blob: 0571983755dcdfa8b86ca89d5ccc2db1ed213027 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/inode.c
3 *
4 * (C) 1997 Linus Torvalds
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/fs.h>
8#include <linux/mm.h>
9#include <linux/dcache.h>
10#include <linux/init.h>
11#include <linux/quotaops.h>
12#include <linux/slab.h>
13#include <linux/writeback.h>
14#include <linux/module.h>
15#include <linux/backing-dev.h>
16#include <linux/wait.h>
17#include <linux/hash.h>
18#include <linux/swap.h>
19#include <linux/security.h>
Mimi Zohar6146f0d2009-02-04 09:06:57 -050020#include <linux/ima.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/pagemap.h>
22#include <linux/cdev.h>
23#include <linux/bootmem.h>
Robert Love0eeca282005-07-12 17:06:03 -040024#include <linux/inotify.h>
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -080025#include <linux/mount.h>
Arjan van de Venefaee192009-01-06 07:20:54 -080026#include <linux/async.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
28/*
29 * This is needed for the following functions:
30 * - inode_has_buffers
31 * - invalidate_inode_buffers
Linus Torvalds1da177e2005-04-16 15:20:36 -070032 * - invalidate_bdev
33 *
34 * FIXME: remove all knowledge of the buffer layer from this file
35 */
36#include <linux/buffer_head.h>
37
38/*
39 * New inode.c implementation.
40 *
41 * This implementation has the basic premise of trying
42 * to be extremely low-overhead and SMP-safe, yet be
43 * simple enough to be "obviously correct".
44 *
45 * Famous last words.
46 */
47
48/* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
49
50/* #define INODE_PARANOIA 1 */
51/* #define INODE_DEBUG 1 */
52
53/*
54 * Inode lookup is no longer as critical as it used to be:
55 * most of the lookups are going to be through the dcache.
56 */
57#define I_HASHBITS i_hash_shift
58#define I_HASHMASK i_hash_mask
59
Eric Dumazetfa3536c2006-03-26 01:37:24 -080060static unsigned int i_hash_mask __read_mostly;
61static unsigned int i_hash_shift __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63/*
64 * Each inode can be on two separate lists. One is
65 * the hash list of the inode, used for lookups. The
66 * other linked list is the "type" list:
67 * "in_use" - valid inode, i_count > 0, i_nlink > 0
68 * "dirty" - as "in_use" but also dirty
69 * "unused" - valid inode, i_count = 0
70 *
71 * A "dirty" list is maintained for each super block,
72 * allowing for low-overhead inode sync() operations.
73 */
74
75LIST_HEAD(inode_in_use);
76LIST_HEAD(inode_unused);
Eric Dumazetfa3536c2006-03-26 01:37:24 -080077static struct hlist_head *inode_hashtable __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
79/*
80 * A simple spinlock to protect the list manipulations.
81 *
82 * NOTE! You also have to own the lock if you change
83 * the i_state of an inode while it is in use..
84 */
85DEFINE_SPINLOCK(inode_lock);
86
87/*
Ingo Molnarf24075b2006-03-23 03:00:34 -080088 * iprune_mutex provides exclusion between the kswapd or try_to_free_pages
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 * icache shrinking path, and the umount path. Without this exclusion,
90 * by the time prune_icache calls iput for the inode whose pages it has
91 * been invalidating, or by the time it calls clear_inode & destroy_inode
92 * from its final dispose_list, the struct super_block they refer to
93 * (for inode->i_sb->s_op) may already have been freed and reused.
94 */
Adrian Bunkbdfc3262006-03-25 03:06:56 -080095static DEFINE_MUTEX(iprune_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97/*
98 * Statistics gathering..
99 */
100struct inodes_stat_t inodes_stat;
101
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530102static struct kmem_cache *inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
Joern Engel1c0eeaf2007-10-16 23:30:44 -0700104static void wake_up_inode(struct inode *inode)
105{
106 /*
107 * Prevent speculative execution through spin_unlock(&inode_lock);
108 */
109 smp_mb();
110 wake_up_bit(&inode->i_state, __I_LOCK);
111}
112
David Chinner2cb15992008-10-30 17:32:23 +1100113/**
114 * inode_init_always - perform inode structure intialisation
Randy Dunlap0bc02f32009-01-06 14:41:13 -0800115 * @sb: superblock inode belongs to
116 * @inode: inode to initialise
David Chinner2cb15992008-10-30 17:32:23 +1100117 *
118 * These are initializations that need to be done on every inode
119 * allocation as the fields are not initialised by slab allocation.
120 */
121struct inode *inode_init_always(struct super_block *sb, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122{
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700123 static const struct address_space_operations empty_aops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 static struct inode_operations empty_iops;
Arjan van de Ven99ac48f2006-03-28 01:56:41 -0800125 static const struct file_operations empty_fops;
David Chinner2cb15992008-10-30 17:32:23 +1100126
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530127 struct address_space *const mapping = &inode->i_data;
David Chinner2cb15992008-10-30 17:32:23 +1100128
129 inode->i_sb = sb;
130 inode->i_blkbits = sb->s_blocksize_bits;
131 inode->i_flags = 0;
132 atomic_set(&inode->i_count, 1);
133 inode->i_op = &empty_iops;
134 inode->i_fop = &empty_fops;
135 inode->i_nlink = 1;
Al Viro56ff5ef2008-12-09 09:34:39 -0500136 inode->i_uid = 0;
137 inode->i_gid = 0;
David Chinner2cb15992008-10-30 17:32:23 +1100138 atomic_set(&inode->i_writecount, 0);
139 inode->i_size = 0;
140 inode->i_blocks = 0;
141 inode->i_bytes = 0;
142 inode->i_generation = 0;
143#ifdef CONFIG_QUOTA
144 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
145#endif
146 inode->i_pipe = NULL;
147 inode->i_bdev = NULL;
148 inode->i_cdev = NULL;
149 inode->i_rdev = 0;
150 inode->dirtied_when = 0;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500151
152 if (security_inode_alloc(inode))
153 goto out_free_inode;
154
155 /* allocate and initialize an i_integrity */
156 if (ima_inode_alloc(inode))
157 goto out_free_security;
David Chinner2cb15992008-10-30 17:32:23 +1100158
159 spin_lock_init(&inode->i_lock);
160 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
161
162 mutex_init(&inode->i_mutex);
163 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
164
165 init_rwsem(&inode->i_alloc_sem);
166 lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key);
167
168 mapping->a_ops = &empty_aops;
169 mapping->host = inode;
170 mapping->flags = 0;
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800171 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
David Chinner2cb15992008-10-30 17:32:23 +1100172 mapping->assoc_mapping = NULL;
173 mapping->backing_dev_info = &default_backing_dev_info;
174 mapping->writeback_index = 0;
175
176 /*
177 * If the block_device provides a backing_dev_info for client
178 * inodes then use that. Otherwise the inode share the bdev's
179 * backing_dev_info.
