skbuff.h 34 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250
  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 <net/checksum.h>
  28. #define HAVE_ALLOC_SKB /* For the drivers to know */
  29. #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
  30. #define SLAB_SKB /* Slabified skbuffs */
  31. #define CHECKSUM_NONE 0
  32. #define CHECKSUM_HW 1
  33. #define CHECKSUM_UNNECESSARY 2
  34. #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
  35. ~(SMP_CACHE_BYTES - 1))
  36. #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
  37. sizeof(struct skb_shared_info)) & \
  38. ~(SMP_CACHE_BYTES - 1))
  39. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  40. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  41. /* A. Checksumming of received packets by device.
  42. *
  43. * NONE: device failed to checksum this packet.
  44. * skb->csum is undefined.
  45. *
  46. * UNNECESSARY: device parsed packet and wouldbe verified checksum.
  47. * skb->csum is undefined.
  48. * It is bad option, but, unfortunately, many of vendors do this.
  49. * Apparently with secret goal to sell you new device, when you
  50. * will add new protocol to your host. F.e. IPv6. 8)
  51. *
  52. * HW: the most generic way. Device supplied checksum of _all_
  53. * the packet as seen by netif_rx in skb->csum.
  54. * NOTE: Even if device supports only some protocols, but
  55. * is able to produce some skb->csum, it MUST use HW,
  56. * not UNNECESSARY.
  57. *
  58. * B. Checksumming on output.
  59. *
  60. * NONE: skb is checksummed by protocol or csum is not required.
  61. *
  62. * HW: device is required to csum packet as seen by hard_start_xmit
  63. * from skb->h.raw to the end and to record the checksum
  64. * at skb->h.raw+skb->csum.
  65. *
  66. * Device must show its capabilities in dev->features, set
  67. * at device setup time.
  68. * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
  69. * everything.
  70. * NETIF_F_NO_CSUM - loopback or reliable single hop media.
  71. * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
  72. * TCP/UDP over IPv4. Sigh. Vendors like this
  73. * way by an unknown reason. Though, see comment above
  74. * about CHECKSUM_UNNECESSARY. 8)
  75. *
  76. * Any questions? No questions, good. --ANK
  77. */
  78. struct net_device;
  79. #ifdef CONFIG_NETFILTER
  80. struct nf_conntrack {
  81. atomic_t use;
  82. void (*destroy)(struct nf_conntrack *);
  83. };
  84. #ifdef CONFIG_BRIDGE_NETFILTER
  85. struct nf_bridge_info {
  86. atomic_t use;
  87. struct net_device *physindev;
  88. struct net_device *physoutdev;
  89. #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
  90. struct net_device *netoutdev;
  91. #endif
  92. unsigned int mask;
  93. unsigned long data[32 / sizeof(unsigned long)];
  94. };
  95. #endif
  96. #endif
  97. struct sk_buff_head {
  98. /* These two members must be first. */
  99. struct sk_buff *next;
  100. struct sk_buff *prev;
  101. __u32 qlen;
  102. spinlock_t lock;
  103. };
  104. struct sk_buff;
  105. /* To allow 64K frame to be packed as single skb without frag_list */
  106. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
  107. typedef struct skb_frag_struct skb_frag_t;
  108. struct skb_frag_struct {
  109. struct page *page;
  110. __u16 page_offset;
  111. __u16 size;
  112. };
  113. /* This data is invariant across clones and lives at
  114. * the end of the header data, ie. at skb->end.
  115. */
  116. struct skb_shared_info {
  117. atomic_t dataref;
  118. unsigned int nr_frags;
  119. unsigned short tso_size;
  120. unsigned short tso_segs;
  121. struct sk_buff *frag_list;
  122. skb_frag_t frags[MAX_SKB_FRAGS];
  123. };
  124. /* We divide dataref into two halves. The higher 16 bits hold references
  125. * to the payload part of skb->data. The lower 16 bits hold references to
  126. * the entire skb->data. It is up to the users of the skb to agree on
  127. * where the payload starts.
  128. *
  129. * All users must obey the rule that the skb->data reference count must be
  130. * greater than or equal to the payload reference count.
  131. *
  132. * Holding a reference to the payload part means that the user does not
  133. * care about modifications to the header part of skb->data.
