blob: 33c963d08ab49ecc4244a5e817db4af525794ce5 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#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>
Nick Piggin88e0fbc2009-09-22 16:43:50 -070016#include <linux/rwsem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/hash.h>
18#include <linux/swap.h>
19#include <linux/security.h>
20#include <linux/pagemap.h>
21#include <linux/cdev.h>
22#include <linux/bootmem.h>
Eric Paris3be25f42009-05-21 17:01:26 -040023#include <linux/fsnotify.h>
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -080024#include <linux/mount.h>
Arjan van de Venefaee192009-01-06 07:20:54 -080025#include <linux/async.h>
Al Virof19d4a82009-06-08 19:50:45 -040026#include <linux/posix_acl.h>
Eric Parisa178d202010-10-25 14:41:59 -040027#include <linux/ima.h>
Serge E. Hallyne795b712011-03-23 16:43:25 -070028#include <linux/cred.h>
Dave Chinnera66979a2011-03-22 22:23:41 +110029#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
31/*
Dave Chinner250df6e2011-03-22 22:23:36 +110032 * inode locking rules.
33 *
34 * inode->i_lock protects:
35 * inode->i_state, inode->i_hash, __iget()
Dave Chinner02afc412011-03-22 22:23:38 +110036 * inode_lru_lock protects:
37 * inode_lru, inode->i_lru
Dave Chinner55fa6092011-03-22 22:23:40 +110038 * inode_sb_list_lock protects:
39 * sb->s_inodes, inode->i_sb_list
Dave Chinnera66979a2011-03-22 22:23:41 +110040 * inode_wb_list_lock protects:
41 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
Dave Chinner67a23c42011-03-22 22:23:42 +110042 * inode_hash_lock protects:
43 * inode_hashtable, inode->i_hash
Dave Chinner250df6e2011-03-22 22:23:36 +110044 *
45 * Lock ordering:
Dave Chinner55fa6092011-03-22 22:23:40 +110046 *
47 * inode_sb_list_lock
48 * inode->i_lock
Dave Chinner02afc412011-03-22 22:23:38 +110049 * inode_lru_lock
Dave Chinnera66979a2011-03-22 22:23:41 +110050 *
51 * inode_wb_list_lock
52 * inode->i_lock
Dave Chinner67a23c42011-03-22 22:23:42 +110053 *
54 * inode_hash_lock
55 * inode_sb_list_lock
56 * inode->i_lock
57 *
58 * iunique_lock
59 * inode_hash_lock
Dave Chinner250df6e2011-03-22 22:23:36 +110060 */
61
62/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 * This is needed for the following functions:
64 * - inode_has_buffers
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 * - invalidate_bdev
66 *
67 * FIXME: remove all knowledge of the buffer layer from this file
68 */
69#include <linux/buffer_head.h>
70
71/*
72 * New inode.c implementation.
73 *
74 * This implementation has the basic premise of trying
75 * to be extremely low-overhead and SMP-safe, yet be
76 * simple enough to be "obviously correct".
77 *
78 * Famous last words.
79 */
80
81/* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
82
83/* #define INODE_PARANOIA 1 */
84/* #define INODE_DEBUG 1 */
85
86/*
87 * Inode lookup is no longer as critical as it used to be:
88 * most of the lookups are going to be through the dcache.
89 */
90#define I_HASHBITS i_hash_shift
91#define I_HASHMASK i_hash_mask
92
Eric Dumazetfa3536c2006-03-26 01:37:24 -080093static unsigned int i_hash_mask __read_mostly;
94static unsigned int i_hash_shift __read_mostly;
Dave Chinner67a23c42011-03-22 22:23:42 +110095static struct hlist_head *inode_hashtable __read_mostly;
96static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
98/*
99 * Each inode can be on two separate lists. One is
100 * the hash list of the inode, used for lookups. The
101 * other linked list is the "type" list:
102 * "in_use" - valid inode, i_count > 0, i_nlink > 0
103 * "dirty" - as "in_use" but also dirty
104 * "unused" - valid inode, i_count = 0
105 *
106 * A "dirty" list is maintained for each super block,
107 * allowing for low-overhead inode sync() operations.
108 */
109
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100110static LIST_HEAD(inode_lru);
Dave Chinner02afc412011-03-22 22:23:38 +1100111static DEFINE_SPINLOCK(inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Dave Chinner55fa6092011-03-22 22:23:40 +1100113__cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
Dave Chinnera66979a2011-03-22 22:23:41 +1100114__cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_wb_list_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +1100115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116/*
Christoph Hellwigbab1d942011-03-15 21:51:24 +0100117 * iprune_sem provides exclusion between the icache shrinking and the
118 * umount path.
Nick Piggin88e0fbc2009-09-22 16:43:50 -0700119 *
Christoph Hellwigbab1d942011-03-15 21:51:24 +0100120 * We don't actually need it to protect anything in the umount path,
121 * but only need to cycle through it to make sure any inode that
122 * prune_icache took off the LRU list has been fully torn down by the
123 * time we are past evict_inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 */
Nick Piggin88e0fbc2009-09-22 16:43:50 -0700125static DECLARE_RWSEM(iprune_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126
127/*
Jens Axboe7dcda1c2011-04-05 23:51:48 +0200128 * Empty aops. Can be used for the cases where the user does not
129 * define any of the address_space operations.
130 */
131const struct address_space_operations empty_aops = {
132};
133EXPORT_SYMBOL(empty_aops);
134
135/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 * Statistics gathering..
137 */
138struct inodes_stat_t inodes_stat;
139
Nick Piggin3e880fb2011-01-07 17:49:19 +1100140static DEFINE_PER_CPU(unsigned int, nr_inodes);
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400141
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530142static struct kmem_cache *inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
Nick Piggin3e880fb2011-01-07 17:49:19 +1100144static int get_nr_inodes(void)
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400145{
Nick Piggin3e880fb2011-01-07 17:49:19 +1100146 int i;
147 int sum = 0;
148 for_each_possible_cpu(i)
149 sum += per_cpu(nr_inodes, i);
150 return sum < 0 ? 0 : sum;
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400151}
152
153static inline int get_nr_inodes_unused(void)
154{
Nick Piggin86c87492011-01-07 17:49:18 +1100155 return inodes_stat.nr_unused;
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400156}
157
158int get_nr_dirty_inodes(void)
159{
Nick Piggin3e880fb2011-01-07 17:49:19 +1100160 /* not actually dirty inodes, but a wild approximation */
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400161 int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
162 return nr_dirty > 0 ? nr_dirty : 0;
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400163}
164
165/*
166 * Handle nr_inode sysctl
167 */
168#ifdef CONFIG_SYSCTL
169int proc_nr_inodes(ctl_table *table, int write,
170 void __user *buffer, size_t *lenp, loff_t *ppos)
171{
172 inodes_stat.nr_inodes = get_nr_inodes();
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400173 return proc_dointvec(table, write, buffer, lenp, ppos);
174}
175#endif
176
David Chinner2cb15992008-10-30 17:32:23 +1100177/**
178 * inode_init_always - perform inode structure intialisation
Randy Dunlap0bc02f32009-01-06 14:41:13 -0800179 * @sb: superblock inode belongs to
180 * @inode: inode to initialise
David Chinner2cb15992008-10-30 17:32:23 +1100181 *
182 * These are initializations that need to be done on every inode
183 * allocation as the fields are not initialised by slab allocation.
