at v2.6.30 58 kB view raw
1/* 2 * Definitions for the 'struct sk_buff' memory handlers. 3 * 4 * Authors: 5 * Alan Cox, <gw4pts@gw4pts.ampr.org> 6 * Florian La Roche, <rzsfl@rz.uni-sb.de> 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/compiler.h> 19#include <linux/time.h> 20#include <linux/cache.h> 21 22#include <asm/atomic.h> 23#include <asm/types.h> 24#include <linux/spinlock.h> 25#include <linux/net.h> 26#include <linux/textsearch.h> 27#include <net/checksum.h> 28#include <linux/rcupdate.h> 29#include <linux/dmaengine.h> 30#include <linux/hrtimer.h> 31 32/* Don't change this without changing skb_csum_unnecessary! */ 33#define CHECKSUM_NONE 0 34#define CHECKSUM_UNNECESSARY 1 35#define CHECKSUM_COMPLETE 2 36#define CHECKSUM_PARTIAL 3 37 38#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \ 39 ~(SMP_CACHE_BYTES - 1)) 40#define SKB_WITH_OVERHEAD(X) \ 41 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 42#define SKB_MAX_ORDER(X, ORDER) \ 43 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 44#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 45#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 46 47/* A. Checksumming of received packets by device. 48 * 49 * NONE: device failed to checksum this packet. 50 * skb->csum is undefined. 51 * 52 * UNNECESSARY: device parsed packet and wouldbe verified checksum. 53 * skb->csum is undefined. 54 * It is bad option, but, unfortunately, many of vendors do this. 55 * Apparently with secret goal to sell you new device, when you 56 * will add new protocol to your host. F.e. IPv6. 8) 57 * 58 * COMPLETE: the most generic way. Device supplied checksum of _all_ 59 * the packet as seen by netif_rx in skb->csum. 60 * NOTE: Even if device supports only some protocols, but 61 * is able to produce some skb->csum, it MUST use COMPLETE, 62 * not UNNECESSARY. 63 * 64 * PARTIAL: identical to the case for output below. This may occur 65 * on a packet received directly from another Linux OS, e.g., 66 * a virtualised Linux kernel on the same host. The packet can 67 * be treated in the same way as UNNECESSARY except that on 68 * output (i.e., forwarding) the checksum must be filled in 69 * by the OS or the hardware. 70 * 71 * B. Checksumming on output. 72 * 73 * NONE: skb is checksummed by protocol or csum is not required. 74 * 75 * PARTIAL: device is required to csum packet as seen by hard_start_xmit 76 * from skb->csum_start to the end and to record the checksum 77 * at skb->csum_start + skb->csum_offset. 78 * 79 * Device must show its capabilities in dev->features, set 80 * at device setup time. 81 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum 82 * everything. 83 * NETIF_F_NO_CSUM - loopback or reliable single hop media. 84 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only 85 * TCP/UDP over IPv4. Sigh. Vendors like this 86 * way by an unknown reason. Though, see comment above 87 * about CHECKSUM_UNNECESSARY. 8) 88 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead. 89 * 90 * Any questions? No questions, good. --ANK 91 */ 92 93struct net_device; 94struct scatterlist; 95struct pipe_inode_info; 96 97#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 98struct nf_conntrack { 99 atomic_t use; 100}; 101#endif 102 103#ifdef CONFIG_BRIDGE_NETFILTER 104struct nf_bridge_info { 105 atomic_t use; 106 struct net_device *physindev; 107 struct net_device *physoutdev; 108 unsigned int mask; 109 unsigned long data[32 / sizeof(unsigned long)]; 110}; 111#endif 112 113struct sk_buff_head { 114 /* These two members must be first. */ 115 struct sk_buff *next; 116 struct sk_buff *prev; 117 118 __u32 qlen; 119 spinlock_t lock; 120}; 121 122struct sk_buff; 123 124/* To allow 64K frame to be packed as single skb without frag_list */ 125#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2) 126 127typedef struct skb_frag_struct skb_frag_t; 128 129struct skb_frag_struct { 130 struct page *page; 131 __u32 page_offset; 132 __u32 size; 133}; 134 135#define HAVE_HW_TIME_STAMP 136 137/** 138 * struct skb_shared_hwtstamps - hardware time stamps 139 * @hwtstamp: hardware time stamp transformed into duration 140 * since arbitrary point in time 141 * @syststamp: hwtstamp transformed to system time base 142 * 143 * Software time stamps generated by ktime_get_real() are stored in 144 * skb->tstamp. The relation between the different kinds of time 145 * stamps is as follows: 146 * 147 * syststamp and tstamp can be compared against each other in 148 * arbitrary combinations. The accuracy of a 149 * syststamp/tstamp/"syststamp from other device" comparison is 150 * limited by the accuracy of the transformation into system time 151 * base. This depends on the device driver and its underlying 152 * hardware. 153 * 154 * hwtstamps can only be compared against other hwtstamps from 155 * the same device. 156 * 157 * This structure is attached to packets as part of the 158 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 159 */ 160struct skb_shared_hwtstamps { 161 ktime_t hwtstamp; 162 ktime_t syststamp; 163}; 164 165/** 166 * struct skb_shared_tx - instructions for time stamping of outgoing packets 167 * @hardware: generate hardware time stamp 168 * @software: generate software time stamp 169 * @in_progress: device driver is going to provide 170 * hardware time stamp 171 * @flags: all shared_tx flags 172 * 173 * These flags are attached to packets as part of the 174 * &skb_shared_info. Use skb_tx() to get a pointer. 175 */ 176union skb_shared_tx { 177 struct { 178 __u8 hardware:1, 179 software:1, 180 in_progress:1; 181 }; 182 __u8 flags; 183}; 184 185/* This data is invariant across clones and lives at 186 * the end of the header data, ie. at skb->end. 187 */ 188struct skb_shared_info { 189 atomic_t dataref; 190 unsigned short nr_frags; 191 unsigned short gso_size; 192 /* Warning: this field is not always filled in (UFO)! */ 193 unsigned short gso_segs; 194 unsigned short gso_type; 195 __be32 ip6_frag_id; 196 union skb_shared_tx tx_flags; 197#ifdef CONFIG_HAS_DMA 198 unsigned int num_dma_maps; 199#endif 200 struct sk_buff *frag_list; 201 struct skb_shared_hwtstamps hwtstamps; 202 skb_frag_t frags[MAX_SKB_FRAGS]; 203#ifdef CONFIG_HAS_DMA 204 dma_addr_t dma_maps[MAX_SKB_FRAGS + 1]; 205#endif 206}; 207 208/* We divide dataref into two halves. The higher 16 bits hold references 209 * to the payload part of skb->data. The lower 16 bits hold references to 210 * the entire skb->data. A clone of a headerless skb holds the length of 211 * the header in skb->hdr_len. 212 * 213 * All users must obey the rule that the skb->data reference count must be 214 * greater than or equal to the payload reference count. 