1 /* 2 * Definitions for the 'struct sk_buff' memory handlers. 3 * 4 * Authors: 5 * Alan Cox, <[email protected]> 6 * Florian La Roche, <[email protected]> 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU General Public License 10 * as published by the Free Software Foundation; either version 11 * 2 of the License, or (at your option) any later version. 12 */ 13 14 #ifndef _LINUX_SKBUFF_H 15 #define _LINUX_SKBUFF_H 16 17 #include <linux/kernel.h> 18 #include <linux/kmemcheck.h> 19 #include <linux/compiler.h> 20 #include <linux/time.h> 21 #include <linux/bug.h> 22 #include <linux/cache.h> 23 24 #include <linux/atomic.h> 25 #include <asm/types.h> 26 #include <linux/spinlock.h> 27 #include <linux/net.h> 28 #include <linux/textsearch.h> 29 #include <net/checksum.h> 30 #include <linux/rcupdate.h> 31 #include <linux/dmaengine.h> 32 #include <linux/hrtimer.h> 33 #include <linux/dma-mapping.h> 34 #include <linux/netdev_features.h> 35 36 /* Don't change this without changing skb_csum_unnecessary! */ 37 #define CHECKSUM_NONE 0 38 #define CHECKSUM_UNNECESSARY 1 39 #define CHECKSUM_COMPLETE 2 40 #define CHECKSUM_PARTIAL 3 41 42 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \ 43 ~(SMP_CACHE_BYTES - 1)) 44 #define SKB_WITH_OVERHEAD(X) \ 45 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 46 #define SKB_MAX_ORDER(X, ORDER) \ 47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) 48 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) 49 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) 50 51 /* return minimum truesize of one skb containing X bytes of data */ 52 #define SKB_TRUESIZE(X) ((X) + \ 53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ 54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 55 56 /* A. Checksumming of received packets by device. 57 * 58 * NONE: device failed to checksum this packet. 59 * skb->csum is undefined. 60 * 61 * UNNECESSARY: device parsed packet and wouldbe verified checksum. 62 * skb->csum is undefined. 63 * It is bad option, but, unfortunately, many of vendors do this. 64 * Apparently with secret goal to sell you new device, when you 65 * will add new protocol to your host. F.e. IPv6. 8) 66 * 67 * COMPLETE: the most generic way. Device supplied checksum of _all_ 68 * the packet as seen by netif_rx in skb->csum. 69 * NOTE: Even if device supports only some protocols, but 70 * is able to produce some skb->csum, it MUST use COMPLETE, 71 * not UNNECESSARY. 72 * 73 * PARTIAL: identical to the case for output below. This may occur 74 * on a packet received directly from another Linux OS, e.g., 75 * a virtualised Linux kernel on the same host. The packet can 76 * be treated in the same way as UNNECESSARY except that on 77 * output (i.e., forwarding) the checksum must be filled in 78 * by the OS or the hardware. 79 * 80 * B. Checksumming on output. 81 * 82 * NONE: skb is checksummed by protocol or csum is not required. 83 * 84 * PARTIAL: device is required to csum packet as seen by hard_start_xmit 85 * from skb->csum_start to the end and to record the checksum 86 * at skb->csum_start + skb->csum_offset. 87 * 88 * Device must show its capabilities in dev->features, set 89 * at device setup time. 90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum 91 * everything. 92 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only 93 * TCP/UDP over IPv4. Sigh. Vendors like this 94 * way by an unknown reason. Though, see comment above 95 * about CHECKSUM_UNNECESSARY. 8) 96 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead. 97 * 98 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers 99 * that do not want net to perform the checksum calculation should use 100 * this flag in their outgoing skbs. 101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC 102 * offload. Correspondingly, the FCoE protocol driver 103 * stack should use CHECKSUM_UNNECESSARY. 104 * 105 * Any questions? No questions, good. --ANK 106 */ 107 108 struct net_device; 109 struct scatterlist; 110 struct pipe_inode_info; 111 112 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 113 struct nf_conntrack { 114 atomic_t use; 115 }; 116 #endif 117 118 #ifdef CONFIG_BRIDGE_NETFILTER 119 struct nf_bridge_info { 120 atomic_t use; 121 unsigned int mask; 122 struct net_device *physindev; 123 struct net_device *physoutdev; 124 unsigned long data[32 / sizeof(unsigned long)]; 125 }; 126 #endif 127 128 struct sk_buff_head { 129 /* These two members must be first. */ 130 struct sk_buff *next; 131 struct sk_buff *prev; 132 133 __u32 qlen; 134 spinlock_t lock; 135 }; 136 137 struct sk_buff; 138 139 /* To allow 64K frame to be packed as single skb without frag_list we 140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for 141 * buffers which do not start on a page boundary. 142 * 143 * Since GRO uses frags we allocate at least 16 regardless of page 144 * size. 145 */ 146 #if (65536/PAGE_SIZE + 1) < 16 147 #define MAX_SKB_FRAGS 16UL 148 #else 149 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) 150 #endif 151 152 typedef struct skb_frag_struct skb_frag_t; 153 154 struct skb_frag_struct { 155 struct { 156 struct page *p; 157 } page; 158 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) 159 __u32 page_offset; 160 __u32 size; 161 #else 162 __u16 page_offset; 163 __u16 size; 164 #endif 165 }; 166 167 static inline unsigned int skb_frag_size(const skb_frag_t *frag) 168 { 169 return frag->size; 170 } 171 172 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) 173 { 174 frag->size = size; 175 } 176 177 static inline void skb_frag_size_add(skb_frag_t *frag, int delta) 178 { 179 frag->size += delta; 180 } 181 182 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) 183 { 184 frag->size -= delta; 185 } 186 187 #define HAVE_HW_TIME_STAMP 188 189 /** 190 * struct skb_shared_hwtstamps - hardware time stamps 191 * @hwtstamp: hardware time stamp transformed into duration 192 * since arbitrary point in time 193 * @syststamp: hwtstamp transformed to system time base 194 * 195 * Software time stamps generated by ktime_get_real() are stored in 196 * skb->tstamp. The relation between the different kinds of time 197 * stamps is as follows: 198 * 199 * syststamp and tstamp can be compared against each other in 200 * arbitrary combinations. The accuracy of a 201 * syststamp/tstamp/"syststamp from other device" comparison is 202 * limited by the accuracy of the transformation into system time 203 * base. This depends on the device driver and its underlying 204 * hardware. 205 * 206 * hwtstamps can only be compared against other hwtstamps from 207 * the same device. 208 * 209 * This structure is attached to packets as part of the 210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer. 211 */ 212 struct skb_shared_hwtstamps { 213 ktime_t hwtstamp; 214 ktime_t syststamp; 215 }; 216 217 /* Definitions for tx_flags in struct skb_shared_info */ 218 enum { 219 /* generate hardware time stamp */ 220 SKBTX_HW_TSTAMP = 1 << 0, 221 222 /* generate software time stamp */ 223 SKBTX_SW_TSTAMP = 1 << 1, 224 225 /* device driver is going to provide hardware time stamp */ 226 SKBTX_IN_PROGRESS = 1 << 2, 227 228 /* device driver supports TX zero-copy buffers */ 229 SKBTX_DEV_ZEROCOPY = 1 << 3, 230 231 /* generate wifi status information (where possible) */ 232 SKBTX_WIFI_STATUS = 1 << 4, 233 }; 234 235 /* 236 * The callback notifies userspace to release buffers when skb DMA is done in 237 * lower device, the skb last reference should be 0 when calling this. 238 * The ctx field is used to track device context. 239 * The desc field is used to track userspace buffer index. 240 */ 241 struct ubuf_info { 242 void (*callback)(struct ubuf_info *); 243 void *ctx; 244 unsigned long desc; 245 }; 246 247 /* This data is invariant across clones and lives at 248 * the end of the header data, ie. at skb->end. 249 */ 250 struct skb_shared_info { 251 unsigned char nr_frags; 252 __u8 tx_flags; 253 unsigned short gso_size; 254 /* Warning: this field is not always filled in (UFO)! */ 255 unsigned short gso_segs; 256 unsigned short gso_type; 257 struct sk_buff *frag_list; 258 struct skb_shared_hwtstamps hwtstamps; 259 __be32 ip6_frag_id; 260 261 /* 262 * Warning : all fields before dataref are cleared in __alloc_skb() 263 */ 264 atomic_t dataref; 265 266 /* Intermediate layers must ensure that destructor_arg 267 * remains valid until skb destructor */ 268 void * destructor_arg; 269 270 /* must be last field, see pskb_expand_head() */ 271 skb_frag_t frags[MAX_SKB_FRAGS]; 272 }; 273 274 /* We divide dataref into two halves. The higher 16 bits hold references 275 * to the payload part of skb->data. The lower 16 bits hold references to 276 * the entire skb->data. A clone of a headerless skb holds the length of 277 * the header in skb->hdr_len. 278 * 279 * All users must obey the rule that the skb->data reference count must be 280 * greater than or equal to the payload reference count. 