184 */
Christoph Hellwig54e34622009-08-07 14:38:25 -0300185int inode_init_always(struct super_block *sb, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186{
Alexey Dobriyan6e1d5dc2009-09-21 17:01:11 -0700187 static const struct inode_operations empty_iops;
Arjan van de Ven99ac48f2006-03-28 01:56:41 -0800188 static const struct file_operations empty_fops;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530189 struct address_space *const mapping = &inode->i_data;
David Chinner2cb15992008-10-30 17:32:23 +1100190
191 inode->i_sb = sb;
192 inode->i_blkbits = sb->s_blocksize_bits;
193 inode->i_flags = 0;
194 atomic_set(&inode->i_count, 1);
195 inode->i_op = &empty_iops;
196 inode->i_fop = &empty_fops;
197 inode->i_nlink = 1;
Al Viro56ff5ef2008-12-09 09:34:39 -0500198 inode->i_uid = 0;
199 inode->i_gid = 0;
David Chinner2cb15992008-10-30 17:32:23 +1100200 atomic_set(&inode->i_writecount, 0);
201 inode->i_size = 0;
202 inode->i_blocks = 0;
203 inode->i_bytes = 0;
204 inode->i_generation = 0;
205#ifdef CONFIG_QUOTA
206 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
207#endif
208 inode->i_pipe = NULL;
209 inode->i_bdev = NULL;
210 inode->i_cdev = NULL;
211 inode->i_rdev = 0;
212 inode->dirtied_when = 0;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500213
214 if (security_inode_alloc(inode))
Christoph Hellwig54e34622009-08-07 14:38:25 -0300215 goto out;
David Chinner2cb15992008-10-30 17:32:23 +1100216 spin_lock_init(&inode->i_lock);
217 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
218
219 mutex_init(&inode->i_mutex);
220 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
221
222 init_rwsem(&inode->i_alloc_sem);
223 lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key);
224
225 mapping->a_ops = &empty_aops;
226 mapping->host = inode;
227 mapping->flags = 0;
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800228 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
David Chinner2cb15992008-10-30 17:32:23 +1100229 mapping->assoc_mapping = NULL;
230 mapping->backing_dev_info = &default_backing_dev_info;
231 mapping->writeback_index = 0;
232
233 /*
234 * If the block_device provides a backing_dev_info for client
235 * inodes then use that. Otherwise the inode share the bdev's
236 * backing_dev_info.
237 */
238 if (sb->s_bdev) {
239 struct backing_dev_info *bdi;
240
Jens Axboe2c96ce92009-09-15 09:43:56 +0200241 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
David Chinner2cb15992008-10-30 17:32:23 +1100242 mapping->backing_dev_info = bdi;
243 }
244 inode->i_private = NULL;
245 inode->i_mapping = mapping;
Al Virof19d4a82009-06-08 19:50:45 -0400246#ifdef CONFIG_FS_POSIX_ACL
247 inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
248#endif
David Chinner2cb15992008-10-30 17:32:23 +1100249
Eric Paris3be25f42009-05-21 17:01:26 -0400250#ifdef CONFIG_FSNOTIFY
251 inode->i_fsnotify_mask = 0;
252#endif
253
Nick Piggin3e880fb2011-01-07 17:49:19 +1100254 this_cpu_inc(nr_inodes);
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400255
Christoph Hellwig54e34622009-08-07 14:38:25 -0300256 return 0;
Christoph Hellwig54e34622009-08-07 14:38:25 -0300257out:
258 return -ENOMEM;
David Chinner2cb15992008-10-30 17:32:23 +1100259}
260EXPORT_SYMBOL(inode_init_always);
261
262static struct inode *alloc_inode(struct super_block *sb)
263{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 struct inode *inode;
265
266 if (sb->s_op->alloc_inode)
267 inode = sb->s_op->alloc_inode(sb);
268 else
David Chinner2cb15992008-10-30 17:32:23 +1100269 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Christoph Hellwig54e34622009-08-07 14:38:25 -0300271 if (!inode)
272 return NULL;
273
274 if (unlikely(inode_init_always(sb, inode))) {
275 if (inode->i_sb->s_op->destroy_inode)
276 inode->i_sb->s_op->destroy_inode(inode);
277 else
278 kmem_cache_free(inode_cachep, inode);
279 return NULL;
280 }
281
282 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283}
284
Nick Pigginff0c7d12011-01-07 17:49:50 +1100285void free_inode_nonrcu(struct inode *inode)
286{
287 kmem_cache_free(inode_cachep, inode);
288}
289EXPORT_SYMBOL(free_inode_nonrcu);
290
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300291void __destroy_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
Eric Sesterhennb7542f82006-04-02 13:38:18 +0200293 BUG_ON(inode_has_buffers(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 security_inode_free(inode);
Eric Paris3be25f42009-05-21 17:01:26 -0400295 fsnotify_inode_delete(inode);
Al Virof19d4a82009-06-08 19:50:45 -0400296#ifdef CONFIG_FS_POSIX_ACL
297 if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED)
298 posix_acl_release(inode->i_acl);
299 if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
300 posix_acl_release(inode->i_default_acl);
301#endif
Nick Piggin3e880fb2011-01-07 17:49:19 +1100302 this_cpu_dec(nr_inodes);
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300303}
304EXPORT_SYMBOL(__destroy_inode);
305
Nick Pigginfa0d7e32011-01-07 17:49:49 +1100306static void i_callback(struct rcu_head *head)
307{
308 struct inode *inode = container_of(head, struct inode, i_rcu);
309 INIT_LIST_HEAD(&inode->i_dentry);
310 kmem_cache_free(inode_cachep, inode);
311}
312
Christoph Hellwig56b0dac2010-10-06 10:48:55 +0200313static void destroy_inode(struct inode *inode)
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300314{
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100315 BUG_ON(!list_empty(&inode->i_lru));
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300316 __destroy_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 if (inode->i_sb->s_op->destroy_inode)
318 inode->i_sb->s_op->destroy_inode(inode);
319 else
Nick Pigginfa0d7e32011-01-07 17:49:49 +1100320 call_rcu(&inode->i_rcu, i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100323void address_space_init_once(struct address_space *mapping)
324{
325 memset(mapping, 0, sizeof(*mapping));
326 INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
327 spin_lock_init(&mapping->tree_lock);
328 spin_lock_init(&mapping->i_mmap_lock);
329 INIT_LIST_HEAD(&mapping->private_list);
330 spin_lock_init(&mapping->private_lock);
331 INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
332 INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
333 mutex_init(&mapping->unmap_mutex);
334}
335EXPORT_SYMBOL(address_space_init_once);
336
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337/*
338 * These are initializations that only need to be done
339 * once, because the fields are idempotent across use
340 * of the inode, so let the slab aware of that.
