1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Definitions for the 'struct sk_buff' memory handlers. 4 * 5 * Authors: 6 * Alan Cox, <gw4pts@gw4pts.ampr.org> 7 * Florian La Roche, <rzsfl@rz.uni-sb.de> 8 */ 9 10 #ifndef _LINUX_SKBUFF_H 11 #define _LINUX_SKBUFF_H 12 13 #include <linux/kernel.h> 14 #include <linux/compiler.h> 15 #include <linux/time.h> 16 #include <linux/bug.h> 17 #include <linux/bvec.h> 18 #include <linux/cache.h> 19 #include <linux/rbtree.h> 20 #include <linux/socket.h> 21 #include <linux/refcount.h> 22 23 #include <linux/atomic.h> 24 #include <asm/types.h> 25 #include <linux/spinlock.h> 26 #include <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 #include <net/page_pool.h> 36 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 37 #include <linux/netfilter/nf_conntrack_common.h> 38 #endif 39 #include <net/net_debug.h> 40 #include <net/dropreason-core.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 struct net_device *physindev; 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 }; 334 #endif 335 336 struct sk_buff_head { 337 /* These two members must be first to match sk_buff. */ 338 struct_group_tagged(sk_buff_list, list, 339 struct sk_buff *next; 340 struct sk_buff *prev; 341 ); 342 343 __u32 qlen; 344 spinlock_t lock; 345 }; 346 347 struct sk_buff; 348 349 #ifndef CONFIG_MAX_SKB_FRAGS 350 # define CONFIG_MAX_SKB_FRAGS 17 351 #endif 352 353 #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS 354 355 extern int sysctl_max_skb_frags; 356 357 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to 358 * segment using its current segmentation instead. 359 */ 360 #define GSO_BY_FRAGS 0xFFFF 361 362 typedef struct bio_vec skb_frag_t; 363 364 /** 365 * skb_frag_size() - Returns the size of a skb fragment 366 * @frag: skb fragment 367 */ 368 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 369 { 370 return frag->bv_len; 371 } 372 373 /** 374 * skb_frag_size_set() - Sets the size of a skb fragment 375 * @frag: skb fragment 376 * @size: size of fragment 377 */ 378 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 379 { 380 frag->bv_len = size; 381 } 382 383 /** 384 * skb_frag_size_add() - Increments the size of a skb fragment by @delta 385 * @frag: skb fragment 386 * @delta: value to add 387 */ 388 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 389 { 390 frag->bv_len += delta; 391 } 392 393 /** 394 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta 395 * @frag: skb fragment 396 * @delta: value to subtract 397 */ 398 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 399 { 400 frag->bv_len -= delta; 401 } 402 403 /** 404 * skb_frag_must_loop - Test if %p is a high memory page 405 * @p: fragment's page 406 */ 407 static inline bool skb_frag_must_loop(struct page *p) 408 { 409 #if defined(CONFIG_HIGHMEM) 410 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) 411 return true; 412 #endif 413 return false; 414 } 415 416 /** 417 * skb_frag_foreach_page - loop over pages in a fragment 418 * 419 * @f: skb frag to operate on 420 * @f_off: offset from start of f->bv_page 421 * @f_len: length from f_off to loop over 422 * @p: (temp var) current page 423 * @p_off: (temp var) offset from start of current page, 424 * non-zero only on first page. 425 * @p_len: (temp var) length in current page, 426 * < PAGE_SIZE only on first and last page. 427 * @copied: (temp var) length so far, excluding current p_len. 428 * 429 * A fragment can hold a compound page, in which case per-page 430 * operations, notably kmap_atomic, must be called for each 431 * regular page. 432 */ 433 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ 434 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ 435 p_off = (f_off) & (PAGE_SIZE - 1), \ 436 p_len = skb_frag_must_loop(p) ? \ 437 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ 438 copied = 0; \ 439 copied < f_len; \ 440 copied += p_len, p++, p_off = 0, \ 441 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ 442 443 #define HAVE_HW_TIME_STAMP 444 445 /** 446 * struct skb_shared_hwtstamps - hardware time stamps 447 * @hwtstamp: hardware time stamp transformed into duration 448 * since arbitrary point in time 449 * @netdev_data: address/cookie of network device driver used as 450 * reference to actual hardware time stamp 451 * 452 * Software time stamps generated by ktime_get_real() are stored in 453 * skb->tstamp. 454 * 455 * hwtstamps can only be compared against other hwtstamps from 456 * the same device. 457 * 458 * This structure is attached to packets as part of the 459 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 460 */ 461 struct skb_shared_hwtstamps { 462 union { 463 ktime_t hwtstamp; 464 void *netdev_data; 465 }; 466 }; 467 468 /* Definitions for tx_flags in struct skb_shared_info */ 469 enum { 470 /* generate hardware time stamp */ 471 SKBTX_HW_TSTAMP = 1 << 0, 472 473 /* generate software time stamp when queueing packet to NIC */ 474 SKBTX_SW_TSTAMP = 1 << 1, 475 476 /* device driver is going to provide hardware time stamp */ 477 SKBTX_IN_PROGRESS = 1 << 2, 478 479 /* generate hardware time stamp based on cycles if supported */ 480 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3, 481 482 /* generate wifi status information (where possible) */ 483 SKBTX_WIFI_STATUS = 1 << 4, 484 485 /* determine hardware time stamp based on time or cycles */ 486 SKBTX_HW_TSTAMP_NETDEV = 1 << 5, 487 488 /* generate software time stamp when entering packet scheduling */ 489 SKBTX_SCHED_TSTAMP = 1 << 6, 490 }; 491 492 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 493 SKBTX_SCHED_TSTAMP) 494 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \ 495 SKBTX_HW_TSTAMP_USE_CYCLES | \ 496 SKBTX_ANY_SW_TSTAMP) 497 498 /* Definitions for flags in struct skb_shared_info */ 499 enum { 500 /* use zcopy routines */ 501 SKBFL_ZEROCOPY_ENABLE = BIT(0), 502 503 /* This indicates at least one fragment might be overwritten 504 * (as in vmsplice(), sendfile() ...) 505 * If we need to compute a TX checksum, we'll need to copy 506 * all frags to avoid possible bad checksum 507 */ 508 SKBFL_SHARED_FRAG = BIT(1), 509 510 /* segment contains only zerocopy data and should not be 511 * charged to the kernel memory. 512 */ 513 SKBFL_PURE_ZEROCOPY = BIT(2), 514 515 SKBFL_DONT_ORPHAN = BIT(3), 516 517 /* page references are managed by the ubuf_info, so it's safe to 518 * use frags only up until ubuf_info is released 519 */ 520 SKBFL_MANAGED_FRAG_REFS = BIT(4), 521 }; 522 523 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) 524 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \ 525 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS) 526 527 /* 528 * The callback notifies userspace to release buffers when skb DMA is done in 529 * lower device, the skb last reference should be 0 when calling this. 530 * The zerocopy_success argument is true if zero copy transmit occurred, 531 * false on data copy or out of memory error caused by data copy attempt. 532 * The ctx field is used to track device context. 533 * The desc field is used to track userspace buffer index. 534 */ 535 struct ubuf_info { 536 void (*callback)(struct sk_buff *, struct ubuf_info *, 537 bool zerocopy_success); 538 refcount_t refcnt; 539 u8 flags; 540 }; 541 542 struct ubuf_info_msgzc { 543 struct ubuf_info ubuf; 544 545 union { 546 struct { 547 unsigned long desc; 548 void *ctx; 549 }; 550 struct { 551 u32 id; 552 u16 len; 553 u16 zerocopy:1; 554 u32 bytelen; 555 }; 556 }; 557 558 struct mmpin { 559 struct user_struct *user; 560 unsigned int num_pg; 561 } mmp; 562 }; 563 564 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) 565 #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \ 566 ubuf) 567 568 int mm_account_pinned_pages(struct mmpin *mmp, size_t size); 569 void mm_unaccount_pinned_pages(struct mmpin *mmp); 570 571 /* This data is invariant across clones and lives at 572 * the end of the header data, ie. at skb->end. 573 */ 574 struct skb_shared_info { 575 __u8 flags; 576 __u8 meta_len; 577 __u8 nr_frags; 578 __u8 tx_flags; 579 unsigned short gso_size; 580 /* Warning: this field is not always filled in (UFO)! */ 581 unsigned short gso_segs; 582 struct sk_buff *frag_list; 583 struct skb_shared_hwtstamps hwtstamps; 584 unsigned int gso_type; 585 u32 tskey; 586 587 /* 588 * Warning : all fields before dataref are cleared in __alloc_skb() 589 */ 590 atomic_t dataref; 591 unsigned int xdp_frags_size; 592 593 /* Intermediate layers must ensure that destructor_arg 594 * remains valid until skb destructor */ 595 void * destructor_arg; 596 597 /* must be last field, see pskb_expand_head() */ 598 skb_frag_t frags[MAX_SKB_FRAGS]; 599 }; 600 601 /** 602 * DOC: dataref and headerless skbs 603 * 604 * Transport layers send out clones of payload skbs they hold for 605 * retransmissions. To allow lower layers of the stack to prepend their headers 606 * we split &skb_shared_info.dataref into two halves. 607 * The lower 16 bits count the overall number of references. 608 * The higher 16 bits indicate how many of the references are payload-only. 609 * skb_header_cloned() checks if skb is allowed to add / write the headers. 610 * 611 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr 612 * (via __skb_header_release()). Any clone created from marked skb will get 613 * &sk_buff.hdr_len populated with the available headroom. 614 * If there's the only clone in existence it's able to modify the headroom 615 * at will. The sequence of calls inside the transport layer is:: 616 * 617 * <alloc skb> 618 * skb_reserve() 619 * __skb_header_release() 620 * skb_clone() 621 * // send the clone down the stack 622 * 623 * This is not a very generic construct and it depends on the transport layers 624 * doing the right thing. In practice there's usually only one payload-only skb. 625 * Having multiple payload-only skbs with different lengths of hdr_len is not 626 * possible. The payload-only skbs should never leave their owner. 627 */ 628 #define SKB_DATAREF_SHIFT 16 629 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 630 631 632 enum { 633 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 634 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 635 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 636 }; 637 638 enum { 639 SKB_GSO_TCPV4 = 1 << 0, 640 641 /* This indicates the skb is from an untrusted source. */ 642 SKB_GSO_DODGY = 1 << 1, 643 644 /* This indicates the tcp segment has CWR set. */ 645 SKB_GSO_TCP_ECN = 1 << 2, 646 647 SKB_GSO_TCP_FIXEDID = 1 << 3, 648 649 SKB_GSO_TCPV6 = 1 << 4, 650 651 SKB_GSO_FCOE = 1 << 5, 652 653 SKB_GSO_GRE = 1 << 6, 654 655 SKB_GSO_GRE_CSUM = 1 << 7, 656 657 SKB_GSO_IPXIP4 = 1 << 8, 658 659 SKB_GSO_IPXIP6 = 1 << 9, 660 661 SKB_GSO_UDP_TUNNEL = 1 << 10, 662 663 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 664 665 SKB_GSO_PARTIAL = 1 << 12, 666 667 SKB_GSO_TUNNEL_REMCSUM = 1 << 13, 668 669 SKB_GSO_SCTP = 1 << 14, 670 671 SKB_GSO_ESP = 1 << 15, 672 673 SKB_GSO_UDP = 1 << 16, 674 675 SKB_GSO_UDP_L4 = 1 << 17, 676 677 SKB_GSO_FRAGLIST = 1 << 18, 678 }; 679 680 #if BITS_PER_LONG > 32 681 #define NET_SKBUFF_DATA_USES_OFFSET 1 682 #endif 683 684 #ifdef NET_SKBUFF_DATA_USES_OFFSET 685 typedef unsigned int sk_buff_data_t; 686 #else 687 typedef unsigned char *sk_buff_data_t; 688 #endif 689 690 /** 691 * DOC: Basic sk_buff geometry 692 * 693 * struct sk_buff itself is a metadata structure and does not hold any packet 694 * data. All the data is held in associated buffers. 695 * 696 * &sk_buff.head points to the main "head" buffer. The head buffer is divided 697 * into two parts: 698 * 699 * - data buffer, containing headers and sometimes payload; 700 * this is the part of the skb operated on by the common helpers 701 * such as skb_put() or skb_pull(); 702 * - shared info (struct skb_shared_info) which holds an array of pointers 703 * to read-only data in the (page, offset, length) format. 704 * 705 * Optionally &skb_shared_info.frag_list may point to another skb. 706 * 707 * Basic diagram may look like this:: 708 * 709 * --------------- 710 * | sk_buff | 711 * --------------- 712 * ,--------------------------- + head 713 * / ,----------------- + data 714 * / / ,----------- + tail 715 * | | | , + end 716 * | | | | 717 * v v v v 718 * ----------------------------------------------- 719 * | headroom | data | tailroom | skb_shared_info | 720 * ----------------------------------------------- 721 * + [page frag] 722 * + [page frag] 723 * + [page frag] 724 * + [page frag] --------- 725 * + frag_list --> | sk_buff | 726 * --------- 727 * 728 */ 729 730 /** 731 * struct sk_buff - socket buffer 732 * @next: Next buffer in list 733 * @prev: Previous buffer in list 734 * @tstamp: Time we arrived/left 735 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point 736 * for retransmit timer 737 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 738 * @list: queue head 739 * @ll_node: anchor in an llist (eg socket defer_list) 740 * @sk: Socket we are owned by 741 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in 742 * fragmentation management 743 * @dev: Device we arrived on/are leaving by 744 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL 745 * @cb: Control buffer. Free for use by every layer. Put private vars here 746 * @_skb_refdst: destination entry (with norefcount bit) 747 * @sp: the security path, used for xfrm 748 * @len: Length of actual data 749 * @data_len: Data length 750 * @mac_len: Length of link layer header 751 * @hdr_len: writable header length of cloned skb 752 * @csum: Checksum (must include start/offset pair) 753 * @csum_start: Offset from skb->head where checksumming should start 754 * @csum_offset: Offset from csum_start where checksum should be stored 755 * @priority: Packet queueing priority 756 * @ignore_df: allow local fragmentation 757 * @cloned: Head may be cloned (check refcnt to be sure) 758 * @ip_summed: Driver fed us an IP checksum 759 * @nohdr: Payload reference only, must not modify header 760 * @pkt_type: Packet class 761 * @fclone: skbuff clone status 762 * @ipvs_property: skbuff is owned by ipvs 763 * @inner_protocol_type: whether the inner protocol is 764 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO 765 * @remcsum_offload: remote checksum offload is enabled 766 * @offload_fwd_mark: Packet was L2-forwarded in hardware 767 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware 768 * @tc_skip_classify: do not classify packet. set by IFB device 769 * @tc_at_ingress: used within tc_classify to distinguish in/egress 770 * @redirected: packet was redirected by packet classifier 771 * @from_ingress: packet was redirected from the ingress path 772 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h 773 * @peeked: this packet has been seen already, so stats have been 774 * done for it, don't do them again 775 * @nf_trace: netfilter packet trace flag 776 * @protocol: Packet protocol from driver 777 * @destructor: Destruct function 778 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) 779 * @_sk_redir: socket redirection information for skmsg 780 * @_nfct: Associated connection, if any (with nfctinfo bits) 781 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 782 * @skb_iif: ifindex of device we arrived on 783 * @tc_index: Traffic control index 784 * @hash: the packet hash 785 * @queue_mapping: Queue mapping for multiqueue devices 786 * @head_frag: skb was allocated from page fragments, 787 * not allocated by kmalloc() or vmalloc(). 788 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves 789 * @pp_recycle: mark the packet for recycling instead of freeing (implies 790 * page_pool support on driver) 791 * @active_extensions: active extensions (skb_ext_id types) 792 * @ndisc_nodetype: router type (from link layer) 793 * @ooo_okay: allow the mapping of a socket to a queue to be changed 794 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 795 * ports. 796 * @sw_hash: indicates hash was computed in software stack 797 * @wifi_acked_valid: wifi_acked was set 798 * @wifi_acked: whether frame was acked on wifi or not 799 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 800 * @encapsulation: indicates the inner headers in the skbuff are valid 801 * @encap_hdr_csum: software checksum is needed 802 * @csum_valid: checksum is already valid 803 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL 804 * @csum_complete_sw: checksum was completed by software 805 * @csum_level: indicates the number of consecutive checksums found in 806 * the packet minus one that have been verified as 807 * CHECKSUM_UNNECESSARY (max 3) 808 * @dst_pending_confirm: need to confirm neighbour 809 * @decrypted: Decrypted SKB 810 * @slow_gro: state present at GRO time, slower prepare step required 811 * @mono_delivery_time: When set, skb->tstamp has the 812 * delivery_time in mono clock base (i.e. EDT). Otherwise, the 813 * skb->tstamp has the (rcv) timestamp at ingress and 814 * delivery_time at egress. 815 * @napi_id: id of the NAPI struct this skb came from 816 * @sender_cpu: (aka @napi_id) source CPU in XPS 817 * @alloc_cpu: CPU which did the skb allocation. 