281 * 282 * Holding a reference to the payload part means that the user does not 283 * care about modifications to the header part of skb->data. 284 */ 285 #define SKB_DATAREF_SHIFT 16 286 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) 287 288 289 enum { 290 SKB_FCLONE_UNAVAILABLE, 291 SKB_FCLONE_ORIG, 292 SKB_FCLONE_CLONE, 293 }; 294 295 enum { 296 SKB_GSO_TCPV4 = 1 << 0, 297 SKB_GSO_UDP = 1 << 1, 298 299 /* This indicates the skb is from an untrusted source. */ 300 SKB_GSO_DODGY = 1 << 2, 301 302 /* This indicates the tcp segment has CWR set. */ 303 SKB_GSO_TCP_ECN = 1 << 3, 304 305 SKB_GSO_TCPV6 = 1 << 4, 306 307 SKB_GSO_FCOE = 1 << 5, 308 }; 309 310 #if BITS_PER_LONG > 32 311 #define NET_SKBUFF_DATA_USES_OFFSET 1 312 #endif 313 314 #ifdef NET_SKBUFF_DATA_USES_OFFSET 315 typedef unsigned int sk_buff_data_t; 316 #else 317 typedef unsigned char *sk_buff_data_t; 318 #endif 319 320 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \ 321 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE) 322 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1 323 #endif 324 325 /** 326 * struct sk_buff - socket buffer 327 * @next: Next buffer in list 328 * @prev: Previous buffer in list 329 * @tstamp: Time we arrived 330 * @sk: Socket we are owned by 331 * @dev: Device we arrived on/are leaving by 332 * @cb: Control buffer. Free for use by every layer. Put private vars here 333 * @_skb_refdst: destination entry (with norefcount bit) 334 * @sp: the security path, used for xfrm 335 * @len: Length of actual data 336 * @data_len: Data length 337 * @mac_len: Length of link layer header 338 * @hdr_len: writable header length of cloned skb 339 * @csum: Checksum (must include start/offset pair) 340 * @csum_start: Offset from skb->head where checksumming should start 341 * @csum_offset: Offset from csum_start where checksum should be stored 342 * @priority: Packet queueing priority 343 * @local_df: allow local fragmentation 344 * @cloned: Head may be cloned (check refcnt to be sure) 345 * @ip_summed: Driver fed us an IP checksum 346 * @nohdr: Payload reference only, must not modify header 347 * @nfctinfo: Relationship of this skb to the connection 348 * @pkt_type: Packet class 349 * @fclone: skbuff clone status 350 * @ipvs_property: skbuff is owned by ipvs 351 * @peeked: this packet has been seen already, so stats have been 352 * done for it, don't do them again 353 * @nf_trace: netfilter packet trace flag 354 * @protocol: Packet protocol from driver 355 * @destructor: Destruct function 356 * @nfct: Associated connection, if any 357 * @nfct_reasm: netfilter conntrack re-assembly pointer 358 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c 359 * @skb_iif: ifindex of device we arrived on 360 * @tc_index: Traffic control index 361 * @tc_verd: traffic control verdict 362 * @rxhash: the packet hash computed on receive 363 * @queue_mapping: Queue mapping for multiqueue devices 364 * @ndisc_nodetype: router type (from link layer) 365 * @ooo_okay: allow the mapping of a socket to a queue to be changed 366 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport 367 * ports. 368 * @wifi_acked_valid: wifi_acked was set 369 * @wifi_acked: whether frame was acked on wifi or not 370 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS 371 * @dma_cookie: a cookie to one of several possible DMA operations 372 * done by skb DMA functions 373 * @secmark: security marking 374 * @mark: Generic packet mark 375 * @dropcount: total number of sk_receive_queue overflows 376 * @vlan_tci: vlan tag control information 377 * @transport_header: Transport layer header 378 * @network_header: Network layer header 379 * @mac_header: Link layer header 380 * @tail: Tail pointer 381 * @end: End pointer 382 * @head: Head of buffer 383 * @data: Data head pointer 384 * @truesize: Buffer size 385 * @users: User count - see {datagram,tcp}.c 386 */ 387 388 struct sk_buff { 389 /* These two members must be first. */ 390 struct sk_buff *next; 391 struct sk_buff *prev; 392 393 ktime_t tstamp; 394 395 struct sock *sk; 396 struct net_device *dev; 397 398 /* 399 * This is the control buffer. It is free to use for every 400 * layer. Please put your private variables there. If you 401 * want to keep them across layers you have to do a skb_clone() 402 * first. This is owned by whoever has the skb queued ATM. 403 */ 404 char cb[48] __aligned(8); 405 406 unsigned long _skb_refdst; 407 #ifdef CONFIG_XFRM 408 struct sec_path *sp; 409 #endif 410 unsigned int len, 411 data_len; 412 __u16 mac_len, 413 hdr_len; 414 union { 415 __wsum csum; 416 struct { 417 __u16 csum_start; 418 __u16 csum_offset; 419 }; 420 }; 421 __u32 priority; 422 kmemcheck_bitfield_begin(flags1); 423 __u8 local_df:1, 424 cloned:1, 425 ip_summed:2, 426 nohdr:1, 427 nfctinfo:3; 428 __u8 pkt_type:3, 429 fclone:2, 430 ipvs_property:1, 431 peeked:1, 432 nf_trace:1; 433 kmemcheck_bitfield_end(flags1); 434 __be16 protocol; 435 436 void (*destructor)(struct sk_buff *skb); 437 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 438 struct nf_conntrack *nfct; 439 #endif 440 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 441 struct sk_buff *nfct_reasm; 442 #endif 443 #ifdef CONFIG_BRIDGE_NETFILTER 444 struct nf_bridge_info *nf_bridge; 445 #endif 446 447 int skb_iif; 448 449 __u32 rxhash; 450 451 __u16 vlan_tci; 452 453 #ifdef CONFIG_NET_SCHED 454 __u16 tc_index; /* traffic control index */ 455 #ifdef CONFIG_NET_CLS_ACT 456 __u16 tc_verd; /* traffic control verdict */ 457 #endif 458 #endif 459 460 __u16 queue_mapping; 461 kmemcheck_bitfield_begin(flags2); 462 #ifdef CONFIG_IPV6_NDISC_NODETYPE 463 __u8 ndisc_nodetype:2; 464 #endif 465 __u8 pfmemalloc:1; 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 __u8 head_frag:1; 472 /* 8/10 bit hole (depending on ndisc_nodetype presence) */ 473 kmemcheck_bitfield_end(flags2); 474 475 #ifdef CONFIG_NET_DMA 476 dma_cookie_t dma_cookie; 477 #endif 478 #ifdef CONFIG_NETWORK_SECMARK 479 __u32 secmark; 480 #endif 481 union { 482 __u32 mark; 483 __u32 dropcount; 484 __u32 avail_size; 485 }; 486 487 sk_buff_data_t transport_header; 488 sk_buff_data_t network_header; 489 sk_buff_data_t mac_header; 490 /* These elements must be at the end, see alloc_skb() for details. */ 491 sk_buff_data_t tail; 492 sk_buff_data_t end; 493 unsigned char *head, 494 *data; 495 unsigned int truesize; 496 atomic_t users; 497 }; 498 499 #ifdef __KERNEL__ 500 /* 501 * Handling routines are only of interest to the kernel 502 */ 503 #include <linux/slab.h> 504 505 506 #define SKB_ALLOC_FCLONE 0x01 507 #define SKB_ALLOC_RX 0x02 508 509 /* Returns true if the skb was allocated from PFMEMALLOC reserves */ 510 static inline bool skb_pfmemalloc(const struct sk_buff *skb) 511 { 512 return unlikely(skb->pfmemalloc); 513 } 514 515 /* 516 * skb might have a dst pointer attached, refcounted or not. 517 * _skb_refdst low order bit is set if refcount was _not_ taken 518 */ 519 #define SKB_DST_NOREF 1UL 520 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) 521 522 /** 523 * skb_dst - returns skb dst_entry 524 * @skb: buffer 525 * 526 * Returns skb dst_entry, regardless of reference taken or not. 527 */ 528 static inline struct dst_entry *skb_dst(const struct sk_buff *skb) 529 { 530 /* If refdst was not refcounted, check we still are in a 531 * rcu_read_lock section 532 */ 533 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && 534 !rcu_read_lock_held() && 535 !rcu_read_lock_bh_held()); 536 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); 537 } 538 539 /** 540 * skb_dst_set - sets skb dst 541 * @skb: buffer 542 * @dst: dst entry 543 * 544 * Sets skb dst, assuming a reference was taken on dst and should 545 * be released by skb_dst_drop() 546 */ 547 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) 548 { 549 skb->_skb_refdst = (unsigned long)dst; 550 } 551 552 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst); 553 554 /** 555 * skb_dst_is_noref - Test if skb dst isn't refcounted 556 * @skb: buffer 557 */ 558 static inline bool skb_dst_is_noref(const struct sk_buff *skb) 559 { 560 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); 561 } 562 563 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 564 { 565 return (struct rtable *)skb_dst(skb); 566 } 567 568 extern void kfree_skb(struct sk_buff *skb); 569 extern void consume_skb(struct sk_buff *skb); 570 extern void __kfree_skb(struct sk_buff *skb); 571 extern struct kmem_cache *skbuff_head_cache; 572 573 extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); 574 extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, 575 bool *fragstolen, int *delta_truesize); 576 577 extern struct sk_buff *__alloc_skb(unsigned int size, 578 gfp_t priority, int flags, int node); 579 extern struct sk_buff *build_skb(void *data, unsigned int frag_size); 580 static inline struct sk_buff *alloc_skb(unsigned int size, 581 gfp_t priority) 582 { 583 return __alloc_skb(size, priority, 0, NUMA_NO_NODE); 584 } 585 586 static inline struct sk_buff *alloc_skb_fclone(unsigned int size, 587 gfp_t priority) 588 { 589 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); 590 } 591 592 extern void skb_recycle(struct sk_buff *skb); 593 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size); 594 595 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); 596 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); 597 extern struct sk_buff *skb_clone(struct sk_buff *skb, 598 gfp_t priority); 599 extern struct sk_buff *skb_copy(const struct sk_buff *skb, 600 gfp_t priority); 601 extern struct sk_buff *__pskb_copy(struct sk_buff *skb, 602 int headroom, gfp_t gfp_mask); 603 604 extern int pskb_expand_head(struct sk_buff *skb, 605 int nhead, int ntail, 606 gfp_t gfp_mask); 607 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, 608 unsigned int headroom); 609 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb, 610 int newheadroom, int newtailroom, 611 gfp_t priority); 612 extern int skb_to_sgvec(struct sk_buff *skb, 613 struct scatterlist *sg, int offset, 614 int len); 615 extern int skb_cow_data(struct sk_buff *skb, int tailbits, 616 struct sk_buff **trailer); 617 extern int skb_pad(struct sk_buff *skb, int pad); 618 #define dev_kfree_skb(a) consume_skb(a) 619 620 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, 621 int getfrag(void *from, char *to, int offset, 622 int len,int odd, struct sk_buff *skb), 623 void *from, int length); 624 625 struct skb_seq_state { 626 __u32 lower_offset; 627 __u32 upper_offset; 628 __u32 frag_idx; 629 __u32 stepped_offset; 630 struct sk_buff *root_skb; 631 struct sk_buff *cur_skb; 632 __u8 *frag_data; 633 }; 634 635 extern void skb_prepare_seq_read(struct sk_buff *skb, 636 unsigned int from, unsigned int to, 637 struct skb_seq_state *st); 638 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data, 639 struct skb_seq_state *st); 640 extern void skb_abort_seq_read(struct skb_seq_state *st); 641 642 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, 643 unsigned int to, struct ts_config *config, 644 struct ts_state *state); 645 646 extern void __skb_get_rxhash(struct sk_buff *skb); 647 static inline __u32 skb_get_rxhash(struct sk_buff *skb) 648 { 649 if (!skb->rxhash) 650 __skb_get_rxhash(skb); 651 652 return skb->rxhash; 653 } 654 655 #ifdef NET_SKBUFF_DATA_USES_OFFSET 656 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 657 { 658 return skb->head + skb->end; 659 } 660 661 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 662 { 663 return skb->end; 664 } 665 #else 666 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) 667 { 668 return skb->end; 669 } 670 671 static inline unsigned int skb_end_offset(const struct sk_buff *skb) 672 { 673 return skb->end - skb->head; 674 } 675 #endif 676 677 /* Internal */ 678 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) 679 680 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) 681 { 682 return &skb_shinfo(skb)->hwtstamps; 683 } 684 685 /** 686 * skb_queue_empty - check if a queue is empty 687 * @list: queue head 688 * 689 * Returns true if the queue is empty, false otherwise. 