341 */
342void inode_init_once(struct inode *inode)
343{
344 memset(inode, 0, sizeof(*inode));
345 INIT_HLIST_NODE(&inode->i_hash);
346 INIT_LIST_HEAD(&inode->i_dentry);
347 INIT_LIST_HEAD(&inode->i_devices);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100348 INIT_LIST_HEAD(&inode->i_wb_list);
349 INIT_LIST_HEAD(&inode->i_lru);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100350 address_space_init_once(&inode->i_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 i_size_ordered_init(inode);
Eric Paris3be25f42009-05-21 17:01:26 -0400352#ifdef CONFIG_FSNOTIFY
Eric Parise61ce862009-12-17 21:24:24 -0500353 INIT_HLIST_HEAD(&inode->i_fsnotify_marks);
Eric Paris3be25f42009-05-21 17:01:26 -0400354#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356EXPORT_SYMBOL(inode_init_once);
357
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700358static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530360 struct inode *inode = (struct inode *) foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
Christoph Lametera35afb82007-05-16 22:10:57 -0700362 inode_init_once(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363}
364
365/*
Dave Chinner250df6e2011-03-22 22:23:36 +1100366 * inode->i_lock must be held
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530368void __iget(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369{
Nick Piggin9e38d862010-10-23 06:55:17 -0400370 atomic_inc(&inode->i_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371}
372
Al Viro7de9c6e2010-10-23 11:11:40 -0400373/*
374 * get additional reference to inode; caller must already hold one.
375 */
376void ihold(struct inode *inode)
377{
378 WARN_ON(atomic_inc_return(&inode->i_count) < 2);
379}
380EXPORT_SYMBOL(ihold);
381
Nick Piggin9e38d862010-10-23 06:55:17 -0400382static void inode_lru_list_add(struct inode *inode)
383{
Dave Chinner02afc412011-03-22 22:23:38 +1100384 spin_lock(&inode_lru_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100385 if (list_empty(&inode->i_lru)) {
386 list_add(&inode->i_lru, &inode_lru);
Nick Piggin86c87492011-01-07 17:49:18 +1100387 inodes_stat.nr_unused++;
Nick Piggin9e38d862010-10-23 06:55:17 -0400388 }
Dave Chinner02afc412011-03-22 22:23:38 +1100389 spin_unlock(&inode_lru_lock);
Nick Piggin9e38d862010-10-23 06:55:17 -0400390}
391
392static void inode_lru_list_del(struct inode *inode)
393{
Dave Chinner02afc412011-03-22 22:23:38 +1100394 spin_lock(&inode_lru_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100395 if (!list_empty(&inode->i_lru)) {
396 list_del_init(&inode->i_lru);
Nick Piggin86c87492011-01-07 17:49:18 +1100397 inodes_stat.nr_unused--;
Nick Piggin9e38d862010-10-23 06:55:17 -0400398 }
Dave Chinner02afc412011-03-22 22:23:38 +1100399 spin_unlock(&inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400}
401
Christoph Hellwig646ec462010-10-23 07:15:32 -0400402/**
403 * inode_sb_list_add - add inode to the superblock list of inodes
404 * @inode: inode to add
405 */
406void inode_sb_list_add(struct inode *inode)
407{
Dave Chinner55fa6092011-03-22 22:23:40 +1100408 spin_lock(&inode_sb_list_lock);
409 list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
410 spin_unlock(&inode_sb_list_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400411}
412EXPORT_SYMBOL_GPL(inode_sb_list_add);
413
Dave Chinner55fa6092011-03-22 22:23:40 +1100414static inline void inode_sb_list_del(struct inode *inode)
Christoph Hellwig646ec462010-10-23 07:15:32 -0400415{
Dave Chinner55fa6092011-03-22 22:23:40 +1100416 spin_lock(&inode_sb_list_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400417 list_del_init(&inode->i_sb_list);
Dave Chinner55fa6092011-03-22 22:23:40 +1100418 spin_unlock(&inode_sb_list_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400419}
420
Dave Chinner4c51acb2010-10-23 06:58:09 -0400421static unsigned long hash(struct super_block *sb, unsigned long hashval)
422{
423 unsigned long tmp;
424
425 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
426 L1_CACHE_BYTES;
427 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
428 return tmp & I_HASHMASK;
429}
430
431/**
432 * __insert_inode_hash - hash an inode
433 * @inode: unhashed inode
434 * @hashval: unsigned long value used to locate this object in the
435 * inode_hashtable.
436 *
437 * Add an inode to the inode hash for this superblock.
438 */
439void __insert_inode_hash(struct inode *inode, unsigned long hashval)
440{
Christoph Hellwig646ec462010-10-23 07:15:32 -0400441 struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
442
Dave Chinner67a23c42011-03-22 22:23:42 +1100443 spin_lock(&inode_hash_lock);
Dave Chinner250df6e2011-03-22 22:23:36 +1100444 spin_lock(&inode->i_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400445 hlist_add_head(&inode->i_hash, b);
Dave Chinner250df6e2011-03-22 22:23:36 +1100446 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100447 spin_unlock(&inode_hash_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400448}
449EXPORT_SYMBOL(__insert_inode_hash);
450
451/**
Dave Chinner4c51acb2010-10-23 06:58:09 -0400452 * remove_inode_hash - remove an inode from the hash
453 * @inode: inode to unhash
454 *
455 * Remove an inode from the superblock.
456 */
457void remove_inode_hash(struct inode *inode)
458{
Dave Chinner67a23c42011-03-22 22:23:42 +1100459 spin_lock(&inode_hash_lock);
Dave Chinner250df6e2011-03-22 22:23:36 +1100460 spin_lock(&inode->i_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400461 hlist_del_init(&inode->i_hash);
Dave Chinner250df6e2011-03-22 22:23:36 +1100462 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100463 spin_unlock(&inode_hash_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400464}
465EXPORT_SYMBOL(remove_inode_hash);
466
Al Virob0683aa2010-06-04 20:55:25 -0400467void end_writeback(struct inode *inode)
468{
469 might_sleep();
470 BUG_ON(inode->i_data.nrpages);
471 BUG_ON(!list_empty(&inode->i_data.private_list));
472 BUG_ON(!(inode->i_state & I_FREEING));
473 BUG_ON(inode->i_state & I_CLEAR);
474 inode_sync_wait(inode);
Nick Pigginfa0d7e32011-01-07 17:49:49 +1100475 /* don't need i_lock here, no concurrent mods to i_state */
Al Virob0683aa2010-06-04 20:55:25 -0400476 inode->i_state = I_FREEING | I_CLEAR;
477}
478EXPORT_SYMBOL(end_writeback);
479
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100480/*
481 * Free the inode passed in, removing it from the lists it is still connected
482 * to. We remove any pages still attached to the inode and wait for any IO that
483 * is still in progress before finally destroying the inode.
484 *
485 * An inode must already be marked I_FREEING so that we avoid the inode being
486 * moved back onto lists if we race with other code that manipulates the lists
487 * (e.g. writeback_single_inode). The caller is responsible for setting this.
488 *
489 * An inode must already be removed from the LRU list before being evicted from
490 * the cache. This should occur atomically with setting the I_FREEING state
491 * flag, so no inodes here should ever be on the LRU when being evicted.
