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