skbuff.h 34 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/config.h>
  16. #include <linux/kernel.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/cache.h>
  20. #include <asm/atomic.h>
  21. #include <asm/types.h>
  22. #include <linux/spinlock.h>
  23. #include <linux/mm.h>
  24. #include <linux/highmem.h>
  25. #include <linux/poll.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #define HAVE_ALLOC_SKB /* For the drivers to know */
  30. #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
  31. #define SLAB_SKB /* Slabified skbuffs */
  32. #define CHECKSUM_NONE 0
  33. #define CHECKSUM_HW 1
  34. #define CHECKSUM_UNNECESSARY 2
  35. #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
  36. ~(SMP_CACHE_BYTES - 1))
  37. #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
  38. sizeof(struct skb_shared_info)) & \
  39. ~(SMP_CACHE_BYTES - 1))
  40. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  41. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  42. /* A. Checksumming of received packets by device.
  43. *
  44. * NONE: device failed to checksum this packet.
  45. * skb->csum is undefined.
  46. *
  47. * UNNECESSARY: device parsed packet and wouldbe verified checksum.
  48. * skb->csum is undefined.
  49. * It is bad option, but, unfortunately, many of vendors do this.
  50. * Apparently with secret goal to sell you new device, when you
  51. * will add new protocol to your host. F.e. IPv6. 8)
  52. *
  53. * HW: the most generic way. Device supplied checksum of _all_
  54. * the packet as seen by netif_rx in skb->csum.
  55. * NOTE: Even if device supports only some protocols, but
  56. * is able to produce some skb->csum, it MUST use HW,
  57. * not UNNECESSARY.
  58. *
  59. * B. Checksumming on output.
  60. *
  61. * NONE: skb is checksummed by protocol or csum is not required.
  62. *
  63. * HW: device is required to csum packet as seen by hard_start_xmit
  64. * from skb->h.raw to the end and to record the checksum
  65. * at skb->h.raw+skb->csum.
  66. *
  67. * Device must show its capabilities in dev->features, set
  68. * at device setup time.
  69. * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
  70. * everything.
  71. * NETIF_F_NO_CSUM - loopback or reliable single hop media.
  72. * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
  73. * TCP/UDP over IPv4. Sigh. Vendors like this
  74. * way by an unknown reason. Though, see comment above
  75. * about CHECKSUM_UNNECESSARY. 8)
  76. *
  77. * Any questions? No questions, good. --ANK
  78. */
  79. struct net_device;
  80. #ifdef CONFIG_NETFILTER
  81. struct nf_conntrack {
  82. atomic_t use;
  83. void (*destroy)(struct nf_conntrack *);
  84. };
  85. #ifdef CONFIG_BRIDGE_NETFILTER
  86. struct nf_bridge_info {
  87. atomic_t use;
  88. struct net_device *physindev;
  89. struct net_device *physoutdev;
  90. #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
  91. struct net_device *netoutdev;
  92. #endif
  93. unsigned int mask;
  94. unsigned long data[32 / sizeof(unsigned long)];
  95. };
  96. #endif
  97. #endif
  98. struct sk_buff_head {
  99. /* These two members must be first. */
  100. struct sk_buff *next;
  101. struct sk_buff *prev;
  102. __u32 qlen;
  103. spinlock_t lock;
  104. };
  105. struct sk_buff;
  106. /* To allow 64K frame to be packed as single skb without frag_list */
  107. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
  108. typedef struct skb_frag_struct skb_frag_t;
  109. struct skb_frag_struct {
  110. struct page *page;
  111. __u16 page_offset;
  112. __u16 size;
  113. };
  114. /* This data is invariant across clones and lives at
  115. * the end of the header data, ie. at skb->end.
  116. */
  117. struct skb_shared_info {
  118. atomic_t dataref;
  119. unsigned int nr_frags;
  120. unsigned short tso_size;
  121. unsigned short tso_segs;
  122. struct sk_buff *frag_list;
  123. skb_frag_t frags[MAX_SKB_FRAGS];
  124. };
  125. /* We divide dataref into two halves. The higher 16 bits hold references
  126. * to the payload part of skb->data. The lower 16 bits hold references to
  127. * the entire skb->data. It is up to the users of the skb to agree on
  128. * where the payload starts.
  129. *
  130. * All users must obey the rule that the skb->data reference count must be
  131. * greater than or equal to the payload reference count.
  132. *
  133. * Holding a reference to the payload part means that the user does not
  134. * care about modifications to the header part of skb->data.
