inode.c 36 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393
  1. /*
  2. * linux/fs/inode.c
  3. *
  4. * (C) 1997 Linus Torvalds
  5. */
  6. #include <linux/fs.h>
  7. #include <linux/mm.h>
  8. #include <linux/dcache.h>
  9. #include <linux/init.h>
  10. #include <linux/quotaops.h>
  11. #include <linux/slab.h>
  12. #include <linux/writeback.h>
  13. #include <linux/module.h>
  14. #include <linux/backing-dev.h>
  15. #include <linux/wait.h>
  16. #include <linux/hash.h>
  17. #include <linux/swap.h>
  18. #include <linux/security.h>
  19. #include <linux/pagemap.h>
  20. #include <linux/cdev.h>
  21. #include <linux/bootmem.h>
  22. #include <linux/inotify.h>
  23. #include <linux/mount.h>
  24. /*
  25. * This is needed for the following functions:
  26. * - inode_has_buffers
  27. * - invalidate_inode_buffers
  28. * - invalidate_bdev
  29. *
  30. * FIXME: remove all knowledge of the buffer layer from this file
  31. */
  32. #include <linux/buffer_head.h>
  33. /*
  34. * New inode.c implementation.
  35. *
  36. * This implementation has the basic premise of trying
  37. * to be extremely low-overhead and SMP-safe, yet be
  38. * simple enough to be "obviously correct".
  39. *
  40. * Famous last words.
  41. */
  42. /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
  43. /* #define INODE_PARANOIA 1 */
  44. /* #define INODE_DEBUG 1 */
  45. /*
  46. * Inode lookup is no longer as critical as it used to be:
  47. * most of the lookups are going to be through the dcache.
  48. */
  49. #define I_HASHBITS i_hash_shift
  50. #define I_HASHMASK i_hash_mask
  51. static unsigned int i_hash_mask __read_mostly;
  52. static unsigned int i_hash_shift __read_mostly;
  53. /*
  54. * Each inode can be on two separate lists. One is
  55. * the hash list of the inode, used for lookups. The
  56. * other linked list is the "type" list:
  57. * "in_use" - valid inode, i_count > 0, i_nlink > 0
  58. * "dirty" - as "in_use" but also dirty
  59. * "unused" - valid inode, i_count = 0
  60. *
  61. * A "dirty" list is maintained for each super block,
  62. * allowing for low-overhead inode sync() operations.
  63. */
  64. LIST_HEAD(inode_in_use);
  65. LIST_HEAD(inode_unused);
  66. static struct hlist_head *inode_hashtable __read_mostly;
  67. /*
  68. * A simple spinlock to protect the list manipulations.
  69. *
  70. * NOTE! You also have to own the lock if you change
  71. * the i_state of an inode while it is in use..
  72. */
  73. DEFINE_SPINLOCK(inode_lock);
  74. /*
  75. * iprune_mutex provides exclusion between the kswapd or try_to_free_pages
  76. * icache shrinking path, and the umount path. Without this exclusion,
  77. * by the time prune_icache calls iput for the inode whose pages it has
  78. * been invalidating, or by the time it calls clear_inode & destroy_inode
  79. * from its final dispose_list, the struct super_block they refer to
  80. * (for inode->i_sb->s_op) may already have been freed and reused.
  81. */
  82. static DEFINE_MUTEX(iprune_mutex);
  83. /*
  84. * Statistics gathering..
  85. */
  86. struct inodes_stat_t inodes_stat;
  87. static kmem_cache_t * inode_cachep __read_mostly;
  88. static struct inode *alloc_inode(struct super_block *sb)
  89. {
  90. static const struct address_space_operations empty_aops;
  91. static struct inode_operations empty_iops;
  92. static const struct file_operations empty_fops;
  93. struct inode *inode;
  94. if (sb->s_op->alloc_inode)
  95. inode = sb->s_op->alloc_inode(sb);
  96. else
  97. inode = (struct inode *) kmem_cache_alloc(inode_cachep, SLAB_KERNEL);
  98. if (inode) {
  99. struct address_space * const mapping = &inode->i_data;
  100. inode->i_sb = sb;
  101. inode->i_blkbits = sb->s_blocksize_bits;
  102. inode->i_flags = 0;
  103. atomic_set(&inode->i_count, 1);
  104. inode->i_op = &empty_iops;
  105. inode->i_fop = &empty_fops;
  106. inode->i_nlink = 1;
  107. atomic_set(&inode->i_writecount, 0);
  108. inode->i_size = 0;
  109. inode->i_blocks = 0;
  110. inode->i_bytes = 0;
  111. inode->i_generation = 0;
  112. #ifdef CONFIG_QUOTA
  113. memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
  114. #endif
  115. inode->i_pipe = NULL;
  116. inode->i_bdev = NULL;
  117. inode->i_cdev = NULL;
  118. inode->i_rdev = 0;
  119. inode->dirtied_when = 0;
  120. if (security_inode_alloc(inode)) {
  121. if (inode->i_sb->s_op->destroy_inode)
  122. inode->i_sb->s_op->destroy_inode(inode);
  123. else
  124. kmem_cache_free(inode_cachep, (inode));
  125. return NULL;
  126. }
  127. mapping->a_ops = &empty_aops;
  128. mapping->host = inode;
  129. mapping->flags = 0;
  130. mapping_set_gfp_mask(mapping, GFP_HIGHUSER);
  131. mapping->assoc_mapping = NULL;
  132. mapping->backing_dev_info = &default_backing_dev_info;
  133. /*
  134. * If the block_device provides a backing_dev_info for client
  135. * inodes then use that. Otherwise the inode share the bdev's
  136. * backing_dev_info.
