skbuff.h 34 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258
  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. unsigned char local_df,
  224. cloned:1,
  225. nohdr:1,
  226. pkt_type,
  227. ip_summed;
  228. __u32 priority;
  229. unsigned short protocol;
  230. void (*destructor)(struct sk_buff *skb);
  231. #ifdef CONFIG_NETFILTER
  232. unsigned long nfmark;
  233. __u32 nfcache;
  234. __u32 nfctinfo;
  235. struct nf_conntrack *nfct;
  236. #ifdef CONFIG_BRIDGE_NETFILTER
  237. struct nf_bridge_info *nf_bridge;
  238. #endif
  239. #endif /* CONFIG_NETFILTER */
  240. #if defined(CONFIG_HIPPI)
  241. union {
  242. __u32 ifield;
  243. } private;
  244. #endif
  245. #ifdef CONFIG_NET_SCHED
  246. __u32 tc_index; /* traffic control index */
  247. #ifdef CONFIG_NET_CLS_ACT
  248. __u32 tc_verd; /* traffic control verdict */
  249. __u32 tc_classid; /* traffic control classid */
  250. #endif
  251. #endif
  252. /* These elements must be at the end, see alloc_skb() for details. */
  253. unsigned int truesize;
  254. atomic_t users;
  255. unsigned char *head,
  256. *data,
  257. *tail,
  258. *end;
  259. };
  260. #ifdef __KERNEL__
  261. /*
  262. * Handling routines are only of interest to the kernel
  263. */
  264. #include <linux/slab.h>
  265. #include <asm/system.h>
  266. extern void __kfree_skb(struct sk_buff *skb);
  267. extern struct sk_buff *alloc_skb(unsigned int size, int priority);
  268. extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
  269. unsigned int size, int priority);
  270. extern void kfree_skbmem(struct sk_buff *skb);
  271. extern struct sk_buff *skb_clone(struct sk_buff *skb, int priority);
  272. extern struct sk_buff *skb_copy(const struct sk_buff *skb, int priority);
  273. extern struct sk_buff *pskb_copy(struct sk_buff *skb, int gfp_mask);
  274. extern int pskb_expand_head(struct sk_buff *skb,
  275. int nhead, int ntail, int gfp_mask);
  276. extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  277. unsigned int headroom);
  278. extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
  279. int newheadroom, int newtailroom,
  280. int priority);
  281. extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
  282. #define dev_kfree_skb(a) kfree_skb(a)
  283. extern void skb_over_panic(struct sk_buff *skb, int len,
  284. void *here);
  285. extern void skb_under_panic(struct sk_buff *skb, int len,
  286. void *here);
  287. struct skb_seq_state
  288. {
  289. __u32 lower_offset;
  290. __u32 upper_offset;
  291. __u32 frag_idx;
  292. __u32 stepped_offset;
  293. struct sk_buff *root_skb;
  294. struct sk_buff *cur_skb;
  295. __u8 *frag_data;
  296. };
  297. extern void skb_prepare_seq_read(struct sk_buff *skb,
  298. unsigned int from, unsigned int to,
  299. struct skb_seq_state *st);
  300. extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  301. struct skb_seq_state *st);
  302. extern void skb_abort_seq_read(struct skb_seq_state *st);
  303. extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  304. unsigned int to, struct ts_config *config,
  305. struct ts_state *state);
  306. /* Internal */
  307. #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
  308. /**
  309. * skb_queue_empty - check if a queue is empty
  310. * @list: queue head
  311. *
  312. * Returns true if the queue is empty, false otherwise.
  313. */
  314. static inline int skb_queue_empty(const struct sk_buff_head *list)
  315. {
  316. return list->next == (struct sk_buff *)list;
  317. }
  318. /**
  319. * skb_get - reference buffer
  320. * @skb: buffer to reference
  321. *
  322. * Makes another reference to a socket buffer and returns a pointer
  323. * to the buffer.
  324. */
  325. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  326. {
  327. atomic_inc(&skb->users);
  328. return skb;
  329. }
  330. /*
  331. * If users == 1, we are the only owner and are can avoid redundant
  332. * atomic change.
  333. */
  334. /**
  335. * kfree_skb - free an sk_buff
  336. * @skb: buffer to free
  337. *
  338. * Drop a reference to the buffer and free it if the usage count has
  339. * hit zero.
