skbuff.h 34 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267
  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. __be16 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,
  269. unsigned int __nocast priority);
  270. extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
  271. unsigned int size,
  272. unsigned int __nocast priority);
  273. extern void kfree_skbmem(struct sk_buff *skb);
  274. extern struct sk_buff *skb_clone(struct sk_buff *skb,
  275. unsigned int __nocast priority);
  276. extern struct sk_buff *skb_copy(const struct sk_buff *skb,
  277. unsigned int __nocast priority);
  278. extern struct sk_buff *pskb_copy(struct sk_buff *skb,
  279. unsigned int __nocast gfp_mask);
  280. extern int pskb_expand_head(struct sk_buff *skb,
  281. int nhead, int ntail,
  282. unsigned int __nocast gfp_mask);
  283. extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  284. unsigned int headroom);
  285. extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
  286. int newheadroom, int newtailroom,
  287. unsigned int __nocast priority);
  288. extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
  289. #define dev_kfree_skb(a) kfree_skb(a)
  290. extern void skb_over_panic(struct sk_buff *skb, int len,
  291. void *here);
  292. extern void skb_under_panic(struct sk_buff *skb, int len,
  293. void *here);
  294. struct skb_seq_state
  295. {
  296. __u32 lower_offset;
  297. __u32 upper_offset;
  298. __u32 frag_idx;
  299. __u32 stepped_offset;
  300. struct sk_buff *root_skb;
  301. struct sk_buff *cur_skb;
  302. __u8 *frag_data;
  303. };
  304. extern void skb_prepare_seq_read(struct sk_buff *skb,
  305. unsigned int from, unsigned int to,
  306. struct skb_seq_state *st);
  307. extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  308. struct skb_seq_state *st);
  309. extern void skb_abort_seq_read(struct skb_seq_state *st);
  310. extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  311. unsigned int to, struct ts_config *config,
  312. struct ts_state *state);
  313. /* Internal */
  314. #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
  315. /**
  316. * skb_queue_empty - check if a queue is empty
  317. * @list: queue head
  318. *
  319. * Returns true if the queue is empty, false otherwise.
  320. */
  321. static inline int skb_queue_empty(const struct sk_buff_head *list)
  322. {
  323. return list->next == (struct sk_buff *)list;
  324. }
  325. /**
  326. * skb_get - reference buffer
  327. * @skb: buffer to reference
  328. *
  329. * Makes another reference to a socket buffer and returns a pointer
  330. * to the buffer.
  331. */
  332. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  333. {
  334. atomic_inc(&skb->users);
  335. return skb;
  336. }
  337. /*
  338. * If users == 1, we are the only owner and are can avoid redundant
  339. * atomic change.
  340. */
  341. /**
  342. * kfree_skb - free an sk_buff
  343. * @skb: buffer to free
  344. *
  345. * Drop a reference to the buffer and free it if the usage count has
  346. * hit zero.
  347. */
  348. static inline void kfree_skb(struct sk_buff *skb)
  349. {
  350. if (likely(atomic_read(&skb->users) == 1))
  351. smp_rmb();
  352. else if (likely(!atomic_dec_and_test(&skb->users)))
  353. return;
  354. __kfree_skb(skb);
  355. }
  356. /**
  357. * skb_cloned - is the buffer a clone
  358. * @skb: buffer to check
  359. *
  360. * Returns true if the buffer was generated with skb_clone() and is
  361. * one of multiple shared copies of the buffer. Cloned buffers are
  362. * shared data so must not be written to under normal circumstances.
  363. */
  364. static inline int skb_cloned(const struct sk_buff *skb)
  365. {
  366. return skb->cloned &&
  367. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  368. }
  369. /**
  370. * skb_header_cloned - is the header a clone
  371. * @skb: buffer to check
  372. *
  373. * Returns true if modifying the header part of the buffer requires
  374. * the data to be copied.
