1 /* 2 * Definitions for the 'struct sk_buff' memory handlers. 3 * 4 * Authors: 5 * Alan Cox, <[email protected]> 6 * Florian La Roche, <[email protected]> 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU General Public License 10 * as published by the Free Software Foundation; either version 11 * 2 of the License, or (at your option) any later version. 12 */ 13 14 #ifndef _LINUX_SKBUFF_H 15 #define _LINUX_SKBUFF_H 16 17 #include <linux/kernel.h> 18 #include <linux/kmemcheck.h> 19 #include <linux/compiler.h> 20 #include <linux/time.h> 21 #include <linux/cache.h> 22 23 #include <linux/atomic.h> 24 #include <asm/types.h> 25 #include <linux/spinlock.h> 26 #include <linux/net.h> 27 #include <linux/textsearch.h> 28 #include <net/checksum.h> 29 #include <linux/rcupdate.h> 30 #include <linux/dmaengine.h> 31 #include <linux/hrtimer.h> 32 #include <linux/dma-mapping.h> 33 #include <linux/netdev_features.h> 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 /* return minimum truesize of one skb containing X bytes of data */ 51 #define SKB_TRUESIZE(X) ((X) + \ 52 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 53 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 54 55 /* A. Checksumming of received packets by device. 56 * 57 * NONE: device failed to checksum this packet. 58 * skb->csum is undefined. 59 * 60 * UNNECESSARY: device parsed packet and wouldbe verified checksum. 61 * skb->csum is undefined. 62 * It is bad option, but, unfortunately, many of vendors do this. 63 * Apparently with secret goal to sell you new device, when you 64 * will add new protocol to your host. F.e. IPv6. 8) 65 * 66 * COMPLETE: the most generic way. Device supplied checksum of _all_ 67 * the packet as seen by netif_rx in skb->csum. 68 * NOTE: Even if device supports only some protocols, but 69 * is able to produce some skb->csum, it MUST use COMPLETE, 70 * not UNNECESSARY. 71 * 72 * PARTIAL: identical to the case for output below. This may occur 73 * on a packet received directly from another Linux OS, e.g., 74 * a virtualised Linux kernel on the same host. The packet can 75 * be treated in the same way as UNNECESSARY except that on 76 * output (i.e., forwarding) the checksum must be filled in 77 * by the OS or the hardware. 78 * 79 * B. Checksumming on output. 80 * 81 * NONE: skb is checksummed by protocol or csum is not required. 82 * 83 * PARTIAL: device is required to csum packet as seen by hard_start_xmit 84 * from skb->csum_start to the end and to record the checksum 85 * at skb->csum_start + skb->csum_offset. 86 * 87 * Device must show its capabilities in dev->features, set 88 * at device setup time. 89 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum 90 * everything. 91 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only 92 * TCP/UDP over IPv4. Sigh. Vendors like this 93 * way by an unknown reason. Though, see comment above 94 * about CHECKSUM_UNNECESSARY. 8) 95 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead. 96 * 97 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers 98 * that do not want net to perform the checksum calculation should use 99 * this flag in their outgoing skbs. 100 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC 101 * offload. Correspondingly, the FCoE protocol driver 102 * stack should use CHECKSUM_UNNECESSARY. 103 * 104 * Any questions? No questions, good. --ANK 105 */ 106 107 struct net_device; 108 struct scatterlist; 109 struct pipe_inode_info; 110 111 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 112 struct nf_conntrack { 113 atomic_t use; 114 }; 115 #endif 116 117 #ifdef CONFIG_BRIDGE_NETFILTER 118 struct nf_bridge_info { 119 atomic_t use; 120 struct net_device *physindev; 121 struct net_device *physoutdev; 122 unsigned int mask; 123 unsigned long data[32 / sizeof(unsigned long)]; 124 }; 125 #endif 126 127 struct sk_buff_head { 128 /* These two members must be first. */ 129 struct sk_buff *next; 130 struct sk_buff *prev; 131 132 __u32 qlen; 133 spinlock_t lock; 134 }; 135 136 struct sk_buff; 137 138 /* To allow 64K frame to be packed as single skb without frag_list we 139 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 140 * buffers which do not start on a page boundary. 141 * 142 * Since GRO uses frags we allocate at least 16 regardless of page 143 * size. 144 */ 145 #if (65536/PAGE_SIZE + 1) < 16 146 #define MAX_SKB_FRAGS 16UL 147 #else 148 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 149 #endif 150 151 typedef struct skb_frag_struct skb_frag_t; 152 153 struct skb_frag_struct { 154 struct { 155 struct page *p; 156 } page; 157 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) 158 __u32 page_offset; 159 __u32 size; 160 #else 161 __u16 page_offset; 162 __u16 size; 163 #endif 164 }; 165 166 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 167 { 168 return frag->size; 169 } 170 171 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 172 { 173 frag->size = size; 174 } 175 176 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 177 { 178 frag->size += delta; 179 } 180 181 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 182 { 183 frag->size -= delta; 184 } 185 186 #define HAVE_HW_TIME_STAMP 187 188 /** 189 * struct skb_shared_hwtstamps - hardware time stamps 190 * @hwtstamp: hardware time stamp transformed into duration 191 * since arbitrary point in time 192 * @syststamp: hwtstamp transformed to system time base 193 * 194 * Software time stamps generated by ktime_get_real() are stored in 195 * skb->tstamp. The relation between the different kinds of time 196 * stamps is as follows: 197 * 198 * syststamp and tstamp can be compared against each other in 199 * arbitrary combinations. The accuracy of a 200 * syststamp/tstamp/"syststamp from other device" comparison is 201 * limited by the accuracy of the transformation into system time 202 * base. This depends on the device driver and its underlying 203 * hardware. 204 * 205 * hwtstamps can only be compared against other hwtstamps from 206 * the same device. 207 * 208 * This structure is attached to packets as part of the 209 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 210 */ 211 struct skb_shared_hwtstamps { 212 ktime_t hwtstamp; 213 ktime_t syststamp; 214 }; 215 216 /* Definitions for tx_flags in struct skb_shared_info */ 217 enum { 218 /* generate hardware time stamp */ 219 SKBTX_HW_TSTAMP = 1 << 0, 220 221 /* generate software time stamp */ 222 SKBTX_SW_TSTAMP = 1 << 1, 223 224 /* device driver is going to provide hardware time stamp */ 225 SKBTX_IN_PROGRESS = 1 << 2, 226 227 /* ensure the originating sk reference is available on driver level */ 228 SKBTX_DRV_NEEDS_SK_REF = 1 << 3, 229 230 /* device driver supports TX zero-copy buffers */ 231 SKBTX_DEV_ZEROCOPY = 1 << 4, 232 233 /* generate wifi status information (where possible) */ 234 SKBTX_WIFI_STATUS = 1 << 5, 235 }; 236 237 /* 238 * The callback notifies userspace to release buffers when skb DMA is done in 239 * lower device, the skb last reference should be 0 when calling this. 240 * The desc is used to track userspace buffer index. 241 */ 242 struct ubuf_info { 243 void (*callback)(void *); 244 void *arg; 245 unsigned long desc; 246 }; 247 248 /* This data is invariant across clones and lives at 249 * the end of the header data, ie. at skb->end. 250 */ 251 struct skb_shared_info { 252 unsigned char nr_frags; 253 __u8 tx_flags; 254 unsigned short gso_size; 255 /* Warning: this field is not always filled in (UFO)! */ 256 unsigned short gso_segs; 257 unsigned short gso_type; 258 struct sk_buff *frag_list; 259 struct skb_shared_hwtstamps hwtstamps; 260 __be32 ip6_frag_id; 261 262 /* 263 * Warning : all fields before dataref are cleared in __alloc_skb() 264 */ 265 atomic_t dataref; 266 267 /* Intermediate layers must ensure that destructor_arg 268 * remains valid until skb destructor */ 269 void * destructor_arg; 270 271 /* must be last field, see pskb_expand_head() */ 272 skb_frag_t frags[MAX_SKB_FRAGS]; 273 }; 274 275 /* We divide dataref into two halves. The higher 16 bits hold references 276 * to the payload part of skb->data. The lower 16 bits hold references to 277 * the entire skb->data. A clone of a headerless skb holds the length of 278 * the header in skb->hdr_len. 279 * 280 * All users must obey the rule that the skb->data reference count must be 281 * greater than or equal to the payload reference count. 