818 * @secmark: security marking 819 * @mark: Generic packet mark 820 * @reserved_tailroom: (aka @mark) number of bytes of free space available 821 * at the tail of an sk_buff 822 * @vlan_all: vlan fields (proto & tci) 823 * @vlan_proto: vlan encapsulation protocol 824 * @vlan_tci: vlan tag control information 825 * @inner_protocol: Protocol (encapsulation) 826 * @inner_ipproto: (aka @inner_protocol) stores ipproto when 827 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; 828 * @inner_transport_header: Inner transport layer header (encapsulation) 829 * @inner_network_header: Network layer header (encapsulation) 830 * @inner_mac_header: Link layer header (encapsulation) 831 * @transport_header: Transport layer header 832 * @network_header: Network layer header 833 * @mac_header: Link layer header 834 * @kcov_handle: KCOV remote handle for remote coverage collection 835 * @tail: Tail pointer 836 * @end: End pointer 837 * @head: Head of buffer 838 * @data: Data head pointer 839 * @truesize: Buffer size 840 * @users: User count - see {datagram,tcp}.c 841 * @extensions: allocated extensions, valid if active_extensions is nonzero 842 */ 843 844 struct sk_buff { 845 union { 846 struct { 847 /* These two members must be first to match sk_buff_head. */ 848 struct sk_buff *next; 849 struct sk_buff *prev; 850 851 union { 852 struct net_device *dev; 853 /* Some protocols might use this space to store information, 854 * while device pointer would be NULL. 855 * UDP receive path is one user. 856 */ 857 unsigned long dev_scratch; 858 }; 859 }; 860 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ 861 struct list_head list; 862 struct llist_node ll_node; 863 }; 864 865 union { 866 struct sock *sk; 867 int ip_defrag_offset; 868 }; 869 870 union { 871 ktime_t tstamp; 872 u64 skb_mstamp_ns; /* earliest departure time */ 873 }; 874 /* 875 * This is the control buffer. It is free to use for every 876 * layer. Please put your private variables there. If you 877 * want to keep them across layers you have to do a skb_clone() 878 * first. This is owned by whoever has the skb queued ATM. 879 */ 880 char cb[48] __aligned(8); 881 882 union { 883 struct { 884 unsigned long _skb_refdst; 885 void (*destructor)(struct sk_buff *skb); 886 }; 887 struct list_head tcp_tsorted_anchor; 888 #ifdef CONFIG_NET_SOCK_MSG 889 unsigned long _sk_redir; 890 #endif 891 }; 892 893 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 894 unsigned long _nfct; 895 #endif 896 unsigned int len, 897 data_len; 898 __u16 mac_len, 899 hdr_len; 900 901 /* Following fields are _not_ copied in __copy_skb_header() 902 * Note that queue_mapping is here mostly to fill a hole. 903 */ 904 __u16 queue_mapping; 905 906 /* if you move cloned around you also must adapt those constants */ 907 #ifdef __BIG_ENDIAN_BITFIELD 908 #define CLONED_MASK (1 << 7) 909 #else 910 #define CLONED_MASK 1 911 #endif 912 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset) 913 914 /* private: */ 915 __u8 __cloned_offset[0]; 916 /* public: */ 917 __u8 cloned:1, 918 nohdr:1, 919 fclone:2, 920 peeked:1, 921 head_frag:1, 922 pfmemalloc:1, 923 pp_recycle:1; /* page_pool recycle indicator */ 924 #ifdef CONFIG_SKB_EXTENSIONS 925 __u8 active_extensions; 926 #endif 927 928 /* Fields enclosed in headers group are copied 929 * using a single memcpy() in __copy_skb_header() 930 */ 931 struct_group(headers, 932 933 /* private: */ 934 __u8 __pkt_type_offset[0]; 935 /* public: */ 936 __u8 pkt_type:3; /* see PKT_TYPE_MAX */ 937 __u8 ignore_df:1; 938 __u8 dst_pending_confirm:1; 939 __u8 ip_summed:2; 940 __u8 ooo_okay:1; 941 942 /* private: */ 943 __u8 __mono_tc_offset[0]; 944 /* public: */ 945 __u8 mono_delivery_time:1; /* See SKB_MONO_DELIVERY_TIME_MASK */ 946 #ifdef CONFIG_NET_CLS_ACT 947 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */ 948 __u8 tc_skip_classify:1; 949 #endif 950 __u8 remcsum_offload:1; 951 __u8 csum_complete_sw:1; 952 __u8 csum_level:2; 953 __u8 inner_protocol_type:1; 954 955 __u8 l4_hash:1; 956 __u8 sw_hash:1; 957 #ifdef CONFIG_WIRELESS 958 __u8 wifi_acked_valid:1; 959 __u8 wifi_acked:1; 960 #endif 961 __u8 no_fcs:1; 962 /* Indicates the inner headers are valid in the skbuff. */ 963 __u8 encapsulation:1; 964 __u8 encap_hdr_csum:1; 965 __u8 csum_valid:1; 966 #ifdef CONFIG_IPV6_NDISC_NODETYPE 967 __u8 ndisc_nodetype:2; 968 #endif 969 970 #if IS_ENABLED(CONFIG_IP_VS) 971 __u8 ipvs_property:1; 972 #endif 973 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) 974 __u8 nf_trace:1; 975 #endif 976 #ifdef CONFIG_NET_SWITCHDEV 977 __u8 offload_fwd_mark:1; 978 __u8 offload_l3_fwd_mark:1; 979 #endif 980 __u8 redirected:1; 981 #ifdef CONFIG_NET_REDIRECT 982 __u8 from_ingress:1; 983 #endif 984 #ifdef CONFIG_NETFILTER_SKIP_EGRESS 985 __u8 nf_skip_egress:1; 986 #endif 987 #ifdef CONFIG_TLS_DEVICE 988 __u8 decrypted:1; 989 #endif 990 __u8 slow_gro:1; 991 #if IS_ENABLED(CONFIG_IP_SCTP) 992 __u8 csum_not_inet:1; 993 #endif 994 995 #ifdef CONFIG_NET_SCHED 996 __u16 tc_index; /* traffic control index */ 997 #endif 998 999 u16 alloc_cpu; 1000 1001 union { 1002 __wsum csum; 1003 struct { 1004 __u16 csum_start; 1005 __u16 csum_offset; 1006 }; 1007 }; 1008 __u32 priority; 1009 int skb_iif; 1010 __u32 hash; 1011 union { 1012 u32 vlan_all; 1013 struct { 1014 __be16 vlan_proto; 1015 __u16 vlan_tci; 1016 }; 1017 }; 1018 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 1019 union { 1020 unsigned int napi_id; 1021 unsigned int sender_cpu; 1022 }; 1023 #endif 1024 #ifdef CONFIG_NETWORK_SECMARK 1025 __u32 secmark; 1026 #endif 1027 1028 union { 1029 __u32 mark; 1030 __u32 reserved_tailroom; 1031 }; 1032 1033 union { 1034 __be16 inner_protocol; 1035 __u8 inner_ipproto; 1036 }; 1037 1038 __u16 inner_transport_header; 1039 __u16 inner_network_header; 1040 __u16 inner_mac_header; 1041 1042 __be16 protocol; 1043 __u16 transport_header; 1044 __u16 network_header; 1045 __u16 mac_header; 1046 1047 #ifdef CONFIG_KCOV 1048 u64 kcov_handle; 1049 #endif 1050 1051 ); /* end headers group */ 1052 1053 /* These elements must be at the end, see alloc_skb() for details. */ 1054 sk_buff_data_t tail; 1055 sk_buff_data_t end; 1056 unsigned char *head, 1057 *data; 1058 unsigned int truesize; 1059 refcount_t users; 1060 1061 #ifdef CONFIG_SKB_EXTENSIONS 1062 /* only useable after checking ->active_extensions != 0 */ 1063 struct skb_ext *extensions; 1064 #endif 1065 }; 1066 1067 /* if you move pkt_type around you also must adapt those constants */ 1068 #ifdef __BIG_ENDIAN_BITFIELD 1069 #define PKT_TYPE_MAX (7 << 5) 1070 #else 1071 #define PKT_TYPE_MAX 7 1072 #endif 1073 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset) 1074 1075 /* if you move tc_at_ingress or mono_delivery_time 1076 * around, you also must adapt these constants. 1077 */ 1078 #ifdef __BIG_ENDIAN_BITFIELD 1079 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 7) 1080 #define TC_AT_INGRESS_MASK (1 << 6) 1081 #else 1082 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 0) 1083 #define TC_AT_INGRESS_MASK (1 << 1) 1084 #endif 1085 #define SKB_BF_MONO_TC_OFFSET offsetof(struct sk_buff, __mono_tc_offset) 1086 1087 #ifdef __KERNEL__ 1088 /* 1089 * Handling routines are only of interest to the kernel 1090 */ 1091 1092 #define SKB_ALLOC_FCLONE 0x01 1093 #define SKB_ALLOC_RX 0x02 1094 #define SKB_ALLOC_NAPI 0x04 1095 1096 /** 1097 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves 1098 * @skb: buffer 1099 */ 1100 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 1101 { 1102 return unlikely(skb->pfmemalloc); 1103 } 1104 1105 /* 1106 * skb might have a dst pointer attached, refcounted or not. 1107 * _skb_refdst low order bit is set if refcount was _not_ taken 1108 */ 1109 #define SKB_DST_NOREF 1UL 1110 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 1111 1112 /** 1113 * skb_dst - returns skb dst_entry 1114 * @skb: buffer 1115 * 1116 * Returns skb dst_entry, regardless of reference taken or not. 1117 */ 1118 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 1119 { 1120 /* If refdst was not refcounted, check we still are in a 1121 * rcu_read_lock section 1122 */ 1123 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 1124 !rcu_read_lock_held() && 1125 !rcu_read_lock_bh_held()); 1126 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 1127 } 1128 1129 /** 1130 * skb_dst_set - sets skb dst 1131 * @skb: buffer 1132 * @dst: dst entry 1133 * 1134 * Sets skb dst, assuming a reference was taken on dst and should 1135 * be released by skb_dst_drop() 1136 */ 1137 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 1138 { 1139 skb->slow_gro |= !!dst; 1140 skb->_skb_refdst = (unsigned long)dst; 1141 } 1142 1143 /** 1144 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 1145 * @skb: buffer 1146 * @dst: dst entry 1147 * 1148 * Sets skb dst, assuming a reference was not taken on dst. 1149 * If dst entry is cached, we do not take reference and dst_release 1150 * will be avoided by refdst_drop. If dst entry is not cached, we take 1151 * reference, so that last dst_release can destroy the dst immediately. 1152 */ 1153 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 1154 { 1155 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 1156 skb->slow_gro |= !!dst; 1157 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 1158 } 1159 1160 /** 1161 * skb_dst_is_noref - Test if skb dst isn't refcounted 1162 * @skb: buffer 1163 */ 1164 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 1165 { 1166 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 1167 } 1168 1169 /** 1170 * skb_rtable - Returns the skb &rtable 1171 * @skb: buffer 1172 */ 1173 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 1174 { 1175 return (struct rtable *)skb_dst(skb); 1176 } 1177 1178 /* For mangling skb->pkt_type from user space side from applications 1179 * such as nft, tc, etc, we only allow a conservative subset of 1180 * possible pkt_types to be set. 1181 */ 1182 static inline bool skb_pkt_type_ok(u32 ptype) 1183 { 1184 return ptype <= PACKET_OTHERHOST; 1185 } 1186 1187 /** 1188 * skb_napi_id - Returns the skb's NAPI id 1189 * @skb: buffer 1190 */ 1191 static inline unsigned int skb_napi_id(const struct sk_buff *skb) 1192 { 1193 #ifdef CONFIG_NET_RX_BUSY_POLL 1194 return skb->napi_id; 1195 #else 1196 return 0; 1197 #endif 1198 } 1199 1200 static inline bool skb_wifi_acked_valid(const struct sk_buff *skb) 1201 { 1202 #ifdef CONFIG_WIRELESS 1203 return skb->wifi_acked_valid; 1204 #else 1205 return 0; 1206 #endif 1207 } 1208 1209 /** 1210 * skb_unref - decrement the skb's reference count 1211 * @skb: buffer 1212 * 1213 * Returns true if we can free the skb. 1214 */ 1215 static inline bool skb_unref(struct sk_buff *skb) 1216 { 1217 if (unlikely(!skb)) 1218 return false; 1219 if (likely(refcount_read(&skb->users) == 1)) 1220 smp_rmb(); 1221 else if (likely(!refcount_dec_and_test(&skb->users))) 1222 return false; 1223 1224 return true; 1225 } 1226 1227 void __fix_address 1228 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason); 1229 1230 /** 1231 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason 1232 * @skb: buffer to free 1233 */ 1234 static inline void kfree_skb(struct sk_buff *skb) 1235 { 1236 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 1237 } 1238 1239 void skb_release_head_state(struct sk_buff *skb); 1240 void kfree_skb_list_reason(struct sk_buff *segs, 1241 enum skb_drop_reason reason); 1242 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); 1243 void skb_tx_error(struct sk_buff *skb); 1244 1245 static inline void kfree_skb_list(struct sk_buff *segs) 1246 { 1247 kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED); 1248 } 1249 1250 #ifdef CONFIG_TRACEPOINTS 1251 void consume_skb(struct sk_buff *skb); 1252 #else 1253 static inline void consume_skb(struct sk_buff *skb) 1254 { 1255 return kfree_skb(skb); 1256 } 1257 #endif 1258 1259 void __consume_stateless_skb(struct sk_buff *skb); 1260 void __kfree_skb(struct sk_buff *skb); 1261 extern struct kmem_cache *skbuff_cache; 1262 1263 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 1264 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 1265 bool *fragstolen, int *delta_truesize); 1266 1267 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 1268 int node); 1269 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 1270 struct sk_buff *build_skb(void *data, unsigned int frag_size); 1271 struct sk_buff *build_skb_around(struct sk_buff *skb, 1272 void *data, unsigned int frag_size); 1273 void skb_attempt_defer_free(struct sk_buff *skb); 1274 1275 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); 1276 struct sk_buff *slab_build_skb(void *data); 1277 1278 /** 1279 * alloc_skb - allocate a network buffer 1280 * @size: size to allocate 1281 * @priority: allocation mask 1282 * 1283 * This function is a convenient wrapper around __alloc_skb(). 1284 */ 1285 static inline struct sk_buff *alloc_skb(unsigned int size, 1286 gfp_t priority) 1287 { 1288 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 1289 } 1290 1291 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 1292 unsigned long data_len, 1293 int max_page_order, 1294 int *errcode, 1295 gfp_t gfp_mask); 1296 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); 1297 1298 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 1299 struct sk_buff_fclones { 1300 struct sk_buff skb1; 1301 1302 struct sk_buff skb2; 1303 1304 refcount_t fclone_ref; 1305 }; 1306 1307 /** 1308 * skb_fclone_busy - check if fclone is busy 1309 * @sk: socket 1310 * @skb: buffer 1311 * 1312 * Returns true if skb is a fast clone, and its clone is not freed. 1313 * Some drivers call skb_orphan() in their ndo_start_xmit(), 1314 * so we also check that this didnt happen. 1315 */ 1316 static inline bool skb_fclone_busy(const struct sock *sk, 1317 const struct sk_buff *skb) 1318 { 1319 const struct sk_buff_fclones *fclones; 1320 1321 fclones = container_of(skb, struct sk_buff_fclones, skb1); 1322 1323 return skb->fclone == SKB_FCLONE_ORIG && 1324 refcount_read(&fclones->fclone_ref) > 1 && 1325 READ_ONCE(fclones->skb2.sk) == sk; 1326 } 1327 1328 /** 1329 * alloc_skb_fclone - allocate a network buffer from fclone cache 1330 * @size: size to allocate 1331 * @priority: allocation mask 1332 * 1333 * This function is a convenient wrapper around __alloc_skb(). 1334 */ 1335 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 1336 gfp_t priority) 1337 { 1338 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 1339 } 1340 1341 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 1342 void skb_headers_offset_update(struct sk_buff *skb, int off); 1343 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 1344 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 1345 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); 1346 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 1347 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 1348 gfp_t gfp_mask, bool fclone); 1349 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 1350 gfp_t gfp_mask) 1351 { 1352 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 1353 } 1354 1355 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 1356 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 1357 unsigned int headroom); 1358 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom); 1359 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 1360 int newtailroom, gfp_t priority); 1361 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 1362 int offset, int len); 1363 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, 1364 int offset, int len); 1365 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 1366 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); 1367 1368 /** 1369 * skb_pad - zero pad the tail of an skb 1370 * @skb: buffer to pad 1371 * @pad: space to pad 1372 * 1373 * Ensure that a buffer is followed by a padding area that is zero 1374 * filled. Used by network drivers which may DMA or transfer data 1375 * beyond the buffer end onto the wire. 1376 * 1377 * May return error in out of memory cases. The skb is freed on error. 1378 */ 1379 static inline int skb_pad(struct sk_buff *skb, int pad) 1380 { 1381 return __skb_pad(skb, pad, true); 1382 } 1383 #define dev_kfree_skb(a) consume_skb(a) 1384 1385 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 1386 int offset, size_t size); 1387 1388 struct skb_seq_state { 1389 __u32 lower_offset; 1390 __u32 upper_offset; 1391 __u32 frag_idx; 1392 __u32 stepped_offset; 1393 struct sk_buff *root_skb; 1394 struct sk_buff *cur_skb; 1395 __u8 *frag_data; 1396 __u32 frag_off; 1397 }; 1398 1399 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 1400 unsigned int to, struct skb_seq_state *st); 1401 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 1402 struct skb_seq_state *st); 1403 void skb_abort_seq_read(struct skb_seq_state *st); 1404 1405 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 1406 unsigned int to, struct ts_config *config); 1407 1408 /* 1409 * Packet hash types specify the type of hash in skb_set_hash. 1410 * 1411 * Hash types refer to the protocol layer addresses which are used to 1412 * construct a packet's hash. The hashes are used to differentiate or identify 1413 * flows of the protocol layer for the hash type. Hash types are either 1414 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 1415 * 1416 * Properties of hashes: 1417 * 1418 * 1) Two packets in different flows have different hash values 1419 * 2) Two packets in the same flow should have the same hash value 1420 * 1421 * A hash at a higher layer is considered to be more specific. A driver should 1422 * set the most specific hash possible. 