  137. */
  138. if (sb->s_bdev) {
  139. struct backing_dev_info *bdi;
  140. bdi = sb->s_bdev->bd_inode_backing_dev_info;
  141. if (!bdi)
  142. bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
  143. mapping->backing_dev_info = bdi;
  144. }
  145. inode->i_private = 0;
  146. inode->i_mapping = mapping;
  147. }
  148. return inode;
  149. }
  150. void destroy_inode(struct inode *inode)
  151. {
  152. BUG_ON(inode_has_buffers(inode));
  153. security_inode_free(inode);
  154. if (inode->i_sb->s_op->destroy_inode)
  155. inode->i_sb->s_op->destroy_inode(inode);
  156. else
  157. kmem_cache_free(inode_cachep, (inode));
  158. }
  159. /*
  160. * These are initializations that only need to be done
  161. * once, because the fields are idempotent across use
  162. * of the inode, so let the slab aware of that.
  163. */
  164. void inode_init_once(struct inode *inode)
  165. {
  166. memset(inode, 0, sizeof(*inode));
  167. INIT_HLIST_NODE(&inode->i_hash);
  168. INIT_LIST_HEAD(&inode->i_dentry);
  169. INIT_LIST_HEAD(&inode->i_devices);
  170. mutex_init(&inode->i_mutex);
  171. init_rwsem(&inode->i_alloc_sem);
  172. INIT_RADIX_TREE(&inode->i_data.page_tree, GFP_ATOMIC);
  173. rwlock_init(&inode->i_data.tree_lock);
  174. spin_lock_init(&inode->i_data.i_mmap_lock);
  175. INIT_LIST_HEAD(&inode->i_data.private_list);
  176. spin_lock_init(&inode->i_data.private_lock);
  177. INIT_RAW_PRIO_TREE_ROOT(&inode->i_data.i_mmap);
  178. INIT_LIST_HEAD(&inode->i_data.i_mmap_nonlinear);
  179. spin_lock_init(&inode->i_lock);
  180. i_size_ordered_init(inode);
  181. #ifdef CONFIG_INOTIFY
  182. INIT_LIST_HEAD(&inode->inotify_watches);
  183. mutex_init(&inode->inotify_mutex);
  184. #endif
  185. }
  186. EXPORT_SYMBOL(inode_init_once);
  187. static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
  188. {
  189. struct inode * inode = (struct inode *) foo;
  190. if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
  191. SLAB_CTOR_CONSTRUCTOR)
  192. inode_init_once(inode);
  193. }
  194. /*
  195. * inode_lock must be held
  196. */
  197. void __iget(struct inode * inode)
  198. {
  199. if (atomic_read(&inode->i_count)) {
  200. atomic_inc(&inode->i_count);
  201. return;
  202. }
  203. atomic_inc(&inode->i_count);
  204. if (!(inode->i_state & (I_DIRTY|I_LOCK)))
  205. list_move(&inode->i_list, &inode_in_use);
  206. inodes_stat.nr_unused--;
  207. }
  208. /**
  209. * clear_inode - clear an inode
  210. * @inode: inode to clear
  211. *
  212. * This is called by the filesystem to tell us
  213. * that the inode is no longer useful. We just
  214. * terminate it with extreme prejudice.
  215. */
  216. void clear_inode(struct inode *inode)
  217. {
  218. might_sleep();
  219. invalidate_inode_buffers(inode);
  220. BUG_ON(inode->i_data.nrpages);
  221. BUG_ON(!(inode->i_state & I_FREEING));
  222. BUG_ON(inode->i_state & I_CLEAR);
  223. wait_on_inode(inode);
  224. DQUOT_DROP(inode);
  225. if (inode->i_sb && inode->i_sb->s_op->clear_inode)
  226. inode->i_sb->s_op->clear_inode(inode);
  227. if (S_ISBLK(inode->i_mode) && inode->i_bdev)
  228. bd_forget(inode);
  229. if (S_ISCHR(inode->i_mode) && inode->i_cdev)
  230. cd_forget(inode);
  231. inode->i_state = I_CLEAR;
  232. }
  233. EXPORT_SYMBOL(clear_inode);
  234. /*
  235. * dispose_list - dispose of the contents of a local list
  236. * @head: the head of the list to free
  237. *
  238. * Dispose-list gets a local list with local inodes in it, so it doesn't
  239. * need to worry about list corruption and SMP locks.
  240. */
  241. static void dispose_list(struct list_head *head)
  242. {
  243. int nr_disposed = 0;
  244. while (!list_empty(head)) {
  245. struct inode *inode;
  246. inode = list_entry(head->next, struct inode, i_list);
  247. list_del(&inode->i_list);
  248. if (inode->i_data.nrpages)
  249. truncate_inode_pages(&inode->i_data, 0);
  250. clear_inode(inode);
  251. spin_lock(&inode_lock);
  252. hlist_del_init(&inode->i_hash);
  253. list_del_init(&inode->i_sb_list);
  254. spin_unlock(&inode_lock);
  255. wake_up_inode(inode);
  256. destroy_inode(inode);
  257. nr_disposed++;
  258. }
  259. spin_lock(&inode_lock);
  260. inodes_stat.nr_inodes -= nr_disposed;
  261. spin_unlock(&inode_lock);
  262. }
  263. /*
  264. * Invalidate all inodes for a device.
  265. */
  266. static int invalidate_list(struct list_head *head, struct list_head *dispose)
  267. {
  268. struct list_head *next;
  269. int busy = 0, count = 0;
  270. next = head->next;
  271. for (;;) {
  272. struct list_head * tmp = next;
  273. struct inode * inode;
  274. /*
  275. * We can reschedule here without worrying about the list's
  276. * consistency because the per-sb list of inodes must not
  277. * change during umount anymore, and because iprune_mutex keeps
  278. * shrink_icache_memory() away.
  279. */
  280. cond_resched_lock(&inode_lock);
  281. next = next->next;
  282. if (tmp == head)
  283. break;
  284. inode = list_entry(tmp, struct inode, i_sb_list);
  285. invalidate_inode_buffers(inode);
  286. if (!atomic_read(&inode->i_count)) {
  287. list_move(&inode->i_list, dispose);
  288. inode->i_state |= I_FREEING;
  289. count++;
  290. continue;
  291. }
  292. busy = 1;
  293. }
  294. /* only unused inodes may be cached with i_count zero */
  295. inodes_stat.nr_unused -= count;
  296. return busy;
  297. }
  298. /**
  299. * invalidate_inodes - discard the inodes on a device
  300. * @sb: superblock
  301. *
  302. * Discard all of the inodes for a given superblock. If the discard
  303. * fails because there are busy inodes then a non zero value is returned.
