 2970 skb_reset_network_header(skb); 2971 skb->network_header += offset; 2972 } 2973 2974 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2975 { 2976 return skb->mac_header != (typeof(skb->mac_header))~0U; 2977 } 2978 2979 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2980 { 2981 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb)); 2982 return skb->head + skb->mac_header; 2983 } 2984 2985 static inline int skb_mac_offset(const struct sk_buff *skb) 2986 { 2987 return skb_mac_header(skb) - skb->data; 2988 } 2989 2990 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2991 { 2992 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb)); 2993 return skb->network_header - skb->mac_header; 2994 } 2995 2996 static inline void skb_unset_mac_header(struct sk_buff *skb) 2997 { 2998 skb->mac_header = (typeof(skb->mac_header))~0U; 2999 } 3000 3001 static inline void skb_reset_mac_header(struct sk_buff *skb) 3002 { 3003 skb->mac_header = skb->data - skb->head; 3004 } 3005 3006 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 3007 { 3008 skb_reset_mac_header(skb); 3009 skb->mac_header += offset; 3010 } 3011 3012 static inline void skb_pop_mac_header(struct sk_buff *skb) 3013 { 3014 skb->mac_header = skb->network_header; 3015 } 3016 3017 static inline void skb_probe_transport_header(struct sk_buff *skb) 3018 { 3019 struct flow_keys_basic keys; 3020 3021 if (skb_transport_header_was_set(skb)) 3022 return; 3023 3024 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 3025 NULL, 0, 0, 0, 0)) 3026 skb_set_transport_header(skb, keys.control.thoff); 3027 } 3028 3029 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 3030 { 3031 if (skb_mac_header_was_set(skb)) { 3032 const unsigned char *old_mac = skb_mac_header(skb); 3033 3034 skb_set_mac_header(skb, -skb->mac_len); 3035 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 3036 } 3037 } 3038 3039 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 3040 { 3041 return skb->csum_start - skb_headroom(skb); 3042 } 3043 3044 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 3045 { 3046 return skb->head + skb->csum_start; 3047 } 3048 3049 static inline int skb_transport_offset(const struct sk_buff *skb) 3050 { 3051 return skb_transport_header(skb) - skb->data; 3052 } 3053 3054 static inline u32 skb_network_header_len(const struct sk_buff *skb) 3055 { 3056 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb)); 3057 return skb->transport_header - skb->network_header; 3058 } 3059 3060 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 3061 { 3062 return skb->inner_transport_header - skb->inner_network_header; 3063 } 3064 3065 static inline int skb_network_offset(const struct sk_buff *skb) 3066 { 3067 return skb_network_header(skb) - skb->data; 3068 } 3069 3070 static inline int skb_inner_network_offset(const struct sk_buff *skb) 3071 { 3072 return skb_inner_network_header(skb) - skb->data; 3073 } 3074 3075 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 3076 { 3077 return pskb_may_pull(skb, skb_network_offset(skb) + len); 3078 } 3079 3080 /* 3081 * CPUs often take a performance hit when accessing unaligned memory 3082 * locations. The actual performance hit varies, it can be small if the 3083 * hardware handles it or large if we have to take an exception and fix it 3084 * in software. 3085 * 3086 * Since an ethernet header is 14 bytes network drivers often end up with 3087 * the IP header at an unaligned offset. The IP header can be aligned by 3088 * shifting the start of the packet by 2 bytes. Drivers should do this 3089 * with: 3090 * 3091 * skb_reserve(skb, NET_IP_ALIGN); 3092 * 3093 * The downside to this alignment of the IP header is that the DMA is now 3094 * unaligned. On some architectures the cost of an unaligned DMA is high 3095 * and this cost outweighs the gains made by aligning the IP header. 3096 * 3097 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 3098 * to be overridden. 3099 */ 3100 #ifndef NET_IP_ALIGN 3101 #define NET_IP_ALIGN 2 3102 #endif 3103 3104 /* 3105 * The networking layer reserves some headroom in skb data (via 3106 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 3107 * the header has to grow. In the default case, if the header has to grow 3108 * 32 bytes or less we avoid the reallocation. 3109 * 3110 * Unfortunately this headroom changes the DMA alignment of the resulting 3111 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 3112 * on some architectures. An architecture can override this value, 3113 * perhaps setting it to a cacheline in size (since that will maintain 3114 * cacheline alignment of the DMA). It must be a power of 2. 3115 * 3116 * Various parts of the networking layer expect at least 32 bytes of 3117 * headroom, you should not reduce this. 3118 * 3119 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 3120 * to reduce average number of cache lines per packet. 3121 * get_rps_cpu() for example only access one 64 bytes aligned block : 3122 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 3123 */ 3124 #ifndef NET_SKB_PAD 3125 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 3126 #endif 3127 3128 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 3129 3130 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 3131 { 3132 if (WARN_ON(skb_is_nonlinear(skb))) 3133 return; 3134 skb->len = len; 3135 skb_set_tail_pointer(skb, len); 3136 } 3137 3138 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 3139 { 3140 __skb_set_length(skb, len); 3141 } 3142 3143 void skb_trim(struct sk_buff *skb, unsigned int len); 3144 3145 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 3146 { 3147 if (skb->data_len) 3148 return ___pskb_trim(skb, len); 3149 __skb_trim(skb, len); 3150 return 0; 3151 } 3152 3153 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 3154 { 3155 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 3156 } 3157 3158 /** 3159 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 3160 * @skb: buffer to alter 3161 * @len: new length 3162 * 3163 * This is identical to pskb_trim except that the caller knows that 3164 * the skb is not cloned so we should never get an error due to out- 3165 * of-memory. 3166 */ 3167 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 3168 { 3169 int err = pskb_trim(skb, len); 3170 BUG_ON(err); 3171 } 3172 3173 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 3174 { 3175 unsigned int diff = len - skb->len; 3176 3177 if (skb_tailroom(skb) < diff) { 3178 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 3179 GFP_ATOMIC); 3180 if (ret) 3181 return ret; 3182 } 3183 __skb_set_length(skb, len); 3184 return 0; 3185 } 3186 3187 /** 3188 * skb_orphan - orphan a buffer 3189 * @skb: buffer to orphan 3190 * 3191 * If a buffer currently has an owner then we call the owner's 3192 * destructor function and make the @skb unowned. The buffer continues 3193 * to exist but is no longer charged to its former owner. 3194 */ 3195 static inline void skb_orphan(struct sk_buff *skb) 3196 { 3197 if (skb->destructor) { 3198 skb->destructor(skb); 3199 skb->destructor = NULL; 3200 skb->sk = NULL; 3201 } else { 3202 BUG_ON(skb->sk); 3203 } 3204 } 3205 3206 /** 3207 * skb_orphan_frags - orphan the frags contained in a buffer 3208 * @skb: buffer to orphan frags from 3209 * @gfp_mask: allocation mask for replacement pages 3210 * 3211 * For each frag in the SKB which needs a destructor (i.e. has an 3212 * owner) create a copy of that frag and release the original 3213 * page by calling the destructor. 