1423 * 1424 * A driver cannot indicate a more specific hash than the layer at which a hash 1425 * was computed. For instance an L3 hash cannot be set as an L4 hash. 1426 * 1427 * A driver may indicate a hash level which is less specific than the 1428 * actual layer the hash was computed on. For instance, a hash computed 1429 * at L4 may be considered an L3 hash. This should only be done if the 1430 * driver can't unambiguously determine that the HW computed the hash at 1431 * the higher layer. Note that the "should" in the second property above 1432 * permits this. 1433 */ 1434 enum pkt_hash_types { 1435 PKT_HASH_TYPE_NONE, /* Undefined type */ 1436 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 1437 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 1438 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 1439 }; 1440 1441 static inline void skb_clear_hash(struct sk_buff *skb) 1442 { 1443 skb->hash = 0; 1444 skb->sw_hash = 0; 1445 skb->l4_hash = 0; 1446 } 1447 1448 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 1449 { 1450 if (!skb->l4_hash) 1451 skb_clear_hash(skb); 1452 } 1453 1454 static inline void 1455 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) 1456 { 1457 skb->l4_hash = is_l4; 1458 skb->sw_hash = is_sw; 1459 skb->hash = hash; 1460 } 1461 1462 static inline void 1463 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 1464 { 1465 /* Used by drivers to set hash from HW */ 1466 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); 1467 } 1468 1469 static inline void 1470 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) 1471 { 1472 __skb_set_hash(skb, hash, true, is_l4); 1473 } 1474 1475 void __skb_get_hash(struct sk_buff *skb); 1476 u32 __skb_get_hash_symmetric(const struct sk_buff *skb); 1477 u32 skb_get_poff(const struct sk_buff *skb); 1478 u32 __skb_get_poff(const struct sk_buff *skb, const void *data, 1479 const struct flow_keys_basic *keys, int hlen); 1480 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, 1481 const void *data, int hlen_proto); 1482 1483 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, 1484 int thoff, u8 ip_proto) 1485 { 1486 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); 1487 } 1488 1489 void skb_flow_dissector_init(struct flow_dissector *flow_dissector, 1490 const struct flow_dissector_key *key, 1491 unsigned int key_count); 1492 1493 struct bpf_flow_dissector; 1494 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, 1495 __be16 proto, int nhoff, int hlen, unsigned int flags); 1496 1497 bool __skb_flow_dissect(const struct net *net, 1498 const struct sk_buff *skb, 1499 struct flow_dissector *flow_dissector, 1500 void *target_container, const void *data, 1501 __be16 proto, int nhoff, int hlen, unsigned int flags); 1502 1503 static inline bool skb_flow_dissect(const struct sk_buff *skb, 1504 struct flow_dissector *flow_dissector, 1505 void *target_container, unsigned int flags) 1506 { 1507 return __skb_flow_dissect(NULL, skb, flow_dissector, 1508 target_container, NULL, 0, 0, 0, flags); 1509 } 1510 1511 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, 1512 struct flow_keys *flow, 1513 unsigned int flags) 1514 { 1515 memset(flow, 0, sizeof(*flow)); 1516 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, 1517 flow, NULL, 0, 0, 0, flags); 1518 } 1519 1520 static inline bool 1521 skb_flow_dissect_flow_keys_basic(const struct net *net, 1522 const struct sk_buff *skb, 1523 struct flow_keys_basic *flow, 1524 const void *data, __be16 proto, 1525 int nhoff, int hlen, unsigned int flags) 1526 { 1527 memset(flow, 0, sizeof(*flow)); 1528 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, 1529 data, proto, nhoff, hlen, flags); 1530 } 1531 1532 void skb_flow_dissect_meta(const struct sk_buff *skb, 1533 struct flow_dissector *flow_dissector, 1534 void *target_container); 1535 1536 /* Gets a skb connection tracking info, ctinfo map should be a 1537 * map of mapsize to translate enum ip_conntrack_info states 1538 * to user states. 1539 */ 1540 void 1541 skb_flow_dissect_ct(const struct sk_buff *skb, 1542 struct flow_dissector *flow_dissector, 1543 void *target_container, 1544 u16 *ctinfo_map, size_t mapsize, 1545 bool post_ct, u16 zone); 1546 void 1547 skb_flow_dissect_tunnel_info(const struct sk_buff *skb, 1548 struct flow_dissector *flow_dissector, 1549 void *target_container); 1550 1551 void skb_flow_dissect_hash(const struct sk_buff *skb, 1552 struct flow_dissector *flow_dissector, 1553 void *target_container); 1554 1555 static inline __u32 skb_get_hash(struct sk_buff *skb) 1556 { 1557 if (!skb->l4_hash && !skb->sw_hash) 1558 __skb_get_hash(skb); 1559 1560 return skb->hash; 1561 } 1562 1563 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) 1564 { 1565 if (!skb->l4_hash && !skb->sw_hash) { 1566 struct flow_keys keys; 1567 __u32 hash = __get_hash_from_flowi6(fl6, &keys); 1568 1569 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); 1570 } 1571 1572 return skb->hash; 1573 } 1574 1575 __u32 skb_get_hash_perturb(const struct sk_buff *skb, 1576 const siphash_key_t *perturb); 1577 1578 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 1579 { 1580 return skb->hash; 1581 } 1582 1583 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 1584 { 1585 to->hash = from->hash; 1586 to->sw_hash = from->sw_hash; 1587 to->l4_hash = from->l4_hash; 1588 }; 1589 1590 static inline int skb_cmp_decrypted(const struct sk_buff *skb1, 1591 const struct sk_buff *skb2) 1592 { 1593 #ifdef CONFIG_TLS_DEVICE 1594 return skb2->decrypted - skb1->decrypted; 1595 #else 1596 return 0; 1597 #endif 1598 } 1599 1600 static inline void skb_copy_decrypted(struct sk_buff *to, 1601 const struct sk_buff *from) 1602 { 1603 #ifdef CONFIG_TLS_DEVICE 1604 to->decrypted = from->decrypted; 1605 #endif 1606 } 1607 1608 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1609 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1610 { 1611 return skb->head + skb->end; 1612 } 1613 1614 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1615 { 1616 return skb->end; 1617 } 1618 1619 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) 1620 { 1621 skb->end = offset; 1622 } 1623 #else 1624 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 1625 { 1626 return skb->end; 1627 } 1628 1629 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 1630 { 1631 return skb->end - skb->head; 1632 } 1633 1634 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) 1635 { 1636 skb->end = skb->head + offset; 1637 } 1638 #endif 1639 1640 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, 1641 struct ubuf_info *uarg); 1642 1643 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); 1644 1645 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, 1646 bool success); 1647 1648 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk, 1649 struct sk_buff *skb, struct iov_iter *from, 1650 size_t length); 1651 1652 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb, 1653 struct msghdr *msg, int len) 1654 { 1655 return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len); 1656 } 1657 1658 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, 1659 struct msghdr *msg, int len, 1660 struct ubuf_info *uarg); 1661 1662 /* Internal */ 1663 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 1664 1665 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1666 { 1667 return &skb_shinfo(skb)->hwtstamps; 1668 } 1669 1670 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) 1671 { 1672 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; 1673 1674 return is_zcopy ? skb_uarg(skb) : NULL; 1675 } 1676 1677 static inline bool skb_zcopy_pure(const struct sk_buff *skb) 1678 { 1679 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY; 1680 } 1681 1682 static inline bool skb_zcopy_managed(const struct sk_buff *skb) 1683 { 1684 return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS; 1685 } 1686 1687 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1, 1688 const struct sk_buff *skb2) 1689 { 1690 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2); 1691 } 1692 1693 static inline void net_zcopy_get(struct ubuf_info *uarg) 1694 { 1695 refcount_inc(&uarg->refcnt); 1696 } 1697 1698 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) 1699 { 1700 skb_shinfo(skb)->destructor_arg = uarg; 1701 skb_shinfo(skb)->flags |= uarg->flags; 1702 } 1703 1704 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, 1705 bool *have_ref) 1706 { 1707 if (skb && uarg && !skb_zcopy(skb)) { 1708 if (unlikely(have_ref && *have_ref)) 1709 *have_ref = false; 1710 else 1711 net_zcopy_get(uarg); 1712 skb_zcopy_init(skb, uarg); 1713 } 1714 } 1715 1716 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) 1717 { 1718 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); 1719 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; 1720 } 1721 1722 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) 1723 { 1724 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; 1725 } 1726 1727 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) 1728 { 1729 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); 1730 } 1731 1732 static inline void net_zcopy_put(struct ubuf_info *uarg) 1733 { 1734 if (uarg) 1735 uarg->callback(NULL, uarg, true); 1736 } 1737 1738 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) 1739 { 1740 if (uarg) { 1741 if (uarg->callback == msg_zerocopy_callback) 1742 msg_zerocopy_put_abort(uarg, have_uref); 1743 else if (have_uref) 1744 net_zcopy_put(uarg); 1745 } 1746 } 1747 1748 /* Release a reference on a zerocopy structure */ 1749 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) 1750 { 1751 struct ubuf_info *uarg = skb_zcopy(skb); 1752 1753 if (uarg) { 1754 if (!skb_zcopy_is_nouarg(skb)) 1755 uarg->callback(skb, uarg, zerocopy_success); 1756 1757 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY; 1758 } 1759 } 1760 1761 void __skb_zcopy_downgrade_managed(struct sk_buff *skb); 1762 1763 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb) 1764 { 1765 if (unlikely(skb_zcopy_managed(skb))) 1766 __skb_zcopy_downgrade_managed(skb); 1767 } 1768 1769 static inline void skb_mark_not_on_list(struct sk_buff *skb) 1770 { 1771 skb->next = NULL; 1772 } 1773 1774 static inline void skb_poison_list(struct sk_buff *skb) 1775 { 1776 #ifdef CONFIG_DEBUG_NET 1777 skb->next = SKB_LIST_POISON_NEXT; 1778 #endif 1779 } 1780 1781 /* Iterate through singly-linked GSO fragments of an skb. */ 1782 #define skb_list_walk_safe(first, skb, next_skb) \ 1783 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ 1784 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) 1785 1786 static inline void skb_list_del_init(struct sk_buff *skb) 1787 { 1788 __list_del_entry(&skb->list); 1789 skb_mark_not_on_list(skb); 1790 } 1791 1792 /** 1793 * skb_queue_empty - check if a queue is empty 1794 * @list: queue head 1795 * 1796 * Returns true if the queue is empty, false otherwise. 1797 */ 1798 static inline int skb_queue_empty(const struct sk_buff_head *list) 1799 { 1800 return list->next == (const struct sk_buff *) list; 1801 } 1802 1803 /** 1804 * skb_queue_empty_lockless - check if a queue is empty 1805 * @list: queue head 1806 * 1807 * Returns true if the queue is empty, false otherwise. 1808 * This variant can be used in lockless contexts. 1809 */ 1810 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) 1811 { 1812 return READ_ONCE(list->next) == (const struct sk_buff *) list; 1813 } 1814 1815 1816 /** 1817 * skb_queue_is_last - check if skb is the last entry in the queue 1818 * @list: queue head 1819 * @skb: buffer 1820 * 1821 * Returns true if @skb is the last buffer on the list. 1822 */ 1823 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1824 const struct sk_buff *skb) 1825 { 1826 return skb->next == (const struct sk_buff *) list; 1827 } 1828 1829 /** 1830 * skb_queue_is_first - check if skb is the first entry in the queue 1831 * @list: queue head 1832 * @skb: buffer 1833 * 1834 * Returns true if @skb is the first buffer on the list. 1835 */ 1836 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1837 const struct sk_buff *skb) 1838 { 1839 return skb->prev == (const struct sk_buff *) list; 1840 } 1841 1842 /** 1843 * skb_queue_next - return the next packet in the queue 1844 * @list: queue head 1845 * @skb: current buffer 1846 * 1847 * Return the next packet in @list after @skb. It is only valid to 1848 * call this if skb_queue_is_last() evaluates to false. 1849 */ 1850 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1851 const struct sk_buff *skb) 1852 { 1853 /* This BUG_ON may seem severe, but if we just return then we 1854 * are going to dereference garbage. 1855 */ 1856 BUG_ON(skb_queue_is_last(list, skb)); 1857 return skb->next; 1858 } 1859 1860 /** 1861 * skb_queue_prev - return the prev packet in the queue 1862 * @list: queue head 1863 * @skb: current buffer 1864 * 1865 * Return the prev packet in @list before @skb. It is only valid to 1866 * call this if skb_queue_is_first() evaluates to false. 1867 */ 1868 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1869 const struct sk_buff *skb) 1870 { 1871 /* This BUG_ON may seem severe, but if we just return then we 1872 * are going to dereference garbage. 1873 */ 1874 BUG_ON(skb_queue_is_first(list, skb)); 1875 return skb->prev; 1876 } 1877 1878 /** 1879 * skb_get - reference buffer 1880 * @skb: buffer to reference 1881 * 1882 * Makes another reference to a socket buffer and returns a pointer 1883 * to the buffer. 1884 */ 1885 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1886 { 1887 refcount_inc(&skb->users); 1888 return skb; 1889 } 1890 1891 /* 1892 * If users == 1, we are the only owner and can avoid redundant atomic changes. 1893 */ 1894 1895 /** 1896 * skb_cloned - is the buffer a clone 1897 * @skb: buffer to check 1898 * 1899 * Returns true if the buffer was generated with skb_clone() and is 1900 * one of multiple shared copies of the buffer. Cloned buffers are 1901 * shared data so must not be written to under normal circumstances. 1902 */ 1903 static inline int skb_cloned(const struct sk_buff *skb) 1904 { 1905 return skb->cloned && 1906 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1907 } 1908 1909 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1910 { 1911 might_sleep_if(gfpflags_allow_blocking(pri)); 1912 1913 if (skb_cloned(skb)) 1914 return pskb_expand_head(skb, 0, 0, pri); 1915 1916 return 0; 1917 } 1918 1919 /* This variant of skb_unclone() makes sure skb->truesize 1920 * and skb_end_offset() are not changed, whenever a new skb->head is needed. 1921 * 1922 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X)) 1923 * when various debugging features are in place. 1924 */ 1925 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri); 1926 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri) 1927 { 1928 might_sleep_if(gfpflags_allow_blocking(pri)); 1929 1930 if (skb_cloned(skb)) 1931 return __skb_unclone_keeptruesize(skb, pri); 1932 return 0; 1933 } 1934 1935 /** 1936 * skb_header_cloned - is the header a clone 1937 * @skb: buffer to check 1938 * 1939 * Returns true if modifying the header part of the buffer requires 1940 * the data to be copied. 1941 */ 1942 static inline int skb_header_cloned(const struct sk_buff *skb) 1943 { 1944 int dataref; 1945 1946 if (!skb->cloned) 1947 return 0; 1948 1949 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1950 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1951 return dataref != 1; 1952 } 1953 1954 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) 1955 { 1956 might_sleep_if(gfpflags_allow_blocking(pri)); 1957 1958 if (skb_header_cloned(skb)) 1959 return pskb_expand_head(skb, 0, 0, pri); 1960 1961 return 0; 1962 } 1963 1964 /** 1965 * __skb_header_release() - allow clones to use the headroom 1966 * @skb: buffer to operate on 1967 * 1968 * See "DOC: dataref and headerless skbs". 1969 */ 1970 static inline void __skb_header_release(struct sk_buff *skb) 1971 { 1972 skb->nohdr = 1; 1973 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1974 } 1975 1976 1977 /** 1978 * skb_shared - is the buffer shared 1979 * @skb: buffer to check 1980 * 1981 * Returns true if more than one person has a reference to this 1982 * buffer. 1983 */ 1984 static inline int skb_shared(const struct sk_buff *skb) 1985 { 1986 return refcount_read(&skb->users) != 1; 1987 } 1988 1989 /** 1990 * skb_share_check - check if buffer is shared and if so clone it 1991 * @skb: buffer to check 1992 * @pri: priority for memory allocation 1993 * 1994 * If the buffer is shared the buffer is cloned and the old copy 1995 * drops a reference. A new clone with a single reference is returned. 1996 * If the buffer is not shared the original buffer is returned. When 1997 * being called from interrupt status or with spinlocks held pri must 1998 * be GFP_ATOMIC. 1999 * 2000 * NULL is returned on a memory allocation failure. 2001 */ 2002 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 2003 { 2004 might_sleep_if(gfpflags_allow_blocking(pri)); 2005 if (skb_shared(skb)) { 2006 struct sk_buff *nskb = skb_clone(skb, pri); 2007 2008 if (likely(nskb)) 2009 consume_skb(skb); 2010 else 2011 kfree_skb(skb); 2012 skb = nskb; 2013 } 2014 return skb; 2015 } 2016 2017 /* 2018 * Copy shared buffers into a new sk_buff. We effectively do COW on 2019 * packets to handle cases where we have a local reader and forward 2020 * and a couple of other messy ones. The normal one is tcpdumping 2021 * a packet thats being forwarded. 2022 */ 2023 2024 /** 2025 * skb_unshare - make a copy of a shared buffer 2026 * @skb: buffer to check 2027 * @pri: priority for memory allocation 2028 * 2029 * If the socket buffer is a clone then this function creates a new 2030 * copy of the data, drops a reference count on the old copy and returns 2031 * the new copy with the reference count at 1. If the buffer is not a clone 2032 * the original buffer is returned. When called with a spinlock held or 2033 * from interrupt state @pri must be %GFP_ATOMIC 2034 * 2035 * %NULL is returned on a memory allocation failure. 2036 */ 2037 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 2038 gfp_t pri) 2039 { 2040 might_sleep_if(gfpflags_allow_blocking(pri)); 2041 if (skb_cloned(skb)) { 2042 struct sk_buff *nskb = skb_copy(skb, pri); 2043 2044 /* Free our shared copy */ 2045 if (likely(nskb)) 2046 consume_skb(skb); 2047 else 2048 kfree_skb(skb); 2049 skb = nskb; 2050 } 2051 return skb; 2052 } 2053 2054 /** 2055 * skb_peek - peek at the head of an &sk_buff_head 2056 * @list_: list to peek at 2057 * 2058 * Peek an &sk_buff. Unlike most other operations you _MUST_ 2059 * be careful with this one. A peek leaves the buffer on the 2060 * list and someone else may run off with it. You must hold 2061 * the appropriate locks or have a private queue to do this. 2062 * 2063 * Returns %NULL for an empty list or a pointer to the head element. 2064 * The reference count is not incremented and the reference is therefore 2065 * volatile. Use with caution. 2066 */ 2067 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 2068 { 2069 struct sk_buff *skb = list_->next; 2070 2071 if (skb == (struct sk_buff *)list_) 2072 skb = NULL; 2073 return skb; 2074 } 2075 2076 /** 2077 * __skb_peek - peek at the head of a non-empty &sk_buff_head 2078 * @list_: list to peek at 2079 * 2080 * Like skb_peek(), but the caller knows that the list is not empty. 2081 */ 2082 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) 2083 { 2084 return list_->next; 2085 } 2086 2087 /** 2088 * skb_peek_next - peek skb following the given one from a queue 2089 * @skb: skb to start from 2090 * @list_: list to peek at 2091 * 2092 * Returns %NULL when the end of the list is met or a pointer to the 2093 * next element. The reference count is not incremented and the 2094 * reference is therefore volatile. Use with caution. 2095 */ 2096 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 2097 const struct sk_buff_head *list_) 2098 { 2099 struct sk_buff *next = skb->next; 2100 2101 if (next == (struct sk_buff *)list_) 2102 next = NULL; 2103 return next; 2104 } 2105 2106 /** 2107 * skb_peek_tail - peek at the tail of an &sk_buff_head 2108 * @list_: list to peek at 2109 * 2110 * Peek an &sk_buff. Unlike most other operations you _MUST_ 2111 * be careful with this one. A peek leaves the buffer on the 2112 * list and someone else may run off with it. You must hold 2113 * the appropriate locks or have a private queue to do this. 2114 * 2115 * Returns %NULL for an empty list or a pointer to the tail element. 