215 * 216 * Holding a reference to the payload part means that the user does not 217 * care about modifications to the header part of skb->data. 218 */ 219#define SKB_DATAREF_SHIFT 16 220#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 221 222 223enum { 224 SKB_FCLONE_UNAVAILABLE, 225 SKB_FCLONE_ORIG, 226 SKB_FCLONE_CLONE, 227}; 228 229enum { 230 SKB_GSO_TCPV4 = 1 << 0, 231 SKB_GSO_UDP = 1 << 1, 232 233 /* This indicates the skb is from an untrusted source. */ 234 SKB_GSO_DODGY = 1 << 2, 235 236 /* This indicates the tcp segment has CWR set. */ 237 SKB_GSO_TCP_ECN = 1 << 3, 238 239 SKB_GSO_TCPV6 = 1 << 4, 240 241 SKB_GSO_FCOE = 1 << 5, 242}; 243 244#if BITS_PER_LONG > 32 245#define NET_SKBUFF_DATA_USES_OFFSET 1 246#endif 247 248#ifdef NET_SKBUFF_DATA_USES_OFFSET 249typedef unsigned int sk_buff_data_t; 250#else 251typedef unsigned char *sk_buff_data_t; 252#endif 253 254/** 255 * struct sk_buff - socket buffer 256 * @next: Next buffer in list 257 * @prev: Previous buffer in list 258 * @sk: Socket we are owned by 259 * @tstamp: Time we arrived 260 * @dev: Device we arrived on/are leaving by 261 * @transport_header: Transport layer header 262 * @network_header: Network layer header 263 * @mac_header: Link layer header 264 * @dst: destination entry 265 * @sp: the security path, used for xfrm 266 * @cb: Control buffer. Free for use by every layer. Put private vars here 267 * @len: Length of actual data 268 * @data_len: Data length 269 * @mac_len: Length of link layer header 270 * @hdr_len: writable header length of cloned skb 271 * @csum: Checksum (must include start/offset pair) 272 * @csum_start: Offset from skb->head where checksumming should start 273 * @csum_offset: Offset from csum_start where checksum should be stored 274 * @local_df: allow local fragmentation 275 * @cloned: Head may be cloned (check refcnt to be sure) 276 * @nohdr: Payload reference only, must not modify header 277 * @pkt_type: Packet class 278 * @fclone: skbuff clone status 279 * @ip_summed: Driver fed us an IP checksum 280 * @priority: Packet queueing priority 281 * @users: User count - see {datagram,tcp}.c 282 * @protocol: Packet protocol from driver 283 * @truesize: Buffer size 284 * @head: Head of buffer 285 * @data: Data head pointer 286 * @tail: Tail pointer 287 * @end: End pointer 288 * @destructor: Destruct function 289 * @mark: Generic packet mark 290 * @nfct: Associated connection, if any 291 * @ipvs_property: skbuff is owned by ipvs 292 * @peeked: this packet has been seen already, so stats have been 293 * done for it, don't do them again 294 * @nf_trace: netfilter packet trace flag 295 * @nfctinfo: Relationship of this skb to the connection 296 * @nfct_reasm: netfilter conntrack re-assembly pointer 297 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 298 * @iif: ifindex of device we arrived on 299 * @queue_mapping: Queue mapping for multiqueue devices 300 * @tc_index: Traffic control index 301 * @tc_verd: traffic control verdict 302 * @ndisc_nodetype: router type (from link layer) 303 * @do_not_encrypt: set to prevent encryption of this frame 304 * @requeue: set to indicate that the wireless core should attempt 305 * a software retry on this frame if we failed to 306 * receive an ACK for it 307 * @dma_cookie: a cookie to one of several possible DMA operations 308 * done by skb DMA functions 309 * @secmark: security marking 310 * @vlan_tci: vlan tag control information 311 */ 312 313struct sk_buff { 314 /* These two members must be first. */ 315 struct sk_buff *next; 316 struct sk_buff *prev; 317 318 struct sock *sk; 319 ktime_t tstamp; 320 struct net_device *dev; 321 322 union { 323 struct dst_entry *dst; 324 struct rtable *rtable; 325 }; 326#ifdef CONFIG_XFRM 327 struct sec_path *sp; 328#endif 329 /* 330 * This is the control buffer. It is free to use for every 331 * layer. Please put your private variables there. If you 332 * want to keep them across layers you have to do a skb_clone() 333 * first. This is owned by whoever has the skb queued ATM. 334 */ 335 char cb[48]; 336 337 unsigned int len, 338 data_len; 339 __u16 mac_len, 340 hdr_len; 341 union { 342 __wsum csum; 343 struct { 344 __u16 csum_start; 345 __u16 csum_offset; 346 }; 347 }; 348 __u32 priority; 349 __u8 local_df:1, 350 cloned:1, 351 ip_summed:2, 352 nohdr:1, 353 nfctinfo:3; 354 __u8 pkt_type:3, 355 fclone:2, 356 ipvs_property:1, 357 peeked:1, 358 nf_trace:1; 359 __be16 protocol; 360 361 void (*destructor)(struct sk_buff *skb); 362#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 363 struct nf_conntrack *nfct; 364 struct sk_buff *nfct_reasm; 365#endif 366#ifdef CONFIG_BRIDGE_NETFILTER 367 struct nf_bridge_info *nf_bridge; 368#endif 369 370 int iif; 371 __u16 queue_mapping; 372#ifdef CONFIG_NET_SCHED 373 __u16 tc_index; /* traffic control index */ 374#ifdef CONFIG_NET_CLS_ACT 375 __u16 tc_verd; /* traffic control verdict */ 376#endif 377#endif 378#ifdef CONFIG_IPV6_NDISC_NODETYPE 379 __u8 ndisc_nodetype:2; 380#endif 381#if defined(CONFIG_MAC80211) || defined(CONFIG_MAC80211_MODULE) 382 __u8 do_not_encrypt:1; 383 __u8 requeue:1; 384#endif 385 /* 0/13/14 bit hole */ 386 387#ifdef CONFIG_NET_DMA 388 dma_cookie_t dma_cookie; 389#endif 390#ifdef CONFIG_NETWORK_SECMARK 391 __u32 secmark; 392#endif 393 394 __u32 mark; 395 396 __u16 vlan_tci; 397 398 sk_buff_data_t transport_header; 399 sk_buff_data_t network_header; 400 sk_buff_data_t mac_header; 401 /* These elements must be at the end, see alloc_skb() for details. */ 402 sk_buff_data_t tail; 403 sk_buff_data_t end; 404 unsigned char *head, 405 *data; 406 unsigned int truesize; 407 atomic_t users; 408}; 409 410#ifdef __KERNEL__ 411/* 412 * Handling routines are only of interest to the kernel 413 */ 414#include <linux/slab.h> 415 416#include <asm/system.h> 417 418#ifdef CONFIG_HAS_DMA 419#include <linux/dma-mapping.h> 420extern int skb_dma_map(struct device *dev, struct sk_buff *skb, 421 enum dma_data_direction dir); 422extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb, 423 enum dma_data_direction dir); 424#endif 425 426extern void kfree_skb(struct sk_buff *skb); 427extern void consume_skb(struct sk_buff *skb); 428extern void __kfree_skb(struct sk_buff *skb); 429extern struct sk_buff *__alloc_skb(unsigned int size, 430 gfp_t priority, int fclone, int node); 431static inline struct sk_buff *alloc_skb(unsigned int size, 432 gfp_t priority) 433{ 434 return __alloc_skb(size, priority, 0, -1); 435} 436 437static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 438 gfp_t priority) 439{ 440 return __alloc_skb(size, priority, 1, -1); 441} 442 443extern int skb_recycle_check(struct sk_buff *skb, int skb_size); 444 445extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 446extern struct sk_buff *skb_clone(struct sk_buff *skb, 447 gfp_t priority); 448extern struct sk_buff *skb_copy(const struct sk_buff *skb, 449 gfp_t priority); 450extern struct sk_buff *pskb_copy(struct sk_buff *skb, 451 gfp_t gfp_mask); 452extern int pskb_expand_head(struct sk_buff *skb, 453 int nhead, int ntail, 454 gfp_t gfp_mask); 455extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 456 unsigned int headroom); 457extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb, 458 int newheadroom, int newtailroom, 459 gfp_t priority); 460extern int skb_to_sgvec(struct sk_buff *skb, 461 struct scatterlist *sg, int offset, 462 int len); 463extern int skb_cow_data(struct sk_buff *skb, int tailbits, 464 struct sk_buff **trailer); 465extern int skb_pad(struct sk_buff *skb, int pad); 466#define dev_kfree_skb(a) consume_skb(a) 467#define dev_consume_skb(a) kfree_skb_clean(a) 468extern void skb_over_panic(struct sk_buff *skb, int len, 469 void *here); 470extern void skb_under_panic(struct sk_buff *skb, int len, 471 void *here); 472 473extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, 474 int getfrag(void *from, char *to, int offset, 475 int len,int odd, struct sk_buff *skb), 476 void *from, int length); 477 478struct skb_seq_state 479{ 480 __u32 lower_offset; 481 __u32 upper_offset; 482 __u32 frag_idx; 483 __u32 stepped_offset; 484 struct sk_buff *root_skb; 485 struct sk_buff *cur_skb; 486 __u8 *frag_data; 487}; 488 489extern void skb_prepare_seq_read(struct sk_buff *skb, 490 unsigned int from, unsigned int to, 491 struct skb_seq_state *st); 492extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 493 struct skb_seq_state *st); 494extern void skb_abort_seq_read(struct skb_seq_state *st); 495 496extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 497 unsigned int to, struct ts_config *config, 498 struct ts_state *state); 499 500#ifdef NET_SKBUFF_DATA_USES_OFFSET 501static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 502{ 503 return skb->head + skb->end; 504} 505#else 506static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 507{ 508 return skb->end; 509} 510#endif 511 512/* Internal */ 513#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 514 515static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 516{ 517 return &skb_shinfo(skb)->hwtstamps; 518} 519 520static inline union skb_shared_tx *skb_tx(struct sk_buff *skb) 521{ 522 return &skb_shinfo(skb)->tx_flags; 523} 524 525/** 526 * skb_queue_empty - check if a queue is empty 527 * @list: queue head 528 * 529 * Returns true if the queue is empty, false otherwise. 