3214 */ 3215 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 3216 { 3217 if (likely(!skb_zcopy(skb))) 3218 return 0; 3219 if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN) 3220 return 0; 3221 return skb_copy_ubufs(skb, gfp_mask); 3222 } 3223 3224 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 3225 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 3226 { 3227 if (likely(!skb_zcopy(skb))) 3228 return 0; 3229 return skb_copy_ubufs(skb, gfp_mask); 3230 } 3231 3232 /** 3233 * __skb_queue_purge_reason - empty a list 3234 * @list: list to empty 3235 * @reason: drop reason 3236 * 3237 * Delete all buffers on an &sk_buff list. Each buffer is removed from 3238 * the list and one reference dropped. This function does not take the 3239 * list lock and the caller must hold the relevant locks to use it. 3240 */ 3241 static inline void __skb_queue_purge_reason(struct sk_buff_head *list, 3242 enum skb_drop_reason reason) 3243 { 3244 struct sk_buff *skb; 3245 3246 while ((skb = __skb_dequeue(list)) != NULL) 3247 kfree_skb_reason(skb, reason); 3248 } 3249 3250 static inline void __skb_queue_purge(struct sk_buff_head *list) 3251 { 3252 __skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE); 3253 } 3254 3255 void skb_queue_purge_reason(struct sk_buff_head *list, 3256 enum skb_drop_reason reason); 3257 3258 static inline void skb_queue_purge(struct sk_buff_head *list) 3259 { 3260 skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE); 3261 } 3262 3263 unsigned int skb_rbtree_purge(struct rb_root *root); 3264 void skb_errqueue_purge(struct sk_buff_head *list); 3265 3266 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 3267 3268 /** 3269 * netdev_alloc_frag - allocate a page fragment 3270 * @fragsz: fragment size 3271 * 3272 * Allocates a frag from a page for receive buffer. 3273 * Uses GFP_ATOMIC allocations. 3274 */ 3275 static inline void *netdev_alloc_frag(unsigned int fragsz) 3276 { 3277 return __netdev_alloc_frag_align(fragsz, ~0u); 3278 } 3279 3280 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 3281 unsigned int align) 3282 { 3283 WARN_ON_ONCE(!is_power_of_2(align)); 3284 return __netdev_alloc_frag_align(fragsz, -align); 3285 } 3286 3287 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 3288 gfp_t gfp_mask); 3289 3290 /** 3291 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 3292 * @dev: network device to receive on 3293 * @length: length to allocate 3294 * 3295 * Allocate a new &sk_buff and assign it a usage count of one. The 3296 * buffer has unspecified headroom built in. Users should allocate 3297 * the headroom they think they need without accounting for the 3298 * built in space. The built in space is used for optimisations. 3299 * 3300 * %NULL is returned if there is no free memory. Although this function 3301 * allocates memory it can be called from an interrupt. 3302 */ 3303 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 3304 unsigned int length) 3305 { 3306 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 3307 } 3308 3309 /* legacy helper around __netdev_alloc_skb() */ 3310 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 3311 gfp_t gfp_mask) 3312 { 3313 return __netdev_alloc_skb(NULL, length, gfp_mask); 3314 } 3315 3316 /* legacy helper around netdev_alloc_skb() */ 3317 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 3318 { 3319 return netdev_alloc_skb(NULL, length); 3320 } 3321 3322 3323 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 3324 unsigned int length, gfp_t gfp) 3325 { 3326 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 3327 3328 if (NET_IP_ALIGN && skb) 3329 skb_reserve(skb, NET_IP_ALIGN); 3330 return skb; 3331 } 3332 3333 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 3334 unsigned int length) 3335 { 3336 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 3337 } 3338 3339 static inline void skb_free_frag(void *addr) 3340 { 3341 page_frag_free(addr); 3342 } 3343 3344 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 3345 3346 static inline void *napi_alloc_frag(unsigned int fragsz) 3347 { 3348 return __napi_alloc_frag_align(fragsz, ~0u); 3349 } 3350 3351 static inline void *napi_alloc_frag_align(unsigned int fragsz, 3352 unsigned int align) 3353 { 3354 WARN_ON_ONCE(!is_power_of_2(align)); 3355 return __napi_alloc_frag_align(fragsz, -align); 3356 } 3357 3358 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 3359 unsigned int length, gfp_t gfp_mask); 3360 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 3361 unsigned int length) 3362 { 3363 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 3364 } 3365 void napi_consume_skb(struct sk_buff *skb, int budget); 3366 3367 void napi_skb_free_stolen_head(struct sk_buff *skb); 3368 void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason); 3369 3370 /** 3371 * __dev_alloc_pages - allocate page for network Rx 3372 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 3373 * @order: size of the allocation 3374 * 3375 * Allocate a new page. 3376 * 3377 * %NULL is returned if there is no free memory. 3378 */ 3379 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 3380 unsigned int order) 3381 { 3382 /* This piece of code contains several assumptions. 3383 * 1. This is for device Rx, therefore a cold page is preferred. 3384 * 2. The expectation is the user wants a compound page. 3385 * 3. If requesting a order 0 page it will not be compound 3386 * due to the check to see if order has a value in prep_new_page 3387 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 3388 * code in gfp_to_alloc_flags that should be enforcing this. 3389 */ 3390 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 3391 3392 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 3393 } 3394 3395 static inline struct page *dev_alloc_pages(unsigned int order) 3396 { 3397 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 3398 } 3399 3400 /** 3401 * __dev_alloc_page - allocate a page for network Rx 3402 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 3403 * 3404 * Allocate a new page. 3405 * 3406 * %NULL is returned if there is no free memory. 3407 */ 3408 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 3409 { 3410 return __dev_alloc_pages(gfp_mask, 0); 3411 } 3412 3413 static inline struct page *dev_alloc_page(void) 3414 { 3415 return dev_alloc_pages(0); 3416 } 3417 3418 /** 3419 * dev_page_is_reusable - check whether a page can be reused for network Rx 3420 * @page: the page to test 3421 * 3422 * A page shouldn't be considered for reusing/recycling if it was allocated 3423 * under memory pressure or at a distant memory node. 3424 * 3425 * Returns false if this page should be returned to page allocator, true 3426 * otherwise. 3427 */ 3428 static inline bool dev_page_is_reusable(const struct page *page) 3429 { 3430 return likely(page_to_nid(page) == numa_mem_id() && 3431 !page_is_pfmemalloc(page)); 3432 } 3433 3434 /** 3435 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 3436 * @page: The page that was allocated from skb_alloc_page 3437 * @skb: The skb that may need pfmemalloc set 3438 */ 3439 static inline void skb_propagate_pfmemalloc(const struct page *page, 3440 struct sk_buff *skb) 3441 { 3442 if (page_is_pfmemalloc(page)) 3443 skb->pfmemalloc = true; 3444 } 3445 3446 /** 3447 * skb_frag_off() - Returns the offset of a skb fragment 3448 * @frag: the paged fragment 3449 */ 3450 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3451 { 3452 return frag->offset; 3453 } 3454 3455 /** 3456 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3457 * @frag: skb fragment 3458 * @delta: value to add 3459 */ 3460 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3461 { 3462 frag->offset += delta; 3463 } 3464 3465 /** 3466 * skb_frag_off_set() - Sets the offset of a skb fragment 3467 * @frag: skb fragment 3468 * @offset: offset of fragment 3469 */ 3470 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3471 { 3472 frag->offset = offset; 3473 } 3474 3475 /** 3476 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3477 * @fragto: skb fragment where offset is set 3478 * @fragfrom: skb fragment offset is copied from 3479 */ 3480 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3481 const skb_frag_t *fragfrom) 3482 { 3483 fragto->offset = fragfrom->offset; 3484 } 3485 3486 /** 3487 * skb_frag_page - retrieve the page referred to by a paged fragment 3488 * @frag: the paged fragment 3489 * 3490 * Returns the &struct page associated with @frag. 3491 */ 3492 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3493 { 3494 return netmem_to_page(frag->netmem); 3495 } 3496 3497 /** 3498 * __skb_frag_ref - take an addition reference on a paged fragment. 3499 * @frag: the paged fragment 3500 * 3501 * Takes an additional reference on the paged fragment @frag. 3502 */ 3503 static inline void __skb_frag_ref(skb_frag_t *frag) 3504 { 3505 get_page(skb_frag_page(frag)); 3506 } 3507 3508 /** 3509 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3510 * @skb: the buffer 3511 * @f: the fragment offset. 