2116 * The reference count is not incremented and the reference is therefore 2117 * volatile. Use with caution. 2118 */ 2119 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 2120 { 2121 struct sk_buff *skb = READ_ONCE(list_->prev); 2122 2123 if (skb == (struct sk_buff *)list_) 2124 skb = NULL; 2125 return skb; 2126 2127 } 2128 2129 /** 2130 * skb_queue_len - get queue length 2131 * @list_: list to measure 2132 * 2133 * Return the length of an &sk_buff queue. 2134 */ 2135 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 2136 { 2137 return list_->qlen; 2138 } 2139 2140 /** 2141 * skb_queue_len_lockless - get queue length 2142 * @list_: list to measure 2143 * 2144 * Return the length of an &sk_buff queue. 2145 * This variant can be used in lockless contexts. 2146 */ 2147 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) 2148 { 2149 return READ_ONCE(list_->qlen); 2150 } 2151 2152 /** 2153 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 2154 * @list: queue to initialize 2155 * 2156 * This initializes only the list and queue length aspects of 2157 * an sk_buff_head object. This allows to initialize the list 2158 * aspects of an sk_buff_head without reinitializing things like 2159 * the spinlock. It can also be used for on-stack sk_buff_head 2160 * objects where the spinlock is known to not be used. 2161 */ 2162 static inline void __skb_queue_head_init(struct sk_buff_head *list) 2163 { 2164 list->prev = list->next = (struct sk_buff *)list; 2165 list->qlen = 0; 2166 } 2167 2168 /* 2169 * This function creates a split out lock class for each invocation; 2170 * this is needed for now since a whole lot of users of the skb-queue 2171 * infrastructure in drivers have different locking usage (in hardirq) 2172 * than the networking core (in softirq only). In the long run either the 2173 * network layer or drivers should need annotation to consolidate the 2174 * main types of usage into 3 classes. 2175 */ 2176 static inline void skb_queue_head_init(struct sk_buff_head *list) 2177 { 2178 spin_lock_init(&list->lock); 2179 __skb_queue_head_init(list); 2180 } 2181 2182 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 2183 struct lock_class_key *class) 2184 { 2185 skb_queue_head_init(list); 2186 lockdep_set_class(&list->lock, class); 2187 } 2188 2189 /* 2190 * Insert an sk_buff on a list. 2191 * 2192 * The "__skb_xxxx()" functions are the non-atomic ones that 2193 * can only be called with interrupts disabled. 2194 */ 2195 static inline void __skb_insert(struct sk_buff *newsk, 2196 struct sk_buff *prev, struct sk_buff *next, 2197 struct sk_buff_head *list) 2198 { 2199 /* See skb_queue_empty_lockless() and skb_peek_tail() 2200 * for the opposite READ_ONCE() 2201 */ 2202 WRITE_ONCE(newsk->next, next); 2203 WRITE_ONCE(newsk->prev, prev); 2204 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk); 2205 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk); 2206 WRITE_ONCE(list->qlen, list->qlen + 1); 2207 } 2208 2209 static inline void __skb_queue_splice(const struct sk_buff_head *list, 2210 struct sk_buff *prev, 2211 struct sk_buff *next) 2212 { 2213 struct sk_buff *first = list->next; 2214 struct sk_buff *last = list->prev; 2215 2216 WRITE_ONCE(first->prev, prev); 2217 WRITE_ONCE(prev->next, first); 2218 2219 WRITE_ONCE(last->next, next); 2220 WRITE_ONCE(next->prev, last); 2221 } 2222 2223 /** 2224 * skb_queue_splice - join two skb lists, this is designed for stacks 2225 * @list: the new list to add 2226 * @head: the place to add it in the first list 2227 */ 2228 static inline void skb_queue_splice(const struct sk_buff_head *list, 2229 struct sk_buff_head *head) 2230 { 2231 if (!skb_queue_empty(list)) { 2232 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 2233 head->qlen += list->qlen; 2234 } 2235 } 2236 2237 /** 2238 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 2239 * @list: the new list to add 2240 * @head: the place to add it in the first list 2241 * 2242 * The list at @list is reinitialised 2243 */ 2244 static inline void skb_queue_splice_init(struct sk_buff_head *list, 2245 struct sk_buff_head *head) 2246 { 2247 if (!skb_queue_empty(list)) { 2248 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 2249 head->qlen += list->qlen; 2250 __skb_queue_head_init(list); 2251 } 2252 } 2253 2254 /** 2255 * skb_queue_splice_tail - join two skb lists, each list being a queue 2256 * @list: the new list to add 2257 * @head: the place to add it in the first list 2258 */ 2259 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 2260 struct sk_buff_head *head) 2261 { 2262 if (!skb_queue_empty(list)) { 2263 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2264 head->qlen += list->qlen; 2265 } 2266 } 2267 2268 /** 2269 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 2270 * @list: the new list to add 2271 * @head: the place to add it in the first list 2272 * 2273 * Each of the lists is a queue. 2274 * The list at @list is reinitialised 2275 */ 2276 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 2277 struct sk_buff_head *head) 2278 { 2279 if (!skb_queue_empty(list)) { 2280 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 2281 head->qlen += list->qlen; 2282 __skb_queue_head_init(list); 2283 } 2284 } 2285 2286 /** 2287 * __skb_queue_after - queue a buffer at the list head 2288 * @list: list to use 2289 * @prev: place after this buffer 2290 * @newsk: buffer to queue 2291 * 2292 * Queue a buffer int the middle of a list. This function takes no locks 2293 * and you must therefore hold required locks before calling it. 2294 * 2295 * A buffer cannot be placed on two lists at the same time. 2296 */ 2297 static inline void __skb_queue_after(struct sk_buff_head *list, 2298 struct sk_buff *prev, 2299 struct sk_buff *newsk) 2300 { 2301 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list); 2302 } 2303 2304 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 2305 struct sk_buff_head *list); 2306 2307 static inline void __skb_queue_before(struct sk_buff_head *list, 2308 struct sk_buff *next, 2309 struct sk_buff *newsk) 2310 { 2311 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list); 2312 } 2313 2314 /** 2315 * __skb_queue_head - queue a buffer at the list head 2316 * @list: list to use 2317 * @newsk: buffer to queue 2318 * 2319 * Queue a buffer at the start of a list. This function takes no locks 2320 * and you must therefore hold required locks before calling it. 2321 * 2322 * A buffer cannot be placed on two lists at the same time. 2323 */ 2324 static inline void __skb_queue_head(struct sk_buff_head *list, 2325 struct sk_buff *newsk) 2326 { 2327 __skb_queue_after(list, (struct sk_buff *)list, newsk); 2328 } 2329 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 2330 2331 /** 2332 * __skb_queue_tail - queue a buffer at the list tail 2333 * @list: list to use 2334 * @newsk: buffer to queue 2335 * 2336 * Queue a buffer at the end of a list. This function takes no locks 2337 * and you must therefore hold required locks before calling it. 2338 * 2339 * A buffer cannot be placed on two lists at the same time. 2340 */ 2341 static inline void __skb_queue_tail(struct sk_buff_head *list, 2342 struct sk_buff *newsk) 2343 { 2344 __skb_queue_before(list, (struct sk_buff *)list, newsk); 2345 } 2346 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 2347 2348 /* 2349 * remove sk_buff from list. _Must_ be called atomically, and with 2350 * the list known.. 2351 */ 2352 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 2353 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 2354 { 2355 struct sk_buff *next, *prev; 2356 2357 WRITE_ONCE(list->qlen, list->qlen - 1); 2358 next = skb->next; 2359 prev = skb->prev; 2360 skb->next = skb->prev = NULL; 2361 WRITE_ONCE(next->prev, prev); 2362 WRITE_ONCE(prev->next, next); 2363 } 2364 2365 /** 2366 * __skb_dequeue - remove from the head of the queue 2367 * @list: list to dequeue from 2368 * 2369 * Remove the head of the list. This function does not take any locks 2370 * so must be used with appropriate locks held only. The head item is 2371 * returned or %NULL if the list is empty. 2372 */ 2373 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 2374 { 2375 struct sk_buff *skb = skb_peek(list); 2376 if (skb) 2377 __skb_unlink(skb, list); 2378 return skb; 2379 } 2380 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 2381 2382 /** 2383 * __skb_dequeue_tail - remove from the tail of the queue 2384 * @list: list to dequeue from 2385 * 2386 * Remove the tail of the list. This function does not take any locks 2387 * so must be used with appropriate locks held only. The tail item is 2388 * returned or %NULL if the list is empty. 2389 */ 2390 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 2391 { 2392 struct sk_buff *skb = skb_peek_tail(list); 2393 if (skb) 2394 __skb_unlink(skb, list); 2395 return skb; 2396 } 2397 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 2398 2399 2400 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 2401 { 2402 return skb->data_len; 2403 } 2404 2405 static inline unsigned int skb_headlen(const struct sk_buff *skb) 2406 { 2407 return skb->len - skb->data_len; 2408 } 2409 2410 static inline unsigned int __skb_pagelen(const struct sk_buff *skb) 2411 { 2412 unsigned int i, len = 0; 2413 2414 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) 2415 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 2416 return len; 2417 } 2418 2419 static inline unsigned int skb_pagelen(const struct sk_buff *skb) 2420 { 2421 return skb_headlen(skb) + __skb_pagelen(skb); 2422 } 2423 2424 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo, 2425 int i, struct page *page, 2426 int off, int size) 2427 { 2428 skb_frag_t *frag = &shinfo->frags[i]; 2429 2430 /* 2431 * Propagate page pfmemalloc to the skb if we can. The problem is 2432 * that not all callers have unique ownership of the page but rely 2433 * on page_is_pfmemalloc doing the right thing(tm). 2434 */ 2435 frag->bv_page = page; 2436 frag->bv_offset = off; 2437 skb_frag_size_set(frag, size); 2438 } 2439 2440 /** 2441 * skb_len_add - adds a number to len fields of skb 2442 * @skb: buffer to add len to 2443 * @delta: number of bytes to add 2444 */ 2445 static inline void skb_len_add(struct sk_buff *skb, int delta) 2446 { 2447 skb->len += delta; 2448 skb->data_len += delta; 2449 skb->truesize += delta; 2450 } 2451 2452 /** 2453 * __skb_fill_page_desc - initialise a paged fragment in an skb 2454 * @skb: buffer containing fragment to be initialised 2455 * @i: paged fragment index to initialise 2456 * @page: the page to use for this fragment 2457 * @off: the offset to the data with @page 2458 * @size: the length of the data 2459 * 2460 * Initialises the @i'th fragment of @skb to point to &size bytes at 2461 * offset @off within @page. 2462 * 2463 * Does not take any additional reference on the fragment. 2464 */ 2465 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 2466 struct page *page, int off, int size) 2467 { 2468 __skb_fill_page_desc_noacc(skb_shinfo(skb), i, page, off, size); 2469 page = compound_head(page); 2470 if (page_is_pfmemalloc(page)) 2471 skb->pfmemalloc = true; 2472 } 2473 2474 /** 2475 * skb_fill_page_desc - initialise a paged fragment in an skb 2476 * @skb: buffer containing fragment to be initialised 2477 * @i: paged fragment index to initialise 2478 * @page: the page to use for this fragment 2479 * @off: the offset to the data with @page 2480 * @size: the length of the data 2481 * 2482 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 2483 * @skb to point to @size bytes at offset @off within @page. In 2484 * addition updates @skb such that @i is the last fragment. 2485 * 2486 * Does not take any additional reference on the fragment. 2487 */ 2488 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 2489 struct page *page, int off, int size) 2490 { 2491 __skb_fill_page_desc(skb, i, page, off, size); 2492 skb_shinfo(skb)->nr_frags = i + 1; 2493 } 2494 2495 /** 2496 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb 2497 * @skb: buffer containing fragment to be initialised 2498 * @i: paged fragment index to initialise 2499 * @page: the page to use for this fragment 2500 * @off: the offset to the data with @page 2501 * @size: the length of the data 2502 * 2503 * Variant of skb_fill_page_desc() which does not deal with 2504 * pfmemalloc, if page is not owned by us. 2505 */ 2506 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i, 2507 struct page *page, int off, 2508 int size) 2509 { 2510 struct skb_shared_info *shinfo = skb_shinfo(skb); 2511 2512 __skb_fill_page_desc_noacc(shinfo, i, page, off, size); 2513 shinfo->nr_frags = i + 1; 2514 } 2515 2516 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 2517 int size, unsigned int truesize); 2518 2519 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 2520 unsigned int truesize); 2521 2522 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 2523 2524 #ifdef NET_SKBUFF_DATA_USES_OFFSET 2525 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2526 { 2527 return skb->head + skb->tail; 2528 } 2529 2530 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2531 { 2532 skb->tail = skb->data - skb->head; 2533 } 2534 2535 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2536 { 2537 skb_reset_tail_pointer(skb); 2538 skb->tail += offset; 2539 } 2540 2541 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 2542 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 2543 { 2544 return skb->tail; 2545 } 2546 2547 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 2548 { 2549 skb->tail = skb->data; 2550 } 2551 2552 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 2553 { 2554 skb->tail = skb->data + offset; 2555 } 2556 2557 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 2558 2559 static inline void skb_assert_len(struct sk_buff *skb) 2560 { 2561 #ifdef CONFIG_DEBUG_NET 2562 if (WARN_ONCE(!skb->len, "%s\n", __func__)) 2563 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false); 2564 #endif /* CONFIG_DEBUG_NET */ 2565 } 2566 2567 /* 2568 * Add data to an sk_buff 2569 */ 2570 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 2571 void *skb_put(struct sk_buff *skb, unsigned int len); 2572 static inline void *__skb_put(struct sk_buff *skb, unsigned int len) 2573 { 2574 void *tmp = skb_tail_pointer(skb); 2575 SKB_LINEAR_ASSERT(skb); 2576 skb->tail += len; 2577 skb->len += len; 2578 return tmp; 2579 } 2580 2581 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) 2582 { 2583 void *tmp = __skb_put(skb, len); 2584 2585 memset(tmp, 0, len); 2586 return tmp; 2587 } 2588 2589 static inline void *__skb_put_data(struct sk_buff *skb, const void *data, 2590 unsigned int len) 2591 { 2592 void *tmp = __skb_put(skb, len); 2593 2594 memcpy(tmp, data, len); 2595 return tmp; 2596 } 2597 2598 static inline void __skb_put_u8(struct sk_buff *skb, u8 val) 2599 { 2600 *(u8 *)__skb_put(skb, 1) = val; 2601 } 2602 2603 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) 2604 { 2605 void *tmp = skb_put(skb, len); 2606 2607 memset(tmp, 0, len); 2608 2609 return tmp; 2610 } 2611 2612 static inline void *skb_put_data(struct sk_buff *skb, const void *data, 2613 unsigned int len) 2614 { 2615 void *tmp = skb_put(skb, len); 2616 2617 memcpy(tmp, data, len); 2618 2619 return tmp; 2620 } 2621 2622 static inline void skb_put_u8(struct sk_buff *skb, u8 val) 2623 { 2624 *(u8 *)skb_put(skb, 1) = val; 2625 } 2626 2627 void *skb_push(struct sk_buff *skb, unsigned int len); 2628 static inline void *__skb_push(struct sk_buff *skb, unsigned int len) 2629 { 2630 skb->data -= len; 2631 skb->len += len; 2632 return skb->data; 2633 } 2634 2635 void *skb_pull(struct sk_buff *skb, unsigned int len); 2636 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) 2637 { 2638 skb->len -= len; 2639 if (unlikely(skb->len < skb->data_len)) { 2640 #if defined(CONFIG_DEBUG_NET) 2641 skb->len += len; 2642 pr_err("__skb_pull(len=%u)\n", len); 2643 skb_dump(KERN_ERR, skb, false); 2644 #endif 2645 BUG(); 2646 } 2647 return skb->data += len; 2648 } 2649 2650 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) 2651 { 2652 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 2653 } 2654 2655 void *skb_pull_data(struct sk_buff *skb, size_t len); 2656 2657 void *__pskb_pull_tail(struct sk_buff *skb, int delta); 2658 2659 static inline enum skb_drop_reason 2660 pskb_may_pull_reason(struct sk_buff *skb, unsigned int len) 2661 { 2662 if (likely(len <= skb_headlen(skb))) 2663 return SKB_NOT_DROPPED_YET; 2664 2665 if (unlikely(len > skb->len)) 2666 return SKB_DROP_REASON_PKT_TOO_SMALL; 2667 2668 if (unlikely(!__pskb_pull_tail(skb, len - skb_headlen(skb)))) 2669 return SKB_DROP_REASON_NOMEM; 2670 2671 return SKB_NOT_DROPPED_YET; 2672 } 2673 2674 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) 2675 { 2676 return pskb_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET; 2677 } 2678 2679 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) 2680 { 2681 if (!pskb_may_pull(skb, len)) 2682 return NULL; 2683 2684 skb->len -= len; 2685 return skb->data += len; 2686 } 2687 2688 void skb_condense(struct sk_buff *skb); 2689 2690 /** 2691 * skb_headroom - bytes at buffer head 2692 * @skb: buffer to check 2693 * 2694 * Return the number of bytes of free space at the head of an &sk_buff. 2695 */ 2696 static inline unsigned int skb_headroom(const struct sk_buff *skb) 2697 { 2698 return skb->data - skb->head; 2699 } 2700 2701 /** 2702 * skb_tailroom - bytes at buffer end 2703 * @skb: buffer to check 2704 * 2705 * Return the number of bytes of free space at the tail of an sk_buff 2706 */ 2707 static inline int skb_tailroom(const struct sk_buff *skb) 2708 { 2709 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 2710 } 2711 2712 /** 2713 * skb_availroom - bytes at buffer end 2714 * @skb: buffer to check 2715 * 2716 * Return the number of bytes of free space at the tail of an sk_buff 2717 * allocated by sk_stream_alloc() 2718 */ 2719 static inline int skb_availroom(const struct sk_buff *skb) 2720 { 2721 if (skb_is_nonlinear(skb)) 2722 return 0; 2723 2724 return skb->end - skb->tail - skb->reserved_tailroom; 2725 } 2726 2727 /** 2728 * skb_reserve - adjust headroom 2729 * @skb: buffer to alter 2730 * @len: bytes to move 2731 * 2732 * Increase the headroom of an empty &sk_buff by reducing the tail 2733 * room. This is only allowed for an empty buffer. 