180 */
181 if (sb->s_bdev) {
182 struct backing_dev_info *bdi;
183
184 bdi = sb->s_bdev->bd_inode_backing_dev_info;
185 if (!bdi)
186 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
187 mapping->backing_dev_info = bdi;
188 }
189 inode->i_private = NULL;
190 inode->i_mapping = mapping;
191
192 return inode;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500193
194out_free_security:
195 security_inode_free(inode);
196out_free_inode:
197 if (inode->i_sb->s_op->destroy_inode)
198 inode->i_sb->s_op->destroy_inode(inode);
199 else
200 kmem_cache_free(inode_cachep, (inode));
201 return NULL;
David Chinner2cb15992008-10-30 17:32:23 +1100202}
203EXPORT_SYMBOL(inode_init_always);
204
205static struct inode *alloc_inode(struct super_block *sb)
206{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 struct inode *inode;
208
209 if (sb->s_op->alloc_inode)
210 inode = sb->s_op->alloc_inode(sb);
211 else
David Chinner2cb15992008-10-30 17:32:23 +1100212 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
David Chinner2cb15992008-10-30 17:32:23 +1100214 if (inode)
215 return inode_init_always(sb, inode);
216 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217}
218
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530219void destroy_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220{
Eric Sesterhennb7542f82006-04-02 13:38:18 +0200221 BUG_ON(inode_has_buffers(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 security_inode_free(inode);
223 if (inode->i_sb->s_op->destroy_inode)
224 inode->i_sb->s_op->destroy_inode(inode);
225 else
226 kmem_cache_free(inode_cachep, (inode));
227}
Christoph Hellwig087e3b02008-10-30 18:24:37 +1100228EXPORT_SYMBOL(destroy_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
230
231/*
232 * These are initializations that only need to be done
233 * once, because the fields are idempotent across use
234 * of the inode, so let the slab aware of that.
235 */
236void inode_init_once(struct inode *inode)
237{
238 memset(inode, 0, sizeof(*inode));
239 INIT_HLIST_NODE(&inode->i_hash);
240 INIT_LIST_HEAD(&inode->i_dentry);
241 INIT_LIST_HEAD(&inode->i_devices);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 INIT_RADIX_TREE(&inode->i_data.page_tree, GFP_ATOMIC);
Nick Piggin19fd6232008-07-25 19:45:32 -0700243 spin_lock_init(&inode->i_data.tree_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 spin_lock_init(&inode->i_data.i_mmap_lock);
245 INIT_LIST_HEAD(&inode->i_data.private_list);
246 spin_lock_init(&inode->i_data.private_lock);
247 INIT_RAW_PRIO_TREE_ROOT(&inode->i_data.i_mmap);
248 INIT_LIST_HEAD(&inode->i_data.i_mmap_nonlinear);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 i_size_ordered_init(inode);
Robert Love0eeca282005-07-12 17:06:03 -0400250#ifdef CONFIG_INOTIFY
251 INIT_LIST_HEAD(&inode->inotify_watches);
Ingo Molnard4f9af92006-03-23 03:00:30 -0800252 mutex_init(&inode->inotify_mutex);
Robert Love0eeca282005-07-12 17:06:03 -0400253#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255EXPORT_SYMBOL(inode_init_once);
256
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700257static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530259 struct inode *inode = (struct inode *) foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
Christoph Lametera35afb82007-05-16 22:10:57 -0700261 inode_init_once(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262}
263
264/*
265 * inode_lock must be held
266 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530267void __iget(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
269 if (atomic_read(&inode->i_count)) {
270 atomic_inc(&inode->i_count);
271 return;
272 }
273 atomic_inc(&inode->i_count);
Joern Engel1c0eeaf2007-10-16 23:30:44 -0700274 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 list_move(&inode->i_list, &inode_in_use);
276 inodes_stat.nr_unused--;
277}
278
279/**
280 * clear_inode - clear an inode
281 * @inode: inode to clear
282 *
283 * This is called by the filesystem to tell us
284 * that the inode is no longer useful. We just
285 * terminate it with extreme prejudice.
286 */
287void clear_inode(struct inode *inode)
288{
289 might_sleep();
290 invalidate_inode_buffers(inode);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530291
Eric Sesterhennb7542f82006-04-02 13:38:18 +0200292 BUG_ON(inode->i_data.nrpages);
293 BUG_ON(!(inode->i_state & I_FREEING));
294 BUG_ON(inode->i_state & I_CLEAR);
Joern Engel1c0eeaf2007-10-16 23:30:44 -0700295 inode_sync_wait(inode);
Jan Kara9e3509e2009-01-26 16:45:12 +0100296 vfs_dq_drop(inode);
Christoph Hellwigacb0c852007-05-08 00:25:52 -0700297 if (inode->i_sb->s_op->clear_inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 inode->i_sb->s_op->clear_inode(inode);
Theodore Ts'oeaf796e2006-09-27 01:50:48 -0700299 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 bd_forget(inode);
Theodore Ts'o577c4eb2006-09-27 01:50:49 -0700301 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 cd_forget(inode);
303 inode->i_state = I_CLEAR;
304}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305EXPORT_SYMBOL(clear_inode);
306
307/*
308 * dispose_list - dispose of the contents of a local list
309 * @head: the head of the list to free
310 *
311 * Dispose-list gets a local list with local inodes in it, so it doesn't
312 * need to worry about list corruption and SMP locks.
313 */
314static void dispose_list(struct list_head *head)
315{
316 int nr_disposed = 0;
317
318 while (!list_empty(head)) {
319 struct inode *inode;
320
Pavel Emelianovb5e61812007-05-08 00:30:19 -0700321 inode = list_first_entry(head, struct inode, i_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 list_del(&inode->i_list);
323
324 if (inode->i_data.nrpages)
325 truncate_inode_pages(&inode->i_data, 0);
326 clear_inode(inode);
Artem B. Bityuckiy4120db42005-07-12 13:58:12 -0700327
328 spin_lock(&inode_lock);
329 hlist_del_init(&inode->i_hash);
330 list_del_init(&inode->i_sb_list);
331 spin_unlock(&inode_lock);
332
333 wake_up_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 destroy_inode(inode);
335 nr_disposed++;
336 }
337 spin_lock(&inode_lock);
338 inodes_stat.nr_inodes -= nr_disposed;
339 spin_unlock(&inode_lock);
340}
341
342/*
343 * Invalidate all inodes for a device.
344 */
345static int invalidate_list(struct list_head *head, struct list_head *dispose)
346{
347 struct list_head *next;
348 int busy = 0, count = 0;
349
350 next = head->next;
351 for (;;) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530352 struct list_head *tmp = next;
353 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354
355 /*
356 * We can reschedule here without worrying about the list's
357 * consistency because the per-sb list of inodes must not
Ingo Molnarf24075b2006-03-23 03:00:34 -0800358 * change during umount anymore, and because iprune_mutex keeps
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 * shrink_icache_memory() away.