3512 * 3513 * Takes an additional reference on the @f'th paged fragment of @skb. 3514 */ 3515 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3516 { 3517 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3518 } 3519 3520 int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb, 3521 unsigned int headroom); 3522 int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb, 3523 struct bpf_prog *prog); 3524 bool napi_pp_put_page(struct page *page, bool napi_safe); 3525 3526 static inline void 3527 skb_page_unref(const struct sk_buff *skb, struct page *page, bool napi_safe) 3528 { 3529 #ifdef CONFIG_PAGE_POOL 3530 if (skb->pp_recycle && napi_pp_put_page(page, napi_safe)) 3531 return; 3532 #endif 3533 put_page(page); 3534 } 3535 3536 static inline void 3537 napi_frag_unref(skb_frag_t *frag, bool recycle, bool napi_safe) 3538 { 3539 struct page *page = skb_frag_page(frag); 3540 3541 #ifdef CONFIG_PAGE_POOL 3542 if (recycle && napi_pp_put_page(page, napi_safe)) 3543 return; 3544 #endif 3545 put_page(page); 3546 } 3547 3548 /** 3549 * __skb_frag_unref - release a reference on a paged fragment. 3550 * @frag: the paged fragment 3551 * @recycle: recycle the page if allocated via page_pool 3552 * 3553 * Releases a reference on the paged fragment @frag 3554 * or recycles the page via the page_pool API. 3555 */ 3556 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) 3557 { 3558 napi_frag_unref(frag, recycle, false); 3559 } 3560 3561 /** 3562 * skb_frag_unref - release a reference on a paged fragment of an skb. 3563 * @skb: the buffer 3564 * @f: the fragment offset 3565 * 3566 * Releases a reference on the @f'th paged fragment of @skb. 3567 */ 3568 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3569 { 3570 struct skb_shared_info *shinfo = skb_shinfo(skb); 3571 3572 if (!skb_zcopy_managed(skb)) 3573 __skb_frag_unref(&shinfo->frags[f], skb->pp_recycle); 3574 } 3575 3576 /** 3577 * skb_frag_address - gets the address of the data contained in a paged fragment 3578 * @frag: the paged fragment buffer 3579 * 3580 * Returns the address of the data within @frag. The page must already 3581 * be mapped. 3582 */ 3583 static inline void *skb_frag_address(const skb_frag_t *frag) 3584 { 3585 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3586 } 3587 3588 /** 3589 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3590 * @frag: the paged fragment buffer 3591 * 3592 * Returns the address of the data within @frag. Checks that the page 3593 * is mapped and returns %NULL otherwise. 3594 */ 3595 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3596 { 3597 void *ptr = page_address(skb_frag_page(frag)); 3598 if (unlikely(!ptr)) 3599 return NULL; 3600 3601 return ptr + skb_frag_off(frag); 3602 } 3603 3604 /** 3605 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3606 * @fragto: skb fragment where page is set 3607 * @fragfrom: skb fragment page is copied from 3608 */ 3609 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3610 const skb_frag_t *fragfrom) 3611 { 3612 fragto->netmem = fragfrom->netmem; 3613 } 3614 3615 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3616 3617 /** 3618 * skb_frag_dma_map - maps a paged fragment via the DMA API 3619 * @dev: the device to map the fragment to 3620 * @frag: the paged fragment to map 3621 * @offset: the offset within the fragment (starting at the 3622 * fragment's own offset) 3623 * @size: the number of bytes to map 3624 * @dir: the direction of the mapping (``PCI_DMA_*``) 3625 * 3626 * Maps the page associated with @frag to @device. 3627 */ 3628 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3629 const skb_frag_t *frag, 3630 size_t offset, size_t size, 3631 enum dma_data_direction dir) 3632 { 3633 return dma_map_page(dev, skb_frag_page(frag), 3634 skb_frag_off(frag) + offset, size, dir); 3635 } 3636 3637 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3638 gfp_t gfp_mask) 3639 { 3640 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3641 } 3642 3643 3644 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3645 gfp_t gfp_mask) 3646 { 3647 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3648 } 3649 3650 3651 /** 3652 * skb_clone_writable - is the header of a clone writable 3653 * @skb: buffer to check 3654 * @len: length up to which to write 3655 * 3656 * Returns true if modifying the header part of the cloned buffer 3657 * does not requires the data to be copied. 3658 */ 3659 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3660 { 3661 return !skb_header_cloned(skb) && 3662 skb_headroom(skb) + len <= skb->hdr_len; 3663 } 3664 3665 static inline int skb_try_make_writable(struct sk_buff *skb, 3666 unsigned int write_len) 3667 { 3668 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3669 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3670 } 3671 3672 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3673 int cloned) 3674 { 3675 int delta = 0; 3676 3677 if (headroom > skb_headroom(skb)) 3678 delta = headroom - skb_headroom(skb); 3679 3680 if (delta || cloned) 3681 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3682 GFP_ATOMIC); 3683 return 0; 3684 } 3685 3686 /** 3687 * skb_cow - copy header of skb when it is required 3688 * @skb: buffer to cow 3689 * @headroom: needed headroom 3690 * 3691 * If the skb passed lacks sufficient headroom or its data part 3692 * is shared, data is reallocated. If reallocation fails, an error 3693 * is returned and original skb is not changed. 3694 * 3695 * The result is skb with writable area skb->head...skb->tail 3696 * and at least @headroom of space at head. 3697 */ 3698 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3699 { 3700 return __skb_cow(skb, headroom, skb_cloned(skb)); 3701 } 3702 3703 /** 3704 * skb_cow_head - skb_cow but only making the head writable 3705 * @skb: buffer to cow 3706 * @headroom: needed headroom 3707 * 3708 * This function is identical to skb_cow except that we replace the 3709 * skb_cloned check by skb_header_cloned. It should be used when 3710 * you only need to push on some header and do not need to modify 3711 * the data. 3712 */ 3713 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3714 { 3715 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3716 } 3717 3718 /** 3719 * skb_padto - pad an skbuff up to a minimal size 3720 * @skb: buffer to pad 3721 * @len: minimal length 3722 * 3723 * Pads up a buffer to ensure the trailing bytes exist and are 3724 * blanked. If the buffer already contains sufficient data it 3725 * is untouched. Otherwise it is extended. Returns zero on 3726 * success. The skb is freed on error. 3727 */ 3728 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3729 { 3730 unsigned int size = skb->len; 3731 if (likely(size >= len)) 3732 return 0; 3733 return skb_pad(skb, len - size); 3734 } 3735 3736 /** 3737 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3738 * @skb: buffer to pad 3739 * @len: minimal length 3740 * @free_on_error: free buffer on error 3741 * 3742 * Pads up a buffer to ensure the trailing bytes exist and are 3743 * blanked. If the buffer already contains sufficient data it 3744 * is untouched. Otherwise it is extended. Returns zero on 3745 * success. The skb is freed on error if @free_on_error is true. 3746 */ 3747 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3748 unsigned int len, 3749 bool free_on_error) 3750 { 3751 unsigned int size = skb->len; 3752 3753 if (unlikely(size < len)) { 3754 len -= size; 3755 if (__skb_pad(skb, len, free_on_error)) 3756 return -ENOMEM; 3757 __skb_put(skb, len); 3758 } 3759 return 0; 3760 } 3761 3762 /** 3763 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3764 * @skb: buffer to pad 3765 * @len: minimal length 3766 * 3767 * Pads up a buffer to ensure the trailing bytes exist and are 3768 * blanked. If the buffer already contains sufficient data it 3769 * is untouched. Otherwise it is extended. Returns zero on 3770 * success. The skb is freed on error. 