  340. */
  341. static inline void kfree_skb(struct sk_buff *skb)
  342. {
  343. if (likely(atomic_read(&skb->users) == 1))
  344. smp_rmb();
  345. else if (likely(!atomic_dec_and_test(&skb->users)))
  346. return;
  347. __kfree_skb(skb);
  348. }
  349. /**
  350. * skb_cloned - is the buffer a clone
  351. * @skb: buffer to check
  352. *
  353. * Returns true if the buffer was generated with skb_clone() and is
  354. * one of multiple shared copies of the buffer. Cloned buffers are
  355. * shared data so must not be written to under normal circumstances.
  356. */
  357. static inline int skb_cloned(const struct sk_buff *skb)
  358. {
  359. return skb->cloned &&
  360. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  361. }
  362. /**
  363. * skb_header_cloned - is the header a clone
  364. * @skb: buffer to check
  365. *
  366. * Returns true if modifying the header part of the buffer requires
  367. * the data to be copied.
  368. */
  369. static inline int skb_header_cloned(const struct sk_buff *skb)
  370. {
  371. int dataref;
  372. if (!skb->cloned)
  373. return 0;
  374. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  375. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  376. return dataref != 1;
  377. }
  378. /**
  379. * skb_header_release - release reference to header
  380. * @skb: buffer to operate on
  381. *
  382. * Drop a reference to the header part of the buffer. This is done
  383. * by acquiring a payload reference. You must not read from the header
  384. * part of skb->data after this.
  385. */
  386. static inline void skb_header_release(struct sk_buff *skb)
  387. {
  388. BUG_ON(skb->nohdr);
  389. skb->nohdr = 1;
  390. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  391. }
  392. /**
  393. * skb_shared - is the buffer shared
  394. * @skb: buffer to check
  395. *
  396. * Returns true if more than one person has a reference to this
  397. * buffer.
  398. */
  399. static inline int skb_shared(const struct sk_buff *skb)
  400. {
  401. return atomic_read(&skb->users) != 1;
  402. }
  403. /**
  404. * skb_share_check - check if buffer is shared and if so clone it
  405. * @skb: buffer to check
  406. * @pri: priority for memory allocation
  407. *
  408. * If the buffer is shared the buffer is cloned and the old copy
  409. * drops a reference. A new clone with a single reference is returned.
  410. * If the buffer is not shared the original buffer is returned. When
  411. * being called from interrupt status or with spinlocks held pri must
  412. * be GFP_ATOMIC.
  413. *
  414. * NULL is returned on a memory allocation failure.
  415. */
  416. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, int pri)
  417. {
  418. might_sleep_if(pri & __GFP_WAIT);
  419. if (skb_shared(skb)) {
  420. struct sk_buff *nskb = skb_clone(skb, pri);
  421. kfree_skb(skb);
  422. skb = nskb;
  423. }
  424. return skb;
  425. }
  426. /*
  427. * Copy shared buffers into a new sk_buff. We effectively do COW on
  428. * packets to handle cases where we have a local reader and forward
  429. * and a couple of other messy ones. The normal one is tcpdumping
  430. * a packet thats being forwarded.
  431. */
  432. /**
  433. * skb_unshare - make a copy of a shared buffer
  434. * @skb: buffer to check
  435. * @pri: priority for memory allocation
  436. *
  437. * If the socket buffer is a clone then this function creates a new
  438. * copy of the data, drops a reference count on the old copy and returns
  439. * the new copy with the reference count at 1. If the buffer is not a clone
  440. * the original buffer is returned. When called with a spinlock held or
  441. * from interrupt state @pri must be %GFP_ATOMIC
  442. *
  443. * %NULL is returned on a memory allocation failure.
  444. */
  445. static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
  446. {
  447. might_sleep_if(pri & __GFP_WAIT);
  448. if (skb_cloned(skb)) {
  449. struct sk_buff *nskb = skb_copy(skb, pri);
  450. kfree_skb(skb); /* Free our shared copy */
  451. skb = nskb;
  452. }
  453. return skb;
  454. }
  455. /**
  456. * skb_peek
  457. * @list_: list to peek at
  458. *
  459. * Peek an &sk_buff. Unlike most other operations you _MUST_
  460. * be careful with this one. A peek leaves the buffer on the
  461. * list and someone else may run off with it. You must hold
  462. * the appropriate locks or have a private queue to do this.