360 */
361 cond_resched_lock(&inode_lock);
362
363 next = next->next;
364 if (tmp == head)
365 break;
366 inode = list_entry(tmp, struct inode, i_sb_list);
Nick Pigginaabb8fd2009-03-11 13:17:36 -0700367 if (inode->i_state & I_NEW)
368 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 invalidate_inode_buffers(inode);
370 if (!atomic_read(&inode->i_count)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 list_move(&inode->i_list, dispose);
Nick Piggin7ef0d732009-03-12 14:31:38 -0700372 WARN_ON(inode->i_state & I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 inode->i_state |= I_FREEING;
374 count++;
375 continue;
376 }
377 busy = 1;
378 }
379 /* only unused inodes may be cached with i_count zero */
380 inodes_stat.nr_unused -= count;
381 return busy;
382}
383
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384/**
385 * invalidate_inodes - discard the inodes on a device
386 * @sb: superblock
387 *
388 * Discard all of the inodes for a given superblock. If the discard
389 * fails because there are busy inodes then a non zero value is returned.
390 * If the discard is successful all the inodes have been discarded.
391 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530392int invalidate_inodes(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393{
394 int busy;
395 LIST_HEAD(throw_away);
396
Ingo Molnarf24075b2006-03-23 03:00:34 -0800397 mutex_lock(&iprune_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 spin_lock(&inode_lock);
Robert Love0eeca282005-07-12 17:06:03 -0400399 inotify_unmount_inodes(&sb->s_inodes);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 busy = invalidate_list(&sb->s_inodes, &throw_away);
401 spin_unlock(&inode_lock);
402
403 dispose_list(&throw_away);
Ingo Molnarf24075b2006-03-23 03:00:34 -0800404 mutex_unlock(&iprune_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
406 return busy;
407}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408EXPORT_SYMBOL(invalidate_inodes);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
410static int can_unuse(struct inode *inode)
411{
412 if (inode->i_state)
413 return 0;
414 if (inode_has_buffers(inode))
415 return 0;
416 if (atomic_read(&inode->i_count))
417 return 0;
418 if (inode->i_data.nrpages)
419 return 0;
420 return 1;
421}
422
423/*
424 * Scan `goal' inodes on the unused list for freeable ones. They are moved to
425 * a temporary list and then are freed outside inode_lock by dispose_list().
426 *
427 * Any inodes which are pinned purely because of attached pagecache have their
428 * pagecache removed. We expect the final iput() on that inode to add it to
429 * the front of the inode_unused list. So look for it there and if the
430 * inode is still freeable, proceed. The right inode is found 99.9% of the
431 * time in testing on a 4-way.
432 *
433 * If the inode has metadata buffers attached to mapping->private_list then
434 * try to remove them.
435 */
436static void prune_icache(int nr_to_scan)
437{
438 LIST_HEAD(freeable);
439 int nr_pruned = 0;
440 int nr_scanned;
441 unsigned long reap = 0;
442
Ingo Molnarf24075b2006-03-23 03:00:34 -0800443 mutex_lock(&iprune_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 spin_lock(&inode_lock);
445 for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) {
446 struct inode *inode;
447
448 if (list_empty(&inode_unused))
449 break;
450
451 inode = list_entry(inode_unused.prev, struct inode, i_list);
452
453 if (inode->i_state || atomic_read(&inode->i_count)) {
454 list_move(&inode->i_list, &inode_unused);
455 continue;
456 }
457 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
458 __iget(inode);
459 spin_unlock(&inode_lock);
460 if (remove_inode_buffers(inode))
Andrew Mortonfc0ecff2007-02-10 01:45:39 -0800461 reap += invalidate_mapping_pages(&inode->i_data,
462 0, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 iput(inode);
464 spin_lock(&inode_lock);
465
466 if (inode != list_entry(inode_unused.next,
467 struct inode, i_list))
468 continue; /* wrong inode or list_empty */
469 if (!can_unuse(inode))
470 continue;
471 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 list_move(&inode->i_list, &freeable);
Nick Piggin7ef0d732009-03-12 14:31:38 -0700473 WARN_ON(inode->i_state & I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 inode->i_state |= I_FREEING;
475 nr_pruned++;
476 }
477 inodes_stat.nr_unused -= nr_pruned;
Christoph Lameterf8891e52006-06-30 01:55:45 -0700478 if (current_is_kswapd())
479 __count_vm_events(KSWAPD_INODESTEAL, reap);
480 else
481 __count_vm_events(PGINODESTEAL, reap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 spin_unlock(&inode_lock);
483
484 dispose_list(&freeable);
Ingo Molnarf24075b2006-03-23 03:00:34 -0800485 mutex_unlock(&iprune_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486}
487
488/*
489 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
490 * "unused" means that no dentries are referring to the inodes: the files are
491 * not open and the dcache references to those inodes have already been
492 * reclaimed.
493 *
494 * This function is passed the number of inodes to scan, and it returns the
495 * total number of remaining possibly-reclaimable inodes.
496 */
Al Viro27496a82005-10-21 03:20:48 -0400497static int shrink_icache_memory(int nr, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498{
499 if (nr) {
500 /*
501 * Nasty deadlock avoidance. We may hold various FS locks,
502 * and we don't want to recurse into the FS that called us
503 * in clear_inode() and friends..
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530504 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 if (!(gfp_mask & __GFP_FS))
506 return -1;
507 prune_icache(nr);
508 }
509 return (inodes_stat.nr_unused / 100) * sysctl_vfs_cache_pressure;
510}
511
Rusty Russell8e1f9362007-07-17 04:03:17 -0700512static struct shrinker icache_shrinker = {
513 .shrink = shrink_icache_memory,
514 .seeks = DEFAULT_SEEKS,
515};
516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517static void __wait_on_freeing_inode(struct inode *inode);
518/*
519 * Called with the inode lock held.
520 * NOTE: we are not increasing the inode-refcount, you must call __iget()
521 * by hand after calling find_inode now! This simplifies iunique and won't
522 * add any additional branch in the common code.
523 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530524static struct inode *find_inode(struct super_block *sb,
525 struct hlist_head *head,
526 int (*test)(struct inode *, void *),
527 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528{
529 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530530 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
532repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700533 hlist_for_each_entry(inode, node, head, i_hash) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 if (inode->i_sb != sb)
535 continue;
536 if (!test(inode, data))
537 continue;
Alexander Viro991114c2005-06-23 00:09:01 -0700538 if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 __wait_on_freeing_inode(inode);
540 goto repeat;
541 }
542 break;
543 }
544 return node ? inode : NULL;
545}
546
547/*
548 * find_inode_fast is the fast path version of find_inode, see the comment at
549 * iget_locked for details.