3771 */ 3772 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3773 { 3774 return __skb_put_padto(skb, len, true); 3775 } 3776 3777 bool csum_and_copy_from_iter_full(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i) 3778 __must_check; 3779 3780 static inline int skb_add_data(struct sk_buff *skb, 3781 struct iov_iter *from, int copy) 3782 { 3783 const int off = skb->len; 3784 3785 if (skb->ip_summed == CHECKSUM_NONE) { 3786 __wsum csum = 0; 3787 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3788 &csum, from)) { 3789 skb->csum = csum_block_add(skb->csum, csum, off); 3790 return 0; 3791 } 3792 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3793 return 0; 3794 3795 __skb_trim(skb, off); 3796 return -EFAULT; 3797 } 3798 3799 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3800 const struct page *page, int off) 3801 { 3802 if (skb_zcopy(skb)) 3803 return false; 3804 if (i) { 3805 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3806 3807 return page == skb_frag_page(frag) && 3808 off == skb_frag_off(frag) + skb_frag_size(frag); 3809 } 3810 return false; 3811 } 3812 3813 static inline int __skb_linearize(struct sk_buff *skb) 3814 { 3815 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3816 } 3817 3818 /** 3819 * skb_linearize - convert paged skb to linear one 3820 * @skb: buffer to linarize 3821 * 3822 * If there is no free memory -ENOMEM is returned, otherwise zero 3823 * is returned and the old skb data released. 3824 */ 3825 static inline int skb_linearize(struct sk_buff *skb) 3826 { 3827 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3828 } 3829 3830 /** 3831 * skb_has_shared_frag - can any frag be overwritten 3832 * @skb: buffer to test 3833 * 3834 * Return true if the skb has at least one frag that might be modified 3835 * by an external entity (as in vmsplice()/sendfile()) 3836 */ 3837 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3838 { 3839 return skb_is_nonlinear(skb) && 3840 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3841 } 3842 3843 /** 3844 * skb_linearize_cow - make sure skb is linear and writable 3845 * @skb: buffer to process 3846 * 3847 * If there is no free memory -ENOMEM is returned, otherwise zero 3848 * is returned and the old skb data released. 3849 */ 3850 static inline int skb_linearize_cow(struct sk_buff *skb) 3851 { 3852 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3853 __skb_linearize(skb) : 0; 3854 } 3855 3856 static __always_inline void 3857 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3858 unsigned int off) 3859 { 3860 if (skb->ip_summed == CHECKSUM_COMPLETE) 3861 skb->csum = csum_block_sub(skb->csum, 3862 csum_partial(start, len, 0), off); 3863 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3864 skb_checksum_start_offset(skb) < 0) 3865 skb->ip_summed = CHECKSUM_NONE; 3866 } 3867 3868 /** 3869 * skb_postpull_rcsum - update checksum for received skb after pull 3870 * @skb: buffer to update 3871 * @start: start of data before pull 3872 * @len: length of data pulled 3873 * 3874 * After doing a pull on a received packet, you need to call this to 3875 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3876 * CHECKSUM_NONE so that it can be recomputed from scratch. 3877 */ 3878 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3879 const void *start, unsigned int len) 3880 { 3881 if (skb->ip_summed == CHECKSUM_COMPLETE) 3882 skb->csum = wsum_negate(csum_partial(start, len, 3883 wsum_negate(skb->csum))); 3884 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3885 skb_checksum_start_offset(skb) < 0) 3886 skb->ip_summed = CHECKSUM_NONE; 3887 } 3888 3889 static __always_inline void 3890 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3891 unsigned int off) 3892 { 3893 if (skb->ip_summed == CHECKSUM_COMPLETE) 3894 skb->csum = csum_block_add(skb->csum, 3895 csum_partial(start, len, 0), off); 3896 } 3897 3898 /** 3899 * skb_postpush_rcsum - update checksum for received skb after push 3900 * @skb: buffer to update 3901 * @start: start of data after push 3902 * @len: length of data pushed 3903 * 3904 * After doing a push on a received packet, you need to call this to 3905 * update the CHECKSUM_COMPLETE checksum. 3906 */ 3907 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3908 const void *start, unsigned int len) 3909 { 3910 __skb_postpush_rcsum(skb, start, len, 0); 3911 } 3912 3913 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3914 3915 /** 3916 * skb_push_rcsum - push skb and update receive checksum 3917 * @skb: buffer to update 3918 * @len: length of data pulled 3919 * 3920 * This function performs an skb_push on the packet and updates 3921 * the CHECKSUM_COMPLETE checksum. It should be used on 3922 * receive path processing instead of skb_push unless you know 3923 * that the checksum difference is zero (e.g., a valid IP header) 3924 * or you are setting ip_summed to CHECKSUM_NONE. 3925 */ 3926 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3927 { 3928 skb_push(skb, len); 3929 skb_postpush_rcsum(skb, skb->data, len); 3930 return skb->data; 3931 } 3932 3933 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3934 /** 3935 * pskb_trim_rcsum - trim received skb and update checksum 3936 * @skb: buffer to trim 3937 * @len: new length 3938 * 3939 * This is exactly the same as pskb_trim except that it ensures the 3940 * checksum of received packets are still valid after the operation. 3941 * It can change skb pointers. 3942 */ 3943 3944 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3945 { 3946 if (likely(len >= skb->len)) 3947 return 0; 3948 return pskb_trim_rcsum_slow(skb, len); 3949 } 3950 3951 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3952 { 3953 if (skb->ip_summed == CHECKSUM_COMPLETE) 3954 skb->ip_summed = CHECKSUM_NONE; 3955 __skb_trim(skb, len); 3956 return 0; 3957 } 3958 3959 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3960 { 3961 if (skb->ip_summed == CHECKSUM_COMPLETE) 3962 skb->ip_summed = CHECKSUM_NONE; 3963 return __skb_grow(skb, len); 3964 } 3965 3966 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3967 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3968 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3969 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3970 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3971 3972 #define skb_queue_walk(queue, skb) \ 3973 for (skb = (queue)->next; \ 3974 skb != (struct sk_buff *)(queue); \ 3975 skb = skb->next) 3976 3977 #define skb_queue_walk_safe(queue, skb, tmp) \ 3978 for (skb = (queue)->next, tmp = skb->next; \ 3979 skb != (struct sk_buff *)(queue); \ 3980 skb = tmp, tmp = skb->next) 3981 3982 #define skb_queue_walk_from(queue, skb) \ 3983 for (; skb != (struct sk_buff *)(queue); \ 3984 skb = skb->next) 3985 3986 #define skb_rbtree_walk(skb, root) \ 3987 for (skb = skb_rb_first(root); skb != NULL; \ 3988 skb = skb_rb_next(skb)) 3989 3990 #define skb_rbtree_walk_from(skb) \ 3991 for (; skb != NULL; \ 3992 skb = skb_rb_next(skb)) 3993 3994 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3995 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3996 skb = tmp) 3997 3998 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3999 for (tmp = skb->next; \ 4000 skb != (struct sk_buff *)(queue); \ 4001 skb = tmp, tmp = skb->next) 4002 4003 #define skb_queue_reverse_walk(queue, skb) \ 4004 for (skb = (queue)->prev; \ 4005 skb != (struct sk_buff *)(queue); \ 4006 skb = skb->prev) 4007 4008 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 4009 for (skb = (queue)->prev, tmp = skb->prev; \ 4010 skb != (struct sk_buff *)(queue); \ 4011 skb = tmp, tmp = skb->prev) 4012 4013 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 4014 for (tmp = skb->prev; \ 4015 skb != (struct sk_buff *)(queue); \ 4016 skb = tmp, tmp = skb->prev) 4017 4018 static inline bool skb_has_frag_list(const struct sk_buff *skb) 4019 { 4020 return skb_shinfo(skb)->frag_list != NULL; 4021 } 4022 4023 static inline void skb_frag_list_init(struct sk_buff *skb) 4024 { 4025 skb_shinfo(skb)->frag_list = NULL; 4026 } 4027 4028 #define skb_walk_frags(skb, iter) \ 4029 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 4030 4031 4032 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 4033 int *err, long *timeo_p, 4034 const struct sk_buff *skb); 4035 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 4036 struct sk_buff_head *queue, 4037 unsigned int flags, 4038 int *off, int *err, 4039 struct sk_buff **last); 4040 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 4041 struct sk_buff_head *queue, 4042 unsigned int flags, int *off, int *err, 4043 struct sk_buff **last); 4044 struct sk_buff *__skb_recv_datagram(struct sock *sk, 4045 struct sk_buff_head *sk_queue, 4046 unsigned int flags, int *off, int *err); 4047 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err); 4048 __poll_t datagram_poll(struct file *file, struct socket *sock, 4049 struct poll_table_struct *wait); 4050 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 4051 struct iov_iter *to, int size); 4052 