  134. */
  135. #define SKB_DATAREF_SHIFT 16
  136. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  137. /**
  138. * struct sk_buff - socket buffer
  139. * @next: Next buffer in list
  140. * @prev: Previous buffer in list
  141. * @list: List we are on
  142. * @sk: Socket we are owned by
  143. * @stamp: Time we arrived
  144. * @dev: Device we arrived on/are leaving by
  145. * @input_dev: Device we arrived on
  146. * @real_dev: The real device we are using
  147. * @h: Transport layer header
  148. * @nh: Network layer header
  149. * @mac: Link layer header
  150. * @dst: destination entry
  151. * @sp: the security path, used for xfrm
  152. * @cb: Control buffer. Free for use by every layer. Put private vars here
  153. * @len: Length of actual data
  154. * @data_len: Data length
  155. * @mac_len: Length of link layer header
  156. * @csum: Checksum
  157. * @local_df: allow local fragmentation
  158. * @cloned: Head may be cloned (check refcnt to be sure)
  159. * @nohdr: Payload reference only, must not modify header
  160. * @pkt_type: Packet class
  161. * @ip_summed: Driver fed us an IP checksum
  162. * @priority: Packet queueing priority
  163. * @users: User count - see {datagram,tcp}.c
  164. * @protocol: Packet protocol from driver
  165. * @security: Security level of packet
  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_debug: Netfilter debugging
  177. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  178. * @private: Data which is private to the HIPPI implementation
  179. * @tc_index: Traffic control index
  180. * @tc_verd: traffic control verdict
  181. * @tc_classid: traffic control classid
  182. */
  183. struct sk_buff {
  184. /* These two members must be first. */
  185. struct sk_buff *next;
  186. struct sk_buff *prev;
  187. struct sk_buff_head *list;
  188. struct sock *sk;
  189. struct timeval stamp;
  190. struct net_device *dev;
  191. struct net_device *input_dev;
  192. struct net_device *real_dev;
  193. union {
  194. struct tcphdr *th;
  195. struct udphdr *uh;
  196. struct icmphdr *icmph;
  197. struct igmphdr *igmph;
  198. struct iphdr *ipiph;
  199. struct ipv6hdr *ipv6h;
  200. unsigned char *raw;
  201. } h;
  202. union {
  203. struct iphdr *iph;
  204. struct ipv6hdr *ipv6h;
  205. struct arphdr *arph;
  206. unsigned char *raw;
  207. } nh;
  208. union {
  209. unsigned char *raw;
  210. } mac;
  211. struct dst_entry *dst;
  212. struct sec_path *sp;
  213. /*
  214. * This is the control buffer. It is free to use for every
  215. * layer. Please put your private variables there. If you
  216. * want to keep them across layers you have to do a skb_clone()
  217. * first. This is owned by whoever has the skb queued ATM.
  218. */
  219. char cb[40];
  220. unsigned int len,
  221. data_len,
  222. mac_len,
  223. csum;
  224. unsigned char local_df,
  225. cloned:1,
  226. nohdr:1,
  227. pkt_type,
  228. ip_summed;
  229. __u32 priority;
  230. unsigned short protocol,
  231. security;
  232. void (*destructor)(struct sk_buff *skb);
  233. #ifdef CONFIG_NETFILTER
  234. unsigned long nfmark;
  235. __u32 nfcache;
  236. __u32 nfctinfo;
  237. struct nf_conntrack *nfct;
  238. #ifdef CONFIG_NETFILTER_DEBUG
  239. unsigned int nf_debug;
  240. #endif
  241. #ifdef CONFIG_BRIDGE_NETFILTER
  242. struct nf_bridge_info *nf_bridge;
  243. #endif
  244. #endif /* CONFIG_NETFILTER */
  245. #if defined(CONFIG_HIPPI)
  246. union {
  247. __u32 ifield;
  248. } private;
  249. #endif
  250. #ifdef CONFIG_NET_SCHED
  251. __u32 tc_index; /* traffic control index */
  252. #ifdef CONFIG_NET_CLS_ACT
  253. __u32 tc_verd; /* traffic control verdict */
  254. __u32 tc_classid; /* traffic control classid */
  255. #endif
  256. #endif
  257. /* These elements must be at the end, see alloc_skb() for details. */
  258. unsigned int truesize;
  259. atomic_t users;
  260. unsigned char *head,
  261. *data,
  262. *tail,
  263. *end;
  264. };
  265. #ifdef __KERNEL__
  266. /*
  267. * Handling routines are only of interest to the kernel
  268. */
  269. #include <linux/slab.h>
  270. #include <asm/system.h>
  271. extern void __kfree_skb(struct sk_buff *skb);
  272. extern struct sk_buff *alloc_skb(unsigned int size, int priority);
  273. extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
  274. unsigned int size, int priority);
  275. extern void kfree_skbmem(struct sk_buff *skb);
  276. extern struct sk_buff *skb_clone(struct sk_buff *skb, int priority);
  277. extern struct sk_buff *skb_copy(const struct sk_buff *skb, int priority);
  278. extern struct sk_buff *pskb_copy(struct sk_buff *skb, int gfp_mask);
  279. extern int pskb_expand_head(struct sk_buff *skb,
  280. int nhead, int ntail, int gfp_mask);
  281. extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  282. unsigned int headroom);
  283. extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
  284. int newheadroom, int newtailroom,
  285. int priority);
  286. extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
  287. #define dev_kfree_skb(a) kfree_skb(a)
  288. extern void skb_over_panic(struct sk_buff *skb, int len,
  289. void *here);
  290. extern void skb_under_panic(struct sk_buff *skb, int len,
  291. void *here);
  292. /* Internal */
  293. #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
  294. /**
  295. * skb_queue_empty - check if a queue is empty
  296. * @list: queue head
  297. *
  298. * Returns true if the queue is empty, false otherwise.