550 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530551static struct inode *find_inode_fast(struct super_block *sb,
552 struct hlist_head *head, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
554 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530555 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
557repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700558 hlist_for_each_entry(inode, node, head, i_hash) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 if (inode->i_ino != ino)
560 continue;
561 if (inode->i_sb != sb)
562 continue;
Alexander Viro991114c2005-06-23 00:09:01 -0700563 if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 __wait_on_freeing_inode(inode);
565 goto repeat;
566 }
567 break;
568 }
569 return node ? inode : NULL;
570}
571
David Chinner8290c352008-10-30 17:35:24 +1100572static unsigned long hash(struct super_block *sb, unsigned long hashval)
573{
574 unsigned long tmp;
575
576 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
577 L1_CACHE_BYTES;
578 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
579 return tmp & I_HASHMASK;
580}
581
582static inline void
583__inode_add_to_lists(struct super_block *sb, struct hlist_head *head,
584 struct inode *inode)
585{
586 inodes_stat.nr_inodes++;
587 list_add(&inode->i_list, &inode_in_use);
588 list_add(&inode->i_sb_list, &sb->s_inodes);
589 if (head)
590 hlist_add_head(&inode->i_hash, head);
591}
592
593/**
594 * inode_add_to_lists - add a new inode to relevant lists
Randy Dunlap0bc02f32009-01-06 14:41:13 -0800595 * @sb: superblock inode belongs to
596 * @inode: inode to mark in use
David Chinner8290c352008-10-30 17:35:24 +1100597 *
598 * When an inode is allocated it needs to be accounted for, added to the in use
599 * list, the owning superblock and the inode hash. This needs to be done under
600 * the inode_lock, so export a function to do this rather than the inode lock
601 * itself. We calculate the hash list to add to here so it is all internal
602 * which requires the caller to have already set up the inode number in the
603 * inode to add.
604 */
605void inode_add_to_lists(struct super_block *sb, struct inode *inode)
606{
607 struct hlist_head *head = inode_hashtable + hash(sb, inode->i_ino);
608
609 spin_lock(&inode_lock);
610 __inode_add_to_lists(sb, head, inode);
611 spin_unlock(&inode_lock);
612}
613EXPORT_SYMBOL_GPL(inode_add_to_lists);
614
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615/**
616 * new_inode - obtain an inode
617 * @sb: superblock
618 *
Mel Gorman769848c2007-07-17 04:03:05 -0700619 * Allocates a new inode for given superblock. The default gfp_mask
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800620 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
Mel Gorman769848c2007-07-17 04:03:05 -0700621 * If HIGHMEM pages are unsuitable or it is known that pages allocated
622 * for the page cache are not reclaimable or migratable,
623 * mapping_set_gfp_mask() must be called with suitable flags on the
624 * newly created inode's mapping
625 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 */
627struct inode *new_inode(struct super_block *sb)
628{
Jeff Layton866b04f2007-05-08 00:32:29 -0700629 /*
630 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
631 * error if st_ino won't fit in target struct field. Use 32bit counter
632 * here to attempt to avoid that.
633 */
634 static unsigned int last_ino;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530635 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637 spin_lock_prefetch(&inode_lock);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530638
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 inode = alloc_inode(sb);
640 if (inode) {
641 spin_lock(&inode_lock);
David Chinner8290c352008-10-30 17:35:24 +1100642 __inode_add_to_lists(sb, NULL, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 inode->i_ino = ++last_ino;
644 inode->i_state = 0;
645 spin_unlock(&inode_lock);
646 }
647 return inode;
648}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649EXPORT_SYMBOL(new_inode);
650
651void unlock_new_inode(struct inode *inode)
652{
Peter Zijlstra14358e62007-10-14 01:38:33 +0200653#ifdef CONFIG_DEBUG_LOCK_ALLOC
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200654 if (inode->i_mode & S_IFDIR) {
655 struct file_system_type *type = inode->i_sb->s_type;
656
657 /*
658 * ensure nobody is actually holding i_mutex
659 */
660 mutex_destroy(&inode->i_mutex);
661 mutex_init(&inode->i_mutex);
Peter Zijlstra14358e62007-10-14 01:38:33 +0200662 lockdep_set_class(&inode->i_mutex, &type->i_mutex_dir_key);
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200663 }
Peter Zijlstra14358e62007-10-14 01:38:33 +0200664#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 /*
666 * This is special! We do not need the spinlock
667 * when clearing I_LOCK, because we're guaranteed
668 * that nobody else tries to do anything about the
669 * state of the inode when it is locked, as we
670 * just created it (so there can be no old holders
671 * that haven't tested I_LOCK).
672 */
Nick Piggin7ef0d732009-03-12 14:31:38 -0700673 WARN_ON((inode->i_state & (I_LOCK|I_NEW)) != (I_LOCK|I_NEW));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 inode->i_state &= ~(I_LOCK|I_NEW);
675 wake_up_inode(inode);
676}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677EXPORT_SYMBOL(unlock_new_inode);
678
679/*
680 * This is called without the inode lock held.. Be careful.
681 *
682 * We no longer cache the sb_flags in i_flags - see fs.h
683 * -- rmk@arm.uk.linux.org
684 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530685static struct inode *get_new_inode(struct super_block *sb,
686 struct hlist_head *head,
687 int (*test)(struct inode *, void *),
688 int (*set)(struct inode *, void *),
689 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530691 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
693 inode = alloc_inode(sb);
694 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530695 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
697 spin_lock(&inode_lock);
698 /* We released the lock, so.. */
699 old = find_inode(sb, head, test, data);
700 if (!old) {
701 if (set(inode, data))
702 goto set_failed;
703
David Chinner8290c352008-10-30 17:35:24 +1100704 __inode_add_to_lists(sb, head, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 inode->i_state = I_LOCK|I_NEW;
706 spin_unlock(&inode_lock);
707
708 /* Return the locked inode with I_NEW set, the
709 * caller is responsible for filling in the contents
710 */
711 return inode;
712 }
713
714 /*
715 * Uhhuh, somebody else created the same inode under
716 * us. Use the old inode instead of the one we just
717 * allocated.
718 */
719 __iget(old);
720 spin_unlock(&inode_lock);
721 destroy_inode(inode);
722 inode = old;
723 wait_on_inode(inode);
724 }
725 return inode;
726
727set_failed:
728 spin_unlock(&inode_lock);
729 destroy_inode(inode);
730 return NULL;
731}
732
733/*
734 * get_new_inode_fast is the fast path version of get_new_inode, see the
735 * comment at iget_locked for details.
736 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530737static struct inode *get_new_inode_fast(struct super_block *sb,
738 struct hlist_head *head, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530740 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
742 inode = alloc_inode(sb);
743 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530744 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
746 spin_lock(&inode_lock);
747 /* We released the lock, so.. */
748 old = find_inode_fast(sb, head, ino);
749 if (!old) {
750 inode->i_ino = ino;
David Chinner8290c352008-10-30 17:35:24 +1100751 __inode_add_to_lists(sb, head, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 inode->i_state = I_LOCK|I_NEW;
753 spin_unlock(&inode_lock);
754
755 /* Return the locked inode with I_NEW set, the
756 * caller is responsible for filling in the contents
757 */
758 return inode;
759 }
760
761 /*
762 * Uhhuh, somebody else created the same inode under
763 * us. Use the old inode instead of the one we just
764 * allocated.
765 */
766 __iget(old);
767 spin_unlock(&inode_lock);
768 destroy_inode(inode);
769 inode = old;
770 wait_on_inode(inode);
771 }
772 return inode;
773}
774
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775/**
776 * iunique - get a unique inode number
777 * @sb: superblock
778 * @max_reserved: highest reserved inode number
779 *
780 * Obtain an inode number that is unique on the system for a given
781 * superblock. This is used by file systems that have no natural
782 * permanent inode numbering system. An inode number is returned that
783 * is higher than the reserved limit but unique.
