1 /* 2 * Definitions for the 'struct sk_buff' memory handlers. 3 * 4 * Authors: 5 * Alan Cox, <[email protected]> 6 * Florian La Roche, <[email protected]> 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU General Public License 10 * as published by the Free Software Foundation; either version 11 * 2 of the License, or (at your option) any later version. 12 */ 13 14 #ifndef _LINUX_SKBUFF_H 15 #define _LINUX_SKBUFF_H 16 17 #include <linux/kernel.h> 18 #include <linux/kmemcheck.h> 19 #include <linux/compiler.h> 20 #include <linux/time.h> 21 #include <linux/bug.h> 22 #include <linux/cache.h> 23 #include <linux/rbtree.h> 24 #include <linux/socket.h> 25 26 #include <linux/atomic.h> 27 #include <asm/types.h> 28 #include <linux/spinlock.h> 29 #include <linux/net.h> 30 #include <linux/textsearch.h> 31 #include <net/checksum.h> 32 #include <linux/rcupdate.h> 33 #include <linux/hrtimer.h> 34 #include <linux/dma-mapping.h> 35 #include <linux/netdev_features.h> 36 #include <linux/sched.h> 37 #include <net/flow_dissector.h> 38 #include <linux/splice.h> 39 #include <linux/in6.h> 40 41 /* A. Checksumming of received packets by device. 42 * 43 * CHECKSUM_NONE: 44 * 45 * Device failed to checksum this packet e.g. due to lack of capabilities. 46 * The packet contains full (though not verified) checksum in packet but 47 * not in skb->csum. Thus, skb->csum is undefined in this case. 48 * 49 * CHECKSUM_UNNECESSARY: 50 * 51 * The hardware you're dealing with doesn't calculate the full checksum 52 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums 53 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY 54 * if their checksums are okay. skb->csum is still undefined in this case 55 * though. It is a bad option, but, unfortunately, nowadays most vendors do 56 * this. Apparently with the secret goal to sell you new devices, when you 57 * will add new protocol to your host, f.e. IPv6 8) 58 * 59 * CHECKSUM_UNNECESSARY is applicable to following protocols: 60 * TCP: IPv6 and IPv4. 61 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a 62 * zero UDP checksum for either IPv4 or IPv6, the networking stack 63 * may perform further validation in this case. 64 * GRE: only if the checksum is present in the header. 65 * SCTP: indicates the CRC in SCTP header has been validated. 66 * 67 * skb->csum_level indicates the number of consecutive checksums found in 68 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. 69 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet 70 * and a device is able to verify the checksums for UDP (possibly zero), 71 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to 72 * two. If the device were only able to verify the UDP checksum and not 73 * GRE, either because it doesn't support GRE checksum of because GRE 74 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is 75 * not considered in this case). 76 * 77 * CHECKSUM_COMPLETE: 78 * 79 * This is the most generic way. The device supplied checksum of the _whole_ 80 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the 81 * hardware doesn't need to parse L3/L4 headers to implement this. 82 * 83 * Note: Even if device supports only some protocols, but is able to produce 84 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. 85 * 86 * CHECKSUM_PARTIAL: 87 * 88 * A checksum is set up to be offloaded to a device as described in the 89 * output description for CHECKSUM_PARTIAL. This may occur on a packet 90 * received directly from another Linux OS, e.g., a virtualized Linux kernel 91 * on the same host, or it may be set in the input path in GRO or remote 92 * checksum offload. For the purposes of checksum verification, the checksum 93 * referred to by skb->csum_start + skb->csum_offset and any preceding 94 * checksums in the packet are considered verified. Any checksums in the 95 * packet that are after the checksum being offloaded are not considered to 96 * be verified. 97 * 98 * B. Checksumming on output. 99 * 100 * CHECKSUM_NONE: 101 * 102 * The skb was already checksummed by the protocol, or a checksum is not 103 * required. 104 * 105 * CHECKSUM_PARTIAL: 106 * 107 * The device is required to checksum the packet as seen by hard_start_xmit() 108 * from skb->csum_start up to the end, and to record/write the checksum at 109 * offset skb->csum_start + skb->csum_offset. 110 * 111 * The device must show its capabilities in dev->features, set up at device 112 * setup time, e.g. netdev_features.h: 113 * 114 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything. 115 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over 116 * IPv4. Sigh. Vendors like this way for an unknown reason. 117 * Though, see comment above about CHECKSUM_UNNECESSARY. 8) 118 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead. 119 * NETIF_F_... - Well, you get the picture. 120 * 121 * CHECKSUM_UNNECESSARY: 122 * 123 * Normally, the device will do per protocol specific checksumming. Protocol 124 * implementations that do not want the NIC to perform the checksum 125 * calculation should use this flag in their outgoing skbs. 126 * 127 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC 128 * offload. Correspondingly, the FCoE protocol driver 129 * stack should use CHECKSUM_UNNECESSARY. 130 * 131 * Any questions? No questions, good. --ANK 132 */ 133 134 /* Don't change this without changing skb_csum_unnecessary! */ 135 #define CHECKSUM_NONE 0 136 #define CHECKSUM_UNNECESSARY 1 137 #define CHECKSUM_COMPLETE 2 138 #define CHECKSUM_PARTIAL 3 139 140 /* Maximum value in skb->csum_level */ 141 #define SKB_MAX_CSUM_LEVEL 3 142 143 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) 144 #define SKB_WITH_OVERHEAD(X) \ 145 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 146 #define SKB_MAX_ORDER(X, ORDER) \ 147 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 148 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 149 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 150 151 /* return minimum truesize of one skb containing X bytes of data */ 152 #define SKB_TRUESIZE(X) ((X) + \ 153 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 154 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 155 156 struct net_device; 157 struct scatterlist; 158 struct pipe_inode_info; 159 struct iov_iter; 160 struct napi_struct; 161 162 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 163 struct nf_conntrack { 164 atomic_t use; 165 }; 166 #endif 167 168 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 169 struct nf_bridge_info { 170 atomic_t use; 171 enum { 172 BRNF_PROTO_UNCHANGED, 173 BRNF_PROTO_8021Q, 174 BRNF_PROTO_PPPOE 175 } orig_proto:8; 176 bool pkt_otherhost; 177 __u16 frag_max_size; 178 unsigned int mask; 179 struct net_device *physindev; 180 union { 181 struct net_device *physoutdev; 182 char neigh_header[8]; 183 }; 184 union { 185 __be32 ipv4_daddr; 186 struct in6_addr ipv6_daddr; 187 }; 188 }; 189 #endif 190 191 struct sk_buff_head { 192 /* These two members must be first. */ 193 struct sk_buff *next; 194 struct sk_buff *prev; 195 196 __u32 qlen; 197 spinlock_t lock; 198 }; 199 200 struct sk_buff; 201 202 /* To allow 64K frame to be packed as single skb without frag_list we 203 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 204 * buffers which do not start on a page boundary. 205 * 206 * Since GRO uses frags we allocate at least 16 regardless of page 207 * size. 208 */ 209 #if (65536/PAGE_SIZE + 1) < 16 210 #define MAX_SKB_FRAGS 16UL 211 #else 212 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 213 #endif 214 215 typedef struct skb_frag_struct skb_frag_t; 216 217 struct skb_frag_struct { 218 struct { 219 struct page *p; 220 } page; 221 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) 222 __u32 page_offset; 223 __u32 size; 224 #else 225 __u16 page_offset; 226 __u16 size; 227 #endif 228 }; 229 230 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 231 { 232 return frag->size; 233 } 234 235 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 236 { 237 frag->size = size; 238 } 239 240 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 241 { 242 frag->size += delta; 243 } 244 245 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 246 { 247 frag->size -= delta; 248 } 249 250 #define HAVE_HW_TIME_STAMP 251 252 /** 253 * struct skb_shared_hwtstamps - hardware time stamps 254 * @hwtstamp: hardware time stamp transformed into duration 255 * since arbitrary point in time 256 * 257 * Software time stamps generated by ktime_get_real() are stored in 258 * skb->tstamp. 259 * 260 * hwtstamps can only be compared against other hwtstamps from 261 * the same device. 262 * 263 * This structure is attached to packets as part of the 264 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 265 */ 266 struct skb_shared_hwtstamps { 267 ktime_t hwtstamp; 268 }; 269 270 /* Definitions for tx_flags in struct skb_shared_info */ 271 enum { 272 /* generate hardware time stamp */ 273 SKBTX_HW_TSTAMP = 1 << 0, 274 275 /* generate software time stamp when queueing packet to NIC */ 276 SKBTX_SW_TSTAMP = 1 << 1, 277 278 /* device driver is going to provide hardware time stamp */ 279 SKBTX_IN_PROGRESS = 1 << 2, 280 281 /* device driver supports TX zero-copy buffers */ 282 SKBTX_DEV_ZEROCOPY = 1 << 3, 283 284 /* generate wifi status information (where possible) */ 285 SKBTX_WIFI_STATUS = 1 << 4, 286 287 /* This indicates at least one fragment might be overwritten 288 * (as in vmsplice(), sendfile() ...) 289 * If we need to compute a TX checksum, we'll need to copy 290 * all frags to avoid possible bad checksum 291 */ 292 SKBTX_SHARED_FRAG = 1 << 5, 293 294 /* generate software time stamp when entering packet scheduling */ 295 SKBTX_SCHED_TSTAMP = 1 << 6, 296 297 /* generate software timestamp on peer data acknowledgment */ 298 SKBTX_ACK_TSTAMP = 1 << 7, 299 }; 300 301 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ 302 SKBTX_SCHED_TSTAMP | \ 303 SKBTX_ACK_TSTAMP) 304 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) 305 306 /* 307 * The callback notifies userspace to release buffers when skb DMA is done in 308 * lower device, the skb last reference should be 0 when calling this. 309 * The zerocopy_success argument is true if zero copy transmit occurred, 310 * false on data copy or out of memory error caused by data copy attempt. 311 * The ctx field is used to track device context. 312 * The desc field is used to track userspace buffer index. 313 */ 314 struct ubuf_info { 315 void (*callback)(struct ubuf_info *, bool zerocopy_success); 316 void *ctx; 317 unsigned long desc; 318 }; 319 320 /* This data is invariant across clones and lives at 321 * the end of the header data, ie. at skb->end. 322 */ 323 struct skb_shared_info { 324 unsigned char nr_frags; 325 __u8 tx_flags; 326 unsigned short gso_size; 327 /* Warning: this field is not always filled in (UFO)! */ 328 unsigned short gso_segs; 329 unsigned short gso_type; 330 struct sk_buff *frag_list; 331 struct skb_shared_hwtstamps hwtstamps; 332 u32 tskey; 333 __be32 ip6_frag_id; 334 335 /* 336 * Warning : all fields before dataref are cleared in __alloc_skb() 337 */ 338 atomic_t dataref; 339 340 /* Intermediate layers must ensure that destructor_arg 341 * remains valid until skb destructor */ 342 void * destructor_arg; 343 344 /* must be last field, see pskb_expand_head() */ 345 skb_frag_t frags[MAX_SKB_FRAGS]; 346 }; 347 348 /* We divide dataref into two halves. The higher 16 bits hold references 349 * to the payload part of skb->data. The lower 16 bits hold references to 350 * the entire skb->data. A clone of a headerless skb holds the length of 351 * the header in skb->hdr_len. 352 * 353 * All users must obey the rule that the skb->data reference count must be 354 * greater than or equal to the payload reference count. 355 * 356 * Holding a reference to the payload part means that the user does not 357 * care about modifications to the header part of skb->data. 358 */ 359 #define SKB_DATAREF_SHIFT 16 360 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 361 362 363 enum { 364 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ 365 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ 366 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ 367 }; 368 369 enum { 370 SKB_GSO_TCPV4 = 1 << 0, 371 SKB_GSO_UDP = 1 << 1, 372 373 /* This indicates the skb is from an untrusted source. */ 374 SKB_GSO_DODGY = 1 << 2, 375 376 /* This indicates the tcp segment has CWR set. */ 377 SKB_GSO_TCP_ECN = 1 << 3, 378 379 SKB_GSO_TCPV6 = 1 << 4, 380 381 SKB_GSO_FCOE = 1 << 5, 382 383 SKB_GSO_GRE = 1 << 6, 384 385 SKB_GSO_GRE_CSUM = 1 << 7, 386 387 SKB_GSO_IPIP = 1 << 8, 388 389 SKB_GSO_SIT = 1 << 9, 390 391 SKB_GSO_UDP_TUNNEL = 1 << 10, 392 393 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, 394 395 SKB_GSO_TUNNEL_REMCSUM = 1 << 12, 396 }; 397 398 #if BITS_PER_LONG > 32 399 #define NET_SKBUFF_DATA_USES_OFFSET 1 400 #endif 401 402 #ifdef NET_SKBUFF_DATA_USES_OFFSET 403 typedef unsigned int sk_buff_data_t; 404 #else 405 typedef unsigned char *sk_buff_data_t; 406 #endif 407 408 /** 409 * struct skb_mstamp - multi resolution time stamps 410 * @stamp_us: timestamp in us resolution 411 * @stamp_jiffies: timestamp in jiffies 412 */ 413 struct skb_mstamp { 414 union { 415 u64 v64; 416 struct { 417 u32 stamp_us; 418 u32 stamp_jiffies; 419 }; 420 }; 421 }; 422 423 /** 424 * skb_mstamp_get - get current timestamp 425 * @cl: place to store timestamps 426 */ 427 static inline void skb_mstamp_get(struct skb_mstamp *cl) 428 { 429 u64 val = local_clock(); 430 431 do_div(val, NSEC_PER_USEC); 432 cl->stamp_us = (u32)val; 433 cl->stamp_jiffies = (u32)jiffies; 434 } 435 436 /** 437 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp 438 * @t1: pointer to newest sample 439 * @t0: pointer to oldest sample 440 */ 441 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1, 442 const struct skb_mstamp *t0) 443 { 444 s32 delta_us = t1->stamp_us - t0->stamp_us; 445 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies; 446 447 /* If delta_us is negative, this might be because interval is too big, 448 * or local_clock() drift is too big : fallback using jiffies. 