282 * 283 * Holding a reference to the payload part means that the user does not 284 * care about modifications to the header part of skb->data. 285 */ 286 #define SKB_DATAREF_SHIFT 16 287 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 288 289 290 enum { 291 SKB_FCLONE_UNAVAILABLE, 292 SKB_FCLONE_ORIG, 293 SKB_FCLONE_CLONE, 294 }; 295 296 enum { 297 SKB_GSO_TCPV4 = 1 << 0, 298 SKB_GSO_UDP = 1 << 1, 299 300 /* This indicates the skb is from an untrusted source. */ 301 SKB_GSO_DODGY = 1 << 2, 302 303 /* This indicates the tcp segment has CWR set. */ 304 SKB_GSO_TCP_ECN = 1 << 3, 305 306 SKB_GSO_TCPV6 = 1 << 4, 307 308 SKB_GSO_FCOE = 1 << 5, 309 }; 310 311 #if BITS_PER_LONG > 32 312 #define NET_SKBUFF_DATA_USES_OFFSET 1 313 #endif 314 315 #ifdef NET_SKBUFF_DATA_USES_OFFSET 316 typedef unsigned int sk_buff_data_t; 317 #else 318 typedef unsigned char *sk_buff_data_t; 319 #endif 320 321 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \ 322 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE) 323 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1 324 #endif 325 326 /** 327 * struct sk_buff - socket buffer 328 * @next: Next buffer in list 329 * @prev: Previous buffer in list 330 * @tstamp: Time we arrived 331 * @sk: Socket we are owned by 332 * @dev: Device we arrived on/are leaving by 333 * @cb: Control buffer. Free for use by every layer. Put private vars here 334 * @_skb_refdst: destination entry (with norefcount bit) 335 * @sp: the security path, used for xfrm 336 * @len: Length of actual data 337 * @data_len: Data length 338 * @mac_len: Length of link layer header 339 * @hdr_len: writable header length of cloned skb 340 * @csum: Checksum (must include start/offset pair) 341 * @csum_start: Offset from skb->head where checksumming should start 342 * @csum_offset: Offset from csum_start where checksum should be stored 343 * @priority: Packet queueing priority 344 * @local_df: allow local fragmentation 345 * @cloned: Head may be cloned (check refcnt to be sure) 346 * @ip_summed: Driver fed us an IP checksum 347 * @nohdr: Payload reference only, must not modify header 348 * @nfctinfo: Relationship of this skb to the connection 349 * @pkt_type: Packet class 350 * @fclone: skbuff clone status 351 * @ipvs_property: skbuff is owned by ipvs 352 * @peeked: this packet has been seen already, so stats have been 353 * done for it, don't do them again 354 * @nf_trace: netfilter packet trace flag 355 * @protocol: Packet protocol from driver 356 * @destructor: Destruct function 357 * @nfct: Associated connection, if any 358 * @nfct_reasm: netfilter conntrack re-assembly pointer 359 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 360 * @skb_iif: ifindex of device we arrived on 361 * @tc_index: Traffic control index 362 * @tc_verd: traffic control verdict 363 * @rxhash: the packet hash computed on receive 364 * @queue_mapping: Queue mapping for multiqueue devices 365 * @ndisc_nodetype: router type (from link layer) 366 * @ooo_okay: allow the mapping of a socket to a queue to be changed 367 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport 368 * ports. 369 * @wifi_acked_valid: wifi_acked was set 370 * @wifi_acked: whether frame was acked on wifi or not 371 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 372 * @dma_cookie: a cookie to one of several possible DMA operations 373 * done by skb DMA functions 374 * @secmark: security marking 375 * @mark: Generic packet mark 376 * @dropcount: total number of sk_receive_queue overflows 377 * @vlan_tci: vlan tag control information 378 * @transport_header: Transport layer header 379 * @network_header: Network layer header 380 * @mac_header: Link layer header 381 * @tail: Tail pointer 382 * @end: End pointer 383 * @head: Head of buffer 384 * @data: Data head pointer 385 * @truesize: Buffer size 386 * @users: User count - see {datagram,tcp}.c 387 */ 388 389 struct sk_buff { 390 /* These two members must be first. */ 391 struct sk_buff *next; 392 struct sk_buff *prev; 393 394 ktime_t tstamp; 395 396 struct sock *sk; 397 struct net_device *dev; 398 399 /* 400 * This is the control buffer. It is free to use for every 401 * layer. Please put your private variables there. If you 402 * want to keep them across layers you have to do a skb_clone() 403 * first. This is owned by whoever has the skb queued ATM. 404 */ 405 char cb[48] __aligned(8); 406 407 unsigned long _skb_refdst; 408 #ifdef CONFIG_XFRM 409 struct sec_path *sp; 410 #endif 411 unsigned int len, 412 data_len; 413 __u16 mac_len, 414 hdr_len; 415 union { 416 __wsum csum; 417 struct { 418 __u16 csum_start; 419 __u16 csum_offset; 420 }; 421 }; 422 __u32 priority; 423 kmemcheck_bitfield_begin(flags1); 424 __u8 local_df:1, 425 cloned:1, 426 ip_summed:2, 427 nohdr:1, 428 nfctinfo:3; 429 __u8 pkt_type:3, 430 fclone:2, 431 ipvs_property:1, 432 peeked:1, 433 nf_trace:1; 434 kmemcheck_bitfield_end(flags1); 435 __be16 protocol; 436 437 void (*destructor)(struct sk_buff *skb); 438 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 439 struct nf_conntrack *nfct; 440 #endif 441 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 442 struct sk_buff *nfct_reasm; 443 #endif 444 #ifdef CONFIG_BRIDGE_NETFILTER 445 struct nf_bridge_info *nf_bridge; 446 #endif 447 448 int skb_iif; 449 450 __u32 rxhash; 451 452 __u16 vlan_tci; 453 454 #ifdef CONFIG_NET_SCHED 455 __u16 tc_index; /* traffic control index */ 456 #ifdef CONFIG_NET_CLS_ACT 457 __u16 tc_verd; /* traffic control verdict */ 458 #endif 459 #endif 460 461 __u16 queue_mapping; 462 kmemcheck_bitfield_begin(flags2); 463 #ifdef CONFIG_IPV6_NDISC_NODETYPE 464 __u8 ndisc_nodetype:2; 465 #endif 466 __u8 ooo_okay:1; 467 __u8 l4_rxhash:1; 468 __u8 wifi_acked_valid:1; 469 __u8 wifi_acked:1; 470 __u8 no_fcs:1; 471 /* 9/11 bit hole (depending on ndisc_nodetype presence) */ 472 kmemcheck_bitfield_end(flags2); 473 474 #ifdef CONFIG_NET_DMA 475 dma_cookie_t dma_cookie; 476 #endif 477 #ifdef CONFIG_NETWORK_SECMARK 478 __u32 secmark; 479 #endif 480 union { 481 __u32 mark; 482 __u32 dropcount; 483 }; 484 485 sk_buff_data_t transport_header; 486 sk_buff_data_t network_header; 487 sk_buff_data_t mac_header; 488 /* These elements must be at the end, see alloc_skb() for details. */ 489 sk_buff_data_t tail; 490 sk_buff_data_t end; 491 unsigned char *head, 492 *data; 493 unsigned int truesize; 494 atomic_t users; 495 }; 496 497 #ifdef __KERNEL__ 498 /* 499 * Handling routines are only of interest to the kernel 500 */ 501 #include <linux/slab.h> 502 503 504 /* 505 * skb might have a dst pointer attached, refcounted or not. 506 * _skb_refdst low order bit is set if refcount was _not_ taken 507 */ 508 #define SKB_DST_NOREF 1UL 509 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 510 511 /** 512 * skb_dst - returns skb dst_entry 513 * @skb: buffer 514 * 515 * Returns skb dst_entry, regardless of reference taken or not. 516 */ 517 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 518 { 519 /* If refdst was not refcounted, check we still are in a 520 * rcu_read_lock section 521 */ 522 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 523 !rcu_read_lock_held() && 524 !rcu_read_lock_bh_held()); 525 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 526 } 527 528 /** 529 * skb_dst_set - sets skb dst 530 * @skb: buffer 531 * @dst: dst entry 532 * 533 * Sets skb dst, assuming a reference was taken on dst and should 534 * be released by skb_dst_drop() 535 */ 536 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 537 { 538 skb->_skb_refdst = (unsigned long)dst; 539 } 540 541 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst); 542 543 /** 544 * skb_dst_is_noref - Test if skb dst isn't refcounted 545 * @skb: buffer 546 */ 547 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 548 { 549 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 550 } 551 552 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 553 { 554 return (struct rtable *)skb_dst(skb); 555 } 556 557 extern void kfree_skb(struct sk_buff *skb); 558 extern void consume_skb(struct sk_buff *skb); 559 extern void __kfree_skb(struct sk_buff *skb); 560 extern struct sk_buff *__alloc_skb(unsigned int size, 561 gfp_t priority, int fclone, int node); 562 extern struct sk_buff *build_skb(void *data); 563 static inline struct sk_buff *alloc_skb(unsigned int size, 564 gfp_t priority) 565 { 566 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 567 } 568 569 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 570 gfp_t priority) 571 { 572 return __alloc_skb(size, priority, 1, NUMA_NO_NODE); 573 } 574 575 extern void skb_recycle(struct sk_buff *skb); 576 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size); 577 578 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 579 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 580 extern struct sk_buff *skb_clone(struct sk_buff *skb, 581 gfp_t priority); 582 extern struct sk_buff *skb_copy(const struct sk_buff *skb, 583 gfp_t priority); 584 extern struct sk_buff *__pskb_copy(struct sk_buff *skb, 585 int headroom, gfp_t gfp_mask); 586 587 extern int pskb_expand_head(struct sk_buff *skb, 588 int nhead, int ntail, 589 gfp_t gfp_mask); 590 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 591 unsigned int headroom); 592 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb, 593 int newheadroom, int newtailroom, 594 gfp_t priority); 595 extern int skb_to_sgvec(struct sk_buff *skb, 596 struct scatterlist *sg, int offset, 597 int len); 598 extern int skb_cow_data(struct sk_buff *skb, int tailbits, 599 struct sk_buff **trailer); 600 extern int skb_pad(struct sk_buff *skb, int pad); 601 #define dev_kfree_skb(a) consume_skb(a) 602 603 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, 604 int getfrag(void *from, char *to, int offset, 605 int len,int odd, struct sk_buff *skb), 606 void *from, int length); 607 608 struct skb_seq_state { 609 __u32 lower_offset; 610 __u32 upper_offset; 611 __u32 frag_idx; 612 __u32 stepped_offset; 613 struct sk_buff *root_skb; 614 struct sk_buff *cur_skb; 615 __u8 *frag_data; 616 }; 617 618 extern void skb_prepare_seq_read(struct sk_buff *skb, 619 unsigned int from, unsigned int to, 620 struct skb_seq_state *st); 621 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 622 struct skb_seq_state *st); 623 extern void skb_abort_seq_read(struct skb_seq_state *st); 624 625 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 626 unsigned int to, struct ts_config *config, 627 struct ts_state *state); 628 629 extern void __skb_get_rxhash(struct sk_buff *skb); 630 static inline __u32 skb_get_rxhash(struct sk_buff *skb) 631 { 632 if (!skb->rxhash) 633 __skb_get_rxhash(skb); 634 635 return skb->rxhash; 636 } 637 638 #ifdef NET_SKBUFF_DATA_USES_OFFSET 639 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 640 { 641 return skb->head + skb->end; 642 } 643 #else 644 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 645 { 646 return skb->end; 647 } 648 #endif 649 650 /* Internal */ 651 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 652 653 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 654 { 655 return &skb_shinfo(skb)->hwtstamps; 656 } 657 658 /** 659 * skb_queue_empty - check if a queue is empty 660 * @list: queue head 661 * 662 * Returns true if the queue is empty, false otherwise. 