  304. * If the discard is successful all the inodes have been discarded.
  305. */
  306. int invalidate_inodes(struct super_block * sb)
  307. {
  308. int busy;
  309. LIST_HEAD(throw_away);
  310. mutex_lock(&iprune_mutex);
  311. spin_lock(&inode_lock);
  312. inotify_unmount_inodes(&sb->s_inodes);
  313. busy = invalidate_list(&sb->s_inodes, &throw_away);
  314. spin_unlock(&inode_lock);
  315. dispose_list(&throw_away);
  316. mutex_unlock(&iprune_mutex);
  317. return busy;
  318. }
  319. EXPORT_SYMBOL(invalidate_inodes);
  320. static int can_unuse(struct inode *inode)
  321. {
  322. if (inode->i_state)
  323. return 0;
  324. if (inode_has_buffers(inode))
  325. return 0;
  326. if (atomic_read(&inode->i_count))
  327. return 0;
  328. if (inode->i_data.nrpages)
  329. return 0;
  330. return 1;
  331. }
  332. /*
  333. * Scan `goal' inodes on the unused list for freeable ones. They are moved to
  334. * a temporary list and then are freed outside inode_lock by dispose_list().
  335. *
  336. * Any inodes which are pinned purely because of attached pagecache have their
  337. * pagecache removed. We expect the final iput() on that inode to add it to
  338. * the front of the inode_unused list. So look for it there and if the
  339. * inode is still freeable, proceed. The right inode is found 99.9% of the
  340. * time in testing on a 4-way.
  341. *
  342. * If the inode has metadata buffers attached to mapping->private_list then
  343. * try to remove them.
  344. */
  345. static void prune_icache(int nr_to_scan)
  346. {
  347. LIST_HEAD(freeable);
  348. int nr_pruned = 0;
  349. int nr_scanned;
  350. unsigned long reap = 0;
  351. mutex_lock(&iprune_mutex);
  352. spin_lock(&inode_lock);
  353. for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) {
  354. struct inode *inode;
  355. if (list_empty(&inode_unused))
  356. break;
  357. inode = list_entry(inode_unused.prev, struct inode, i_list);
  358. if (inode->i_state || atomic_read(&inode->i_count)) {
  359. list_move(&inode->i_list, &inode_unused);
  360. continue;
  361. }
  362. if (inode_has_buffers(inode) || inode->i_data.nrpages) {
  363. __iget(inode);
  364. spin_unlock(&inode_lock);
  365. if (remove_inode_buffers(inode))
  366. reap += invalidate_inode_pages(&inode->i_data);
  367. iput(inode);
  368. spin_lock(&inode_lock);
  369. if (inode != list_entry(inode_unused.next,
  370. struct inode, i_list))
  371. continue; /* wrong inode or list_empty */
  372. if (!can_unuse(inode))
  373. continue;
  374. }
  375. list_move(&inode->i_list, &freeable);
  376. inode->i_state |= I_FREEING;
  377. nr_pruned++;
  378. }
  379. inodes_stat.nr_unused -= nr_pruned;
  380. if (current_is_kswapd())
  381. __count_vm_events(KSWAPD_INODESTEAL, reap);
  382. else
  383. __count_vm_events(PGINODESTEAL, reap);
  384. spin_unlock(&inode_lock);
  385. dispose_list(&freeable);
  386. mutex_unlock(&iprune_mutex);
  387. }
  388. /*
  389. * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
  390. * "unused" means that no dentries are referring to the inodes: the files are
  391. * not open and the dcache references to those inodes have already been
  392. * reclaimed.
  393. *
  394. * This function is passed the number of inodes to scan, and it returns the
  395. * total number of remaining possibly-reclaimable inodes.
  396. */
  397. static int shrink_icache_memory(int nr, gfp_t gfp_mask)
  398. {
  399. if (nr) {
  400. /*
  401. * Nasty deadlock avoidance. We may hold various FS locks,
  402. * and we don't want to recurse into the FS that called us
  403. * in clear_inode() and friends..
  404. */
  405. if (!(gfp_mask & __GFP_FS))
  406. return -1;
  407. prune_icache(nr);
  408. }
  409. return (inodes_stat.nr_unused / 100) * sysctl_vfs_cache_pressure;
  410. }
  411. static void __wait_on_freeing_inode(struct inode *inode);
  412. /*
  413. * Called with the inode lock held.
  414. * NOTE: we are not increasing the inode-refcount, you must call __iget()
  415. * by hand after calling find_inode now! This simplifies iunique and won't
  416. * add any additional branch in the common code.
  417. */
  418. static struct inode * find_inode(struct super_block * sb, struct hlist_head *head, int (*test)(struct inode *, void *), void *data)
  419. {
  420. struct hlist_node *node;
  421. struct inode * inode = NULL;
  422. repeat:
  423. hlist_for_each (node, head) {
  424. inode = hlist_entry(node, struct inode, i_hash);
  425. if (inode->i_sb != sb)
  426. continue;
  427. if (!test(inode, data))
  428. continue;
  429. if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) {
  430. __wait_on_freeing_inode(inode);
  431. goto repeat;
  432. }
  433. break;
  434. }
  435. return node ? inode : NULL;
  436. }
  437. /*
  438. * find_inode_fast is the fast path version of find_inode, see the comment at
  439. * iget_locked for details.
  440. */
  441. static struct inode * find_inode_fast(struct super_block * sb, struct hlist_head *head, unsigned long ino)
  442. {
  443. struct hlist_node *node;
  444. struct inode * inode = NULL;
  445. repeat:
  446. hlist_for_each (node, head) {
  447. inode = hlist_entry(node, struct inode, i_hash);
  448. if (inode->i_ino != ino)
  449. continue;
  450. if (inode->i_sb != sb)
  451. continue;
  452. if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) {
  453. __wait_on_freeing_inode(inode);
  454. goto repeat;
  455. }
  456. break;
  457. }
  458. return node ? inode : NULL;
  459. }
  460. /**
  461. * new_inode - obtain an inode
  462. * @sb: superblock
  463. *
  464. * Allocates a new inode for given superblock.