690 */ 691 static inline int skb_queue_empty(const struct sk_buff_head *list) 692 { 693 return list->next == (struct sk_buff *)list; 694 } 695 696 /** 697 * skb_queue_is_last - check if skb is the last entry in the queue 698 * @list: queue head 699 * @skb: buffer 700 * 701 * Returns true if @skb is the last buffer on the list. 702 */ 703 static inline bool skb_queue_is_last(const struct sk_buff_head *list, 704 const struct sk_buff *skb) 705 { 706 return skb->next == (struct sk_buff *)list; 707 } 708 709 /** 710 * skb_queue_is_first - check if skb is the first entry in the queue 711 * @list: queue head 712 * @skb: buffer 713 * 714 * Returns true if @skb is the first buffer on the list. 715 */ 716 static inline bool skb_queue_is_first(const struct sk_buff_head *list, 717 const struct sk_buff *skb) 718 { 719 return skb->prev == (struct sk_buff *)list; 720 } 721 722 /** 723 * skb_queue_next - return the next packet in the queue 724 * @list: queue head 725 * @skb: current buffer 726 * 727 * Return the next packet in @list after @skb. It is only valid to 728 * call this if skb_queue_is_last() evaluates to false. 729 */ 730 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, 731 const struct sk_buff *skb) 732 { 733 /* This BUG_ON may seem severe, but if we just return then we 734 * are going to dereference garbage. 735 */ 736 BUG_ON(skb_queue_is_last(list, skb)); 737 return skb->next; 738 } 739 740 /** 741 * skb_queue_prev - return the prev packet in the queue 742 * @list: queue head 743 * @skb: current buffer 744 * 745 * Return the prev packet in @list before @skb. It is only valid to 746 * call this if skb_queue_is_first() evaluates to false. 747 */ 748 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, 749 const struct sk_buff *skb) 750 { 751 /* This BUG_ON may seem severe, but if we just return then we 752 * are going to dereference garbage. 753 */ 754 BUG_ON(skb_queue_is_first(list, skb)); 755 return skb->prev; 756 } 757 758 /** 759 * skb_get - reference buffer 760 * @skb: buffer to reference 761 * 762 * Makes another reference to a socket buffer and returns a pointer 763 * to the buffer. 764 */ 765 static inline struct sk_buff *skb_get(struct sk_buff *skb) 766 { 767 atomic_inc(&skb->users); 768 return skb; 769 } 770 771 /* 772 * If users == 1, we are the only owner and are can avoid redundant 773 * atomic change. 774 */ 775 776 /** 777 * skb_cloned - is the buffer a clone 778 * @skb: buffer to check 779 * 780 * Returns true if the buffer was generated with skb_clone() and is 781 * one of multiple shared copies of the buffer. Cloned buffers are 782 * shared data so must not be written to under normal circumstances. 783 */ 784 static inline int skb_cloned(const struct sk_buff *skb) 785 { 786 return skb->cloned && 787 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; 788 } 789 790 /** 791 * skb_header_cloned - is the header a clone 792 * @skb: buffer to check 793 * 794 * Returns true if modifying the header part of the buffer requires 795 * the data to be copied. 796 */ 797 static inline int skb_header_cloned(const struct sk_buff *skb) 798 { 799 int dataref; 800 801 if (!skb->cloned) 802 return 0; 803 804 dataref = atomic_read(&skb_shinfo(skb)->dataref); 805 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); 806 return dataref != 1; 807 } 808 809 /** 810 * skb_header_release - release reference to header 811 * @skb: buffer to operate on 812 * 813 * Drop a reference to the header part of the buffer. This is done 814 * by acquiring a payload reference. You must not read from the header 815 * part of skb->data after this. 816 */ 817 static inline void skb_header_release(struct sk_buff *skb) 818 { 819 BUG_ON(skb->nohdr); 820 skb->nohdr = 1; 821 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); 822 } 823 824 /** 825 * skb_shared - is the buffer shared 826 * @skb: buffer to check 827 * 828 * Returns true if more than one person has a reference to this 829 * buffer. 830 */ 831 static inline int skb_shared(const struct sk_buff *skb) 832 { 833 return atomic_read(&skb->users) != 1; 834 } 835 836 /** 837 * skb_share_check - check if buffer is shared and if so clone it 838 * @skb: buffer to check 839 * @pri: priority for memory allocation 840 * 841 * If the buffer is shared the buffer is cloned and the old copy 842 * drops a reference. A new clone with a single reference is returned. 843 * If the buffer is not shared the original buffer is returned. When 844 * being called from interrupt status or with spinlocks held pri must 845 * be GFP_ATOMIC. 846 * 847 * NULL is returned on a memory allocation failure. 848 */ 849 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) 850 { 851 might_sleep_if(pri & __GFP_WAIT); 852 if (skb_shared(skb)) { 853 struct sk_buff *nskb = skb_clone(skb, pri); 854 855 if (likely(nskb)) 856 consume_skb(skb); 857 else 858 kfree_skb(skb); 859 skb = nskb; 860 } 861 return skb; 862 } 863 864 /* 865 * Copy shared buffers into a new sk_buff. We effectively do COW on 866 * packets to handle cases where we have a local reader and forward 867 * and a couple of other messy ones. The normal one is tcpdumping 868 * a packet thats being forwarded. 869 */ 870 871 /** 872 * skb_unshare - make a copy of a shared buffer 873 * @skb: buffer to check 874 * @pri: priority for memory allocation 875 * 876 * If the socket buffer is a clone then this function creates a new 877 * copy of the data, drops a reference count on the old copy and returns 878 * the new copy with the reference count at 1. If the buffer is not a clone 879 * the original buffer is returned. When called with a spinlock held or 880 * from interrupt state @pri must be %GFP_ATOMIC 881 * 882 * %NULL is returned on a memory allocation failure. 883 */ 884 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, 885 gfp_t pri) 886 { 887 might_sleep_if(pri & __GFP_WAIT); 888 if (skb_cloned(skb)) { 889 struct sk_buff *nskb = skb_copy(skb, pri); 890 kfree_skb(skb); /* Free our shared copy */ 891 skb = nskb; 892 } 893 return skb; 894 } 895 896 /** 897 * skb_peek - peek at the head of an &sk_buff_head 898 * @list_: list to peek at 899 * 900 * Peek an &sk_buff. Unlike most other operations you _MUST_ 901 * be careful with this one. A peek leaves the buffer on the 902 * list and someone else may run off with it. You must hold 903 * the appropriate locks or have a private queue to do this. 904 * 905 * Returns %NULL for an empty list or a pointer to the head element. 906 * The reference count is not incremented and the reference is therefore 907 * volatile. Use with caution. 908 */ 909 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) 910 { 911 struct sk_buff *skb = list_->next; 912 913 if (skb == (struct sk_buff *)list_) 914 skb = NULL; 915 return skb; 916 } 917 918 /** 919 * skb_peek_next - peek skb following the given one from a queue 920 * @skb: skb to start from 921 * @list_: list to peek at 922 * 923 * Returns %NULL when the end of the list is met or a pointer to the 924 * next element. The reference count is not incremented and the 925 * reference is therefore volatile. Use with caution. 926 */ 927 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, 928 const struct sk_buff_head *list_) 929 { 930 struct sk_buff *next = skb->next; 931 932 if (next == (struct sk_buff *)list_) 933 next = NULL; 934 return next; 935 } 936 937 /** 938 * skb_peek_tail - peek at the tail of an &sk_buff_head 939 * @list_: list to peek at 940 * 941 * Peek an &sk_buff. Unlike most other operations you _MUST_ 942 * be careful with this one. A peek leaves the buffer on the 943 * list and someone else may run off with it. You must hold 944 * the appropriate locks or have a private queue to do this. 945 * 946 * Returns %NULL for an empty list or a pointer to the tail element. 947 * The reference count is not incremented and the reference is therefore 948 * volatile. Use with caution. 949 */ 950 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) 951 { 952 struct sk_buff *skb = list_->prev; 953 954 if (skb == (struct sk_buff *)list_) 955 skb = NULL; 956 return skb; 957 958 } 959 960 /** 961 * skb_queue_len - get queue length 962 * @list_: list to measure 963 * 964 * Return the length of an &sk_buff queue. 965 */ 966 static inline __u32 skb_queue_len(const struct sk_buff_head *list_) 967 { 968 return list_->qlen; 969 } 970 971 /** 972 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head 973 * @list: queue to initialize 974 * 975 * This initializes only the list and queue length aspects of 976 * an sk_buff_head object. This allows to initialize the list 977 * aspects of an sk_buff_head without reinitializing things like 978 * the spinlock. It can also be used for on-stack sk_buff_head 979 * objects where the spinlock is known to not be used. 980 */ 981 static inline void __skb_queue_head_init(struct sk_buff_head *list) 982 { 983 list->prev = list->next = (struct sk_buff *)list; 984 list->qlen = 0; 985 } 986 987 /* 988 * This function creates a split out lock class for each invocation; 989 * this is needed for now since a whole lot of users of the skb-queue 990 * infrastructure in drivers have different locking usage (in hardirq) 991 * than the networking core (in softirq only). In the long run either the 992 * network layer or drivers should need annotation to consolidate the 993 * main types of usage into 3 classes. 994 */ 995 static inline void skb_queue_head_init(struct sk_buff_head *list) 996 { 997 spin_lock_init(&list->lock); 998 __skb_queue_head_init(list); 999 } 1000 1001 static inline void skb_queue_head_init_class(struct sk_buff_head *list, 1002 struct lock_class_key *class) 1003 { 1004 skb_queue_head_init(list); 1005 lockdep_set_class(&list->lock, class); 1006 } 1007 1008 /* 1009 * Insert an sk_buff on a list. 1010 * 1011 * The "__skb_xxxx()" functions are the non-atomic ones that 1012 * can only be called with interrupts disabled. 