492 */
Al Viro644da592010-06-07 13:21:05 -0400493static void evict(struct inode *inode)
Al Virob4272d42010-06-04 19:33:20 -0400494{
495 const struct super_operations *op = inode->i_sb->s_op;
496
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100497 BUG_ON(!(inode->i_state & I_FREEING));
498 BUG_ON(!list_empty(&inode->i_lru));
499
Dave Chinnera66979a2011-03-22 22:23:41 +1100500 inode_wb_list_del(inode);
Dave Chinner55fa6092011-03-22 22:23:40 +1100501 inode_sb_list_del(inode);
502
Al Virobe7ce412010-06-04 19:40:39 -0400503 if (op->evict_inode) {
504 op->evict_inode(inode);
Al Virob4272d42010-06-04 19:33:20 -0400505 } else {
506 if (inode->i_data.nrpages)
507 truncate_inode_pages(&inode->i_data, 0);
Al Viro30140832010-06-07 13:23:20 -0400508 end_writeback(inode);
Al Virob4272d42010-06-04 19:33:20 -0400509 }
Al Viro661074e2010-06-04 20:19:55 -0400510 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
511 bd_forget(inode);
512 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
513 cd_forget(inode);
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100514
515 remove_inode_hash(inode);
516
517 spin_lock(&inode->i_lock);
518 wake_up_bit(&inode->i_state, __I_NEW);
519 BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
520 spin_unlock(&inode->i_lock);
521
522 destroy_inode(inode);
Al Virob4272d42010-06-04 19:33:20 -0400523}
524
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525/*
526 * dispose_list - dispose of the contents of a local list
527 * @head: the head of the list to free
528 *
529 * Dispose-list gets a local list with local inodes in it, so it doesn't
530 * need to worry about list corruption and SMP locks.
531 */
532static void dispose_list(struct list_head *head)
533{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 while (!list_empty(head)) {
535 struct inode *inode;
536
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100537 inode = list_first_entry(head, struct inode, i_lru);
538 list_del_init(&inode->i_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Al Viro644da592010-06-07 13:21:05 -0400540 evict(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542}
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544/**
Al Viro63997e92010-10-25 20:49:35 -0400545 * evict_inodes - evict all evictable inodes for a superblock
546 * @sb: superblock to operate on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 *
Al Viro63997e92010-10-25 20:49:35 -0400548 * Make sure that no inodes with zero refcount are retained. This is
549 * called by superblock shutdown after having MS_ACTIVE flag removed,
550 * so any inode reaching zero refcount during or after that call will
551 * be immediately evicted.
552 */
553void evict_inodes(struct super_block *sb)
554{
555 struct inode *inode, *next;
556 LIST_HEAD(dispose);
557
Dave Chinner55fa6092011-03-22 22:23:40 +1100558 spin_lock(&inode_sb_list_lock);
Al Viro63997e92010-10-25 20:49:35 -0400559 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
560 if (atomic_read(&inode->i_count))
561 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100562
563 spin_lock(&inode->i_lock);
564 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
565 spin_unlock(&inode->i_lock);
Al Viro63997e92010-10-25 20:49:35 -0400566 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100567 }
Al Viro63997e92010-10-25 20:49:35 -0400568
569 inode->i_state |= I_FREEING;
Dave Chinner02afc412011-03-22 22:23:38 +1100570 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100571 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100572 list_add(&inode->i_lru, &dispose);
Al Viro63997e92010-10-25 20:49:35 -0400573 }
Dave Chinner55fa6092011-03-22 22:23:40 +1100574 spin_unlock(&inode_sb_list_lock);
Al Viro63997e92010-10-25 20:49:35 -0400575
576 dispose_list(&dispose);
Christoph Hellwigbab1d942011-03-15 21:51:24 +0100577
578 /*
579 * Cycle through iprune_sem to make sure any inode that prune_icache
580 * moved off the list before we took the lock has been fully torn
581 * down.
582 */
583 down_write(&iprune_sem);
Al Viro63997e92010-10-25 20:49:35 -0400584 up_write(&iprune_sem);
585}
586
587/**
Christoph Hellwiga0318782010-10-24 19:40:33 +0200588 * invalidate_inodes - attempt to free all inodes on a superblock
589 * @sb: superblock to operate on
NeilBrown93b270f2011-02-24 17:25:47 +1100590 * @kill_dirty: flag to guide handling of dirty inodes
Christoph Hellwiga0318782010-10-24 19:40:33 +0200591 *
592 * Attempts to free all inodes for a given superblock. If there were any
593 * busy inodes return a non-zero value, else zero.
NeilBrown93b270f2011-02-24 17:25:47 +1100594 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
595 * them as busy.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 */
NeilBrown93b270f2011-02-24 17:25:47 +1100597int invalidate_inodes(struct super_block *sb, bool kill_dirty)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598{
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400599 int busy = 0;
Christoph Hellwiga0318782010-10-24 19:40:33 +0200600 struct inode *inode, *next;
601 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
Dave Chinner55fa6092011-03-22 22:23:40 +1100603 spin_lock(&inode_sb_list_lock);
Christoph Hellwiga0318782010-10-24 19:40:33 +0200604 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100605 spin_lock(&inode->i_lock);
606 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
607 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100609 }
NeilBrown93b270f2011-02-24 17:25:47 +1100610 if (inode->i_state & I_DIRTY && !kill_dirty) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100611 spin_unlock(&inode->i_lock);
NeilBrown93b270f2011-02-24 17:25:47 +1100612 busy = 1;
613 continue;
614 }
Christoph Hellwig99a38912010-10-23 19:07:20 +0200615 if (atomic_read(&inode->i_count)) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100616 spin_unlock(&inode->i_lock);
Christoph Hellwig99a38912010-10-23 19:07:20 +0200617 busy = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 continue;
619 }
Christoph Hellwig99a38912010-10-23 19:07:20 +0200620
Christoph Hellwig99a38912010-10-23 19:07:20 +0200621 inode->i_state |= I_FREEING;
Dave Chinner02afc412011-03-22 22:23:38 +1100622 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100623 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100624 list_add(&inode->i_lru, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 }
Dave Chinner55fa6092011-03-22 22:23:40 +1100626 spin_unlock(&inode_sb_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Christoph Hellwiga0318782010-10-24 19:40:33 +0200628 dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
630 return busy;
631}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632
633static int can_unuse(struct inode *inode)
634{
Nick Piggin9e38d862010-10-23 06:55:17 -0400635 if (inode->i_state & ~I_REFERENCED)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 return 0;
637 if (inode_has_buffers(inode))
638 return 0;
639 if (atomic_read(&inode->i_count))
640 return 0;
641 if (inode->i_data.nrpages)
642 return 0;
643 return 1;
644}
645
646/*
Nick Piggin9e38d862010-10-23 06:55:17 -0400647 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
Dave Chinner02afc412011-03-22 22:23:38 +1100648 * temporary list and then are freed outside inode_lru_lock by dispose_list().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 *
650 * Any inodes which are pinned purely because of attached pagecache have their
Nick Piggin9e38d862010-10-23 06:55:17 -0400651 * pagecache removed. If the inode has metadata buffers attached to
652 * mapping->private_list then try to remove them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 *
Nick Piggin9e38d862010-10-23 06:55:17 -0400654 * If the inode has the I_REFERENCED flag set, then it means that it has been
655 * used recently - the flag is set in iput_final(). When we encounter such an
656 * inode, clear the flag and move it to the back of the LRU so it gets another
657 * pass through the LRU before it gets reclaimed. This is necessary because of
658 * the fact we are doing lazy LRU updates to minimise lock contention so the
659 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
660 * with this flag set because they are the inodes that are out of order.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 */
662static void prune_icache(int nr_to_scan)
663{
664 LIST_HEAD(freeable);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 int nr_scanned;
666 unsigned long reap = 0;
667
Nick Piggin88e0fbc2009-09-22 16:43:50 -0700668 down_read(&iprune_sem);
Dave Chinner02afc412011-03-22 22:23:38 +1100669 spin_lock(&inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) {
671 struct inode *inode;
672
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100673 if (list_empty(&inode_lru))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 break;
675
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100676 inode = list_entry(inode_lru.prev, struct inode, i_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Nick Piggin9e38d862010-10-23 06:55:17 -0400678 /*
Dave Chinner02afc412011-03-22 22:23:38 +1100679 * we are inverting the inode_lru_lock/inode->i_lock here,
680 * so use a trylock. If we fail to get the lock, just move the
681 * inode to the back of the list so we don't spin on it.