  465. */
  466. struct inode *new_inode(struct super_block *sb)
  467. {
  468. static unsigned long last_ino;
  469. struct inode * inode;
  470. spin_lock_prefetch(&inode_lock);
  471. inode = alloc_inode(sb);
  472. if (inode) {
  473. spin_lock(&inode_lock);
  474. inodes_stat.nr_inodes++;
  475. list_add(&inode->i_list, &inode_in_use);
  476. list_add(&inode->i_sb_list, &sb->s_inodes);
  477. inode->i_ino = ++last_ino;
  478. inode->i_state = 0;
  479. spin_unlock(&inode_lock);
  480. }
  481. return inode;
  482. }
  483. EXPORT_SYMBOL(new_inode);
  484. void unlock_new_inode(struct inode *inode)
  485. {
  486. /*
  487. * This is special! We do not need the spinlock
  488. * when clearing I_LOCK, because we're guaranteed
  489. * that nobody else tries to do anything about the
  490. * state of the inode when it is locked, as we
  491. * just created it (so there can be no old holders
  492. * that haven't tested I_LOCK).
  493. */
  494. inode->i_state &= ~(I_LOCK|I_NEW);
  495. wake_up_inode(inode);
  496. }
  497. EXPORT_SYMBOL(unlock_new_inode);
  498. /*
  499. * This is called without the inode lock held.. Be careful.
  500. *
  501. * We no longer cache the sb_flags in i_flags - see fs.h
  502. * -- rmk@arm.uk.linux.org
  503. */
  504. static struct inode * get_new_inode(struct super_block *sb, struct hlist_head *head, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *data)
  505. {
  506. struct inode * inode;
  507. inode = alloc_inode(sb);
  508. if (inode) {
  509. struct inode * old;
  510. spin_lock(&inode_lock);
  511. /* We released the lock, so.. */
  512. old = find_inode(sb, head, test, data);
  513. if (!old) {
  514. if (set(inode, data))
  515. goto set_failed;
  516. inodes_stat.nr_inodes++;
  517. list_add(&inode->i_list, &inode_in_use);
  518. list_add(&inode->i_sb_list, &sb->s_inodes);
  519. hlist_add_head(&inode->i_hash, head);
  520. inode->i_state = I_LOCK|I_NEW;
  521. spin_unlock(&inode_lock);
  522. /* Return the locked inode with I_NEW set, the
  523. * caller is responsible for filling in the contents
  524. */
  525. return inode;
  526. }
  527. /*
  528. * Uhhuh, somebody else created the same inode under
  529. * us. Use the old inode instead of the one we just
  530. * allocated.
  531. */
  532. __iget(old);
  533. spin_unlock(&inode_lock);
  534. destroy_inode(inode);
  535. inode = old;
  536. wait_on_inode(inode);
  537. }
  538. return inode;
  539. set_failed:
  540. spin_unlock(&inode_lock);
  541. destroy_inode(inode);
  542. return NULL;
  543. }
  544. /*
  545. * get_new_inode_fast is the fast path version of get_new_inode, see the
  546. * comment at iget_locked for details.
  547. */
  548. static struct inode * get_new_inode_fast(struct super_block *sb, struct hlist_head *head, unsigned long ino)
  549. {
  550. struct inode * inode;
  551. inode = alloc_inode(sb);
  552. if (inode) {
  553. struct inode * old;
  554. spin_lock(&inode_lock);
  555. /* We released the lock, so.. */
  556. old = find_inode_fast(sb, head, ino);
  557. if (!old) {
  558. inode->i_ino = ino;
  559. inodes_stat.nr_inodes++;
  560. list_add(&inode->i_list, &inode_in_use);
  561. list_add(&inode->i_sb_list, &sb->s_inodes);
  562. hlist_add_head(&inode->i_hash, head);
  563. inode->i_state = I_LOCK|I_NEW;
  564. spin_unlock(&inode_lock);
  565. /* Return the locked inode with I_NEW set, the
  566. * caller is responsible for filling in the contents
  567. */
  568. return inode;
  569. }
  570. /*
  571. * Uhhuh, somebody else created the same inode under
  572. * us. Use the old inode instead of the one we just
  573. * allocated.
  574. */
  575. __iget(old);
  576. spin_unlock(&inode_lock);
  577. destroy_inode(inode);
  578. inode = old;
  579. wait_on_inode(inode);
  580. }
  581. return inode;
  582. }
  583. static inline unsigned long hash(struct super_block *sb, unsigned long hashval)
  584. {
  585. unsigned long tmp;
  586. tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
  587. L1_CACHE_BYTES;
  588. tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
  589. return tmp & I_HASHMASK;
  590. }
  591. /**
  592. * iunique - get a unique inode number
  593. * @sb: superblock
  594. * @max_reserved: highest reserved inode number
  595. *
  596. * Obtain an inode number that is unique on the system for a given
  597. * superblock. This is used by file systems that have no natural
  598. * permanent inode numbering system. An inode number is returned that
  599. * is higher than the reserved limit but unique.
  600. *
  601. * BUGS:
  602. * With a large number of inodes live on the file system this function
  603. * currently becomes quite slow.
  604. */
  605. ino_t iunique(struct super_block *sb, ino_t max_reserved)
  606. {
  607. static ino_t counter;
  608. struct inode *inode;
  609. struct hlist_head * head;
  610. ino_t res;
  611. spin_lock(&inode_lock);
  612. retry:
  613. if (counter > max_reserved) {
  614. head = inode_hashtable + hash(sb,counter);
  615. res = counter++;
  616. inode = find_inode_fast(sb, head, res);
  617. if (!inode) {
  618. spin_unlock(&inode_lock);
  619. return res;
  620. }
  621. } else {
  622. counter = max_reserved + 1;
  623. }
  624. goto retry;
  625. }
  626. EXPORT_SYMBOL(iunique);
  627. struct inode *igrab(struct inode *inode)
  628. {
  629. spin_lock(&inode_lock);
  630. if (!(inode->i_state & (I_FREEING|I_WILL_FREE)))
  631. __iget(inode);
  632. else
  633. /*
  634. * Handle the case where s_op->clear_inode is not been
  635. * called yet, and somebody is calling igrab
  636. * while the inode is getting freed.