  299. */
  300. static inline int skb_queue_empty(const struct sk_buff_head *list)
  301. {
  302. return list->next == (struct sk_buff *)list;
  303. }
  304. /**
  305. * skb_get - reference buffer
  306. * @skb: buffer to reference
  307. *
  308. * Makes another reference to a socket buffer and returns a pointer
  309. * to the buffer.
  310. */
  311. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  312. {
  313. atomic_inc(&skb->users);
  314. return skb;
  315. }
  316. /*
  317. * If users == 1, we are the only owner and are can avoid redundant
  318. * atomic change.
  319. */
  320. /**
  321. * kfree_skb - free an sk_buff
  322. * @skb: buffer to free
  323. *
  324. * Drop a reference to the buffer and free it if the usage count has
  325. * hit zero.
  326. */
  327. static inline void kfree_skb(struct sk_buff *skb)
  328. {
  329. if (likely(atomic_read(&skb->users) == 1))
  330. smp_rmb();
  331. else if (likely(!atomic_dec_and_test(&skb->users)))
  332. return;
  333. __kfree_skb(skb);
  334. }
  335. /**
  336. * skb_cloned - is the buffer a clone
  337. * @skb: buffer to check
  338. *
  339. * Returns true if the buffer was generated with skb_clone() and is
  340. * one of multiple shared copies of the buffer. Cloned buffers are
  341. * shared data so must not be written to under normal circumstances.
  342. */
  343. static inline int skb_cloned(const struct sk_buff *skb)
  344. {
  345. return skb->cloned &&
  346. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  347. }
  348. /**
  349. * skb_header_cloned - is the header a clone
  350. * @skb: buffer to check
  351. *
  352. * Returns true if modifying the header part of the buffer requires
  353. * the data to be copied.
  354. */
  355. static inline int skb_header_cloned(const struct sk_buff *skb)
  356. {
  357. int dataref;
  358. if (!skb->cloned)
  359. return 0;
  360. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  361. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  362. return dataref != 1;
  363. }
  364. /**
  365. * skb_header_release - release reference to header
  366. * @skb: buffer to operate on
  367. *
  368. * Drop a reference to the header part of the buffer. This is done
  369. * by acquiring a payload reference. You must not read from the header
  370. * part of skb->data after this.
  371. */
  372. static inline void skb_header_release(struct sk_buff *skb)
  373. {
  374. BUG_ON(skb->nohdr);
  375. skb->nohdr = 1;
  376. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  377. }
  378. /**
  379. * skb_shared - is the buffer shared
  380. * @skb: buffer to check
  381. *
  382. * Returns true if more than one person has a reference to this
  383. * buffer.
  384. */
  385. static inline int skb_shared(const struct sk_buff *skb)
  386. {
  387. return atomic_read(&skb->users) != 1;
  388. }
  389. /**
  390. * skb_share_check - check if buffer is shared and if so clone it
  391. * @skb: buffer to check
  392. * @pri: priority for memory allocation
  393. *
  394. * If the buffer is shared the buffer is cloned and the old copy
  395. * drops a reference. A new clone with a single reference is returned.
  396. * If the buffer is not shared the original buffer is returned. When
  397. * being called from interrupt status or with spinlocks held pri must
  398. * be GFP_ATOMIC.
  399. *
  400. * NULL is returned on a memory allocation failure.
  401. */
  402. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, int pri)
  403. {
  404. might_sleep_if(pri & __GFP_WAIT);
  405. if (skb_shared(skb)) {
  406. struct sk_buff *nskb = skb_clone(skb, pri);
  407. kfree_skb(skb);
  408. skb = nskb;
  409. }
  410. return skb;
  411. }
  412. /*
  413. * Copy shared buffers into a new sk_buff. We effectively do COW on
  414. * packets to handle cases where we have a local reader and forward
  415. * and a couple of other messy ones. The normal one is tcpdumping
  416. * a packet thats being forwarded.