  375. */
  376. static inline int skb_header_cloned(const struct sk_buff *skb)
  377. {
  378. int dataref;
  379. if (!skb->cloned)
  380. return 0;
  381. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  382. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  383. return dataref != 1;
  384. }
  385. /**
  386. * skb_header_release - release reference to header
  387. * @skb: buffer to operate on
  388. *
  389. * Drop a reference to the header part of the buffer. This is done
  390. * by acquiring a payload reference. You must not read from the header
  391. * part of skb->data after this.
  392. */
  393. static inline void skb_header_release(struct sk_buff *skb)
  394. {
  395. BUG_ON(skb->nohdr);
  396. skb->nohdr = 1;
  397. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  398. }
  399. /**
  400. * skb_shared - is the buffer shared
  401. * @skb: buffer to check
  402. *
  403. * Returns true if more than one person has a reference to this
  404. * buffer.
  405. */
  406. static inline int skb_shared(const struct sk_buff *skb)
  407. {
  408. return atomic_read(&skb->users) != 1;
  409. }
  410. /**
  411. * skb_share_check - check if buffer is shared and if so clone it
  412. * @skb: buffer to check
  413. * @pri: priority for memory allocation
  414. *
  415. * If the buffer is shared the buffer is cloned and the old copy
  416. * drops a reference. A new clone with a single reference is returned.
  417. * If the buffer is not shared the original buffer is returned. When
  418. * being called from interrupt status or with spinlocks held pri must
  419. * be GFP_ATOMIC.
  420. *
  421. * NULL is returned on a memory allocation failure.
  422. */
  423. static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
  424. unsigned int __nocast pri)
  425. {
  426. might_sleep_if(pri & __GFP_WAIT);
  427. if (skb_shared(skb)) {
  428. struct sk_buff *nskb = skb_clone(skb, pri);
  429. kfree_skb(skb);
  430. skb = nskb;
  431. }
  432. return skb;
  433. }
  434. /*
  435. * Copy shared buffers into a new sk_buff. We effectively do COW on
  436. * packets to handle cases where we have a local reader and forward
  437. * and a couple of other messy ones. The normal one is tcpdumping
  438. * a packet thats being forwarded.
  439. */
  440. /**
  441. * skb_unshare - make a copy of a shared buffer
  442. * @skb: buffer to check
  443. * @pri: priority for memory allocation
  444. *
  445. * If the socket buffer is a clone then this function creates a new
  446. * copy of the data, drops a reference count on the old copy and returns
  447. * the new copy with the reference count at 1. If the buffer is not a clone
  448. * the original buffer is returned. When called with a spinlock held or
  449. * from interrupt state @pri must be %GFP_ATOMIC
  450. *
  451. * %NULL is returned on a memory allocation failure.
  452. */
  453. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  454. unsigned int __nocast pri)
  455. {
  456. might_sleep_if(pri & __GFP_WAIT);
  457. if (skb_cloned(skb)) {
  458. struct sk_buff *nskb = skb_copy(skb, pri);
  459. kfree_skb(skb); /* Free our shared copy */
  460. skb = nskb;
  461. }
  462. return skb;
  463. }
  464. /**
  465. * skb_peek
  466. * @list_: list to peek at
  467. *
  468. * Peek an &sk_buff. Unlike most other operations you _MUST_
  469. * be careful with this one. A peek leaves the buffer on the
  470. * list and someone else may run off with it. You must hold
  471. * the appropriate locks or have a private queue to do this.
  472. *
  473. * Returns %NULL for an empty list or a pointer to the head element.
  474. * The reference count is not incremented and the reference is therefore
  475. * volatile. Use with caution.
  476. */
  477. static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
  478. {
  479. struct sk_buff *list = ((struct sk_buff *)list_)->next;
  480. if (list == (struct sk_buff *)list_)
  481. list = NULL;
  482. return list;
  483. }
  484. /**
  485. * skb_peek_tail
  486. * @list_: list to peek at
  487. *
  488. * Peek an &sk_buff. Unlike most other operations you _MUST_
  489. * be careful with this one. A peek leaves the buffer on the
  490. * list and someone else may run off with it. You must hold
  491. * the appropriate locks or have a private queue to do this.
  492. *
  493. * Returns %NULL for an empty list or a pointer to the tail element.
  494. * The reference count is not incremented and the reference is therefore
  495. * volatile. Use with caution.