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 4053 struct msghdr *msg, int size) 4054 { 4055 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 4056 } 4057 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 4058 struct msghdr *msg); 4059 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 4060 struct iov_iter *to, int len, 4061 struct ahash_request *hash); 4062 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 4063 struct iov_iter *from, int len); 4064 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 4065 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 4066 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 4067 static inline void skb_free_datagram_locked(struct sock *sk, 4068 struct sk_buff *skb) 4069 { 4070 __skb_free_datagram_locked(sk, skb, 0); 4071 } 4072 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 4073 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 4074 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 4075 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 4076 int len); 4077 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 4078 struct pipe_inode_info *pipe, unsigned int len, 4079 unsigned int flags); 4080 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 4081 int len); 4082 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len); 4083 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 4084 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 4085 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 4086 int len, int hlen); 4087 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 4088 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 4089 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 4090 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 4091 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 4092 unsigned int offset); 4093 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 4094 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len); 4095 int skb_ensure_writable_head_tail(struct sk_buff *skb, struct net_device *dev); 4096 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 4097 int skb_vlan_pop(struct sk_buff *skb); 4098 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 4099 int skb_eth_pop(struct sk_buff *skb); 4100 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 4101 const unsigned char *src); 4102 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 4103 int mac_len, bool ethernet); 4104 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 4105 bool ethernet); 4106 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 4107 int skb_mpls_dec_ttl(struct sk_buff *skb); 4108 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 4109 gfp_t gfp); 4110 4111 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 4112 { 4113 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 4114 } 4115 4116 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 4117 { 4118 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 4119 } 4120 4121 struct skb_checksum_ops { 4122 __wsum (*update)(const void *mem, int len, __wsum wsum); 4123 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 4124 }; 4125 4126 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 4127 4128 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 4129 __wsum csum, const struct skb_checksum_ops *ops); 4130 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 4131 __wsum csum); 4132 4133 static inline void * __must_check 4134 __skb_header_pointer(const struct sk_buff *skb, int offset, int len, 4135 const void *data, int hlen, void *buffer) 4136 { 4137 if (likely(hlen - offset >= len)) 4138 return (void *)data + offset; 4139 4140 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) 4141 return NULL; 4142 4143 return buffer; 4144 } 4145 4146 static inline void * __must_check 4147 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 4148 { 4149 return __skb_header_pointer(skb, offset, len, skb->data, 4150 skb_headlen(skb), buffer); 4151 } 4152 4153 static inline void * __must_check 4154 skb_pointer_if_linear(const struct sk_buff *skb, int offset, int len) 4155 { 4156 if (likely(skb_headlen(skb) - offset >= len)) 4157 return skb->data + offset; 4158 return NULL; 4159 } 4160 4161 /** 4162 * skb_needs_linearize - check if we need to linearize a given skb 4163 * depending on the given device features. 4164 * @skb: socket buffer to check 4165 * @features: net device features 4166 * 4167 * Returns true if either: 4168 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 4169 * 2. skb is fragmented and the device does not support SG. 4170 */ 4171 static inline bool skb_needs_linearize(struct sk_buff *skb, 4172 netdev_features_t features) 4173 { 4174 return skb_is_nonlinear(skb) && 4175 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 4176 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 4177 } 4178 4179 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 4180 void *to, 4181 const unsigned int len) 4182 { 4183 memcpy(to, skb->data, len); 4184 } 4185 4186 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 4187 const int offset, void *to, 4188 const unsigned int len) 4189 { 4190 memcpy(to, skb->data + offset, len); 4191 } 4192 4193 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 4194 const void *from, 4195 const unsigned int len) 4196 { 4197 memcpy(skb->data, from, len); 4198 } 4199 4200 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 4201 const int offset, 4202 const void *from, 4203 const unsigned int len) 4204 { 4205 memcpy(skb->data + offset, from, len); 4206 } 4207 4208 void skb_init(void); 4209 4210 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 4211 { 4212 return skb->tstamp; 4213 } 4214 4215 /** 4216 * skb_get_timestamp - get timestamp from a skb 4217 * @skb: skb to get stamp from 4218 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 4219 * 4220 * Timestamps are stored in the skb as offsets to a base timestamp. 4221 * This function converts the offset back to a struct timeval and stores 4222 * it in stamp. 4223 */ 4224 static inline void skb_get_timestamp(const struct sk_buff *skb, 4225 struct __kernel_old_timeval *stamp) 4226 { 4227 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 4228 } 4229 4230 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 4231 struct __kernel_sock_timeval *stamp) 4232 { 4233 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 4234 4235 stamp->tv_sec = ts.tv_sec; 4236 stamp->tv_usec = ts.tv_nsec / 1000; 4237 } 4238 4239 static inline void skb_get_timestampns(const struct sk_buff *skb, 4240 struct __kernel_old_timespec *stamp) 4241 { 4242 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 4243 4244 stamp->tv_sec = ts.tv_sec; 4245 stamp->tv_nsec = ts.tv_nsec; 4246 } 4247 4248 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 4249 struct __kernel_timespec *stamp) 4250 { 4251 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 4252 4253 stamp->tv_sec = ts.tv_sec; 4254 stamp->tv_nsec = ts.tv_nsec; 4255 } 4256 4257 static inline void __net_timestamp(struct sk_buff *skb) 4258 { 4259 skb->tstamp = ktime_get_real(); 4260 skb->mono_delivery_time = 0; 4261 } 4262 4263 static inline ktime_t net_timedelta(ktime_t t) 4264 { 4265 return ktime_sub(ktime_get_real(), t); 4266 } 4267 4268 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt, 4269 bool mono) 4270 { 4271 skb->tstamp = kt; 4272 skb->mono_delivery_time = kt && mono; 4273 } 4274 4275 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key); 4276 4277 /* It is used in the ingress path to clear the delivery_time. 4278 * If needed, set the skb->tstamp to the (rcv) timestamp. 