449 */ 450 if (delta_us <= 0 || 451 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ))) 452 453 delta_us = jiffies_to_usecs(delta_jiffies); 454 455 return delta_us; 456 } 457 458 459 /** 460 * struct sk_buff - socket buffer 461 * @next: Next buffer in list 462 * @prev: Previous buffer in list 463 * @tstamp: Time we arrived/left 464 * @rbnode: RB tree node, alternative to next/prev for netem/tcp 465 * @sk: Socket we are owned by 466 * @dev: Device we arrived on/are leaving by 467 * @cb: Control buffer. Free for use by every layer. Put private vars here 468 * @_skb_refdst: destination entry (with norefcount bit) 469 * @sp: the security path, used for xfrm 470 * @len: Length of actual data 471 * @data_len: Data length 472 * @mac_len: Length of link layer header 473 * @hdr_len: writable header length of cloned skb 474 * @csum: Checksum (must include start/offset pair) 475 * @csum_start: Offset from skb->head where checksumming should start 476 * @csum_offset: Offset from csum_start where checksum should be stored 477 * @priority: Packet queueing priority 478 * @ignore_df: allow local fragmentation 479 * @cloned: Head may be cloned (check refcnt to be sure) 480 * @ip_summed: Driver fed us an IP checksum 481 * @nohdr: Payload reference only, must not modify header 482 * @nfctinfo: Relationship of this skb to the connection 483 * @pkt_type: Packet class 484 * @fclone: skbuff clone status 485 * @ipvs_property: skbuff is owned by ipvs 486 * @peeked: this packet has been seen already, so stats have been 487 * done for it, don't do them again 488 * @nf_trace: netfilter packet trace flag 489 * @protocol: Packet protocol from driver 490 * @destructor: Destruct function 491 * @nfct: Associated connection, if any 492 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 493 * @skb_iif: ifindex of device we arrived on 494 * @tc_index: Traffic control index 495 * @tc_verd: traffic control verdict 496 * @hash: the packet hash 497 * @queue_mapping: Queue mapping for multiqueue devices 498 * @xmit_more: More SKBs are pending for this queue 499 * @ndisc_nodetype: router type (from link layer) 500 * @ooo_okay: allow the mapping of a socket to a queue to be changed 501 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport 502 * ports. 503 * @sw_hash: indicates hash was computed in software stack 504 * @wifi_acked_valid: wifi_acked was set 505 * @wifi_acked: whether frame was acked on wifi or not 506 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 507 * @napi_id: id of the NAPI struct this skb came from 508 * @secmark: security marking 509 * @mark: Generic packet mark 510 * @vlan_proto: vlan encapsulation protocol 511 * @vlan_tci: vlan tag control information 512 * @inner_protocol: Protocol (encapsulation) 513 * @inner_transport_header: Inner transport layer header (encapsulation) 514 * @inner_network_header: Network layer header (encapsulation) 515 * @inner_mac_header: Link layer header (encapsulation) 516 * @transport_header: Transport layer header 517 * @network_header: Network layer header 518 * @mac_header: Link layer header 519 * @tail: Tail pointer 520 * @end: End pointer 521 * @head: Head of buffer 522 * @data: Data head pointer 523 * @truesize: Buffer size 524 * @users: User count - see {datagram,tcp}.c 525 */ 526 527 struct sk_buff { 528 union { 529 struct { 530 /* These two members must be first. */ 531 struct sk_buff *next; 532 struct sk_buff *prev; 533 534 union { 535 ktime_t tstamp; 536 struct skb_mstamp skb_mstamp; 537 }; 538 }; 539 struct rb_node rbnode; /* used in netem & tcp stack */ 540 }; 541 struct sock *sk; 542 struct net_device *dev; 543 544 /* 545 * This is the control buffer. It is free to use for every 546 * layer. Please put your private variables there. If you 547 * want to keep them across layers you have to do a skb_clone() 548 * first. This is owned by whoever has the skb queued ATM. 549 */ 550 char cb[48] __aligned(8); 551 552 unsigned long _skb_refdst; 553 void (*destructor)(struct sk_buff *skb); 554 #ifdef CONFIG_XFRM 555 struct sec_path *sp; 556 #endif 557 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 558 struct nf_conntrack *nfct; 559 #endif 560 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 561 struct nf_bridge_info *nf_bridge; 562 #endif 563 unsigned int len, 564 data_len; 565 __u16 mac_len, 566 hdr_len; 567 568 /* Following fields are _not_ copied in __copy_skb_header() 569 * Note that queue_mapping is here mostly to fill a hole. 570 */ 571 kmemcheck_bitfield_begin(flags1); 572 __u16 queue_mapping; 573 __u8 cloned:1, 574 nohdr:1, 575 fclone:2, 576 peeked:1, 577 head_frag:1, 578 xmit_more:1; 579 /* one bit hole */ 580 kmemcheck_bitfield_end(flags1); 581 582 /* fields enclosed in headers_start/headers_end are copied 583 * using a single memcpy() in __copy_skb_header() 584 */ 585 /* private: */ 586 __u32 headers_start[0]; 587 /* public: */ 588 589 /* if you move pkt_type around you also must adapt those constants */ 590 #ifdef __BIG_ENDIAN_BITFIELD 591 #define PKT_TYPE_MAX (7 << 5) 592 #else 593 #define PKT_TYPE_MAX 7 594 #endif 595 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) 596 597 __u8 __pkt_type_offset[0]; 598 __u8 pkt_type:3; 599 __u8 pfmemalloc:1; 600 __u8 ignore_df:1; 601 __u8 nfctinfo:3; 602 603 __u8 nf_trace:1; 604 __u8 ip_summed:2; 605 __u8 ooo_okay:1; 606 __u8 l4_hash:1; 607 __u8 sw_hash:1; 608 __u8 wifi_acked_valid:1; 609 __u8 wifi_acked:1; 610 611 __u8 no_fcs:1; 612 /* Indicates the inner headers are valid in the skbuff. */ 613 __u8 encapsulation:1; 614 __u8 encap_hdr_csum:1; 615 __u8 csum_valid:1; 616 __u8 csum_complete_sw:1; 617 __u8 csum_level:2; 618 __u8 csum_bad:1; 619 620 #ifdef CONFIG_IPV6_NDISC_NODETYPE 621 __u8 ndisc_nodetype:2; 622 #endif 623 __u8 ipvs_property:1; 624 __u8 inner_protocol_type:1; 625 __u8 remcsum_offload:1; 626 /* 3 or 5 bit hole */ 627 628 #ifdef CONFIG_NET_SCHED 629 __u16 tc_index; /* traffic control index */ 630 #ifdef CONFIG_NET_CLS_ACT 631 __u16 tc_verd; /* traffic control verdict */ 632 #endif 633 #endif 634 635 union { 636 __wsum csum; 637 struct { 638 __u16 csum_start; 639 __u16 csum_offset; 640 }; 641 }; 642 __u32 priority; 643 int skb_iif; 644 __u32 hash; 645 __be16 vlan_proto; 646 __u16 vlan_tci; 647 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) 648 union { 649 unsigned int napi_id; 650 unsigned int sender_cpu; 651 }; 652 #endif 653 #ifdef CONFIG_NETWORK_SECMARK 654 __u32 secmark; 655 #endif 656 union { 657 __u32 mark; 658 __u32 reserved_tailroom; 659 }; 660 661 union { 662 __be16 inner_protocol; 663 __u8 inner_ipproto; 664 }; 665 666 __u16 inner_transport_header; 667 __u16 inner_network_header; 668 __u16 inner_mac_header; 669 670 __be16 protocol; 671 __u16 transport_header; 672 __u16 network_header; 673 __u16 mac_header; 674 675 /* private: */ 676 __u32 headers_end[0]; 677 /* public: */ 678 679 /* These elements must be at the end, see alloc_skb() for details. */ 680 sk_buff_data_t tail; 681 sk_buff_data_t end; 682 unsigned char *head, 683 *data; 684 unsigned int truesize; 685 atomic_t users; 686 }; 687 688 #ifdef __KERNEL__ 689 /* 690 * Handling routines are only of interest to the kernel 691 */ 692 #include <linux/slab.h> 693 694 695 #define SKB_ALLOC_FCLONE 0x01 696 #define SKB_ALLOC_RX 0x02 697 #define SKB_ALLOC_NAPI 0x04 698 699 /* Returns true if the skb was allocated from PFMEMALLOC reserves */ 700 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 701 { 702 return unlikely(skb->pfmemalloc); 703 } 704 705 /* 706 * skb might have a dst pointer attached, refcounted or not. 707 * _skb_refdst low order bit is set if refcount was _not_ taken 708 */ 709 #define SKB_DST_NOREF 1UL 710 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 711 712 /** 713 * skb_dst - returns skb dst_entry 714 * @skb: buffer 715 * 716 * Returns skb dst_entry, regardless of reference taken or not. 717 */ 718 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 719 { 720 /* If refdst was not refcounted, check we still are in a 721 * rcu_read_lock section 722 */ 723 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 724 !rcu_read_lock_held() && 725 !rcu_read_lock_bh_held()); 726 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 727 } 728 729 /** 730 * skb_dst_set - sets skb dst 731 * @skb: buffer 732 * @dst: dst entry 733 * 734 * Sets skb dst, assuming a reference was taken on dst and should 735 * be released by skb_dst_drop() 736 */ 737 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 738 { 739 skb->_skb_refdst = (unsigned long)dst; 740 } 741 742 /** 743 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference 744 * @skb: buffer 745 * @dst: dst entry 746 * 747 * Sets skb dst, assuming a reference was not taken on dst. 748 * If dst entry is cached, we do not take reference and dst_release 749 * will be avoided by refdst_drop. If dst entry is not cached, we take 750 * reference, so that last dst_release can destroy the dst immediately. 751 */ 752 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) 753 { 754 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); 755 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; 756 } 757 758 /** 759 * skb_dst_is_noref - Test if skb dst isn't refcounted 760 * @skb: buffer 761 */ 762 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 763 { 764 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 765 } 766 767 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 768 { 769 return (struct rtable *)skb_dst(skb); 770 } 771 772 void kfree_skb(struct sk_buff *skb); 773 void kfree_skb_list(struct sk_buff *segs); 774 void skb_tx_error(struct sk_buff *skb); 775 void consume_skb(struct sk_buff *skb); 776 void __kfree_skb(struct sk_buff *skb); 777 extern struct kmem_cache *skbuff_head_cache; 778 779 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 780 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 781 bool *fragstolen, int *delta_truesize); 782 783 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, 784 int node); 785 struct sk_buff *__build_skb(void *data, unsigned int frag_size); 786 struct sk_buff *build_skb(void *data, unsigned int frag_size); 787 static inline struct sk_buff *alloc_skb(unsigned int size, 788 gfp_t priority) 789 { 790 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 791 } 792 793 struct sk_buff *alloc_skb_with_frags(unsigned long header_len, 794 unsigned long data_len, 795 int max_page_order, 796 int *errcode, 797 gfp_t gfp_mask); 798 799 /* Layout of fast clones : [skb1][skb2][fclone_ref] */ 800 struct sk_buff_fclones { 801 struct sk_buff skb1; 802 803 struct sk_buff skb2; 804 805 atomic_t fclone_ref; 806 }; 807 808 /** 809 * skb_fclone_busy - check if fclone is busy 810 * @skb: buffer 811 * 812 * Returns true is skb is a fast clone, and its clone is not freed. 813 * Some drivers call skb_orphan() in their ndo_start_xmit(), 814 * so we also check that this didnt happen. 815 */ 816 static inline bool skb_fclone_busy(const struct sock *sk, 817 const struct sk_buff *skb) 818 { 819 const struct sk_buff_fclones *fclones; 820 821 fclones = container_of(skb, struct sk_buff_fclones, skb1); 822 823 return skb->fclone == SKB_FCLONE_ORIG && 824 atomic_read(&fclones->fclone_ref) > 1 && 825 fclones->skb2.sk == sk; 826 } 827 828 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 829 gfp_t priority) 830 { 831 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 832 } 833 834 struct sk_buff *__alloc_skb_head(gfp_t priority, int node); 835 static inline struct sk_buff *alloc_skb_head(gfp_t priority) 836 { 837 return __alloc_skb_head(priority, -1); 838 } 839 840 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 841 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 842 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); 843 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); 844 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, 845 gfp_t gfp_mask, bool fclone); 846 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, 847 gfp_t gfp_mask) 848 { 849 return __pskb_copy_fclone(skb, headroom, gfp_mask, false); 850 } 851 852 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); 853 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 854 unsigned int headroom); 855 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, 856 int newtailroom, gfp_t priority); 857 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, 858 int offset, int len); 859 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset, 860 int len); 861 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); 862 int skb_pad(struct sk_buff *skb, int pad); 863 #define dev_kfree_skb(a) consume_skb(a) 864 865 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, 866 int getfrag(void *from, char *to, int offset, 867 int len, int odd, struct sk_buff *skb), 868 void *from, int length); 869 870 int skb_append_pagefrags(struct sk_buff *skb, struct page *page, 871 int offset, size_t size); 872 873 struct skb_seq_state { 874 __u32 lower_offset; 875 __u32 upper_offset; 876 __u32 frag_idx; 877 __u32 stepped_offset; 878 struct sk_buff *root_skb; 879 struct sk_buff *cur_skb; 880 __u8 *frag_data; 881 }; 882 883 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, 884 unsigned int to, struct skb_seq_state *st); 885 unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 886 struct skb_seq_state *st); 887 void skb_abort_seq_read(struct skb_seq_state *st); 888 889 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 890 unsigned int to, struct ts_config *config); 891 892 /* 893 * Packet hash types specify the type of hash in skb_set_hash. 894 * 895 * Hash types refer to the protocol layer addresses which are used to 896 * construct a packet's hash. The hashes are used to differentiate or identify 897 * flows of the protocol layer for the hash type. Hash types are either 898 * layer-2 (L2), layer-3 (L3), or layer-4 (L4). 