530 */ 531static inline int skb_queue_empty(const struct sk_buff_head *list) 532{ 533 return list->next == (struct sk_buff *)list; 534} 535 536/** 537 * skb_queue_is_last - check if skb is the last entry in the queue 538 * @list: queue head 539 * @skb: buffer 540 * 541 * Returns true if @skb is the last buffer on the list. 542 */ 543static inline bool skb_queue_is_last(const struct sk_buff_head *list, 544 const struct sk_buff *skb) 545{ 546 return (skb->next == (struct sk_buff *) list); 547} 548 549/** 550 * skb_queue_is_first - check if skb is the first entry in the queue 551 * @list: queue head 552 * @skb: buffer 553 * 554 * Returns true if @skb is the first buffer on the list. 555 */ 556static inline bool skb_queue_is_first(const struct sk_buff_head *list, 557 const struct sk_buff *skb) 558{ 559 return (skb->prev == (struct sk_buff *) list); 560} 561 562/** 563 * skb_queue_next - return the next packet in the queue 564 * @list: queue head 565 * @skb: current buffer 566 * 567 * Return the next packet in @list after @skb. It is only valid to 568 * call this if skb_queue_is_last() evaluates to false. 569 */ 570static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 571 const struct sk_buff *skb) 572{ 573 /* This BUG_ON may seem severe, but if we just return then we 574 * are going to dereference garbage. 575 */ 576 BUG_ON(skb_queue_is_last(list, skb)); 577 return skb->next; 578} 579 580/** 581 * skb_queue_prev - return the prev packet in the queue 582 * @list: queue head 583 * @skb: current buffer 584 * 585 * Return the prev packet in @list before @skb. It is only valid to 586 * call this if skb_queue_is_first() evaluates to false. 587 */ 588static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 589 const struct sk_buff *skb) 590{ 591 /* This BUG_ON may seem severe, but if we just return then we 592 * are going to dereference garbage. 593 */ 594 BUG_ON(skb_queue_is_first(list, skb)); 595 return skb->prev; 596} 597 598/** 599 * skb_get - reference buffer 600 * @skb: buffer to reference 601 * 602 * Makes another reference to a socket buffer and returns a pointer 603 * to the buffer. 604 */ 605static inline struct sk_buff *skb_get(struct sk_buff *skb) 606{ 607 atomic_inc(&skb->users); 608 return skb; 609} 610 611/* 612 * If users == 1, we are the only owner and are can avoid redundant 613 * atomic change. 614 */ 615 616/** 617 * skb_cloned - is the buffer a clone 618 * @skb: buffer to check 619 * 620 * Returns true if the buffer was generated with skb_clone() and is 621 * one of multiple shared copies of the buffer. Cloned buffers are 622 * shared data so must not be written to under normal circumstances. 623 */ 624static inline int skb_cloned(const struct sk_buff *skb) 625{ 626 return skb->cloned && 627 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 628} 629 630/** 631 * skb_header_cloned - is the header a clone 632 * @skb: buffer to check 633 * 634 * Returns true if modifying the header part of the buffer requires 635 * the data to be copied. 636 */ 637static inline int skb_header_cloned(const struct sk_buff *skb) 638{ 639 int dataref; 640 641 if (!skb->cloned) 642 return 0; 643 644 dataref = atomic_read(&skb_shinfo(skb)->dataref); 645 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 646 return dataref != 1; 647} 648 649/** 650 * skb_header_release - release reference to header 651 * @skb: buffer to operate on 652 * 653 * Drop a reference to the header part of the buffer. This is done 654 * by acquiring a payload reference. You must not read from the header 655 * part of skb->data after this. 656 */ 657static inline void skb_header_release(struct sk_buff *skb) 658{ 659 BUG_ON(skb->nohdr); 660 skb->nohdr = 1; 661 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); 662} 663 664/** 665 * skb_shared - is the buffer shared 666 * @skb: buffer to check 667 * 668 * Returns true if more than one person has a reference to this 669 * buffer. 670 */ 671static inline int skb_shared(const struct sk_buff *skb) 672{ 673 return atomic_read(&skb->users) != 1; 674} 675 676/** 677 * skb_share_check - check if buffer is shared and if so clone it 678 * @skb: buffer to check 679 * @pri: priority for memory allocation 680 * 681 * If the buffer is shared the buffer is cloned and the old copy 682 * drops a reference. A new clone with a single reference is returned. 683 * If the buffer is not shared the original buffer is returned. When 684 * being called from interrupt status or with spinlocks held pri must 685 * be GFP_ATOMIC. 686 * 687 * NULL is returned on a memory allocation failure. 688 */ 689static inline struct sk_buff *skb_share_check(struct sk_buff *skb, 690 gfp_t pri) 691{ 692 might_sleep_if(pri & __GFP_WAIT); 693 if (skb_shared(skb)) { 694 struct sk_buff *nskb = skb_clone(skb, pri); 695 kfree_skb(skb); 696 skb = nskb; 697 } 698 return skb; 699} 700 701/* 702 * Copy shared buffers into a new sk_buff. We effectively do COW on 703 * packets to handle cases where we have a local reader and forward 704 * and a couple of other messy ones. The normal one is tcpdumping 705 * a packet thats being forwarded. 706 */ 707 708/** 709 * skb_unshare - make a copy of a shared buffer 710 * @skb: buffer to check 711 * @pri: priority for memory allocation 712 * 713 * If the socket buffer is a clone then this function creates a new 714 * copy of the data, drops a reference count on the old copy and returns 715 * the new copy with the reference count at 1. If the buffer is not a clone 716 * the original buffer is returned. When called with a spinlock held or 717 * from interrupt state @pri must be %GFP_ATOMIC 718 * 719 * %NULL is returned on a memory allocation failure. 720 */ 721static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 722 gfp_t pri) 723{ 724 might_sleep_if(pri & __GFP_WAIT); 725 if (skb_cloned(skb)) { 726 struct sk_buff *nskb = skb_copy(skb, pri); 727 kfree_skb(skb); /* Free our shared copy */ 728 skb = nskb; 729 } 730 return skb; 731} 732 733/** 734 * skb_peek 735 * @list_: list to peek at 736 * 737 * Peek an &sk_buff. Unlike most other operations you _MUST_ 738 * be careful with this one. A peek leaves the buffer on the 739 * list and someone else may run off with it. You must hold 740 * the appropriate locks or have a private queue to do this. 741 * 742 * Returns %NULL for an empty list or a pointer to the head element. 743 * The reference count is not incremented and the reference is therefore 744 * volatile. Use with caution. 745 */ 746static inline struct sk_buff *skb_peek(struct sk_buff_head *list_) 747{ 748 struct sk_buff *list = ((struct sk_buff *)list_)->next; 749 if (list == (struct sk_buff *)list_) 750 list = NULL; 751 return list; 752} 753 754/** 755 * skb_peek_tail 756 * @list_: list to peek at 757 * 758 * Peek an &sk_buff. Unlike most other operations you _MUST_ 759 * be careful with this one. A peek leaves the buffer on the 760 * list and someone else may run off with it. You must hold 761 * the appropriate locks or have a private queue to do this. 762 * 763 * Returns %NULL for an empty list or a pointer to the tail element. 764 * The reference count is not incremented and the reference is therefore 765 * volatile. Use with caution. 