1013 */ 1014 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list); 1015 static inline void __skb_insert(struct sk_buff *newsk, 1016 struct sk_buff *prev, struct sk_buff *next, 1017 struct sk_buff_head *list) 1018 { 1019 newsk->next = next; 1020 newsk->prev = prev; 1021 next->prev = prev->next = newsk; 1022 list->qlen++; 1023 } 1024 1025 static inline void __skb_queue_splice(const struct sk_buff_head *list, 1026 struct sk_buff *prev, 1027 struct sk_buff *next) 1028 { 1029 struct sk_buff *first = list->next; 1030 struct sk_buff *last = list->prev; 1031 1032 first->prev = prev; 1033 prev->next = first; 1034 1035 last->next = next; 1036 next->prev = last; 1037 } 1038 1039 /** 1040 * skb_queue_splice - join two skb lists, this is designed for stacks 1041 * @list: the new list to add 1042 * @head: the place to add it in the first list 1043 */ 1044 static inline void skb_queue_splice(const struct sk_buff_head *list, 1045 struct sk_buff_head *head) 1046 { 1047 if (!skb_queue_empty(list)) { 1048 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1049 head->qlen += list->qlen; 1050 } 1051 } 1052 1053 /** 1054 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list 1055 * @list: the new list to add 1056 * @head: the place to add it in the first list 1057 * 1058 * The list at @list is reinitialised 1059 */ 1060 static inline void skb_queue_splice_init(struct sk_buff_head *list, 1061 struct sk_buff_head *head) 1062 { 1063 if (!skb_queue_empty(list)) { 1064 __skb_queue_splice(list, (struct sk_buff *) head, head->next); 1065 head->qlen += list->qlen; 1066 __skb_queue_head_init(list); 1067 } 1068 } 1069 1070 /** 1071 * skb_queue_splice_tail - join two skb lists, each list being a queue 1072 * @list: the new list to add 1073 * @head: the place to add it in the first list 1074 */ 1075 static inline void skb_queue_splice_tail(const struct sk_buff_head *list, 1076 struct sk_buff_head *head) 1077 { 1078 if (!skb_queue_empty(list)) { 1079 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1080 head->qlen += list->qlen; 1081 } 1082 } 1083 1084 /** 1085 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list 1086 * @list: the new list to add 1087 * @head: the place to add it in the first list 1088 * 1089 * Each of the lists is a queue. 1090 * The list at @list is reinitialised 1091 */ 1092 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, 1093 struct sk_buff_head *head) 1094 { 1095 if (!skb_queue_empty(list)) { 1096 __skb_queue_splice(list, head->prev, (struct sk_buff *) head); 1097 head->qlen += list->qlen; 1098 __skb_queue_head_init(list); 1099 } 1100 } 1101 1102 /** 1103 * __skb_queue_after - queue a buffer at the list head 1104 * @list: list to use 1105 * @prev: place after this buffer 1106 * @newsk: buffer to queue 1107 * 1108 * Queue a buffer int the middle of a list. This function takes no locks 1109 * and you must therefore hold required locks before calling it. 1110 * 1111 * A buffer cannot be placed on two lists at the same time. 1112 */ 1113 static inline void __skb_queue_after(struct sk_buff_head *list, 1114 struct sk_buff *prev, 1115 struct sk_buff *newsk) 1116 { 1117 __skb_insert(newsk, prev, prev->next, list); 1118 } 1119 1120 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, 1121 struct sk_buff_head *list); 1122 1123 static inline void __skb_queue_before(struct sk_buff_head *list, 1124 struct sk_buff *next, 1125 struct sk_buff *newsk) 1126 { 1127 __skb_insert(newsk, next->prev, next, list); 1128 } 1129 1130 /** 1131 * __skb_queue_head - queue a buffer at the list head 1132 * @list: list to use 1133 * @newsk: buffer to queue 1134 * 1135 * Queue a buffer at the start of a list. This function takes no locks 1136 * and you must therefore hold required locks before calling it. 1137 * 1138 * A buffer cannot be placed on two lists at the same time. 1139 */ 1140 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); 1141 static inline void __skb_queue_head(struct sk_buff_head *list, 1142 struct sk_buff *newsk) 1143 { 1144 __skb_queue_after(list, (struct sk_buff *)list, newsk); 1145 } 1146 1147 /** 1148 * __skb_queue_tail - queue a buffer at the list tail 1149 * @list: list to use 1150 * @newsk: buffer to queue 1151 * 1152 * Queue a buffer at the end of a list. This function takes no locks 1153 * and you must therefore hold required locks before calling it. 1154 * 1155 * A buffer cannot be placed on two lists at the same time. 1156 */ 1157 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); 1158 static inline void __skb_queue_tail(struct sk_buff_head *list, 1159 struct sk_buff *newsk) 1160 { 1161 __skb_queue_before(list, (struct sk_buff *)list, newsk); 1162 } 1163 1164 /* 1165 * remove sk_buff from list. _Must_ be called atomically, and with 1166 * the list known.. 1167 */ 1168 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); 1169 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) 1170 { 1171 struct sk_buff *next, *prev; 1172 1173 list->qlen--; 1174 next = skb->next; 1175 prev = skb->prev; 1176 skb->next = skb->prev = NULL; 1177 next->prev = prev; 1178 prev->next = next; 1179 } 1180 1181 /** 1182 * __skb_dequeue - remove from the head of the queue 1183 * @list: list to dequeue from 1184 * 1185 * Remove the head of the list. This function does not take any locks 1186 * so must be used with appropriate locks held only. The head item is 1187 * returned or %NULL if the list is empty. 1188 */ 1189 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list); 1190 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) 1191 { 1192 struct sk_buff *skb = skb_peek(list); 1193 if (skb) 1194 __skb_unlink(skb, list); 1195 return skb; 1196 } 1197 1198 /** 1199 * __skb_dequeue_tail - remove from the tail of the queue 1200 * @list: list to dequeue from 1201 * 1202 * Remove the tail of the list. This function does not take any locks 1203 * so must be used with appropriate locks held only. The tail item is 1204 * returned or %NULL if the list is empty. 1205 */ 1206 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); 1207 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) 1208 { 1209 struct sk_buff *skb = skb_peek_tail(list); 1210 if (skb) 1211 __skb_unlink(skb, list); 1212 return skb; 1213 } 1214 1215 1216 static inline bool skb_is_nonlinear(const struct sk_buff *skb) 1217 { 1218 return skb->data_len; 1219 } 1220 1221 static inline unsigned int skb_headlen(const struct sk_buff *skb) 1222 { 1223 return skb->len - skb->data_len; 1224 } 1225 1226 static inline int skb_pagelen(const struct sk_buff *skb) 1227 { 1228 int i, len = 0; 1229 1230 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) 1231 len += skb_frag_size(&skb_shinfo(skb)->frags[i]); 1232 return len + skb_headlen(skb); 1233 } 1234 1235 /** 1236 * __skb_fill_page_desc - initialise a paged fragment in an skb 1237 * @skb: buffer containing fragment to be initialised 1238 * @i: paged fragment index to initialise 1239 * @page: the page to use for this fragment 1240 * @off: the offset to the data with @page 1241 * @size: the length of the data 1242 * 1243 * Initialises the @i'th fragment of @skb to point to &size bytes at 1244 * offset @off within @page. 1245 * 1246 * Does not take any additional reference on the fragment. 1247 */ 1248 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, 1249 struct page *page, int off, int size) 1250 { 1251 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 1252 1253 /* 1254 * Propagate page->pfmemalloc to the skb if we can. The problem is 1255 * that not all callers have unique ownership of the page. If 1256 * pfmemalloc is set, we check the mapping as a mapping implies 1257 * page->index is set (index and pfmemalloc share space). 1258 * If it's a valid mapping, we cannot use page->pfmemalloc but we 1259 * do not lose pfmemalloc information as the pages would not be 1260 * allocated using __GFP_MEMALLOC. 1261 */ 1262 if (page->pfmemalloc && !page->mapping) 1263 skb->pfmemalloc = true; 1264 frag->page.p = page; 1265 frag->page_offset = off; 1266 skb_frag_size_set(frag, size); 1267 } 1268 1269 /** 1270 * skb_fill_page_desc - initialise a paged fragment in an skb 1271 * @skb: buffer containing fragment to be initialised 1272 * @i: paged fragment index to initialise 1273 * @page: the page to use for this fragment 1274 * @off: the offset to the data with @page 1275 * @size: the length of the data 1276 * 1277 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of 1278 * @skb to point to &size bytes at offset @off within @page. In 1279 * addition updates @skb such that @i is the last fragment. 1280 * 1281 * Does not take any additional reference on the fragment. 1282 */ 1283 static inline void skb_fill_page_desc(struct sk_buff *skb, int i, 1284 struct page *page, int off, int size) 1285 { 1286 __skb_fill_page_desc(skb, i, page, off, size); 1287 skb_shinfo(skb)->nr_frags = i + 1; 1288 } 1289 1290 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, 1291 int off, int size, unsigned int truesize); 1292 1293 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) 1294 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb)) 1295 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) 1296 1297 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1298 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1299 { 1300 return skb->head + skb->tail; 1301 } 1302 1303 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1304 { 1305 skb->tail = skb->data - skb->head; 1306 } 1307 1308 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1309 { 1310 skb_reset_tail_pointer(skb); 1311 skb->tail += offset; 1312 } 1313 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 1314 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) 1315 { 1316 return skb->tail; 1317 } 1318 1319 static inline void skb_reset_tail_pointer(struct sk_buff *skb) 1320 { 1321 skb->tail = skb->data; 1322 } 1323 1324 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) 1325 { 1326 skb->tail = skb->data + offset; 1327 } 1328 1329 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1330 1331 /* 1332 * Add data to an sk_buff 1333 */ 1334 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len); 1335 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) 1336 { 1337 unsigned char *tmp = skb_tail_pointer(skb); 1338 SKB_LINEAR_ASSERT(skb); 1339 skb->tail += len; 1340 skb->len += len; 1341 return tmp; 1342 } 1343 1344 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len); 1345 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) 1346 { 1347 skb->data -= len; 1348 skb->len += len; 1349 return skb->data; 1350 } 1351 1352 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); 1353 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) 1354 { 1355 skb->len -= len; 1356 BUG_ON(skb->len < skb->data_len); 1357 return skb->data += len; 1358 } 1359 1360 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len) 1361 { 1362 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); 1363 } 1364 1365 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); 1366 1367 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) 1368 { 1369 if (len > skb_headlen(skb) && 1370 !