2734 */ 2735 static inline void skb_reserve(struct sk_buff *skb, int len) 2736 { 2737 skb->data += len; 2738 skb->tail += len; 2739 } 2740 2741 /** 2742 * skb_tailroom_reserve - adjust reserved_tailroom 2743 * @skb: buffer to alter 2744 * @mtu: maximum amount of headlen permitted 2745 * @needed_tailroom: minimum amount of reserved_tailroom 2746 * 2747 * Set reserved_tailroom so that headlen can be as large as possible but 2748 * not larger than mtu and tailroom cannot be smaller than 2749 * needed_tailroom. 2750 * The required headroom should already have been reserved before using 2751 * this function. 2752 */ 2753 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, 2754 unsigned int needed_tailroom) 2755 { 2756 SKB_LINEAR_ASSERT(skb); 2757 if (mtu < skb_tailroom(skb) - needed_tailroom) 2758 /* use at most mtu */ 2759 skb->reserved_tailroom = skb_tailroom(skb) - mtu; 2760 else 2761 /* use up to all available space */ 2762 skb->reserved_tailroom = needed_tailroom; 2763 } 2764 2765 #define ENCAP_TYPE_ETHER 0 2766 #define ENCAP_TYPE_IPPROTO 1 2767 2768 static inline void skb_set_inner_protocol(struct sk_buff *skb, 2769 __be16 protocol) 2770 { 2771 skb->inner_protocol = protocol; 2772 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 2773 } 2774 2775 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 2776 __u8 ipproto) 2777 { 2778 skb->inner_ipproto = ipproto; 2779 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 2780 } 2781 2782 static inline void skb_reset_inner_headers(struct sk_buff *skb) 2783 { 2784 skb->inner_mac_header = skb->mac_header; 2785 skb->inner_network_header = skb->network_header; 2786 skb->inner_transport_header = skb->transport_header; 2787 } 2788 2789 static inline void skb_reset_mac_len(struct sk_buff *skb) 2790 { 2791 skb->mac_len = skb->network_header - skb->mac_header; 2792 } 2793 2794 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 2795 *skb) 2796 { 2797 return skb->head + skb->inner_transport_header; 2798 } 2799 2800 static inline int skb_inner_transport_offset(const struct sk_buff *skb) 2801 { 2802 return skb_inner_transport_header(skb) - skb->data; 2803 } 2804 2805 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 2806 { 2807 skb->inner_transport_header = skb->data - skb->head; 2808 } 2809 2810 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 2811 const int offset) 2812 { 2813 skb_reset_inner_transport_header(skb); 2814 skb->inner_transport_header += offset; 2815 } 2816 2817 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 2818 { 2819 return skb->head + skb->inner_network_header; 2820 } 2821 2822 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 2823 { 2824 skb->inner_network_header = skb->data - skb->head; 2825 } 2826 2827 static inline void skb_set_inner_network_header(struct sk_buff *skb, 2828 const int offset) 2829 { 2830 skb_reset_inner_network_header(skb); 2831 skb->inner_network_header += offset; 2832 } 2833 2834 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 2835 { 2836 return skb->head + skb->inner_mac_header; 2837 } 2838 2839 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 2840 { 2841 skb->inner_mac_header = skb->data - skb->head; 2842 } 2843 2844 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 2845 const int offset) 2846 { 2847 skb_reset_inner_mac_header(skb); 2848 skb->inner_mac_header += offset; 2849 } 2850 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 2851 { 2852 return skb->transport_header != (typeof(skb->transport_header))~0U; 2853 } 2854 2855 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 2856 { 2857 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb)); 2858 return skb->head + skb->transport_header; 2859 } 2860 2861 static inline void skb_reset_transport_header(struct sk_buff *skb) 2862 { 2863 skb->transport_header = skb->data - skb->head; 2864 } 2865 2866 static inline void skb_set_transport_header(struct sk_buff *skb, 2867 const int offset) 2868 { 2869 skb_reset_transport_header(skb); 2870 skb->transport_header += offset; 2871 } 2872 2873 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 2874 { 2875 return skb->head + skb->network_header; 2876 } 2877 2878 static inline void skb_reset_network_header(struct sk_buff *skb) 2879 { 2880 skb->network_header = skb->data - skb->head; 2881 } 2882 2883 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 2884 { 2885 skb_reset_network_header(skb); 2886 skb->network_header += offset; 2887 } 2888 2889 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 2890 { 2891 return skb->mac_header != (typeof(skb->mac_header))~0U; 2892 } 2893 2894 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 2895 { 2896 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb)); 2897 return skb->head + skb->mac_header; 2898 } 2899 2900 static inline int skb_mac_offset(const struct sk_buff *skb) 2901 { 2902 return skb_mac_header(skb) - skb->data; 2903 } 2904 2905 static inline u32 skb_mac_header_len(const struct sk_buff *skb) 2906 { 2907 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb)); 2908 return skb->network_header - skb->mac_header; 2909 } 2910 2911 static inline void skb_unset_mac_header(struct sk_buff *skb) 2912 { 2913 skb->mac_header = (typeof(skb->mac_header))~0U; 2914 } 2915 2916 static inline void skb_reset_mac_header(struct sk_buff *skb) 2917 { 2918 skb->mac_header = skb->data - skb->head; 2919 } 2920 2921 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 2922 { 2923 skb_reset_mac_header(skb); 2924 skb->mac_header += offset; 2925 } 2926 2927 static inline void skb_pop_mac_header(struct sk_buff *skb) 2928 { 2929 skb->mac_header = skb->network_header; 2930 } 2931 2932 static inline void skb_probe_transport_header(struct sk_buff *skb) 2933 { 2934 struct flow_keys_basic keys; 2935 2936 if (skb_transport_header_was_set(skb)) 2937 return; 2938 2939 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, 2940 NULL, 0, 0, 0, 0)) 2941 skb_set_transport_header(skb, keys.control.thoff); 2942 } 2943 2944 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 2945 { 2946 if (skb_mac_header_was_set(skb)) { 2947 const unsigned char *old_mac = skb_mac_header(skb); 2948 2949 skb_set_mac_header(skb, -skb->mac_len); 2950 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 2951 } 2952 } 2953 2954 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 2955 { 2956 return skb->csum_start - skb_headroom(skb); 2957 } 2958 2959 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) 2960 { 2961 return skb->head + skb->csum_start; 2962 } 2963 2964 static inline int skb_transport_offset(const struct sk_buff *skb) 2965 { 2966 return skb_transport_header(skb) - skb->data; 2967 } 2968 2969 static inline u32 skb_network_header_len(const struct sk_buff *skb) 2970 { 2971 return skb->transport_header - skb->network_header; 2972 } 2973 2974 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 2975 { 2976 return skb->inner_transport_header - skb->inner_network_header; 2977 } 2978 2979 static inline int skb_network_offset(const struct sk_buff *skb) 2980 { 2981 return skb_network_header(skb) - skb->data; 2982 } 2983 2984 static inline int skb_inner_network_offset(const struct sk_buff *skb) 2985 { 2986 return skb_inner_network_header(skb) - skb->data; 2987 } 2988 2989 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 2990 { 2991 return pskb_may_pull(skb, skb_network_offset(skb) + len); 2992 } 2993 2994 /* 2995 * CPUs often take a performance hit when accessing unaligned memory 2996 * locations. The actual performance hit varies, it can be small if the 2997 * hardware handles it or large if we have to take an exception and fix it 2998 * in software. 2999 * 3000 * Since an ethernet header is 14 bytes network drivers often end up with 3001 * the IP header at an unaligned offset. The IP header can be aligned by 3002 * shifting the start of the packet by 2 bytes. Drivers should do this 3003 * with: 3004 * 3005 * skb_reserve(skb, NET_IP_ALIGN); 3006 * 3007 * The downside to this alignment of the IP header is that the DMA is now 3008 * unaligned. On some architectures the cost of an unaligned DMA is high 3009 * and this cost outweighs the gains made by aligning the IP header. 3010 * 3011 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 3012 * to be overridden. 3013 */ 3014 #ifndef NET_IP_ALIGN 3015 #define NET_IP_ALIGN 2 3016 #endif 3017 3018 /* 3019 * The networking layer reserves some headroom in skb data (via 3020 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 3021 * the header has to grow. In the default case, if the header has to grow 3022 * 32 bytes or less we avoid the reallocation. 3023 * 3024 * Unfortunately this headroom changes the DMA alignment of the resulting 3025 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 3026 * on some architectures. An architecture can override this value, 3027 * perhaps setting it to a cacheline in size (since that will maintain 3028 * cacheline alignment of the DMA). It must be a power of 2. 3029 * 3030 * Various parts of the networking layer expect at least 32 bytes of 3031 * headroom, you should not reduce this. 3032 * 3033 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 3034 * to reduce average number of cache lines per packet. 3035 * get_rps_cpu() for example only access one 64 bytes aligned block : 3036 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 3037 */ 3038 #ifndef NET_SKB_PAD 3039 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 3040 #endif 3041 3042 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 3043 3044 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) 3045 { 3046 if (WARN_ON(skb_is_nonlinear(skb))) 3047 return; 3048 skb->len = len; 3049 skb_set_tail_pointer(skb, len); 3050 } 3051 3052 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 3053 { 3054 __skb_set_length(skb, len); 3055 } 3056 3057 void skb_trim(struct sk_buff *skb, unsigned int len); 3058 3059 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 3060 { 3061 if (skb->data_len) 3062 return ___pskb_trim(skb, len); 3063 __skb_trim(skb, len); 3064 return 0; 3065 } 3066 3067 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 3068 { 3069 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 3070 } 3071 3072 /** 3073 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 3074 * @skb: buffer to alter 3075 * @len: new length 3076 * 3077 * This is identical to pskb_trim except that the caller knows that 3078 * the skb is not cloned so we should never get an error due to out- 3079 * of-memory. 3080 */ 3081 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 3082 { 3083 int err = pskb_trim(skb, len); 3084 BUG_ON(err); 3085 } 3086 3087 static inline int __skb_grow(struct sk_buff *skb, unsigned int len) 3088 { 3089 unsigned int diff = len - skb->len; 3090 3091 if (skb_tailroom(skb) < diff) { 3092 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), 3093 GFP_ATOMIC); 3094 if (ret) 3095 return ret; 3096 } 3097 __skb_set_length(skb, len); 3098 return 0; 3099 } 3100 3101 /** 3102 * skb_orphan - orphan a buffer 3103 * @skb: buffer to orphan 3104 * 3105 * If a buffer currently has an owner then we call the owner's 3106 * destructor function and make the @skb unowned. The buffer continues 3107 * to exist but is no longer charged to its former owner. 3108 */ 3109 static inline void skb_orphan(struct sk_buff *skb) 3110 { 3111 if (skb->destructor) { 3112 skb->destructor(skb); 3113 skb->destructor = NULL; 3114 skb->sk = NULL; 3115 } else { 3116 BUG_ON(skb->sk); 3117 } 3118 } 3119 3120 /** 3121 * skb_orphan_frags - orphan the frags contained in a buffer 3122 * @skb: buffer to orphan frags from 3123 * @gfp_mask: allocation mask for replacement pages 3124 * 3125 * For each frag in the SKB which needs a destructor (i.e. has an 3126 * owner) create a copy of that frag and release the original 3127 * page by calling the destructor. 3128 */ 3129 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 3130 { 3131 if (likely(!skb_zcopy(skb))) 3132 return 0; 3133 if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN) 3134 return 0; 3135 return skb_copy_ubufs(skb, gfp_mask); 3136 } 3137 3138 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ 3139 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) 3140 { 3141 if (likely(!skb_zcopy(skb))) 3142 return 0; 3143 return skb_copy_ubufs(skb, gfp_mask); 3144 } 3145 3146 /** 3147 * __skb_queue_purge - empty a list 3148 * @list: list to empty 3149 * 3150 * Delete all buffers on an &sk_buff list. Each buffer is removed from 3151 * the list and one reference dropped. This function does not take the 3152 * list lock and the caller must hold the relevant locks to use it. 3153 */ 3154 static inline void __skb_queue_purge(struct sk_buff_head *list) 3155 { 3156 struct sk_buff *skb; 3157 while ((skb = __skb_dequeue(list)) != NULL) 3158 kfree_skb(skb); 3159 } 3160 void skb_queue_purge(struct sk_buff_head *list); 3161 3162 unsigned int skb_rbtree_purge(struct rb_root *root); 3163 3164 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 3165 3166 /** 3167 * netdev_alloc_frag - allocate a page fragment 3168 * @fragsz: fragment size 3169 * 3170 * Allocates a frag from a page for receive buffer. 3171 * Uses GFP_ATOMIC allocations. 3172 */ 3173 static inline void *netdev_alloc_frag(unsigned int fragsz) 3174 { 3175 return __netdev_alloc_frag_align(fragsz, ~0u); 3176 } 3177 3178 static inline void *netdev_alloc_frag_align(unsigned int fragsz, 3179 unsigned int align) 3180 { 3181 WARN_ON_ONCE(!is_power_of_2(align)); 3182 return __netdev_alloc_frag_align(fragsz, -align); 3183 } 3184 3185 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 3186 gfp_t gfp_mask); 3187 3188 /** 3189 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 3190 * @dev: network device to receive on 3191 * @length: length to allocate 3192 * 3193 * Allocate a new &sk_buff and assign it a usage count of one. The 3194 * buffer has unspecified headroom built in. Users should allocate 3195 * the headroom they think they need without accounting for the 3196 * built in space. The built in space is used for optimisations. 3197 * 3198 * %NULL is returned if there is no free memory. Although this function 3199 * allocates memory it can be called from an interrupt. 3200 */ 3201 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 3202 unsigned int length) 3203 { 3204 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 3205 } 3206 3207 /* legacy helper around __netdev_alloc_skb() */ 3208 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 3209 gfp_t gfp_mask) 3210 { 3211 return __netdev_alloc_skb(NULL, length, gfp_mask); 3212 } 3213 3214 /* legacy helper around netdev_alloc_skb() */ 3215 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 3216 { 3217 return netdev_alloc_skb(NULL, length); 3218 } 3219 3220 3221 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 3222 unsigned int length, gfp_t gfp) 3223 { 3224 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 3225 3226 if (NET_IP_ALIGN && skb) 3227 skb_reserve(skb, NET_IP_ALIGN); 3228 return skb; 3229 } 3230 3231 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 3232 unsigned int length) 3233 { 3234 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 3235 } 3236 3237 static inline void skb_free_frag(void *addr) 3238 { 3239 page_frag_free(addr); 3240 } 3241 3242 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); 3243 3244 static inline void *napi_alloc_frag(unsigned int fragsz) 3245 { 3246 return __napi_alloc_frag_align(fragsz, ~0u); 3247 } 3248 3249 static inline void *napi_alloc_frag_align(unsigned int fragsz, 3250 unsigned int align) 3251 { 3252 WARN_ON_ONCE(!is_power_of_2(align)); 3253 return __napi_alloc_frag_align(fragsz, -align); 3254 } 3255 3256 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 3257 unsigned int length, gfp_t gfp_mask); 3258 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 3259 unsigned int length) 3260 { 3261 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 3262 } 3263 void napi_consume_skb(struct sk_buff *skb, int budget); 3264 3265 void napi_skb_free_stolen_head(struct sk_buff *skb); 3266 void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason); 3267 3268 /** 3269 * __dev_alloc_pages - allocate page for network Rx 3270 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 3271 * @order: size of the allocation 3272 * 3273 * Allocate a new page. 3274 * 3275 * %NULL is returned if there is no free memory. 3276 */ 3277 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 3278 unsigned int order) 3279 { 3280 /* This piece of code contains several assumptions. 3281 * 1. This is for device Rx, therefor a cold page is preferred. 3282 * 2. The expectation is the user wants a compound page. 3283 * 3. If requesting a order 0 page it will not be compound 3284 * due to the check to see if order has a value in prep_new_page 3285 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 3286 * code in gfp_to_alloc_flags that should be enforcing this. 3287 */ 3288 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; 3289 3290 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 3291 } 3292 3293 static inline struct page *dev_alloc_pages(unsigned int order) 3294 { 3295 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); 3296 } 3297 3298 /** 3299 * __dev_alloc_page - allocate a page for network Rx 3300 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 3301 * 3302 * Allocate a new page. 3303 * 3304 * %NULL is returned if there is no free memory. 3305 */ 3306 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 3307 { 3308 return __dev_alloc_pages(gfp_mask, 0); 3309 } 3310 3311 static inline struct page *dev_alloc_page(void) 3312 { 3313 return dev_alloc_pages(0); 3314 } 3315 3316 /** 3317 * dev_page_is_reusable - check whether a page can be reused for network Rx 3318 * @page: the page to test 3319 * 3320 * A page shouldn't be considered for reusing/recycling if it was allocated 3321 * under memory pressure or at a distant memory node. 3322 * 3323 * Returns false if this page should be returned to page allocator, true 3324 * otherwise. 