663 */ 664 static inline int skb_queue_empty(const struct sk_buff_head *list) 665 { 666 return list->next == (struct sk_buff *)list; 667 } 668 669 /** 670 * skb_queue_is_last - check if skb is the last entry in the queue 671 * @list: queue head 672 * @skb: buffer 673 * 674 * Returns true if @skb is the last buffer on the list. 675 */ 676 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 677 const struct sk_buff *skb) 678 { 679 return skb->next == (struct sk_buff *)list; 680 } 681 682 /** 683 * skb_queue_is_first - check if skb is the first entry in the queue 684 * @list: queue head 685 * @skb: buffer 686 * 687 * Returns true if @skb is the first buffer on the list. 688 */ 689 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 690 const struct sk_buff *skb) 691 { 692 return skb->prev == (struct sk_buff *)list; 693 } 694 695 /** 696 * skb_queue_next - return the next packet in the queue 697 * @list: queue head 698 * @skb: current buffer 699 * 700 * Return the next packet in @list after @skb. It is only valid to 701 * call this if skb_queue_is_last() evaluates to false. 702 */ 703 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 704 const struct sk_buff *skb) 705 { 706 /* This BUG_ON may seem severe, but if we just return then we 707 * are going to dereference garbage. 708 */ 709 BUG_ON(skb_queue_is_last(list, skb)); 710 return skb->next; 711 } 712 713 /** 714 * skb_queue_prev - return the prev packet in the queue 715 * @list: queue head 716 * @skb: current buffer 717 * 718 * Return the prev packet in @list before @skb. It is only valid to 719 * call this if skb_queue_is_first() evaluates to false. 720 */ 721 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 722 const struct sk_buff *skb) 723 { 724 /* This BUG_ON may seem severe, but if we just return then we 725 * are going to dereference garbage. 726 */ 727 BUG_ON(skb_queue_is_first(list, skb)); 728 return skb->prev; 729 } 730 731 /** 732 * skb_get - reference buffer 733 * @skb: buffer to reference 734 * 735 * Makes another reference to a socket buffer and returns a pointer 736 * to the buffer. 737 */ 738 static inline struct sk_buff *skb_get(struct sk_buff *skb) 739 { 740 atomic_inc(&skb->users); 741 return skb; 742 } 743 744 /* 745 * If users == 1, we are the only owner and are can avoid redundant 746 * atomic change. 747 */ 748 749 /** 750 * skb_cloned - is the buffer a clone 751 * @skb: buffer to check 752 * 753 * Returns true if the buffer was generated with skb_clone() and is 754 * one of multiple shared copies of the buffer. Cloned buffers are 755 * shared data so must not be written to under normal circumstances. 756 */ 757 static inline int skb_cloned(const struct sk_buff *skb) 758 { 759 return skb->cloned && 760 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 761 } 762 763 /** 764 * skb_header_cloned - is the header a clone 765 * @skb: buffer to check 766 * 767 * Returns true if modifying the header part of the buffer requires 768 * the data to be copied. 769 */ 770 static inline int skb_header_cloned(const struct sk_buff *skb) 771 { 772 int dataref; 773 774 if (!skb->cloned) 775 return 0; 776 777 dataref = atomic_read(&skb_shinfo(skb)->dataref); 778 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 779 return dataref != 1; 780 } 781 782 /** 783 * skb_header_release - release reference to header 784 * @skb: buffer to operate on 785 * 786 * Drop a reference to the header part of the buffer. This is done 787 * by acquiring a payload reference. You must not read from the header 788 * part of skb->data after this. 789 */ 790 static inline void skb_header_release(struct sk_buff *skb) 791 { 792 BUG_ON(skb->nohdr); 793 skb->nohdr = 1; 794 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); 795 } 796 797 /** 798 * skb_shared - is the buffer shared 799 * @skb: buffer to check 800 * 801 * Returns true if more than one person has a reference to this 802 * buffer. 803 */ 804 static inline int skb_shared(const struct sk_buff *skb) 805 { 806 return atomic_read(&skb->users) != 1; 807 } 808 809 /** 810 * skb_share_check - check if buffer is shared and if so clone it 811 * @skb: buffer to check 812 * @pri: priority for memory allocation 813 * 814 * If the buffer is shared the buffer is cloned and the old copy 815 * drops a reference. A new clone with a single reference is returned. 816 * If the buffer is not shared the original buffer is returned. When 817 * being called from interrupt status or with spinlocks held pri must 818 * be GFP_ATOMIC. 819 * 820 * NULL is returned on a memory allocation failure. 821 */ 822 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, 823 gfp_t pri) 824 { 825 might_sleep_if(pri & __GFP_WAIT); 826 if (skb_shared(skb)) { 827 struct sk_buff *nskb = skb_clone(skb, pri); 828 kfree_skb(skb); 829 skb = nskb; 830 } 831 return skb; 832 } 833 834 /* 835 * Copy shared buffers into a new sk_buff. We effectively do COW on 836 * packets to handle cases where we have a local reader and forward 837 * and a couple of other messy ones. The normal one is tcpdumping 838 * a packet thats being forwarded. 839 */ 840 841 /** 842 * skb_unshare - make a copy of a shared buffer 843 * @skb: buffer to check 844 * @pri: priority for memory allocation 845 * 846 * If the socket buffer is a clone then this function creates a new 847 * copy of the data, drops a reference count on the old copy and returns 848 * the new copy with the reference count at 1. If the buffer is not a clone 849 * the original buffer is returned. When called with a spinlock held or 850 * from interrupt state @pri must be %GFP_ATOMIC 851 * 852 * %NULL is returned on a memory allocation failure. 853 */ 854 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 855 gfp_t pri) 856 { 857 might_sleep_if(pri & __GFP_WAIT); 858 if (skb_cloned(skb)) { 859 struct sk_buff *nskb = skb_copy(skb, pri); 860 kfree_skb(skb); /* Free our shared copy */ 861 skb = nskb; 862 } 863 return skb; 864 } 865 866 /** 867 * skb_peek - peek at the head of an &sk_buff_head 868 * @list_: list to peek at 869 * 870 * Peek an &sk_buff. Unlike most other operations you _MUST_ 871 * be careful with this one. A peek leaves the buffer on the 872 * list and someone else may run off with it. You must hold 873 * the appropriate locks or have a private queue to do this. 874 * 875 * Returns %NULL for an empty list or a pointer to the head element. 876 * The reference count is not incremented and the reference is therefore 877 * volatile. Use with caution. 878 */ 879 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 880 { 881 struct sk_buff *list = ((const struct sk_buff *)list_)->next; 882 if (list == (struct sk_buff *)list_) 883 list = NULL; 884 return list; 885 } 886 887 /** 888 * skb_peek_next - peek skb following the given one from a queue 889 * @skb: skb to start from 890 * @list_: list to peek at 891 * 892 * Returns %NULL when the end of the list is met or a pointer to the 893 * next element. The reference count is not incremented and the 894 * reference is therefore volatile. Use with caution. 895 */ 896 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 897 const struct sk_buff_head *list_) 898 { 899 struct sk_buff *next = skb->next; 900 if (next == (struct sk_buff *)list_) 901 next = NULL; 902 return next; 903 } 904 905 /** 906 * skb_peek_tail - peek at the tail of an &sk_buff_head 907 * @list_: list to peek at 908 * 909 * Peek an &sk_buff. Unlike most other operations you _MUST_ 910 * be careful with this one. A peek leaves the buffer on the 911 * list and someone else may run off with it. You must hold 912 * the appropriate locks or have a private queue to do this. 913 * 914 * Returns %NULL for an empty list or a pointer to the tail element. 915 * The reference count is not incremented and the reference is therefore 916 * volatile. Use with caution. 917 */ 918 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 919 { 920 struct sk_buff *list = ((const struct sk_buff *)list_)->prev; 921 if (list == (struct sk_buff *)list_) 922 list = NULL; 923 return list; 924 } 925 926 /** 927 * skb_queue_len - get queue length 928 * @list_: list to measure 929 * 930 * Return the length of an &sk_buff queue. 931 */ 932 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 933 { 934 return list_->qlen; 935 } 936 937 /** 938 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 939 * @list: queue to initialize 940 * 941 * This initializes only the list and queue length aspects of 942 * an sk_buff_head object. This allows to initialize the list 943 * aspects of an sk_buff_head without reinitializing things like 944 * the spinlock. It can also be used for on-stack sk_buff_head 945 * objects where the spinlock is known to not be used. 946 */ 947 static inline void __skb_queue_head_init(struct sk_buff_head *list) 948 { 949 list->prev = list->next = (struct sk_buff *)list; 950 list->qlen = 0; 951 } 952 953 /* 954 * This function creates a split out lock class for each invocation; 955 * this is needed for now since a whole lot of users of the skb-queue 956 * infrastructure in drivers have different locking usage (in hardirq) 957 * than the networking core (in softirq only). In the long run either the 958 * network layer or drivers should need annotation to consolidate the 959 * main types of usage into 3 classes. 