682 */
683 if (!spin_trylock(&inode->i_lock)) {
684 list_move(&inode->i_lru, &inode_lru);
685 continue;
686 }
687
688 /*
Nick Piggin9e38d862010-10-23 06:55:17 -0400689 * Referenced or dirty inodes are still in use. Give them
690 * another pass through the LRU as we canot reclaim them now.
691 */
692 if (atomic_read(&inode->i_count) ||
693 (inode->i_state & ~I_REFERENCED)) {
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100694 list_del_init(&inode->i_lru);
Dave Chinnerf283c862011-03-22 22:23:39 +1100695 spin_unlock(&inode->i_lock);
Nick Piggin86c87492011-01-07 17:49:18 +1100696 inodes_stat.nr_unused--;
Nick Piggin9e38d862010-10-23 06:55:17 -0400697 continue;
698 }
699
700 /* recently referenced inodes get one more pass */
701 if (inode->i_state & I_REFERENCED) {
Nick Piggin9e38d862010-10-23 06:55:17 -0400702 inode->i_state &= ~I_REFERENCED;
Dave Chinner250df6e2011-03-22 22:23:36 +1100703 list_move(&inode->i_lru, &inode_lru);
Dave Chinnerf283c862011-03-22 22:23:39 +1100704 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 continue;
706 }
707 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
708 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100709 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100710 spin_unlock(&inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 if (remove_inode_buffers(inode))
Andrew Mortonfc0ecff2007-02-10 01:45:39 -0800712 reap += invalidate_mapping_pages(&inode->i_data,
713 0, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 iput(inode);
Dave Chinner02afc412011-03-22 22:23:38 +1100715 spin_lock(&inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100717 if (inode != list_entry(inode_lru.next,
718 struct inode, i_lru))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 continue; /* wrong inode or list_empty */
Dave Chinner02afc412011-03-22 22:23:38 +1100720 /* avoid lock inversions with trylock */
721 if (!spin_trylock(&inode->i_lock))
722 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100723 if (!can_unuse(inode)) {
724 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100726 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 }
Nick Piggin7ef0d732009-03-12 14:31:38 -0700728 WARN_ON(inode->i_state & I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 inode->i_state |= I_FREEING;
Dave Chinner250df6e2011-03-22 22:23:36 +1100730 spin_unlock(&inode->i_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100731
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100732 list_move(&inode->i_lru, &freeable);
Nick Piggin86c87492011-01-07 17:49:18 +1100733 inodes_stat.nr_unused--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 }
Christoph Lameterf8891e52006-06-30 01:55:45 -0700735 if (current_is_kswapd())
736 __count_vm_events(KSWAPD_INODESTEAL, reap);
737 else
738 __count_vm_events(PGINODESTEAL, reap);
Dave Chinner02afc412011-03-22 22:23:38 +1100739 spin_unlock(&inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740
741 dispose_list(&freeable);
Nick Piggin88e0fbc2009-09-22 16:43:50 -0700742 up_read(&iprune_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743}
744
745/*
746 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
747 * "unused" means that no dentries are referring to the inodes: the files are
748 * not open and the dcache references to those inodes have already been
749 * reclaimed.
750 *
751 * This function is passed the number of inodes to scan, and it returns the
752 * total number of remaining possibly-reclaimable inodes.
753 */
Dave Chinner7f8275d2010-07-19 14:56:17 +1000754static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
756 if (nr) {
757 /*
758 * Nasty deadlock avoidance. We may hold various FS locks,
759 * and we don't want to recurse into the FS that called us
760 * in clear_inode() and friends..
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530761 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 if (!(gfp_mask & __GFP_FS))
763 return -1;
764 prune_icache(nr);
765 }
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400766 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767}
768
Rusty Russell8e1f9362007-07-17 04:03:17 -0700769static struct shrinker icache_shrinker = {
770 .shrink = shrink_icache_memory,
771 .seeks = DEFAULT_SEEKS,
772};
773
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774static void __wait_on_freeing_inode(struct inode *inode);
775/*
776 * Called with the inode lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530778static struct inode *find_inode(struct super_block *sb,
779 struct hlist_head *head,
780 int (*test)(struct inode *, void *),
781 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782{
783 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530784 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785
786repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700787 hlist_for_each_entry(inode, node, head, i_hash) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100788 spin_lock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100789 if (inode->i_sb != sb) {
790 spin_unlock(&inode->i_lock);
791 continue;
792 }
793 if (!test(inode, data)) {
794 spin_unlock(&inode->i_lock);
795 continue;
796 }
Al Viroa4ffdde2010-06-02 17:38:30 -0400797 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 __wait_on_freeing_inode(inode);
799 goto repeat;
800 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400801 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100802 spin_unlock(&inode->i_lock);
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400803 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400805 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806}
807
808/*
809 * find_inode_fast is the fast path version of find_inode, see the comment at
810 * iget_locked for details.
811 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530812static struct inode *find_inode_fast(struct super_block *sb,
813 struct hlist_head *head, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814{
815 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530816 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
818repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700819 hlist_for_each_entry(inode, node, head, i_hash) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100820 spin_lock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100821 if (inode->i_ino != ino) {
822 spin_unlock(&inode->i_lock);
823 continue;
824 }
825 if (inode->i_sb != sb) {
826 spin_unlock(&inode->i_lock);
827 continue;
828 }
Al Viroa4ffdde2010-06-02 17:38:30 -0400829 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 __wait_on_freeing_inode(inode);
831 goto repeat;
832 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400833 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100834 spin_unlock(&inode->i_lock);
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400835 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400837 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
839
Eric Dumazetf991bd22010-10-23 11:18:01 -0400840/*
841 * Each cpu owns a range of LAST_INO_BATCH numbers.
842 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
843 * to renew the exhausted range.
David Chinner8290c352008-10-30 17:35:24 +1100844 *
Eric Dumazetf991bd22010-10-23 11:18:01 -0400845 * This does not significantly increase overflow rate because every CPU can
846 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
847 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
848 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
849 * overflow rate by 2x, which does not seem too significant.
850 *
851 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
852 * error if st_ino won't fit in target struct field. Use 32bit counter
853 * here to attempt to avoid that.