899 * 900 * Properties of hashes: 901 * 902 * 1) Two packets in different flows have different hash values 903 * 2) Two packets in the same flow should have the same hash value 904 * 905 * A hash at a higher layer is considered to be more specific. A driver should 906 * set the most specific hash possible. 907 * 908 * A driver cannot indicate a more specific hash than the layer at which a hash 909 * was computed. For instance an L3 hash cannot be set as an L4 hash. 910 * 911 * A driver may indicate a hash level which is less specific than the 912 * actual layer the hash was computed on. For instance, a hash computed 913 * at L4 may be considered an L3 hash. This should only be done if the 914 * driver can't unambiguously determine that the HW computed the hash at 915 * the higher layer. Note that the "should" in the second property above 916 * permits this. 917 */ 918 enum pkt_hash_types { 919 PKT_HASH_TYPE_NONE, /* Undefined type */ 920 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ 921 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ 922 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ 923 }; 924 925 static inline void 926 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) 927 { 928 skb->l4_hash = (type == PKT_HASH_TYPE_L4); 929 skb->sw_hash = 0; 930 skb->hash = hash; 931 } 932 933 static inline __u32 skb_get_hash(struct sk_buff *skb) 934 { 935 if (!skb->l4_hash && !skb->sw_hash) 936 __skb_get_hash(skb); 937 938 return skb->hash; 939 } 940 941 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb); 942 943 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) 944 { 945 return skb->hash; 946 } 947 948 static inline void skb_clear_hash(struct sk_buff *skb) 949 { 950 skb->hash = 0; 951 skb->sw_hash = 0; 952 skb->l4_hash = 0; 953 } 954 955 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) 956 { 957 if (!skb->l4_hash) 958 skb_clear_hash(skb); 959 } 960 961 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) 962 { 963 to->hash = from->hash; 964 to->sw_hash = from->sw_hash; 965 to->l4_hash = from->l4_hash; 966 }; 967 968 static inline void skb_sender_cpu_clear(struct sk_buff *skb) 969 { 970 #ifdef CONFIG_XPS 971 skb->sender_cpu = 0; 972 #endif 973 } 974 975 #ifdef NET_SKBUFF_DATA_USES_OFFSET 976 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 977 { 978 return skb->head + skb->end; 979 } 980 981 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 982 { 983 return skb->end; 984 } 985 #else 986 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 987 { 988 return skb->end; 989 } 990 991 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 992 { 993 return skb->end - skb->head; 994 } 995 #endif 996 997 /* Internal */ 998 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 999 1000 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 1001 { 1002 return &skb_shinfo(skb)->hwtstamps; 1003 } 1004 1005 /** 1006 * skb_queue_empty - check if a queue is empty 1007 * @list: queue head 1008 * 1009 * Returns true if the queue is empty, false otherwise. 1010 */ 1011 static inline int skb_queue_empty(const struct sk_buff_head *list) 1012 { 1013 return list->next == (const struct sk_buff *) list; 1014 } 1015 1016 /** 1017 * skb_queue_is_last - check if skb is the last entry in the queue 1018 * @list: queue head 1019 * @skb: buffer 1020 * 1021 * Returns true if @skb is the last buffer on the list. 1022 */ 1023 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 1024 const struct sk_buff *skb) 1025 { 1026 return skb->next == (const struct sk_buff *) list; 1027 } 1028 1029 /** 1030 * skb_queue_is_first - check if skb is the first entry in the queue 1031 * @list: queue head 1032 * @skb: buffer 1033 * 1034 * Returns true if @skb is the first buffer on the list. 1035 */ 1036 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 1037 const struct sk_buff *skb) 1038 { 1039 return skb->prev == (const struct sk_buff *) list; 1040 } 1041 1042 /** 1043 * skb_queue_next - return the next packet in the queue 1044 * @list: queue head 1045 * @skb: current buffer 1046 * 1047 * Return the next packet in @list after @skb. It is only valid to 1048 * call this if skb_queue_is_last() evaluates to false. 1049 */ 1050 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 1051 const struct sk_buff *skb) 1052 { 1053 /* This BUG_ON may seem severe, but if we just return then we 1054 * are going to dereference garbage. 1055 */ 1056 BUG_ON(skb_queue_is_last(list, skb)); 1057 return skb->next; 1058 } 1059 1060 /** 1061 * skb_queue_prev - return the prev packet in the queue 1062 * @list: queue head 1063 * @skb: current buffer 1064 * 1065 * Return the prev packet in @list before @skb. It is only valid to 1066 * call this if skb_queue_is_first() evaluates to false. 1067 */ 1068 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 1069 const struct sk_buff *skb) 1070 { 1071 /* This BUG_ON may seem severe, but if we just return then we 1072 * are going to dereference garbage. 1073 */ 1074 BUG_ON(skb_queue_is_first(list, skb)); 1075 return skb->prev; 1076 } 1077 1078 /** 1079 * skb_get - reference buffer 1080 * @skb: buffer to reference 1081 * 1082 * Makes another reference to a socket buffer and returns a pointer 1083 * to the buffer. 1084 */ 1085 static inline struct sk_buff *skb_get(struct sk_buff *skb) 1086 { 1087 atomic_inc(&skb->users); 1088 return skb; 1089 } 1090 1091 /* 1092 * If users == 1, we are the only owner and are can avoid redundant 1093 * atomic change. 1094 */ 1095 1096 /** 1097 * skb_cloned - is the buffer a clone 1098 * @skb: buffer to check 1099 * 1100 * Returns true if the buffer was generated with skb_clone() and is 1101 * one of multiple shared copies of the buffer. Cloned buffers are 1102 * shared data so must not be written to under normal circumstances. 1103 */ 1104 static inline int skb_cloned(const struct sk_buff *skb) 1105 { 1106 return skb->cloned && 1107 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 1108 } 1109 1110 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) 1111 { 1112 might_sleep_if(pri & __GFP_WAIT); 1113 1114 if (skb_cloned(skb)) 1115 return pskb_expand_head(skb, 0, 0, pri); 1116 1117 return 0; 1118 } 1119 1120 /** 1121 * skb_header_cloned - is the header a clone 1122 * @skb: buffer to check 1123 * 1124 * Returns true if modifying the header part of the buffer requires 1125 * the data to be copied. 1126 */ 1127 static inline int skb_header_cloned(const struct sk_buff *skb) 1128 { 1129 int dataref; 1130 1131 if (!skb->cloned) 1132 return 0; 1133 1134 dataref = atomic_read(&skb_shinfo(skb)->dataref); 1135 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 1136 return dataref != 1; 1137 } 1138 1139 /** 1140 * skb_header_release - release reference to header 1141 * @skb: buffer to operate on 1142 * 1143 * Drop a reference to the header part of the buffer. This is done 1144 * by acquiring a payload reference. You must not read from the header 1145 * part of skb->data after this. 1146 * Note : Check if you can use __skb_header_release() instead. 1147 */ 1148 static inline void skb_header_release(struct sk_buff *skb) 1149 { 1150 BUG_ON(skb->nohdr); 1151 skb->nohdr = 1; 1152 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); 1153 } 1154 1155 /** 1156 * __skb_header_release - release reference to header 1157 * @skb: buffer to operate on 1158 * 1159 * Variant of skb_header_release() assuming skb is private to caller. 1160 * We can avoid one atomic operation. 1161 */ 1162 static inline void __skb_header_release(struct sk_buff *skb) 1163 { 1164 skb->nohdr = 1; 1165 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); 1166 } 1167 1168 1169 /** 1170 * skb_shared - is the buffer shared 1171 * @skb: buffer to check 1172 * 1173 * Returns true if more than one person has a reference to this 1174 * buffer. 1175 */ 1176 static inline int skb_shared(const struct sk_buff *skb) 1177 { 1178 return atomic_read(&skb->users) != 1; 1179 } 1180 1181 /** 1182 * skb_share_check - check if buffer is shared and if so clone it 1183 * @skb: buffer to check 1184 * @pri: priority for memory allocation 1185 * 1186 * If the buffer is shared the buffer is cloned and the old copy 1187 * drops a reference. A new clone with a single reference is returned. 1188 * If the buffer is not shared the original buffer is returned. When 1189 * being called from interrupt status or with spinlocks held pri must 1190 * be GFP_ATOMIC. 1191 * 1192 * NULL is returned on a memory allocation failure. 1193 */ 1194 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 1195 { 1196 might_sleep_if(pri & __GFP_WAIT); 1197 if (skb_shared(skb)) { 1198 struct sk_buff *nskb = skb_clone(skb, pri); 1199 1200 if (likely(nskb)) 1201 consume_skb(skb); 1202 else 1203 kfree_skb(skb); 1204 skb = nskb; 1205 } 1206 return skb; 1207 } 1208 1209 /* 1210 * Copy shared buffers into a new sk_buff. We effectively do COW on 1211 * packets to handle cases where we have a local reader and forward 1212 * and a couple of other messy ones. The normal one is tcpdumping 1213 * a packet thats being forwarded. 1214 */ 1215 1216 /** 1217 * skb_unshare - make a copy of a shared buffer 1218 * @skb: buffer to check 1219 * @pri: priority for memory allocation 1220 * 1221 * If the socket buffer is a clone then this function creates a new 1222 * copy of the data, drops a reference count on the old copy and returns 1223 * the new copy with the reference count at 1. If the buffer is not a clone 1224 * the original buffer is returned. When called with a spinlock held or 1225 * from interrupt state @pri must be %GFP_ATOMIC 1226 * 1227 * %NULL is returned on a memory allocation failure. 1228 */ 1229 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 1230 gfp_t pri) 1231 { 1232 might_sleep_if(pri & __GFP_WAIT); 1233 if (skb_cloned(skb)) { 1234 struct sk_buff *nskb = skb_copy(skb, pri); 1235 1236 /* Free our shared copy */ 1237 if (likely(nskb)) 1238 consume_skb(skb); 1239 else 1240 kfree_skb(skb); 1241 skb = nskb; 1242 } 1243 return skb; 1244 } 1245 1246 /** 1247 * skb_peek - peek at the head of an &sk_buff_head 1248 * @list_: list to peek at 1249 * 1250 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1251 * be careful with this one. A peek leaves the buffer on the 1252 * list and someone else may run off with it. You must hold 1253 * the appropriate locks or have a private queue to do this. 1254 * 1255 * Returns %NULL for an empty list or a pointer to the head element. 1256 * The reference count is not incremented and the reference is therefore 1257 * volatile. Use with caution. 1258 */ 1259 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 1260 { 1261 struct sk_buff *skb = list_->next; 1262 1263 if (skb == (struct sk_buff *)list_) 1264 skb = NULL; 1265 return skb; 1266 } 1267 1268 /** 1269 * skb_peek_next - peek skb following the given one from a queue 1270 * @skb: skb to start from 1271 * @list_: list to peek at 1272 * 1273 * Returns %NULL when the end of the list is met or a pointer to the 1274 * next element. The reference count is not incremented and the 1275 * reference is therefore volatile. Use with caution. 1276 */ 1277 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 1278 const struct sk_buff_head *list_) 1279 { 1280 struct sk_buff *next = skb->next; 1281 1282 if (next == (struct sk_buff *)list_) 1283 next = NULL; 1284 return next; 1285 } 1286 1287 /** 1288 * skb_peek_tail - peek at the tail of an &sk_buff_head 1289 * @list_: list to peek at 1290 * 1291 * Peek an &sk_buff. Unlike most other operations you _MUST_ 1292 * be careful with this one. A peek leaves the buffer on the 1293 * list and someone else may run off with it. You must hold 1294 * the appropriate locks or have a private queue to do this. 1295 * 1296 * Returns %NULL for an empty list or a pointer to the tail element. 1297 * The reference count is not incremented and the reference is therefore 1298 * volatile. Use with caution. 1299 */ 1300 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 1301 { 1302 struct sk_buff *skb = list_->prev; 1303 1304 if (skb == (struct sk_buff *)list_) 1305 skb = NULL; 1306 return skb; 1307 1308 } 1309 1310 /** 1311 * skb_queue_len - get queue length 1312 * @list_: list to measure 1313 * 1314 * Return the length of an &sk_buff queue. 