  135. */
  136. #define SKB_DATAREF_SHIFT 16
  137. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  138. /**
  139. * struct sk_buff - socket buffer
  140. * @next: Next buffer in list
  141. * @prev: Previous buffer in list
  142. * @list: List we are on
  143. * @sk: Socket we are owned by
  144. * @stamp: Time we arrived
  145. * @dev: Device we arrived on/are leaving by
  146. * @input_dev: Device we arrived on
  147. * @real_dev: The real device we are using
  148. * @h: Transport layer header
  149. * @nh: Network layer header
  150. * @mac: Link layer header
  151. * @dst: destination entry
  152. * @sp: the security path, used for xfrm
  153. * @cb: Control buffer. Free for use by every layer. Put private vars here
  154. * @len: Length of actual data
  155. * @data_len: Data length
  156. * @mac_len: Length of link layer header
  157. * @csum: Checksum
  158. * @local_df: allow local fragmentation
  159. * @cloned: Head may be cloned (check refcnt to be sure)
  160. * @nohdr: Payload reference only, must not modify header
  161. * @pkt_type: Packet class
  162. * @ip_summed: Driver fed us an IP checksum
  163. * @priority: Packet queueing priority
  164. * @users: User count - see {datagram,tcp}.c
  165. * @protocol: Packet protocol from driver
  166. * @truesize: Buffer size
  167. * @head: Head of buffer
  168. * @data: Data head pointer
  169. * @tail: Tail pointer
  170. * @end: End pointer
  171. * @destructor: Destruct function
  172. * @nfmark: Can be used for communication between hooks
  173. * @nfcache: Cache info
  174. * @nfct: Associated connection, if any
  175. * @nfctinfo: Relationship of this skb to the connection
  176. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  177. * @private: Data which is private to the HIPPI implementation
  178. * @tc_index: Traffic control index
  179. * @tc_verd: traffic control verdict
  180. * @tc_classid: traffic control classid
  181. */
  182. struct sk_buff {
  183. /* These two members must be first. */
  184. struct sk_buff *next;
  185. struct sk_buff *prev;
  186. struct sk_buff_head *list;
  187. struct sock *sk;
  188. struct timeval stamp;
  189. struct net_device *dev;
  190. struct net_device *input_dev;
  191. struct net_device *real_dev;
  192. union {
  193. struct tcphdr *th;
  194. struct udphdr *uh;
  195. struct icmphdr *icmph;
  196. struct igmphdr *igmph;
  197. struct iphdr *ipiph;
  198. struct ipv6hdr *ipv6h;
  199. unsigned char *raw;
  200. } h;
  201. union {
  202. struct iphdr *iph;
  203. struct ipv6hdr *ipv6h;
  204. struct arphdr *arph;
  205. unsigned char *raw;
  206. } nh;
  207. union {
  208. unsigned char *raw;
  209. } mac;
  210. struct dst_entry *dst;
  211. struct sec_path *sp;
  212. /*
  213. * This is the control buffer. It is free to use for every
  214. * layer. Please put your private variables there. If you
  215. * want to keep them across layers you have to do a skb_clone()
  216. * first. This is owned by whoever has the skb queued ATM.
  217. */
  218. char cb[40];
  219. unsigned int len,
  220. data_len,
  221. mac_len,
  222. csum;
  223. __u32 priority;
  224. __u8 local_df:1,
  225. cloned:1,
  226. ip_summed:2,
  227. nohdr:1;
  228. /* 3 bits spare */
  229. __u8 pkt_type;
  230. __u16 protocol;
  231. void (*destructor)(struct sk_buff *skb);
  232. #ifdef CONFIG_NETFILTER
  233. unsigned long nfmark;
  234. __u32 nfcache;
  235. __u32 nfctinfo;
  236. struct nf_conntrack *nfct;
  237. #ifdef CONFIG_BRIDGE_NETFILTER
  238. struct nf_bridge_info *nf_bridge;
  239. #endif
  240. #endif /* CONFIG_NETFILTER */
  241. #if defined(CONFIG_HIPPI)
  242. union {
  243. __u32 ifield;
  244. } private;
  245. #endif
  246. #ifdef CONFIG_NET_SCHED
  247. __u32 tc_index; /* traffic control index */
  248. #ifdef CONFIG_NET_CLS_ACT
  249. __u32 tc_verd; /* traffic control verdict */
  250. __u32 tc_classid; /* traffic control classid */
  251. #endif
  252. #endif
  253. /* These elements must be at the end, see alloc_skb() for details. */
  254. unsigned int truesize;
  255. atomic_t users;
  256. unsigned char *head,
  257. *data,
  258. *tail,
  259. *end;
  260. };
  261. #ifdef __KERNEL__
  262. /*
  263. * Handling routines are only of interest to the kernel
  264. */
  265. #include <linux/slab.h>
  266. #include <asm/system.h>
  267. extern void __kfree_skb(struct sk_buff *skb);
  268. extern struct sk_buff *alloc_skb(unsigned int size, int priority);
  269. extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
  270. unsigned int size, int priority);
  271. extern void kfree_skbmem(struct sk_buff *skb);
  272. extern struct sk_buff *skb_clone(struct sk_buff *skb, int priority);
  273. extern struct sk_buff *skb_copy(const struct sk_buff *skb, int priority);