  417. */
  418. /**
  419. * skb_unshare - make a copy of a shared buffer
  420. * @skb: buffer to check
  421. * @pri: priority for memory allocation
  422. *
  423. * If the socket buffer is a clone then this function creates a new
  424. * copy of the data, drops a reference count on the old copy and returns
  425. * the new copy with the reference count at 1. If the buffer is not a clone
  426. * the original buffer is returned. When called with a spinlock held or
  427. * from interrupt state @pri must be %GFP_ATOMIC
  428. *
  429. * %NULL is returned on a memory allocation failure.
  430. */
  431. static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
  432. {
  433. might_sleep_if(pri & __GFP_WAIT);
  434. if (skb_cloned(skb)) {
  435. struct sk_buff *nskb = skb_copy(skb, pri);
  436. kfree_skb(skb); /* Free our shared copy */
  437. skb = nskb;
  438. }
  439. return skb;
  440. }
  441. /**
  442. * skb_peek
  443. * @list_: list to peek at
  444. *
  445. * Peek an &sk_buff. Unlike most other operations you _MUST_
  446. * be careful with this one. A peek leaves the buffer on the
  447. * list and someone else may run off with it. You must hold
  448. * the appropriate locks or have a private queue to do this.
  449. *
  450. * Returns %NULL for an empty list or a pointer to the head element.
  451. * The reference count is not incremented and the reference is therefore
  452. * volatile. Use with caution.
  453. */
  454. static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
  455. {
  456. struct sk_buff *list = ((struct sk_buff *)list_)->next;
  457. if (list == (struct sk_buff *)list_)
  458. list = NULL;
  459. return list;
  460. }
  461. /**
  462. * skb_peek_tail
  463. * @list_: list to peek at
  464. *
  465. * Peek an &sk_buff. Unlike most other operations you _MUST_
  466. * be careful with this one. A peek leaves the buffer on the
  467. * list and someone else may run off with it. You must hold
  468. * the appropriate locks or have a private queue to do this.
  469. *
  470. * Returns %NULL for an empty list or a pointer to the tail element.
  471. * The reference count is not incremented and the reference is therefore
  472. * volatile. Use with caution.
  473. */
  474. static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
  475. {
  476. struct sk_buff *list = ((struct sk_buff *)list_)->prev;
  477. if (list == (struct sk_buff *)list_)
  478. list = NULL;
  479. return list;
  480. }
  481. /**
  482. * skb_queue_len - get queue length
  483. * @list_: list to measure
  484. *
  485. * Return the length of an &sk_buff queue.
  486. */
  487. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  488. {
  489. return list_->qlen;
  490. }
  491. static inline void skb_queue_head_init(struct sk_buff_head *list)
  492. {
  493. spin_lock_init(&list->lock);
  494. list->prev = list->next = (struct sk_buff *)list;
  495. list->qlen = 0;
  496. }
  497. /*
  498. * Insert an sk_buff at the start of a list.
  499. *
  500. * The "__skb_xxxx()" functions are the non-atomic ones that
  501. * can only be called with interrupts disabled.
  502. */
  503. /**
  504. * __skb_queue_head - queue a buffer at the list head
  505. * @list: list to use
  506. * @newsk: buffer to queue
  507. *
  508. * Queue a buffer at the start of a list. This function takes no locks
  509. * and you must therefore hold required locks before calling it.
  510. *
  511. * A buffer cannot be placed on two lists at the same time.
  512. */
  513. extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  514. static inline void __skb_queue_head(struct sk_buff_head *list,
  515. struct sk_buff *newsk)
  516. {
  517. struct sk_buff *prev, *next;
  518. newsk->list = list;
  519. list->qlen++;
  520. prev = (struct sk_buff *)list;
  521. next = prev->next;
  522. newsk->next = next;
  523. newsk->prev = prev;
  524. next->prev = prev->next = newsk;
  525. }
  526. /**
  527. * __skb_queue_tail - queue a buffer at the list tail
  528. * @list: list to use
  529. * @newsk: buffer to queue
  530. *
  531. * Queue a buffer at the end of a list. This function takes no locks
  532. * and you must therefore hold required locks before calling it.
  533. *
  534. * A buffer cannot be placed on two lists at the same time.
  535. */
  536. extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  537. static inline void __skb_queue_tail(struct sk_buff_head *list,
  538. struct sk_buff *newsk)
  539. {
  540. struct sk_buff *prev, *next;
  541. newsk->list = list;
  542. list->qlen++;
  543. next = (struct sk_buff *)list;
  544. prev = next->prev;
  545. newsk->next = next;
  546. newsk->prev = prev;
  547. next->prev = prev->next = newsk;
  548. }
  549. /**
  550. * __skb_dequeue - remove from the head of the queue
  551. * @list: list to dequeue from
  552. *
  553. * Remove the head of the list. This function does not take any locks
  554. * so must be used with appropriate locks held only. The head item is
  555. * returned or %NULL if the list is empty.