784 *
785 * BUGS:
786 * With a large number of inodes live on the file system this function
787 * currently becomes quite slow.
788 */
789ino_t iunique(struct super_block *sb, ino_t max_reserved)
790{
Jeff Layton866b04f2007-05-08 00:32:29 -0700791 /*
792 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
793 * error if st_ino won't fit in target struct field. Use 32bit counter
794 * here to attempt to avoid that.
795 */
796 static unsigned int counter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 struct inode *inode;
Jeffrey Layton3361c7b2007-05-08 00:29:48 -0700798 struct hlist_head *head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 ino_t res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
Jeffrey Layton3361c7b2007-05-08 00:29:48 -0700801 spin_lock(&inode_lock);
802 do {
803 if (counter <= max_reserved)
804 counter = max_reserved + 1;
805 res = counter++;
806 head = inode_hashtable + hash(sb, res);
807 inode = find_inode_fast(sb, head, res);
808 } while (inode != NULL);
809 spin_unlock(&inode_lock);
810
811 return res;
812}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813EXPORT_SYMBOL(iunique);
814
815struct inode *igrab(struct inode *inode)
816{
817 spin_lock(&inode_lock);
Jan Blunck4a3b0a42007-02-10 01:44:59 -0800818 if (!(inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 __iget(inode);
820 else
821 /*
822 * Handle the case where s_op->clear_inode is not been
823 * called yet, and somebody is calling igrab
824 * while the inode is getting freed.
825 */
826 inode = NULL;
827 spin_unlock(&inode_lock);
828 return inode;
829}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830EXPORT_SYMBOL(igrab);
831
832/**
833 * ifind - internal function, you want ilookup5() or iget5().
834 * @sb: super block of file system to search
835 * @head: the head of the list to search
836 * @test: callback used for comparisons between inodes
837 * @data: opaque data pointer to pass to @test
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700838 * @wait: if true wait for the inode to be unlocked, if false do not
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 *
840 * ifind() searches for the inode specified by @data in the inode
841 * cache. This is a generalized version of ifind_fast() for file systems where
842 * the inode number is not sufficient for unique identification of an inode.
843 *
844 * If the inode is in the cache, the inode is returned with an incremented
845 * reference count.
846 *
847 * Otherwise NULL is returned.
848 *
849 * Note, @test is called with the inode_lock held, so can't sleep.
850 */
Matt Mackall5d2bea42006-01-08 01:05:21 -0800851static struct inode *ifind(struct super_block *sb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 struct hlist_head *head, int (*test)(struct inode *, void *),
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700853 void *data, const int wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854{
855 struct inode *inode;
856
857 spin_lock(&inode_lock);
858 inode = find_inode(sb, head, test, data);
859 if (inode) {
860 __iget(inode);
861 spin_unlock(&inode_lock);
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700862 if (likely(wait))
863 wait_on_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 return inode;
865 }
866 spin_unlock(&inode_lock);
867 return NULL;
868}
869
870/**
871 * ifind_fast - internal function, you want ilookup() or iget().
872 * @sb: super block of file system to search
873 * @head: head of the list to search
874 * @ino: inode number to search for
875 *
876 * ifind_fast() searches for the inode @ino in the inode cache. This is for
877 * file systems where the inode number is sufficient for unique identification
878 * of an inode.
879 *
880 * If the inode is in the cache, the inode is returned with an incremented
881 * reference count.
882 *
883 * Otherwise NULL is returned.
884 */
Matt Mackall5d2bea42006-01-08 01:05:21 -0800885static struct inode *ifind_fast(struct super_block *sb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 struct hlist_head *head, unsigned long ino)
887{
888 struct inode *inode;
889
890 spin_lock(&inode_lock);
891 inode = find_inode_fast(sb, head, ino);
892 if (inode) {
893 __iget(inode);
894 spin_unlock(&inode_lock);
895 wait_on_inode(inode);
896 return inode;
897 }
898 spin_unlock(&inode_lock);
899 return NULL;
900}
901
902/**
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700903 * ilookup5_nowait - search for an inode in the inode cache
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 * @sb: super block of file system to search
905 * @hashval: hash value (usually inode number) to search for
906 * @test: callback used for comparisons between inodes
907 * @data: opaque data pointer to pass to @test
908 *
909 * ilookup5() uses ifind() to search for the inode specified by @hashval and
910 * @data in the inode cache. This is a generalized version of ilookup() for
911 * file systems where the inode number is not sufficient for unique
912 * identification of an inode.
913 *
914 * If the inode is in the cache, the inode is returned with an incremented
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700915 * reference count. Note, the inode lock is not waited upon so you have to be
916 * very careful what you do with the returned inode. You probably should be
917 * using ilookup5() instead.
918 *
919 * Otherwise NULL is returned.
920 *
921 * Note, @test is called with the inode_lock held, so can't sleep.
922 */
923struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
924 int (*test)(struct inode *, void *), void *data)
925{
926 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
927
928 return ifind(sb, head, test, data, 0);
929}
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700930EXPORT_SYMBOL(ilookup5_nowait);
931
932/**
933 * ilookup5 - search for an inode in the inode cache
934 * @sb: super block of file system to search
935 * @hashval: hash value (usually inode number) to search for
936 * @test: callback used for comparisons between inodes
937 * @data: opaque data pointer to pass to @test
938 *
939 * ilookup5() uses ifind() to search for the inode specified by @hashval and
940 * @data in the inode cache. This is a generalized version of ilookup() for
941 * file systems where the inode number is not sufficient for unique
942 * identification of an inode.
943 *
944 * If the inode is in the cache, the inode lock is waited upon and the inode is
945 * returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 *
947 * Otherwise NULL is returned.
948 *
949 * Note, @test is called with the inode_lock held, so can't sleep.
950 */
951struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
952 int (*test)(struct inode *, void *), void *data)
953{
954 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
955
Anton Altaparmakov88bd5122005-07-13 01:10:44 -0700956 return ifind(sb, head, test, data, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958EXPORT_SYMBOL(ilookup5);
959
960/**
961 * ilookup - search for an inode in the inode cache
962 * @sb: super block of file system to search
963 * @ino: inode number to search for
964 *
965 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
966 * This is for file systems where the inode number is sufficient for unique
967 * identification of an inode.
968 *
969 * If the inode is in the cache, the inode is returned with an incremented
970 * reference count.
971 *
972 * Otherwise NULL is returned.
973 */
974struct inode *ilookup(struct super_block *sb, unsigned long ino)
975{
976 struct hlist_head *head = inode_hashtable + hash(sb, ino);
977
978 return ifind_fast(sb, head, ino);
979}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980EXPORT_SYMBOL(ilookup);
981
982/**
983 * iget5_locked - obtain an inode from a mounted file system
984 * @sb: super block of file system
985 * @hashval: hash value (usually inode number) to get
986 * @test: callback used for comparisons between inodes
987 * @set: callback used to initialize a new struct inode
988 * @data: opaque data pointer to pass to @test and @set
989 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 * iget5_locked() uses ifind() to search for the inode specified by @hashval
991 * and @data in the inode cache and if present it is returned with an increased
992 * reference count. This is a generalized version of iget_locked() for file
993 * systems where the inode number is not sufficient for unique identification
994 * of an inode.