4279 */ 4280 static inline void skb_clear_delivery_time(struct sk_buff *skb) 4281 { 4282 if (skb->mono_delivery_time) { 4283 skb->mono_delivery_time = 0; 4284 if (static_branch_unlikely(&netstamp_needed_key)) 4285 skb->tstamp = ktime_get_real(); 4286 else 4287 skb->tstamp = 0; 4288 } 4289 } 4290 4291 static inline void skb_clear_tstamp(struct sk_buff *skb) 4292 { 4293 if (skb->mono_delivery_time) 4294 return; 4295 4296 skb->tstamp = 0; 4297 } 4298 4299 static inline ktime_t skb_tstamp(const struct sk_buff *skb) 4300 { 4301 if (skb->mono_delivery_time) 4302 return 0; 4303 4304 return skb->tstamp; 4305 } 4306 4307 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond) 4308 { 4309 if (!skb->mono_delivery_time && skb->tstamp) 4310 return skb->tstamp; 4311 4312 if (static_branch_unlikely(&netstamp_needed_key) || cond) 4313 return ktime_get_real(); 4314 4315 return 0; 4316 } 4317 4318 static inline u8 skb_metadata_len(const struct sk_buff *skb) 4319 { 4320 return skb_shinfo(skb)->meta_len; 4321 } 4322 4323 static inline void *skb_metadata_end(const struct sk_buff *skb) 4324 { 4325 return skb_mac_header(skb); 4326 } 4327 4328 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 4329 const struct sk_buff *skb_b, 4330 u8 meta_len) 4331 { 4332 const void *a = skb_metadata_end(skb_a); 4333 const void *b = skb_metadata_end(skb_b); 4334 u64 diffs = 0; 4335 4336 if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) || 4337 BITS_PER_LONG != 64) 4338 goto slow; 4339 4340 /* Using more efficient variant than plain call to memcmp(). */ 4341 switch (meta_len) { 4342 #define __it(x, op) (x -= sizeof(u##op)) 4343 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 4344 case 32: diffs |= __it_diff(a, b, 64); 4345 fallthrough; 4346 case 24: diffs |= __it_diff(a, b, 64); 4347 fallthrough; 4348 case 16: diffs |= __it_diff(a, b, 64); 4349 fallthrough; 4350 case 8: diffs |= __it_diff(a, b, 64); 4351 break; 4352 case 28: diffs |= __it_diff(a, b, 64); 4353 fallthrough; 4354 case 20: diffs |= __it_diff(a, b, 64); 4355 fallthrough; 4356 case 12: diffs |= __it_diff(a, b, 64); 4357 fallthrough; 4358 case 4: diffs |= __it_diff(a, b, 32); 4359 break; 4360 default: 4361 slow: 4362 return memcmp(a - meta_len, b - meta_len, meta_len); 4363 } 4364 return diffs; 4365 } 4366 4367 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 4368 const struct sk_buff *skb_b) 4369 { 4370 u8 len_a = skb_metadata_len(skb_a); 4371 u8 len_b = skb_metadata_len(skb_b); 4372 4373 if (!(len_a | len_b)) 4374 return false; 4375 4376 return len_a != len_b ? 4377 true : __skb_metadata_differs(skb_a, skb_b, len_a); 4378 } 4379 4380 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 4381 { 4382 skb_shinfo(skb)->meta_len = meta_len; 4383 } 4384 4385 static inline void skb_metadata_clear(struct sk_buff *skb) 4386 { 4387 skb_metadata_set(skb, 0); 4388 } 4389 4390 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 4391 4392 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 4393 4394 void skb_clone_tx_timestamp(struct sk_buff *skb); 4395 bool skb_defer_rx_timestamp(struct sk_buff *skb); 4396 4397 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 4398 4399 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 4400 { 4401 } 4402 4403 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 4404 { 4405 return false; 4406 } 4407 4408 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 4409 4410 /** 4411 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 4412 * 4413 * PHY drivers may accept clones of transmitted packets for 4414 * timestamping via their phy_driver.txtstamp method. These drivers 4415 * must call this function to return the skb back to the stack with a 4416 * timestamp. 4417 * 4418 * @skb: clone of the original outgoing packet 4419 * @hwtstamps: hardware time stamps 4420 * 4421 */ 4422 void skb_complete_tx_timestamp(struct sk_buff *skb, 4423 struct skb_shared_hwtstamps *hwtstamps); 4424 4425 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 4426 struct skb_shared_hwtstamps *hwtstamps, 4427 struct sock *sk, int tstype); 4428 4429 /** 4430 * skb_tstamp_tx - queue clone of skb with send time stamps 4431 * @orig_skb: the original outgoing packet 4432 * @hwtstamps: hardware time stamps, may be NULL if not available 4433 * 4434 * If the skb has a socket associated, then this function clones the 4435 * skb (thus sharing the actual data and optional structures), stores 4436 * the optional hardware time stamping information (if non NULL) or 4437 * generates a software time stamp (otherwise), then queues the clone 4438 * to the error queue of the socket. Errors are silently ignored. 4439 */ 4440 void skb_tstamp_tx(struct sk_buff *orig_skb, 4441 struct skb_shared_hwtstamps *hwtstamps); 4442 4443 /** 4444 * skb_tx_timestamp() - Driver hook for transmit timestamping 4445 * 4446 * Ethernet MAC Drivers should call this function in their hard_xmit() 4447 * function immediately before giving the sk_buff to the MAC hardware. 4448 * 4449 * Specifically, one should make absolutely sure that this function is 4450 * called before TX completion of this packet can trigger. Otherwise 4451 * the packet could potentially already be freed. 4452 * 4453 * @skb: A socket buffer. 4454 */ 4455 static inline void skb_tx_timestamp(struct sk_buff *skb) 4456 { 4457 skb_clone_tx_timestamp(skb); 4458 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 4459 skb_tstamp_tx(skb, NULL); 4460 } 4461 4462 /** 4463 * skb_complete_wifi_ack - deliver skb with wifi status 4464 * 4465 * @skb: the original outgoing packet 4466 * @acked: ack status 4467 * 4468 */ 4469 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 4470 4471 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 4472 __sum16 __skb_checksum_complete(struct sk_buff *skb); 4473 4474 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 4475 { 4476 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 4477 skb->csum_valid || 4478 (skb->ip_summed == CHECKSUM_PARTIAL && 4479 skb_checksum_start_offset(skb) >= 0)); 4480 } 4481 4482 /** 4483 * skb_checksum_complete - Calculate checksum of an entire packet 4484 * @skb: packet to process 4485 * 4486 * This function calculates the checksum over the entire packet plus 4487 * the value of skb->csum. The latter can be used to supply the 4488 * checksum of a pseudo header as used by TCP/UDP. It returns the 4489 * checksum. 4490 * 4491 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 4492 * this function can be used to verify that checksum on received 4493 * packets. In that case the function should return zero if the 4494 * checksum is correct. In particular, this function will return zero 4495 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 4496 * hardware has already verified the correctness of the checksum. 4497 */ 4498 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 4499 { 4500 return skb_csum_unnecessary(skb) ? 4501 0 : __skb_checksum_complete(skb); 4502 } 4503 4504 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 4505 { 4506 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4507 if (skb->csum_level == 0) 4508 skb->ip_summed = CHECKSUM_NONE; 4509 else 4510 skb->csum_level--; 4511 } 4512 } 4513 4514 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4515 { 4516 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4517 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4518 skb->csum_level++; 4519 } else if (skb->ip_summed == CHECKSUM_NONE) { 4520 skb->ip_summed = CHECKSUM_UNNECESSARY; 4521 skb->csum_level = 0; 4522 } 4523 } 4524 4525 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4526 { 4527 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4528 skb->ip_summed = CHECKSUM_NONE; 4529 skb->csum_level = 0; 4530 } 4531 } 4532 4533 /* Check if we need to perform checksum complete validation. 4534 * 4535 * Returns true if checksum complete is needed, false otherwise 4536 * (either checksum is unnecessary or zero checksum is allowed). 4537 */ 4538 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4539 bool zero_okay, 4540 __sum16 check) 4541 { 4542 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4543 skb->csum_valid = 1; 4544 __skb_decr_checksum_unnecessary(skb); 4545 return false; 4546 } 4547 4548 return true; 4549 } 4550 4551 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4552 * in checksum_init. 4553 */ 4554 #define CHECKSUM_BREAK 76 4555 4556 /* Unset checksum-complete 4557 * 4558 * Unset checksum complete can be done when packet is being modified 4559 * (uncompressed for instance) and checksum-complete value is 4560 * invalidated. 4561 */ 4562 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4563 { 4564 if (skb->ip_summed == CHECKSUM_COMPLETE) 4565 skb->ip_summed = CHECKSUM_NONE; 4566 } 4567 4568 /* Validate (init) checksum based on checksum complete. 