  274. extern struct sk_buff *pskb_copy(struct sk_buff *skb, int gfp_mask);
  275. extern int pskb_expand_head(struct sk_buff *skb,
  276. int nhead, int ntail, int gfp_mask);
  277. extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  278. unsigned int headroom);
  279. extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
  280. int newheadroom, int newtailroom,
  281. int priority);
  282. extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
  283. #define dev_kfree_skb(a) kfree_skb(a)
  284. extern void skb_over_panic(struct sk_buff *skb, int len,
  285. void *here);
  286. extern void skb_under_panic(struct sk_buff *skb, int len,
  287. void *here);
  288. struct skb_seq_state
  289. {
  290. __u32 lower_offset;
  291. __u32 upper_offset;
  292. __u32 frag_idx;
  293. __u32 stepped_offset;
  294. struct sk_buff *root_skb;
  295. struct sk_buff *cur_skb;
  296. __u8 *frag_data;
  297. };
  298. extern void skb_prepare_seq_read(struct sk_buff *skb,
  299. unsigned int from, unsigned int to,
  300. struct skb_seq_state *st);
  301. extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  302. struct skb_seq_state *st);
  303. extern void skb_abort_seq_read(struct skb_seq_state *st);
  304. extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  305. unsigned int to, struct ts_config *config,
  306. struct ts_state *state);
  307. /* Internal */
  308. #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
  309. /**
  310. * skb_queue_empty - check if a queue is empty
  311. * @list: queue head
  312. *
  313. * Returns true if the queue is empty, false otherwise.
  314. */
  315. static inline int skb_queue_empty(const struct sk_buff_head *list)
  316. {
  317. return list->next == (struct sk_buff *)list;
  318. }
  319. /**
  320. * skb_get - reference buffer
  321. * @skb: buffer to reference
  322. *
  323. * Makes another reference to a socket buffer and returns a pointer
  324. * to the buffer.
  325. */
  326. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  327. {
  328. atomic_inc(&skb->users);
  329. return skb;
  330. }
  331. /*
  332. * If users == 1, we are the only owner and are can avoid redundant
  333. * atomic change.
  334. */
  335. /**
  336. * kfree_skb - free an sk_buff
  337. * @skb: buffer to free
  338. *
  339. * Drop a reference to the buffer and free it if the usage count has
  340. * hit zero.
  341. */
  342. static inline void kfree_skb(struct sk_buff *skb)
  343. {
  344. if (likely(atomic_read(&skb->users) == 1))
  345. smp_rmb();
  346. else if (likely(!atomic_dec_and_test(&skb->users)))
  347. return;
  348. __kfree_skb(skb);
  349. }
  350. /**
  351. * skb_cloned - is the buffer a clone
  352. * @skb: buffer to check
  353. *
  354. * Returns true if the buffer was generated with skb_clone() and is
  355. * one of multiple shared copies of the buffer. Cloned buffers are
  356. * shared data so must not be written to under normal circumstances.
  357. */
  358. static inline int skb_cloned(const struct sk_buff *skb)
  359. {
  360. return skb->cloned &&
  361. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  362. }
  363. /**
  364. * skb_header_cloned - is the header a clone
  365. * @skb: buffer to check
  366. *
  367. * Returns true if modifying the header part of the buffer requires
  368. * the data to be copied.
  369. */
  370. static inline int skb_header_cloned(const struct sk_buff *skb)
  371. {
  372. int dataref;
  373. if (!skb->cloned)
  374. return 0;
  375. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  376. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  377. return dataref != 1;
  378. }
  379. /**
  380. * skb_header_release - release reference to header
  381. * @skb: buffer to operate on
  382. *
  383. * Drop a reference to the header part of the buffer. This is done
  384. * by acquiring a payload reference. You must not read from the header
  385. * part of skb->data after this.
  386. */
  387. static inline void skb_header_release(struct sk_buff *skb)
  388. {
  389. BUG_ON(skb->nohdr);
  390. skb->nohdr = 1;
  391. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  392. }
  393. /**
  394. * skb_shared - is the buffer shared
  395. * @skb: buffer to check
  396. *
  397. * Returns true if more than one person has a reference to this
  398. * buffer.
  399. */
  400. static inline int skb_shared(const struct sk_buff *skb)
  401. {
  402. return atomic_read(&skb->users) != 1;
  403. }
  404. /**
  405. * skb_share_check - check if buffer is shared and if so clone it
  406. * @skb: buffer to check
  407. * @pri: priority for memory allocation
  408. *
  409. * If the buffer is shared the buffer is cloned and the old copy
  410. * drops a reference. A new clone with a single reference is returned.