  496. */
  497. static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
  498. {
  499. struct sk_buff *list = ((struct sk_buff *)list_)->prev;
  500. if (list == (struct sk_buff *)list_)
  501. list = NULL;
  502. return list;
  503. }
  504. /**
  505. * skb_queue_len - get queue length
  506. * @list_: list to measure
  507. *
  508. * Return the length of an &sk_buff queue.
  509. */
  510. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  511. {
  512. return list_->qlen;
  513. }
  514. static inline void skb_queue_head_init(struct sk_buff_head *list)
  515. {
  516. spin_lock_init(&list->lock);
  517. list->prev = list->next = (struct sk_buff *)list;
  518. list->qlen = 0;
  519. }
  520. /*
  521. * Insert an sk_buff at the start of a list.
  522. *
  523. * The "__skb_xxxx()" functions are the non-atomic ones that
  524. * can only be called with interrupts disabled.
  525. */
  526. /**
  527. * __skb_queue_head - queue a buffer at the list head
  528. * @list: list to use
  529. * @newsk: buffer to queue
  530. *
  531. * Queue a buffer at the start 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_head(struct sk_buff_head *list, struct sk_buff *newsk);
  537. static inline void __skb_queue_head(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. prev = (struct sk_buff *)list;
  544. next = prev->next;
  545. newsk->next = next;
  546. newsk->prev = prev;
  547. next->prev = prev->next = newsk;
  548. }
  549. /**
  550. * __skb_queue_tail - queue a buffer at the list tail
  551. * @list: list to use
  552. * @newsk: buffer to queue
  553. *
  554. * Queue a buffer at the end of a list. This function takes no locks
  555. * and you must therefore hold required locks before calling it.
  556. *
  557. * A buffer cannot be placed on two lists at the same time.
  558. */
  559. extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  560. static inline void __skb_queue_tail(struct sk_buff_head *list,
  561. struct sk_buff *newsk)
  562. {
  563. struct sk_buff *prev, *next;
  564. newsk->list = list;
  565. list->qlen++;
  566. next = (struct sk_buff *)list;
  567. prev = next->prev;
  568. newsk->next = next;
  569. newsk->prev = prev;
  570. next->prev = prev->next = newsk;
  571. }
  572. /**
  573. * __skb_dequeue - remove from the head of the queue
  574. * @list: list to dequeue from
  575. *
  576. * Remove the head of the list. This function does not take any locks
  577. * so must be used with appropriate locks held only. The head item is
  578. * returned or %NULL if the list is empty.
  579. */
  580. extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  581. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  582. {
  583. struct sk_buff *next, *prev, *result;
  584. prev = (struct sk_buff *) list;
  585. next = prev->next;
  586. result = NULL;
  587. if (next != prev) {
  588. result = next;
  589. next = next->next;
  590. list->qlen--;
  591. next->prev = prev;
  592. prev->next = next;
  593. result->next = result->prev = NULL;
  594. result->list = NULL;
  595. }
  596. return result;
  597. }
  598. /*
  599. * Insert a packet on a list.
  600. */
  601. extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
  602. static inline void __skb_insert(struct sk_buff *newsk,
  603. struct sk_buff *prev, struct sk_buff *next,
  604. struct sk_buff_head *list)
  605. {
  606. newsk->next = next;
  607. newsk->prev = prev;
  608. next->prev = prev->next = newsk;
  609. newsk->list = list;
  610. list->qlen++;
  611. }
  612. /*
  613. * Place a packet after a given packet in a list.
  614. */
  615. extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
  616. static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
  617. {
  618. __skb_insert(newsk, old, old->next, old->list);
  619. }
  620. /*
  621. * remove sk_buff from list. _Must_ be called atomically, and with
  622. * the list known..
  623. */
  624. extern void skb_unlink(struct sk_buff *skb);
  625. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  626. {
  627. struct sk_buff *next, *prev;
  628. list->qlen--;
  629. next = skb->next;
  630. prev = skb->prev;
  631. skb->next = skb->prev = NULL;
  632. skb->list = NULL;
  633. next->prev = prev;
  634. prev->next = next;
  635. }
  636. /* XXX: more streamlined implementation */
  637. /**
  638. * __skb_dequeue_tail - remove from the tail of the queue
  639. * @list: list to dequeue from
  640. *
  641. * Remove the tail of the list. This function does not take any locks
  642. * so must be used with appropriate locks held only. The tail item is
  643. * returned or %NULL if the list is empty.