David Chinner8290c352008-10-30 17:35:24 +1100854 */
Eric Dumazetf991bd22010-10-23 11:18:01 -0400855#define LAST_INO_BATCH 1024
856static DEFINE_PER_CPU(unsigned int, last_ino);
David Chinner8290c352008-10-30 17:35:24 +1100857
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400858unsigned int get_next_ino(void)
Eric Dumazetf991bd22010-10-23 11:18:01 -0400859{
860 unsigned int *p = &get_cpu_var(last_ino);
861 unsigned int res = *p;
862
863#ifdef CONFIG_SMP
864 if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
865 static atomic_t shared_last_ino;
866 int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
867
868 res = next - LAST_INO_BATCH;
869 }
870#endif
871
872 *p = ++res;
873 put_cpu_var(last_ino);
874 return res;
David Chinner8290c352008-10-30 17:35:24 +1100875}
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400876EXPORT_SYMBOL(get_next_ino);
David Chinner8290c352008-10-30 17:35:24 +1100877
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878/**
879 * new_inode - obtain an inode
880 * @sb: superblock
881 *
Mel Gorman769848c2007-07-17 04:03:05 -0700882 * Allocates a new inode for given superblock. The default gfp_mask
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800883 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
Mel Gorman769848c2007-07-17 04:03:05 -0700884 * If HIGHMEM pages are unsuitable or it is known that pages allocated
885 * for the page cache are not reclaimable or migratable,
886 * mapping_set_gfp_mask() must be called with suitable flags on the
887 * newly created inode's mapping
888 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 */
890struct inode *new_inode(struct super_block *sb)
891{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530892 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
Dave Chinner55fa6092011-03-22 22:23:40 +1100894 spin_lock_prefetch(&inode_sb_list_lock);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530895
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 inode = alloc_inode(sb);
897 if (inode) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100898 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 inode->i_state = 0;
Dave Chinner250df6e2011-03-22 22:23:36 +1100900 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +1100901 inode_sb_list_add(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 }
903 return inode;
904}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905EXPORT_SYMBOL(new_inode);
906
Dave Chinner250df6e2011-03-22 22:23:36 +1100907/**
908 * unlock_new_inode - clear the I_NEW state and wake up any waiters
909 * @inode: new inode to unlock
910 *
911 * Called when the inode is fully initialised to clear the new state of the
912 * inode and wake up anyone waiting for the inode to finish initialisation.
913 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914void unlock_new_inode(struct inode *inode)
915{
Peter Zijlstra14358e62007-10-14 01:38:33 +0200916#ifdef CONFIG_DEBUG_LOCK_ALLOC
Namhyung Kima3314a02010-10-11 22:38:00 +0900917 if (S_ISDIR(inode->i_mode)) {
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200918 struct file_system_type *type = inode->i_sb->s_type;
919
Jan Kara9a7aa122009-06-04 15:26:49 +0200920 /* Set new key only if filesystem hasn't already changed it */
921 if (!lockdep_match_class(&inode->i_mutex,
922 &type->i_mutex_key)) {
923 /*
924 * ensure nobody is actually holding i_mutex
925 */
926 mutex_destroy(&inode->i_mutex);
927 mutex_init(&inode->i_mutex);
928 lockdep_set_class(&inode->i_mutex,
929 &type->i_mutex_dir_key);
930 }
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200931 }
Peter Zijlstra14358e62007-10-14 01:38:33 +0200932#endif
Dave Chinner250df6e2011-03-22 22:23:36 +1100933 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +0100934 WARN_ON(!(inode->i_state & I_NEW));
935 inode->i_state &= ~I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +1100936 wake_up_bit(&inode->i_state, __I_NEW);
937 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939EXPORT_SYMBOL(unlock_new_inode);
940
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400941/**
942 * iget5_locked - obtain an inode from a mounted file system
943 * @sb: super block of file system
944 * @hashval: hash value (usually inode number) to get
945 * @test: callback used for comparisons between inodes
946 * @set: callback used to initialize a new struct inode
947 * @data: opaque data pointer to pass to @test and @set
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400949 * Search for the inode specified by @hashval and @data in the inode cache,
950 * and if present it is return it with an increased reference count. This is
951 * a generalized version of iget_locked() for file systems where the inode
952 * number is not sufficient for unique identification of an inode.
953 *
954 * If the inode is not in cache, allocate a new inode and return it locked,
955 * hashed, and with the I_NEW flag set. The file system gets to fill it in
956 * before unlocking it via unlock_new_inode().
957 *
958 * Note both @test and @set are called with the inode_hash_lock held, so can't
959 * sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 */
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400961struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
962 int (*test)(struct inode *, void *),
963 int (*set)(struct inode *, void *), void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400965 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530966 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400968 spin_lock(&inode_hash_lock);
969 inode = find_inode(sb, head, test, data);
970 spin_unlock(&inode_hash_lock);
971
972 if (inode) {
973 wait_on_inode(inode);
974 return inode;
975 }
976
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 inode = alloc_inode(sb);
978 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530979 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
Dave Chinner67a23c42011-03-22 22:23:42 +1100981 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 /* We released the lock, so.. */
983 old = find_inode(sb, head, test, data);
984 if (!old) {
985 if (set(inode, data))
986 goto set_failed;
987
Dave Chinner250df6e2011-03-22 22:23:36 +1100988 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +0100989 inode->i_state = I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +1100990 hlist_add_head(&inode->i_hash, head);
991 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +1100992 inode_sb_list_add(inode);
Dave Chinner67a23c42011-03-22 22:23:42 +1100993 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
995 /* Return the locked inode with I_NEW set, the
996 * caller is responsible for filling in the contents
997 */
998 return inode;
999 }
1000
1001 /*
1002 * Uhhuh, somebody else created the same inode under
1003 * us. Use the old inode instead of the one we just
1004 * allocated.
1005 */
Dave Chinner67a23c42011-03-22 22:23:42 +11001006 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 destroy_inode(inode);
1008 inode = old;
1009 wait_on_inode(inode);
1010 }
1011 return inode;
1012
1013set_failed:
Dave Chinner67a23c42011-03-22 22:23:42 +11001014 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 destroy_inode(inode);
1016 return NULL;
1017}
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001018EXPORT_SYMBOL(iget5_locked);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001020/**
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 *
1025 * Search for the inode specified by @ino in the inode cache and if present
1026 * return it with an increased reference count. This is for file systems
1027 * where the inode number is sufficient for unique identification of an inode.
1028 *
1029 * If the inode is not in cache, allocate a new inode and return it locked,
1030 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1031 * before unlocking it via unlock_new_inode().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 */
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001033struct inode *iget_locked(struct super_block *sb, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001035 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301036 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001038 spin_lock(&inode_hash_lock);
1039 inode = find_inode_fast(sb, head, ino);
1040 spin_unlock(&inode_hash_lock);
1041 if (inode) {
1042 wait_on_inode(inode);
1043 return inode;
1044 }
1045
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 inode = alloc_inode(sb);
1047 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301048 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049
Dave Chinner67a23c42011-03-22 22:23:42 +11001050 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 /* We released the lock, so.. */
1052 old = find_inode_fast(sb, head, ino);
1053 if (!old) {
1054 inode->i_ino = ino;
Dave Chinner250df6e2011-03-22 22:23:36 +11001055 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001056 inode->i_state = I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +11001057 hlist_add_head(&inode->i_hash, head);
1058 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +11001059 inode_sb_list_add(inode);
Dave Chinner67a23c42011-03-22 22:23:42 +11001060 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
1062 /* Return the locked inode with I_NEW set, the
1063 * caller is responsible for filling in the contents
1064 */
1065 return inode;
1066 }
1067
1068 /*
1069 * Uhhuh, somebody else created the same inode under
1070 * us. Use the old inode instead of the one we just
1071 * allocated.