766 */ 767static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_) 768{ 769 struct sk_buff *list = ((struct sk_buff *)list_)->prev; 770 if (list == (struct sk_buff *)list_) 771 list = NULL; 772 return list; 773} 774 775/** 776 * skb_queue_len - get queue length 777 * @list_: list to measure 778 * 779 * Return the length of an &sk_buff queue. 780 */ 781static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 782{ 783 return list_->qlen; 784} 785 786/** 787 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 788 * @list: queue to initialize 789 * 790 * This initializes only the list and queue length aspects of 791 * an sk_buff_head object. This allows to initialize the list 792 * aspects of an sk_buff_head without reinitializing things like 793 * the spinlock. It can also be used for on-stack sk_buff_head 794 * objects where the spinlock is known to not be used. 795 */ 796static inline void __skb_queue_head_init(struct sk_buff_head *list) 797{ 798 list->prev = list->next = (struct sk_buff *)list; 799 list->qlen = 0; 800} 801 802/* 803 * This function creates a split out lock class for each invocation; 804 * this is needed for now since a whole lot of users of the skb-queue 805 * infrastructure in drivers have different locking usage (in hardirq) 806 * than the networking core (in softirq only). In the long run either the 807 * network layer or drivers should need annotation to consolidate the 808 * main types of usage into 3 classes. 809 */ 810static inline void skb_queue_head_init(struct sk_buff_head *list) 811{ 812 spin_lock_init(&list->lock); 813 __skb_queue_head_init(list); 814} 815 816static inline void skb_queue_head_init_class(struct sk_buff_head *list, 817 struct lock_class_key *class) 818{ 819 skb_queue_head_init(list); 820 lockdep_set_class(&list->lock, class); 821} 822 823/* 824 * Insert an sk_buff on a list. 825 * 826 * The "__skb_xxxx()" functions are the non-atomic ones that 827 * can only be called with interrupts disabled. 828 */ 829extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list); 830static inline void __skb_insert(struct sk_buff *newsk, 831 struct sk_buff *prev, struct sk_buff *next, 832 struct sk_buff_head *list) 833{ 834 newsk->next = next; 835 newsk->prev = prev; 836 next->prev = prev->next = newsk; 837 list->qlen++; 838} 839 840static inline void __skb_queue_splice(const struct sk_buff_head *list, 841 struct sk_buff *prev, 842 struct sk_buff *next) 843{ 844 struct sk_buff *first = list->next; 845 struct sk_buff *last = list->prev; 846 847 first->prev = prev; 848 prev->next = first; 849 850 last->next = next; 851 next->prev = last; 852} 853 854/** 855 * skb_queue_splice - join two skb lists, this is designed for stacks 856 * @list: the new list to add 857 * @head: the place to add it in the first list 858 */ 859static inline void skb_queue_splice(const struct sk_buff_head *list, 860 struct sk_buff_head *head) 861{ 862 if (!skb_queue_empty(list)) { 863 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 864 head->qlen += list->qlen; 865 } 866} 867 868/** 869 * skb_queue_splice - join two skb lists and reinitialise the emptied list 870 * @list: the new list to add 871 * @head: the place to add it in the first list 872 * 873 * The list at @list is reinitialised 874 */ 875static inline void skb_queue_splice_init(struct sk_buff_head *list, 876 struct sk_buff_head *head) 877{ 878 if (!skb_queue_empty(list)) { 879 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 880 head->qlen += list->qlen; 881 __skb_queue_head_init(list); 882 } 883} 884 885/** 886 * skb_queue_splice_tail - join two skb lists, each list being a queue 887 * @list: the new list to add 888 * @head: the place to add it in the first list 889 */ 890static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 891 struct sk_buff_head *head) 892{ 893 if (!skb_queue_empty(list)) { 894 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 895 head->qlen += list->qlen; 896 } 897} 898 899/** 900 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list 901 * @list: the new list to add 902 * @head: the place to add it in the first list 903 * 904 * Each of the lists is a queue. 905 * The list at @list is reinitialised 906 */ 907static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 908 struct sk_buff_head *head) 909{ 910 if (!skb_queue_empty(list)) { 911 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 912 head->qlen += list->qlen; 913 __skb_queue_head_init(list); 914 } 915} 916 917/** 918 * __skb_queue_after - queue a buffer at the list head 919 * @list: list to use 920 * @prev: place after this buffer 921 * @newsk: buffer to queue 922 * 923 * Queue a buffer int the middle of a list. This function takes no locks 924 * and you must therefore hold required locks before calling it. 925 * 926 * A buffer cannot be placed on two lists at the same time. 927 */ 928static inline void __skb_queue_after(struct sk_buff_head *list, 929 struct sk_buff *prev, 930 struct sk_buff *newsk) 931{ 932 __skb_insert(newsk, prev, prev->next, list); 933} 934 935extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, 936 struct sk_buff_head *list); 937 938static inline void __skb_queue_before(struct sk_buff_head *list, 939 struct sk_buff *next, 940 struct sk_buff *newsk) 941{ 942 __skb_insert(newsk, next->prev, next, list); 943} 944 945/** 946 * __skb_queue_head - queue a buffer at the list head 947 * @list: list to use 948 * @newsk: buffer to queue 949 * 950 * Queue a buffer at the start of a list. This function takes no locks 951 * and you must therefore hold required locks before calling it. 952 * 953 * A buffer cannot be placed on two lists at the same time. 954 */ 955extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 956static inline void __skb_queue_head(struct sk_buff_head *list, 957 struct sk_buff *newsk) 958{ 959 __skb_queue_after(list, (struct sk_buff *)list, newsk); 960} 961 962/** 963 * __skb_queue_tail - queue a buffer at the list tail 964 * @list: list to use 965 * @newsk: buffer to queue 966 * 967 * Queue a buffer at the end of a list. This function takes no locks 968 * and you must therefore hold required locks before calling it. 969 * 970 * A buffer cannot be placed on two lists at the same time. 971 */ 972extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 973static inline void __skb_queue_tail(struct sk_buff_head *list, 974 struct sk_buff *newsk) 975{ 976 __skb_queue_before(list, (struct sk_buff *)list, newsk); 977} 978 979/* 980 * remove sk_buff from list. _Must_ be called atomically, and with 981 * the list known.. 982 */ 983extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 984static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 985{ 986 struct sk_buff *next, *prev; 987 988 list->qlen--; 989 next = skb->next; 990 prev = skb->prev; 991 skb->next = skb->prev = NULL; 992 next->prev = prev; 993 prev->next = next; 994} 995 996/** 997 * __skb_dequeue - remove from the head of the queue 998 * @list: list to dequeue from 999 * 1000 * Remove the head of the list. This function does not take any locks 1001 * so must be used with appropriate locks held only. The head item is 1002 * returned or %NULL if the list is empty. 1003 */ 1004extern struct sk_buff *skb_dequeue(struct sk_buff_head *list); 1005static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 1006{ 1007 struct sk_buff *skb = skb_peek(list); 1008 if (skb) 1009 __skb_unlink(skb, list); 1010 return skb; 1011} 1012 1013/** 1014 * __skb_dequeue_tail - remove from the tail of the queue 1015 * @list: list to dequeue from 1016 * 1017 * Remove the tail of the list. This function does not take any locks 1018 * so must be used with appropriate locks held only. The tail item is 1019 * returned or %NULL if the list is empty. 