  644. */
  645. extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  646. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  647. {
  648. struct sk_buff *skb = skb_peek_tail(list);
  649. if (skb)
  650. __skb_unlink(skb, list);
  651. return skb;
  652. }
  653. static inline int skb_is_nonlinear(const struct sk_buff *skb)
  654. {
  655. return skb->data_len;
  656. }
  657. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  658. {
  659. return skb->len - skb->data_len;
  660. }
  661. static inline int skb_pagelen(const struct sk_buff *skb)
  662. {
  663. int i, len = 0;
  664. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  665. len += skb_shinfo(skb)->frags[i].size;
  666. return len + skb_headlen(skb);
  667. }
  668. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  669. struct page *page, int off, int size)
  670. {
  671. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  672. frag->page = page;
  673. frag->page_offset = off;
  674. frag->size = size;
  675. skb_shinfo(skb)->nr_frags = i + 1;
  676. }
  677. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  678. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
  679. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  680. /*
  681. * Add data to an sk_buff
  682. */
  683. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  684. {
  685. unsigned char *tmp = skb->tail;
  686. SKB_LINEAR_ASSERT(skb);
  687. skb->tail += len;
  688. skb->len += len;
  689. return tmp;
  690. }
  691. /**
  692. * skb_put - add data to a buffer
  693. * @skb: buffer to use
  694. * @len: amount of data to add
  695. *
  696. * This function extends the used data area of the buffer. If this would
  697. * exceed the total buffer size the kernel will panic. A pointer to the
  698. * first byte of the extra data is returned.
  699. */
  700. static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
  701. {
  702. unsigned char *tmp = skb->tail;
  703. SKB_LINEAR_ASSERT(skb);
  704. skb->tail += len;
  705. skb->len += len;
  706. if (unlikely(skb->tail>skb->end))
  707. skb_over_panic(skb, len, current_text_addr());
  708. return tmp;
  709. }
  710. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  711. {
  712. skb->data -= len;
  713. skb->len += len;
  714. return skb->data;
  715. }
  716. /**
  717. * skb_push - add data to the start of a buffer
  718. * @skb: buffer to use
  719. * @len: amount of data to add
  720. *
  721. * This function extends the used data area of the buffer at the buffer
  722. * start. If this would exceed the total buffer headroom the kernel will
  723. * panic. A pointer to the first byte of the extra data is returned.
  724. */
  725. static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
  726. {
  727. skb->data -= len;
  728. skb->len += len;
  729. if (unlikely(skb->data<skb->head))
  730. skb_under_panic(skb, len, current_text_addr());
  731. return skb->data;
  732. }
  733. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  734. {
  735. skb->len -= len;
  736. BUG_ON(skb->len < skb->data_len);
  737. return skb->data += len;
  738. }
  739. /**
  740. * skb_pull - remove data from the start of a buffer
  741. * @skb: buffer to use
  742. * @len: amount of data to remove
  743. *
  744. * This function removes data from the start of a buffer, returning
  745. * the memory to the headroom. A pointer to the next data in the buffer
  746. * is returned. Once the data has been pulled future pushes will overwrite
  747. * the old data.
  748. */
  749. static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
  750. {
  751. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  752. }
  753. extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  754. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  755. {
  756. if (len > skb_headlen(skb) &&
  757. !__pskb_pull_tail(skb, len-skb_headlen(skb)))
  758. return NULL;
  759. skb->len -= len;
  760. return skb->data += len;
  761. }
  762. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  763. {
  764. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  765. }
  766. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  767. {
  768. if (likely(len <= skb_headlen(skb)))
  769. return 1;
  770. if (unlikely(len > skb->len))
  771. return 0;
  772. return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
  773. }
  774. /**
  775. * skb_headroom - bytes at buffer head
  776. * @skb: buffer to check
  777. *
  778. * Return the number of bytes of free space at the head of an &sk_buff.
