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