1072 */
Dave Chinner67a23c42011-03-22 22:23:42 +11001073 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 destroy_inode(inode);
1075 inode = old;
1076 wait_on_inode(inode);
1077 }
1078 return inode;
1079}
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001080EXPORT_SYMBOL(iget_locked);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001082/*
1083 * search the inode cache for a matching inode number.
1084 * If we find one, then the inode number we are trying to
1085 * allocate is not unique and so we should not use it.
1086 *
1087 * Returns 1 if the inode number is unique, 0 if it is not.
1088 */
1089static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1090{
1091 struct hlist_head *b = inode_hashtable + hash(sb, ino);
1092 struct hlist_node *node;
1093 struct inode *inode;
1094
Dave Chinner67a23c42011-03-22 22:23:42 +11001095 spin_lock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001096 hlist_for_each_entry(inode, node, b, i_hash) {
Dave Chinner67a23c42011-03-22 22:23:42 +11001097 if (inode->i_ino == ino && inode->i_sb == sb) {
1098 spin_unlock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001099 return 0;
Dave Chinner67a23c42011-03-22 22:23:42 +11001100 }
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001101 }
Dave Chinner67a23c42011-03-22 22:23:42 +11001102 spin_unlock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001103
1104 return 1;
1105}
1106
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107/**
1108 * iunique - get a unique inode number
1109 * @sb: superblock
1110 * @max_reserved: highest reserved inode number
1111 *
1112 * Obtain an inode number that is unique on the system for a given
1113 * superblock. This is used by file systems that have no natural
1114 * permanent inode numbering system. An inode number is returned that
1115 * is higher than the reserved limit but unique.
1116 *
1117 * BUGS:
1118 * With a large number of inodes live on the file system this function
1119 * currently becomes quite slow.
1120 */
1121ino_t iunique(struct super_block *sb, ino_t max_reserved)
1122{
Jeff Layton866b04f2007-05-08 00:32:29 -07001123 /*
1124 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1125 * error if st_ino won't fit in target struct field. Use 32bit counter
1126 * here to attempt to avoid that.
1127 */
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001128 static DEFINE_SPINLOCK(iunique_lock);
Jeff Layton866b04f2007-05-08 00:32:29 -07001129 static unsigned int counter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 ino_t res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001132 spin_lock(&iunique_lock);
Jeffrey Layton3361c7b2007-05-08 00:29:48 -07001133 do {
1134 if (counter <= max_reserved)
1135 counter = max_reserved + 1;
1136 res = counter++;
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001137 } while (!test_inode_iunique(sb, res));
1138 spin_unlock(&iunique_lock);
Jeffrey Layton3361c7b2007-05-08 00:29:48 -07001139
1140 return res;
1141}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142EXPORT_SYMBOL(iunique);
1143
1144struct inode *igrab(struct inode *inode)
1145{
Dave Chinner250df6e2011-03-22 22:23:36 +11001146 spin_lock(&inode->i_lock);
1147 if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +11001149 spin_unlock(&inode->i_lock);
1150 } else {
1151 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 /*
1153 * Handle the case where s_op->clear_inode is not been
1154 * called yet, and somebody is calling igrab
1155 * while the inode is getting freed.
1156 */
1157 inode = NULL;
Dave Chinner250df6e2011-03-22 22:23:36 +11001158 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 return inode;
1160}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161EXPORT_SYMBOL(igrab);
1162
1163/**
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001164 * ilookup5_nowait - search for an inode in the inode cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 * @sb: super block of file system to search
1166 * @hashval: hash value (usually inode number) to search for
1167 * @test: callback used for comparisons between inodes
1168 * @data: opaque data pointer to pass to @test
1169 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001170 * Search for the inode specified by @hashval and @data in the inode cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 * If the inode is in the cache, the inode is returned with an incremented
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001172 * reference count.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001173 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001174 * Note: I_NEW is not waited upon so you have to be very careful what you do
1175 * with the returned inode. You probably should be using ilookup5() instead.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001176 *
Randy Dunlapb6d0ad62011-03-26 13:27:47 -07001177 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001178 */
1179struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1180 int (*test)(struct inode *, void *), void *data)
1181{
1182 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001183 struct inode *inode;
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001184
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001185 spin_lock(&inode_hash_lock);
1186 inode = find_inode(sb, head, test, data);
1187 spin_unlock(&inode_hash_lock);
1188
1189 return inode;
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001190}
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001191EXPORT_SYMBOL(ilookup5_nowait);
1192
1193/**
1194 * ilookup5 - search for an inode in the inode cache
1195 * @sb: super block of file system to search
1196 * @hashval: hash value (usually inode number) to search for
1197 * @test: callback used for comparisons between inodes
1198 * @data: opaque data pointer to pass to @test
1199 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001200 * Search for the inode specified by @hashval and @data in the inode cache,
1201 * and if the inode is in the cache, return the inode with an incremented
1202 * reference count. Waits on I_NEW before returning the inode.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001203 * returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001205 * This is a generalized version of ilookup() for file systems where the
1206 * inode number is not sufficient for unique identification of an inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001208 * Note: @test is called with the inode_hash_lock held, so can't sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 */
1210struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1211 int (*test)(struct inode *, void *), void *data)
1212{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001213 struct inode *inode = ilookup5_nowait(sb, hashval, test, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001215 if (inode)
1216 wait_on_inode(inode);
1217 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219EXPORT_SYMBOL(ilookup5);
1220
1221/**
1222 * ilookup - search for an inode in the inode cache
1223 * @sb: super block of file system to search
1224 * @ino: inode number to search for
1225 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001226 * Search for the inode @ino in the inode cache, and if the inode is in the
1227 * cache, the inode is returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 */
1229struct inode *ilookup(struct super_block *sb, unsigned long ino)
1230{
1231 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001232 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001234 spin_lock(&inode_hash_lock);
1235 inode = find_inode_fast(sb, head, ino);
1236 spin_unlock(&inode_hash_lock);
1237
1238 if (inode)
1239 wait_on_inode(inode);
1240 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242EXPORT_SYMBOL(ilookup);
1243
Al Viro261bca82008-12-30 01:48:21 -05001244int insert_inode_locked(struct inode *inode)
1245{
1246 struct super_block *sb = inode->i_sb;
1247 ino_t ino = inode->i_ino;
1248 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Al Viro261bca82008-12-30 01:48:21 -05001249
Al Viro261bca82008-12-30 01:48:21 -05001250 while (1) {
Al Viro72a43d62009-05-13 19:13:40 +01001251 struct hlist_node *node;
1252 struct inode *old = NULL;
Dave Chinner67a23c42011-03-22 22:23:42 +11001253 spin_lock(&inode_hash_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001254 hlist_for_each_entry(old, node, head, i_hash) {