  779. */
  780. static inline int skb_headroom(const struct sk_buff *skb)
  781. {
  782. return skb->data - skb->head;
  783. }
  784. /**
  785. * skb_tailroom - bytes at buffer end
  786. * @skb: buffer to check
  787. *
  788. * Return the number of bytes of free space at the tail of an sk_buff
  789. */
  790. static inline int skb_tailroom(const struct sk_buff *skb)
  791. {
  792. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  793. }
  794. /**
  795. * skb_reserve - adjust headroom
  796. * @skb: buffer to alter
  797. * @len: bytes to move
  798. *
  799. * Increase the headroom of an empty &sk_buff by reducing the tail
  800. * room. This is only allowed for an empty buffer.
  801. */
  802. static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
  803. {
  804. skb->data += len;
  805. skb->tail += len;
  806. }
  807. /*
  808. * CPUs often take a performance hit when accessing unaligned memory
  809. * locations. The actual performance hit varies, it can be small if the
  810. * hardware handles it or large if we have to take an exception and fix it
  811. * in software.
  812. *
  813. * Since an ethernet header is 14 bytes network drivers often end up with
  814. * the IP header at an unaligned offset. The IP header can be aligned by
  815. * shifting the start of the packet by 2 bytes. Drivers should do this
  816. * with:
  817. *
  818. * skb_reserve(NET_IP_ALIGN);
  819. *
  820. * The downside to this alignment of the IP header is that the DMA is now
  821. * unaligned. On some architectures the cost of an unaligned DMA is high
  822. * and this cost outweighs the gains made by aligning the IP header.
  823. *
  824. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  825. * to be overridden.
  826. */
  827. #ifndef NET_IP_ALIGN
  828. #define NET_IP_ALIGN 2
  829. #endif
  830. extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
  831. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  832. {
  833. if (!skb->data_len) {
  834. skb->len = len;
  835. skb->tail = skb->data + len;
  836. } else
  837. ___pskb_trim(skb, len, 0);
  838. }
  839. /**
  840. * skb_trim - remove end from a buffer
  841. * @skb: buffer to alter
  842. * @len: new length
  843. *
  844. * Cut the length of a buffer down by removing data from the tail. If
  845. * the buffer is already under the length specified it is not modified.
  846. */
  847. static inline void skb_trim(struct sk_buff *skb, unsigned int len)
  848. {
  849. if (skb->len > len)
  850. __skb_trim(skb, len);
  851. }
  852. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  853. {
  854. if (!skb->data_len) {
  855. skb->len = len;
  856. skb->tail = skb->data+len;
  857. return 0;
  858. }
  859. return ___pskb_trim(skb, len, 1);
  860. }
  861. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  862. {
  863. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  864. }
  865. /**
  866. * skb_orphan - orphan a buffer
  867. * @skb: buffer to orphan
  868. *
  869. * If a buffer currently has an owner then we call the owner's
  870. * destructor function and make the @skb unowned. The buffer continues
  871. * to exist but is no longer charged to its former owner.
  872. */
  873. static inline void skb_orphan(struct sk_buff *skb)
  874. {
  875. if (skb->destructor)
  876. skb->destructor(skb);
  877. skb->destructor = NULL;
  878. skb->sk = NULL;
  879. }
  880. /**
  881. * __skb_queue_purge - empty a list
  882. * @list: list to empty
  883. *
  884. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  885. * the list and one reference dropped. This function does not take the
  886. * list lock and the caller must hold the relevant locks to use it.
  887. */
  888. extern void skb_queue_purge(struct sk_buff_head *list);
  889. static inline void __skb_queue_purge(struct sk_buff_head *list)
  890. {
  891. struct sk_buff *skb;
  892. while ((skb = __skb_dequeue(list)) != NULL)
  893. kfree_skb(skb);
  894. }
  895. #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
  896. /**
  897. * __dev_alloc_skb - allocate an skbuff for sending
  898. * @length: length to allocate
  899. * @gfp_mask: get_free_pages mask, passed to alloc_skb
  900. *
  901. * Allocate a new &sk_buff and assign it a usage count of one. The
  902. * buffer has unspecified headroom built in. Users should allocate
  903. * the headroom they think they need without accounting for the
  904. * built in space. The built in space is used for optimisations.