1020 */ 1021extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 1022static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 1023{ 1024 struct sk_buff *skb = skb_peek_tail(list); 1025 if (skb) 1026 __skb_unlink(skb, list); 1027 return skb; 1028} 1029 1030 1031static inline int skb_is_nonlinear(const struct sk_buff *skb) 1032{ 1033 return skb->data_len; 1034} 1035 1036static inline unsigned int skb_headlen(const struct sk_buff *skb) 1037{ 1038 return skb->len - skb->data_len; 1039} 1040 1041static inline int skb_pagelen(const struct sk_buff *skb) 1042{ 1043 int i, len = 0; 1044 1045 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) 1046 len += skb_shinfo(skb)->frags[i].size; 1047 return len + skb_headlen(skb); 1048} 1049 1050static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 1051 struct page *page, int off, int size) 1052{ 1053 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 1054 1055 frag->page = page; 1056 frag->page_offset = off; 1057 frag->size = size; 1058 skb_shinfo(skb)->nr_frags = i + 1; 1059} 1060 1061extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, 1062 int off, int size); 1063 1064#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) 1065#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list) 1066#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 1067 1068#ifdef NET_SKBUFF_DATA_USES_OFFSET 1069static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1070{ 1071 return skb->head + skb->tail; 1072} 1073 1074static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1075{ 1076 skb->tail = skb->data - skb->head; 1077} 1078 1079static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1080{ 1081 skb_reset_tail_pointer(skb); 1082 skb->tail += offset; 1083} 1084#else /* NET_SKBUFF_DATA_USES_OFFSET */ 1085static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1086{ 1087 return skb->tail; 1088} 1089 1090static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1091{ 1092 skb->tail = skb->data; 1093} 1094 1095static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1096{ 1097 skb->tail = skb->data + offset; 1098} 1099 1100#endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1101 1102/* 1103 * Add data to an sk_buff 1104 */ 1105extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len); 1106static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) 1107{ 1108 unsigned char *tmp = skb_tail_pointer(skb); 1109 SKB_LINEAR_ASSERT(skb); 1110 skb->tail += len; 1111 skb->len += len; 1112 return tmp; 1113} 1114 1115extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len); 1116static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) 1117{ 1118 skb->data -= len; 1119 skb->len += len; 1120 return skb->data; 1121} 1122 1123extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); 1124static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) 1125{ 1126 skb->len -= len; 1127 BUG_ON(skb->len < skb->data_len); 1128 return skb->data += len; 1129} 1130 1131extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); 1132 1133static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) 1134{ 1135 if (len > skb_headlen(skb) && 1136 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 1137 return NULL; 1138 skb->len -= len; 1139 return skb->data += len; 1140} 1141 1142static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) 1143{ 1144 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 1145} 1146 1147static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) 1148{ 1149 if (likely(len <= skb_headlen(skb))) 1150 return 1; 1151 if (unlikely(len > skb->len)) 1152 return 0; 1153 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 1154} 1155 1156/** 1157 * skb_headroom - bytes at buffer head 1158 * @skb: buffer to check 1159 * 1160 * Return the number of bytes of free space at the head of an &sk_buff. 1161 */ 1162static inline unsigned int skb_headroom(const struct sk_buff *skb) 1163{ 1164 return skb->data - skb->head; 1165} 1166 1167/** 1168 * skb_tailroom - bytes at buffer end 1169 * @skb: buffer to check 1170 * 1171 * Return the number of bytes of free space at the tail of an sk_buff 1172 */ 1173static inline int skb_tailroom(const struct sk_buff *skb) 1174{ 1175 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 1176} 1177 1178/** 1179 * skb_reserve - adjust headroom 1180 * @skb: buffer to alter 1181 * @len: bytes to move 1182 * 1183 * Increase the headroom of an empty &sk_buff by reducing the tail 1184 * room. This is only allowed for an empty buffer. 1185 */ 1186static inline void skb_reserve(struct sk_buff *skb, int len) 1187{ 1188 skb->data += len; 1189 skb->tail += len; 1190} 1191 1192#ifdef NET_SKBUFF_DATA_USES_OFFSET 1193static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1194{ 1195 return skb->head + skb->transport_header; 1196} 1197 1198static inline void skb_reset_transport_header(struct sk_buff *skb) 1199{ 1200 skb->transport_header = skb->data - skb->head; 1201} 1202 1203static inline void skb_set_transport_header(struct sk_buff *skb, 1204 const int offset) 1205{ 1206 skb_reset_transport_header(skb); 1207 skb->transport_header += offset; 1208} 1209 1210static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1211{ 1212 return skb->head + skb->network_header; 1213} 1214 1215static inline void skb_reset_network_header(struct sk_buff *skb) 1216{ 1217 skb->network_header = skb->data - skb->head; 1218} 1219 1220static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1221{ 1222 skb_reset_network_header(skb); 1223 skb->network_header += offset; 1224} 1225 1226static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1227{ 1228 return skb->head + skb->mac_header; 1229} 1230 1231static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1232{ 1233 return skb->mac_header != ~0U; 1234} 1235 1236static inline void skb_reset_mac_header(struct sk_buff *skb) 1237{ 1238 skb->mac_header = skb->data - skb->head; 1239} 1240 1241static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1242{ 1243 skb_reset_mac_header(skb); 1244 skb->mac_header += offset; 1245} 1246 1247#else /* NET_SKBUFF_DATA_USES_OFFSET */ 1248 1249static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1250{ 1251 return skb->transport_header; 1252} 1253 1254static inline void skb_reset_transport_header(struct sk_buff *skb) 1255{ 1256 skb->transport_header = skb->data; 1257} 1258 1259static inline void skb_set_transport_header(struct sk_buff *skb, 1260 const int offset) 1261{ 1262 skb->transport_header = skb->data + offset; 1263} 1264 1265static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1266{ 1267 return skb->network_header; 1268} 1269 1270static inline void skb_reset_network_header(struct sk_buff *skb) 1271{ 1272 skb->network_header = skb->data; 1273} 1274 1275static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1276{ 1277 skb->network_header = skb->data + offset; 1278} 1279 1280static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1281{ 1282 return skb->mac_header; 1283} 1284 1285static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1286{ 1287 return skb->mac_header != NULL; 1288} 1289 1290static inline void skb_reset_mac_header(struct sk_buff *skb) 1291{ 1292 skb->mac_header = skb->data; 1293} 1294 1295static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1296{ 1297 skb->mac_header = skb->data + offset; 1298} 1299#endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1300 1301static inline int skb_transport_offset(const struct sk_buff *skb) 1302{ 1303 return skb_transport_header(skb) - skb->data; 1304} 1305 1306static inline u32 skb_network_header_len(const struct sk_buff *skb) 1307{ 1308 return skb->transport_header - skb->network_header; 1309} 1310 1311static inline int skb_network_offset(const struct sk_buff *skb) 1312{ 1313 return skb_network_header(skb) - skb->data; 1314} 1315 1316/* 1317 * CPUs often take a performance hit when accessing unaligned memory 1318 * locations. The actual performance hit varies, it can be small if the 1319 * hardware handles it or large if we have to take an exception and fix it 1320 * in software. 