4569 * 4570 * Return values: 4571 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4572 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4573 * checksum is stored in skb->csum for use in __skb_checksum_complete 4574 * non-zero: value of invalid checksum 4575 * 4576 */ 4577 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4578 bool complete, 4579 __wsum psum) 4580 { 4581 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4582 if (!csum_fold(csum_add(psum, skb->csum))) { 4583 skb->csum_valid = 1; 4584 return 0; 4585 } 4586 } 4587 4588 skb->csum = psum; 4589 4590 if (complete || skb->len <= CHECKSUM_BREAK) { 4591 __sum16 csum; 4592 4593 csum = __skb_checksum_complete(skb); 4594 skb->csum_valid = !csum; 4595 return csum; 4596 } 4597 4598 return 0; 4599 } 4600 4601 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4602 { 4603 return 0; 4604 } 4605 4606 /* Perform checksum validate (init). Note that this is a macro since we only 4607 * want to calculate the pseudo header which is an input function if necessary. 4608 * First we try to validate without any computation (checksum unnecessary) and 4609 * then calculate based on checksum complete calling the function to compute 4610 * pseudo header. 4611 * 4612 * Return values: 4613 * 0: checksum is validated or try to in skb_checksum_complete 4614 * non-zero: value of invalid checksum 4615 */ 4616 #define __skb_checksum_validate(skb, proto, complete, \ 4617 zero_okay, check, compute_pseudo) \ 4618 ({ \ 4619 __sum16 __ret = 0; \ 4620 skb->csum_valid = 0; \ 4621 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4622 __ret = __skb_checksum_validate_complete(skb, \ 4623 complete, compute_pseudo(skb, proto)); \ 4624 __ret; \ 4625 }) 4626 4627 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4628 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4629 4630 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4631 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4632 4633 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4634 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4635 4636 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4637 compute_pseudo) \ 4638 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4639 4640 #define skb_checksum_simple_validate(skb) \ 4641 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4642 4643 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4644 { 4645 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4646 } 4647 4648 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4649 { 4650 skb->csum = ~pseudo; 4651 skb->ip_summed = CHECKSUM_COMPLETE; 4652 } 4653 4654 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4655 do { \ 4656 if (__skb_checksum_convert_check(skb)) \ 4657 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4658 } while (0) 4659 4660 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4661 u16 start, u16 offset) 4662 { 4663 skb->ip_summed = CHECKSUM_PARTIAL; 4664 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4665 skb->csum_offset = offset - start; 4666 } 4667 4668 /* Update skbuf and packet to reflect the remote checksum offload operation. 4669 * When called, ptr indicates the starting point for skb->csum when 4670 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4671 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4672 */ 4673 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4674 int start, int offset, bool nopartial) 4675 { 4676 __wsum delta; 4677 4678 if (!nopartial) { 4679 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4680 return; 4681 } 4682 4683 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4684 __skb_checksum_complete(skb); 4685 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4686 } 4687 4688 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4689 4690 /* Adjust skb->csum since we changed the packet */ 4691 skb->csum = csum_add(skb->csum, delta); 4692 } 4693 4694 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4695 { 4696 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4697 return (void *)(skb->_nfct & NFCT_PTRMASK); 4698 #else 4699 return NULL; 4700 #endif 4701 } 4702 4703 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4704 { 4705 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4706 return skb->_nfct; 4707 #else 4708 return 0UL; 4709 #endif 4710 } 4711 4712 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4713 { 4714 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4715 skb->slow_gro |= !!nfct; 4716 skb->_nfct = nfct; 4717 #endif 4718 } 4719 4720 #ifdef CONFIG_SKB_EXTENSIONS 4721 enum skb_ext_id { 4722 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4723 SKB_EXT_BRIDGE_NF, 4724 #endif 4725 #ifdef CONFIG_XFRM 4726 SKB_EXT_SEC_PATH, 4727 #endif 4728 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4729 TC_SKB_EXT, 4730 #endif 4731 #if IS_ENABLED(CONFIG_MPTCP) 4732 SKB_EXT_MPTCP, 4733 #endif 4734 #if IS_ENABLED(CONFIG_MCTP_FLOWS) 4735 SKB_EXT_MCTP, 4736 #endif 4737 SKB_EXT_NUM, /* must be last */ 4738 }; 4739 4740 /** 4741 * struct skb_ext - sk_buff extensions 4742 * @refcnt: 1 on allocation, deallocated on 0 4743 * @offset: offset to add to @data to obtain extension address 4744 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4745 * @data: start of extension data, variable sized 4746 * 4747 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4748 * to use 'u8' types while allowing up to 2kb worth of extension data. 4749 */ 4750 struct skb_ext { 4751 refcount_t refcnt; 4752 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4753 u8 chunks; /* same */ 4754 char data[] __aligned(8); 4755 }; 4756 4757 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4758 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4759 struct skb_ext *ext); 4760 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4761 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4762 void __skb_ext_put(struct skb_ext *ext); 4763 4764 static inline void skb_ext_put(struct sk_buff *skb) 4765 { 4766 if (skb->active_extensions) 4767 __skb_ext_put(skb->extensions); 4768 } 4769 4770 static inline void __skb_ext_copy(struct sk_buff *dst, 4771 const struct sk_buff *src) 4772 { 4773 dst->active_extensions = src->active_extensions; 4774 4775 if (src->active_extensions) { 4776 struct skb_ext *ext = src->extensions; 4777 4778 refcount_inc(&ext->refcnt); 4779 dst->extensions = ext; 4780 } 4781 } 4782 4783 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4784 { 4785 skb_ext_put(dst); 4786 __skb_ext_copy(dst, src); 4787 } 4788 4789 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4790 { 4791 return !!ext->offset[i]; 4792 } 4793 4794 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4795 { 4796 return skb->active_extensions & (1 << id); 4797 } 4798 4799 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4800 { 4801 if (skb_ext_exist(skb, id)) 4802 __skb_ext_del(skb, id); 4803 } 4804 4805 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4806 { 4807 if (skb_ext_exist(skb, id)) { 4808 struct skb_ext *ext = skb->extensions; 4809 4810 return (void *)ext + (ext->offset[id] << 3); 4811 } 4812 4813 return NULL; 4814 } 4815 4816 static inline void skb_ext_reset(struct sk_buff *skb) 4817 { 4818 if (unlikely(skb->active_extensions)) { 4819 __skb_ext_put(skb->extensions); 4820 skb->active_extensions = 0; 4821 } 4822 } 4823 4824 static inline bool skb_has_extensions(struct sk_buff *skb) 4825 { 4826 return unlikely(skb->active_extensions); 4827 } 4828 #else 4829 static inline void skb_ext_put(struct sk_buff *skb) {} 4830 static inline void skb_ext_reset(struct sk_buff *skb) {} 4831 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4832 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4833 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4834 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4835 #endif /* CONFIG_SKB_EXTENSIONS */ 4836 4837 static inline void nf_reset_ct(struct sk_buff *skb) 4838 { 4839 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4840 nf_conntrack_put(skb_nfct(skb)); 4841 skb->_nfct = 0; 4842 #endif 4843 } 4844 4845 static inline void nf_reset_trace(struct sk_buff *skb) 4846 { 4847 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) 4848 skb->nf_trace = 0; 4849 #endif 4850 } 4851 4852 static inline void ipvs_reset(struct sk_buff *skb) 4853 { 4854 #if IS_ENABLED(CONFIG_IP_VS) 4855 skb->ipvs_property = 0; 4856 #endif 4857 } 4858 4859 /* Note: This doesn't put any conntrack info in dst. */ 4860 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4861 bool copy) 4862 { 4863 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4864 dst->_nfct = src->_nfct; 4865 nf_conntrack_get(skb_nfct(src)); 4866 #endif 4867 