  905. *
  906. * %NULL is returned in there is no free memory.
  907. */
  908. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  909. unsigned int __nocast gfp_mask)
  910. {
  911. struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
  912. if (likely(skb))
  913. skb_reserve(skb, 16);
  914. return skb;
  915. }
  916. #else
  917. extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
  918. #endif
  919. /**
  920. * dev_alloc_skb - allocate an skbuff for sending
  921. * @length: length to allocate
  922. *
  923. * Allocate a new &sk_buff and assign it a usage count of one. The
  924. * buffer has unspecified headroom built in. Users should allocate
  925. * the headroom they think they need without accounting for the
  926. * built in space. The built in space is used for optimisations.
  927. *
  928. * %NULL is returned in there is no free memory. Although this function
  929. * allocates memory it can be called from an interrupt.
  930. */
  931. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  932. {
  933. return __dev_alloc_skb(length, GFP_ATOMIC);
  934. }
  935. /**
  936. * skb_cow - copy header of skb when it is required
  937. * @skb: buffer to cow
  938. * @headroom: needed headroom
  939. *
  940. * If the skb passed lacks sufficient headroom or its data part
  941. * is shared, data is reallocated. If reallocation fails, an error
  942. * is returned and original skb is not changed.
  943. *
  944. * The result is skb with writable area skb->head...skb->tail
  945. * and at least @headroom of space at head.
  946. */
  947. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  948. {
  949. int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
  950. if (delta < 0)
  951. delta = 0;
  952. if (delta || skb_cloned(skb))
  953. return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
  954. return 0;
  955. }
  956. /**
  957. * skb_padto - pad an skbuff up to a minimal size
  958. * @skb: buffer to pad
  959. * @len: minimal length
  960. *
  961. * Pads up a buffer to ensure the trailing bytes exist and are
  962. * blanked. If the buffer already contains sufficient data it
  963. * is untouched. Returns the buffer, which may be a replacement
  964. * for the original, or NULL for out of memory - in which case
  965. * the original buffer is still freed.
  966. */
  967. static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
  968. {
  969. unsigned int size = skb->len;
  970. if (likely(size >= len))
  971. return skb;
  972. return skb_pad(skb, len-size);
  973. }
  974. static inline int skb_add_data(struct sk_buff *skb,
  975. char __user *from, int copy)
  976. {
  977. const int off = skb->len;
  978. if (skb->ip_summed == CHECKSUM_NONE) {
  979. int err = 0;
  980. unsigned int csum = csum_and_copy_from_user(from,
  981. skb_put(skb, copy),
  982. copy, 0, &err);
  983. if (!err) {
  984. skb->csum = csum_block_add(skb->csum, csum, off);
  985. return 0;
  986. }
  987. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  988. return 0;
  989. __skb_trim(skb, off);
  990. return -EFAULT;
  991. }
  992. static inline int skb_can_coalesce(struct sk_buff *skb, int i,
  993. struct page *page, int off)
  994. {
  995. if (i) {
  996. struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  997. return page == frag->page &&
  998. off == frag->page_offset + frag->size;
  999. }
  1000. return 0;
  1001. }
  1002. /**
  1003. * skb_linearize - convert paged skb to linear one
  1004. * @skb: buffer to linarize
  1005. * @gfp: allocation mode
  1006. *
  1007. * If there is no free memory -ENOMEM is returned, otherwise zero
  1008. * is returned and the old skb data released.
  1009. */
  1010. extern int __skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp);
  1011. static inline int skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp)
  1012. {
  1013. return __skb_linearize(skb, gfp);
  1014. }
  1015. /**
  1016. * skb_postpull_rcsum - update checksum for received skb after pull
  1017. * @skb: buffer to update
  1018. * @start: start of data before pull
  1019. * @len: length of data pulled
  1020. *
  1021. * After doing a pull on a received packet, you need to call this to
  1022. * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
  1023. * so that it can be recomputed from scratch.