995 *
996 * If the inode is not in cache, get_new_inode() is called to allocate a new
997 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
998 * file system gets to fill it in before unlocking it via unlock_new_inode().
999 *
1000 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1001 */
1002struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
1003 int (*test)(struct inode *, void *),
1004 int (*set)(struct inode *, void *), void *data)
1005{
1006 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1007 struct inode *inode;
1008
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001009 inode = ifind(sb, head, test, data, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 if (inode)
1011 return inode;
1012 /*
1013 * get_new_inode() will do the right thing, re-trying the search
1014 * in case it had to block at any point.
1015 */
1016 return get_new_inode(sb, head, test, set, data);
1017}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018EXPORT_SYMBOL(iget5_locked);
1019
1020/**
1021 * iget_locked - obtain an inode from a mounted file system
1022 * @sb: super block of file system
1023 * @ino: inode number to get
1024 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1026 * the inode cache and if present it is returned with an increased reference
1027 * count. This is for file systems where the inode number is sufficient for
1028 * unique identification of an inode.
1029 *
1030 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1031 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1032 * The file system gets to fill it in before unlocking it via
1033 * unlock_new_inode().
1034 */
1035struct inode *iget_locked(struct super_block *sb, unsigned long ino)
1036{
1037 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1038 struct inode *inode;
1039
1040 inode = ifind_fast(sb, head, ino);
1041 if (inode)
1042 return inode;
1043 /*
1044 * get_new_inode_fast() will do the right thing, re-trying the search
1045 * in case it had to block at any point.
1046 */
1047 return get_new_inode_fast(sb, head, ino);
1048}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049EXPORT_SYMBOL(iget_locked);
1050
Al Viro261bca82008-12-30 01:48:21 -05001051int insert_inode_locked(struct inode *inode)
1052{
1053 struct super_block *sb = inode->i_sb;
1054 ino_t ino = inode->i_ino;
1055 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1056 struct inode *old;
1057
1058 inode->i_state |= I_LOCK|I_NEW;
1059 while (1) {
1060 spin_lock(&inode_lock);
1061 old = find_inode_fast(sb, head, ino);
1062 if (likely(!old)) {
1063 hlist_add_head(&inode->i_hash, head);
1064 spin_unlock(&inode_lock);
1065 return 0;
1066 }
1067 __iget(old);
1068 spin_unlock(&inode_lock);
1069 wait_on_inode(old);
1070 if (unlikely(!hlist_unhashed(&old->i_hash))) {
1071 iput(old);
1072 return -EBUSY;
1073 }
1074 iput(old);
1075 }
1076}
Al Viro261bca82008-12-30 01:48:21 -05001077EXPORT_SYMBOL(insert_inode_locked);
1078
1079int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1080 int (*test)(struct inode *, void *), void *data)
1081{
1082 struct super_block *sb = inode->i_sb;
1083 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1084 struct inode *old;
1085
1086 inode->i_state |= I_LOCK|I_NEW;
1087
1088 while (1) {
1089 spin_lock(&inode_lock);
1090 old = find_inode(sb, head, test, data);
1091 if (likely(!old)) {
1092 hlist_add_head(&inode->i_hash, head);
1093 spin_unlock(&inode_lock);
1094 return 0;
1095 }
1096 __iget(old);
1097 spin_unlock(&inode_lock);
1098 wait_on_inode(old);
1099 if (unlikely(!hlist_unhashed(&old->i_hash))) {
1100 iput(old);
1101 return -EBUSY;
1102 }
1103 iput(old);
1104 }
1105}
Al Viro261bca82008-12-30 01:48:21 -05001106EXPORT_SYMBOL(insert_inode_locked4);
1107
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108/**
1109 * __insert_inode_hash - hash an inode
1110 * @inode: unhashed inode
1111 * @hashval: unsigned long value used to locate this object in the
1112 * inode_hashtable.
1113 *
1114 * Add an inode to the inode hash for this superblock.
1115 */
1116void __insert_inode_hash(struct inode *inode, unsigned long hashval)
1117{
1118 struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval);
1119 spin_lock(&inode_lock);
1120 hlist_add_head(&inode->i_hash, head);
1121 spin_unlock(&inode_lock);
1122}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123EXPORT_SYMBOL(__insert_inode_hash);
1124
1125/**
1126 * remove_inode_hash - remove an inode from the hash
1127 * @inode: inode to unhash
1128 *
1129 * Remove an inode from the superblock.
1130 */
1131void remove_inode_hash(struct inode *inode)
1132{
1133 spin_lock(&inode_lock);
1134 hlist_del_init(&inode->i_hash);
1135 spin_unlock(&inode_lock);
1136}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137EXPORT_SYMBOL(remove_inode_hash);
1138
1139/*
1140 * Tell the filesystem that this inode is no longer of any interest and should
1141 * be completely destroyed.
1142 *
1143 * We leave the inode in the inode hash table until *after* the filesystem's
1144 * ->delete_inode completes. This ensures that an iget (such as nfsd might
1145 * instigate) will always find up-to-date information either in the hash or on
1146 * disk.
1147 *
1148 * I_FREEING is set so that no-one will take a new reference to the inode while
1149 * it is being deleted.