1315 */ 1316 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 1317 { 1318 return list_->qlen; 1319 } 1320 1321 /** 1322 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 1323 * @list: queue to initialize 1324 * 1325 * This initializes only the list and queue length aspects of 1326 * an sk_buff_head object. This allows to initialize the list 1327 * aspects of an sk_buff_head without reinitializing things like 1328 * the spinlock. It can also be used for on-stack sk_buff_head 1329 * objects where the spinlock is known to not be used. 1330 */ 1331 static inline void __skb_queue_head_init(struct sk_buff_head *list) 1332 { 1333 list->prev = list->next = (struct sk_buff *)list; 1334 list->qlen = 0; 1335 } 1336 1337 /* 1338 * This function creates a split out lock class for each invocation; 1339 * this is needed for now since a whole lot of users of the skb-queue 1340 * infrastructure in drivers have different locking usage (in hardirq) 1341 * than the networking core (in softirq only). In the long run either the 1342 * network layer or drivers should need annotation to consolidate the 1343 * main types of usage into 3 classes. 1344 */ 1345 static inline void skb_queue_head_init(struct sk_buff_head *list) 1346 { 1347 spin_lock_init(&list->lock); 1348 __skb_queue_head_init(list); 1349 } 1350 1351 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1352 struct lock_class_key *class) 1353 { 1354 skb_queue_head_init(list); 1355 lockdep_set_class(&list->lock, class); 1356 } 1357 1358 /* 1359 * Insert an sk_buff on a list. 1360 * 1361 * The "__skb_xxxx()" functions are the non-atomic ones that 1362 * can only be called with interrupts disabled. 1363 */ 1364 void skb_insert(struct sk_buff *old, struct sk_buff *newsk, 1365 struct sk_buff_head *list); 1366 static inline void __skb_insert(struct sk_buff *newsk, 1367 struct sk_buff *prev, struct sk_buff *next, 1368 struct sk_buff_head *list) 1369 { 1370 newsk->next = next; 1371 newsk->prev = prev; 1372 next->prev = prev->next = newsk; 1373 list->qlen++; 1374 } 1375 1376 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1377 struct sk_buff *prev, 1378 struct sk_buff *next) 1379 { 1380 struct sk_buff *first = list->next; 1381 struct sk_buff *last = list->prev; 1382 1383 first->prev = prev; 1384 prev->next = first; 1385 1386 last->next = next; 1387 next->prev = last; 1388 } 1389 1390 /** 1391 * skb_queue_splice - join two skb lists, this is designed for stacks 1392 * @list: the new list to add 1393 * @head: the place to add it in the first list 1394 */ 1395 static inline void skb_queue_splice(const struct sk_buff_head *list, 1396 struct sk_buff_head *head) 1397 { 1398 if (!skb_queue_empty(list)) { 1399 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1400 head->qlen += list->qlen; 1401 } 1402 } 1403 1404 /** 1405 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1406 * @list: the new list to add 1407 * @head: the place to add it in the first list 1408 * 1409 * The list at @list is reinitialised 1410 */ 1411 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1412 struct sk_buff_head *head) 1413 { 1414 if (!skb_queue_empty(list)) { 1415 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1416 head->qlen += list->qlen; 1417 __skb_queue_head_init(list); 1418 } 1419 } 1420 1421 /** 1422 * skb_queue_splice_tail - join two skb lists, each list being a queue 1423 * @list: the new list to add 1424 * @head: the place to add it in the first list 1425 */ 1426 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1427 struct sk_buff_head *head) 1428 { 1429 if (!skb_queue_empty(list)) { 1430 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1431 head->qlen += list->qlen; 1432 } 1433 } 1434 1435 /** 1436 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1437 * @list: the new list to add 1438 * @head: the place to add it in the first list 1439 * 1440 * Each of the lists is a queue. 1441 * The list at @list is reinitialised 1442 */ 1443 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 1444 struct sk_buff_head *head) 1445 { 1446 if (!skb_queue_empty(list)) { 1447 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1448 head->qlen += list->qlen; 1449 __skb_queue_head_init(list); 1450 } 1451 } 1452 1453 /** 1454 * __skb_queue_after - queue a buffer at the list head 1455 * @list: list to use 1456 * @prev: place after this buffer 1457 * @newsk: buffer to queue 1458 * 1459 * Queue a buffer int the middle of a list. This function takes no locks 1460 * and you must therefore hold required locks before calling it. 1461 * 1462 * A buffer cannot be placed on two lists at the same time. 1463 */ 1464 static inline void __skb_queue_after(struct sk_buff_head *list, 1465 struct sk_buff *prev, 1466 struct sk_buff *newsk) 1467 { 1468 __skb_insert(newsk, prev, prev->next, list); 1469 } 1470 1471 void skb_append(struct sk_buff *old, struct sk_buff *newsk, 1472 struct sk_buff_head *list); 1473 1474 static inline void __skb_queue_before(struct sk_buff_head *list, 1475 struct sk_buff *next, 1476 struct sk_buff *newsk) 1477 { 1478 __skb_insert(newsk, next->prev, next, list); 1479 } 1480 1481 /** 1482 * __skb_queue_head - queue a buffer at the list head 1483 * @list: list to use 1484 * @newsk: buffer to queue 1485 * 1486 * Queue a buffer at the start of a list. This function takes no locks 1487 * and you must therefore hold required locks before calling it. 1488 * 1489 * A buffer cannot be placed on two lists at the same time. 1490 */ 1491 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 1492 static inline void __skb_queue_head(struct sk_buff_head *list, 1493 struct sk_buff *newsk) 1494 { 1495 __skb_queue_after(list, (struct sk_buff *)list, newsk); 1496 } 1497 1498 /** 1499 * __skb_queue_tail - queue a buffer at the list tail 1500 * @list: list to use 1501 * @newsk: buffer to queue 1502 * 1503 * Queue a buffer at the end of a list. This function takes no locks 1504 * and you must therefore hold required locks before calling it. 1505 * 1506 * A buffer cannot be placed on two lists at the same time. 1507 */ 1508 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 1509 static inline void __skb_queue_tail(struct sk_buff_head *list, 1510 struct sk_buff *newsk) 1511 { 1512 __skb_queue_before(list, (struct sk_buff *)list, newsk); 1513 } 1514 1515 /* 1516 * remove sk_buff from list. _Must_ be called atomically, and with 1517 * the list known.. 1518 */ 1519 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 1520 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 1521 { 1522 struct sk_buff *next, *prev; 1523 1524 list->qlen--; 1525 next = skb->next; 1526 prev = skb->prev; 1527 skb->next = skb->prev = NULL; 1528 next->prev = prev; 1529 prev->next = next; 1530 } 1531 1532 /** 1533 * __skb_dequeue - remove from the head of the queue 1534 * @list: list to dequeue from 1535 * 1536 * Remove the head of the list. This function does not take any locks 1537 * so must be used with appropriate locks held only. The head item is 1538 * returned or %NULL if the list is empty. 1539 */ 1540 struct sk_buff *skb_dequeue(struct sk_buff_head *list); 1541 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 1542 { 1543 struct sk_buff *skb = skb_peek(list); 1544 if (skb) 1545 __skb_unlink(skb, list); 1546 return skb; 1547 } 1548 1549 /** 1550 * __skb_dequeue_tail - remove from the tail of the queue 1551 * @list: list to dequeue from 1552 * 1553 * Remove the tail of the list. This function does not take any locks 1554 * so must be used with appropriate locks held only. The tail item is 1555 * returned or %NULL if the list is empty. 1556 */ 1557 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 1558 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 1559 { 1560 struct sk_buff *skb = skb_peek_tail(list); 1561 if (skb) 1562 __skb_unlink(skb, list); 1563 return skb; 1564 } 1565 1566 1567 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 1568 { 1569 return skb->data_len; 1570 } 1571 1572 static inline unsigned int skb_headlen(const struct sk_buff *skb) 1573 { 1574 return skb->len - skb->data_len; 1575 } 1576 1577 static inline int skb_pagelen(const struct sk_buff *skb) 1578 { 1579 int i, len = 0; 1580 1581 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) 1582 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 1583 return len + skb_headlen(skb); 1584 } 1585 1586 /** 1587 * __skb_fill_page_desc - initialise a paged fragment in an skb 1588 * @skb: buffer containing fragment to be initialised 1589 * @i: paged fragment index to initialise 1590 * @page: the page to use for this fragment 1591 * @off: the offset to the data with @page 1592 * @size: the length of the data 1593 * 1594 * Initialises the @i'th fragment of @skb to point to &size bytes at 1595 * offset @off within @page. 1596 * 1597 * Does not take any additional reference on the fragment. 1598 */ 1599 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 1600 struct page *page, int off, int size) 1601 { 1602 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 1603 1604 /* 1605 * Propagate page pfmemalloc to the skb if we can. The problem is 1606 * that not all callers have unique ownership of the page but rely 1607 * on page_is_pfmemalloc doing the right thing(tm). 1608 */ 1609 frag->page.p = page; 1610 frag->page_offset = off; 1611 skb_frag_size_set(frag, size); 1612 1613 page = compound_head(page); 1614 if (page_is_pfmemalloc(page)) 1615 skb->pfmemalloc = true; 1616 } 1617 1618 /** 1619 * skb_fill_page_desc - initialise a paged fragment in an skb 1620 * @skb: buffer containing fragment to be initialised 1621 * @i: paged fragment index to initialise 1622 * @page: the page to use for this fragment 1623 * @off: the offset to the data with @page 1624 * @size: the length of the data 1625 * 1626 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 1627 * @skb to point to @size bytes at offset @off within @page. In 1628 * addition updates @skb such that @i is the last fragment. 1629 * 1630 * Does not take any additional reference on the fragment. 1631 */ 1632 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 1633 struct page *page, int off, int size) 1634 { 1635 __skb_fill_page_desc(skb, i, page, off, size); 1636 skb_shinfo(skb)->nr_frags = i + 1; 1637 } 1638 1639 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, 1640 int size, unsigned int truesize); 1641 1642 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, 1643 unsigned int truesize); 1644 1645 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) 1646 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb)) 1647 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 1648 1649 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1650 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1651 { 1652 return skb->head + skb->tail; 1653 } 1654 1655 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1656 { 1657 skb->tail = skb->data - skb->head; 1658 } 1659 1660 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1661 { 1662 skb_reset_tail_pointer(skb); 1663 skb->tail += offset; 1664 } 1665 1666 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 1667 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1668 { 1669 return skb->tail; 1670 } 1671 1672 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1673 { 1674 skb->tail = skb->data; 1675 } 1676 1677 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1678 { 1679 skb->tail = skb->data + offset; 1680 } 1681 1682 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1683 1684 /* 1685 * Add data to an sk_buff 1686 */ 1687 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); 1688 unsigned char *skb_put(struct sk_buff *skb, unsigned int len); 1689 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) 1690 { 1691 unsigned char *tmp = skb_tail_pointer(skb); 1692 SKB_LINEAR_ASSERT(skb); 1693 skb->tail += len; 1694 skb->len += len; 1695 return tmp; 1696 } 1697 1698 unsigned char *skb_push(struct sk_buff *skb, unsigned int len); 1699 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) 1700 { 1701 skb->data -= len; 1702 skb->len += len; 1703 return skb->data; 1704 } 1705 1706 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); 1707 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) 1708 { 1709 skb->len -= len; 1710 BUG_ON(skb->len < skb->data_len); 1711 return skb->data += len; 1712 } 1713 1714 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len) 1715 { 1716 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 1717 } 1718 1719 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); 1720 1721 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) 1722 { 1723 if (len > skb_headlen(skb) && 1724 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 1725 return NULL; 1726 skb->len -= len; 1727 return skb->data += len; 1728 } 1729 1730 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) 1731 { 1732 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 1733 } 1734 1735 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) 1736 { 1737 if (likely(len <= skb_headlen(skb))) 1738 return 1; 1739 if (unlikely(len > skb->len)) 1740 return 0; 1741 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 1742 } 1743 1744 /** 1745 * skb_headroom - bytes at buffer head 1746 * @skb: buffer to check 1747 * 1748 * Return the number of bytes of free space at the head of an &sk_buff. 1749 */ 1750 static inline unsigned int skb_headroom(const struct sk_buff *skb) 1751 { 1752 return skb->data - skb->head; 1753 } 1754 1755 /** 1756 * skb_tailroom - bytes at buffer end 1757 * @skb: buffer to check 1758 * 1759 * Return the number of bytes of free space at the tail of an sk_buff 1760 */ 1761 static inline int skb_tailroom(const struct sk_buff *skb) 1762 { 1763 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 1764 } 1765 1766 /** 1767 * skb_availroom - bytes at buffer end 1768 * @skb: buffer to check 1769 * 1770 * Return the number of bytes of free space at the tail of an sk_buff 1771 * allocated by sk_stream_alloc() 1772 */ 1773 static inline int skb_availroom(const struct sk_buff *skb) 1774 { 1775 if (skb_is_nonlinear(skb)) 1776 return 0; 1777 1778 return skb->end - skb->tail - skb->reserved_tailroom; 1779 } 1780 1781 /** 1782 * skb_reserve - adjust headroom 1783 * @skb: buffer to alter 1784 * @len: bytes to move 1785 * 1786 * Increase the headroom of an empty &sk_buff by reducing the tail 1787 * room. This is only allowed for an empty buffer. 