3325 */ 3326 static inline bool dev_page_is_reusable(const struct page *page) 3327 { 3328 return likely(page_to_nid(page) == numa_mem_id() && 3329 !page_is_pfmemalloc(page)); 3330 } 3331 3332 /** 3333 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 3334 * @page: The page that was allocated from skb_alloc_page 3335 * @skb: The skb that may need pfmemalloc set 3336 */ 3337 static inline void skb_propagate_pfmemalloc(const struct page *page, 3338 struct sk_buff *skb) 3339 { 3340 if (page_is_pfmemalloc(page)) 3341 skb->pfmemalloc = true; 3342 } 3343 3344 /** 3345 * skb_frag_off() - Returns the offset of a skb fragment 3346 * @frag: the paged fragment 3347 */ 3348 static inline unsigned int skb_frag_off(const skb_frag_t *frag) 3349 { 3350 return frag->bv_offset; 3351 } 3352 3353 /** 3354 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta 3355 * @frag: skb fragment 3356 * @delta: value to add 3357 */ 3358 static inline void skb_frag_off_add(skb_frag_t *frag, int delta) 3359 { 3360 frag->bv_offset += delta; 3361 } 3362 3363 /** 3364 * skb_frag_off_set() - Sets the offset of a skb fragment 3365 * @frag: skb fragment 3366 * @offset: offset of fragment 3367 */ 3368 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) 3369 { 3370 frag->bv_offset = offset; 3371 } 3372 3373 /** 3374 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment 3375 * @fragto: skb fragment where offset is set 3376 * @fragfrom: skb fragment offset is copied from 3377 */ 3378 static inline void skb_frag_off_copy(skb_frag_t *fragto, 3379 const skb_frag_t *fragfrom) 3380 { 3381 fragto->bv_offset = fragfrom->bv_offset; 3382 } 3383 3384 /** 3385 * skb_frag_page - retrieve the page referred to by a paged fragment 3386 * @frag: the paged fragment 3387 * 3388 * Returns the &struct page associated with @frag. 3389 */ 3390 static inline struct page *skb_frag_page(const skb_frag_t *frag) 3391 { 3392 return frag->bv_page; 3393 } 3394 3395 /** 3396 * __skb_frag_ref - take an addition reference on a paged fragment. 3397 * @frag: the paged fragment 3398 * 3399 * Takes an additional reference on the paged fragment @frag. 3400 */ 3401 static inline void __skb_frag_ref(skb_frag_t *frag) 3402 { 3403 get_page(skb_frag_page(frag)); 3404 } 3405 3406 /** 3407 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 3408 * @skb: the buffer 3409 * @f: the fragment offset. 3410 * 3411 * Takes an additional reference on the @f'th paged fragment of @skb. 3412 */ 3413 static inline void skb_frag_ref(struct sk_buff *skb, int f) 3414 { 3415 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 3416 } 3417 3418 static inline void 3419 napi_frag_unref(skb_frag_t *frag, bool recycle, bool napi_safe) 3420 { 3421 struct page *page = skb_frag_page(frag); 3422 3423 #ifdef CONFIG_PAGE_POOL 3424 if (recycle && page_pool_return_skb_page(page, napi_safe)) 3425 return; 3426 #endif 3427 put_page(page); 3428 } 3429 3430 /** 3431 * __skb_frag_unref - release a reference on a paged fragment. 3432 * @frag: the paged fragment 3433 * @recycle: recycle the page if allocated via page_pool 3434 * 3435 * Releases a reference on the paged fragment @frag 3436 * or recycles the page via the page_pool API. 3437 */ 3438 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) 3439 { 3440 napi_frag_unref(frag, recycle, false); 3441 } 3442 3443 /** 3444 * skb_frag_unref - release a reference on a paged fragment of an skb. 3445 * @skb: the buffer 3446 * @f: the fragment offset 3447 * 3448 * Releases a reference on the @f'th paged fragment of @skb. 3449 */ 3450 static inline void skb_frag_unref(struct sk_buff *skb, int f) 3451 { 3452 struct skb_shared_info *shinfo = skb_shinfo(skb); 3453 3454 if (!skb_zcopy_managed(skb)) 3455 __skb_frag_unref(&shinfo->frags[f], skb->pp_recycle); 3456 } 3457 3458 /** 3459 * skb_frag_address - gets the address of the data contained in a paged fragment 3460 * @frag: the paged fragment buffer 3461 * 3462 * Returns the address of the data within @frag. The page must already 3463 * be mapped. 3464 */ 3465 static inline void *skb_frag_address(const skb_frag_t *frag) 3466 { 3467 return page_address(skb_frag_page(frag)) + skb_frag_off(frag); 3468 } 3469 3470 /** 3471 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 3472 * @frag: the paged fragment buffer 3473 * 3474 * Returns the address of the data within @frag. Checks that the page 3475 * is mapped and returns %NULL otherwise. 3476 */ 3477 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 3478 { 3479 void *ptr = page_address(skb_frag_page(frag)); 3480 if (unlikely(!ptr)) 3481 return NULL; 3482 3483 return ptr + skb_frag_off(frag); 3484 } 3485 3486 /** 3487 * skb_frag_page_copy() - sets the page in a fragment from another fragment 3488 * @fragto: skb fragment where page is set 3489 * @fragfrom: skb fragment page is copied from 3490 */ 3491 static inline void skb_frag_page_copy(skb_frag_t *fragto, 3492 const skb_frag_t *fragfrom) 3493 { 3494 fragto->bv_page = fragfrom->bv_page; 3495 } 3496 3497 /** 3498 * __skb_frag_set_page - sets the page contained in a paged fragment 3499 * @frag: the paged fragment 3500 * @page: the page to set 3501 * 3502 * Sets the fragment @frag to contain @page. 3503 */ 3504 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 3505 { 3506 frag->bv_page = page; 3507 } 3508 3509 /** 3510 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 3511 * @skb: the buffer 3512 * @f: the fragment offset 3513 * @page: the page to set 3514 * 3515 * Sets the @f'th fragment of @skb to contain @page. 3516 */ 3517 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 3518 struct page *page) 3519 { 3520 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 3521 } 3522 3523 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 3524 3525 /** 3526 * skb_frag_dma_map - maps a paged fragment via the DMA API 3527 * @dev: the device to map the fragment to 3528 * @frag: the paged fragment to map 3529 * @offset: the offset within the fragment (starting at the 3530 * fragment's own offset) 3531 * @size: the number of bytes to map 3532 * @dir: the direction of the mapping (``PCI_DMA_*``) 3533 * 3534 * Maps the page associated with @frag to @device. 3535 */ 3536 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 3537 const skb_frag_t *frag, 3538 size_t offset, size_t size, 3539 enum dma_data_direction dir) 3540 { 3541 return dma_map_page(dev, skb_frag_page(frag), 3542 skb_frag_off(frag) + offset, size, dir); 3543 } 3544 3545 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 3546 gfp_t gfp_mask) 3547 { 3548 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 3549 } 3550 3551 3552 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 3553 gfp_t gfp_mask) 3554 { 3555 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 3556 } 3557 3558 3559 /** 3560 * skb_clone_writable - is the header of a clone writable 3561 * @skb: buffer to check 3562 * @len: length up to which to write 3563 * 3564 * Returns true if modifying the header part of the cloned buffer 3565 * does not requires the data to be copied. 3566 */ 3567 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 3568 { 3569 return !skb_header_cloned(skb) && 3570 skb_headroom(skb) + len <= skb->hdr_len; 3571 } 3572 3573 static inline int skb_try_make_writable(struct sk_buff *skb, 3574 unsigned int write_len) 3575 { 3576 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && 3577 pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 3578 } 3579 3580 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 3581 int cloned) 3582 { 3583 int delta = 0; 3584 3585 if (headroom > skb_headroom(skb)) 3586 delta = headroom - skb_headroom(skb); 3587 3588 if (delta || cloned) 3589 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 3590 GFP_ATOMIC); 3591 return 0; 3592 } 3593 3594 /** 3595 * skb_cow - copy header of skb when it is required 3596 * @skb: buffer to cow 3597 * @headroom: needed headroom 3598 * 3599 * If the skb passed lacks sufficient headroom or its data part 3600 * is shared, data is reallocated. If reallocation fails, an error 3601 * is returned and original skb is not changed. 3602 * 3603 * The result is skb with writable area skb->head...skb->tail 3604 * and at least @headroom of space at head. 3605 */ 3606 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 3607 { 3608 return __skb_cow(skb, headroom, skb_cloned(skb)); 3609 } 3610 3611 /** 3612 * skb_cow_head - skb_cow but only making the head writable 3613 * @skb: buffer to cow 3614 * @headroom: needed headroom 3615 * 3616 * This function is identical to skb_cow except that we replace the 3617 * skb_cloned check by skb_header_cloned. It should be used when 3618 * you only need to push on some header and do not need to modify 3619 * the data. 3620 */ 3621 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 3622 { 3623 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 3624 } 3625 3626 /** 3627 * skb_padto - pad an skbuff up to a minimal size 3628 * @skb: buffer to pad 3629 * @len: minimal length 3630 * 3631 * Pads up a buffer to ensure the trailing bytes exist and are 3632 * blanked. If the buffer already contains sufficient data it 3633 * is untouched. Otherwise it is extended. Returns zero on 3634 * success. The skb is freed on error. 3635 */ 3636 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 3637 { 3638 unsigned int size = skb->len; 3639 if (likely(size >= len)) 3640 return 0; 3641 return skb_pad(skb, len - size); 3642 } 3643 3644 /** 3645 * __skb_put_padto - increase size and pad an skbuff up to a minimal size 3646 * @skb: buffer to pad 3647 * @len: minimal length 3648 * @free_on_error: free buffer on error 3649 * 3650 * Pads up a buffer to ensure the trailing bytes exist and are 3651 * blanked. If the buffer already contains sufficient data it 3652 * is untouched. Otherwise it is extended. Returns zero on 3653 * success. The skb is freed on error if @free_on_error is true. 3654 */ 3655 static inline int __must_check __skb_put_padto(struct sk_buff *skb, 3656 unsigned int len, 3657 bool free_on_error) 3658 { 3659 unsigned int size = skb->len; 3660 3661 if (unlikely(size < len)) { 3662 len -= size; 3663 if (__skb_pad(skb, len, free_on_error)) 3664 return -ENOMEM; 3665 __skb_put(skb, len); 3666 } 3667 return 0; 3668 } 3669 3670 /** 3671 * skb_put_padto - increase size and pad an skbuff up to a minimal size 3672 * @skb: buffer to pad 3673 * @len: minimal length 3674 * 3675 * Pads up a buffer to ensure the trailing bytes exist and are 3676 * blanked. If the buffer already contains sufficient data it 3677 * is untouched. Otherwise it is extended. Returns zero on 3678 * success. The skb is freed on error. 3679 */ 3680 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) 3681 { 3682 return __skb_put_padto(skb, len, true); 3683 } 3684 3685 static inline int skb_add_data(struct sk_buff *skb, 3686 struct iov_iter *from, int copy) 3687 { 3688 const int off = skb->len; 3689 3690 if (skb->ip_summed == CHECKSUM_NONE) { 3691 __wsum csum = 0; 3692 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, 3693 &csum, from)) { 3694 skb->csum = csum_block_add(skb->csum, csum, off); 3695 return 0; 3696 } 3697 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) 3698 return 0; 3699 3700 __skb_trim(skb, off); 3701 return -EFAULT; 3702 } 3703 3704 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 3705 const struct page *page, int off) 3706 { 3707 if (skb_zcopy(skb)) 3708 return false; 3709 if (i) { 3710 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; 3711 3712 return page == skb_frag_page(frag) && 3713 off == skb_frag_off(frag) + skb_frag_size(frag); 3714 } 3715 return false; 3716 } 3717 3718 static inline int __skb_linearize(struct sk_buff *skb) 3719 { 3720 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 3721 } 3722 3723 /** 3724 * skb_linearize - convert paged skb to linear one 3725 * @skb: buffer to linarize 3726 * 3727 * If there is no free memory -ENOMEM is returned, otherwise zero 3728 * is returned and the old skb data released. 3729 */ 3730 static inline int skb_linearize(struct sk_buff *skb) 3731 { 3732 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 3733 } 3734 3735 /** 3736 * skb_has_shared_frag - can any frag be overwritten 3737 * @skb: buffer to test 3738 * 3739 * Return true if the skb has at least one frag that might be modified 3740 * by an external entity (as in vmsplice()/sendfile()) 3741 */ 3742 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 3743 { 3744 return skb_is_nonlinear(skb) && 3745 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; 3746 } 3747 3748 /** 3749 * skb_linearize_cow - make sure skb is linear and writable 3750 * @skb: buffer to process 3751 * 3752 * If there is no free memory -ENOMEM is returned, otherwise zero 3753 * is returned and the old skb data released. 3754 */ 3755 static inline int skb_linearize_cow(struct sk_buff *skb) 3756 { 3757 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 3758 __skb_linearize(skb) : 0; 3759 } 3760 3761 static __always_inline void 3762 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3763 unsigned int off) 3764 { 3765 if (skb->ip_summed == CHECKSUM_COMPLETE) 3766 skb->csum = csum_block_sub(skb->csum, 3767 csum_partial(start, len, 0), off); 3768 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3769 skb_checksum_start_offset(skb) < 0) 3770 skb->ip_summed = CHECKSUM_NONE; 3771 } 3772 3773 /** 3774 * skb_postpull_rcsum - update checksum for received skb after pull 3775 * @skb: buffer to update 3776 * @start: start of data before pull 3777 * @len: length of data pulled 3778 * 3779 * After doing a pull on a received packet, you need to call this to 3780 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 3781 * CHECKSUM_NONE so that it can be recomputed from scratch. 3782 */ 3783 static inline void skb_postpull_rcsum(struct sk_buff *skb, 3784 const void *start, unsigned int len) 3785 { 3786 if (skb->ip_summed == CHECKSUM_COMPLETE) 3787 skb->csum = wsum_negate(csum_partial(start, len, 3788 wsum_negate(skb->csum))); 3789 else if (skb->ip_summed == CHECKSUM_PARTIAL && 3790 skb_checksum_start_offset(skb) < 0) 3791 skb->ip_summed = CHECKSUM_NONE; 3792 } 3793 3794 static __always_inline void 3795 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, 3796 unsigned int off) 3797 { 3798 if (skb->ip_summed == CHECKSUM_COMPLETE) 3799 skb->csum = csum_block_add(skb->csum, 3800 csum_partial(start, len, 0), off); 3801 } 3802 3803 /** 3804 * skb_postpush_rcsum - update checksum for received skb after push 3805 * @skb: buffer to update 3806 * @start: start of data after push 3807 * @len: length of data pushed 3808 * 3809 * After doing a push on a received packet, you need to call this to 3810 * update the CHECKSUM_COMPLETE checksum. 3811 */ 3812 static inline void skb_postpush_rcsum(struct sk_buff *skb, 3813 const void *start, unsigned int len) 3814 { 3815 __skb_postpush_rcsum(skb, start, len, 0); 3816 } 3817 3818 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 3819 3820 /** 3821 * skb_push_rcsum - push skb and update receive checksum 3822 * @skb: buffer to update 3823 * @len: length of data pulled 3824 * 3825 * This function performs an skb_push on the packet and updates 3826 * the CHECKSUM_COMPLETE checksum. It should be used on 3827 * receive path processing instead of skb_push unless you know 3828 * that the checksum difference is zero (e.g., a valid IP header) 3829 * or you are setting ip_summed to CHECKSUM_NONE. 3830 */ 3831 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) 3832 { 3833 skb_push(skb, len); 3834 skb_postpush_rcsum(skb, skb->data, len); 3835 return skb->data; 3836 } 3837 3838 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); 3839 /** 3840 * pskb_trim_rcsum - trim received skb and update checksum 3841 * @skb: buffer to trim 3842 * @len: new length 3843 * 3844 * This is exactly the same as pskb_trim except that it ensures the 3845 * checksum of received packets are still valid after the operation. 3846 * It can change skb pointers. 