  463. *
  464. * Returns %NULL for an empty list or a pointer to the head element.
  465. * The reference count is not incremented and the reference is therefore
  466. * volatile. Use with caution.
  467. */
  468. static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
  469. {
  470. struct sk_buff *list = ((struct sk_buff *)list_)->next;
  471. if (list == (struct sk_buff *)list_)
  472. list = NULL;
  473. return list;
  474. }
  475. /**
  476. * skb_peek_tail
  477. * @list_: list to peek at
  478. *
  479. * Peek an &sk_buff. Unlike most other operations you _MUST_
  480. * be careful with this one. A peek leaves the buffer on the
  481. * list and someone else may run off with it. You must hold
  482. * the appropriate locks or have a private queue to do this.
  483. *
  484. * Returns %NULL for an empty list or a pointer to the tail element.
  485. * The reference count is not incremented and the reference is therefore
  486. * volatile. Use with caution.
  487. */
  488. static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
  489. {
  490. struct sk_buff *list = ((struct sk_buff *)list_)->prev;
  491. if (list == (struct sk_buff *)list_)
  492. list = NULL;
  493. return list;
  494. }
  495. /**
  496. * skb_queue_len - get queue length
  497. * @list_: list to measure
  498. *
  499. * Return the length of an &sk_buff queue.
  500. */
  501. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  502. {
  503. return list_->qlen;
  504. }
  505. static inline void skb_queue_head_init(struct sk_buff_head *list)
  506. {
  507. spin_lock_init(&list->lock);
  508. list->prev = list->next = (struct sk_buff *)list;
  509. list->qlen = 0;
  510. }
  511. /*
  512. * Insert an sk_buff at the start of a list.
  513. *
  514. * The "__skb_xxxx()" functions are the non-atomic ones that
  515. * can only be called with interrupts disabled.
  516. */
  517. /**
  518. * __skb_queue_head - queue a buffer at the list head
  519. * @list: list to use
  520. * @newsk: buffer to queue
  521. *
  522. * Queue a buffer at the start of a list. This function takes no locks
  523. * and you must therefore hold required locks before calling it.
  524. *
  525. * A buffer cannot be placed on two lists at the same time.
  526. */
  527. extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  528. static inline void __skb_queue_head(struct sk_buff_head *list,
  529. struct sk_buff *newsk)
  530. {
  531. struct sk_buff *prev, *next;
  532. newsk->list = list;
  533. list->qlen++;
  534. prev = (struct sk_buff *)list;
  535. next = prev->next;
  536. newsk->next = next;
  537. newsk->prev = prev;
  538. next->prev = prev->next = newsk;
  539. }
  540. /**
  541. * __skb_queue_tail - queue a buffer at the list tail
  542. * @list: list to use
  543. * @newsk: buffer to queue
  544. *
  545. * Queue a buffer at the end of a list. This function takes no locks
  546. * and you must therefore hold required locks before calling it.
  547. *
  548. * A buffer cannot be placed on two lists at the same time.
  549. */
  550. extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  551. static inline void __skb_queue_tail(struct sk_buff_head *list,
  552. struct sk_buff *newsk)
  553. {
  554. struct sk_buff *prev, *next;
  555. newsk->list = list;
  556. list->qlen++;
  557. next = (struct sk_buff *)list;
  558. prev = next->prev;
  559. newsk->next = next;
  560. newsk->prev = prev;
  561. next->prev = prev->next = newsk;
  562. }
  563. /**
  564. * __skb_dequeue - remove from the head of the queue
  565. * @list: list to dequeue from
  566. *
  567. * Remove the head of the list. This function does not take any locks
  568. * so must be used with appropriate locks held only. The head item is
  569. * returned or %NULL if the list is empty.
  570. */
  571. extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  572. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  573. {
  574. struct sk_buff *next, *prev, *result;
  575. prev = (struct sk_buff *) list;
  576. next = prev->next;
  577. result = NULL;
  578. if (next != prev) {
  579. result = next;
  580. next = next->next;
  581. list->qlen--;
  582. next->prev = prev;
  583. prev->next = next;
  584. result->next = result->prev = NULL;
  585. result->list = NULL;
  586. }
  587. return result;
  588. }
  589. /*
  590. * Insert a packet on a list.