960 */ 961 static inline void skb_queue_head_init(struct sk_buff_head *list) 962 { 963 spin_lock_init(&list->lock); 964 __skb_queue_head_init(list); 965 } 966 967 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 968 struct lock_class_key *class) 969 { 970 skb_queue_head_init(list); 971 lockdep_set_class(&list->lock, class); 972 } 973 974 /* 975 * Insert an sk_buff on a list. 976 * 977 * The "__skb_xxxx()" functions are the non-atomic ones that 978 * can only be called with interrupts disabled. 979 */ 980 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list); 981 static inline void __skb_insert(struct sk_buff *newsk, 982 struct sk_buff *prev, struct sk_buff *next, 983 struct sk_buff_head *list) 984 { 985 newsk->next = next; 986 newsk->prev = prev; 987 next->prev = prev->next = newsk; 988 list->qlen++; 989 } 990 991 static inline void __skb_queue_splice(const struct sk_buff_head *list, 992 struct sk_buff *prev, 993 struct sk_buff *next) 994 { 995 struct sk_buff *first = list->next; 996 struct sk_buff *last = list->prev; 997 998 first->prev = prev; 999 prev->next = first; 1000 1001 last->next = next; 1002 next->prev = last; 1003 } 1004 1005 /** 1006 * skb_queue_splice - join two skb lists, this is designed for stacks 1007 * @list: the new list to add 1008 * @head: the place to add it in the first list 1009 */ 1010 static inline void skb_queue_splice(const struct sk_buff_head *list, 1011 struct sk_buff_head *head) 1012 { 1013 if (!skb_queue_empty(list)) { 1014 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1015 head->qlen += list->qlen; 1016 } 1017 } 1018 1019 /** 1020 * skb_queue_splice - join two skb lists and reinitialise the emptied list 1021 * @list: the new list to add 1022 * @head: the place to add it in the first list 1023 * 1024 * The list at @list is reinitialised 1025 */ 1026 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1027 struct sk_buff_head *head) 1028 { 1029 if (!skb_queue_empty(list)) { 1030 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1031 head->qlen += list->qlen; 1032 __skb_queue_head_init(list); 1033 } 1034 } 1035 1036 /** 1037 * skb_queue_splice_tail - join two skb lists, each list being a queue 1038 * @list: the new list to add 1039 * @head: the place to add it in the first list 1040 */ 1041 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1042 struct sk_buff_head *head) 1043 { 1044 if (!skb_queue_empty(list)) { 1045 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1046 head->qlen += list->qlen; 1047 } 1048 } 1049 1050 /** 1051 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list 1052 * @list: the new list to add 1053 * @head: the place to add it in the first list 1054 * 1055 * Each of the lists is a queue. 1056 * The list at @list is reinitialised 1057 */ 1058 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 1059 struct sk_buff_head *head) 1060 { 1061 if (!skb_queue_empty(list)) { 1062 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1063 head->qlen += list->qlen; 1064 __skb_queue_head_init(list); 1065 } 1066 } 1067 1068 /** 1069 * __skb_queue_after - queue a buffer at the list head 1070 * @list: list to use 1071 * @prev: place after this buffer 1072 * @newsk: buffer to queue 1073 * 1074 * Queue a buffer int the middle of a list. This function takes no locks 1075 * and you must therefore hold required locks before calling it. 1076 * 1077 * A buffer cannot be placed on two lists at the same time. 1078 */ 1079 static inline void __skb_queue_after(struct sk_buff_head *list, 1080 struct sk_buff *prev, 1081 struct sk_buff *newsk) 1082 { 1083 __skb_insert(newsk, prev, prev->next, list); 1084 } 1085 1086 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, 1087 struct sk_buff_head *list); 1088 1089 static inline void __skb_queue_before(struct sk_buff_head *list, 1090 struct sk_buff *next, 1091 struct sk_buff *newsk) 1092 { 1093 __skb_insert(newsk, next->prev, next, list); 1094 } 1095 1096 /** 1097 * __skb_queue_head - queue a buffer at the list head 1098 * @list: list to use 1099 * @newsk: buffer to queue 1100 * 1101 * Queue a buffer at the start of a list. This function takes no locks 1102 * and you must therefore hold required locks before calling it. 1103 * 1104 * A buffer cannot be placed on two lists at the same time. 1105 */ 1106 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 1107 static inline void __skb_queue_head(struct sk_buff_head *list, 1108 struct sk_buff *newsk) 1109 { 1110 __skb_queue_after(list, (struct sk_buff *)list, newsk); 1111 } 1112 1113 /** 1114 * __skb_queue_tail - queue a buffer at the list tail 1115 * @list: list to use 1116 * @newsk: buffer to queue 1117 * 1118 * Queue a buffer at the end of a list. This function takes no locks 1119 * and you must therefore hold required locks before calling it. 1120 * 1121 * A buffer cannot be placed on two lists at the same time. 1122 */ 1123 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 1124 static inline void __skb_queue_tail(struct sk_buff_head *list, 1125 struct sk_buff *newsk) 1126 { 1127 __skb_queue_before(list, (struct sk_buff *)list, newsk); 1128 } 1129 1130 /* 1131 * remove sk_buff from list. _Must_ be called atomically, and with 1132 * the list known.. 1133 */ 1134 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 1135 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 1136 { 1137 struct sk_buff *next, *prev; 1138 1139 list->qlen--; 1140 next = skb->next; 1141 prev = skb->prev; 1142 skb->next = skb->prev = NULL; 1143 next->prev = prev; 1144 prev->next = next; 1145 } 1146 1147 /** 1148 * __skb_dequeue - remove from the head of the queue 1149 * @list: list to dequeue from 1150 * 1151 * Remove the head of the list. This function does not take any locks 1152 * so must be used with appropriate locks held only. The head item is 1153 * returned or %NULL if the list is empty. 1154 */ 1155 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list); 1156 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 1157 { 1158 struct sk_buff *skb = skb_peek(list); 1159 if (skb) 1160 __skb_unlink(skb, list); 1161 return skb; 1162 } 1163 1164 /** 1165 * __skb_dequeue_tail - remove from the tail of the queue 1166 * @list: list to dequeue from 1167 * 1168 * Remove the tail of the list. This function does not take any locks 1169 * so must be used with appropriate locks held only. The tail item is 1170 * returned or %NULL if the list is empty. 1171 */ 1172 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 1173 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 1174 { 1175 struct sk_buff *skb = skb_peek_tail(list); 1176 if (skb) 1177 __skb_unlink(skb, list); 1178 return skb; 1179 } 1180 1181 1182 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 1183 { 1184 return skb->data_len; 1185 } 1186 1187 static inline unsigned int skb_headlen(const struct sk_buff *skb) 1188 { 1189 return skb->len - skb->data_len; 1190 } 1191 1192 static inline int skb_pagelen(const struct sk_buff *skb) 1193 { 1194 int i, len = 0; 1195 1196 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) 1197 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 1198 return len + skb_headlen(skb); 1199 } 1200 1201 /** 1202 * __skb_fill_page_desc - initialise a paged fragment in an skb 1203 * @skb: buffer containing fragment to be initialised 1204 * @i: paged fragment index to initialise 1205 * @page: the page to use for this fragment 1206 * @off: the offset to the data with @page 1207 * @size: the length of the data 1208 * 1209 * Initialises the @i'th fragment of @skb to point to &size bytes at 1210 * offset @off within @page. 1211 * 1212 * Does not take any additional reference on the fragment. 1213 */ 1214 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 1215 struct page *page, int off, int size) 1216 { 1217 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 1218 1219 frag->page.p = page; 1220 frag->page_offset = off; 1221 skb_frag_size_set(frag, size); 1222 } 1223 1224 /** 1225 * skb_fill_page_desc - initialise a paged fragment in an skb 1226 * @skb: buffer containing fragment to be initialised 1227 * @i: paged fragment index to initialise 1228 * @page: the page to use for this fragment 1229 * @off: the offset to the data with @page 1230 * @size: the length of the data 1231 * 1232 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 1233 * @skb to point to &size bytes at offset @off within @page. In 1234 * addition updates @skb such that @i is the last fragment. 1235 * 1236 * Does not take any additional reference on the fragment. 