__pskb_pull_tail(skb, len - skb_headlen(skb))) 1371 return NULL; 1372 skb->len -= len; 1373 return skb->data += len; 1374 } 1375 1376 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) 1377 { 1378 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); 1379 } 1380 1381 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) 1382 { 1383 if (likely(len <= skb_headlen(skb))) 1384 return 1; 1385 if (unlikely(len > skb->len)) 1386 return 0; 1387 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; 1388 } 1389 1390 /** 1391 * skb_headroom - bytes at buffer head 1392 * @skb: buffer to check 1393 * 1394 * Return the number of bytes of free space at the head of an &sk_buff. 1395 */ 1396 static inline unsigned int skb_headroom(const struct sk_buff *skb) 1397 { 1398 return skb->data - skb->head; 1399 } 1400 1401 /** 1402 * skb_tailroom - bytes at buffer end 1403 * @skb: buffer to check 1404 * 1405 * Return the number of bytes of free space at the tail of an sk_buff 1406 */ 1407 static inline int skb_tailroom(const struct sk_buff *skb) 1408 { 1409 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; 1410 } 1411 1412 /** 1413 * skb_availroom - bytes at buffer end 1414 * @skb: buffer to check 1415 * 1416 * Return the number of bytes of free space at the tail of an sk_buff 1417 * allocated by sk_stream_alloc() 1418 */ 1419 static inline int skb_availroom(const struct sk_buff *skb) 1420 { 1421 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len; 1422 } 1423 1424 /** 1425 * skb_reserve - adjust headroom 1426 * @skb: buffer to alter 1427 * @len: bytes to move 1428 * 1429 * Increase the headroom of an empty &sk_buff by reducing the tail 1430 * room. This is only allowed for an empty buffer. 1431 */ 1432 static inline void skb_reserve(struct sk_buff *skb, int len) 1433 { 1434 skb->data += len; 1435 skb->tail += len; 1436 } 1437 1438 static inline void skb_reset_mac_len(struct sk_buff *skb) 1439 { 1440 skb->mac_len = skb->network_header - skb->mac_header; 1441 } 1442 1443 #ifdef NET_SKBUFF_DATA_USES_OFFSET 1444 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1445 { 1446 return skb->head + skb->transport_header; 1447 } 1448 1449 static inline void skb_reset_transport_header(struct sk_buff *skb) 1450 { 1451 skb->transport_header = skb->data - skb->head; 1452 } 1453 1454 static inline void skb_set_transport_header(struct sk_buff *skb, 1455 const int offset) 1456 { 1457 skb_reset_transport_header(skb); 1458 skb->transport_header += offset; 1459 } 1460 1461 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1462 { 1463 return skb->head + skb->network_header; 1464 } 1465 1466 static inline void skb_reset_network_header(struct sk_buff *skb) 1467 { 1468 skb->network_header = skb->data - skb->head; 1469 } 1470 1471 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1472 { 1473 skb_reset_network_header(skb); 1474 skb->network_header += offset; 1475 } 1476 1477 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1478 { 1479 return skb->head + skb->mac_header; 1480 } 1481 1482 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1483 { 1484 return skb->mac_header != ~0U; 1485 } 1486 1487 static inline void skb_reset_mac_header(struct sk_buff *skb) 1488 { 1489 skb->mac_header = skb->data - skb->head; 1490 } 1491 1492 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1493 { 1494 skb_reset_mac_header(skb); 1495 skb->mac_header += offset; 1496 } 1497 1498 #else /* NET_SKBUFF_DATA_USES_OFFSET */ 1499 1500 static inline unsigned char *skb_transport_header(const struct sk_buff *skb) 1501 { 1502 return skb->transport_header; 1503 } 1504 1505 static inline void skb_reset_transport_header(struct sk_buff *skb) 1506 { 1507 skb->transport_header = skb->data; 1508 } 1509 1510 static inline void skb_set_transport_header(struct sk_buff *skb, 1511 const int offset) 1512 { 1513 skb->transport_header = skb->data + offset; 1514 } 1515 1516 static inline unsigned char *skb_network_header(const struct sk_buff *skb) 1517 { 1518 return skb->network_header; 1519 } 1520 1521 static inline void skb_reset_network_header(struct sk_buff *skb) 1522 { 1523 skb->network_header = skb->data; 1524 } 1525 1526 static inline void skb_set_network_header(struct sk_buff *skb, const int offset) 1527 { 1528 skb->network_header = skb->data + offset; 1529 } 1530 1531 static inline unsigned char *skb_mac_header(const struct sk_buff *skb) 1532 { 1533 return skb->mac_header; 1534 } 1535 1536 static inline int skb_mac_header_was_set(const struct sk_buff *skb) 1537 { 1538 return skb->mac_header != NULL; 1539 } 1540 1541 static inline void skb_reset_mac_header(struct sk_buff *skb) 1542 { 1543 skb->mac_header = skb->data; 1544 } 1545 1546 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) 1547 { 1548 skb->mac_header = skb->data + offset; 1549 } 1550 #endif /* NET_SKBUFF_DATA_USES_OFFSET */ 1551 1552 static inline void skb_mac_header_rebuild(struct sk_buff *skb) 1553 { 1554 if (skb_mac_header_was_set(skb)) { 1555 const unsigned char *old_mac = skb_mac_header(skb); 1556 1557 skb_set_mac_header(skb, -skb->mac_len); 1558 memmove(skb_mac_header(skb), old_mac, skb->mac_len); 1559 } 1560 } 1561 1562 static inline int skb_checksum_start_offset(const struct sk_buff *skb) 1563 { 1564 return skb->csum_start - skb_headroom(skb); 1565 } 1566 1567 static inline int skb_transport_offset(const struct sk_buff *skb) 1568 { 1569 return skb_transport_header(skb) - skb->data; 1570 } 1571 1572 static inline u32 skb_network_header_len(const struct sk_buff *skb) 1573 { 1574 return skb->transport_header - skb->network_header; 1575 } 1576 1577 static inline int skb_network_offset(const struct sk_buff *skb) 1578 { 1579 return skb_network_header(skb) - skb->data; 1580 } 1581 1582 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) 1583 { 1584 return pskb_may_pull(skb, skb_network_offset(skb) + len); 1585 } 1586 1587 /* 1588 * CPUs often take a performance hit when accessing unaligned memory 1589 * locations. The actual performance hit varies, it can be small if the 1590 * hardware handles it or large if we have to take an exception and fix it 1591 * in software. 1592 * 1593 * Since an ethernet header is 14 bytes network drivers often end up with 1594 * the IP header at an unaligned offset. The IP header can be aligned by 1595 * shifting the start of the packet by 2 bytes. Drivers should do this 1596 * with: 1597 * 1598 * skb_reserve(skb, NET_IP_ALIGN); 1599 * 1600 * The downside to this alignment of the IP header is that the DMA is now 1601 * unaligned. On some architectures the cost of an unaligned DMA is high 1602 * and this cost outweighs the gains made by aligning the IP header. 1603 * 1604 * Since this trade off varies between architectures, we allow NET_IP_ALIGN 1605 * to be overridden. 1606 */ 1607 #ifndef NET_IP_ALIGN 1608 #define NET_IP_ALIGN 2 1609 #endif 1610 1611 /* 1612 * The networking layer reserves some headroom in skb data (via 1613 * dev_alloc_skb). This is used to avoid having to reallocate skb data when 1614 * the header has to grow. In the default case, if the header has to grow 1615 * 32 bytes or less we avoid the reallocation. 1616 * 1617 * Unfortunately this headroom changes the DMA alignment of the resulting 1618 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive 1619 * on some architectures. An architecture can override this value, 1620 * perhaps setting it to a cacheline in size (since that will maintain 1621 * cacheline alignment of the DMA). It must be a power of 2. 1622 * 1623 * Various parts of the networking layer expect at least 32 bytes of 1624 * headroom, you should not reduce this. 1625 * 1626 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) 1627 * to reduce average number of cache lines per packet. 1628 * get_rps_cpus() for example only access one 64 bytes aligned block : 1629 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) 1630 */ 1631 #ifndef NET_SKB_PAD 1632 #define NET_SKB_PAD max(32, L1_CACHE_BYTES) 1633 #endif 1634 1635 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len); 1636 1637 static inline void __skb_trim(struct sk_buff *skb, unsigned int len) 1638 { 1639 if (unlikely(skb_is_nonlinear(skb))) { 1640 WARN_ON(1); 1641 return; 1642 } 1643 skb->len = len; 1644 skb_set_tail_pointer(skb, len); 1645 } 1646 1647 extern void skb_trim(struct sk_buff *skb, unsigned int len); 1648 1649 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) 1650 { 1651 if (skb->data_len) 1652 return ___pskb_trim(skb, len); 1653 __skb_trim(skb, len); 1654 return 0; 1655 } 1656 1657 static inline int pskb_trim(struct sk_buff *skb, unsigned int len) 1658 { 1659 return (len < skb->len) ? __pskb_trim(skb, len) : 0; 1660 } 1661 1662 /** 1663 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer 1664 * @skb: buffer to alter 1665 * @len: new length 1666 * 1667 * This is identical to pskb_trim except that the caller knows that 1668 * the skb is not cloned so we should never get an error due to out- 1669 * of-memory. 1670 */ 1671 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) 1672 { 1673 int err = pskb_trim(skb, len); 1674 BUG_ON(err); 1675 } 1676 1677 /** 1678 * skb_orphan - orphan a buffer 1679 * @skb: buffer to orphan 1680 * 1681 * If a buffer currently has an owner then we call the owner's 1682 * destructor function and make the @skb unowned. The buffer continues 1683 * to exist but is no longer charged to its former owner. 