  637. */
  638. inode = NULL;
  639. spin_unlock(&inode_lock);
  640. return inode;
  641. }
  642. EXPORT_SYMBOL(igrab);
  643. /**
  644. * ifind - internal function, you want ilookup5() or iget5().
  645. * @sb: super block of file system to search
  646. * @head: the head of the list to search
  647. * @test: callback used for comparisons between inodes
  648. * @data: opaque data pointer to pass to @test
  649. * @wait: if true wait for the inode to be unlocked, if false do not
  650. *
  651. * ifind() searches for the inode specified by @data in the inode
  652. * cache. This is a generalized version of ifind_fast() for file systems where
  653. * the inode number is not sufficient for unique identification of an inode.
  654. *
  655. * If the inode is in the cache, the inode is returned with an incremented
  656. * reference count.
  657. *
  658. * Otherwise NULL is returned.
  659. *
  660. * Note, @test is called with the inode_lock held, so can't sleep.
  661. */
  662. static struct inode *ifind(struct super_block *sb,
  663. struct hlist_head *head, int (*test)(struct inode *, void *),
  664. void *data, const int wait)
  665. {
  666. struct inode *inode;
  667. spin_lock(&inode_lock);
  668. inode = find_inode(sb, head, test, data);
  669. if (inode) {
  670. __iget(inode);
  671. spin_unlock(&inode_lock);
  672. if (likely(wait))
  673. wait_on_inode(inode);
  674. return inode;
  675. }
  676. spin_unlock(&inode_lock);
  677. return NULL;
  678. }
  679. /**
  680. * ifind_fast - internal function, you want ilookup() or iget().
  681. * @sb: super block of file system to search
  682. * @head: head of the list to search
  683. * @ino: inode number to search for
  684. *
  685. * ifind_fast() searches for the inode @ino in the inode cache. This is for
  686. * file systems where the inode number is sufficient for unique identification
  687. * of an inode.
  688. *
  689. * If the inode is in the cache, the inode is returned with an incremented
  690. * reference count.
  691. *
  692. * Otherwise NULL is returned.
  693. */
  694. static struct inode *ifind_fast(struct super_block *sb,
  695. struct hlist_head *head, unsigned long ino)
  696. {
  697. struct inode *inode;
  698. spin_lock(&inode_lock);
  699. inode = find_inode_fast(sb, head, ino);
  700. if (inode) {
  701. __iget(inode);
  702. spin_unlock(&inode_lock);
  703. wait_on_inode(inode);
  704. return inode;
  705. }
  706. spin_unlock(&inode_lock);
  707. return NULL;
  708. }
  709. /**
  710. * ilookup5_nowait - search for an inode in the inode cache
  711. * @sb: super block of file system to search
  712. * @hashval: hash value (usually inode number) to search for
  713. * @test: callback used for comparisons between inodes
  714. * @data: opaque data pointer to pass to @test
  715. *
  716. * ilookup5() uses ifind() to search for the inode specified by @hashval and
  717. * @data in the inode cache. This is a generalized version of ilookup() for
  718. * file systems where the inode number is not sufficient for unique
  719. * identification of an inode.
  720. *
  721. * If the inode is in the cache, the inode is returned with an incremented
  722. * reference count. Note, the inode lock is not waited upon so you have to be
  723. * very careful what you do with the returned inode. You probably should be
  724. * using ilookup5() instead.
  725. *
  726. * Otherwise NULL is returned.
  727. *
  728. * Note, @test is called with the inode_lock held, so can't sleep.
  729. */
  730. struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
  731. int (*test)(struct inode *, void *), void *data)
  732. {
  733. struct hlist_head *head = inode_hashtable + hash(sb, hashval);
  734. return ifind(sb, head, test, data, 0);
  735. }
  736. EXPORT_SYMBOL(ilookup5_nowait);
  737. /**
  738. * ilookup5 - search for an inode in the inode cache
  739. * @sb: super block of file system to search
  740. * @hashval: hash value (usually inode number) to search for
  741. * @test: callback used for comparisons between inodes
  742. * @data: opaque data pointer to pass to @test
  743. *
  744. * ilookup5() uses ifind() to search for the inode specified by @hashval and
  745. * @data in the inode cache. This is a generalized version of ilookup() for
  746. * file systems where the inode number is not sufficient for unique
  747. * identification of an inode.
  748. *
  749. * If the inode is in the cache, the inode lock is waited upon and the inode is
  750. * returned with an incremented reference count.
  751. *
  752. * Otherwise NULL is returned.
  753. *
  754. * Note, @test is called with the inode_lock held, so can't sleep.
  755. */
  756. struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
  757. int (*test)(struct inode *, void *), void *data)
  758. {
  759. struct hlist_head *head = inode_hashtable + hash(sb, hashval);
  760. return ifind(sb, head, test, data, 1);
  761. }
  762. EXPORT_SYMBOL(ilookup5);
  763. /**
  764. * ilookup - search for an inode in the inode cache
  765. * @sb: super block of file system to search
  766. * @ino: inode number to search for
  767. *
  768. * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
  769. * This is for file systems where the inode number is sufficient for unique
  770. * identification of an inode.
  771. *
  772. * If the inode is in the cache, the inode is returned with an incremented
  773. * reference count.
  774. *
  775. * Otherwise NULL is returned.