  591. */
  592. extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
  593. static inline void __skb_insert(struct sk_buff *newsk,
  594. struct sk_buff *prev, struct sk_buff *next,
  595. struct sk_buff_head *list)
  596. {
  597. newsk->next = next;
  598. newsk->prev = prev;
  599. next->prev = prev->next = newsk;
  600. newsk->list = list;
  601. list->qlen++;
  602. }
  603. /*
  604. * Place a packet after a given packet in a list.
  605. */
  606. extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
  607. static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
  608. {
  609. __skb_insert(newsk, old, old->next, old->list);
  610. }
  611. /*
  612. * remove sk_buff from list. _Must_ be called atomically, and with
  613. * the list known..
  614. */
  615. extern void skb_unlink(struct sk_buff *skb);
  616. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  617. {
  618. struct sk_buff *next, *prev;
  619. list->qlen--;
  620. next = skb->next;
  621. prev = skb->prev;
  622. skb->next = skb->prev = NULL;
  623. skb->list = NULL;
  624. next->prev = prev;
  625. prev->next = next;
  626. }
  627. /* XXX: more streamlined implementation */
  628. /**
  629. * __skb_dequeue_tail - remove from the tail of the queue
  630. * @list: list to dequeue from
  631. *
  632. * Remove the tail of the list. This function does not take any locks
  633. * so must be used with appropriate locks held only. The tail item is
  634. * returned or %NULL if the list is empty.
  635. */
  636. extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  637. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  638. {
  639. struct sk_buff *skb = skb_peek_tail(list);
  640. if (skb)
  641. __skb_unlink(skb, list);
  642. return skb;
  643. }
  644. static inline int skb_is_nonlinear(const struct sk_buff *skb)
  645. {
  646. return skb->data_len;
  647. }
  648. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  649. {
  650. return skb->len - skb->data_len;
  651. }
  652. static inline int skb_pagelen(const struct sk_buff *skb)
  653. {
  654. int i, len = 0;
  655. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  656. len += skb_shinfo(skb)->frags[i].size;
  657. return len + skb_headlen(skb);
  658. }
  659. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  660. struct page *page, int off, int size)
  661. {
  662. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  663. frag->page = page;
  664. frag->page_offset = off;
  665. frag->size = size;
  666. skb_shinfo(skb)->nr_frags = i + 1;
  667. }
  668. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  669. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
  670. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  671. /*
  672. * Add data to an sk_buff
  673. */
  674. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  675. {
  676. unsigned char *tmp = skb->tail;
  677. SKB_LINEAR_ASSERT(skb);
  678. skb->tail += len;
  679. skb->len += len;
  680. return tmp;
  681. }
  682. /**
  683. * skb_put - add data to a buffer
  684. * @skb: buffer to use
  685. * @len: amount of data to add
  686. *
  687. * This function extends the used data area of the buffer. If this would
  688. * exceed the total buffer size the kernel will panic. A pointer to the
  689. * first byte of the extra data is returned.
  690. */
  691. static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
  692. {
  693. unsigned char *tmp = skb->tail;
  694. SKB_LINEAR_ASSERT(skb);
  695. skb->tail += len;
  696. skb->len += len;
  697. if (unlikely(skb->tail>skb->end))
  698. skb_over_panic(skb, len, current_text_addr());
  699. return tmp;
  700. }
  701. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  702. {
  703. skb->data -= len;
  704. skb->len += len;
  705. return skb->data;
  706. }
  707. /**
  708. * skb_push - add data to the start of a buffer
  709. * @skb: buffer to use
  710. * @len: amount of data to add
  711. *
  712. * This function extends the used data area of the buffer at the buffer
  713. * start. If this would exceed the total buffer headroom the kernel will
  714. * panic. A pointer to the first byte of the extra data is returned.
  715. */
  716. static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
  717. {
  718. skb->data -= len;
  719. skb->len += len;
  720. if (unlikely(skb->data<skb->head))
  721. skb_under_panic(skb, len, current_text_addr());
  722. return skb->data;
  723. }
  724. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  725. {
  726. skb->len -= len;
  727. BUG_ON(skb->len < skb->data_len);
  728. return skb->data += len;
  729. }
  730. /**
  731. * skb_pull - remove data from the start of a buffer
  732. * @skb: buffer to use
  733. * @len: amount of data to remove
  734. *
  735. * This function removes data from the start of a buffer, returning
  736. * the memory to the headroom. A pointer to the next data in the buffer
  737. * is returned. Once the data has been pulled future pushes will overwrite
  738. * the old data.