  411. * If the buffer is not shared the original buffer is returned. When
  412. * being called from interrupt status or with spinlocks held pri must
  413. * be GFP_ATOMIC.
  414. *
  415. * NULL is returned on a memory allocation failure.
  416. */
  417. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, int pri)
  418. {
  419. might_sleep_if(pri & __GFP_WAIT);
  420. if (skb_shared(skb)) {
  421. struct sk_buff *nskb = skb_clone(skb, pri);
  422. kfree_skb(skb);
  423. skb = nskb;
  424. }
  425. return skb;
  426. }
  427. /*
  428. * Copy shared buffers into a new sk_buff. We effectively do COW on
  429. * packets to handle cases where we have a local reader and forward
  430. * and a couple of other messy ones. The normal one is tcpdumping
  431. * a packet thats being forwarded.
  432. */
  433. /**
  434. * skb_unshare - make a copy of a shared buffer
  435. * @skb: buffer to check
  436. * @pri: priority for memory allocation
  437. *
  438. * If the socket buffer is a clone then this function creates a new
  439. * copy of the data, drops a reference count on the old copy and returns
  440. * the new copy with the reference count at 1. If the buffer is not a clone
  441. * the original buffer is returned. When called with a spinlock held or
  442. * from interrupt state @pri must be %GFP_ATOMIC
  443. *
  444. * %NULL is returned on a memory allocation failure.
  445. */
  446. static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
  447. {
  448. might_sleep_if(pri & __GFP_WAIT);
  449. if (skb_cloned(skb)) {
  450. struct sk_buff *nskb = skb_copy(skb, pri);
  451. kfree_skb(skb); /* Free our shared copy */
  452. skb = nskb;
  453. }
  454. return skb;
  455. }
  456. /**
  457. * skb_peek
  458. * @list_: list to peek at
  459. *
  460. * Peek an &sk_buff. Unlike most other operations you _MUST_
  461. * be careful with this one. A peek leaves the buffer on the
  462. * list and someone else may run off with it. You must hold
  463. * the appropriate locks or have a private queue to do this.
  464. *
  465. * Returns %NULL for an empty list or a pointer to the head element.
  466. * The reference count is not incremented and the reference is therefore
  467. * volatile. Use with caution.
  468. */
  469. static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
  470. {
  471. struct sk_buff *list = ((struct sk_buff *)list_)->next;
  472. if (list == (struct sk_buff *)list_)
  473. list = NULL;
  474. return list;
  475. }
  476. /**
  477. * skb_peek_tail
  478. * @list_: list to peek at
  479. *
  480. * Peek an &sk_buff. Unlike most other operations you _MUST_
  481. * be careful with this one. A peek leaves the buffer on the
  482. * list and someone else may run off with it. You must hold
  483. * the appropriate locks or have a private queue to do this.
  484. *
  485. * Returns %NULL for an empty list or a pointer to the tail element.
  486. * The reference count is not incremented and the reference is therefore
  487. * volatile. Use with caution.
  488. */
  489. static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
  490. {
  491. struct sk_buff *list = ((struct sk_buff *)list_)->prev;
  492. if (list == (struct sk_buff *)list_)
  493. list = NULL;
  494. return list;
  495. }
  496. /**
  497. * skb_queue_len - get queue length
  498. * @list_: list to measure
  499. *
  500. * Return the length of an &sk_buff queue.
  501. */
  502. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  503. {
  504. return list_->qlen;
  505. }
  506. static inline void skb_queue_head_init(struct sk_buff_head *list)
  507. {
  508. spin_lock_init(&list->lock);
  509. list->prev = list->next = (struct sk_buff *)list;
  510. list->qlen = 0;
  511. }
  512. /*
  513. * Insert an sk_buff at the start of a list.
  514. *
  515. * The "__skb_xxxx()" functions are the non-atomic ones that
  516. * can only be called with interrupts disabled.
  517. */
  518. /**
  519. * __skb_queue_head - queue a buffer at the list head
  520. * @list: list to use
  521. * @newsk: buffer to queue
  522. *
  523. * Queue a buffer at the start of a list. This function takes no locks
  524. * and you must therefore hold required locks before calling it.
  525. *
  526. * A buffer cannot be placed on two lists at the same time.
  527. */
  528. extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  529. static inline void __skb_queue_head(struct sk_buff_head *list,
  530. struct sk_buff *newsk)
  531. {
  532. struct sk_buff *prev, *next;
  533. newsk->list = list;
  534. list->qlen++;
  535. prev = (struct sk_buff *)list;
  536. next = prev->next;
  537. newsk->next = next;
  538. newsk->prev = prev;
  539. next->prev = prev->next = newsk;
  540. }
  541. /**
  542. * __skb_queue_tail - queue a buffer at the list tail
  543. * @list: list to use
  544. * @newsk: buffer to queue
  545. *
  546. * Queue a buffer at the end of a list. This function takes no locks
  547. * and you must therefore hold required locks before calling it.