1237 */ 1238 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 1239 struct page *page, int off, int size) 1240 { 1241 __skb_fill_page_desc(skb, i, page, off, size); 1242 skb_shinfo(skb)->nr_frags = i + 1; 1243 } 1244 1245 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, 1246 int off, int size); 1247 1248 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) 1249 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb)) 1250 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 1251 1252 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1253 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1254 { 1255 return skb->head + skb->tail; 1256 } 1257 1258 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1259 { 1260 skb->tail = skb->data - skb->head; 1261 } 1262 1263 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1264 { 1265 skb_reset_tail_pointer(skb); 1266 skb->tail += offset; 1267 } 1268 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 1269 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1270 { 1271 return skb->tail; 1272 } 1273 1274 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1275 { 1276 skb->tail = skb->data; 1277 } 1278 1279 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1280 { 1281 skb->tail = skb->data + offset; 1282 } 1283 1284 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1285 1286 /* 1287 * Add data to an sk_buff 1288 */ 1289 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len); 1290 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) 1291 { 1292 unsigned char *tmp = skb_tail_pointer(skb); 1293 SKB_LINEAR_ASSERT(skb); 1294 skb->tail += len; 1295 skb->len += len; 1296 return tmp; 1297 } 1298 1299 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len); 1300 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) 1301 { 1302 skb->data -= len; 1303 skb->len += len; 1304 return skb->data; 1305 } 1306 1307 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); 1308 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) 1309 { 1310 skb->len -= len; 1311 BUG_ON(skb->len < skb->data_len); 1312 return skb->data += len; 1313 } 1314 1315 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len) 1316 { 1317 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 1318 } 1319 1320 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); 1321 1322 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) 1323 { 1324 if (len > skb_headlen(skb) && 1325 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 1326 return NULL; 1327 skb->len -= len; 1328 return skb->data += len; 1329 } 1330 1331 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) 1332 { 1333 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 1334 } 1335 1336 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) 1337 { 1338 if (likely(len <= skb_headlen(skb))) 1339 return 1; 1340 if (unlikely(len > skb->len)) 1341 return 0; 1342 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 1343 } 1344 1345 /** 1346 * skb_headroom - bytes at buffer head 1347 * @skb: buffer to check 1348 * 1349 * Return the number of bytes of free space at the head of an &sk_buff. 1350 */ 1351 static inline unsigned int skb_headroom(const struct sk_buff *skb) 1352 { 1353 return skb->data - skb->head; 1354 } 1355 1356 /** 1357 * skb_tailroom - bytes at buffer end 1358 * @skb: buffer to check 1359 * 1360 * Return the number of bytes of free space at the tail of an sk_buff 1361 */ 1362 static inline int skb_tailroom(const struct sk_buff *skb) 1363 { 1364 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 1365 } 1366 1367 /** 1368 * skb_reserve - adjust headroom 1369 * @skb: buffer to alter 1370 * @len: bytes to move 1371 * 1372 * Increase the headroom of an empty &sk_buff by reducing the tail 1373 * room. This is only allowed for an empty buffer. 1374 */ 1375 static inline void skb_reserve(struct sk_buff *skb, int len) 1376 { 1377 skb->data += len; 1378 skb->tail += len; 1379 } 1380 1381 static inline void skb_reset_mac_len(struct sk_buff *skb) 1382 { 1383 skb->mac_len = skb->network_header - skb->mac_header; 1384 } 1385 1386 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1387 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1388 { 1389 return skb->head + skb->transport_header; 1390 } 1391 1392 static inline void skb_reset_transport_header(struct sk_buff *skb) 1393 { 1394 skb->transport_header = skb->data - skb->head; 1395 } 1396 1397 static inline void skb_set_transport_header(struct sk_buff *skb, 1398 const int offset) 1399 { 1400 skb_reset_transport_header(skb); 1401 skb->transport_header += offset; 1402 } 1403 1404 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1405 { 1406 return skb->head + skb->network_header; 1407 } 1408 1409 static inline void skb_reset_network_header(struct sk_buff *skb) 1410 { 1411 skb->network_header = skb->data - skb->head; 1412 } 1413 1414 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1415 { 1416 skb_reset_network_header(skb); 1417 skb->network_header += offset; 1418 } 1419 1420 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1421 { 1422 return skb->head + skb->mac_header; 1423 } 1424 1425 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1426 { 1427 return skb->mac_header != ~0U; 1428 } 1429 1430 static inline void skb_reset_mac_header(struct sk_buff *skb) 1431 { 1432 skb->mac_header = skb->data - skb->head; 1433 } 1434 1435 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1436 { 1437 skb_reset_mac_header(skb); 1438 skb->mac_header += offset; 1439 } 1440 1441 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 1442 1443 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1444 { 1445 return skb->transport_header; 1446 } 1447 1448 static inline void skb_reset_transport_header(struct sk_buff *skb) 1449 { 1450 skb->transport_header = skb->data; 1451 } 1452 1453 static inline void skb_set_transport_header(struct sk_buff *skb, 1454 const int offset) 1455 { 1456 skb->transport_header = skb->data + offset; 1457 } 1458 1459 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1460 { 1461 return skb->network_header; 1462 } 1463 1464 static inline void skb_reset_network_header(struct sk_buff *skb) 1465 { 1466 skb->network_header = skb->data; 1467 } 1468 1469 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1470 { 1471 skb->network_header = skb->data + offset; 1472 } 1473 1474 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1475 { 1476 return skb->mac_header; 1477 } 1478 1479 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1480 { 1481 return skb->mac_header != NULL; 1482 } 1483 1484 static inline void skb_reset_mac_header(struct sk_buff *skb) 1485 { 1486 skb->mac_header = skb->data; 1487 } 1488 1489 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1490 { 1491 skb->mac_header = skb->data + offset; 1492 } 1493 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1494 1495 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 1496 { 1497 if (skb_mac_header_was_set(skb)) { 1498 const unsigned char *old_mac = skb_mac_header(skb); 1499 1500 skb_set_mac_header(skb, -skb->mac_len); 1501 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 1502 } 1503 } 1504 1505 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 1506 { 1507 return skb->csum_start - skb_headroom(skb); 1508 } 1509 1510 static inline int skb_transport_offset(const struct sk_buff *skb) 1511 { 1512 return skb_transport_header(skb) - skb->data; 1513 } 1514 1515 static inline u32 skb_network_header_len(const struct sk_buff *skb) 1516 { 1517 return skb->transport_header - skb->network_header; 1518 } 1519 1520 static inline int skb_network_offset(const struct sk_buff *skb) 1521 { 1522 return skb_network_header(skb) - skb->data; 1523 } 1524 1525 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 1526 { 1527 return pskb_may_pull(skb, skb_network_offset(skb) + len); 1528 } 1529 1530 /* 1531 * CPUs often take a performance hit when accessing unaligned memory 1532 * locations. The actual performance hit varies, it can be small if the 1533 * hardware handles it or large if we have to take an exception and fix it 1534 * in software. 1535 * 1536 * Since an ethernet header is 14 bytes network drivers often end up with 1537 * the IP header at an unaligned offset. The IP header can be aligned by 1538 * shifting the start of the packet by 2 bytes. Drivers should do this 1539 * with: 1540 * 1541 * skb_reserve(skb, NET_IP_ALIGN); 1542 * 1543 * The downside to this alignment of the IP header is that the DMA is now 1544 * unaligned. On some architectures the cost of an unaligned DMA is high 1545 * and this cost outweighs the gains made by aligning the IP header. 1546 * 1547 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 1548 * to be overridden. 1549 */ 1550 #ifndef NET_IP_ALIGN 1551 #define NET_IP_ALIGN 2 1552 #endif 1553 1554 /* 1555 * The networking layer reserves some headroom in skb data (via 1556 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 1557 * the header has to grow. In the default case, if the header has to grow 1558 * 32 bytes or less we avoid the reallocation. 1559 * 1560 * Unfortunately this headroom changes the DMA alignment of the resulting 1561 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 1562 * on some architectures. An architecture can override this value, 1563 * perhaps setting it to a cacheline in size (since that will maintain 1564 * cacheline alignment of the DMA). It must be a power of 2. 1565 * 1566 * Various parts of the networking layer expect at least 32 bytes of 1567 * headroom, you should not reduce this. 1568 * 1569 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 1570 * to reduce average number of cache lines per packet. 1571 * get_rps_cpus() for example only access one 64 bytes aligned block : 1572 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 1573 */ 1574 #ifndef NET_SKB_PAD 1575 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 1576 #endif 1577 1578 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len); 1579 1580 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 1581 { 1582 if (unlikely(skb_is_nonlinear(skb))) { 1583 WARN_ON(1); 1584 return; 1585 } 1586 skb->len = len; 1587 skb_set_tail_pointer(skb, len); 1588 } 1589 1590 extern void skb_trim(struct sk_buff *skb, unsigned int len); 1591 1592 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 1593 { 1594 if (skb->data_len) 1595 return ___pskb_trim(skb, len); 1596 __skb_trim(skb, len); 1597 return 0; 1598 } 1599 1600 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 1601 { 1602 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 1603 } 1604 1605 /** 1606 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 1607 * @skb: buffer to alter 1608 * @len: new length 1609 * 1610 * This is identical to pskb_trim except that the caller knows that 1611 * the skb is not cloned so we should never get an error due to out- 1612 * of-memory. 