  739. */
  740. static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
  741. {
  742. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  743. }
  744. extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  745. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  746. {
  747. if (len > skb_headlen(skb) &&
  748. !__pskb_pull_tail(skb, len-skb_headlen(skb)))
  749. return NULL;
  750. skb->len -= len;
  751. return skb->data += len;
  752. }
  753. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  754. {
  755. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  756. }
  757. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  758. {
  759. if (likely(len <= skb_headlen(skb)))
  760. return 1;
  761. if (unlikely(len > skb->len))
  762. return 0;
  763. return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
  764. }
  765. /**
  766. * skb_headroom - bytes at buffer head
  767. * @skb: buffer to check
  768. *
  769. * Return the number of bytes of free space at the head of an &sk_buff.
  770. */
  771. static inline int skb_headroom(const struct sk_buff *skb)
  772. {
  773. return skb->data - skb->head;
  774. }
  775. /**
  776. * skb_tailroom - bytes at buffer end
  777. * @skb: buffer to check
  778. *
  779. * Return the number of bytes of free space at the tail of an sk_buff
  780. */
  781. static inline int skb_tailroom(const struct sk_buff *skb)
  782. {
  783. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  784. }
  785. /**
  786. * skb_reserve - adjust headroom
  787. * @skb: buffer to alter
  788. * @len: bytes to move
  789. *
  790. * Increase the headroom of an empty &sk_buff by reducing the tail
  791. * room. This is only allowed for an empty buffer.
  792. */
  793. static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
  794. {
  795. skb->data += len;
  796. skb->tail += len;
  797. }
  798. /*
  799. * CPUs often take a performance hit when accessing unaligned memory
  800. * locations. The actual performance hit varies, it can be small if the
  801. * hardware handles it or large if we have to take an exception and fix it
  802. * in software.
  803. *
  804. * Since an ethernet header is 14 bytes network drivers often end up with
  805. * the IP header at an unaligned offset. The IP header can be aligned by
  806. * shifting the start of the packet by 2 bytes. Drivers should do this
  807. * with:
  808. *
  809. * skb_reserve(NET_IP_ALIGN);
  810. *
  811. * The downside to this alignment of the IP header is that the DMA is now
  812. * unaligned. On some architectures the cost of an unaligned DMA is high
  813. * and this cost outweighs the gains made by aligning the IP header.
  814. *
  815. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  816. * to be overridden.
  817. */
  818. #ifndef NET_IP_ALIGN
  819. #define NET_IP_ALIGN 2
  820. #endif
  821. extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
  822. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  823. {
  824. if (!skb->data_len) {
  825. skb->len = len;
  826. skb->tail = skb->data + len;
  827. } else
  828. ___pskb_trim(skb, len, 0);
  829. }
  830. /**
  831. * skb_trim - remove end from a buffer
  832. * @skb: buffer to alter
  833. * @len: new length
  834. *
  835. * Cut the length of a buffer down by removing data from the tail. If
  836. * the buffer is already under the length specified it is not modified.
  837. */
  838. static inline void skb_trim(struct sk_buff *skb, unsigned int len)
  839. {
  840. if (skb->len > len)
  841. __skb_trim(skb, len);
  842. }
  843. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  844. {
  845. if (!skb->data_len) {
  846. skb->len = len;
  847. skb->tail = skb->data+len;
  848. return 0;
  849. }
  850. return ___pskb_trim(skb, len, 1);
  851. }
  852. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  853. {
  854. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  855. }
  856. /**
  857. * skb_orphan - orphan a buffer
  858. * @skb: buffer to orphan
  859. *
  860. * If a buffer currently has an owner then we call the owner's
  861. * destructor function and make the @skb unowned. The buffer continues
  862. * to exist but is no longer charged to its former owner.
  863. */
  864. static inline void skb_orphan(struct sk_buff *skb)
  865. {
  866. if (skb->destructor)
  867. skb->destructor(skb);
  868. skb->destructor = NULL;
  869. skb->sk = NULL;
  870. }
  871. /**
  872. * __skb_queue_purge - empty a list
  873. * @list: list to empty
  874. *
  875. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  876. * the list and one reference dropped. This function does not take the
  877. * list lock and the caller must hold the relevant locks to use it.