1684 */ 1685 static inline void skb_orphan(struct sk_buff *skb) 1686 { 1687 if (skb->destructor) 1688 skb->destructor(skb); 1689 skb->destructor = NULL; 1690 skb->sk = NULL; 1691 } 1692 1693 /** 1694 * skb_orphan_frags - orphan the frags contained in a buffer 1695 * @skb: buffer to orphan frags from 1696 * @gfp_mask: allocation mask for replacement pages 1697 * 1698 * For each frag in the SKB which needs a destructor (i.e. has an 1699 * owner) create a copy of that frag and release the original 1700 * page by calling the destructor. 1701 */ 1702 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) 1703 { 1704 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY))) 1705 return 0; 1706 return skb_copy_ubufs(skb, gfp_mask); 1707 } 1708 1709 /** 1710 * __skb_queue_purge - empty a list 1711 * @list: list to empty 1712 * 1713 * Delete all buffers on an &sk_buff list. Each buffer is removed from 1714 * the list and one reference dropped. This function does not take the 1715 * list lock and the caller must hold the relevant locks to use it. 1716 */ 1717 extern void skb_queue_purge(struct sk_buff_head *list); 1718 static inline void __skb_queue_purge(struct sk_buff_head *list) 1719 { 1720 struct sk_buff *skb; 1721 while ((skb = __skb_dequeue(list)) != NULL) 1722 kfree_skb(skb); 1723 } 1724 1725 extern void *netdev_alloc_frag(unsigned int fragsz); 1726 1727 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev, 1728 unsigned int length, 1729 gfp_t gfp_mask); 1730 1731 /** 1732 * netdev_alloc_skb - allocate an skbuff for rx on a specific device 1733 * @dev: network device to receive on 1734 * @length: length to allocate 1735 * 1736 * Allocate a new &sk_buff and assign it a usage count of one. The 1737 * buffer has unspecified headroom built in. Users should allocate 1738 * the headroom they think they need without accounting for the 1739 * built in space. The built in space is used for optimisations. 1740 * 1741 * %NULL is returned if there is no free memory. Although this function 1742 * allocates memory it can be called from an interrupt. 1743 */ 1744 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, 1745 unsigned int length) 1746 { 1747 return __netdev_alloc_skb(dev, length, GFP_ATOMIC); 1748 } 1749 1750 /* legacy helper around __netdev_alloc_skb() */ 1751 static inline struct sk_buff *__dev_alloc_skb(unsigned int length, 1752 gfp_t gfp_mask) 1753 { 1754 return __netdev_alloc_skb(NULL, length, gfp_mask); 1755 } 1756 1757 /* legacy helper around netdev_alloc_skb() */ 1758 static inline struct sk_buff *dev_alloc_skb(unsigned int length) 1759 { 1760 return netdev_alloc_skb(NULL, length); 1761 } 1762 1763 1764 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, 1765 unsigned int length, gfp_t gfp) 1766 { 1767 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); 1768 1769 if (NET_IP_ALIGN && skb) 1770 skb_reserve(skb, NET_IP_ALIGN); 1771 return skb; 1772 } 1773 1774 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, 1775 unsigned int length) 1776 { 1777 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); 1778 } 1779 1780 /* 1781 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data 1782 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX 1783 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used 1784 * @order: size of the allocation 1785 * 1786 * Allocate a new page. 1787 * 1788 * %NULL is returned if there is no free memory. 1789 */ 1790 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask, 1791 struct sk_buff *skb, 1792 unsigned int order) 1793 { 1794 struct page *page; 1795 1796 gfp_mask |= __GFP_COLD; 1797 1798 if (!(gfp_mask & __GFP_NOMEMALLOC)) 1799 gfp_mask |= __GFP_MEMALLOC; 1800 1801 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); 1802 if (skb && page && page->pfmemalloc) 1803 skb->pfmemalloc = true; 1804 1805 return page; 1806 } 1807 1808 /** 1809 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data 1810 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX 1811 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used 1812 * 1813 * Allocate a new page. 1814 * 1815 * %NULL is returned if there is no free memory. 1816 */ 1817 static inline struct page *__skb_alloc_page(gfp_t gfp_mask, 1818 struct sk_buff *skb) 1819 { 1820 return __skb_alloc_pages(gfp_mask, skb, 0); 1821 } 1822 1823 /** 1824 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page 1825 * @page: The page that was allocated from skb_alloc_page 1826 * @skb: The skb that may need pfmemalloc set 1827 */ 1828 static inline void skb_propagate_pfmemalloc(struct page *page, 1829 struct sk_buff *skb) 1830 { 1831 if (page && page->pfmemalloc) 1832 skb->pfmemalloc = true; 1833 } 1834 1835 /** 1836 * skb_frag_page - retrieve the page refered to by a paged fragment 1837 * @frag: the paged fragment 1838 * 1839 * Returns the &struct page associated with @frag. 1840 */ 1841 static inline struct page *skb_frag_page(const skb_frag_t *frag) 1842 { 1843 return frag->page.p; 1844 } 1845 1846 /** 1847 * __skb_frag_ref - take an addition reference on a paged fragment. 1848 * @frag: the paged fragment 1849 * 1850 * Takes an additional reference on the paged fragment @frag. 1851 */ 1852 static inline void __skb_frag_ref(skb_frag_t *frag) 1853 { 1854 get_page(skb_frag_page(frag)); 1855 } 1856 1857 /** 1858 * skb_frag_ref - take an addition reference on a paged fragment of an skb. 1859 * @skb: the buffer 1860 * @f: the fragment offset. 1861 * 1862 * Takes an additional reference on the @f'th paged fragment of @skb. 1863 */ 1864 static inline void skb_frag_ref(struct sk_buff *skb, int f) 1865 { 1866 __skb_frag_ref(&skb_shinfo(skb)->frags[f]); 1867 } 1868 1869 /** 1870 * __skb_frag_unref - release a reference on a paged fragment. 1871 * @frag: the paged fragment 1872 * 1873 * Releases a reference on the paged fragment @frag. 1874 */ 1875 static inline void __skb_frag_unref(skb_frag_t *frag) 1876 { 1877 put_page(skb_frag_page(frag)); 1878 } 1879 1880 /** 1881 * skb_frag_unref - release a reference on a paged fragment of an skb. 1882 * @skb: the buffer 1883 * @f: the fragment offset 1884 * 1885 * Releases a reference on the @f'th paged fragment of @skb. 1886 */ 1887 static inline void skb_frag_unref(struct sk_buff *skb, int f) 1888 { 1889 __skb_frag_unref(&skb_shinfo(skb)->frags[f]); 1890 } 1891 1892 /** 1893 * skb_frag_address - gets the address of the data contained in a paged fragment 1894 * @frag: the paged fragment buffer 1895 * 1896 * Returns the address of the data within @frag. The page must already 1897 * be mapped. 1898 */ 1899 static inline void *skb_frag_address(const skb_frag_t *frag) 1900 { 1901 return page_address(skb_frag_page(frag)) + frag->page_offset; 1902 } 1903 1904 /** 1905 * skb_frag_address_safe - gets the address of the data contained in a paged fragment 1906 * @frag: the paged fragment buffer 1907 * 1908 * Returns the address of the data within @frag. Checks that the page 1909 * is mapped and returns %NULL otherwise. 1910 */ 1911 static inline void *skb_frag_address_safe(const skb_frag_t *frag) 1912 { 1913 void *ptr = page_address(skb_frag_page(frag)); 1914 if (unlikely(!ptr)) 1915 return NULL; 1916 1917 return ptr + frag->page_offset; 1918 } 1919 1920 /** 1921 * __skb_frag_set_page - sets the page contained in a paged fragment 1922 * @frag: the paged fragment 1923 * @page: the page to set 1924 * 1925 * Sets the fragment @frag to contain @page. 1926 */ 1927 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) 1928 { 1929 frag->page.p = page; 1930 } 1931 1932 /** 1933 * skb_frag_set_page - sets the page contained in a paged fragment of an skb 1934 * @skb: the buffer 1935 * @f: the fragment offset 1936 * @page: the page to set 1937 * 1938 * Sets the @f'th fragment of @skb to contain @page. 1939 */ 1940 static inline void skb_frag_set_page(struct sk_buff *skb, int f, 1941 struct page *page) 1942 { 1943 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); 1944 } 1945 1946 /** 1947 * skb_frag_dma_map - maps a paged fragment via the DMA API 1948 * @dev: the device to map the fragment to 1949 * @frag: the paged fragment to map 1950 * @offset: the offset within the fragment (starting at the 1951 * fragment's own offset) 1952 * @size: the number of bytes to map 1953 * @dir: the direction of the mapping (%PCI_DMA_*) 1954 * 1955 * Maps the page associated with @frag to @device. 1956 */ 1957 static inline dma_addr_t skb_frag_dma_map(struct device *dev, 1958 const skb_frag_t *frag, 1959 size_t offset, size_t size, 1960 enum dma_data_direction dir) 1961 { 1962 return dma_map_page(dev, skb_frag_page(frag), 1963 frag->page_offset + offset, size, dir); 1964 } 1965 1966 static inline struct sk_buff *pskb_copy(struct sk_buff *skb, 1967 gfp_t gfp_mask) 1968 { 1969 return __pskb_copy(skb, skb_headroom(skb), gfp_mask); 1970 } 1971 1972 /** 1973 * skb_clone_writable - is the header of a clone writable 1974 * @skb: buffer to check 1975 * @len: length up to which to write 1976 * 1977 * Returns true if modifying the header part of the cloned buffer 1978 * does not requires the data to be copied. 1979 */ 1980 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) 1981 { 1982 return !skb_header_cloned(skb) && 1983 skb_headroom(skb) + len <= skb->hdr_len; 1984 } 1985 1986 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, 1987 int cloned) 1988 { 1989 int delta = 0; 1990 1991 if (headroom > skb_headroom(skb)) 1992 delta = headroom - skb_headroom(skb); 1993 1994 if (delta || cloned) 1995 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, 1996 GFP_ATOMIC); 1997 return 0; 1998 } 1999 2000 /** 2001 * skb_cow - copy header of skb when it is required 2002 * @skb: buffer to cow 2003 * @headroom: needed headroom 2004 * 2005 * If the skb passed lacks sufficient headroom or its data part 2006 * is shared, data is reallocated. If reallocation fails, an error 2007 * is returned and original skb is not changed. 