1788 */ 1789 static inline void skb_reserve(struct sk_buff *skb, int len) 1790 { 1791 skb->data += len; 1792 skb->tail += len; 1793 } 1794 1795 #define ENCAP_TYPE_ETHER 0 1796 #define ENCAP_TYPE_IPPROTO 1 1797 1798 static inline void skb_set_inner_protocol(struct sk_buff *skb, 1799 __be16 protocol) 1800 { 1801 skb->inner_protocol = protocol; 1802 skb->inner_protocol_type = ENCAP_TYPE_ETHER; 1803 } 1804 1805 static inline void skb_set_inner_ipproto(struct sk_buff *skb, 1806 __u8 ipproto) 1807 { 1808 skb->inner_ipproto = ipproto; 1809 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; 1810 } 1811 1812 static inline void skb_reset_inner_headers(struct sk_buff *skb) 1813 { 1814 skb->inner_mac_header = skb->mac_header; 1815 skb->inner_network_header = skb->network_header; 1816 skb->inner_transport_header = skb->transport_header; 1817 } 1818 1819 static inline void skb_reset_mac_len(struct sk_buff *skb) 1820 { 1821 skb->mac_len = skb->network_header - skb->mac_header; 1822 } 1823 1824 static inline unsigned char *skb_inner_transport_header(const struct sk_buff 1825 *skb) 1826 { 1827 return skb->head + skb->inner_transport_header; 1828 } 1829 1830 static inline void skb_reset_inner_transport_header(struct sk_buff *skb) 1831 { 1832 skb->inner_transport_header = skb->data - skb->head; 1833 } 1834 1835 static inline void skb_set_inner_transport_header(struct sk_buff *skb, 1836 const int offset) 1837 { 1838 skb_reset_inner_transport_header(skb); 1839 skb->inner_transport_header += offset; 1840 } 1841 1842 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) 1843 { 1844 return skb->head + skb->inner_network_header; 1845 } 1846 1847 static inline void skb_reset_inner_network_header(struct sk_buff *skb) 1848 { 1849 skb->inner_network_header = skb->data - skb->head; 1850 } 1851 1852 static inline void skb_set_inner_network_header(struct sk_buff *skb, 1853 const int offset) 1854 { 1855 skb_reset_inner_network_header(skb); 1856 skb->inner_network_header += offset; 1857 } 1858 1859 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) 1860 { 1861 return skb->head + skb->inner_mac_header; 1862 } 1863 1864 static inline void skb_reset_inner_mac_header(struct sk_buff *skb) 1865 { 1866 skb->inner_mac_header = skb->data - skb->head; 1867 } 1868 1869 static inline void skb_set_inner_mac_header(struct sk_buff *skb, 1870 const int offset) 1871 { 1872 skb_reset_inner_mac_header(skb); 1873 skb->inner_mac_header += offset; 1874 } 1875 static inline bool skb_transport_header_was_set(const struct sk_buff *skb) 1876 { 1877 return skb->transport_header != (typeof(skb->transport_header))~0U; 1878 } 1879 1880 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1881 { 1882 return skb->head + skb->transport_header; 1883 } 1884 1885 static inline void skb_reset_transport_header(struct sk_buff *skb) 1886 { 1887 skb->transport_header = skb->data - skb->head; 1888 } 1889 1890 static inline void skb_set_transport_header(struct sk_buff *skb, 1891 const int offset) 1892 { 1893 skb_reset_transport_header(skb); 1894 skb->transport_header += offset; 1895 } 1896 1897 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1898 { 1899 return skb->head + skb->network_header; 1900 } 1901 1902 static inline void skb_reset_network_header(struct sk_buff *skb) 1903 { 1904 skb->network_header = skb->data - skb->head; 1905 } 1906 1907 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1908 { 1909 skb_reset_network_header(skb); 1910 skb->network_header += offset; 1911 } 1912 1913 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1914 { 1915 return skb->head + skb->mac_header; 1916 } 1917 1918 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1919 { 1920 return skb->mac_header != (typeof(skb->mac_header))~0U; 1921 } 1922 1923 static inline void skb_reset_mac_header(struct sk_buff *skb) 1924 { 1925 skb->mac_header = skb->data - skb->head; 1926 } 1927 1928 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1929 { 1930 skb_reset_mac_header(skb); 1931 skb->mac_header += offset; 1932 } 1933 1934 static inline void skb_pop_mac_header(struct sk_buff *skb) 1935 { 1936 skb->mac_header = skb->network_header; 1937 } 1938 1939 static inline void skb_probe_transport_header(struct sk_buff *skb, 1940 const int offset_hint) 1941 { 1942 struct flow_keys keys; 1943 1944 if (skb_transport_header_was_set(skb)) 1945 return; 1946 else if (skb_flow_dissect_flow_keys(skb, &keys)) 1947 skb_set_transport_header(skb, keys.control.thoff); 1948 else 1949 skb_set_transport_header(skb, offset_hint); 1950 } 1951 1952 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 1953 { 1954 if (skb_mac_header_was_set(skb)) { 1955 const unsigned char *old_mac = skb_mac_header(skb); 1956 1957 skb_set_mac_header(skb, -skb->mac_len); 1958 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 1959 } 1960 } 1961 1962 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 1963 { 1964 return skb->csum_start - skb_headroom(skb); 1965 } 1966 1967 static inline int skb_transport_offset(const struct sk_buff *skb) 1968 { 1969 return skb_transport_header(skb) - skb->data; 1970 } 1971 1972 static inline u32 skb_network_header_len(const struct sk_buff *skb) 1973 { 1974 return skb->transport_header - skb->network_header; 1975 } 1976 1977 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) 1978 { 1979 return skb->inner_transport_header - skb->inner_network_header; 1980 } 1981 1982 static inline int skb_network_offset(const struct sk_buff *skb) 1983 { 1984 return skb_network_header(skb) - skb->data; 1985 } 1986 1987 static inline int skb_inner_network_offset(const struct sk_buff *skb) 1988 { 1989 return skb_inner_network_header(skb) - skb->data; 1990 } 1991 1992 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 1993 { 1994 return pskb_may_pull(skb, skb_network_offset(skb) + len); 1995 } 1996 1997 /* 1998 * CPUs often take a performance hit when accessing unaligned memory 1999 * locations. The actual performance hit varies, it can be small if the 2000 * hardware handles it or large if we have to take an exception and fix it 2001 * in software. 2002 * 2003 * Since an ethernet header is 14 bytes network drivers often end up with 2004 * the IP header at an unaligned offset. The IP header can be aligned by 2005 * shifting the start of the packet by 2 bytes. Drivers should do this 2006 * with: 2007 * 2008 * skb_reserve(skb, NET_IP_ALIGN); 2009 * 2010 * The downside to this alignment of the IP header is that the DMA is now 2011 * unaligned. On some architectures the cost of an unaligned DMA is high 2012 * and this cost outweighs the gains made by aligning the IP header. 2013 * 2014 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 2015 * to be overridden. 2016 */ 2017 #ifndef NET_IP_ALIGN 2018 #define NET_IP_ALIGN 2 2019 #endif 2020 2021 /* 2022 * The networking layer reserves some headroom in skb data (via 2023 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 2024 * the header has to grow. In the default case, if the header has to grow 2025 * 32 bytes or less we avoid the reallocation. 2026 * 2027 * Unfortunately this headroom changes the DMA alignment of the resulting 2028 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 2029 * on some architectures. An architecture can override this value, 2030 * perhaps setting it to a cacheline in size (since that will maintain 2031 * cacheline alignment of the DMA). It must be a power of 2. 2032 * 2033 * Various parts of the networking layer expect at least 32 bytes of 2034 * headroom, you should not reduce this. 2035 * 2036 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 2037 * to reduce average number of cache lines per packet. 2038 * get_rps_cpus() for example only access one 64 bytes aligned block : 2039 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 2040 */ 2041 #ifndef NET_SKB_PAD 2042 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 2043 #endif 2044 2045 int ___pskb_trim(struct sk_buff *skb, unsigned int len); 2046 2047 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 2048 { 2049 if (unlikely(skb_is_nonlinear(skb))) { 2050 WARN_ON(1); 2051 return; 2052 } 2053 skb->len = len; 2054 skb_set_tail_pointer(skb, len); 2055 } 2056 2057 void skb_trim(struct sk_buff *skb, unsigned int len); 2058 2059 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 2060 { 2061 if (skb->data_len) 2062 return ___pskb_trim(skb, len); 2063 __skb_trim(skb, len); 2064 return 0; 2065 } 2066 2067 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 2068 { 2069 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 2070 } 2071 2072 /** 2073 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 2074 * @skb: buffer to alter 2075 * @len: new length 2076 * 2077 * This is identical to pskb_trim except that the caller knows that 2078 * the skb is not cloned so we should never get an error due to out- 2079 * of-memory. 2080 */ 2081 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 2082 { 2083 int err = pskb_trim(skb, len); 2084 BUG_ON(err); 2085 } 2086 2087 /** 2088 * skb_orphan - orphan a buffer 2089 * @skb: buffer to orphan 2090 * 2091 * If a buffer currently has an owner then we call the owner's 2092 * destructor function and make the @skb unowned. The buffer continues 2093 * to exist but is no longer charged to its former owner. 2094 */ 2095 static inline void skb_orphan(struct sk_buff *skb) 2096 { 2097 if (skb->destructor) { 2098 skb->destructor(skb); 2099 skb->destructor = NULL; 2100 skb->sk = NULL; 2101 } else { 2102 BUG_ON(skb->sk); 2103 } 2104 } 2105 2106 /** 2107 * skb_orphan_frags - orphan the frags contained in a buffer 2108 * @skb: buffer to orphan frags from 2109 * @gfp_mask: allocation mask for replacement pages 2110 * 2111 * For each frag in the SKB which needs a destructor (i.e. has an 2112 * owner) create a copy of that frag and release the original 2113 * page by calling the destructor. 2114 */ 2115 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 2116 { 2117 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY))) 2118 return 0; 2119 return skb_copy_ubufs(skb, gfp_mask); 2120 } 2121 2122 /** 2123 * __skb_queue_purge - empty a list 2124 * @list: list to empty 2125 * 2126 * Delete all buffers on an &sk_buff list. Each buffer is removed from 2127 * the list and one reference dropped. This function does not take the 2128 * list lock and the caller must hold the relevant locks to use it. 2129 */ 2130 void skb_queue_purge(struct sk_buff_head *list); 2131 static inline void __skb_queue_purge(struct sk_buff_head *list) 2132 { 2133 struct sk_buff *skb; 2134 while ((skb = __skb_dequeue(list)) != NULL) 2135 kfree_skb(skb); 2136 } 2137 2138 void *netdev_alloc_frag(unsigned int fragsz); 2139 2140 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, 2141 gfp_t gfp_mask); 2142 2143 /** 2144 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 2145 * @dev: network device to receive on 2146 * @length: length to allocate 2147 * 2148 * Allocate a new &sk_buff and assign it a usage count of one. The 2149 * buffer has unspecified headroom built in. Users should allocate 2150 * the headroom they think they need without accounting for the 2151 * built in space. The built in space is used for optimisations. 2152 * 2153 * %NULL is returned if there is no free memory. Although this function 2154 * allocates memory it can be called from an interrupt. 2155 */ 2156 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 2157 unsigned int length) 2158 { 2159 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 2160 } 2161 2162 /* legacy helper around __netdev_alloc_skb() */ 2163 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 2164 gfp_t gfp_mask) 2165 { 2166 return __netdev_alloc_skb(NULL, length, gfp_mask); 2167 } 2168 2169 /* legacy helper around netdev_alloc_skb() */ 2170 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 2171 { 2172 return netdev_alloc_skb(NULL, length); 2173 } 2174 2175 2176 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 2177 unsigned int length, gfp_t gfp) 2178 { 2179 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 2180 2181 if (NET_IP_ALIGN && skb) 2182 skb_reserve(skb, NET_IP_ALIGN); 2183 return skb; 2184 } 2185 2186 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 2187 unsigned int length) 2188 { 2189 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 2190 } 2191 2192 static inline void skb_free_frag(void *addr) 2193 { 2194 __free_page_frag(addr); 2195 } 2196 2197 void *napi_alloc_frag(unsigned int fragsz); 2198 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, 2199 unsigned int length, gfp_t gfp_mask); 2200 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, 2201 unsigned int length) 2202 { 2203 return __napi_alloc_skb(napi, length, GFP_ATOMIC); 2204 } 2205 2206 /** 2207 * __dev_alloc_pages - allocate page for network Rx 2208 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2209 * @order: size of the allocation 2210 * 2211 * Allocate a new page. 2212 * 2213 * %NULL is returned if there is no free memory. 