  878. */
  879. extern void skb_queue_purge(struct sk_buff_head *list);
  880. static inline void __skb_queue_purge(struct sk_buff_head *list)
  881. {
  882. struct sk_buff *skb;
  883. while ((skb = __skb_dequeue(list)) != NULL)
  884. kfree_skb(skb);
  885. }
  886. #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
  887. /**
  888. * __dev_alloc_skb - allocate an skbuff for sending
  889. * @length: length to allocate
  890. * @gfp_mask: get_free_pages mask, passed to alloc_skb
  891. *
  892. * Allocate a new &sk_buff and assign it a usage count of one. The
  893. * buffer has unspecified headroom built in. Users should allocate
  894. * the headroom they think they need without accounting for the
  895. * built in space. The built in space is used for optimisations.
  896. *
  897. * %NULL is returned in there is no free memory.
  898. */
  899. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  900. int gfp_mask)
  901. {
  902. struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
  903. if (likely(skb))
  904. skb_reserve(skb, 16);
  905. return skb;
  906. }
  907. #else
  908. extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
  909. #endif
  910. /**
  911. * dev_alloc_skb - allocate an skbuff for sending
  912. * @length: length to allocate
  913. *
  914. * Allocate a new &sk_buff and assign it a usage count of one. The
  915. * buffer has unspecified headroom built in. Users should allocate
  916. * the headroom they think they need without accounting for the
  917. * built in space. The built in space is used for optimisations.
  918. *
  919. * %NULL is returned in there is no free memory. Although this function
  920. * allocates memory it can be called from an interrupt.
  921. */
  922. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  923. {
  924. return __dev_alloc_skb(length, GFP_ATOMIC);
  925. }
  926. /**
  927. * skb_cow - copy header of skb when it is required
  928. * @skb: buffer to cow
  929. * @headroom: needed headroom
  930. *
  931. * If the skb passed lacks sufficient headroom or its data part
  932. * is shared, data is reallocated. If reallocation fails, an error
  933. * is returned and original skb is not changed.
  934. *
  935. * The result is skb with writable area skb->head...skb->tail
  936. * and at least @headroom of space at head.
  937. */
  938. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  939. {
  940. int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
  941. if (delta < 0)
  942. delta = 0;
  943. if (delta || skb_cloned(skb))
  944. return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
  945. return 0;
  946. }
  947. /**
  948. * skb_padto - pad an skbuff up to a minimal size
  949. * @skb: buffer to pad
  950. * @len: minimal length
  951. *
  952. * Pads up a buffer to ensure the trailing bytes exist and are
  953. * blanked. If the buffer already contains sufficient data it
  954. * is untouched. Returns the buffer, which may be a replacement
  955. * for the original, or NULL for out of memory - in which case
  956. * the original buffer is still freed.
  957. */
  958. static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
  959. {
  960. unsigned int size = skb->len;
  961. if (likely(size >= len))
  962. return skb;
  963. return skb_pad(skb, len-size);
  964. }
  965. static inline int skb_add_data(struct sk_buff *skb,
  966. char __user *from, int copy)
  967. {
  968. const int off = skb->len;
  969. if (skb->ip_summed == CHECKSUM_NONE) {
  970. int err = 0;
  971. unsigned int csum = csum_and_copy_from_user(from,
  972. skb_put(skb, copy),
  973. copy, 0, &err);
  974. if (!err) {
  975. skb->csum = csum_block_add(skb->csum, csum, off);
  976. return 0;
  977. }
  978. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  979. return 0;
  980. __skb_trim(skb, off);
  981. return -EFAULT;
  982. }
  983. static inline int skb_can_coalesce(struct sk_buff *skb, int i,
  984. struct page *page, int off)
  985. {
  986. if (i) {
  987. struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  988. return page == frag->page &&
  989. off == frag->page_offset + frag->size;
  990. }
  991. return 0;
  992. }
  993. /**
  994. * skb_linearize - convert paged skb to linear one
  995. * @skb: buffer to linarize
  996. * @gfp: allocation mode
  997. *
  998. * If there is no free memory -ENOMEM is returned, otherwise zero
  999. * is returned and the old skb data released.
  1000. */
  1001. extern int __skb_linearize(struct sk_buff *skb, int gfp);
  1002. static inline int skb_linearize(struct sk_buff *skb, int gfp)
  1003. {
  1004. return __skb_linearize(skb, gfp);
  1005. }
  1006. /**
  1007. * skb_postpull_rcsum - update checksum for received skb after pull
  1008. * @skb: buffer to update
  1009. * @start: start of data before pull
  1010. * @len: length of data pulled
  1011. *
  1012. * After doing a pull on a received packet, you need to call this to
  1013. * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
  1014. * so that it can be recomputed from scratch.