  556. */
  557. extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  558. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  559. {
  560. struct sk_buff *next, *prev, *result;
  561. prev = (struct sk_buff *) list;
  562. next = prev->next;
  563. result = NULL;
  564. if (next != prev) {
  565. result = next;
  566. next = next->next;
  567. list->qlen--;
  568. next->prev = prev;
  569. prev->next = next;
  570. result->next = result->prev = NULL;
  571. result->list = NULL;
  572. }
  573. return result;
  574. }
  575. /*
  576. * Insert a packet on a list.
  577. */
  578. extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
  579. static inline void __skb_insert(struct sk_buff *newsk,
  580. struct sk_buff *prev, struct sk_buff *next,
  581. struct sk_buff_head *list)
  582. {
  583. newsk->next = next;
  584. newsk->prev = prev;
  585. next->prev = prev->next = newsk;
  586. newsk->list = list;
  587. list->qlen++;
  588. }
  589. /*
  590. * Place a packet after a given packet in a list.
  591. */
  592. extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
  593. static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
  594. {
  595. __skb_insert(newsk, old, old->next, old->list);
  596. }
  597. /*
  598. * remove sk_buff from list. _Must_ be called atomically, and with
  599. * the list known..
  600. */
  601. extern void skb_unlink(struct sk_buff *skb);
  602. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  603. {
  604. struct sk_buff *next, *prev;
  605. list->qlen--;
  606. next = skb->next;
  607. prev = skb->prev;
  608. skb->next = skb->prev = NULL;
  609. skb->list = NULL;
  610. next->prev = prev;
  611. prev->next = next;
  612. }
  613. /* XXX: more streamlined implementation */
  614. /**
  615. * __skb_dequeue_tail - remove from the tail of the queue
  616. * @list: list to dequeue from
  617. *
  618. * Remove the tail of the list. This function does not take any locks
  619. * so must be used with appropriate locks held only. The tail item is
  620. * returned or %NULL if the list is empty.
  621. */
  622. extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  623. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  624. {
  625. struct sk_buff *skb = skb_peek_tail(list);
  626. if (skb)
  627. __skb_unlink(skb, list);
  628. return skb;
  629. }
  630. static inline int skb_is_nonlinear(const struct sk_buff *skb)
  631. {
  632. return skb->data_len;
  633. }
  634. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  635. {
  636. return skb->len - skb->data_len;
  637. }
  638. static inline int skb_pagelen(const struct sk_buff *skb)
  639. {
  640. int i, len = 0;
  641. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  642. len += skb_shinfo(skb)->frags[i].size;
  643. return len + skb_headlen(skb);
  644. }
  645. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  646. struct page *page, int off, int size)
  647. {
  648. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  649. frag->page = page;
  650. frag->page_offset = off;
  651. frag->size = size;
  652. skb_shinfo(skb)->nr_frags = i + 1;
  653. }
  654. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  655. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
  656. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  657. /*
  658. * Add data to an sk_buff
  659. */
  660. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  661. {
  662. unsigned char *tmp = skb->tail;
  663. SKB_LINEAR_ASSERT(skb);
  664. skb->tail += len;
  665. skb->len += len;
  666. return tmp;
  667. }
  668. /**
  669. * skb_put - add data to a buffer
  670. * @skb: buffer to use
  671. * @len: amount of data to add
  672. *
  673. * This function extends the used data area of the buffer. If this would
  674. * exceed the total buffer size the kernel will panic. A pointer to the
  675. * first byte of the extra data is returned.
  676. */
  677. static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
  678. {
  679. unsigned char *tmp = skb->tail;
  680. SKB_LINEAR_ASSERT(skb);
  681. skb->tail += len;
  682. skb->len += len;
  683. if (unlikely(skb->tail>skb->end))
  684. skb_over_panic(skb, len, current_text_addr());
  685. return tmp;
  686. }
  687. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  688. {
  689. skb->data -= len;
  690. skb->len += len;
  691. return skb->data;
  692. }
  693. /**
  694. * skb_push - add data to the start of a buffer
  695. * @skb: buffer to use
  696. * @len: amount of data to add
  697. *
  698. * This function extends the used data area of the buffer at the buffer
  699. * start. If this would exceed the total buffer headroom the kernel will
  700. * panic. A pointer to the first byte of the extra data is returned.