2214 */ 2215 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, 2216 unsigned int order) 2217 { 2218 /* This piece of code contains several assumptions. 2219 * 1. This is for device Rx, therefor a cold page is preferred. 2220 * 2. The expectation is the user wants a compound page. 2221 * 3. If requesting a order 0 page it will not be compound 2222 * due to the check to see if order has a value in prep_new_page 2223 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to 2224 * code in gfp_to_alloc_flags that should be enforcing this. 2225 */ 2226 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC; 2227 2228 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 2229 } 2230 2231 static inline struct page *dev_alloc_pages(unsigned int order) 2232 { 2233 return __dev_alloc_pages(GFP_ATOMIC, order); 2234 } 2235 2236 /** 2237 * __dev_alloc_page - allocate a page for network Rx 2238 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx 2239 * 2240 * Allocate a new page. 2241 * 2242 * %NULL is returned if there is no free memory. 2243 */ 2244 static inline struct page *__dev_alloc_page(gfp_t gfp_mask) 2245 { 2246 return __dev_alloc_pages(gfp_mask, 0); 2247 } 2248 2249 static inline struct page *dev_alloc_page(void) 2250 { 2251 return __dev_alloc_page(GFP_ATOMIC); 2252 } 2253 2254 /** 2255 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 2256 * @page: The page that was allocated from skb_alloc_page 2257 * @skb: The skb that may need pfmemalloc set 2258 */ 2259 static inline void skb_propagate_pfmemalloc(struct page *page, 2260 struct sk_buff *skb) 2261 { 2262 if (page_is_pfmemalloc(page)) 2263 skb->pfmemalloc = true; 2264 } 2265 2266 /** 2267 * skb_frag_page - retrieve the page referred to by a paged fragment 2268 * @frag: the paged fragment 2269 * 2270 * Returns the &struct page associated with @frag. 2271 */ 2272 static inline struct page *skb_frag_page(const skb_frag_t *frag) 2273 { 2274 return frag->page.p; 2275 } 2276 2277 /** 2278 * __skb_frag_ref - take an addition reference on a paged fragment. 2279 * @frag: the paged fragment 2280 * 2281 * Takes an additional reference on the paged fragment @frag. 2282 */ 2283 static inline void __skb_frag_ref(skb_frag_t *frag) 2284 { 2285 get_page(skb_frag_page(frag)); 2286 } 2287 2288 /** 2289 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 2290 * @skb: the buffer 2291 * @f: the fragment offset. 2292 * 2293 * Takes an additional reference on the @f'th paged fragment of @skb. 2294 */ 2295 static inline void skb_frag_ref(struct sk_buff *skb, int f) 2296 { 2297 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 2298 } 2299 2300 /** 2301 * __skb_frag_unref - release a reference on a paged fragment. 2302 * @frag: the paged fragment 2303 * 2304 * Releases a reference on the paged fragment @frag. 2305 */ 2306 static inline void __skb_frag_unref(skb_frag_t *frag) 2307 { 2308 put_page(skb_frag_page(frag)); 2309 } 2310 2311 /** 2312 * skb_frag_unref - release a reference on a paged fragment of an skb. 2313 * @skb: the buffer 2314 * @f: the fragment offset 2315 * 2316 * Releases a reference on the @f'th paged fragment of @skb. 2317 */ 2318 static inline void skb_frag_unref(struct sk_buff *skb, int f) 2319 { 2320 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 2321 } 2322 2323 /** 2324 * skb_frag_address - gets the address of the data contained in a paged fragment 2325 * @frag: the paged fragment buffer 2326 * 2327 * Returns the address of the data within @frag. The page must already 2328 * be mapped. 2329 */ 2330 static inline void *skb_frag_address(const skb_frag_t *frag) 2331 { 2332 return page_address(skb_frag_page(frag)) + frag->page_offset; 2333 } 2334 2335 /** 2336 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 2337 * @frag: the paged fragment buffer 2338 * 2339 * Returns the address of the data within @frag. Checks that the page 2340 * is mapped and returns %NULL otherwise. 2341 */ 2342 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 2343 { 2344 void *ptr = page_address(skb_frag_page(frag)); 2345 if (unlikely(!ptr)) 2346 return NULL; 2347 2348 return ptr + frag->page_offset; 2349 } 2350 2351 /** 2352 * __skb_frag_set_page - sets the page contained in a paged fragment 2353 * @frag: the paged fragment 2354 * @page: the page to set 2355 * 2356 * Sets the fragment @frag to contain @page. 2357 */ 2358 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 2359 { 2360 frag->page.p = page; 2361 } 2362 2363 /** 2364 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 2365 * @skb: the buffer 2366 * @f: the fragment offset 2367 * @page: the page to set 2368 * 2369 * Sets the @f'th fragment of @skb to contain @page. 2370 */ 2371 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 2372 struct page *page) 2373 { 2374 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 2375 } 2376 2377 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); 2378 2379 /** 2380 * skb_frag_dma_map - maps a paged fragment via the DMA API 2381 * @dev: the device to map the fragment to 2382 * @frag: the paged fragment to map 2383 * @offset: the offset within the fragment (starting at the 2384 * fragment's own offset) 2385 * @size: the number of bytes to map 2386 * @dir: the direction of the mapping (%PCI_DMA_*) 2387 * 2388 * Maps the page associated with @frag to @device. 2389 */ 2390 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 2391 const skb_frag_t *frag, 2392 size_t offset, size_t size, 2393 enum dma_data_direction dir) 2394 { 2395 return dma_map_page(dev, skb_frag_page(frag), 2396 frag->page_offset + offset, size, dir); 2397 } 2398 2399 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 2400 gfp_t gfp_mask) 2401 { 2402 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 2403 } 2404 2405 2406 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, 2407 gfp_t gfp_mask) 2408 { 2409 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); 2410 } 2411 2412 2413 /** 2414 * skb_clone_writable - is the header of a clone writable 2415 * @skb: buffer to check 2416 * @len: length up to which to write 2417 * 2418 * Returns true if modifying the header part of the cloned buffer 2419 * does not requires the data to be copied. 2420 */ 2421 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 2422 { 2423 return !skb_header_cloned(skb) && 2424 skb_headroom(skb) + len <= skb->hdr_len; 2425 } 2426 2427 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 2428 int cloned) 2429 { 2430 int delta = 0; 2431 2432 if (headroom > skb_headroom(skb)) 2433 delta = headroom - skb_headroom(skb); 2434 2435 if (delta || cloned) 2436 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 2437 GFP_ATOMIC); 2438 return 0; 2439 } 2440 2441 /** 2442 * skb_cow - copy header of skb when it is required 2443 * @skb: buffer to cow 2444 * @headroom: needed headroom 2445 * 2446 * If the skb passed lacks sufficient headroom or its data part 2447 * is shared, data is reallocated. If reallocation fails, an error 2448 * is returned and original skb is not changed. 2449 * 2450 * The result is skb with writable area skb->head...skb->tail 2451 * and at least @headroom of space at head. 2452 */ 2453 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 2454 { 2455 return __skb_cow(skb, headroom, skb_cloned(skb)); 2456 } 2457 2458 /** 2459 * skb_cow_head - skb_cow but only making the head writable 2460 * @skb: buffer to cow 2461 * @headroom: needed headroom 2462 * 2463 * This function is identical to skb_cow except that we replace the 2464 * skb_cloned check by skb_header_cloned. It should be used when 2465 * you only need to push on some header and do not need to modify 2466 * the data. 2467 */ 2468 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 2469 { 2470 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 2471 } 2472 2473 /** 2474 * skb_padto - pad an skbuff up to a minimal size 2475 * @skb: buffer to pad 2476 * @len: minimal length 2477 * 2478 * Pads up a buffer to ensure the trailing bytes exist and are 2479 * blanked. If the buffer already contains sufficient data it 2480 * is untouched. Otherwise it is extended. Returns zero on 2481 * success. The skb is freed on error. 2482 */ 2483 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 2484 { 2485 unsigned int size = skb->len; 2486 if (likely(size >= len)) 2487 return 0; 2488 return skb_pad(skb, len - size); 2489 } 2490 2491 /** 2492 * skb_put_padto - increase size and pad an skbuff up to a minimal size 2493 * @skb: buffer to pad 2494 * @len: minimal length 2495 * 2496 * Pads up a buffer to ensure the trailing bytes exist and are 2497 * blanked. If the buffer already contains sufficient data it 2498 * is untouched. Otherwise it is extended. Returns zero on 2499 * success. The skb is freed on error. 2500 */ 2501 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len) 2502 { 2503 unsigned int size = skb->len; 2504 2505 if (unlikely(size < len)) { 2506 len -= size; 2507 if (skb_pad(skb, len)) 2508 return -ENOMEM; 2509 __skb_put(skb, len); 2510 } 2511 return 0; 2512 } 2513 2514 static inline int skb_add_data(struct sk_buff *skb, 2515 struct iov_iter *from, int copy) 2516 { 2517 const int off = skb->len; 2518 2519 if (skb->ip_summed == CHECKSUM_NONE) { 2520 __wsum csum = 0; 2521 if (csum_and_copy_from_iter(skb_put(skb, copy), copy, 2522 &csum, from) == copy) { 2523 skb->csum = csum_block_add(skb->csum, csum, off); 2524 return 0; 2525 } 2526 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy) 2527 return 0; 2528 2529 __skb_trim(skb, off); 2530 return -EFAULT; 2531 } 2532 2533 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 2534 const struct page *page, int off) 2535 { 2536 if (i) { 2537 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; 2538 2539 return page == skb_frag_page(frag) && 2540 off == frag->page_offset + skb_frag_size(frag); 2541 } 2542 return false; 2543 } 2544 2545 static inline int __skb_linearize(struct sk_buff *skb) 2546 { 2547 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 2548 } 2549 2550 /** 2551 * skb_linearize - convert paged skb to linear one 2552 * @skb: buffer to linarize 2553 * 2554 * If there is no free memory -ENOMEM is returned, otherwise zero 2555 * is returned and the old skb data released. 2556 */ 2557 static inline int skb_linearize(struct sk_buff *skb) 2558 { 2559 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 2560 } 2561 2562 /** 2563 * skb_has_shared_frag - can any frag be overwritten 2564 * @skb: buffer to test 2565 * 2566 * Return true if the skb has at least one frag that might be modified 2567 * by an external entity (as in vmsplice()/sendfile()) 2568 */ 2569 static inline bool skb_has_shared_frag(const struct sk_buff *skb) 2570 { 2571 return skb_is_nonlinear(skb) && 2572 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG; 2573 } 2574 2575 /** 2576 * skb_linearize_cow - make sure skb is linear and writable 2577 * @skb: buffer to process 2578 * 2579 * If there is no free memory -ENOMEM is returned, otherwise zero 2580 * is returned and the old skb data released. 2581 */ 2582 static inline int skb_linearize_cow(struct sk_buff *skb) 2583 { 2584 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 2585 __skb_linearize(skb) : 0; 2586 } 2587 2588 /** 2589 * skb_postpull_rcsum - update checksum for received skb after pull 2590 * @skb: buffer to update 2591 * @start: start of data before pull 2592 * @len: length of data pulled 2593 * 2594 * After doing a pull on a received packet, you need to call this to 2595 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 2596 * CHECKSUM_NONE so that it can be recomputed from scratch. 2597 */ 2598 2599 static inline void skb_postpull_rcsum(struct sk_buff *skb, 2600 const void *start, unsigned int len) 2601 { 2602 if (skb->ip_summed == CHECKSUM_COMPLETE) 2603 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); 2604 } 2605 2606 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 2607 2608 /** 2609 * pskb_trim_rcsum - trim received skb and update checksum 2610 * @skb: buffer to trim 2611 * @len: new length 2612 * 2613 * This is exactly the same as pskb_trim except that it ensures the 2614 * checksum of received packets are still valid after the operation. 