1321 * 1322 * Since an ethernet header is 14 bytes network drivers often end up with 1323 * the IP header at an unaligned offset. The IP header can be aligned by 1324 * shifting the start of the packet by 2 bytes. Drivers should do this 1325 * with: 1326 * 1327 * skb_reserve(NET_IP_ALIGN); 1328 * 1329 * The downside to this alignment of the IP header is that the DMA is now 1330 * unaligned. On some architectures the cost of an unaligned DMA is high 1331 * and this cost outweighs the gains made by aligning the IP header. 1332 * 1333 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 1334 * to be overridden. 1335 */ 1336#ifndef NET_IP_ALIGN 1337#define NET_IP_ALIGN 2 1338#endif 1339 1340/* 1341 * The networking layer reserves some headroom in skb data (via 1342 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 1343 * the header has to grow. In the default case, if the header has to grow 1344 * 32 bytes or less we avoid the reallocation. 1345 * 1346 * Unfortunately this headroom changes the DMA alignment of the resulting 1347 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 1348 * on some architectures. An architecture can override this value, 1349 * perhaps setting it to a cacheline in size (since that will maintain 1350 * cacheline alignment of the DMA). It must be a power of 2. 1351 * 1352 * Various parts of the networking layer expect at least 32 bytes of 1353 * headroom, you should not reduce this. 1354 */ 1355#ifndef NET_SKB_PAD 1356#define NET_SKB_PAD 32 1357#endif 1358 1359extern int ___pskb_trim(struct sk_buff *skb, unsigned int len); 1360 1361static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 1362{ 1363 if (unlikely(skb->data_len)) { 1364 WARN_ON(1); 1365 return; 1366 } 1367 skb->len = len; 1368 skb_set_tail_pointer(skb, len); 1369} 1370 1371extern void skb_trim(struct sk_buff *skb, unsigned int len); 1372 1373static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 1374{ 1375 if (skb->data_len) 1376 return ___pskb_trim(skb, len); 1377 __skb_trim(skb, len); 1378 return 0; 1379} 1380 1381static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 1382{ 1383 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 1384} 1385 1386/** 1387 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 1388 * @skb: buffer to alter 1389 * @len: new length 1390 * 1391 * This is identical to pskb_trim except that the caller knows that 1392 * the skb is not cloned so we should never get an error due to out- 1393 * of-memory. 1394 */ 1395static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 1396{ 1397 int err = pskb_trim(skb, len); 1398 BUG_ON(err); 1399} 1400 1401/** 1402 * skb_orphan - orphan a buffer 1403 * @skb: buffer to orphan 1404 * 1405 * If a buffer currently has an owner then we call the owner's 1406 * destructor function and make the @skb unowned. The buffer continues 1407 * to exist but is no longer charged to its former owner. 1408 */ 1409static inline void skb_orphan(struct sk_buff *skb) 1410{ 1411 if (skb->destructor) 1412 skb->destructor(skb); 1413 skb->destructor = NULL; 1414 skb->sk = NULL; 1415} 1416 1417/** 1418 * __skb_queue_purge - empty a list 1419 * @list: list to empty 1420 * 1421 * Delete all buffers on an &sk_buff list. Each buffer is removed from 1422 * the list and one reference dropped. This function does not take the 1423 * list lock and the caller must hold the relevant locks to use it. 1424 */ 1425extern void skb_queue_purge(struct sk_buff_head *list); 1426static inline void __skb_queue_purge(struct sk_buff_head *list) 1427{ 1428 struct sk_buff *skb; 1429 while ((skb = __skb_dequeue(list)) != NULL) 1430 kfree_skb(skb); 1431} 1432 1433/** 1434 * __dev_alloc_skb - allocate an skbuff for receiving 1435 * @length: length to allocate 1436 * @gfp_mask: get_free_pages mask, passed to alloc_skb 1437 * 1438 * Allocate a new &sk_buff and assign it a usage count of one. The 1439 * buffer has unspecified headroom built in. Users should allocate 1440 * the headroom they think they need without accounting for the 1441 * built in space. The built in space is used for optimisations. 1442 * 1443 * %NULL is returned if there is no free memory. 1444 */ 1445static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 1446 gfp_t gfp_mask) 1447{ 1448 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask); 1449 if (likely(skb)) 1450 skb_reserve(skb, NET_SKB_PAD); 1451 return skb; 1452} 1453 1454extern struct sk_buff *dev_alloc_skb(unsigned int length); 1455 1456extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev, 1457 unsigned int length, gfp_t gfp_mask); 1458 1459/** 1460 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 1461 * @dev: network device to receive on 1462 * @length: length to allocate 1463 * 1464 * Allocate a new &sk_buff and assign it a usage count of one. The 1465 * buffer has unspecified headroom built in. Users should allocate 1466 * the headroom they think they need without accounting for the 1467 * built in space. The built in space is used for optimisations. 1468 * 1469 * %NULL is returned if there is no free memory. Although this function 1470 * allocates memory it can be called from an interrupt. 1471 */ 1472static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 1473 unsigned int length) 1474{ 1475 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 1476} 1477 1478extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask); 1479 1480/** 1481 * netdev_alloc_page - allocate a page for ps-rx on a specific device 1482 * @dev: network device to receive on 1483 * 1484 * Allocate a new page node local to the specified device. 1485 * 1486 * %NULL is returned if there is no free memory. 1487 */ 1488static inline struct page *netdev_alloc_page(struct net_device *dev) 1489{ 1490 return __netdev_alloc_page(dev, GFP_ATOMIC); 1491} 1492 1493static inline void netdev_free_page(struct net_device *dev, struct page *page) 1494{ 1495 __free_page(page); 1496} 1497 1498/** 1499 * skb_clone_writable - is the header of a clone writable 1500 * @skb: buffer to check 1501 * @len: length up to which to write 1502 * 1503 * Returns true if modifying the header part of the cloned buffer 1504 * does not requires the data to be copied. 1505 */ 1506static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len) 1507{ 1508 return !skb_header_cloned(skb) && 1509 skb_headroom(skb) + len <= skb->hdr_len; 1510} 1511 1512static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 1513 int cloned) 1514{ 1515 int delta = 0; 1516 1517 if (headroom < NET_SKB_PAD) 1518 headroom = NET_SKB_PAD; 1519 if (headroom > skb_headroom(skb)) 1520 delta = headroom - skb_headroom(skb); 1521 1522 if (delta || cloned) 1523 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 1524 GFP_ATOMIC); 1525 return 0; 1526} 1527 1528/** 1529 * skb_cow - copy header of skb when it is required 1530 * @skb: buffer to cow 1531 * @headroom: needed headroom 1532 * 1533 * If the skb passed lacks sufficient headroom or its data part 1534 * is shared, data is reallocated. If reallocation fails, an error 1535 * is returned and original skb is not changed. 