  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. if (unlikely(skb->data<skb->head))
  707. skb_under_panic(skb, len, current_text_addr());
  708. return skb->data;
  709. }
  710. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  711. {
  712. skb->len -= len;
  713. BUG_ON(skb->len < skb->data_len);
  714. return skb->data += len;
  715. }
  716. /**
  717. * skb_pull - remove data from the start of a buffer
  718. * @skb: buffer to use
  719. * @len: amount of data to remove
  720. *
  721. * This function removes data from the start of a buffer, returning
  722. * the memory to the headroom. A pointer to the next data in the buffer
  723. * is returned. Once the data has been pulled future pushes will overwrite
  724. * the old data.
  725. */
  726. static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
  727. {
  728. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  729. }
  730. extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  731. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  732. {
  733. if (len > skb_headlen(skb) &&
  734. !__pskb_pull_tail(skb, len-skb_headlen(skb)))
  735. return NULL;
  736. skb->len -= len;
  737. return skb->data += len;
  738. }
  739. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  740. {
  741. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  742. }
  743. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  744. {
  745. if (likely(len <= skb_headlen(skb)))
  746. return 1;
  747. if (unlikely(len > skb->len))
  748. return 0;
  749. return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
  750. }
  751. /**
  752. * skb_headroom - bytes at buffer head
  753. * @skb: buffer to check
  754. *
  755. * Return the number of bytes of free space at the head of an &sk_buff.
  756. */
  757. static inline int skb_headroom(const struct sk_buff *skb)
  758. {
  759. return skb->data - skb->head;
  760. }
  761. /**
  762. * skb_tailroom - bytes at buffer end
  763. * @skb: buffer to check
  764. *
  765. * Return the number of bytes of free space at the tail of an sk_buff
  766. */
  767. static inline int skb_tailroom(const struct sk_buff *skb)
  768. {
  769. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  770. }
  771. /**
  772. * skb_reserve - adjust headroom
  773. * @skb: buffer to alter
  774. * @len: bytes to move
  775. *
  776. * Increase the headroom of an empty &sk_buff by reducing the tail
  777. * room. This is only allowed for an empty buffer.
  778. */
  779. static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
  780. {
  781. skb->data += len;
  782. skb->tail += len;
  783. }
  784. /*
  785. * CPUs often take a performance hit when accessing unaligned memory
  786. * locations. The actual performance hit varies, it can be small if the
  787. * hardware handles it or large if we have to take an exception and fix it
  788. * in software.
  789. *
  790. * Since an ethernet header is 14 bytes network drivers often end up with
  791. * the IP header at an unaligned offset. The IP header can be aligned by
  792. * shifting the start of the packet by 2 bytes. Drivers should do this
  793. * with:
  794. *
  795. * skb_reserve(NET_IP_ALIGN);
  796. *
  797. * The downside to this alignment of the IP header is that the DMA is now
  798. * unaligned. On some architectures the cost of an unaligned DMA is high
  799. * and this cost outweighs the gains made by aligning the IP header.
  800. *
  801. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  802. * to be overridden.
  803. */
  804. #ifndef NET_IP_ALIGN
  805. #define NET_IP_ALIGN 2
  806. #endif
  807. extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
  808. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  809. {
  810. if (!skb->data_len) {
  811. skb->len = len;
  812. skb->tail = skb->data + len;
  813. } else
  814. ___pskb_trim(skb, len, 0);
  815. }
  816. /**
  817. * skb_trim - remove end from a buffer
  818. * @skb: buffer to alter
  819. * @len: new length
  820. *
  821. * Cut the length of a buffer down by removing data from the tail. If
  822. * the buffer is already under the length specified it is not modified.
  823. */
  824. static inline void skb_trim(struct sk_buff *skb, unsigned int len)
  825. {
  826. if (skb->len > len)
  827. __skb_trim(skb, len);
  828. }
  829. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  830. {
  831. if (!skb->data_len) {
  832. skb->len = len;
  833. skb->tail = skb->data+len;
  834. return 0;
  835. }
  836. return ___pskb_trim(skb, len, 1);
  837. }
  838. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  839. {
  840. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  841. }
  842. /**
  843. * skb_orphan - orphan a buffer
  844. * @skb: buffer to orphan
  845. *
  846. * If a buffer currently has an owner then we call the owner's
  847. * destructor function and make the @skb unowned. The buffer continues
  848. * to exist but is no longer charged to its former owner.
  849. */
  850. static inline void skb_orphan(struct sk_buff *skb)
  851. {
  852. if (skb->destructor)
  853. skb->destructor(skb);
  854. skb->destructor = NULL;
  855. skb->sk = NULL;
  856. }
  857. /**
  858. * __skb_queue_purge - empty a list
  859. * @list: list to empty
  860. *
  861. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  862. * the list and one reference dropped. This function does not take the
  863. * list lock and the caller must hold the relevant locks to use it.