  1024. */
  1025. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  1026. const void *start, int len)
  1027. {
  1028. if (skb->ip_summed == CHECKSUM_HW)
  1029. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  1030. }
  1031. /**
  1032. * pskb_trim_rcsum - trim received skb and update checksum
  1033. * @skb: buffer to trim
  1034. * @len: new length
  1035. *
  1036. * This is exactly the same as pskb_trim except that it ensures the
  1037. * checksum of received packets are still valid after the operation.
  1038. */
  1039. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  1040. {
  1041. if (len >= skb->len)
  1042. return 0;
  1043. if (skb->ip_summed == CHECKSUM_HW)
  1044. skb->ip_summed = CHECKSUM_NONE;
  1045. return __pskb_trim(skb, len);
  1046. }
  1047. static inline void *kmap_skb_frag(const skb_frag_t *frag)
  1048. {
  1049. #ifdef CONFIG_HIGHMEM
  1050. BUG_ON(in_irq());
  1051. local_bh_disable();
  1052. #endif
  1053. return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
  1054. }
  1055. static inline void kunmap_skb_frag(void *vaddr)
  1056. {
  1057. kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
  1058. #ifdef CONFIG_HIGHMEM
  1059. local_bh_enable();
  1060. #endif
  1061. }
  1062. #define skb_queue_walk(queue, skb) \
  1063. for (skb = (queue)->next; \
  1064. prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
  1065. skb = skb->next)
  1066. extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
  1067. int noblock, int *err);
  1068. extern unsigned int datagram_poll(struct file *file, struct socket *sock,
  1069. struct poll_table_struct *wait);
  1070. extern int skb_copy_datagram_iovec(const struct sk_buff *from,
  1071. int offset, struct iovec *to,
  1072. int size);
  1073. extern int skb_copy_and_csum_datagram_iovec(const
  1074. struct sk_buff *skb,
  1075. int hlen,
  1076. struct iovec *iov);
  1077. extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  1078. extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
  1079. int len, unsigned int csum);
  1080. extern int skb_copy_bits(const struct sk_buff *skb, int offset,
  1081. void *to, int len);
  1082. extern int skb_store_bits(const struct sk_buff *skb, int offset,
  1083. void *from, int len);
  1084. extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
  1085. int offset, u8 *to, int len,
  1086. unsigned int csum);
  1087. extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  1088. extern void skb_split(struct sk_buff *skb,
  1089. struct sk_buff *skb1, const u32 len);
  1090. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  1091. int len, void *buffer)
  1092. {
  1093. int hlen = skb_headlen(skb);
  1094. if (hlen - offset >= len)
  1095. return skb->data + offset;
  1096. if (skb_copy_bits(skb, offset, buffer, len) < 0)
  1097. return NULL;
  1098. return buffer;
  1099. }
  1100. extern void skb_init(void);
  1101. extern void skb_add_mtu(int mtu);
  1102. #ifdef CONFIG_NETFILTER
  1103. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  1104. {
  1105. if (nfct && atomic_dec_and_test(&nfct->use))
  1106. nfct->destroy(nfct);
  1107. }
  1108. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  1109. {
  1110. if (nfct)
  1111. atomic_inc(&nfct->use);
  1112. }
  1113. static inline void nf_reset(struct sk_buff *skb)
  1114. {
  1115. nf_conntrack_put(skb->nfct);
  1116. skb->nfct = NULL;
  1117. }
  1118. #ifdef CONFIG_BRIDGE_NETFILTER
  1119. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  1120. {
  1121. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  1122. kfree(nf_bridge);
  1123. }
  1124. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  1125. {
  1126. if (nf_bridge)
  1127. atomic_inc(&nf_bridge->use);
  1128. }
  1129. #endif /* CONFIG_BRIDGE_NETFILTER */
  1130. #else /* CONFIG_NETFILTER */
  1131. static inline void nf_reset(struct sk_buff *skb) {}
  1132. #endif /* CONFIG_NETFILTER */
  1133. #endif /* __KERNEL__ */
  1134. #endif /* _LINUX_SKBUFF_H */