1536 * 1537 * The result is skb with writable area skb->head...skb->tail 1538 * and at least @headroom of space at head. 1539 */ 1540static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 1541{ 1542 return __skb_cow(skb, headroom, skb_cloned(skb)); 1543} 1544 1545/** 1546 * skb_cow_head - skb_cow but only making the head writable 1547 * @skb: buffer to cow 1548 * @headroom: needed headroom 1549 * 1550 * This function is identical to skb_cow except that we replace the 1551 * skb_cloned check by skb_header_cloned. It should be used when 1552 * you only need to push on some header and do not need to modify 1553 * the data. 1554 */ 1555static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 1556{ 1557 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 1558} 1559 1560/** 1561 * skb_padto - pad an skbuff up to a minimal size 1562 * @skb: buffer to pad 1563 * @len: minimal length 1564 * 1565 * Pads up a buffer to ensure the trailing bytes exist and are 1566 * blanked. If the buffer already contains sufficient data it 1567 * is untouched. Otherwise it is extended. Returns zero on 1568 * success. The skb is freed on error. 1569 */ 1570 1571static inline int skb_padto(struct sk_buff *skb, unsigned int len) 1572{ 1573 unsigned int size = skb->len; 1574 if (likely(size >= len)) 1575 return 0; 1576 return skb_pad(skb, len - size); 1577} 1578 1579static inline int skb_add_data(struct sk_buff *skb, 1580 char __user *from, int copy) 1581{ 1582 const int off = skb->len; 1583 1584 if (skb->ip_summed == CHECKSUM_NONE) { 1585 int err = 0; 1586 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy), 1587 copy, 0, &err); 1588 if (!err) { 1589 skb->csum = csum_block_add(skb->csum, csum, off); 1590 return 0; 1591 } 1592 } else if (!copy_from_user(skb_put(skb, copy), from, copy)) 1593 return 0; 1594 1595 __skb_trim(skb, off); 1596 return -EFAULT; 1597} 1598 1599static inline int skb_can_coalesce(struct sk_buff *skb, int i, 1600 struct page *page, int off) 1601{ 1602 if (i) { 1603 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; 1604 1605 return page == frag->page && 1606 off == frag->page_offset + frag->size; 1607 } 1608 return 0; 1609} 1610 1611static inline int __skb_linearize(struct sk_buff *skb) 1612{ 1613 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 1614} 1615 1616/** 1617 * skb_linearize - convert paged skb to linear one 1618 * @skb: buffer to linarize 1619 * 1620 * If there is no free memory -ENOMEM is returned, otherwise zero 1621 * is returned and the old skb data released. 1622 */ 1623static inline int skb_linearize(struct sk_buff *skb) 1624{ 1625 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 1626} 1627 1628/** 1629 * skb_linearize_cow - make sure skb is linear and writable 1630 * @skb: buffer to process 1631 * 1632 * If there is no free memory -ENOMEM is returned, otherwise zero 1633 * is returned and the old skb data released. 1634 */ 1635static inline int skb_linearize_cow(struct sk_buff *skb) 1636{ 1637 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 1638 __skb_linearize(skb) : 0; 1639} 1640 1641/** 1642 * skb_postpull_rcsum - update checksum for received skb after pull 1643 * @skb: buffer to update 1644 * @start: start of data before pull 1645 * @len: length of data pulled 1646 * 1647 * After doing a pull on a received packet, you need to call this to 1648 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 1649 * CHECKSUM_NONE so that it can be recomputed from scratch. 1650 */ 1651 1652static inline void skb_postpull_rcsum(struct sk_buff *skb, 1653 const void *start, unsigned int len) 1654{ 1655 if (skb->ip_summed == CHECKSUM_COMPLETE) 1656 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); 1657} 1658 1659unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 1660 1661/** 1662 * pskb_trim_rcsum - trim received skb and update checksum 1663 * @skb: buffer to trim 1664 * @len: new length 1665 * 1666 * This is exactly the same as pskb_trim except that it ensures the 1667 * checksum of received packets are still valid after the operation. 1668 */ 1669 1670static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 1671{ 1672 if (likely(len >= skb->len)) 1673 return 0; 1674 if (skb->ip_summed == CHECKSUM_COMPLETE) 1675 skb->ip_summed = CHECKSUM_NONE; 1676 return __pskb_trim(skb, len); 1677} 1678 1679#define skb_queue_walk(queue, skb) \ 1680 for (skb = (queue)->next; \ 1681 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \ 1682 skb = skb->next) 1683 1684#define skb_queue_walk_safe(queue, skb, tmp) \ 1685 for (skb = (queue)->next, tmp = skb->next; \ 1686 skb != (struct sk_buff *)(queue); \ 1687 skb = tmp, tmp = skb->next) 1688 1689#define skb_queue_walk_from(queue, skb) \ 1690 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \ 1691 skb = skb->next) 1692 1693#define skb_queue_walk_from_safe(queue, skb, tmp) \ 1694 for (tmp = skb->next; \ 1695 skb != (struct sk_buff *)(queue); \ 1696 skb = tmp, tmp = skb->next) 1697 1698#define skb_queue_reverse_walk(queue, skb) \ 1699 for (skb = (queue)->prev; \ 1700 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \ 1701 skb = skb->prev) 1702 1703 1704extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, 1705 int *peeked, int *err); 1706extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, 1707 int noblock, int *err); 1708extern unsigned int datagram_poll(struct file *file, struct socket *sock, 1709 struct poll_table_struct *wait); 1710extern int skb_copy_datagram_iovec(const struct sk_buff *from, 1711 int offset, struct iovec *to, 1712 int size); 1713extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, 1714 int hlen, 1715 struct iovec *iov); 1716extern int skb_copy_datagram_from_iovec(struct sk_buff *skb, 1717 int offset, 1718 struct iovec *from, 1719 int len); 1720extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 1721extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, 1722 unsigned int flags); 1723extern __wsum skb_checksum(const struct sk_buff *skb, int offset, 1724 int len, __wsum csum); 1725extern int skb_copy_bits(const struct sk_buff *skb, int offset, 1726 void *to, int len); 1727extern int skb_store_bits(struct sk_buff *skb, int offset, 1728 const void *from, int len); 1729extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, 1730 int offset, u8 *to, int len, 1731 __wsum csum); 1732extern int skb_splice_bits(struct sk_buff *skb, 1733 unsigned int offset, 1734 struct pipe_inode_info *pipe, 1735 unsigned int len, 1736 unsigned int flags); 1737extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 1738extern void skb_split(struct sk_buff *skb, 1739 struct sk_buff *skb1, const u32 len); 1740extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, 1741 int shiftlen); 1742 1743extern struct sk_buff *skb_segment(struct sk_buff *skb, int features); 1744 1745static inline void *skb_header_pointer(const struct sk_buff *skb, int offset, 1746 int len, void *buffer) 1747{ 1748 int hlen = skb_headlen(skb); 1749 1750 if (hlen - offset >= len) 1751 return skb->data + offset; 1752 1753 if (skb_copy_bits(skb, offset, buffer, len) < 0) 1754 return NULL; 1755 1756 return buffer; 1757} 1758 1759static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 1760 void *to, 1761 const unsigned int len) 1762{ 1763 memcpy(to, skb->data, len); 1764} 1765 1766static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 1767 const int offset, void *to, 1768 const unsigned int len) 1769{ 1770 memcpy(to, skb->data + offset, len); 1771} 1772 1773static inline void skb_copy_to_linear_data(struct sk_buff *skb, 1774 const void *from, 1775 const unsigned int len) 1776{ 1777 memcpy(skb->data, from, len); 1778} 1779 1780static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 1781 const int offset, 1782 const void *from, 1783 const unsigned int len) 1784{ 1785 memcpy(skb->data + offset, from, len); 1786} 1787 1788extern void skb_init(void); 1789 1790static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 1791{ 1792 return skb->tstamp; 1793} 1794 1795/** 1796 * skb_get_timestamp - get timestamp from a skb 1797 * @skb: skb to get stamp from 1798 * @stamp: pointer to struct timeval to store stamp in 1799 * 1800 * Timestamps are stored in the skb as offsets to a base timestamp. 