1150 */
Arjan van de Venb32714b2009-01-09 07:04:15 -08001151void generic_delete_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001153 const struct super_operations *op = inode->i_sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
Arjan van de Venb32714b2009-01-09 07:04:15 -08001155 list_del_init(&inode->i_list);
1156 list_del_init(&inode->i_sb_list);
Nick Piggin7ef0d732009-03-12 14:31:38 -07001157 WARN_ON(inode->i_state & I_NEW);
Arjan van de Venb32714b2009-01-09 07:04:15 -08001158 inode->i_state |= I_FREEING;
1159 inodes_stat.nr_inodes--;
1160 spin_unlock(&inode_lock);
1161
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 security_inode_delete(inode);
1163
1164 if (op->delete_inode) {
1165 void (*delete)(struct inode *) = op->delete_inode;
1166 if (!is_bad_inode(inode))
Jan Kara9e3509e2009-01-26 16:45:12 +01001167 vfs_dq_init(inode);
Mark Fashehe85b5652005-09-09 13:01:29 -07001168 /* Filesystems implementing their own
1169 * s_op->delete_inode are required to call
1170 * truncate_inode_pages and clear_inode()
1171 * internally */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 delete(inode);
Mark Fashehe85b5652005-09-09 13:01:29 -07001173 } else {
1174 truncate_inode_pages(&inode->i_data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 clear_inode(inode);
Mark Fashehe85b5652005-09-09 13:01:29 -07001176 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 spin_lock(&inode_lock);
1178 hlist_del_init(&inode->i_hash);
1179 spin_unlock(&inode_lock);
1180 wake_up_inode(inode);
Eric Sesterhennb7542f82006-04-02 13:38:18 +02001181 BUG_ON(inode->i_state != I_CLEAR);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 destroy_inode(inode);
1183}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184EXPORT_SYMBOL(generic_delete_inode);
1185
1186static void generic_forget_inode(struct inode *inode)
1187{
1188 struct super_block *sb = inode->i_sb;
1189
1190 if (!hlist_unhashed(&inode->i_hash)) {
Joern Engel1c0eeaf2007-10-16 23:30:44 -07001191 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 list_move(&inode->i_list, &inode_unused);
1193 inodes_stat.nr_unused++;
Christoph Hellwigacb0c852007-05-08 00:25:52 -07001194 if (sb->s_flags & MS_ACTIVE) {
Alexander Viro991114c2005-06-23 00:09:01 -07001195 spin_unlock(&inode_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 return;
Alexander Viro991114c2005-06-23 00:09:01 -07001197 }
Nick Piggin7ef0d732009-03-12 14:31:38 -07001198 WARN_ON(inode->i_state & I_NEW);
Alexander Viro991114c2005-06-23 00:09:01 -07001199 inode->i_state |= I_WILL_FREE;
1200 spin_unlock(&inode_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 write_inode_now(inode, 1);
1202 spin_lock(&inode_lock);
Nick Piggin7ef0d732009-03-12 14:31:38 -07001203 WARN_ON(inode->i_state & I_NEW);
Alexander Viro991114c2005-06-23 00:09:01 -07001204 inode->i_state &= ~I_WILL_FREE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 inodes_stat.nr_unused--;
1206 hlist_del_init(&inode->i_hash);
1207 }
1208 list_del_init(&inode->i_list);
1209 list_del_init(&inode->i_sb_list);
Nick Piggin7ef0d732009-03-12 14:31:38 -07001210 WARN_ON(inode->i_state & I_NEW);
Alexander Viro991114c2005-06-23 00:09:01 -07001211 inode->i_state |= I_FREEING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 inodes_stat.nr_inodes--;
1213 spin_unlock(&inode_lock);
1214 if (inode->i_data.nrpages)
1215 truncate_inode_pages(&inode->i_data, 0);
1216 clear_inode(inode);
Andrea Arcangeli7f04c262005-10-30 15:03:05 -08001217 wake_up_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 destroy_inode(inode);
1219}
1220
1221/*
1222 * Normal UNIX filesystem behaviour: delete the
1223 * inode when the usage count drops to zero, and
1224 * i_nlink is zero.
1225 */
Mark Fashehcb2c0232005-07-07 17:56:03 -07001226void generic_drop_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227{
1228 if (!inode->i_nlink)
1229 generic_delete_inode(inode);
1230 else
1231 generic_forget_inode(inode);
1232}
Mark Fashehcb2c0232005-07-07 17:56:03 -07001233EXPORT_SYMBOL_GPL(generic_drop_inode);
1234
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235/*
1236 * Called when we're dropping the last reference
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301237 * to an inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 *
1239 * Call the FS "drop()" function, defaulting to
1240 * the legacy UNIX filesystem behaviour..
1241 *
1242 * NOTE! NOTE! NOTE! We're called with the inode lock
1243 * held, and the drop function is supposed to release
1244 * the lock!
1245 */
1246static inline void iput_final(struct inode *inode)
1247{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001248 const struct super_operations *op = inode->i_sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 void (*drop)(struct inode *) = generic_drop_inode;
1250
1251 if (op && op->drop_inode)
1252 drop = op->drop_inode;
1253 drop(inode);
1254}
1255
1256/**
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301257 * iput - put an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 * @inode: inode to put
1259 *
1260 * Puts an inode, dropping its usage count. If the inode use count hits
1261 * zero, the inode is then freed and may also be destroyed.
1262 *
1263 * Consequently, iput() can sleep.
1264 */
1265void iput(struct inode *inode)
1266{
1267 if (inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 BUG_ON(inode->i_state == I_CLEAR);
1269
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 if (atomic_dec_and_lock(&inode->i_count, &inode_lock))
1271 iput_final(inode);
1272 }
1273}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274EXPORT_SYMBOL(iput);
1275
1276/**
1277 * bmap - find a block number in a file
1278 * @inode: inode of file
1279 * @block: block to find
1280 *
1281 * Returns the block number on the device holding the inode that
1282 * is the disk block number for the block of the file requested.
1283 * That is, asked for block 4 of inode 1 the function will return the
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301284 * disk block relative to the disk start that holds that block of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 * file.
1286 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301287sector_t bmap(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288{
1289 sector_t res = 0;
1290 if (inode->i_mapping->a_ops->bmap)
1291 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1292 return res;
1293}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294EXPORT_SYMBOL(bmap);
1295
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001296/*
1297 * With relative atime, only update atime if the previous atime is
1298 * earlier than either the ctime or mtime or if at least a day has
1299 * passed since the last atime update.
1300 */
1301static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1302 struct timespec now)
1303{
1304
1305 if (!(mnt->mnt_flags & MNT_RELATIME))
1306 return 1;
1307 /*
1308 * Is mtime younger than atime? If yes, update atime:
1309 */
1310 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1311 return 1;
1312 /*
1313 * Is ctime younger than atime? If yes, update atime:
1314 */
1315 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1316 return 1;
1317
1318 /*
1319 * Is the previous atime value older than a day? If yes,
1320 * update atime:
1321 */
1322 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1323 return 1;
1324 /*
1325 * Good, we can skip the atime update:
1326 */
1327 return 0;
1328}
1329
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330/**
Christoph Hellwig869243a2006-01-09 20:52:03 -08001331 * touch_atime - update the access time
1332 * @mnt: mount the inode is accessed on
Martin Waitz7045f372006-02-01 03:06:57 -08001333 * @dentry: dentry accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 *
1335 * Update the accessed time on an inode and mark it for writeback.
1336 * This function automatically handles read only file systems and media,
1337 * as well as the "noatime" flag and inode specific "noatime" markers.