2008 * 2009 * The result is skb with writable area skb->head...skb->tail 2010 * and at least @headroom of space at head. 2011 */ 2012 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) 2013 { 2014 return __skb_cow(skb, headroom, skb_cloned(skb)); 2015 } 2016 2017 /** 2018 * skb_cow_head - skb_cow but only making the head writable 2019 * @skb: buffer to cow 2020 * @headroom: needed headroom 2021 * 2022 * This function is identical to skb_cow except that we replace the 2023 * skb_cloned check by skb_header_cloned. It should be used when 2024 * you only need to push on some header and do not need to modify 2025 * the data. 2026 */ 2027 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) 2028 { 2029 return __skb_cow(skb, headroom, skb_header_cloned(skb)); 2030 } 2031 2032 /** 2033 * skb_padto - pad an skbuff up to a minimal size 2034 * @skb: buffer to pad 2035 * @len: minimal length 2036 * 2037 * Pads up a buffer to ensure the trailing bytes exist and are 2038 * blanked. If the buffer already contains sufficient data it 2039 * is untouched. Otherwise it is extended. Returns zero on 2040 * success. The skb is freed on error. 2041 */ 2042 2043 static inline int skb_padto(struct sk_buff *skb, unsigned int len) 2044 { 2045 unsigned int size = skb->len; 2046 if (likely(size >= len)) 2047 return 0; 2048 return skb_pad(skb, len - size); 2049 } 2050 2051 static inline int skb_add_data(struct sk_buff *skb, 2052 char __user *from, int copy) 2053 { 2054 const int off = skb->len; 2055 2056 if (skb->ip_summed == CHECKSUM_NONE) { 2057 int err = 0; 2058 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy), 2059 copy, 0, &err); 2060 if (!err) { 2061 skb->csum = csum_block_add(skb->csum, csum, off); 2062 return 0; 2063 } 2064 } else if (!copy_from_user(skb_put(skb, copy), from, copy)) 2065 return 0; 2066 2067 __skb_trim(skb, off); 2068 return -EFAULT; 2069 } 2070 2071 static inline bool skb_can_coalesce(struct sk_buff *skb, int i, 2072 const struct page *page, int off) 2073 { 2074 if (i) { 2075 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; 2076 2077 return page == skb_frag_page(frag) && 2078 off == frag->page_offset + skb_frag_size(frag); 2079 } 2080 return false; 2081 } 2082 2083 static inline int __skb_linearize(struct sk_buff *skb) 2084 { 2085 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; 2086 } 2087 2088 /** 2089 * skb_linearize - convert paged skb to linear one 2090 * @skb: buffer to linarize 2091 * 2092 * If there is no free memory -ENOMEM is returned, otherwise zero 2093 * is returned and the old skb data released. 2094 */ 2095 static inline int skb_linearize(struct sk_buff *skb) 2096 { 2097 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; 2098 } 2099 2100 /** 2101 * skb_linearize_cow - make sure skb is linear and writable 2102 * @skb: buffer to process 2103 * 2104 * If there is no free memory -ENOMEM is returned, otherwise zero 2105 * is returned and the old skb data released. 2106 */ 2107 static inline int skb_linearize_cow(struct sk_buff *skb) 2108 { 2109 return skb_is_nonlinear(skb) || skb_cloned(skb) ? 2110 __skb_linearize(skb) : 0; 2111 } 2112 2113 /** 2114 * skb_postpull_rcsum - update checksum for received skb after pull 2115 * @skb: buffer to update 2116 * @start: start of data before pull 2117 * @len: length of data pulled 2118 * 2119 * After doing a pull on a received packet, you need to call this to 2120 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to 2121 * CHECKSUM_NONE so that it can be recomputed from scratch. 2122 */ 2123 2124 static inline void skb_postpull_rcsum(struct sk_buff *skb, 2125 const void *start, unsigned int len) 2126 { 2127 if (skb->ip_summed == CHECKSUM_COMPLETE) 2128 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); 2129 } 2130 2131 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); 2132 2133 /** 2134 * pskb_trim_rcsum - trim received skb and update checksum 2135 * @skb: buffer to trim 2136 * @len: new length 2137 * 2138 * This is exactly the same as pskb_trim except that it ensures the 2139 * checksum of received packets are still valid after the operation. 2140 */ 2141 2142 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) 2143 { 2144 if (likely(len >= skb->len)) 2145 return 0; 2146 if (skb->ip_summed == CHECKSUM_COMPLETE) 2147 skb->ip_summed = CHECKSUM_NONE; 2148 return __pskb_trim(skb, len); 2149 } 2150 2151 #define skb_queue_walk(queue, skb) \ 2152 for (skb = (queue)->next; \ 2153 skb != (struct sk_buff *)(queue); \ 2154 skb = skb->next) 2155 2156 #define skb_queue_walk_safe(queue, skb, tmp) \ 2157 for (skb = (queue)->next, tmp = skb->next; \ 2158 skb != (struct sk_buff *)(queue); \ 2159 skb = tmp, tmp = skb->next) 2160 2161 #define skb_queue_walk_from(queue, skb) \ 2162 for (; skb != (struct sk_buff *)(queue); \ 2163 skb = skb->next) 2164 2165 #define skb_queue_walk_from_safe(queue, skb, tmp) \ 2166 for (tmp = skb->next; \ 2167 skb != (struct sk_buff *)(queue); \ 2168 skb = tmp, tmp = skb->next) 2169 2170 #define skb_queue_reverse_walk(queue, skb) \ 2171 for (skb = (queue)->prev; \ 2172 skb != (struct sk_buff *)(queue); \ 2173 skb = skb->prev) 2174 2175 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ 2176 for (skb = (queue)->prev, tmp = skb->prev; \ 2177 skb != (struct sk_buff *)(queue); \ 2178 skb = tmp, tmp = skb->prev) 2179 2180 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ 2181 for (tmp = skb->prev; \ 2182 skb != (struct sk_buff *)(queue); \ 2183 skb = tmp, tmp = skb->prev) 2184 2185 static inline bool skb_has_frag_list(const struct sk_buff *skb) 2186 { 2187 return skb_shinfo(skb)->frag_list != NULL; 2188 } 2189 2190 static inline void skb_frag_list_init(struct sk_buff *skb) 2191 { 2192 skb_shinfo(skb)->frag_list = NULL; 2193 } 2194 2195 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag) 2196 { 2197 frag->next = skb_shinfo(skb)->frag_list; 2198 skb_shinfo(skb)->frag_list = frag; 2199 } 2200 2201 #define skb_walk_frags(skb, iter) \ 2202 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) 2203 2204 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, 2205 int *peeked, int *off, int *err); 2206 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, 2207 int noblock, int *err); 2208 extern unsigned int datagram_poll(struct file *file, struct socket *sock, 2209 struct poll_table_struct *wait); 2210 extern int skb_copy_datagram_iovec(const struct sk_buff *from, 2211 int offset, struct iovec *to, 2212 int size); 2213 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, 2214 int hlen, 2215 struct iovec *iov); 2216 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb, 2217 int offset, 2218 const struct iovec *from, 2219 int from_offset, 2220 int len); 2221 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from, 2222 int offset, 2223 const struct iovec *to, 2224 int to_offset, 2225 int size); 2226 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb); 2227 extern void skb_free_datagram_locked(struct sock *sk, 2228 struct sk_buff *skb); 2229 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, 2230 unsigned int flags); 2231 extern __wsum skb_checksum(const struct sk_buff *skb, int offset, 2232 int len, __wsum csum); 2233 extern int skb_copy_bits(const struct sk_buff *skb, int offset, 2234 void *to, int len); 2235 extern int skb_store_bits(struct sk_buff *skb, int offset, 2236 const void *from, int len); 2237 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, 2238 int offset, u8 *to, int len, 2239 __wsum csum); 2240 extern int skb_splice_bits(struct sk_buff *skb, 2241 unsigned int offset, 2242 struct pipe_inode_info *pipe, 2243 unsigned int len, 2244 unsigned int flags); 2245 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); 2246 extern void skb_split(struct sk_buff *skb, 2247 struct sk_buff *skb1, const u32 len); 2248 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, 2249 int shiftlen); 2250 2251 extern struct sk_buff *skb_segment(struct sk_buff *skb, 2252 netdev_features_t features); 2253 2254 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset, 2255 int len, void *buffer) 2256 { 2257 int hlen = skb_headlen(skb); 2258 2259 if (hlen - offset >= len) 2260 return skb->data + offset; 2261 2262 if (skb_copy_bits(skb, offset, buffer, len) < 0) 2263 return NULL; 2264 2265 return buffer; 2266 } 2267 2268 static inline void skb_copy_from_linear_data(const struct sk_buff *skb, 2269 void *to, 2270 const unsigned int len) 2271 { 2272 memcpy(to, skb->data, len); 2273 } 2274 2275 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, 2276 const int offset, void *to, 2277 const unsigned int len) 2278 { 2279 memcpy(to, skb->data + offset, len); 2280 } 2281 2282 static inline void skb_copy_to_linear_data(struct sk_buff *skb, 2283 const void *from, 2284 const unsigned int len) 2285 { 2286 memcpy(skb->data, from, len); 2287 } 2288 2289 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, 2290 const int offset, 2291 const void *from, 2292 const unsigned int len) 2293 { 2294 memcpy(skb->data + offset, from, len); 2295 } 2296 2297 extern void skb_init(void); 2298 2299 static inline ktime_t skb_get_ktime(const struct sk_buff *skb) 2300 { 2301 return skb->tstamp; 2302 } 2303 2304 /** 2305 * skb_get_timestamp - get timestamp from a skb 2306 * @skb: skb to get stamp from 2307 * @stamp: pointer to struct timeval to store stamp in 2308 * 2309 * Timestamps are stored in the skb as offsets to a base timestamp. 2310 * This function converts the offset back to a struct timeval and stores 2311 * it in stamp. 2312 */ 2313 static inline void skb_get_timestamp(const struct sk_buff *skb, 2314 struct timeval *stamp) 2315 { 2316 *stamp = ktime_to_timeval(skb->tstamp); 2317 } 2318 2319 static inline void skb_get_timestampns(const struct sk_buff *skb, 2320 struct timespec *stamp) 2321 { 2322 *stamp = ktime_to_timespec(skb->tstamp); 2323 } 2324 2325 static inline void __net_timestamp(struct sk_buff *skb) 2326 { 2327 skb->tstamp = ktime_get_real(); 2328 } 2329 2330 static inline ktime_t net_timedelta(ktime_t t) 2331 { 2332 return ktime_sub(ktime_get_real(), t); 2333 } 2334 2335 static inline ktime_t net_invalid_timestamp(void) 2336 { 2337 return ktime_set(0, 0); 2338 } 2339 2340 extern void skb_timestamping_init(void); 2341 2342 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 2343 2344 extern void skb_clone_tx_timestamp(struct sk_buff *skb); 2345 extern bool skb_defer_rx_timestamp(struct sk_buff *skb); 2346 2347 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ 2348 2349 static inline void skb_clone_tx_timestamp(struct sk_buff *skb) 2350 { 2351 } 2352 2353 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) 2354 { 2355 return false; 2356 } 2357 2358 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ 2359 2360 /** 2361 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps 2362 * 2363 * PHY drivers may accept clones of transmitted packets for 2364 * timestamping via their phy_driver.txtstamp method. These drivers 2365 * must call this function to return the skb back to the stack, with 2366 * or without a timestamp. 