1613 */ 1614 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 1615 { 1616 int err = pskb_trim(skb, len); 1617 BUG_ON(err); 1618 } 1619 1620 /** 1621 * skb_orphan - orphan a buffer 1622 * @skb: buffer to orphan 1623 * 1624 * If a buffer currently has an owner then we call the owner's 1625 * destructor function and make the @skb unowned. The buffer continues 1626 * to exist but is no longer charged to its former owner. 1627 */ 1628 static inline void skb_orphan(struct sk_buff *skb) 1629 { 1630 if (skb->destructor) 1631 skb->destructor(skb); 1632 skb->destructor = NULL; 1633 skb->sk = NULL; 1634 } 1635 1636 /** 1637 * __skb_queue_purge - empty a list 1638 * @list: list to empty 1639 * 1640 * Delete all buffers on an &sk_buff list. Each buffer is removed from 1641 * the list and one reference dropped. This function does not take the 1642 * list lock and the caller must hold the relevant locks to use it. 1643 */ 1644 extern void skb_queue_purge(struct sk_buff_head *list); 1645 static inline void __skb_queue_purge(struct sk_buff_head *list) 1646 { 1647 struct sk_buff *skb; 1648 while ((skb = __skb_dequeue(list)) != NULL) 1649 kfree_skb(skb); 1650 } 1651 1652 /** 1653 * __dev_alloc_skb - allocate an skbuff for receiving 1654 * @length: length to allocate 1655 * @gfp_mask: get_free_pages mask, passed to alloc_skb 1656 * 1657 * Allocate a new &sk_buff and assign it a usage count of one. The 1658 * buffer has unspecified headroom built in. Users should allocate 1659 * the headroom they think they need without accounting for the 1660 * built in space. The built in space is used for optimisations. 1661 * 1662 * %NULL is returned if there is no free memory. 1663 */ 1664 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 1665 gfp_t gfp_mask) 1666 { 1667 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask); 1668 if (likely(skb)) 1669 skb_reserve(skb, NET_SKB_PAD); 1670 return skb; 1671 } 1672 1673 extern struct sk_buff *dev_alloc_skb(unsigned int length); 1674 1675 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev, 1676 unsigned int length, gfp_t gfp_mask); 1677 1678 /** 1679 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 1680 * @dev: network device to receive on 1681 * @length: length to allocate 1682 * 1683 * Allocate a new &sk_buff and assign it a usage count of one. The 1684 * buffer has unspecified headroom built in. Users should allocate 1685 * the headroom they think they need without accounting for the 1686 * built in space. The built in space is used for optimisations. 1687 * 1688 * %NULL is returned if there is no free memory. Although this function 1689 * allocates memory it can be called from an interrupt. 1690 */ 1691 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 1692 unsigned int length) 1693 { 1694 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 1695 } 1696 1697 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 1698 unsigned int length, gfp_t gfp) 1699 { 1700 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 1701 1702 if (NET_IP_ALIGN && skb) 1703 skb_reserve(skb, NET_IP_ALIGN); 1704 return skb; 1705 } 1706 1707 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 1708 unsigned int length) 1709 { 1710 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 1711 } 1712 1713 /** 1714 * skb_frag_page - retrieve the page refered to by a paged fragment 1715 * @frag: the paged fragment 1716 * 1717 * Returns the &struct page associated with @frag. 1718 */ 1719 static inline struct page *skb_frag_page(const skb_frag_t *frag) 1720 { 1721 return frag->page.p; 1722 } 1723 1724 /** 1725 * __skb_frag_ref - take an addition reference on a paged fragment. 1726 * @frag: the paged fragment 1727 * 1728 * Takes an additional reference on the paged fragment @frag. 1729 */ 1730 static inline void __skb_frag_ref(skb_frag_t *frag) 1731 { 1732 get_page(skb_frag_page(frag)); 1733 } 1734 1735 /** 1736 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 1737 * @skb: the buffer 1738 * @f: the fragment offset. 1739 * 1740 * Takes an additional reference on the @f'th paged fragment of @skb. 1741 */ 1742 static inline void skb_frag_ref(struct sk_buff *skb, int f) 1743 { 1744 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 1745 } 1746 1747 /** 1748 * __skb_frag_unref - release a reference on a paged fragment. 1749 * @frag: the paged fragment 1750 * 1751 * Releases a reference on the paged fragment @frag. 1752 */ 1753 static inline void __skb_frag_unref(skb_frag_t *frag) 1754 { 1755 put_page(skb_frag_page(frag)); 1756 } 1757 1758 /** 1759 * skb_frag_unref - release a reference on a paged fragment of an skb. 1760 * @skb: the buffer 1761 * @f: the fragment offset 1762 * 1763 * Releases a reference on the @f'th paged fragment of @skb. 1764 */ 1765 static inline void skb_frag_unref(struct sk_buff *skb, int f) 1766 { 1767 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 1768 } 1769 1770 /** 1771 * skb_frag_address - gets the address of the data contained in a paged fragment 1772 * @frag: the paged fragment buffer 1773 * 1774 * Returns the address of the data within @frag. The page must already 1775 * be mapped. 1776 */ 1777 static inline void *skb_frag_address(const skb_frag_t *frag) 1778 { 1779 return page_address(skb_frag_page(frag)) + frag->page_offset; 1780 } 1781 1782 /** 1783 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 1784 * @frag: the paged fragment buffer 1785 * 1786 * Returns the address of the data within @frag. Checks that the page 1787 * is mapped and returns %NULL otherwise. 1788 */ 1789 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 1790 { 1791 void *ptr = page_address(skb_frag_page(frag)); 1792 if (unlikely(!ptr)) 1793 return NULL; 1794 1795 return ptr + frag->page_offset; 1796 } 1797 1798 /** 1799 * __skb_frag_set_page - sets the page contained in a paged fragment 1800 * @frag: the paged fragment 1801 * @page: the page to set 1802 * 1803 * Sets the fragment @frag to contain @page. 1804 */ 1805 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 1806 { 1807 frag->page.p = page; 1808 } 1809 1810 /** 1811 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 1812 * @skb: the buffer 1813 * @f: the fragment offset 1814 * @page: the page to set 1815 * 1816 * Sets the @f'th fragment of @skb to contain @page. 1817 */ 1818 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 1819 struct page *page) 1820 { 1821 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 1822 } 1823 1824 /** 1825 * skb_frag_dma_map - maps a paged fragment via the DMA API 1826 * @dev: the device to map the fragment to 1827 * @frag: the paged fragment to map 1828 * @offset: the offset within the fragment (starting at the 1829 * fragment's own offset) 1830 * @size: the number of bytes to map 1831 * @dir: the direction of the mapping (%PCI_DMA_*) 1832 * 1833 * Maps the page associated with @frag to @device. 1834 */ 1835 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 1836 const skb_frag_t *frag, 1837 size_t offset, size_t size, 1838 enum dma_data_direction dir) 1839 { 1840 return dma_map_page(dev, skb_frag_page(frag), 1841 frag->page_offset + offset, size, dir); 1842 } 1843 1844 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 1845 gfp_t gfp_mask) 1846 { 1847 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 1848 } 1849 1850 /** 1851 * skb_clone_writable - is the header of a clone writable 1852 * @skb: buffer to check 1853 * @len: length up to which to write 1854 * 1855 * Returns true if modifying the header part of the cloned buffer 1856 * does not requires the data to be copied. 1857 */ 1858 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 1859 { 1860 return !skb_header_cloned(skb) && 1861 skb_headroom(skb) + len <= skb->hdr_len; 1862 } 1863 1864 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 1865 int cloned) 1866 { 1867 int delta = 0; 1868 1869 if (headroom < NET_SKB_PAD) 1870 headroom = NET_SKB_PAD; 1871 if (headroom > skb_headroom(skb)) 1872 delta = headroom - skb_headroom(skb); 1873 1874 if (delta || cloned) 1875 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 1876 GFP_ATOMIC); 1877 return 0; 1878 } 1879 1880 /** 1881 * skb_cow - copy header of skb when it is required 1882 * @skb: buffer to cow 1883 * @headroom: needed headroom 1884 * 1885 * If the skb passed lacks sufficient headroom or its data part 1886 * is shared, data is reallocated. If reallocation fails, an error 1887 * is returned and original skb is not changed. 1888 * 1889 * The result is skb with writable area skb->head...skb->tail 1890 * and at least @headroom of space at head. 1891 */ 1892 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 1893 { 1894 return __skb_cow(skb, headroom, skb_cloned(skb)); 1895 } 1896 1897 /** 1898 * skb_cow_head - skb_cow but only making the head writable 1899 * @skb: buffer to cow 1900 * @headroom: needed headroom 1901 * 1902 * This function is identical to skb_cow except that we replace the 1903 * skb_cloned check by skb_header_cloned. It should be used when 1904 * you only need to push on some header and do not need to modify 1905 * the data. 1906 */ 1907 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 1908 { 1909 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 1910 } 1911 1912 /** 1913 * skb_padto - pad an skbuff up to a minimal size 1914 * @skb: buffer to pad 1915 * @len: minimal length 1916 * 1917 * Pads up a buffer to ensure the trailing bytes exist and are 1918 * blanked. If the buffer already contains sufficient data it 1919 * is untouched. Otherwise it is extended. Returns zero on 1920 * success. The skb is freed on error. 