  864. */
  865. extern void skb_queue_purge(struct sk_buff_head *list);
  866. static inline void __skb_queue_purge(struct sk_buff_head *list)
  867. {
  868. struct sk_buff *skb;
  869. while ((skb = __skb_dequeue(list)) != NULL)
  870. kfree_skb(skb);
  871. }
  872. #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
  873. /**
  874. * __dev_alloc_skb - allocate an skbuff for sending
  875. * @length: length to allocate
  876. * @gfp_mask: get_free_pages mask, passed to alloc_skb
  877. *
  878. * Allocate a new &sk_buff and assign it a usage count of one. The
  879. * buffer has unspecified headroom built in. Users should allocate
  880. * the headroom they think they need without accounting for the
  881. * built in space. The built in space is used for optimisations.
  882. *
  883. * %NULL is returned in there is no free memory.
  884. */
  885. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  886. int gfp_mask)
  887. {
  888. struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
  889. if (likely(skb))
  890. skb_reserve(skb, 16);
  891. return skb;
  892. }
  893. #else
  894. extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
  895. #endif
  896. /**
  897. * dev_alloc_skb - allocate an skbuff for sending
  898. * @length: length to allocate
  899. *
  900. * Allocate a new &sk_buff and assign it a usage count of one. The
  901. * buffer has unspecified headroom built in. Users should allocate
  902. * the headroom they think they need without accounting for the
  903. * built in space. The built in space is used for optimisations.
  904. *
  905. * %NULL is returned in there is no free memory. Although this function
  906. * allocates memory it can be called from an interrupt.
  907. */
  908. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  909. {
  910. return __dev_alloc_skb(length, GFP_ATOMIC);
  911. }
  912. /**
  913. * skb_cow - copy header of skb when it is required
  914. * @skb: buffer to cow
  915. * @headroom: needed headroom
  916. *
  917. * If the skb passed lacks sufficient headroom or its data part
  918. * is shared, data is reallocated. If reallocation fails, an error
  919. * is returned and original skb is not changed.
  920. *
  921. * The result is skb with writable area skb->head...skb->tail
  922. * and at least @headroom of space at head.
  923. */
  924. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  925. {
  926. int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
  927. if (delta < 0)
  928. delta = 0;
  929. if (delta || skb_cloned(skb))
  930. return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
  931. return 0;
  932. }
  933. /**
  934. * skb_padto - pad an skbuff up to a minimal size
  935. * @skb: buffer to pad
  936. * @len: minimal length
  937. *
  938. * Pads up a buffer to ensure the trailing bytes exist and are
  939. * blanked. If the buffer already contains sufficient data it
  940. * is untouched. Returns the buffer, which may be a replacement
  941. * for the original, or NULL for out of memory - in which case
  942. * the original buffer is still freed.
  943. */
  944. static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
  945. {
  946. unsigned int size = skb->len;
  947. if (likely(size >= len))
  948. return skb;
  949. return skb_pad(skb, len-size);
  950. }
  951. static inline int skb_add_data(struct sk_buff *skb,
  952. char __user *from, int copy)
  953. {
  954. const int off = skb->len;
  955. if (skb->ip_summed == CHECKSUM_NONE) {
  956. int err = 0;
  957. unsigned int csum = csum_and_copy_from_user(from,
  958. skb_put(skb, copy),
  959. copy, 0, &err);
  960. if (!err) {
  961. skb->csum = csum_block_add(skb->csum, csum, off);
  962. return 0;
  963. }
  964. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  965. return 0;
  966. __skb_trim(skb, off);
  967. return -EFAULT;
  968. }
  969. static inline int skb_can_coalesce(struct sk_buff *skb, int i,
  970. struct page *page, int off)
  971. {
  972. if (i) {
  973. struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  974. return page == frag->page &&
  975. off == frag->page_offset + frag->size;
  976. }
  977. return 0;
  978. }
  979. /**
  980. * skb_linearize - convert paged skb to linear one
  981. * @skb: buffer to linarize
  982. * @gfp: allocation mode
  983. *
  984. * If there is no free memory -ENOMEM is returned, otherwise zero
  985. * is returned and the old skb data released.
  986. */
  987. extern int __skb_linearize(struct sk_buff *skb, int gfp);
  988. static inline int skb_linearize(struct sk_buff *skb, int gfp)
  989. {
  990. return __skb_linearize(skb, gfp);
  991. }
  992. /**
  993. * skb_postpull_rcsum - update checksum for received skb after pull
  994. * @skb: buffer to update
  995. * @start: start of data before pull
  996. * @len: length of data pulled
  997. *
  998. * After doing a pull on a received packet, you need to call this to
  999. * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
  1000. * so that it can be recomputed from scratch.