2615 */ 2616 2617 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 2618 { 2619 if (likely(len >= skb->len)) 2620 return 0; 2621 if (skb->ip_summed == CHECKSUM_COMPLETE) 2622 skb->ip_summed = CHECKSUM_NONE; 2623 return __pskb_trim(skb, len); 2624 } 2625 2626 #define skb_queue_walk(queue, skb) \ 2627 for (skb = (queue)->next; \ 2628 skb != (struct sk_buff *)(queue); \ 2629 skb = skb->next) 2630 2631 #define skb_queue_walk_safe(queue, skb, tmp) \ 2632 for (skb = (queue)->next, tmp = skb->next; \ 2633 skb != (struct sk_buff *)(queue); \ 2634 skb = tmp, tmp = skb->next) 2635 2636 #define skb_queue_walk_from(queue, skb) \ 2637 for (; skb != (struct sk_buff *)(queue); \ 2638 skb = skb->next) 2639 2640 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 2641 for (tmp = skb->next; \ 2642 skb != (struct sk_buff *)(queue); \ 2643 skb = tmp, tmp = skb->next) 2644 2645 #define skb_queue_reverse_walk(queue, skb) \ 2646 for (skb = (queue)->prev; \ 2647 skb != (struct sk_buff *)(queue); \ 2648 skb = skb->prev) 2649 2650 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 2651 for (skb = (queue)->prev, tmp = skb->prev; \ 2652 skb != (struct sk_buff *)(queue); \ 2653 skb = tmp, tmp = skb->prev) 2654 2655 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 2656 for (tmp = skb->prev; \ 2657 skb != (struct sk_buff *)(queue); \ 2658 skb = tmp, tmp = skb->prev) 2659 2660 static inline bool skb_has_frag_list(const struct sk_buff *skb) 2661 { 2662 return skb_shinfo(skb)->frag_list != NULL; 2663 } 2664 2665 static inline void skb_frag_list_init(struct sk_buff *skb) 2666 { 2667 skb_shinfo(skb)->frag_list = NULL; 2668 } 2669 2670 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag) 2671 { 2672 frag->next = skb_shinfo(skb)->frag_list; 2673 skb_shinfo(skb)->frag_list = frag; 2674 } 2675 2676 #define skb_walk_frags(skb, iter) \ 2677 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 2678 2679 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, 2680 int *peeked, int *off, int *err); 2681 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, 2682 int *err); 2683 unsigned int datagram_poll(struct file *file, struct socket *sock, 2684 struct poll_table_struct *wait); 2685 int skb_copy_datagram_iter(const struct sk_buff *from, int offset, 2686 struct iov_iter *to, int size); 2687 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, 2688 struct msghdr *msg, int size) 2689 { 2690 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); 2691 } 2692 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, 2693 struct msghdr *msg); 2694 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, 2695 struct iov_iter *from, int len); 2696 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); 2697 void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 2698 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb); 2699 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); 2700 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); 2701 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); 2702 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, 2703 int len, __wsum csum); 2704 ssize_t skb_socket_splice(struct sock *sk, 2705 struct pipe_inode_info *pipe, 2706 struct splice_pipe_desc *spd); 2707 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, 2708 struct pipe_inode_info *pipe, unsigned int len, 2709 unsigned int flags, 2710 ssize_t (*splice_cb)(struct sock *, 2711 struct pipe_inode_info *, 2712 struct splice_pipe_desc *)); 2713 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 2714 unsigned int skb_zerocopy_headlen(const struct sk_buff *from); 2715 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, 2716 int len, int hlen); 2717 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); 2718 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); 2719 void skb_scrub_packet(struct sk_buff *skb, bool xnet); 2720 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb); 2721 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); 2722 struct sk_buff *skb_vlan_untag(struct sk_buff *skb); 2723 int skb_ensure_writable(struct sk_buff *skb, int write_len); 2724 int skb_vlan_pop(struct sk_buff *skb); 2725 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); 2726 2727 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) 2728 { 2729 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 2730 } 2731 2732 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) 2733 { 2734 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; 2735 } 2736 2737 struct skb_checksum_ops { 2738 __wsum (*update)(const void *mem, int len, __wsum wsum); 2739 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); 2740 }; 2741 2742 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, 2743 __wsum csum, const struct skb_checksum_ops *ops); 2744 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, 2745 __wsum csum); 2746 2747 static inline void * __must_check 2748 __skb_header_pointer(const struct sk_buff *skb, int offset, 2749 int len, void *data, int hlen, void *buffer) 2750 { 2751 if (hlen - offset >= len) 2752 return data + offset; 2753 2754 if (!skb || 2755 skb_copy_bits(skb, offset, buffer, len) < 0) 2756 return NULL; 2757 2758 return buffer; 2759 } 2760 2761 static inline void * __must_check 2762 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) 2763 { 2764 return __skb_header_pointer(skb, offset, len, skb->data, 2765 skb_headlen(skb), buffer); 2766 } 2767 2768 /** 2769 * skb_needs_linearize - check if we need to linearize a given skb 2770 * depending on the given device features. 2771 * @skb: socket buffer to check 2772 * @features: net device features 2773 * 2774 * Returns true if either: 2775 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 2776 * 2. skb is fragmented and the device does not support SG. 2777 */ 2778 static inline bool skb_needs_linearize(struct sk_buff *skb, 2779 netdev_features_t features) 2780 { 2781 return skb_is_nonlinear(skb) && 2782 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || 2783 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); 2784 } 2785 2786 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 2787 void *to, 2788 const unsigned int len) 2789 { 2790 memcpy(to, skb->data, len); 2791 } 2792 2793 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 2794 const int offset, void *to, 2795 const unsigned int len) 2796 { 2797 memcpy(to, skb->data + offset, len); 2798 } 2799 2800 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 2801 const void *from, 2802 const unsigned int len) 2803 { 2804 memcpy(skb->data, from, len); 2805 } 2806 2807 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 2808 const int offset, 2809 const void *from, 2810 const unsigned int len) 2811 { 2812 memcpy(skb->data + offset, from, len); 2813 } 2814 2815 void skb_init(void); 2816 2817 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 2818 { 2819 return skb->tstamp; 2820 } 2821 2822 /** 2823 * skb_get_timestamp - get timestamp from a skb 2824 * @skb: skb to get stamp from 2825 * @stamp: pointer to struct timeval to store stamp in 2826 * 2827 * Timestamps are stored in the skb as offsets to a base timestamp. 2828 * This function converts the offset back to a struct timeval and stores 2829 * it in stamp. 2830 */ 2831 static inline void skb_get_timestamp(const struct sk_buff *skb, 2832 struct timeval *stamp) 2833 { 2834 *stamp = ktime_to_timeval(skb->tstamp); 2835 } 2836 2837 static inline void skb_get_timestampns(const struct sk_buff *skb, 2838 struct timespec *stamp) 2839 { 2840 *stamp = ktime_to_timespec(skb->tstamp); 2841 } 2842 2843 static inline void __net_timestamp(struct sk_buff *skb) 2844 { 2845 skb->tstamp = ktime_get_real(); 2846 } 2847 2848 static inline ktime_t net_timedelta(ktime_t t) 2849 { 2850 return ktime_sub(ktime_get_real(), t); 2851 } 2852 2853 static inline ktime_t net_invalid_timestamp(void) 2854 { 2855 return ktime_set(0, 0); 2856 } 2857 2858 struct sk_buff *skb_clone_sk(struct sk_buff *skb); 2859 2860 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 2861 2862 void skb_clone_tx_timestamp(struct sk_buff *skb); 2863 bool skb_defer_rx_timestamp(struct sk_buff *skb); 2864 2865 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 2866 2867 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 2868 { 2869 } 2870 2871 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 2872 { 2873 return false; 2874 } 2875 2876 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 2877 2878 /** 2879 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 2880 * 2881 * PHY drivers may accept clones of transmitted packets for 2882 * timestamping via their phy_driver.txtstamp method. These drivers 2883 * must call this function to return the skb back to the stack with a 2884 * timestamp. 2885 * 2886 * @skb: clone of the the original outgoing packet 2887 * @hwtstamps: hardware time stamps 2888 * 2889 */ 2890 void skb_complete_tx_timestamp(struct sk_buff *skb, 2891 struct skb_shared_hwtstamps *hwtstamps); 2892 2893 void __skb_tstamp_tx(struct sk_buff *orig_skb, 2894 struct skb_shared_hwtstamps *hwtstamps, 2895 struct sock *sk, int tstype); 2896 2897 /** 2898 * skb_tstamp_tx - queue clone of skb with send time stamps 2899 * @orig_skb: the original outgoing packet 2900 * @hwtstamps: hardware time stamps, may be NULL if not available 2901 * 2902 * If the skb has a socket associated, then this function clones the 2903 * skb (thus sharing the actual data and optional structures), stores 2904 * the optional hardware time stamping information (if non NULL) or 2905 * generates a software time stamp (otherwise), then queues the clone 2906 * to the error queue of the socket. Errors are silently ignored. 2907 */ 2908 void skb_tstamp_tx(struct sk_buff *orig_skb, 2909 struct skb_shared_hwtstamps *hwtstamps); 2910 2911 static inline void sw_tx_timestamp(struct sk_buff *skb) 2912 { 2913 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP && 2914 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) 2915 skb_tstamp_tx(skb, NULL); 2916 } 2917 2918 /** 2919 * skb_tx_timestamp() - Driver hook for transmit timestamping 2920 * 2921 * Ethernet MAC Drivers should call this function in their hard_xmit() 2922 * function immediately before giving the sk_buff to the MAC hardware. 2923 * 2924 * Specifically, one should make absolutely sure that this function is 2925 * called before TX completion of this packet can trigger. Otherwise 2926 * the packet could potentially already be freed. 2927 * 2928 * @skb: A socket buffer. 2929 */ 2930 static inline void skb_tx_timestamp(struct sk_buff *skb) 2931 { 2932 skb_clone_tx_timestamp(skb); 2933 sw_tx_timestamp(skb); 2934 } 2935 2936 /** 2937 * skb_complete_wifi_ack - deliver skb with wifi status 2938 * 2939 * @skb: the original outgoing packet 2940 * @acked: ack status 2941 * 2942 */ 2943 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 2944 2945 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 2946 __sum16 __skb_checksum_complete(struct sk_buff *skb); 2947 2948 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 2949 { 2950 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || 2951 skb->csum_valid || 2952 (skb->ip_summed == CHECKSUM_PARTIAL && 2953 skb_checksum_start_offset(skb) >= 0)); 2954 } 2955 2956 /** 2957 * skb_checksum_complete - Calculate checksum of an entire packet 2958 * @skb: packet to process 2959 * 2960 * This function calculates the checksum over the entire packet plus 2961 * the value of skb->csum. The latter can be used to supply the 2962 * checksum of a pseudo header as used by TCP/UDP. It returns the 2963 * checksum. 2964 * 2965 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 2966 * this function can be used to verify that checksum on received 2967 * packets. In that case the function should return zero if the 2968 * checksum is correct. In particular, this function will return zero 2969 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 2970 * hardware has already verified the correctness of the checksum. 2971 */ 2972 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 2973 { 2974 return skb_csum_unnecessary(skb) ? 2975 0 : __skb_checksum_complete(skb); 2976 } 2977 2978 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) 2979 { 2980 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 2981 if (skb->csum_level == 0) 2982 skb->ip_summed = CHECKSUM_NONE; 2983 else 2984 skb->csum_level--; 2985 } 2986 } 2987 2988 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) 2989 { 2990 if (skb->ip_summed == CHECKSUM_UNNECESSARY) { 2991 if (skb->csum_level < SKB_MAX_CSUM_LEVEL) 2992 skb->csum_level++; 2993 } else if (skb->ip_summed == CHECKSUM_NONE) { 2994 skb->ip_summed = CHECKSUM_UNNECESSARY; 2995 skb->csum_level = 0; 2996 } 2997 } 2998 2999 static inline void __skb_mark_checksum_bad(struct sk_buff *skb) 3000 { 3001 /* Mark current checksum as bad (typically called from GRO 3002 * path). In the case that ip_summed is CHECKSUM_NONE 3003 * this must be the first checksum encountered in the packet. 3004 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first 3005 * checksum after the last one validated. For UDP, a zero 3006 * checksum can not be marked as bad. 3007 */ 3008 3009 if (skb->ip_summed == CHECKSUM_NONE || 3010 skb->ip_summed == CHECKSUM_UNNECESSARY) 3011 skb->csum_bad = 1; 3012 } 3013 3014 /* Check if we need to perform checksum complete validation. 3015 * 3016 * Returns true if checksum complete is needed, false otherwise 3017 * (either checksum is unnecessary or zero checksum is allowed). 3018 */ 3019 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, 3020 bool zero_okay, 3021 __sum16 check) 3022 { 3023 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { 3024 skb->csum_valid = 1; 3025 __skb_decr_checksum_unnecessary(skb); 3026 return false; 3027 } 3028 3029 return true; 3030 } 3031 3032 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly 3033 * in checksum_init. 3034 */ 3035 #define CHECKSUM_BREAK 76 3036 3037 /* Unset checksum-complete 3038 * 3039 * Unset checksum complete can be done when packet is being modified 3040 * (uncompressed for instance) and checksum-complete value is 3041 * invalidated. 