  548. *
  549. * A buffer cannot be placed on two lists at the same time.
  550. */
  551. extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  552. static inline void __skb_queue_tail(struct sk_buff_head *list,
  553. struct sk_buff *newsk)
  554. {
  555. struct sk_buff *prev, *next;
  556. newsk->list = list;
  557. list->qlen++;
  558. next = (struct sk_buff *)list;
  559. prev = next->prev;
  560. newsk->next = next;
  561. newsk->prev = prev;
  562. next->prev = prev->next = newsk;
  563. }
  564. /**
  565. * __skb_dequeue - remove from the head of the queue
  566. * @list: list to dequeue from
  567. *
  568. * Remove the head of the list. This function does not take any locks
  569. * so must be used with appropriate locks held only. The head item is
  570. * returned or %NULL if the list is empty.
  571. */
  572. extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  573. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  574. {
  575. struct sk_buff *next, *prev, *result;
  576. prev = (struct sk_buff *) list;
  577. next = prev->next;
  578. result = NULL;
  579. if (next != prev) {
  580. result = next;
  581. next = next->next;
  582. list->qlen--;
  583. next->prev = prev;
  584. prev->next = next;
  585. result->next = result->prev = NULL;
  586. result->list = NULL;
  587. }
  588. return result;
  589. }
  590. /*
  591. * Insert a packet on a list.
  592. */
  593. extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
  594. static inline void __skb_insert(struct sk_buff *newsk,
  595. struct sk_buff *prev, struct sk_buff *next,
  596. struct sk_buff_head *list)
  597. {
  598. newsk->next = next;
  599. newsk->prev = prev;
  600. next->prev = prev->next = newsk;
  601. newsk->list = list;
  602. list->qlen++;
  603. }
  604. /*
  605. * Place a packet after a given packet in a list.
  606. */
  607. extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
  608. static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
  609. {
  610. __skb_insert(newsk, old, old->next, old->list);
  611. }
  612. /*
  613. * remove sk_buff from list. _Must_ be called atomically, and with
  614. * the list known..
  615. */
  616. extern void skb_unlink(struct sk_buff *skb);
  617. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  618. {
  619. struct sk_buff *next, *prev;
  620. list->qlen--;
  621. next = skb->next;
  622. prev = skb->prev;
  623. skb->next = skb->prev = NULL;
  624. skb->list = NULL;
  625. next->prev = prev;
  626. prev->next = next;
  627. }
  628. /* XXX: more streamlined implementation */
  629. /**
  630. * __skb_dequeue_tail - remove from the tail of the queue
  631. * @list: list to dequeue from
  632. *
  633. * Remove the tail of the list. This function does not take any locks
  634. * so must be used with appropriate locks held only. The tail item is
  635. * returned or %NULL if the list is empty.
  636. */
  637. extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  638. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  639. {
  640. struct sk_buff *skb = skb_peek_tail(list);
  641. if (skb)
  642. __skb_unlink(skb, list);
  643. return skb;
  644. }
  645. static inline int skb_is_nonlinear(const struct sk_buff *skb)
  646. {
  647. return skb->data_len;
  648. }
  649. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  650. {
  651. return skb->len - skb->data_len;
  652. }
  653. static inline int skb_pagelen(const struct sk_buff *skb)
  654. {
  655. int i, len = 0;
  656. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  657. len += skb_shinfo(skb)->frags[i].size;
  658. return len + skb_headlen(skb);
  659. }
  660. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  661. struct page *page, int off, int size)
  662. {
  663. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  664. frag->page = page;
  665. frag->page_offset = off;
  666. frag->size = size;
  667. skb_shinfo(skb)->nr_frags = i + 1;
  668. }
  669. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  670. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
  671. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  672. /*
  673. * Add data to an sk_buff
  674. */
  675. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  676. {
  677. unsigned char *tmp = skb->tail;
  678. SKB_LINEAR_ASSERT(skb);
  679. skb->tail += len;
  680. skb->len += len;
  681. return tmp;
  682. }
  683. /**
  684. * skb_put - add data to a buffer
  685. * @skb: buffer to use
  686. * @len: amount of data to add
  687. *
  688. * This function extends the used data area of the buffer. If this would
  689. * exceed the total buffer size the kernel will panic. A pointer to the
  690. * first byte of the extra data is returned.