  1001. */
  1002. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  1003. const void *start, int len)
  1004. {
  1005. if (skb->ip_summed == CHECKSUM_HW)
  1006. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  1007. }
  1008. /**
  1009. * pskb_trim_rcsum - trim received skb and update checksum
  1010. * @skb: buffer to trim
  1011. * @len: new length
  1012. *
  1013. * This is exactly the same as pskb_trim except that it ensures the
  1014. * checksum of received packets are still valid after the operation.
  1015. */
  1016. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  1017. {
  1018. if (len >= skb->len)
  1019. return 0;
  1020. if (skb->ip_summed == CHECKSUM_HW)
  1021. skb->ip_summed = CHECKSUM_NONE;
  1022. return __pskb_trim(skb, len);
  1023. }
  1024. static inline void *kmap_skb_frag(const skb_frag_t *frag)
  1025. {
  1026. #ifdef CONFIG_HIGHMEM
  1027. BUG_ON(in_irq());
  1028. local_bh_disable();
  1029. #endif
  1030. return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
  1031. }
  1032. static inline void kunmap_skb_frag(void *vaddr)
  1033. {
  1034. kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
  1035. #ifdef CONFIG_HIGHMEM
  1036. local_bh_enable();
  1037. #endif
  1038. }
  1039. #define skb_queue_walk(queue, skb) \
  1040. for (skb = (queue)->next; \
  1041. prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
  1042. skb = skb->next)
  1043. extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
  1044. int noblock, int *err);
  1045. extern unsigned int datagram_poll(struct file *file, struct socket *sock,
  1046. struct poll_table_struct *wait);
  1047. extern int skb_copy_datagram_iovec(const struct sk_buff *from,
  1048. int offset, struct iovec *to,
  1049. int size);
  1050. extern int skb_copy_and_csum_datagram_iovec(const
  1051. struct sk_buff *skb,
  1052. int hlen,
  1053. struct iovec *iov);
  1054. extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  1055. extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
  1056. int len, unsigned int csum);
  1057. extern int skb_copy_bits(const struct sk_buff *skb, int offset,
  1058. void *to, int len);
  1059. extern int skb_store_bits(const struct sk_buff *skb, int offset,
  1060. void *from, int len);
  1061. extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
  1062. int offset, u8 *to, int len,
  1063. unsigned int csum);
  1064. extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  1065. extern void skb_split(struct sk_buff *skb,
  1066. struct sk_buff *skb1, const u32 len);
  1067. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  1068. int len, void *buffer)
  1069. {
  1070. int hlen = skb_headlen(skb);
  1071. if (offset + len <= hlen)
  1072. return skb->data + offset;
  1073. if (skb_copy_bits(skb, offset, buffer, len) < 0)
  1074. return NULL;
  1075. return buffer;
  1076. }
  1077. extern void skb_init(void);
  1078. extern void skb_add_mtu(int mtu);
  1079. #ifdef CONFIG_NETFILTER
  1080. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  1081. {
  1082. if (nfct && atomic_dec_and_test(&nfct->use))
  1083. nfct->destroy(nfct);
  1084. }
  1085. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  1086. {
  1087. if (nfct)
  1088. atomic_inc(&nfct->use);
  1089. }
  1090. static inline void nf_reset(struct sk_buff *skb)
  1091. {
  1092. nf_conntrack_put(skb->nfct);
  1093. skb->nfct = NULL;
  1094. #ifdef CONFIG_NETFILTER_DEBUG
  1095. skb->nf_debug = 0;
  1096. #endif
  1097. }
  1098. static inline void nf_reset_debug(struct sk_buff *skb)
  1099. {
  1100. #ifdef CONFIG_NETFILTER_DEBUG
  1101. skb->nf_debug = 0;
  1102. #endif
  1103. }
  1104. #ifdef CONFIG_BRIDGE_NETFILTER
  1105. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  1106. {
  1107. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  1108. kfree(nf_bridge);
  1109. }
  1110. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  1111. {
  1112. if (nf_bridge)
  1113. atomic_inc(&nf_bridge->use);
  1114. }
  1115. #endif /* CONFIG_BRIDGE_NETFILTER */
  1116. #else /* CONFIG_NETFILTER */
  1117. static inline void nf_reset(struct sk_buff *skb) {}
  1118. #endif /* CONFIG_NETFILTER */
  1119. #endif /* __KERNEL__ */
  1120. #endif /* _LINUX_SKBUFF_H */