3847 */ 3848 3849 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3850 { 3851 if (likely(len >= skb->len)) 3852 return 0; 3853 return pskb_trim_rcsum_slow(skb, len); 3854 } 3855 3856 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) 3857 { 3858 if (skb->ip_summed == CHECKSUM_COMPLETE) 3859 skb->ip_summed = CHECKSUM_NONE; 3860 __skb_trim(skb, len); 3861 return 0; 3862 } 3863 3864 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) 3865 { 3866 if (skb->ip_summed == CHECKSUM_COMPLETE) 3867 skb->ip_summed = CHECKSUM_NONE; 3868 return __skb_grow(skb, len); 3869 } 3870 3871 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) 3872 #define skb_rb_first(root) rb_to_skb(rb_first(root)) 3873 #define skb_rb_last(root) rb_to_skb(rb_last(root)) 3874 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) 3875 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) 3876 3877 #define skb_queue_walk(queue, skb) \ 3878 for (skb = (queue)->next; \ 3879 skb != (struct sk_buff *)(queue); \ 3880 skb = skb->next) 3881 3882 #define skb_queue_walk_safe(queue, skb, tmp) \ 3883 for (skb = (queue)->next, tmp = skb->next; \ 3884 skb != (struct sk_buff *)(queue); \ 3885 skb = tmp, tmp = skb->next) 3886 3887 #define skb_queue_walk_from(queue, skb) \ 3888 for (; skb != (struct sk_buff *)(queue); \ 3889 skb = skb->next) 3890 3891 #define skb_rbtree_walk(skb, root) \ 3892 for (skb = skb_rb_first(root); skb != NULL; \ 3893 skb = skb_rb_next(skb)) 3894 3895 #define skb_rbtree_walk_from(skb) \ 3896 for (; skb != NULL; \ 3897 skb = skb_rb_next(skb)) 3898 3899 #define skb_rbtree_walk_from_safe(skb, tmp) \ 3900 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ 3901 skb = tmp) 3902 3903 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 3904 for (tmp = skb->next; \ 3905 skb != (struct sk_buff *)(queue); \ 3906 skb = tmp, tmp = skb->next) 3907 3908 #define skb_queue_reverse_walk(queue, skb) \ 3909 for (skb = (queue)->prev; \ 3910 skb != (struct sk_buff *)(queue); \ 3911 skb = skb->prev) 3912 3913 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 3914 for (skb = (queue)->prev, tmp = skb->prev; \ 3915 skb != (struct sk_buff *)(queue); \ 3916 skb = tmp, tmp = skb->prev) 3917 3918 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 3919 for (tmp = skb->prev; \ 3920 skb != (struct sk_buff *)(queue); \ 3921 skb = tmp, tmp = skb->prev) 3922 3923 static inline bool skb_has_frag_list(const struct sk_buff *skb) 3924 { 3925 return skb_shinfo(skb)->frag_list != NULL; 3926 } 3927 3928 static inline void skb_frag_list_init(struct sk_buff *skb) 3929 { 3930 skb_shinfo(skb)->frag_list = NULL; 3931 } 3932 3933 #define skb_walk_frags(skb, iter) \ 3934 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 3935 3936 3937 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, 3938 int *err, long *timeo_p, 3939 const struct sk_buff *skb); 3940 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, 3941 struct sk_buff_head *queue, 3942 unsigned int flags, 3943 int *off, int *err, 3944 struct sk_buff **last); 3945 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 3946 struct sk_buff_head *queue, 3947 unsigned int flags, int *off, int *err, 3948 struct sk_buff **last); 3949 struct sk_buff *__skb_recv_datagram(struct sock *sk, 3950 struct sk_buff_head *sk_queue, 3951 unsigned int flags, int *off, int *err); 3952 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err); 3953 __poll_t datagram_poll(struct file *file, struct socket *sock, 3954 struct poll_table_struct *wait); 3955 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 3956 struct iov_iter *to, int size); 3957 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 3958 struct msghdr *msg, int size) 3959 { 3960 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 3961 } 3962 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 3963 struct msghdr *msg); 3964 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, 3965 struct iov_iter *to, int len, 3966 struct ahash_request *hash); 3967 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 3968 struct iov_iter *from, int len); 3969 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 3970 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 3971 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); 3972 static inline void skb_free_datagram_locked(struct sock *sk, 3973 struct sk_buff *skb) 3974 { 3975 __skb_free_datagram_locked(sk, skb, 0); 3976 } 3977 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 3978 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 3979 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 3980 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 3981 int len); 3982 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 3983 struct pipe_inode_info *pipe, unsigned int len, 3984 unsigned int flags); 3985 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, 3986 int len); 3987 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len); 3988 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 3989 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 3990 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 3991 int len, int hlen); 3992 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 3993 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 3994 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 3995 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu); 3996 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len); 3997 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 3998 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, 3999 unsigned int offset); 4000 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 4001 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len); 4002 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); 4003 int skb_vlan_pop(struct sk_buff *skb); 4004 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 4005 int skb_eth_pop(struct sk_buff *skb); 4006 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, 4007 const unsigned char *src); 4008 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, 4009 int mac_len, bool ethernet); 4010 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, 4011 bool ethernet); 4012 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); 4013 int skb_mpls_dec_ttl(struct sk_buff *skb); 4014 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, 4015 gfp_t gfp); 4016 4017 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 4018 { 4019 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; 4020 } 4021 4022 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 4023 { 4024 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 4025 } 4026 4027 struct skb_checksum_ops { 4028 __wsum (*update)(const void *mem, int len, __wsum wsum); 4029 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 4030 }; 4031 4032 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; 4033 4034 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 4035 __wsum csum, const struct skb_checksum_ops *ops); 4036 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 4037 __wsum csum); 4038 4039 static inline void * __must_check 4040 __skb_header_pointer(const struct sk_buff *skb, int offset, int len, 4041 const void *data, int hlen, void *buffer) 4042 { 4043 if (likely(hlen - offset >= len)) 4044 return (void *)data + offset; 4045 4046 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) 4047 return NULL; 4048 4049 return buffer; 4050 } 4051 4052 static inline void * __must_check 4053 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 4054 { 4055 return __skb_header_pointer(skb, offset, len, skb->data, 4056 skb_headlen(skb), buffer); 4057 } 4058 4059 /** 4060 * skb_needs_linearize - check if we need to linearize a given skb 4061 * depending on the given device features. 4062 * @skb: socket buffer to check 4063 * @features: net device features 4064 * 4065 * Returns true if either: 4066 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 4067 * 2. skb is fragmented and the device does not support SG. 4068 */ 4069 static inline bool skb_needs_linearize(struct sk_buff *skb, 4070 netdev_features_t features) 4071 { 4072 return skb_is_nonlinear(skb) && 4073 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 4074 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 4075 } 4076 4077 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 4078 void *to, 4079 const unsigned int len) 4080 { 4081 memcpy(to, skb->data, len); 4082 } 4083 4084 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 4085 const int offset, void *to, 4086 const unsigned int len) 4087 { 4088 memcpy(to, skb->data + offset, len); 4089 } 4090 4091 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 4092 const void *from, 4093 const unsigned int len) 4094 { 4095 memcpy(skb->data, from, len); 4096 } 4097 4098 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 4099 const int offset, 4100 const void *from, 4101 const unsigned int len) 4102 { 4103 memcpy(skb->data + offset, from, len); 4104 } 4105 4106 void skb_init(void); 4107 4108 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 4109 { 4110 return skb->tstamp; 4111 } 4112 4113 /** 4114 * skb_get_timestamp - get timestamp from a skb 4115 * @skb: skb to get stamp from 4116 * @stamp: pointer to struct __kernel_old_timeval to store stamp in 4117 * 4118 * Timestamps are stored in the skb as offsets to a base timestamp. 4119 * This function converts the offset back to a struct timeval and stores 4120 * it in stamp. 4121 */ 4122 static inline void skb_get_timestamp(const struct sk_buff *skb, 4123 struct __kernel_old_timeval *stamp) 4124 { 4125 *stamp = ns_to_kernel_old_timeval(skb->tstamp); 4126 } 4127 4128 static inline void skb_get_new_timestamp(const struct sk_buff *skb, 4129 struct __kernel_sock_timeval *stamp) 4130 { 4131 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 4132 4133 stamp->tv_sec = ts.tv_sec; 4134 stamp->tv_usec = ts.tv_nsec / 1000; 4135 } 4136 4137 static inline void skb_get_timestampns(const struct sk_buff *skb, 4138 struct __kernel_old_timespec *stamp) 4139 { 4140 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 4141 4142 stamp->tv_sec = ts.tv_sec; 4143 stamp->tv_nsec = ts.tv_nsec; 4144 } 4145 4146 static inline void skb_get_new_timestampns(const struct sk_buff *skb, 4147 struct __kernel_timespec *stamp) 4148 { 4149 struct timespec64 ts = ktime_to_timespec64(skb->tstamp); 4150 4151 stamp->tv_sec = ts.tv_sec; 4152 stamp->tv_nsec = ts.tv_nsec; 4153 } 4154 4155 static inline void __net_timestamp(struct sk_buff *skb) 4156 { 4157 skb->tstamp = ktime_get_real(); 4158 skb->mono_delivery_time = 0; 4159 } 4160 4161 static inline ktime_t net_timedelta(ktime_t t) 4162 { 4163 return ktime_sub(ktime_get_real(), t); 4164 } 4165 4166 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt, 4167 bool mono) 4168 { 4169 skb->tstamp = kt; 4170 skb->mono_delivery_time = kt && mono; 4171 } 4172 4173 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key); 4174 4175 /* It is used in the ingress path to clear the delivery_time. 4176 * If needed, set the skb->tstamp to the (rcv) timestamp. 4177 */ 4178 static inline void skb_clear_delivery_time(struct sk_buff *skb) 4179 { 4180 if (skb->mono_delivery_time) { 4181 skb->mono_delivery_time = 0; 4182 if (static_branch_unlikely(&netstamp_needed_key)) 4183 skb->tstamp = ktime_get_real(); 4184 else 4185 skb->tstamp = 0; 4186 } 4187 } 4188 4189 static inline void skb_clear_tstamp(struct sk_buff *skb) 4190 { 4191 if (skb->mono_delivery_time) 4192 return; 4193 4194 skb->tstamp = 0; 4195 } 4196 4197 static inline ktime_t skb_tstamp(const struct sk_buff *skb) 4198 { 4199 if (skb->mono_delivery_time) 4200 return 0; 4201 4202 return skb->tstamp; 4203 } 4204 4205 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond) 4206 { 4207 if (!skb->mono_delivery_time && skb->tstamp) 4208 return skb->tstamp; 4209 4210 if (static_branch_unlikely(&netstamp_needed_key) || cond) 4211 return ktime_get_real(); 4212 4213 return 0; 4214 } 4215 4216 static inline u8 skb_metadata_len(const struct sk_buff *skb) 4217 { 4218 return skb_shinfo(skb)->meta_len; 4219 } 4220 4221 static inline void *skb_metadata_end(const struct sk_buff *skb) 4222 { 4223 return skb_mac_header(skb); 4224 } 4225 4226 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, 4227 const struct sk_buff *skb_b, 4228 u8 meta_len) 4229 { 4230 const void *a = skb_metadata_end(skb_a); 4231 const void *b = skb_metadata_end(skb_b); 4232 /* Using more efficient varaiant than plain call to memcmp(). */ 4233 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 4234 u64 diffs = 0; 4235 4236 switch (meta_len) { 4237 #define __it(x, op) (x -= sizeof(u##op)) 4238 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) 4239 case 32: diffs |= __it_diff(a, b, 64); 4240 fallthrough; 4241 case 24: diffs |= __it_diff(a, b, 64); 4242 fallthrough; 4243 case 16: diffs |= __it_diff(a, b, 64); 4244 fallthrough; 4245 case 8: diffs |= __it_diff(a, b, 64); 4246 break; 4247 case 28: diffs |= __it_diff(a, b, 64); 4248 fallthrough; 4249 case 20: diffs |= __it_diff(a, b, 64); 4250 fallthrough; 4251 case 12: diffs |= __it_diff(a, b, 64); 4252 fallthrough; 4253 case 4: diffs |= __it_diff(a, b, 32); 4254 break; 4255 } 4256 return diffs; 4257 #else 4258 return memcmp(a - meta_len, b - meta_len, meta_len); 4259 #endif 4260 } 4261 4262 static inline bool skb_metadata_differs(const struct sk_buff *skb_a, 4263 const struct sk_buff *skb_b) 4264 { 4265 u8 len_a = skb_metadata_len(skb_a); 4266 u8 len_b = skb_metadata_len(skb_b); 4267 4268 if (!(len_a | len_b)) 4269 return false; 4270 4271 return len_a != len_b ? 4272 true : __skb_metadata_differs(skb_a, skb_b, len_a); 4273 } 4274 4275 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) 4276 { 4277 skb_shinfo(skb)->meta_len = meta_len; 4278 } 4279 4280 static inline void skb_metadata_clear(struct sk_buff *skb) 4281 { 4282 skb_metadata_set(skb, 0); 4283 } 4284 4285 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 4286 4287 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 4288 4289 void skb_clone_tx_timestamp(struct sk_buff *skb); 4290 bool skb_defer_rx_timestamp(struct sk_buff *skb); 4291 4292 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 4293 4294 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 4295 { 4296 } 4297 4298 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 4299 { 4300 return false; 4301 } 4302 4303 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 4304 4305 /** 4306 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 4307 * 4308 * PHY drivers may accept clones of transmitted packets for 4309 * timestamping via their phy_driver.txtstamp method. These drivers 4310 * must call this function to return the skb back to the stack with a 4311 * timestamp. 4312 * 4313 * @skb: clone of the original outgoing packet 4314 * @hwtstamps: hardware time stamps 4315 * 4316 */ 4317 void skb_complete_tx_timestamp(struct sk_buff *skb, 4318 struct skb_shared_hwtstamps *hwtstamps); 4319 4320 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, 4321 struct skb_shared_hwtstamps *hwtstamps, 4322 struct sock *sk, int tstype); 4323 4324 /** 4325 * skb_tstamp_tx - queue clone of skb with send time stamps 4326 * @orig_skb: the original outgoing packet 4327 * @hwtstamps: hardware time stamps, may be NULL if not available 4328 * 4329 * If the skb has a socket associated, then this function clones the 4330 * skb (thus sharing the actual data and optional structures), stores 4331 * the optional hardware time stamping information (if non NULL) or 4332 * generates a software time stamp (otherwise), then queues the clone 4333 * to the error queue of the socket. Errors are silently ignored. 4334 */ 4335 void skb_tstamp_tx(struct sk_buff *orig_skb, 4336 struct skb_shared_hwtstamps *hwtstamps); 4337 4338 /** 4339 * skb_tx_timestamp() - Driver hook for transmit timestamping 4340 * 4341 * Ethernet MAC Drivers should call this function in their hard_xmit() 4342 * function immediately before giving the sk_buff to the MAC hardware. 4343 * 4344 * Specifically, one should make absolutely sure that this function is 4345 * called before TX completion of this packet can trigger. Otherwise 4346 * the packet could potentially already be freed. 4347 * 4348 * @skb: A socket buffer. 4349 */ 4350 static inline void skb_tx_timestamp(struct sk_buff *skb) 4351 { 4352 skb_clone_tx_timestamp(skb); 4353 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) 4354 skb_tstamp_tx(skb, NULL); 4355 } 4356 4357 /** 4358 * skb_complete_wifi_ack - deliver skb with wifi status 4359 * 4360 * @skb: the original outgoing packet 4361 * @acked: ack status 4362 * 4363 */ 4364 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 4365 4366 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 4367 __sum16 __skb_checksum_complete(struct sk_buff *skb); 4368 4369 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 4370 { 4371 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 4372 skb->csum_valid || 4373 (skb->ip_summed == CHECKSUM_PARTIAL && 4374 skb_checksum_start_offset(skb) >= 0)); 4375 } 4376 4377 /** 4378 * skb_checksum_complete - Calculate checksum of an entire packet 4379 * @skb: packet to process 4380 * 4381 * This function calculates the checksum over the entire packet plus 4382 * the value of skb->csum. The latter can be used to supply the 4383 * checksum of a pseudo header as used by TCP/UDP. It returns the 4384 * checksum. 4385 * 4386 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 4387 * this function can be used to verify that checksum on received 4388 * packets. In that case the function should return zero if the 4389 * checksum is correct. In particular, this function will return zero 4390 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 4391 * hardware has already verified the correctness of the checksum. 4392 */ 4393 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 4394 { 4395 return skb_csum_unnecessary(skb) ? 4396 0 : __skb_checksum_complete(skb); 4397 } 4398 4399 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 4400 { 4401 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4402 if (skb->csum_level == 0) 4403 skb->ip_summed = CHECKSUM_NONE; 4404 else 4405 skb->csum_level--; 4406 } 4407 } 4408 4409 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 4410 { 4411 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4412 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 4413 skb->csum_level++; 4414 } else if (skb->ip_summed == CHECKSUM_NONE) { 4415 skb->ip_summed = CHECKSUM_UNNECESSARY; 4416 skb->csum_level = 0; 4417 } 4418 } 4419 4420 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) 4421 { 4422 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 4423 skb->ip_summed = CHECKSUM_NONE; 4424 skb->csum_level = 0; 4425 } 4426 } 4427 4428 /* Check if we need to perform checksum complete validation. 4429 * 4430 * Returns true if checksum complete is needed, false otherwise 4431 * (either checksum is unnecessary or zero checksum is allowed). 4432 */ 4433 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 4434 bool zero_okay, 4435 __sum16 check) 4436 { 4437 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 4438 skb->csum_valid = 1; 4439 __skb_decr_checksum_unnecessary(skb); 4440 return false; 4441 } 4442 4443 return true; 4444 } 4445 4446 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly 4447 * in checksum_init. 4448 */ 4449 #define CHECKSUM_BREAK 76 4450 4451 /* Unset checksum-complete 4452 * 4453 * Unset checksum complete can be done when packet is being modified 4454 * (uncompressed for instance) and checksum-complete value is 4455 * invalidated. 4456 */ 4457 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 4458 { 4459 if (skb->ip_summed == CHECKSUM_COMPLETE) 4460 skb->ip_summed = CHECKSUM_NONE; 4461 } 4462 4463 /* Validate (init) checksum based on checksum complete. 4464 * 4465 * Return values: 4466 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 4467 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 4468 * checksum is stored in skb->csum for use in __skb_checksum_complete 4469 * non-zero: value of invalid checksum 4470 * 4471 */ 4472 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 4473 bool complete, 4474 __wsum psum) 4475 { 4476 if (skb->ip_summed == CHECKSUM_COMPLETE) { 4477 if (!csum_fold(csum_add(psum, skb->csum))) { 4478 skb->csum_valid = 1; 4479 return 0; 4480 } 4481 } 4482 4483 skb->csum = psum; 4484 4485 if (complete || skb->len <= CHECKSUM_BREAK) { 4486 __sum16 csum; 4487 4488 csum = __skb_checksum_complete(skb); 4489 skb->csum_valid = !csum; 4490 return csum; 4491 } 4492 4493 return 0; 4494 } 4495 4496 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 4497 { 4498 return 0; 4499 } 4500 4501 /* Perform checksum validate (init). Note that this is a macro since we only 4502 * want to calculate the pseudo header which is an input function if necessary. 