  776. */
  777. struct inode *ilookup(struct super_block *sb, unsigned long ino)
  778. {
  779. struct hlist_head *head = inode_hashtable + hash(sb, ino);
  780. return ifind_fast(sb, head, ino);
  781. }
  782. EXPORT_SYMBOL(ilookup);
  783. /**
  784. * iget5_locked - obtain an inode from a mounted file system
  785. * @sb: super block of file system
  786. * @hashval: hash value (usually inode number) to get
  787. * @test: callback used for comparisons between inodes
  788. * @set: callback used to initialize a new struct inode
  789. * @data: opaque data pointer to pass to @test and @set
  790. *
  791. * This is iget() without the read_inode() portion of get_new_inode().
  792. *
  793. * iget5_locked() uses ifind() to search for the inode specified by @hashval
  794. * and @data in the inode cache and if present it is returned with an increased
  795. * reference count. This is a generalized version of iget_locked() for file
  796. * systems where the inode number is not sufficient for unique identification
  797. * of an inode.
  798. *
  799. * If the inode is not in cache, get_new_inode() is called to allocate a new
  800. * inode and this is returned locked, hashed, and with the I_NEW flag set. The
  801. * file system gets to fill it in before unlocking it via unlock_new_inode().
  802. *
  803. * Note both @test and @set are called with the inode_lock held, so can't sleep.
  804. */
  805. struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
  806. int (*test)(struct inode *, void *),
  807. int (*set)(struct inode *, void *), void *data)
  808. {
  809. struct hlist_head *head = inode_hashtable + hash(sb, hashval);
  810. struct inode *inode;
  811. inode = ifind(sb, head, test, data, 1);
  812. if (inode)
  813. return inode;
  814. /*
  815. * get_new_inode() will do the right thing, re-trying the search
  816. * in case it had to block at any point.
  817. */
  818. return get_new_inode(sb, head, test, set, data);
  819. }
  820. EXPORT_SYMBOL(iget5_locked);
  821. /**
  822. * iget_locked - obtain an inode from a mounted file system
  823. * @sb: super block of file system
  824. * @ino: inode number to get
  825. *
  826. * This is iget() without the read_inode() portion of get_new_inode_fast().
  827. *
  828. * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
  829. * the inode cache and if present it is returned with an increased reference
  830. * count. This is for file systems where the inode number is sufficient for
  831. * unique identification of an inode.
  832. *
  833. * If the inode is not in cache, get_new_inode_fast() is called to allocate a
  834. * new inode and this is returned locked, hashed, and with the I_NEW flag set.
  835. * The file system gets to fill it in before unlocking it via
  836. * unlock_new_inode().
  837. */
  838. struct inode *iget_locked(struct super_block *sb, unsigned long ino)
  839. {
  840. struct hlist_head *head = inode_hashtable + hash(sb, ino);
  841. struct inode *inode;
  842. inode = ifind_fast(sb, head, ino);
  843. if (inode)
  844. return inode;
  845. /*
  846. * get_new_inode_fast() will do the right thing, re-trying the search
  847. * in case it had to block at any point.
  848. */
  849. return get_new_inode_fast(sb, head, ino);
  850. }
  851. EXPORT_SYMBOL(iget_locked);
  852. /**
  853. * __insert_inode_hash - hash an inode
  854. * @inode: unhashed inode
  855. * @hashval: unsigned long value used to locate this object in the
  856. * inode_hashtable.
  857. *
  858. * Add an inode to the inode hash for this superblock.
  859. */
  860. void __insert_inode_hash(struct inode *inode, unsigned long hashval)
  861. {
  862. struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval);
  863. spin_lock(&inode_lock);
  864. hlist_add_head(&inode->i_hash, head);
  865. spin_unlock(&inode_lock);
  866. }
  867. EXPORT_SYMBOL(__insert_inode_hash);
  868. /**
  869. * remove_inode_hash - remove an inode from the hash
  870. * @inode: inode to unhash
  871. *
  872. * Remove an inode from the superblock.
  873. */
  874. void remove_inode_hash(struct inode *inode)
  875. {
  876. spin_lock(&inode_lock);
  877. hlist_del_init(&inode->i_hash);
  878. spin_unlock(&inode_lock);
  879. }
  880. EXPORT_SYMBOL(remove_inode_hash);
  881. /*
  882. * Tell the filesystem that this inode is no longer of any interest and should
  883. * be completely destroyed.
  884. *
  885. * We leave the inode in the inode hash table until *after* the filesystem's
  886. * ->delete_inode completes. This ensures that an iget (such as nfsd might
  887. * instigate) will always find up-to-date information either in the hash or on
  888. * disk.
  889. *
  890. * I_FREEING is set so that no-one will take a new reference to the inode while
  891. * it is being deleted.
  892. */
  893. void generic_delete_inode(struct inode *inode)
  894. {
  895. struct super_operations *op = inode->i_sb->s_op;
  896. list_del_init(&inode->i_list);
  897. list_del_init(&inode->i_sb_list);
  898. inode->i_state|=I_FREEING;
  899. inodes_stat.nr_inodes--;
  900. spin_unlock(&inode_lock);
  901. security_inode_delete(inode);
  902. if (op->delete_inode) {
  903. void (*delete)(struct inode *) = op->delete_inode;
  904. if (!is_bad_inode(inode))
  905. DQUOT_INIT(inode);
  906. /* Filesystems implementing their own
  907. * s_op->delete_inode are required to call
  908. * truncate_inode_pages and clear_inode()
  909. * internally */
  910. delete(inode);
  911. } else {
  912. truncate_inode_pages(&inode->i_data, 0);
  913. clear_inode(inode);
  914. }
  915. spin_lock(&inode_lock);
  916. hlist_del_init(&inode->i_hash);
  917. spin_unlock(&inode_lock);
  918. wake_up_inode(inode);
  919. BUG_ON(inode->i_state != I_CLEAR);
  920. destroy_inode(inode);
  921. }
  922. EXPORT_SYMBOL(generic_delete_inode);
  923. static void generic_forget_inode(struct inode *inode)
  924. {
  925. struct super_block *sb = inode->i_sb;
  926. if (!hlist_unhashed(&inode->i_hash)) {
  927. if (!(inode->i_state & (I_DIRTY|I_LOCK)))
  928. list_move(&inode->i_list, &inode_unused);
  929. inodes_stat.nr_unused++;
  930. if (!sb || (sb->s_flags & MS_ACTIVE)) {
  931. spin_unlock(&inode_lock);
  932. return;
  933. }
  934. inode->i_state |= I_WILL_FREE;
  935. spin_unlock(&inode_lock);
  936. write_inode_now(inode, 1);
  937. spin_lock(&inode_lock);
  938. inode->i_state &= ~I_WILL_FREE;
  939. inodes_stat.nr_unused--;
  940. hlist_del_init(&inode->i_hash);
  941. }
  942. list_del_init(&inode->i_list);
  943. list_del_init(&inode->i_sb_list);
  944. inode->i_state |= I_FREEING;
  945. inodes_stat.nr_inodes--;
  946. spin_unlock(&inode_lock);
  947. if (inode->i_data.nrpages)
  948. truncate_inode_pages(&inode->i_data, 0);
  949. clear_inode(inode);
  950. wake_up_inode(inode);
  951. destroy_inode(inode);
  952. }
  953. /*
  954. * Normal UNIX filesystem behaviour: delete the
  955. * inode when the usage count drops to zero, and
  956. * i_nlink is zero.