3042 */ 3043 static inline void skb_checksum_complete_unset(struct sk_buff *skb) 3044 { 3045 if (skb->ip_summed == CHECKSUM_COMPLETE) 3046 skb->ip_summed = CHECKSUM_NONE; 3047 } 3048 3049 /* Validate (init) checksum based on checksum complete. 3050 * 3051 * Return values: 3052 * 0: checksum is validated or try to in skb_checksum_complete. In the latter 3053 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo 3054 * checksum is stored in skb->csum for use in __skb_checksum_complete 3055 * non-zero: value of invalid checksum 3056 * 3057 */ 3058 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, 3059 bool complete, 3060 __wsum psum) 3061 { 3062 if (skb->ip_summed == CHECKSUM_COMPLETE) { 3063 if (!csum_fold(csum_add(psum, skb->csum))) { 3064 skb->csum_valid = 1; 3065 return 0; 3066 } 3067 } else if (skb->csum_bad) { 3068 /* ip_summed == CHECKSUM_NONE in this case */ 3069 return (__force __sum16)1; 3070 } 3071 3072 skb->csum = psum; 3073 3074 if (complete || skb->len <= CHECKSUM_BREAK) { 3075 __sum16 csum; 3076 3077 csum = __skb_checksum_complete(skb); 3078 skb->csum_valid = !csum; 3079 return csum; 3080 } 3081 3082 return 0; 3083 } 3084 3085 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) 3086 { 3087 return 0; 3088 } 3089 3090 /* Perform checksum validate (init). Note that this is a macro since we only 3091 * want to calculate the pseudo header which is an input function if necessary. 3092 * First we try to validate without any computation (checksum unnecessary) and 3093 * then calculate based on checksum complete calling the function to compute 3094 * pseudo header. 3095 * 3096 * Return values: 3097 * 0: checksum is validated or try to in skb_checksum_complete 3098 * non-zero: value of invalid checksum 3099 */ 3100 #define __skb_checksum_validate(skb, proto, complete, \ 3101 zero_okay, check, compute_pseudo) \ 3102 ({ \ 3103 __sum16 __ret = 0; \ 3104 skb->csum_valid = 0; \ 3105 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ 3106 __ret = __skb_checksum_validate_complete(skb, \ 3107 complete, compute_pseudo(skb, proto)); \ 3108 __ret; \ 3109 }) 3110 3111 #define skb_checksum_init(skb, proto, compute_pseudo) \ 3112 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) 3113 3114 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ 3115 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) 3116 3117 #define skb_checksum_validate(skb, proto, compute_pseudo) \ 3118 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) 3119 3120 #define skb_checksum_validate_zero_check(skb, proto, check, \ 3121 compute_pseudo) \ 3122 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) 3123 3124 #define skb_checksum_simple_validate(skb) \ 3125 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) 3126 3127 static inline bool __skb_checksum_convert_check(struct sk_buff *skb) 3128 { 3129 return (skb->ip_summed == CHECKSUM_NONE && 3130 skb->csum_valid && !skb->csum_bad); 3131 } 3132 3133 static inline void __skb_checksum_convert(struct sk_buff *skb, 3134 __sum16 check, __wsum pseudo) 3135 { 3136 skb->csum = ~pseudo; 3137 skb->ip_summed = CHECKSUM_COMPLETE; 3138 } 3139 3140 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \ 3141 do { \ 3142 if (__skb_checksum_convert_check(skb)) \ 3143 __skb_checksum_convert(skb, check, \ 3144 compute_pseudo(skb, proto)); \ 3145 } while (0) 3146 3147 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, 3148 u16 start, u16 offset) 3149 { 3150 skb->ip_summed = CHECKSUM_PARTIAL; 3151 skb->csum_start = ((unsigned char *)ptr + start) - skb->head; 3152 skb->csum_offset = offset - start; 3153 } 3154 3155 /* Update skbuf and packet to reflect the remote checksum offload operation. 3156 * When called, ptr indicates the starting point for skb->csum when 3157 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete 3158 * here, skb_postpull_rcsum is done so skb->csum start is ptr. 3159 */ 3160 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, 3161 int start, int offset, bool nopartial) 3162 { 3163 __wsum delta; 3164 3165 if (!nopartial) { 3166 skb_remcsum_adjust_partial(skb, ptr, start, offset); 3167 return; 3168 } 3169 3170 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { 3171 __skb_checksum_complete(skb); 3172 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); 3173 } 3174 3175 delta = remcsum_adjust(ptr, skb->csum, start, offset); 3176 3177 /* Adjust skb->csum since we changed the packet */ 3178 skb->csum = csum_add(skb->csum, delta); 3179 } 3180 3181 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 3182 void nf_conntrack_destroy(struct nf_conntrack *nfct); 3183 static inline void nf_conntrack_put(struct nf_conntrack *nfct) 3184 { 3185 if (nfct && atomic_dec_and_test(&nfct->use)) 3186 nf_conntrack_destroy(nfct); 3187 } 3188 static inline void nf_conntrack_get(struct nf_conntrack *nfct) 3189 { 3190 if (nfct) 3191 atomic_inc(&nfct->use); 3192 } 3193 #endif 3194 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 3195 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) 3196 { 3197 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) 3198 kfree(nf_bridge); 3199 } 3200 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) 3201 { 3202 if (nf_bridge) 3203 atomic_inc(&nf_bridge->use); 3204 } 3205 #endif /* CONFIG_BRIDGE_NETFILTER */ 3206 static inline void nf_reset(struct sk_buff *skb) 3207 { 3208 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 3209 nf_conntrack_put(skb->nfct); 3210 skb->nfct = NULL; 3211 #endif 3212 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 3213 nf_bridge_put(skb->nf_bridge); 3214 skb->nf_bridge = NULL; 3215 #endif 3216 } 3217 3218 static inline void nf_reset_trace(struct sk_buff *skb) 3219 { 3220 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 3221 skb->nf_trace = 0; 3222 #endif 3223 } 3224 3225 /* Note: This doesn't put any conntrack and bridge info in dst. */ 3226 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, 3227 bool copy) 3228 { 3229 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 3230 dst->nfct = src->nfct; 3231 nf_conntrack_get(src->nfct); 3232 if (copy) 3233 dst->nfctinfo = src->nfctinfo; 3234 #endif 3235 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 3236 dst->nf_bridge = src->nf_bridge; 3237 nf_bridge_get(src->nf_bridge); 3238 #endif 3239 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) 3240 if (copy) 3241 dst->nf_trace = src->nf_trace; 3242 #endif 3243 } 3244 3245 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 3246 { 3247 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 3248 nf_conntrack_put(dst->nfct); 3249 #endif 3250 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 3251 nf_bridge_put(dst->nf_bridge); 3252 #endif 3253 __nf_copy(dst, src, true); 3254 } 3255 3256 #ifdef CONFIG_NETWORK_SECMARK 3257 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 3258 { 3259 to->secmark = from->secmark; 3260 } 3261 3262 static inline void skb_init_secmark(struct sk_buff *skb) 3263 { 3264 skb->secmark = 0; 3265 } 3266 #else 3267 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 3268 { } 3269 3270 static inline void skb_init_secmark(struct sk_buff *skb) 3271 { } 3272 #endif 3273 3274 static inline bool skb_irq_freeable(const struct sk_buff *skb) 3275 { 3276 return !skb->destructor && 3277 #if IS_ENABLED(CONFIG_XFRM) 3278 !skb->sp && 3279 #endif 3280 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 3281 !skb->nfct && 3282 #endif 3283 !skb->_skb_refdst && 3284 !skb_has_frag_list(skb); 3285 } 3286 3287 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 3288 { 3289 skb->queue_mapping = queue_mapping; 3290 } 3291 3292 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 3293 { 3294 return skb->queue_mapping; 3295 } 3296 3297 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 3298 { 3299 to->queue_mapping = from->queue_mapping; 3300 } 3301 3302 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 3303 { 3304 skb->queue_mapping = rx_queue + 1; 3305 } 3306 3307 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 3308 { 3309 return skb->queue_mapping - 1; 3310 } 3311 3312 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 3313 { 3314 return skb->queue_mapping != 0; 3315 } 3316 3317 static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 3318 { 3319 #ifdef CONFIG_XFRM 3320 return skb->sp; 3321 #else 3322 return NULL; 3323 #endif 3324 } 3325 3326 /* Keeps track of mac header offset relative to skb->head. 3327 * It is useful for TSO of Tunneling protocol. e.g. GRE. 3328 * For non-tunnel skb it points to skb_mac_header() and for 3329 * tunnel skb it points to outer mac header. 3330 * Keeps track of level of encapsulation of network headers. 3331 */ 3332 struct skb_gso_cb { 3333 int mac_offset; 3334 int encap_level; 3335 __u16 csum_start; 3336 }; 3337 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb) 3338 3339 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) 3340 { 3341 return (skb_mac_header(inner_skb) - inner_skb->head) - 3342 SKB_GSO_CB(inner_skb)->mac_offset; 3343 } 3344 3345 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) 3346 { 3347 int new_headroom, headroom; 3348 int ret; 3349 3350 headroom = skb_headroom(skb); 3351 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); 3352 if (ret) 3353 return ret; 3354 3355 new_headroom = skb_headroom(skb); 3356 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); 3357 return 0; 3358 } 3359 3360 /* Compute the checksum for a gso segment. First compute the checksum value 3361 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and 3362 * then add in skb->csum (checksum from csum_start to end of packet). 3363 * skb->csum and csum_start are then updated to reflect the checksum of the 3364 * resultant packet starting from the transport header-- the resultant checksum 3365 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo 3366 * header. 3367 */ 3368 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) 3369 { 3370 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) - 3371 skb_transport_offset(skb); 3372 __wsum partial; 3373 3374 partial = csum_partial(skb_transport_header(skb), plen, skb->csum); 3375 skb->csum = res; 3376 SKB_GSO_CB(skb)->csum_start -= plen; 3377 3378 return csum_fold(partial); 3379 } 3380 3381 static inline bool skb_is_gso(const struct sk_buff *skb) 3382 { 3383 return skb_shinfo(skb)->gso_size; 3384 } 3385 3386 /* Note: Should be called only if skb_is_gso(skb) is true */ 3387 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 3388 { 3389 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 3390 } 3391 3392 void __skb_warn_lro_forwarding(const struct sk_buff *skb); 3393 3394 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 3395 { 3396 /* LRO sets gso_size but not gso_type, whereas if GSO is really 3397 * wanted then gso_type will be set. */ 3398 const struct skb_shared_info *shinfo = skb_shinfo(skb); 3399 3400 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 3401 unlikely(shinfo->gso_type == 0)) { 3402 __skb_warn_lro_forwarding(skb); 3403 return true; 3404 } 3405 return false; 3406 } 3407 3408 static inline void skb_forward_csum(struct sk_buff *skb) 3409 { 3410 /* Unfortunately we don't support this one. Any brave souls? */ 3411 if (skb->ip_summed == CHECKSUM_COMPLETE) 3412 skb->ip_summed = CHECKSUM_NONE; 3413 } 3414 3415 /** 3416 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 3417 * @skb: skb to check 3418 * 3419 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 3420 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 3421 * use this helper, to document places where we make this assertion. 3422 */ 3423 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 3424 { 3425 #ifdef DEBUG 3426 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 3427 #endif 3428 } 3429 3430 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 3431 3432 int skb_checksum_setup(struct sk_buff *skb, bool recalculate); 3433 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, 3434 unsigned int transport_len, 3435 __sum16(*skb_chkf)(struct sk_buff *skb)); 3436 3437 /** 3438 * skb_head_is_locked - Determine if the skb->head is locked down 3439 * @skb: skb to check 3440 * 3441 * The head on skbs build around a head frag can be removed if they are 3442 * not cloned. This function returns true if the skb head is locked down 3443 * due to either being allocated via kmalloc, or by being a clone with 3444 * multiple references to the head. 3445 */ 3446 static inline bool skb_head_is_locked(const struct sk_buff *skb) 3447 { 3448 return !skb->head_frag || skb_cloned(skb); 3449 } 3450 3451 /** 3452 * skb_gso_network_seglen - Return length of individual segments of a gso packet 3453 * 3454 * @skb: GSO skb 3455 * 3456 * skb_gso_network_seglen is used to determine the real size of the 3457 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP). 3458 * 3459 * The MAC/L2 header is not accounted for. 3460 */ 3461 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb) 3462 { 3463 unsigned int hdr_len = skb_transport_header(skb) - 3464 skb_network_header(skb); 3465 return hdr_len + skb_gso_transport_seglen(skb); 3466 } 3467 #endif /* __KERNEL__ */ 3468 #endif /* _LINUX_SKBUFF_H */ 3469