1801 * This function converts the offset back to a struct timeval and stores 1802 * it in stamp. 1803 */ 1804static inline void skb_get_timestamp(const struct sk_buff *skb, 1805 struct timeval *stamp) 1806{ 1807 *stamp = ktime_to_timeval(skb->tstamp); 1808} 1809 1810static inline void skb_get_timestampns(const struct sk_buff *skb, 1811 struct timespec *stamp) 1812{ 1813 *stamp = ktime_to_timespec(skb->tstamp); 1814} 1815 1816static inline void __net_timestamp(struct sk_buff *skb) 1817{ 1818 skb->tstamp = ktime_get_real(); 1819} 1820 1821static inline ktime_t net_timedelta(ktime_t t) 1822{ 1823 return ktime_sub(ktime_get_real(), t); 1824} 1825 1826static inline ktime_t net_invalid_timestamp(void) 1827{ 1828 return ktime_set(0, 0); 1829} 1830 1831/** 1832 * skb_tstamp_tx - queue clone of skb with send time stamps 1833 * @orig_skb: the original outgoing packet 1834 * @hwtstamps: hardware time stamps, may be NULL if not available 1835 * 1836 * If the skb has a socket associated, then this function clones the 1837 * skb (thus sharing the actual data and optional structures), stores 1838 * the optional hardware time stamping information (if non NULL) or 1839 * generates a software time stamp (otherwise), then queues the clone 1840 * to the error queue of the socket. Errors are silently ignored. 1841 */ 1842extern void skb_tstamp_tx(struct sk_buff *orig_skb, 1843 struct skb_shared_hwtstamps *hwtstamps); 1844 1845extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 1846extern __sum16 __skb_checksum_complete(struct sk_buff *skb); 1847 1848static inline int skb_csum_unnecessary(const struct sk_buff *skb) 1849{ 1850 return skb->ip_summed & CHECKSUM_UNNECESSARY; 1851} 1852 1853/** 1854 * skb_checksum_complete - Calculate checksum of an entire packet 1855 * @skb: packet to process 1856 * 1857 * This function calculates the checksum over the entire packet plus 1858 * the value of skb->csum. The latter can be used to supply the 1859 * checksum of a pseudo header as used by TCP/UDP. It returns the 1860 * checksum. 1861 * 1862 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 1863 * this function can be used to verify that checksum on received 1864 * packets. In that case the function should return zero if the 1865 * checksum is correct. In particular, this function will return zero 1866 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 1867 * hardware has already verified the correctness of the checksum. 1868 */ 1869static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 1870{ 1871 return skb_csum_unnecessary(skb) ? 1872 0 : __skb_checksum_complete(skb); 1873} 1874 1875#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1876extern void nf_conntrack_destroy(struct nf_conntrack *nfct); 1877static inline void nf_conntrack_put(struct nf_conntrack *nfct) 1878{ 1879 if (nfct && atomic_dec_and_test(&nfct->use)) 1880 nf_conntrack_destroy(nfct); 1881} 1882static inline void nf_conntrack_get(struct nf_conntrack *nfct) 1883{ 1884 if (nfct) 1885 atomic_inc(&nfct->use); 1886} 1887static inline void nf_conntrack_get_reasm(struct sk_buff *skb) 1888{ 1889 if (skb) 1890 atomic_inc(&skb->users); 1891} 1892static inline void nf_conntrack_put_reasm(struct sk_buff *skb) 1893{ 1894 if (skb) 1895 kfree_skb(skb); 1896} 1897#endif 1898#ifdef CONFIG_BRIDGE_NETFILTER 1899static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) 1900{ 1901 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) 1902 kfree(nf_bridge); 1903} 1904static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) 1905{ 1906 if (nf_bridge) 1907 atomic_inc(&nf_bridge->use); 1908} 1909#endif /* CONFIG_BRIDGE_NETFILTER */ 1910static inline void nf_reset(struct sk_buff *skb) 1911{ 1912#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1913 nf_conntrack_put(skb->nfct); 1914 skb->nfct = NULL; 1915 nf_conntrack_put_reasm(skb->nfct_reasm); 1916 skb->nfct_reasm = NULL; 1917#endif 1918#ifdef CONFIG_BRIDGE_NETFILTER 1919 nf_bridge_put(skb->nf_bridge); 1920 skb->nf_bridge = NULL; 1921#endif 1922} 1923 1924/* Note: This doesn't put any conntrack and bridge info in dst. */ 1925static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src) 1926{ 1927#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1928 dst->nfct = src->nfct; 1929 nf_conntrack_get(src->nfct); 1930 dst->nfctinfo = src->nfctinfo; 1931 dst->nfct_reasm = src->nfct_reasm; 1932 nf_conntrack_get_reasm(src->nfct_reasm); 1933#endif 1934#ifdef CONFIG_BRIDGE_NETFILTER 1935 dst->nf_bridge = src->nf_bridge; 1936 nf_bridge_get(src->nf_bridge); 1937#endif 1938} 1939 1940static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 1941{ 1942#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1943 nf_conntrack_put(dst->nfct); 1944 nf_conntrack_put_reasm(dst->nfct_reasm); 1945#endif 1946#ifdef CONFIG_BRIDGE_NETFILTER 1947 nf_bridge_put(dst->nf_bridge); 1948#endif 1949 __nf_copy(dst, src); 1950} 1951 1952#ifdef CONFIG_NETWORK_SECMARK 1953static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 1954{ 1955 to->secmark = from->secmark; 1956} 1957 1958static inline void skb_init_secmark(struct sk_buff *skb) 1959{ 1960 skb->secmark = 0; 1961} 1962#else 1963static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 1964{ } 1965 1966static inline void skb_init_secmark(struct sk_buff *skb) 1967{ } 1968#endif 1969 1970static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 1971{ 1972 skb->queue_mapping = queue_mapping; 1973} 1974 1975static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 1976{ 1977 return skb->queue_mapping; 1978} 1979 1980static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 1981{ 1982 to->queue_mapping = from->queue_mapping; 1983} 1984 1985static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 1986{ 1987 skb->queue_mapping = rx_queue + 1; 1988} 1989 1990static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 1991{ 1992 return skb->queue_mapping - 1; 1993} 1994 1995static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 1996{ 1997 return (skb->queue_mapping != 0); 1998} 1999 2000extern u16 skb_tx_hash(const struct net_device *dev, 2001 const struct sk_buff *skb); 2002 2003#ifdef CONFIG_XFRM 2004static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 2005{ 2006 return skb->sp; 2007} 2008#else 2009static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 2010{ 2011 return NULL; 2012} 2013#endif 2014 2015static inline int skb_is_gso(const struct sk_buff *skb) 2016{ 2017 return skb_shinfo(skb)->gso_size; 2018} 2019 2020static inline int skb_is_gso_v6(const struct sk_buff *skb) 2021{ 2022 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 2023} 2024 2025extern void __skb_warn_lro_forwarding(const struct sk_buff *skb); 2026 2027static inline bool skb_warn_if_lro(const struct sk_buff *skb) 2028{ 2029 /* LRO sets gso_size but not gso_type, whereas if GSO is really 2030 * wanted then gso_type will be set. */ 2031 struct skb_shared_info *shinfo = skb_shinfo(skb); 2032 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) { 2033 __skb_warn_lro_forwarding(skb); 2034 return true; 2035 } 2036 return false; 2037} 2038 2039static inline void skb_forward_csum(struct sk_buff *skb) 2040{ 2041 /* Unfortunately we don't support this one. Any brave souls? */ 2042 if (skb->ip_summed == CHECKSUM_COMPLETE) 2043 skb->ip_summed = CHECKSUM_NONE; 2044} 2045 2046bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 2047#endif /* __KERNEL__ */ 2048#endif /* _LINUX_SKBUFF_H */