  957. */
  958. void generic_drop_inode(struct inode *inode)
  959. {
  960. if (!inode->i_nlink)
  961. generic_delete_inode(inode);
  962. else
  963. generic_forget_inode(inode);
  964. }
  965. EXPORT_SYMBOL_GPL(generic_drop_inode);
  966. /*
  967. * Called when we're dropping the last reference
  968. * to an inode.
  969. *
  970. * Call the FS "drop()" function, defaulting to
  971. * the legacy UNIX filesystem behaviour..
  972. *
  973. * NOTE! NOTE! NOTE! We're called with the inode lock
  974. * held, and the drop function is supposed to release
  975. * the lock!
  976. */
  977. static inline void iput_final(struct inode *inode)
  978. {
  979. struct super_operations *op = inode->i_sb->s_op;
  980. void (*drop)(struct inode *) = generic_drop_inode;
  981. if (op && op->drop_inode)
  982. drop = op->drop_inode;
  983. drop(inode);
  984. }
  985. /**
  986. * iput - put an inode
  987. * @inode: inode to put
  988. *
  989. * Puts an inode, dropping its usage count. If the inode use count hits
  990. * zero, the inode is then freed and may also be destroyed.
  991. *
  992. * Consequently, iput() can sleep.
  993. */
  994. void iput(struct inode *inode)
  995. {
  996. if (inode) {
  997. struct super_operations *op = inode->i_sb->s_op;
  998. BUG_ON(inode->i_state == I_CLEAR);
  999. if (op && op->put_inode)
  1000. op->put_inode(inode);
  1001. if (atomic_dec_and_lock(&inode->i_count, &inode_lock))
  1002. iput_final(inode);
  1003. }
  1004. }
  1005. EXPORT_SYMBOL(iput);
  1006. /**
  1007. * bmap - find a block number in a file
  1008. * @inode: inode of file
  1009. * @block: block to find
  1010. *
  1011. * Returns the block number on the device holding the inode that
  1012. * is the disk block number for the block of the file requested.
  1013. * That is, asked for block 4 of inode 1 the function will return the
  1014. * disk block relative to the disk start that holds that block of the
  1015. * file.
  1016. */
  1017. sector_t bmap(struct inode * inode, sector_t block)
  1018. {
  1019. sector_t res = 0;
  1020. if (inode->i_mapping->a_ops->bmap)
  1021. res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
  1022. return res;
  1023. }
  1024. EXPORT_SYMBOL(bmap);
  1025. /**
  1026. * touch_atime - update the access time
  1027. * @mnt: mount the inode is accessed on
  1028. * @dentry: dentry accessed
  1029. *
  1030. * Update the accessed time on an inode and mark it for writeback.
  1031. * This function automatically handles read only file systems and media,
  1032. * as well as the "noatime" flag and inode specific "noatime" markers.
  1033. */
  1034. void touch_atime(struct vfsmount *mnt, struct dentry *dentry)
  1035. {
  1036. struct inode *inode = dentry->d_inode;
  1037. struct timespec now;
  1038. if (IS_RDONLY(inode))
  1039. return;
  1040. if ((inode->i_flags & S_NOATIME) ||
  1041. (inode->i_sb->s_flags & MS_NOATIME) ||
  1042. ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)))
  1043. return;
  1044. /*
  1045. * We may have a NULL vfsmount when coming from NFSD
  1046. */
  1047. if (mnt &&
  1048. ((mnt->mnt_flags & MNT_NOATIME) ||
  1049. ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))))
  1050. return;
  1051. now = current_fs_time(inode->i_sb);
  1052. if (!timespec_equal(&inode->i_atime, &now)) {
  1053. inode->i_atime = now;
  1054. mark_inode_dirty_sync(inode);
  1055. }
  1056. }
  1057. EXPORT_SYMBOL(touch_atime);
  1058. /**
  1059. * file_update_time - update mtime and ctime time
  1060. * @file: file accessed
  1061. *
  1062. * Update the mtime and ctime members of an inode and mark the inode
  1063. * for writeback. Note that this function is meant exclusively for
  1064. * usage in the file write path of filesystems, and filesystems may
  1065. * choose to explicitly ignore update via this function with the
  1066. * S_NOCTIME inode flag, e.g. for network filesystem where these
  1067. * timestamps are handled by the server.