  691. */
  692. static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
  693. {
  694. unsigned char *tmp = skb->tail;
  695. SKB_LINEAR_ASSERT(skb);
  696. skb->tail += len;
  697. skb->len += len;
  698. if (unlikely(skb->tail>skb->end))
  699. skb_over_panic(skb, len, current_text_addr());
  700. return tmp;
  701. }
  702. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  703. {
  704. skb->data -= len;
  705. skb->len += len;
  706. return skb->data;
  707. }
  708. /**
  709. * skb_push - add data to the start of a buffer
  710. * @skb: buffer to use
  711. * @len: amount of data to add
  712. *
  713. * This function extends the used data area of the buffer at the buffer
  714. * start. If this would exceed the total buffer headroom the kernel will
  715. * panic. A pointer to the first byte of the extra data is returned.
  716. */
  717. static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
  718. {
  719. skb->data -= len;
  720. skb->len += len;
  721. if (unlikely(skb->data<skb->head))
  722. skb_under_panic(skb, len, current_text_addr());
  723. return skb->data;
  724. }
  725. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  726. {
  727. skb->len -= len;
  728. BUG_ON(skb->len < skb->data_len);
  729. return skb->data += len;
  730. }
  731. /**
  732. * skb_pull - remove data from the start of a buffer
  733. * @skb: buffer to use
  734. * @len: amount of data to remove
  735. *
  736. * This function removes data from the start of a buffer, returning
  737. * the memory to the headroom. A pointer to the next data in the buffer
  738. * is returned. Once the data has been pulled future pushes will overwrite
  739. * the old data.
  740. */
  741. static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
  742. {
  743. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  744. }
  745. extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  746. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  747. {
  748. if (len > skb_headlen(skb) &&
  749. !__pskb_pull_tail(skb, len-skb_headlen(skb)))
  750. return NULL;
  751. skb->len -= len;
  752. return skb->data += len;
  753. }
  754. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  755. {
  756. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  757. }
  758. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  759. {
  760. if (likely(len <= skb_headlen(skb)))
  761. return 1;
  762. if (unlikely(len > skb->len))
  763. return 0;
  764. return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
  765. }
  766. /**
  767. * skb_headroom - bytes at buffer head
  768. * @skb: buffer to check
  769. *
  770. * Return the number of bytes of free space at the head of an &sk_buff.
  771. */
  772. static inline int skb_headroom(const struct sk_buff *skb)
  773. {
  774. return skb->data - skb->head;
  775. }
  776. /**
  777. * skb_tailroom - bytes at buffer end
  778. * @skb: buffer to check
  779. *
  780. * Return the number of bytes of free space at the tail of an sk_buff
  781. */
  782. static inline int skb_tailroom(const struct sk_buff *skb)
  783. {
  784. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  785. }
  786. /**
  787. * skb_reserve - adjust headroom
  788. * @skb: buffer to alter
  789. * @len: bytes to move
  790. *
  791. * Increase the headroom of an empty &sk_buff by reducing the tail
  792. * room. This is only allowed for an empty buffer.
  793. */
  794. static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
  795. {
  796. skb->data += len;
  797. skb->tail += len;
  798. }
  799. /*
  800. * CPUs often take a performance hit when accessing unaligned memory
  801. * locations. The actual performance hit varies, it can be small if the
  802. * hardware handles it or large if we have to take an exception and fix it
  803. * in software.
  804. *
  805. * Since an ethernet header is 14 bytes network drivers often end up with
  806. * the IP header at an unaligned offset. The IP header can be aligned by
  807. * shifting the start of the packet by 2 bytes. Drivers should do this
  808. * with:
  809. *
  810. * skb_reserve(NET_IP_ALIGN);
  811. *
  812. * The downside to this alignment of the IP header is that the DMA is now
  813. * unaligned. On some architectures the cost of an unaligned DMA is high
  814. * and this cost outweighs the gains made by aligning the IP header.
  815. *
  816. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  817. * to be overridden.
  818. */
  819. #ifndef NET_IP_ALIGN
  820. #define NET_IP_ALIGN 2
  821. #endif
  822. extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
  823. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  824. {
  825. if (!skb->data_len) {
  826. skb->len = len;
  827. skb->tail = skb->data + len;
  828. } else
  829. ___pskb_trim(skb, len, 0);
  830. }
  831. /**
  832. * skb_trim - remove end from a buffer
  833. * @skb: buffer to alter
  834. * @len: new length
  835. *
  836. * Cut the length of a buffer down by removing data from the tail. If
  837. * the buffer is already under the length specified it is not modified.
  838. */
  839. static inline void skb_trim(struct sk_buff *skb, unsigned int len)
  840. {
  841. if (skb->len > len)
  842. __skb_trim(skb, len);
  843. }
  844. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  845. {
  846. if (!skb->data_len) {
  847. skb->len = len;
  848. skb->tail = skb->data+len;
  849. return 0;
  850. }
  851. return ___pskb_trim(skb, len, 1);
  852. }
  853. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  854. {
  855. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  856. }
  857. /**
  858. * skb_orphan - orphan a buffer
  859. * @skb: buffer to orphan
  860. *
  861. * If a buffer currently has an owner then we call the owner's
  862. * destructor function and make the @skb unowned. The buffer continues
  863. * to exist but is no longer charged to its former owner.