  1015. */
  1016. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  1017. const void *start, int len)
  1018. {
  1019. if (skb->ip_summed == CHECKSUM_HW)
  1020. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  1021. }
  1022. /**
  1023. * pskb_trim_rcsum - trim received skb and update checksum
  1024. * @skb: buffer to trim
  1025. * @len: new length
  1026. *
  1027. * This is exactly the same as pskb_trim except that it ensures the
  1028. * checksum of received packets are still valid after the operation.
  1029. */
  1030. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  1031. {
  1032. if (len >= skb->len)
  1033. return 0;
  1034. if (skb->ip_summed == CHECKSUM_HW)
  1035. skb->ip_summed = CHECKSUM_NONE;
  1036. return __pskb_trim(skb, len);
  1037. }
  1038. static inline void *kmap_skb_frag(const skb_frag_t *frag)
  1039. {
  1040. #ifdef CONFIG_HIGHMEM
  1041. BUG_ON(in_irq());
  1042. local_bh_disable();
  1043. #endif
  1044. return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
  1045. }
  1046. static inline void kunmap_skb_frag(void *vaddr)
  1047. {
  1048. kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
  1049. #ifdef CONFIG_HIGHMEM
  1050. local_bh_enable();
  1051. #endif
  1052. }
  1053. #define skb_queue_walk(queue, skb) \
  1054. for (skb = (queue)->next; \
  1055. prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
  1056. skb = skb->next)
  1057. extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
  1058. int noblock, int *err);
  1059. extern unsigned int datagram_poll(struct file *file, struct socket *sock,
  1060. struct poll_table_struct *wait);
  1061. extern int skb_copy_datagram_iovec(const struct sk_buff *from,
  1062. int offset, struct iovec *to,
  1063. int size);
  1064. extern int skb_copy_and_csum_datagram_iovec(const
  1065. struct sk_buff *skb,
  1066. int hlen,
  1067. struct iovec *iov);
  1068. extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  1069. extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
  1070. int len, unsigned int csum);
  1071. extern int skb_copy_bits(const struct sk_buff *skb, int offset,
  1072. void *to, int len);
  1073. extern int skb_store_bits(const struct sk_buff *skb, int offset,
  1074. void *from, int len);
  1075. extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
  1076. int offset, u8 *to, int len,
  1077. unsigned int csum);
  1078. extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  1079. extern void skb_split(struct sk_buff *skb,
  1080. struct sk_buff *skb1, const u32 len);
  1081. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  1082. int len, void *buffer)
  1083. {
  1084. int hlen = skb_headlen(skb);
  1085. if (hlen - offset >= len)
  1086. return skb->data + offset;
  1087. if (skb_copy_bits(skb, offset, buffer, len) < 0)
  1088. return NULL;
  1089. return buffer;
  1090. }
  1091. extern void skb_init(void);
  1092. extern void skb_add_mtu(int mtu);
  1093. #ifdef CONFIG_NETFILTER
  1094. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  1095. {
  1096. if (nfct && atomic_dec_and_test(&nfct->use))
  1097. nfct->destroy(nfct);
  1098. }
  1099. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  1100. {
  1101. if (nfct)
  1102. atomic_inc(&nfct->use);
  1103. }
  1104. static inline void nf_reset(struct sk_buff *skb)
  1105. {
  1106. nf_conntrack_put(skb->nfct);
  1107. skb->nfct = NULL;
  1108. }
  1109. #ifdef CONFIG_BRIDGE_NETFILTER
  1110. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  1111. {
  1112. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  1113. kfree(nf_bridge);
  1114. }
  1115. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  1116. {
  1117. if (nf_bridge)
  1118. atomic_inc(&nf_bridge->use);
  1119. }
  1120. #endif /* CONFIG_BRIDGE_NETFILTER */
  1121. #else /* CONFIG_NETFILTER */
  1122. static inline void nf_reset(struct sk_buff *skb) {}
  1123. #endif /* CONFIG_NETFILTER */
  1124. #endif /* __KERNEL__ */
  1125. #endif /* _LINUX_SKBUFF_H */