1921 */ 1922 1923 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 1924 { 1925 unsigned int size = skb->len; 1926 if (likely(size >= len)) 1927 return 0; 1928 return skb_pad(skb, len - size); 1929 } 1930 1931 static inline int skb_add_data(struct sk_buff *skb, 1932 char __user *from, int copy) 1933 { 1934 const int off = skb->len; 1935 1936 if (skb->ip_summed == CHECKSUM_NONE) { 1937 int err = 0; 1938 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy), 1939 copy, 0, &err); 1940 if (!err) { 1941 skb->csum = csum_block_add(skb->csum, csum, off); 1942 return 0; 1943 } 1944 } else if (!copy_from_user(skb_put(skb, copy), from, copy)) 1945 return 0; 1946 1947 __skb_trim(skb, off); 1948 return -EFAULT; 1949 } 1950 1951 static inline int skb_can_coalesce(struct sk_buff *skb, int i, 1952 const struct page *page, int off) 1953 { 1954 if (i) { 1955 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; 1956 1957 return page == skb_frag_page(frag) && 1958 off == frag->page_offset + skb_frag_size(frag); 1959 } 1960 return 0; 1961 } 1962 1963 static inline int __skb_linearize(struct sk_buff *skb) 1964 { 1965 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 1966 } 1967 1968 /** 1969 * skb_linearize - convert paged skb to linear one 1970 * @skb: buffer to linarize 1971 * 1972 * If there is no free memory -ENOMEM is returned, otherwise zero 1973 * is returned and the old skb data released. 1974 */ 1975 static inline int skb_linearize(struct sk_buff *skb) 1976 { 1977 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 1978 } 1979 1980 /** 1981 * skb_linearize_cow - make sure skb is linear and writable 1982 * @skb: buffer to process 1983 * 1984 * If there is no free memory -ENOMEM is returned, otherwise zero 1985 * is returned and the old skb data released. 1986 */ 1987 static inline int skb_linearize_cow(struct sk_buff *skb) 1988 { 1989 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 1990 __skb_linearize(skb) : 0; 1991 } 1992 1993 /** 1994 * skb_postpull_rcsum - update checksum for received skb after pull 1995 * @skb: buffer to update 1996 * @start: start of data before pull 1997 * @len: length of data pulled 1998 * 1999 * After doing a pull on a received packet, you need to call this to 2000 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 2001 * CHECKSUM_NONE so that it can be recomputed from scratch. 2002 */ 2003 2004 static inline void skb_postpull_rcsum(struct sk_buff *skb, 2005 const void *start, unsigned int len) 2006 { 2007 if (skb->ip_summed == CHECKSUM_COMPLETE) 2008 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); 2009 } 2010 2011 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 2012 2013 /** 2014 * pskb_trim_rcsum - trim received skb and update checksum 2015 * @skb: buffer to trim 2016 * @len: new length 2017 * 2018 * This is exactly the same as pskb_trim except that it ensures the 2019 * checksum of received packets are still valid after the operation. 2020 */ 2021 2022 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 2023 { 2024 if (likely(len >= skb->len)) 2025 return 0; 2026 if (skb->ip_summed == CHECKSUM_COMPLETE) 2027 skb->ip_summed = CHECKSUM_NONE; 2028 return __pskb_trim(skb, len); 2029 } 2030 2031 #define skb_queue_walk(queue, skb) \ 2032 for (skb = (queue)->next; \ 2033 skb != (struct sk_buff *)(queue); \ 2034 skb = skb->next) 2035 2036 #define skb_queue_walk_safe(queue, skb, tmp) \ 2037 for (skb = (queue)->next, tmp = skb->next; \ 2038 skb != (struct sk_buff *)(queue); \ 2039 skb = tmp, tmp = skb->next) 2040 2041 #define skb_queue_walk_from(queue, skb) \ 2042 for (; skb != (struct sk_buff *)(queue); \ 2043 skb = skb->next) 2044 2045 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 2046 for (tmp = skb->next; \ 2047 skb != (struct sk_buff *)(queue); \ 2048 skb = tmp, tmp = skb->next) 2049 2050 #define skb_queue_reverse_walk(queue, skb) \ 2051 for (skb = (queue)->prev; \ 2052 skb != (struct sk_buff *)(queue); \ 2053 skb = skb->prev) 2054 2055 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 2056 for (skb = (queue)->prev, tmp = skb->prev; \ 2057 skb != (struct sk_buff *)(queue); \ 2058 skb = tmp, tmp = skb->prev) 2059 2060 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 2061 for (tmp = skb->prev; \ 2062 skb != (struct sk_buff *)(queue); \ 2063 skb = tmp, tmp = skb->prev) 2064 2065 static inline bool skb_has_frag_list(const struct sk_buff *skb) 2066 { 2067 return skb_shinfo(skb)->frag_list != NULL; 2068 } 2069 2070 static inline void skb_frag_list_init(struct sk_buff *skb) 2071 { 2072 skb_shinfo(skb)->frag_list = NULL; 2073 } 2074 2075 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag) 2076 { 2077 frag->next = skb_shinfo(skb)->frag_list; 2078 skb_shinfo(skb)->frag_list = frag; 2079 } 2080 2081 #define skb_walk_frags(skb, iter) \ 2082 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 2083 2084 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, 2085 int *peeked, int *off, int *err); 2086 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, 2087 int noblock, int *err); 2088 extern unsigned int datagram_poll(struct file *file, struct socket *sock, 2089 struct poll_table_struct *wait); 2090 extern int skb_copy_datagram_iovec(const struct sk_buff *from, 2091 int offset, struct iovec *to, 2092 int size); 2093 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, 2094 int hlen, 2095 struct iovec *iov); 2096 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb, 2097 int offset, 2098 const struct iovec *from, 2099 int from_offset, 2100 int len); 2101 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from, 2102 int offset, 2103 const struct iovec *to, 2104 int to_offset, 2105 int size); 2106 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 2107 extern void skb_free_datagram_locked(struct sock *sk, 2108 struct sk_buff *skb); 2109 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, 2110 unsigned int flags); 2111 extern __wsum skb_checksum(const struct sk_buff *skb, int offset, 2112 int len, __wsum csum); 2113 extern int skb_copy_bits(const struct sk_buff *skb, int offset, 2114 void *to, int len); 2115 extern int skb_store_bits(struct sk_buff *skb, int offset, 2116 const void *from, int len); 2117 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, 2118 int offset, u8 *to, int len, 2119 __wsum csum); 2120 extern int skb_splice_bits(struct sk_buff *skb, 2121 unsigned int offset, 2122 struct pipe_inode_info *pipe, 2123 unsigned int len, 2124 unsigned int flags); 2125 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 2126 extern void skb_split(struct sk_buff *skb, 2127 struct sk_buff *skb1, const u32 len); 2128 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, 2129 int shiftlen); 2130 2131 extern struct sk_buff *skb_segment(struct sk_buff *skb, 2132 netdev_features_t features); 2133 2134 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset, 2135 int len, void *buffer) 2136 { 2137 int hlen = skb_headlen(skb); 2138 2139 if (hlen - offset >= len) 2140 return skb->data + offset; 2141 2142 if (skb_copy_bits(skb, offset, buffer, len) < 0) 2143 return NULL; 2144 2145 return buffer; 2146 } 2147 2148 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 2149 void *to, 2150 const unsigned int len) 2151 { 2152 memcpy(to, skb->data, len); 2153 } 2154 2155 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 2156 const int offset, void *to, 2157 const unsigned int len) 2158 { 2159 memcpy(to, skb->data + offset, len); 2160 } 2161 2162 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 2163 const void *from, 2164 const unsigned int len) 2165 { 2166 memcpy(skb->data, from, len); 2167 } 2168 2169 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 2170 const int offset, 2171 const void *from, 2172 const unsigned int len) 2173 { 2174 memcpy(skb->data + offset, from, len); 2175 } 2176 2177 extern void skb_init(void); 2178 2179 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 2180 { 2181 return skb->tstamp; 2182 } 2183 2184 /** 2185 * skb_get_timestamp - get timestamp from a skb 2186 * @skb: skb to get stamp from 2187 * @stamp: pointer to struct timeval to store stamp in 2188 * 2189 * Timestamps are stored in the skb as offsets to a base timestamp. 2190 * This function converts the offset back to a struct timeval and stores 2191 * it in stamp. 2192 */ 2193 static inline void skb_get_timestamp(const struct sk_buff *skb, 2194 struct timeval *stamp) 2195 { 2196 *stamp = ktime_to_timeval(skb->tstamp); 2197 } 2198 2199 static inline void skb_get_timestampns(const struct sk_buff *skb, 2200 struct timespec *stamp) 2201 { 2202 *stamp = ktime_to_timespec(skb->tstamp); 2203 } 2204 2205 static inline void __net_timestamp(struct sk_buff *skb) 2206 { 2207 skb->tstamp = ktime_get_real(); 2208 } 2209 2210 static inline ktime_t net_timedelta(ktime_t t) 2211 { 2212 return ktime_sub(ktime_get_real(), t); 2213 } 2214 2215 static inline ktime_t net_invalid_timestamp(void) 2216 { 2217 return ktime_set(0, 0); 2218 } 2219 2220 extern void skb_timestamping_init(void); 2221 2222 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 2223 2224 extern void skb_clone_tx_timestamp(struct sk_buff *skb); 2225 extern bool skb_defer_rx_timestamp(struct sk_buff *skb); 2226 2227 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 2228 2229 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 2230 { 2231 } 2232 2233 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 2234 { 2235 return false; 2236 } 2237 2238 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 2239 2240 /** 2241 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 2242 * 2243 * PHY drivers may accept clones of transmitted packets for 2244 * timestamping via their phy_driver.txtstamp method. These drivers 2245 * must call this function to return the skb back to the stack, with 2246 * or without a timestamp. 