  864. */
  865. static inline void skb_orphan(struct sk_buff *skb)
  866. {
  867. if (skb->destructor)
  868. skb->destructor(skb);
  869. skb->destructor = NULL;
  870. skb->sk = NULL;
  871. }
  872. /**
  873. * __skb_queue_purge - empty a list
  874. * @list: list to empty
  875. *
  876. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  877. * the list and one reference dropped. This function does not take the
  878. * list lock and the caller must hold the relevant locks to use it.
  879. */
  880. extern void skb_queue_purge(struct sk_buff_head *list);
  881. static inline void __skb_queue_purge(struct sk_buff_head *list)
  882. {
  883. struct sk_buff *skb;
  884. while ((skb = __skb_dequeue(list)) != NULL)
  885. kfree_skb(skb);
  886. }
  887. #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
  888. /**
  889. * __dev_alloc_skb - allocate an skbuff for sending
  890. * @length: length to allocate
  891. * @gfp_mask: get_free_pages mask, passed to alloc_skb
  892. *
  893. * Allocate a new &sk_buff and assign it a usage count of one. The
  894. * buffer has unspecified headroom built in. Users should allocate
  895. * the headroom they think they need without accounting for the
  896. * built in space. The built in space is used for optimisations.
  897. *
  898. * %NULL is returned in there is no free memory.
  899. */
  900. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  901. int gfp_mask)
  902. {
  903. struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
  904. if (likely(skb))
  905. skb_reserve(skb, 16);
  906. return skb;
  907. }
  908. #else
  909. extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
  910. #endif
  911. /**
  912. * dev_alloc_skb - allocate an skbuff for sending
  913. * @length: length to allocate
  914. *
  915. * Allocate a new &sk_buff and assign it a usage count of one. The
  916. * buffer has unspecified headroom built in. Users should allocate
  917. * the headroom they think they need without accounting for the
  918. * built in space. The built in space is used for optimisations.
  919. *
  920. * %NULL is returned in there is no free memory. Although this function
  921. * allocates memory it can be called from an interrupt.
  922. */
  923. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  924. {
  925. return __dev_alloc_skb(length, GFP_ATOMIC);
  926. }
  927. /**
  928. * skb_cow - copy header of skb when it is required
  929. * @skb: buffer to cow
  930. * @headroom: needed headroom
  931. *
  932. * If the skb passed lacks sufficient headroom or its data part
  933. * is shared, data is reallocated. If reallocation fails, an error
  934. * is returned and original skb is not changed.
  935. *
  936. * The result is skb with writable area skb->head...skb->tail
  937. * and at least @headroom of space at head.
  938. */
  939. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  940. {
  941. int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
  942. if (delta < 0)
  943. delta = 0;
  944. if (delta || skb_cloned(skb))
  945. return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
  946. return 0;
  947. }
  948. /**
  949. * skb_padto - pad an skbuff up to a minimal size
  950. * @skb: buffer to pad
  951. * @len: minimal length
  952. *
  953. * Pads up a buffer to ensure the trailing bytes exist and are
  954. * blanked. If the buffer already contains sufficient data it
  955. * is untouched. Returns the buffer, which may be a replacement
  956. * for the original, or NULL for out of memory - in which case
  957. * the original buffer is still freed.
  958. */
  959. static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
  960. {
  961. unsigned int size = skb->len;
  962. if (likely(size >= len))
  963. return skb;
  964. return skb_pad(skb, len-size);
  965. }
  966. static inline int skb_add_data(struct sk_buff *skb,
  967. char __user *from, int copy)
  968. {
  969. const int off = skb->len;
  970. if (skb->ip_summed == CHECKSUM_NONE) {
  971. int err = 0;
  972. unsigned int csum = csum_and_copy_from_user(from,
  973. skb_put(skb, copy),
  974. copy, 0, &err);
  975. if (!err) {
  976. skb->csum = csum_block_add(skb->csum, csum, off);
  977. return 0;
  978. }
  979. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  980. return 0;
  981. __skb_trim(skb, off);
  982. return -EFAULT;
  983. }
  984. static inline int skb_can_coalesce(struct sk_buff *skb, int i,
  985. struct page *page, int off)
  986. {
  987. if (i) {
  988. struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  989. return page == frag->page &&
  990. off == frag->page_offset + frag->size;
  991. }
  992. return 0;
  993. }
  994. /**
  995. * skb_linearize - convert paged skb to linear one
  996. * @skb: buffer to linarize
  997. * @gfp: allocation mode
  998. *
  999. * If there is no free memory -ENOMEM is returned, otherwise zero
  1000. * is returned and the old skb data released.