4503 * First we try to validate without any computation (checksum unnecessary) and 4504 * then calculate based on checksum complete calling the function to compute 4505 * pseudo header. 4506 * 4507 * Return values: 4508 * 0: checksum is validated or try to in skb_checksum_complete 4509 * non-zero: value of invalid checksum 4510 */ 4511 #define __skb_checksum_validate(skb, proto, complete, \ 4512 zero_okay, check, compute_pseudo) \ 4513 ({ \ 4514 __sum16 __ret = 0; \ 4515 skb->csum_valid = 0; \ 4516 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 4517 __ret = __skb_checksum_validate_complete(skb, \ 4518 complete, compute_pseudo(skb, proto)); \ 4519 __ret; \ 4520 }) 4521 4522 #define skb_checksum_init(skb, proto, compute_pseudo) \ 4523 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 4524 4525 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 4526 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 4527 4528 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 4529 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 4530 4531 #define skb_checksum_validate_zero_check(skb, proto, check, \ 4532 compute_pseudo) \ 4533 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 4534 4535 #define skb_checksum_simple_validate(skb) \ 4536 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 4537 4538 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 4539 { 4540 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); 4541 } 4542 4543 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) 4544 { 4545 skb->csum = ~pseudo; 4546 skb->ip_summed = CHECKSUM_COMPLETE; 4547 } 4548 4549 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ 4550 do { \ 4551 if (__skb_checksum_convert_check(skb)) \ 4552 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ 4553 } while (0) 4554 4555 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 4556 u16 start, u16 offset) 4557 { 4558 skb->ip_summed = CHECKSUM_PARTIAL; 4559 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 4560 skb->csum_offset = offset - start; 4561 } 4562 4563 /* Update skbuf and packet to reflect the remote checksum offload operation. 4564 * When called, ptr indicates the starting point for skb->csum when 4565 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 4566 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 4567 */ 4568 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 4569 int start, int offset, bool nopartial) 4570 { 4571 __wsum delta; 4572 4573 if (!nopartial) { 4574 skb_remcsum_adjust_partial(skb, ptr, start, offset); 4575 return; 4576 } 4577 4578 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 4579 __skb_checksum_complete(skb); 4580 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 4581 } 4582 4583 delta = remcsum_adjust(ptr, skb->csum, start, offset); 4584 4585 /* Adjust skb->csum since we changed the packet */ 4586 skb->csum = csum_add(skb->csum, delta); 4587 } 4588 4589 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) 4590 { 4591 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4592 return (void *)(skb->_nfct & NFCT_PTRMASK); 4593 #else 4594 return NULL; 4595 #endif 4596 } 4597 4598 static inline unsigned long skb_get_nfct(const struct sk_buff *skb) 4599 { 4600 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4601 return skb->_nfct; 4602 #else 4603 return 0UL; 4604 #endif 4605 } 4606 4607 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) 4608 { 4609 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 4610 skb->slow_gro |= !!nfct; 4611 skb->_nfct = nfct; 4612 #endif 4613 } 4614 4615 #ifdef CONFIG_SKB_EXTENSIONS 4616 enum skb_ext_id { 4617 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 4618 SKB_EXT_BRIDGE_NF, 4619 #endif 4620 #ifdef CONFIG_XFRM 4621 SKB_EXT_SEC_PATH, 4622 #endif 4623 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) 4624 TC_SKB_EXT, 4625 #endif 4626 #if IS_ENABLED(CONFIG_MPTCP) 4627 SKB_EXT_MPTCP, 4628 #endif 4629 #if IS_ENABLED(CONFIG_MCTP_FLOWS) 4630 SKB_EXT_MCTP, 4631 #endif 4632 SKB_EXT_NUM, /* must be last */ 4633 }; 4634 4635 /** 4636 * struct skb_ext - sk_buff extensions 4637 * @refcnt: 1 on allocation, deallocated on 0 4638 * @offset: offset to add to @data to obtain extension address 4639 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units 4640 * @data: start of extension data, variable sized 4641 * 4642 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows 4643 * to use 'u8' types while allowing up to 2kb worth of extension data. 4644 */ 4645 struct skb_ext { 4646 refcount_t refcnt; 4647 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ 4648 u8 chunks; /* same */ 4649 char data[] __aligned(8); 4650 }; 4651 4652 struct skb_ext *__skb_ext_alloc(gfp_t flags); 4653 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, 4654 struct skb_ext *ext); 4655 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); 4656 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); 4657 void __skb_ext_put(struct skb_ext *ext); 4658 4659 static inline void skb_ext_put(struct sk_buff *skb) 4660 { 4661 if (skb->active_extensions) 4662 __skb_ext_put(skb->extensions); 4663 } 4664 4665 static inline void __skb_ext_copy(struct sk_buff *dst, 4666 const struct sk_buff *src) 4667 { 4668 dst->active_extensions = src->active_extensions; 4669 4670 if (src->active_extensions) { 4671 struct skb_ext *ext = src->extensions; 4672 4673 refcount_inc(&ext->refcnt); 4674 dst->extensions = ext; 4675 } 4676 } 4677 4678 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) 4679 { 4680 skb_ext_put(dst); 4681 __skb_ext_copy(dst, src); 4682 } 4683 4684 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) 4685 { 4686 return !!ext->offset[i]; 4687 } 4688 4689 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) 4690 { 4691 return skb->active_extensions & (1 << id); 4692 } 4693 4694 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) 4695 { 4696 if (skb_ext_exist(skb, id)) 4697 __skb_ext_del(skb, id); 4698 } 4699 4700 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) 4701 { 4702 if (skb_ext_exist(skb, id)) { 4703 struct skb_ext *ext = skb->extensions; 4704 4705 return (void *)ext + (ext->offset[id] << 3); 4706 } 4707 4708 return NULL; 4709 } 4710 4711 static inline void skb_ext_reset(struct sk_buff *skb) 4712 { 4713 if (unlikely(skb->active_extensions)) { 4714 __skb_ext_put(skb->extensions); 4715 skb->active_extensions = 0; 4716 } 4717 } 4718 4719 static inline bool skb_has_extensions(struct sk_buff *skb) 4720 { 4721 return unlikely(skb->active_extensions); 4722 } 4723 #else 4724 static inline void skb_ext_put(struct sk_buff *skb) {} 4725 static inline void skb_ext_reset(struct sk_buff *skb) {} 4726 static inline void skb_ext_del(struct sk_buff *skb, int unused) {} 4727 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} 4728 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} 4729 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } 4730 #endif /* CONFIG_SKB_EXTENSIONS */ 4731 4732 static inline void nf_reset_ct(struct sk_buff *skb) 4733 { 4734 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4735 nf_conntrack_put(skb_nfct(skb)); 4736 skb->_nfct = 0; 4737 #endif 4738 } 4739 4740 static inline void nf_reset_trace(struct sk_buff *skb) 4741 { 4742 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) 4743 skb->nf_trace = 0; 4744 #endif 4745 } 4746 4747 static inline void ipvs_reset(struct sk_buff *skb) 4748 { 4749 #if IS_ENABLED(CONFIG_IP_VS) 4750 skb->ipvs_property = 0; 4751 #endif 4752 } 4753 4754 /* Note: This doesn't put any conntrack info in dst. */ 4755 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 4756 bool copy) 4757 { 4758 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4759 dst->_nfct = src->_nfct; 4760 nf_conntrack_get(skb_nfct(src)); 4761 #endif 4762 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) 4763 if (copy) 4764 dst->nf_trace = src->nf_trace; 4765 #endif 4766 } 4767 4768 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 4769 { 4770 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 4771 nf_conntrack_put(skb_nfct(dst)); 4772 #endif 4773 dst->slow_gro = src->slow_gro; 4774 __nf_copy(dst, src, true); 4775 } 4776 4777 #ifdef CONFIG_NETWORK_SECMARK 4778 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4779 { 4780 to->secmark = from->secmark; 4781 } 4782 4783 static inline void skb_init_secmark(struct sk_buff *skb) 4784 { 4785 skb->secmark = 0; 4786 } 4787 #else 4788 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 4789 { } 4790 4791 static inline void skb_init_secmark(struct sk_buff *skb) 4792 { } 4793 #endif 4794 4795 static inline int secpath_exists(const struct sk_buff *skb) 4796 { 4797 #ifdef CONFIG_XFRM 4798 return skb_ext_exist(skb, SKB_EXT_SEC_PATH); 4799 #else 4800 return 0; 4801 #endif 4802 } 4803 4804 static inline bool skb_irq_freeable(const struct sk_buff *skb) 4805 { 4806 return !skb->destructor && 4807 !secpath_exists(skb) && 4808 !skb_nfct(skb) && 4809 !skb->_skb_refdst && 4810 !skb_has_frag_list(skb); 4811 } 4812 4813 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 4814 { 4815 skb->queue_mapping = queue_mapping; 4816 } 4817 4818 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 4819 { 4820 return skb->queue_mapping; 4821 } 4822 4823 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 4824 { 4825 to->queue_mapping = from->queue_mapping; 4826 } 4827 4828 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 4829 { 4830 skb->queue_mapping = rx_queue + 1; 4831 } 4832 4833 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 4834 { 4835 return skb->queue_mapping - 1; 4836 } 4837 4838 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 4839 { 4840 return skb->queue_mapping != 0; 4841 } 4842 4843 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) 4844 { 4845 skb->dst_pending_confirm = val; 4846 } 4847 4848 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) 4849 { 4850 return skb->dst_pending_confirm != 0; 4851 } 4852 4853 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) 4854 { 4855 #ifdef CONFIG_XFRM 4856 return skb_ext_find(skb, SKB_EXT_SEC_PATH); 4857 #else 4858 return NULL; 4859 #endif 4860 } 4861 4862 /* Keeps track of mac header offset relative to skb->head. 4863 * It is useful for TSO of Tunneling protocol. e.g. GRE. 4864 * For non-tunnel skb it points to skb_mac_header() and for 4865 * tunnel skb it points to outer mac header. 4866 * Keeps track of level of encapsulation of network headers. 4867 */ 4868 struct skb_gso_cb { 4869 union { 4870 int mac_offset; 4871 int data_offset; 4872 }; 4873 int encap_level; 4874 __wsum csum; 4875 __u16 csum_start; 4876 }; 4877 #define SKB_GSO_CB_OFFSET 32 4878 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET)) 4879 4880 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 4881 { 4882 return (skb_mac_header(inner_skb) - inner_skb->head) - 4883 SKB_GSO_CB(inner_skb)->mac_offset; 4884 } 4885 4886 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 4887 { 4888 int new_headroom, headroom; 4889 int ret; 4890 4891 headroom = skb_headroom(skb); 4892 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 4893 if (ret) 4894 return ret; 4895 4896 new_headroom = skb_headroom(skb); 4897 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 4898 return 0; 4899 } 4900 4901 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) 4902 { 4903 /* Do not update partial checksums if remote checksum is enabled. */ 4904 if (skb->remcsum_offload) 4905 return; 4906 4907 SKB_GSO_CB(skb)->csum = res; 4908 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; 4909 } 4910 4911 /* Compute the checksum for a gso segment. First compute the checksum value 4912 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 4913 * then add in skb->csum (checksum from csum_start to end of packet). 4914 * skb->csum and csum_start are then updated to reflect the checksum of the 4915 * resultant packet starting from the transport header-- the resultant checksum 4916 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 4917 * header. 4918 */ 4919 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 4920 { 4921 unsigned char *csum_start = skb_transport_header(skb); 4922 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; 4923 __wsum partial = SKB_GSO_CB(skb)->csum; 4924 4925 SKB_GSO_CB(skb)->csum = res; 4926 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; 4927 4928 return csum_fold(csum_partial(csum_start, plen, partial)); 4929 } 4930 4931 static inline bool skb_is_gso(const struct sk_buff *skb) 4932 { 4933 return skb_shinfo(skb)->gso_size; 4934 } 4935 4936 /* Note: Should be called only if skb_is_gso(skb) is true */ 4937 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 4938 { 4939 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 4940 } 4941 4942 /* Note: Should be called only if skb_is_gso(skb) is true */ 4943 static inline bool skb_is_gso_sctp(const struct sk_buff *skb) 4944 { 4945 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; 4946 } 4947 4948 /* Note: Should be called only if skb_is_gso(skb) is true */ 4949 static inline bool skb_is_gso_tcp(const struct sk_buff *skb) 4950 { 4951 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); 4952 } 4953 4954 static inline void skb_gso_reset(struct sk_buff *skb) 4955 { 4956 skb_shinfo(skb)->gso_size = 0; 4957 skb_shinfo(skb)->gso_segs = 0; 4958 skb_shinfo(skb)->gso_type = 0; 4959 } 4960 4961 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, 4962 u16 increment) 4963 { 4964 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4965 return; 4966 shinfo->gso_size += increment; 4967 } 4968 4969 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, 4970 u16 decrement) 4971 { 4972 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) 4973 return; 4974 shinfo->gso_size -= decrement; 4975 } 4976 4977 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 4978 4979 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 4980 { 4981 /* LRO sets gso_size but not gso_type, whereas if GSO is really 4982 * wanted then gso_type will be set. */ 4983 const struct skb_shared_info *shinfo = skb_shinfo(skb); 4984 4985 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 4986 unlikely(shinfo->gso_type == 0)) { 4987 __skb_warn_lro_forwarding(skb); 4988 return true; 4989 } 4990 return false; 4991 } 4992 4993 static inline void skb_forward_csum(struct sk_buff *skb) 4994 { 4995 /* Unfortunately we don't support this one. Any brave souls? */ 4996 if (skb->ip_summed == CHECKSUM_COMPLETE) 4997 skb->ip_summed = CHECKSUM_NONE; 4998 } 4999 5000 /** 5001 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 5002 * @skb: skb to check 5003 * 5004 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 5005 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 5006 * use this helper, to document places where we make this assertion. 5007 */ 5008 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 5009 { 5010 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE); 5011 } 5012 5013 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 5014 5015 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 5016 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 5017 unsigned int transport_len, 5018 __sum16(*skb_chkf)(struct sk_buff *skb)); 5019 5020 /** 5021 * skb_head_is_locked - Determine if the skb->head is locked down 5022 * @skb: skb to check 5023 * 5024 * The head on skbs build around a head frag can be removed if they are 5025 * not cloned. This function returns true if the skb head is locked down 5026 * due to either being allocated via kmalloc, or by being a clone with 5027 * multiple references to the head. 5028 */ 5029 static inline bool skb_head_is_locked(const struct sk_buff *skb) 5030 { 5031 return !skb->head_frag || skb_cloned(skb); 5032 } 5033 5034 /* Local Checksum Offload. 5035 * Compute outer checksum based on the assumption that the 5036 * inner checksum will be offloaded later. 5037 * See Documentation/networking/checksum-offloads.rst for 5038 * explanation of how this works. 5039 * Fill in outer checksum adjustment (e.g. with sum of outer 5040 * pseudo-header) before calling. 5041 * Also ensure that inner checksum is in linear data area. 5042 */ 5043 static inline __wsum lco_csum(struct sk_buff *skb) 5044 { 5045 unsigned char *csum_start = skb_checksum_start(skb); 5046 unsigned char *l4_hdr = skb_transport_header(skb); 5047 __wsum partial; 5048 5049 /* Start with complement of inner checksum adjustment */ 5050 partial = ~csum_unfold(*(__force __sum16 *)(csum_start + 5051 skb->csum_offset)); 5052 5053 /* Add in checksum of our headers (incl. outer checksum 5054 * adjustment filled in by caller) and return result. 5055 */ 5056 return csum_partial(l4_hdr, csum_start - l4_hdr, partial); 5057 } 5058 5059 static inline bool skb_is_redirected(const struct sk_buff *skb) 5060 { 5061 return skb->redirected; 5062 } 5063 5064 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) 5065 { 5066 skb->redirected = 1; 5067 #ifdef CONFIG_NET_REDIRECT 5068 skb->from_ingress = from_ingress; 5069 if (skb->from_ingress) 5070 skb_clear_tstamp(skb); 5071 #endif 5072 } 5073 5074 static inline void skb_reset_redirect(struct sk_buff *skb) 5075 { 5076 skb->redirected = 0; 5077 } 5078 5079 static inline void skb_set_redirected_noclear(struct sk_buff *skb, 5080 bool from_ingress) 5081 { 5082 skb->redirected = 1; 5083 #ifdef CONFIG_NET_REDIRECT 5084 skb->from_ingress = from_ingress; 5085 #endif 5086 } 5087 5088 static inline bool skb_csum_is_sctp(struct sk_buff *skb) 5089 { 5090 #if IS_ENABLED(CONFIG_IP_SCTP) 5091 return skb->csum_not_inet; 5092 #else 5093 return 0; 5094 #endif 5095 } 5096 5097 static inline void skb_reset_csum_not_inet(struct sk_buff *skb) 5098 { 5099 skb->ip_summed = CHECKSUM_NONE; 5100 #if IS_ENABLED(CONFIG_IP_SCTP) 5101 skb->csum_not_inet = 0; 5102 #endif 5103 } 5104 5105 static inline void skb_set_kcov_handle(struct sk_buff *skb, 5106 const u64 kcov_handle) 5107 { 5108 #ifdef CONFIG_KCOV 5109 skb->kcov_handle = kcov_handle; 5110 #endif 5111 } 5112 5113 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) 5114 { 5115 #ifdef CONFIG_KCOV 5116 return skb->kcov_handle; 5117 #else 5118 return 0; 5119 #endif 5120 } 5121 5122 static inline void skb_mark_for_recycle(struct sk_buff *skb) 5123 { 5124 #ifdef CONFIG_PAGE_POOL 5125 skb->pp_recycle = 1; 5126 #endif 5127 } 5128 5129 #endif /* __KERNEL__ */ 5130 #endif /* _LINUX_SKBUFF_H */ 5131