1338 */
Christoph Hellwig869243a2006-01-09 20:52:03 -08001339void touch_atime(struct vfsmount *mnt, struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340{
Christoph Hellwig869243a2006-01-09 20:52:03 -08001341 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 struct timespec now;
1343
Dave Hansencdb70f32008-02-15 14:37:41 -08001344 if (mnt_want_write(mnt))
1345 return;
Andrew Mortonb2276132006-12-13 00:34:33 -08001346 if (inode->i_flags & S_NOATIME)
Dave Hansencdb70f32008-02-15 14:37:41 -08001347 goto out;
Eric Dumazet37756ce2007-02-10 01:44:49 -08001348 if (IS_NOATIME(inode))
Dave Hansencdb70f32008-02-15 14:37:41 -08001349 goto out;
Andrew Mortonb2276132006-12-13 00:34:33 -08001350 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
Dave Hansencdb70f32008-02-15 14:37:41 -08001351 goto out;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001352
Dave Hansencdb70f32008-02-15 14:37:41 -08001353 if (mnt->mnt_flags & MNT_NOATIME)
1354 goto out;
1355 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
1356 goto out;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 now = current_fs_time(inode->i_sb);
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001359
1360 if (!relatime_need_update(mnt, inode, now))
1361 goto out;
1362
Valerie Henson47ae32d2006-12-13 00:34:34 -08001363 if (timespec_equal(&inode->i_atime, &now))
Dave Hansencdb70f32008-02-15 14:37:41 -08001364 goto out;
Valerie Henson47ae32d2006-12-13 00:34:34 -08001365
1366 inode->i_atime = now;
1367 mark_inode_dirty_sync(inode);
Dave Hansencdb70f32008-02-15 14:37:41 -08001368out:
1369 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370}
Christoph Hellwig869243a2006-01-09 20:52:03 -08001371EXPORT_SYMBOL(touch_atime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
1373/**
Christoph Hellwig870f4812006-01-09 20:52:01 -08001374 * file_update_time - update mtime and ctime time
1375 * @file: file accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 *
Christoph Hellwig870f4812006-01-09 20:52:01 -08001377 * Update the mtime and ctime members of an inode and mark the inode
1378 * for writeback. Note that this function is meant exclusively for
1379 * usage in the file write path of filesystems, and filesystems may
1380 * choose to explicitly ignore update via this function with the
1381 * S_NOCTIME inode flag, e.g. for network filesystem where these
1382 * timestamps are handled by the server.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 */
1384
Christoph Hellwig870f4812006-01-09 20:52:01 -08001385void file_update_time(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001387 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 struct timespec now;
1389 int sync_it = 0;
Dave Hansen20ddee22008-02-15 14:37:43 -08001390 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391
1392 if (IS_NOCMTIME(inode))
1393 return;
Dave Hansen20ddee22008-02-15 14:37:43 -08001394
1395 err = mnt_want_write(file->f_path.mnt);
1396 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 return;
1398
1399 now = current_fs_time(inode->i_sb);
Andreas Mohred97bd32006-10-02 02:17:17 -07001400 if (!timespec_equal(&inode->i_mtime, &now)) {
1401 inode->i_mtime = now;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 sync_it = 1;
Andreas Mohred97bd32006-10-02 02:17:17 -07001403 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404
Andreas Mohred97bd32006-10-02 02:17:17 -07001405 if (!timespec_equal(&inode->i_ctime, &now)) {
1406 inode->i_ctime = now;
Christoph Hellwig870f4812006-01-09 20:52:01 -08001407 sync_it = 1;
Andreas Mohred97bd32006-10-02 02:17:17 -07001408 }
Christoph Hellwig870f4812006-01-09 20:52:01 -08001409
Jean Noel Cordenner7a224222008-01-28 23:58:27 -05001410 if (IS_I_VERSION(inode)) {
1411 inode_inc_iversion(inode);
1412 sync_it = 1;
1413 }
1414
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 if (sync_it)
1416 mark_inode_dirty_sync(inode);
Dave Hansen20ddee22008-02-15 14:37:43 -08001417 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418}
Christoph Hellwig870f4812006-01-09 20:52:01 -08001419EXPORT_SYMBOL(file_update_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
1421int inode_needs_sync(struct inode *inode)
1422{
1423 if (IS_SYNC(inode))
1424 return 1;
1425 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1426 return 1;
1427 return 0;
1428}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429EXPORT_SYMBOL(inode_needs_sync);
1430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431int inode_wait(void *word)
1432{
1433 schedule();
1434 return 0;
1435}
Stephen Rothwelld44dab82008-11-10 17:06:05 +11001436EXPORT_SYMBOL(inode_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
1438/*
Miklos Szeredi168a9fd2005-07-12 13:58:10 -07001439 * If we try to find an inode in the inode hash while it is being
1440 * deleted, we have to wait until the filesystem completes its
1441 * deletion before reporting that it isn't found. This function waits
1442 * until the deletion _might_ have completed. Callers are responsible
1443 * to recheck inode state.
1444 *
1445 * It doesn't matter if I_LOCK is not set initially, a call to
1446 * wake_up_inode() after removing from the hash list will DTRT.
1447 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 * This is called with inode_lock held.
1449 */
1450static void __wait_on_freeing_inode(struct inode *inode)
1451{
1452 wait_queue_head_t *wq;
1453 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_LOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 wq = bit_waitqueue(&inode->i_state, __I_LOCK);
1455 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
1456 spin_unlock(&inode_lock);
1457 schedule();
1458 finish_wait(wq, &wait.wait);
1459 spin_lock(&inode_lock);
1460}
1461
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462static __initdata unsigned long ihash_entries;
1463static int __init set_ihash_entries(char *str)
1464{
1465 if (!str)
1466 return 0;
1467 ihash_entries = simple_strtoul(str, &str, 0);
1468 return 1;
1469}
1470__setup("ihash_entries=", set_ihash_entries);
1471
1472/*
1473 * Initialize the waitqueues and inode hash table.
1474 */
1475void __init inode_init_early(void)
1476{
1477 int loop;
1478
1479 /* If hashes are distributed across NUMA nodes, defer
1480 * hash allocation until vmalloc space is available.
1481 */
1482 if (hashdist)
1483 return;
1484
1485 inode_hashtable =
1486 alloc_large_system_hash("Inode-cache",
1487 sizeof(struct hlist_head),
1488 ihash_entries,
1489 14,
1490 HASH_EARLY,
1491 &i_hash_shift,
1492 &i_hash_mask,
1493 0);
1494
1495 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1496 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1497}
1498
Denis Cheng74bf17c2007-10-16 23:26:30 -07001499void __init inode_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500{
1501 int loop;
1502
1503 /* inode slab cache */
Paul Jacksonb0196002006-03-24 03:16:09 -08001504 inode_cachep = kmem_cache_create("inode_cache",
1505 sizeof(struct inode),
1506 0,
1507 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1508 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +09001509 init_once);
Rusty Russell8e1f9362007-07-17 04:03:17 -07001510 register_shrinker(&icache_shrinker);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511
1512 /* Hash may have been set up in inode_init_early */
1513 if (!hashdist)
1514 return;
1515
1516 inode_hashtable =
1517 alloc_large_system_hash("Inode-cache",
1518 sizeof(struct hlist_head),
1519 ihash_entries,
1520 14,
1521 0,
1522 &i_hash_shift,
1523 &i_hash_mask,
1524 0);
1525
1526 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1527 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1528}
1529
1530void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1531{
1532 inode->i_mode = mode;
1533 if (S_ISCHR(mode)) {
1534 inode->i_fop = &def_chr_fops;
1535 inode->i_rdev = rdev;
1536 } else if (S_ISBLK(mode)) {
1537 inode->i_fop = &def_blk_fops;
1538 inode->i_rdev = rdev;
1539 } else if (S_ISFIFO(mode))
1540 inode->i_fop = &def_fifo_fops;
1541 else if (S_ISSOCK(mode))
1542 inode->i_fop = &bad_sock_fops;
1543 else
1544 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o)\n",
1545 mode);
1546}
1547EXPORT_SYMBOL(init_special_inode);