  1068. */
  1069. void file_update_time(struct file *file)
  1070. {
  1071. struct inode *inode = file->f_dentry->d_inode;
  1072. struct timespec now;
  1073. int sync_it = 0;
  1074. if (IS_NOCMTIME(inode))
  1075. return;
  1076. if (IS_RDONLY(inode))
  1077. return;
  1078. now = current_fs_time(inode->i_sb);
  1079. if (!timespec_equal(&inode->i_mtime, &now))
  1080. sync_it = 1;
  1081. inode->i_mtime = now;
  1082. if (!timespec_equal(&inode->i_ctime, &now))
  1083. sync_it = 1;
  1084. inode->i_ctime = now;
  1085. if (sync_it)
  1086. mark_inode_dirty_sync(inode);
  1087. }
  1088. EXPORT_SYMBOL(file_update_time);
  1089. int inode_needs_sync(struct inode *inode)
  1090. {
  1091. if (IS_SYNC(inode))
  1092. return 1;
  1093. if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
  1094. return 1;
  1095. return 0;
  1096. }
  1097. EXPORT_SYMBOL(inode_needs_sync);
  1098. /*
  1099. * Quota functions that want to walk the inode lists..
  1100. */
  1101. #ifdef CONFIG_QUOTA
  1102. /* Function back in dquot.c */
  1103. int remove_inode_dquot_ref(struct inode *, int, struct list_head *);
  1104. void remove_dquot_ref(struct super_block *sb, int type,
  1105. struct list_head *tofree_head)
  1106. {
  1107. struct inode *inode;
  1108. if (!sb->dq_op)
  1109. return; /* nothing to do */
  1110. spin_lock(&inode_lock); /* This lock is for inodes code */
  1111. /*
  1112. * We don't have to lock against quota code - test IS_QUOTAINIT is
  1113. * just for speedup...
  1114. */
  1115. list_for_each_entry(inode, &sb->s_inodes, i_sb_list)
  1116. if (!IS_NOQUOTA(inode))
  1117. remove_inode_dquot_ref(inode, type, tofree_head);
  1118. spin_unlock(&inode_lock);
  1119. }
  1120. #endif
  1121. int inode_wait(void *word)
  1122. {
  1123. schedule();
  1124. return 0;
  1125. }
  1126. /*
  1127. * If we try to find an inode in the inode hash while it is being
  1128. * deleted, we have to wait until the filesystem completes its
  1129. * deletion before reporting that it isn't found. This function waits
  1130. * until the deletion _might_ have completed. Callers are responsible
  1131. * to recheck inode state.
  1132. *
  1133. * It doesn't matter if I_LOCK is not set initially, a call to
  1134. * wake_up_inode() after removing from the hash list will DTRT.
  1135. *
  1136. * This is called with inode_lock held.
  1137. */
  1138. static void __wait_on_freeing_inode(struct inode *inode)
  1139. {
  1140. wait_queue_head_t *wq;
  1141. DEFINE_WAIT_BIT(wait, &inode->i_state, __I_LOCK);
  1142. wq = bit_waitqueue(&inode->i_state, __I_LOCK);
  1143. prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
  1144. spin_unlock(&inode_lock);
  1145. schedule();
  1146. finish_wait(wq, &wait.wait);
  1147. spin_lock(&inode_lock);
  1148. }
  1149. void wake_up_inode(struct inode *inode)
  1150. {
  1151. /*
  1152. * Prevent speculative execution through spin_unlock(&inode_lock);
  1153. */
  1154. smp_mb();
  1155. wake_up_bit(&inode->i_state, __I_LOCK);
  1156. }
  1157. static __initdata unsigned long ihash_entries;
  1158. static int __init set_ihash_entries(char *str)
  1159. {
  1160. if (!str)
  1161. return 0;
  1162. ihash_entries = simple_strtoul(str, &str, 0);
  1163. return 1;
  1164. }
  1165. __setup("ihash_entries=", set_ihash_entries);
  1166. /*
  1167. * Initialize the waitqueues and inode hash table.
  1168. */
  1169. void __init inode_init_early(void)
  1170. {
  1171. int loop;
  1172. /* If hashes are distributed across NUMA nodes, defer
  1173. * hash allocation until vmalloc space is available.
  1174. */
  1175. if (hashdist)
  1176. return;
  1177. inode_hashtable =
  1178. alloc_large_system_hash("Inode-cache",
  1179. sizeof(struct hlist_head),
  1180. ihash_entries,
  1181. 14,
  1182. HASH_EARLY,
  1183. &i_hash_shift,
  1184. &i_hash_mask,
  1185. 0);
  1186. for (loop = 0; loop < (1 << i_hash_shift); loop++)
  1187. INIT_HLIST_HEAD(&inode_hashtable[loop]);
  1188. }
  1189. void __init inode_init(unsigned long mempages)
  1190. {
  1191. int loop;
  1192. /* inode slab cache */
  1193. inode_cachep = kmem_cache_create("inode_cache",
  1194. sizeof(struct inode),
  1195. 0,
  1196. (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
  1197. SLAB_MEM_SPREAD),
  1198. init_once,
  1199. NULL);
  1200. set_shrinker(DEFAULT_SEEKS, shrink_icache_memory);
  1201. /* Hash may have been set up in inode_init_early */
  1202. if (!hashdist)
  1203. return;
  1204. inode_hashtable =
  1205. alloc_large_system_hash("Inode-cache",
  1206. sizeof(struct hlist_head),
  1207. ihash_entries,
  1208. 14,
  1209. 0,
  1210. &i_hash_shift,
  1211. &i_hash_mask,
  1212. 0);
  1213. for (loop = 0; loop < (1 << i_hash_shift); loop++)
  1214. INIT_HLIST_HEAD(&inode_hashtable[loop]);
  1215. }
  1216. void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
  1217. {
  1218. inode->i_mode = mode;
  1219. if (S_ISCHR(mode)) {
  1220. inode->i_fop = &def_chr_fops;
  1221. inode->i_rdev = rdev;
  1222. } else if (S_ISBLK(mode)) {
  1223. inode->i_fop = &def_blk_fops;
  1224. inode->i_rdev = rdev;
  1225. } else if (S_ISFIFO(mode))
  1226. inode->i_fop = &def_fifo_fops;
  1227. else if (S_ISSOCK(mode))
  1228. inode->i_fop = &bad_sock_fops;
  1229. else
  1230. printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o)\n",
  1231. mode);
  1232. }
  1233. EXPORT_SYMBOL(init_special_inode);