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) 4868 if (copy) 4869 dst->nf_trace = src->nf_trace; 4870 #endif 4871 } 4872 4873 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4874 { 4875 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4876 nf_conntrack_put(skb_nfct(dst)); 4877 #endif 4878 dst->slow_gro = src->slow_gro; 4879 __nf_copy(dst, src, true); 4880 } 4881 4882 #ifdef CONFIG_NETWORK_SECMARK 4883 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4884 { 4885 to->secmark = from->secmark; 4886 } 4887 4888 static inline void skb_init_secmark(struct sk_buff *skb) 4889 { 4890 skb->secmark = 0; 4891 } 4892 #else 4893 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4894 { } 4895 4896 static inline void skb_init_secmark(struct sk_buff *skb) 4897 { } 4898 #endif 4899 4900 static inline int secpath_exists(const struct sk_buff *skb) 4901 { 4902 #ifdef CONFIG_XFRM 4903 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4904 #else 4905 return 0; 4906 #endif 4907 } 4908 4909 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4910 { 4911 return !skb->destructor && 4912 !secpath_exists(skb) && 4913 !skb_nfct(skb) && 4914 !skb->_skb_refdst && 4915 !skb_has_frag_list(skb); 4916 } 4917 4918 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4919 { 4920 skb->queue_mapping = queue_mapping; 4921 } 4922 4923 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4924 { 4925 return skb->queue_mapping; 4926 } 4927 4928 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4929 { 4930 to->queue_mapping = from->queue_mapping; 4931 } 4932 4933 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4934 { 4935 skb->queue_mapping = rx_queue + 1; 4936 } 4937 4938 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4939 { 4940 return skb->queue_mapping - 1; 4941 } 4942 4943 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4944 { 4945 return skb->queue_mapping != 0; 4946 } 4947 4948 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4949 { 4950 skb->dst_pending_confirm = val; 4951 } 4952 4953 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4954 { 4955 return skb->dst_pending_confirm != 0; 4956 } 4957 4958 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4959 { 4960 #ifdef CONFIG_XFRM 4961 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4962 #else 4963 return NULL; 4964 #endif 4965 } 4966 4967 static inline bool skb_is_gso(const struct sk_buff *skb) 4968 { 4969 return skb_shinfo(skb)->gso_size; 4970 } 4971 4972 /* Note: Should be called only if skb_is_gso(skb) is true */ 4973 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4974 { 4975 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4976 } 4977 4978 /* Note: Should be called only if skb_is_gso(skb) is true */ 4979 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4980 { 4981 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4982 } 4983 4984 /* Note: Should be called only if skb_is_gso(skb) is true */ 4985 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4986 { 4987 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4988 } 4989 4990 static inline void skb_gso_reset(struct sk_buff *skb) 4991 { 4992 skb_shinfo(skb)->gso_size = 0; 4993 skb_shinfo(skb)->gso_segs = 0; 4994 skb_shinfo(skb)->gso_type = 0; 4995 } 4996 4997 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4998 u16 increment) 4999 { 5000 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 5001 return; 5002 shinfo->gso_size += increment; 5003 } 5004 5005 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 5006 u16 decrement) 5007 { 5008 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 5009 return; 5010 shinfo->gso_size -= decrement; 5011 } 5012 5013 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 5014 5015 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 5016 { 5017 /* LRO sets gso_size but not gso_type, whereas if GSO is really 5018 * wanted then gso_type will be set. */ 5019 const struct skb_shared_info *shinfo = skb_shinfo(skb); 5020 5021 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 5022 unlikely(shinfo->gso_type == 0)) { 5023 __skb_warn_lro_forwarding(skb); 5024 return true; 5025 } 5026 return false; 5027 } 5028 5029 static inline void skb_forward_csum(struct sk_buff *skb) 5030 { 5031 /* Unfortunately we don't support this one. Any brave souls? */ 5032 if (skb->ip_summed == CHECKSUM_COMPLETE) 5033 skb->ip_summed = CHECKSUM_NONE; 5034 } 5035 5036 /** 5037 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 5038 * @skb: skb to check 5039 * 5040 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 5041 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 5042 * use this helper, to document places where we make this assertion. 5043 */ 5044 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 5045 { 5046 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE); 5047 } 5048 5049 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 5050 5051 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 5052 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 5053 unsigned int transport_len, 5054 __sum16(*skb_chkf)(struct sk_buff *skb)); 5055 5056 /** 5057 * skb_head_is_locked - Determine if the skb->head is locked down 5058 * @skb: skb to check 5059 * 5060 * The head on skbs build around a head frag can be removed if they are 5061 * not cloned. This function returns true if the skb head is locked down 5062 * due to either being allocated via kmalloc, or by being a clone with 5063 * multiple references to the head. 5064 */ 5065 static inline bool skb_head_is_locked(const struct sk_buff *skb) 5066 { 5067 return !skb->head_frag || skb_cloned(skb); 5068 } 5069 5070 /* Local Checksum Offload. 5071 * Compute outer checksum based on the assumption that the 5072 * inner checksum will be offloaded later. 5073 * See Documentation/networking/checksum-offloads.rst for 5074 * explanation of how this works. 5075 * Fill in outer checksum adjustment (e.g. with sum of outer 5076 * pseudo-header) before calling. 5077 * Also ensure that inner checksum is in linear data area. 5078 */ 5079 static inline __wsum lco_csum(struct sk_buff *skb) 5080 { 5081 unsigned char *csum_start = skb_checksum_start(skb); 5082 unsigned char *l4_hdr = skb_transport_header(skb); 5083 __wsum partial; 5084 5085 /* Start with complement of inner checksum adjustment */ 5086 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 5087 skb->csum_offset)); 5088 5089 /* Add in checksum of our headers (incl. outer checksum 5090 * adjustment filled in by caller) and return result. 5091 */ 5092 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 5093 } 5094 5095 static inline bool skb_is_redirected(const struct sk_buff *skb) 5096 { 5097 return skb->redirected; 5098 } 5099 5100 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 5101 { 5102 skb->redirected = 1; 5103 #ifdef CONFIG_NET_REDIRECT 5104 skb->from_ingress = from_ingress; 5105 if (skb->from_ingress) 5106 skb_clear_tstamp(skb); 5107 #endif 5108 } 5109 5110 static inline void skb_reset_redirect(struct sk_buff *skb) 5111 { 5112 skb->redirected = 0; 5113 } 5114 5115 static inline void skb_set_redirected_noclear(struct sk_buff *skb, 5116 bool from_ingress) 5117 { 5118 skb->redirected = 1; 5119 #ifdef CONFIG_NET_REDIRECT 5120 skb->from_ingress = from_ingress; 5121 #endif 5122 } 5123 5124 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 5125 { 5126 #if IS_ENABLED(CONFIG_IP_SCTP) 5127 return skb->csum_not_inet; 5128 #else 5129 return 0; 5130 #endif 5131 } 5132 5133 static inline void skb_reset_csum_not_inet(struct sk_buff *skb) 5134 { 5135 skb->ip_summed = CHECKSUM_NONE; 5136 #if IS_ENABLED(CONFIG_IP_SCTP) 5137 skb->csum_not_inet = 0; 5138 #endif 5139 } 5140 5141 static inline void skb_set_kcov_handle(struct sk_buff *skb, 5142 const u64 kcov_handle) 5143 { 5144 #ifdef CONFIG_KCOV 5145 skb->kcov_handle = kcov_handle; 5146 #endif 5147 } 5148 5149 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 5150 { 5151 #ifdef CONFIG_KCOV 5152 return skb->kcov_handle; 5153 #else 5154 return 0; 5155 #endif 5156 } 5157 5158 static inline void skb_mark_for_recycle(struct sk_buff *skb) 5159 { 5160 #ifdef CONFIG_PAGE_POOL 5161 skb->pp_recycle = 1; 5162 #endif 5163 } 5164 5165 ssize_t skb_splice_from_iter(struct sk_buff *skb, struct iov_iter *iter, 5166 ssize_t maxsize, gfp_t gfp); 5167 5168 #endif /* __KERNEL__ */ 5169 #endif /* _LINUX_SKBUFF_H */ 5170