2367 * 2368 * @skb: clone of the the original outgoing packet 2369 * @hwtstamps: hardware time stamps, may be NULL if not available 2370 * 2371 */ 2372 void skb_complete_tx_timestamp(struct sk_buff *skb, 2373 struct skb_shared_hwtstamps *hwtstamps); 2374 2375 /** 2376 * skb_tstamp_tx - queue clone of skb with send time stamps 2377 * @orig_skb: the original outgoing packet 2378 * @hwtstamps: hardware time stamps, may be NULL if not available 2379 * 2380 * If the skb has a socket associated, then this function clones the 2381 * skb (thus sharing the actual data and optional structures), stores 2382 * the optional hardware time stamping information (if non NULL) or 2383 * generates a software time stamp (otherwise), then queues the clone 2384 * to the error queue of the socket. Errors are silently ignored. 2385 */ 2386 extern void skb_tstamp_tx(struct sk_buff *orig_skb, 2387 struct skb_shared_hwtstamps *hwtstamps); 2388 2389 static inline void sw_tx_timestamp(struct sk_buff *skb) 2390 { 2391 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP && 2392 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) 2393 skb_tstamp_tx(skb, NULL); 2394 } 2395 2396 /** 2397 * skb_tx_timestamp() - Driver hook for transmit timestamping 2398 * 2399 * Ethernet MAC Drivers should call this function in their hard_xmit() 2400 * function immediately before giving the sk_buff to the MAC hardware. 2401 * 2402 * @skb: A socket buffer. 2403 */ 2404 static inline void skb_tx_timestamp(struct sk_buff *skb) 2405 { 2406 skb_clone_tx_timestamp(skb); 2407 sw_tx_timestamp(skb); 2408 } 2409 2410 /** 2411 * skb_complete_wifi_ack - deliver skb with wifi status 2412 * 2413 * @skb: the original outgoing packet 2414 * @acked: ack status 2415 * 2416 */ 2417 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); 2418 2419 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); 2420 extern __sum16 __skb_checksum_complete(struct sk_buff *skb); 2421 2422 static inline int skb_csum_unnecessary(const struct sk_buff *skb) 2423 { 2424 return skb->ip_summed & CHECKSUM_UNNECESSARY; 2425 } 2426 2427 /** 2428 * skb_checksum_complete - Calculate checksum of an entire packet 2429 * @skb: packet to process 2430 * 2431 * This function calculates the checksum over the entire packet plus 2432 * the value of skb->csum. The latter can be used to supply the 2433 * checksum of a pseudo header as used by TCP/UDP. It returns the 2434 * checksum. 2435 * 2436 * For protocols that contain complete checksums such as ICMP/TCP/UDP, 2437 * this function can be used to verify that checksum on received 2438 * packets. In that case the function should return zero if the 2439 * checksum is correct. In particular, this function will return zero 2440 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the 2441 * hardware has already verified the correctness of the checksum. 2442 */ 2443 static inline __sum16 skb_checksum_complete(struct sk_buff *skb) 2444 { 2445 return skb_csum_unnecessary(skb) ? 2446 0 : __skb_checksum_complete(skb); 2447 } 2448 2449 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2450 extern void nf_conntrack_destroy(struct nf_conntrack *nfct); 2451 static inline void nf_conntrack_put(struct nf_conntrack *nfct) 2452 { 2453 if (nfct && atomic_dec_and_test(&nfct->use)) 2454 nf_conntrack_destroy(nfct); 2455 } 2456 static inline void nf_conntrack_get(struct nf_conntrack *nfct) 2457 { 2458 if (nfct) 2459 atomic_inc(&nfct->use); 2460 } 2461 #endif 2462 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2463 static inline void nf_conntrack_get_reasm(struct sk_buff *skb) 2464 { 2465 if (skb) 2466 atomic_inc(&skb->users); 2467 } 2468 static inline void nf_conntrack_put_reasm(struct sk_buff *skb) 2469 { 2470 if (skb) 2471 kfree_skb(skb); 2472 } 2473 #endif 2474 #ifdef CONFIG_BRIDGE_NETFILTER 2475 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) 2476 { 2477 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) 2478 kfree(nf_bridge); 2479 } 2480 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) 2481 { 2482 if (nf_bridge) 2483 atomic_inc(&nf_bridge->use); 2484 } 2485 #endif /* CONFIG_BRIDGE_NETFILTER */ 2486 static inline void nf_reset(struct sk_buff *skb) 2487 { 2488 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2489 nf_conntrack_put(skb->nfct); 2490 skb->nfct = NULL; 2491 #endif 2492 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2493 nf_conntrack_put_reasm(skb->nfct_reasm); 2494 skb->nfct_reasm = NULL; 2495 #endif 2496 #ifdef CONFIG_BRIDGE_NETFILTER 2497 nf_bridge_put(skb->nf_bridge); 2498 skb->nf_bridge = NULL; 2499 #endif 2500 } 2501 2502 /* Note: This doesn't put any conntrack and bridge info in dst. */ 2503 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src) 2504 { 2505 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2506 dst->nfct = src->nfct; 2507 nf_conntrack_get(src->nfct); 2508 dst->nfctinfo = src->nfctinfo; 2509 #endif 2510 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2511 dst->nfct_reasm = src->nfct_reasm; 2512 nf_conntrack_get_reasm(src->nfct_reasm); 2513 #endif 2514 #ifdef CONFIG_BRIDGE_NETFILTER 2515 dst->nf_bridge = src->nf_bridge; 2516 nf_bridge_get(src->nf_bridge); 2517 #endif 2518 } 2519 2520 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) 2521 { 2522 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 2523 nf_conntrack_put(dst->nfct); 2524 #endif 2525 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED 2526 nf_conntrack_put_reasm(dst->nfct_reasm); 2527 #endif 2528 #ifdef CONFIG_BRIDGE_NETFILTER 2529 nf_bridge_put(dst->nf_bridge); 2530 #endif 2531 __nf_copy(dst, src); 2532 } 2533 2534 #ifdef CONFIG_NETWORK_SECMARK 2535 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 2536 { 2537 to->secmark = from->secmark; 2538 } 2539 2540 static inline void skb_init_secmark(struct sk_buff *skb) 2541 { 2542 skb->secmark = 0; 2543 } 2544 #else 2545 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) 2546 { } 2547 2548 static inline void skb_init_secmark(struct sk_buff *skb) 2549 { } 2550 #endif 2551 2552 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) 2553 { 2554 skb->queue_mapping = queue_mapping; 2555 } 2556 2557 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) 2558 { 2559 return skb->queue_mapping; 2560 } 2561 2562 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) 2563 { 2564 to->queue_mapping = from->queue_mapping; 2565 } 2566 2567 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) 2568 { 2569 skb->queue_mapping = rx_queue + 1; 2570 } 2571 2572 static inline u16 skb_get_rx_queue(const struct sk_buff *skb) 2573 { 2574 return skb->queue_mapping - 1; 2575 } 2576 2577 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) 2578 { 2579 return skb->queue_mapping != 0; 2580 } 2581 2582 extern u16 __skb_tx_hash(const struct net_device *dev, 2583 const struct sk_buff *skb, 2584 unsigned int num_tx_queues); 2585 2586 #ifdef CONFIG_XFRM 2587 static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 2588 { 2589 return skb->sp; 2590 } 2591 #else 2592 static inline struct sec_path *skb_sec_path(struct sk_buff *skb) 2593 { 2594 return NULL; 2595 } 2596 #endif 2597 2598 static inline bool skb_is_gso(const struct sk_buff *skb) 2599 { 2600 return skb_shinfo(skb)->gso_size; 2601 } 2602 2603 static inline bool skb_is_gso_v6(const struct sk_buff *skb) 2604 { 2605 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; 2606 } 2607 2608 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb); 2609 2610 static inline bool skb_warn_if_lro(const struct sk_buff *skb) 2611 { 2612 /* LRO sets gso_size but not gso_type, whereas if GSO is really 2613 * wanted then gso_type will be set. */ 2614 const struct skb_shared_info *shinfo = skb_shinfo(skb); 2615 2616 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && 2617 unlikely(shinfo->gso_type == 0)) { 2618 __skb_warn_lro_forwarding(skb); 2619 return true; 2620 } 2621 return false; 2622 } 2623 2624 static inline void skb_forward_csum(struct sk_buff *skb) 2625 { 2626 /* Unfortunately we don't support this one. Any brave souls? */ 2627 if (skb->ip_summed == CHECKSUM_COMPLETE) 2628 skb->ip_summed = CHECKSUM_NONE; 2629 } 2630 2631 /** 2632 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE 2633 * @skb: skb to check 2634 * 2635 * fresh skbs have their ip_summed set to CHECKSUM_NONE. 2636 * Instead of forcing ip_summed to CHECKSUM_NONE, we can 2637 * use this helper, to document places where we make this assertion. 2638 */ 2639 static inline void skb_checksum_none_assert(const struct sk_buff *skb) 2640 { 2641 #ifdef DEBUG 2642 BUG_ON(skb->ip_summed != CHECKSUM_NONE); 2643 #endif 2644 } 2645 2646 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); 2647 2648 static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size) 2649 { 2650 if (irqs_disabled()) 2651 return false; 2652 2653 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) 2654 return false; 2655 2656 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE) 2657 return false; 2658 2659 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD); 2660 if (skb_end_offset(skb) < skb_size) 2661 return false; 2662 2663 if (skb_shared(skb) || skb_cloned(skb)) 2664 return false; 2665 2666 return true; 2667 } 2668 2669 /** 2670 * skb_head_is_locked - Determine if the skb->head is locked down 2671 * @skb: skb to check 2672 * 2673 * The head on skbs build around a head frag can be removed if they are 2674 * not cloned. This function returns true if the skb head is locked down 2675 * due to either being allocated via kmalloc, or by being a clone with 2676 * multiple references to the head. 2677 */ 2678 static inline bool skb_head_is_locked(const struct sk_buff *skb) 2679 { 2680 return !skb->head_frag || skb_cloned(skb); 2681 } 2682 #endif /* __KERNEL__ */ 2683 #endif /* _LINUX_SKBUFF_H */ 2684