  1001. */
  1002. extern int __skb_linearize(struct sk_buff *skb, int gfp);
  1003. static inline int skb_linearize(struct sk_buff *skb, int gfp)
  1004. {
  1005. return __skb_linearize(skb, gfp);
  1006. }
  1007. /**
  1008. * skb_postpull_rcsum - update checksum for received skb after pull
  1009. * @skb: buffer to update
  1010. * @start: start of data before pull
  1011. * @len: length of data pulled
  1012. *
  1013. * After doing a pull on a received packet, you need to call this to
  1014. * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
  1015. * so that it can be recomputed from scratch.
  1016. */
  1017. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  1018. const void *start, int len)
  1019. {
  1020. if (skb->ip_summed == CHECKSUM_HW)
  1021. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  1022. }
  1023. /**
  1024. * pskb_trim_rcsum - trim received skb and update checksum
  1025. * @skb: buffer to trim
  1026. * @len: new length
  1027. *
  1028. * This is exactly the same as pskb_trim except that it ensures the
  1029. * checksum of received packets are still valid after the operation.
  1030. */
  1031. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  1032. {
  1033. if (len >= skb->len)
  1034. return 0;
  1035. if (skb->ip_summed == CHECKSUM_HW)
  1036. skb->ip_summed = CHECKSUM_NONE;
  1037. return __pskb_trim(skb, len);
  1038. }
  1039. static inline void *kmap_skb_frag(const skb_frag_t *frag)
  1040. {
  1041. #ifdef CONFIG_HIGHMEM
  1042. BUG_ON(in_irq());
  1043. local_bh_disable();
  1044. #endif
  1045. return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
  1046. }
  1047. static inline void kunmap_skb_frag(void *vaddr)
  1048. {
  1049. kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
  1050. #ifdef CONFIG_HIGHMEM
  1051. local_bh_enable();
  1052. #endif
  1053. }
  1054. #define skb_queue_walk(queue, skb) \
  1055. for (skb = (queue)->next; \
  1056. prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
  1057. skb = skb->next)
  1058. extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
  1059. int noblock, int *err);
  1060. extern unsigned int datagram_poll(struct file *file, struct socket *sock,
  1061. struct poll_table_struct *wait);
  1062. extern int skb_copy_datagram_iovec(const struct sk_buff *from,
  1063. int offset, struct iovec *to,
  1064. int size);
  1065. extern int skb_copy_and_csum_datagram_iovec(const
  1066. struct sk_buff *skb,
  1067. int hlen,
  1068. struct iovec *iov);
  1069. extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  1070. extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
  1071. int len, unsigned int csum);
  1072. extern int skb_copy_bits(const struct sk_buff *skb, int offset,
  1073. void *to, int len);
  1074. extern int skb_store_bits(const struct sk_buff *skb, int offset,
  1075. void *from, int len);
  1076. extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
  1077. int offset, u8 *to, int len,
  1078. unsigned int csum);
  1079. extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  1080. extern void skb_split(struct sk_buff *skb,
  1081. struct sk_buff *skb1, const u32 len);
  1082. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  1083. int len, void *buffer)
  1084. {
  1085. int hlen = skb_headlen(skb);
  1086. if (hlen - offset >= len)
  1087. return skb->data + offset;
  1088. if (skb_copy_bits(skb, offset, buffer, len) < 0)
  1089. return NULL;
  1090. return buffer;
  1091. }
  1092. extern void skb_init(void);
  1093. extern void skb_add_mtu(int mtu);
  1094. #ifdef CONFIG_NETFILTER
  1095. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  1096. {
  1097. if (nfct && atomic_dec_and_test(&nfct->use))
  1098. nfct->destroy(nfct);
  1099. }
  1100. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  1101. {
  1102. if (nfct)
  1103. atomic_inc(&nfct->use);
  1104. }
  1105. static inline void nf_reset(struct sk_buff *skb)
  1106. {
  1107. nf_conntrack_put(skb->nfct);
  1108. skb->nfct = NULL;
  1109. }
  1110. #ifdef CONFIG_BRIDGE_NETFILTER
  1111. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  1112. {
  1113. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  1114. kfree(nf_bridge);
  1115. }
  1116. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  1117. {
  1118. if (nf_bridge)
  1119. atomic_inc(&nf_bridge->use);
  1120. }
  1121. #endif /* CONFIG_BRIDGE_NETFILTER */
  1122. #else /* CONFIG_NETFILTER */
  1123. static inline void nf_reset(struct sk_buff *skb) {}
  1124. #endif /* CONFIG_NETFILTER */
  1125. #endif /* __KERNEL__ */
  1126. #endif /* _LINUX_SKBUFF_H */