2247 * 2248 * @skb: clone of the the original outgoing packet 2249 * @hwtstamps: hardware time stamps, may be NULL if not available 2250 * 2251 */ 2252 void skb_complete_tx_timestamp(struct sk_buff *skb, 2253 struct skb_shared_hwtstamps *hwtstamps); 2254 2255 /** 2256 * skb_tstamp_tx - queue clone of skb with send time stamps 2257 * @orig_skb: the original outgoing packet 2258 * @hwtstamps: hardware time stamps, may be NULL if not available 2259 * 2260 * If the skb has a socket associated, then this function clones the 2261 * skb (thus sharing the actual data and optional structures), stores 2262 * the optional hardware time stamping information (if non NULL) or 2263 * generates a software time stamp (otherwise), then queues the clone 2264 * to the error queue of the socket. Errors are silently ignored. 2265 */ 2266 extern void skb_tstamp_tx(struct sk_buff *orig_skb, 2267 struct skb_shared_hwtstamps *hwtstamps); 2268 2269 static inline void sw_tx_timestamp(struct sk_buff *skb) 2270 { 2271 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP && 2272 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) 2273 skb_tstamp_tx(skb, NULL); 2274 } 2275 2276 /** 2277 * skb_tx_timestamp() - Driver hook for transmit timestamping 2278 * 2279 * Ethernet MAC Drivers should call this function in their hard_xmit() 2280 * function immediately before giving the sk_buff to the MAC hardware. 2281 * 2282 * @skb: A socket buffer. 2283 */ 2284 static inline void skb_tx_timestamp(struct sk_buff *skb) 2285 { 2286 skb_clone_tx_timestamp(skb); 2287 sw_tx_timestamp(skb); 2288 } 2289 2290 /** 2291 * skb_complete_wifi_ack - deliver skb with wifi status 2292 * 2293 * @skb: the original outgoing packet 2294 * @acked: ack status 2295 * 2296 */ 2297 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 2298 2299 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 2300 extern __sum16 __skb_checksum_complete(struct sk_buff *skb); 2301 2302 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 2303 { 2304 return skb->ip_summed & CHECKSUM_UNNECESSARY; 2305 } 2306 2307 /** 2308 * skb_checksum_complete - Calculate checksum of an entire packet 2309 * @skb: packet to process 2310 * 2311 * This function calculates the checksum over the entire packet plus 2312 * the value of skb->csum. The latter can be used to supply the 2313 * checksum of a pseudo header as used by TCP/UDP. It returns the 2314 * checksum. 2315 * 2316 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 2317 * this function can be used to verify that checksum on received 2318 * packets. In that case the function should return zero if the 2319 * checksum is correct. In particular, this function will return zero 2320 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 2321 * hardware has already verified the correctness of the checksum. 2322 */ 2323 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 2324 { 2325 return skb_csum_unnecessary(skb) ? 2326 0 : __skb_checksum_complete(skb); 2327 } 2328 2329 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2330 extern void nf_conntrack_destroy(struct nf_conntrack *nfct); 2331 static inline void nf_conntrack_put(struct nf_conntrack *nfct) 2332 { 2333 if (nfct && atomic_dec_and_test(&nfct->use)) 2334 nf_conntrack_destroy(nfct); 2335 } 2336 static inline void nf_conntrack_get(struct nf_conntrack *nfct) 2337 { 2338 if (nfct) 2339 atomic_inc(&nfct->use); 2340 } 2341 #endif 2342 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2343 static inline void nf_conntrack_get_reasm(struct sk_buff *skb) 2344 { 2345 if (skb) 2346 atomic_inc(&skb->users); 2347 } 2348 static inline void nf_conntrack_put_reasm(struct sk_buff *skb) 2349 { 2350 if (skb) 2351 kfree_skb(skb); 2352 } 2353 #endif 2354 #ifdef CONFIG_BRIDGE_NETFILTER 2355 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) 2356 { 2357 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) 2358 kfree(nf_bridge); 2359 } 2360 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) 2361 { 2362 if (nf_bridge) 2363 atomic_inc(&nf_bridge->use); 2364 } 2365 #endif /* CONFIG_BRIDGE_NETFILTER */ 2366 static inline void nf_reset(struct sk_buff *skb) 2367 { 2368 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2369 nf_conntrack_put(skb->nfct); 2370 skb->nfct = NULL; 2371 #endif 2372 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2373 nf_conntrack_put_reasm(skb->nfct_reasm); 2374 skb->nfct_reasm = NULL; 2375 #endif 2376 #ifdef CONFIG_BRIDGE_NETFILTER 2377 nf_bridge_put(skb->nf_bridge); 2378 skb->nf_bridge = NULL; 2379 #endif 2380 } 2381 2382 /* Note: This doesn't put any conntrack and bridge info in dst. */ 2383 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src) 2384 { 2385 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2386 dst->nfct = src->nfct; 2387 nf_conntrack_get(src->nfct); 2388 dst->nfctinfo = src->nfctinfo; 2389 #endif 2390 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2391 dst->nfct_reasm = src->nfct_reasm; 2392 nf_conntrack_get_reasm(src->nfct_reasm); 2393 #endif 2394 #ifdef CONFIG_BRIDGE_NETFILTER 2395 dst->nf_bridge = src->nf_bridge; 2396 nf_bridge_get(src->nf_bridge); 2397 #endif 2398 } 2399 2400 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 2401 { 2402 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2403 nf_conntrack_put(dst->nfct); 2404 #endif 2405 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2406 nf_conntrack_put_reasm(dst->nfct_reasm); 2407 #endif 2408 #ifdef CONFIG_BRIDGE_NETFILTER 2409 nf_bridge_put(dst->nf_bridge); 2410 #endif 2411 __nf_copy(dst, src); 2412 } 2413 2414 #ifdef CONFIG_NETWORK_SECMARK 2415 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 2416 { 2417 to->secmark = from->secmark; 2418 } 2419 2420 static inline void skb_init_secmark(struct sk_buff *skb) 2421 { 2422 skb->secmark = 0; 2423 } 2424 #else 2425 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 2426 { } 2427 2428 static inline void skb_init_secmark(struct sk_buff *skb) 2429 { } 2430 #endif 2431 2432 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 2433 { 2434 skb->queue_mapping = queue_mapping; 2435 } 2436 2437 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 2438 { 2439 return skb->queue_mapping; 2440 } 2441 2442 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 2443 { 2444 to->queue_mapping = from->queue_mapping; 2445 } 2446 2447 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 2448 { 2449 skb->queue_mapping = rx_queue + 1; 2450 } 2451 2452 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 2453 { 2454 return skb->queue_mapping - 1; 2455 } 2456 2457 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 2458 { 2459 return skb->queue_mapping != 0; 2460 } 2461 2462 extern u16 __skb_tx_hash(const struct net_device *dev, 2463 const struct sk_buff *skb, 2464 unsigned int num_tx_queues); 2465 2466 #ifdef CONFIG_XFRM 2467 static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 2468 { 2469 return skb->sp; 2470 } 2471 #else 2472 static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 2473 { 2474 return NULL; 2475 } 2476 #endif 2477 2478 static inline bool skb_is_gso(const struct sk_buff *skb) 2479 { 2480 return skb_shinfo(skb)->gso_size; 2481 } 2482 2483 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 2484 { 2485 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 2486 } 2487 2488 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb); 2489 2490 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 2491 { 2492 /* LRO sets gso_size but not gso_type, whereas if GSO is really 2493 * wanted then gso_type will be set. */ 2494 const struct skb_shared_info *shinfo = skb_shinfo(skb); 2495 2496 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 2497 unlikely(shinfo->gso_type == 0)) { 2498 __skb_warn_lro_forwarding(skb); 2499 return true; 2500 } 2501 return false; 2502 } 2503 2504 static inline void skb_forward_csum(struct sk_buff *skb) 2505 { 2506 /* Unfortunately we don't support this one. Any brave souls? */ 2507 if (skb->ip_summed == CHECKSUM_COMPLETE) 2508 skb->ip_summed = CHECKSUM_NONE; 2509 } 2510 2511 /** 2512 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 2513 * @skb: skb to check 2514 * 2515 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 2516 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 2517 * use this helper, to document places where we make this assertion. 2518 */ 2519 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 2520 { 2521 #ifdef DEBUG 2522 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 2523 #endif 2524 } 2525 2526 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 2527 2528 static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size) 2529 { 2530 if (irqs_disabled()) 2531 return false; 2532 2533 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) 2534 return false; 2535 2536 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE) 2537 return false; 2538 2539 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD); 2540 if (skb_end_pointer(skb) - skb->head < skb_size) 2541 return false; 2542 2543 if (skb_shared(skb) || skb_cloned(skb)) 2544 return false; 2545 2546 return true; 2547 } 2548 #endif /* __KERNEL__ */ 2549 #endif /* _LINUX_SKBUFF_H */ 2550