1255 if (old->i_ino != ino)
1256 continue;
1257 if (old->i_sb != sb)
1258 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001259 spin_lock(&old->i_lock);
1260 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1261 spin_unlock(&old->i_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001262 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001263 }
Al Viro72a43d62009-05-13 19:13:40 +01001264 break;
1265 }
1266 if (likely(!node)) {
Dave Chinner250df6e2011-03-22 22:23:36 +11001267 spin_lock(&inode->i_lock);
1268 inode->i_state |= I_NEW;
Al Viro261bca82008-12-30 01:48:21 -05001269 hlist_add_head(&inode->i_hash, head);
Dave Chinner250df6e2011-03-22 22:23:36 +11001270 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001271 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001272 return 0;
1273 }
1274 __iget(old);
Dave Chinner250df6e2011-03-22 22:23:36 +11001275 spin_unlock(&old->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001276 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001277 wait_on_inode(old);
Al Viro1d3382c2010-10-23 15:19:20 -04001278 if (unlikely(!inode_unhashed(old))) {
Al Viro261bca82008-12-30 01:48:21 -05001279 iput(old);
1280 return -EBUSY;
1281 }
1282 iput(old);
1283 }
1284}
Al Viro261bca82008-12-30 01:48:21 -05001285EXPORT_SYMBOL(insert_inode_locked);
1286
1287int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1288 int (*test)(struct inode *, void *), void *data)
1289{
1290 struct super_block *sb = inode->i_sb;
1291 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Al Viro261bca82008-12-30 01:48:21 -05001292
Al Viro261bca82008-12-30 01:48:21 -05001293 while (1) {
Al Viro72a43d62009-05-13 19:13:40 +01001294 struct hlist_node *node;
1295 struct inode *old = NULL;
1296
Dave Chinner67a23c42011-03-22 22:23:42 +11001297 spin_lock(&inode_hash_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001298 hlist_for_each_entry(old, node, head, i_hash) {
1299 if (old->i_sb != sb)
1300 continue;
1301 if (!test(old, data))
1302 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001303 spin_lock(&old->i_lock);
1304 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1305 spin_unlock(&old->i_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001306 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001307 }
Al Viro72a43d62009-05-13 19:13:40 +01001308 break;
1309 }
1310 if (likely(!node)) {
Dave Chinner250df6e2011-03-22 22:23:36 +11001311 spin_lock(&inode->i_lock);
1312 inode->i_state |= I_NEW;
Al Viro261bca82008-12-30 01:48:21 -05001313 hlist_add_head(&inode->i_hash, head);
Dave Chinner250df6e2011-03-22 22:23:36 +11001314 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001315 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001316 return 0;
1317 }
1318 __iget(old);
Dave Chinner250df6e2011-03-22 22:23:36 +11001319 spin_unlock(&old->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001320 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001321 wait_on_inode(old);
Al Viro1d3382c2010-10-23 15:19:20 -04001322 if (unlikely(!inode_unhashed(old))) {
Al Viro261bca82008-12-30 01:48:21 -05001323 iput(old);
1324 return -EBUSY;
1325 }
1326 iput(old);
1327 }
1328}
Al Viro261bca82008-12-30 01:48:21 -05001329EXPORT_SYMBOL(insert_inode_locked4);
1330
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331
Al Viro45321ac2010-06-07 13:43:19 -04001332int generic_delete_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333{
Al Viro45321ac2010-06-07 13:43:19 -04001334 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336EXPORT_SYMBOL(generic_delete_inode);
1337
Al Viro45321ac2010-06-07 13:43:19 -04001338/*
1339 * Normal UNIX filesystem behaviour: delete the
1340 * inode when the usage count drops to zero, and
1341 * i_nlink is zero.
Jan Kara22fe40422009-09-18 13:05:44 -07001342 */
Al Viro45321ac2010-06-07 13:43:19 -04001343int generic_drop_inode(struct inode *inode)
1344{
Al Viro1d3382c2010-10-23 15:19:20 -04001345 return !inode->i_nlink || inode_unhashed(inode);
Al Viro45321ac2010-06-07 13:43:19 -04001346}
1347EXPORT_SYMBOL_GPL(generic_drop_inode);
1348
1349/*
1350 * Called when we're dropping the last reference
1351 * to an inode.
1352 *
1353 * Call the FS "drop_inode()" function, defaulting to
1354 * the legacy UNIX filesystem behaviour. If it tells
1355 * us to evict inode, do so. Otherwise, retain inode
1356 * in cache if fs is alive, sync and evict if fs is
1357 * shutting down.
1358 */
1359static void iput_final(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360{
1361 struct super_block *sb = inode->i_sb;
Al Viro45321ac2010-06-07 13:43:19 -04001362 const struct super_operations *op = inode->i_sb->s_op;
1363 int drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
Dave Chinner250df6e2011-03-22 22:23:36 +11001365 WARN_ON(inode->i_state & I_NEW);
1366
Al Viro45321ac2010-06-07 13:43:19 -04001367 if (op && op->drop_inode)
1368 drop = op->drop_inode(inode);
1369 else
1370 drop = generic_drop_inode(inode);
1371
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001372 if (!drop && (sb->s_flags & MS_ACTIVE)) {
1373 inode->i_state |= I_REFERENCED;
1374 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
1375 inode_lru_list_add(inode);
1376 spin_unlock(&inode->i_lock);
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001377 return;
1378 }
1379
Al Viro45321ac2010-06-07 13:43:19 -04001380 if (!drop) {
Alexander Viro991114c2005-06-23 00:09:01 -07001381 inode->i_state |= I_WILL_FREE;
Dave Chinner250df6e2011-03-22 22:23:36 +11001382 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 write_inode_now(inode, 1);
Dave Chinner250df6e2011-03-22 22:23:36 +11001384 spin_lock(&inode->i_lock);
Nick Piggin7ef0d732009-03-12 14:31:38 -07001385 WARN_ON(inode->i_state & I_NEW);
Alexander Viro991114c2005-06-23 00:09:01 -07001386 inode->i_state &= ~I_WILL_FREE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 }
Nick Piggin7ccf19a2010-10-21 11:49:30 +11001388
Alexander Viro991114c2005-06-23 00:09:01 -07001389 inode->i_state |= I_FREEING;
Nick Piggin9e38d862010-10-23 06:55:17 -04001390 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +11001391 spin_unlock(&inode->i_lock);
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001392
1393 evict(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394}
1395
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396/**
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301397 * iput - put an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 * @inode: inode to put
1399 *
1400 * Puts an inode, dropping its usage count. If the inode use count hits
1401 * zero, the inode is then freed and may also be destroyed.
1402 *
1403 * Consequently, iput() can sleep.
1404 */
1405void iput(struct inode *inode)
1406{
1407 if (inode) {
Al Viroa4ffdde2010-06-02 17:38:30 -04001408 BUG_ON(inode->i_state & I_CLEAR);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409
Dave Chinnerf283c862011-03-22 22:23:39 +11001410 if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 iput_final(inode);
1412 }
1413}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414EXPORT_SYMBOL(iput);
1415
1416/**
1417 * bmap - find a block number in a file
1418 * @inode: inode of file
1419 * @block: block to find
1420 *
1421 * Returns the block number on the device holding the inode that
1422 * is the disk block number for the block of the file requested.
1423 * That is, asked for block 4 of inode 1 the function will return the
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301424 * disk block relative to the disk start that holds that block of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 * file.
1426 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301427sector_t bmap(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428{
1429 sector_t res = 0;
1430 if (inode->i_mapping->a_ops->bmap)
1431 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1432 return res;
1433}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434EXPORT_SYMBOL(bmap);
1435
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001436/*
1437 * With relative atime, only update atime if the previous atime is
1438 * earlier than either the ctime or mtime or if at least a day has
1439 * passed since the last atime update.
1440 */
1441static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1442 struct timespec now)
1443{
1444
1445 if (!(mnt->mnt_flags & MNT_RELATIME))
1446 return 1;
1447 /*
1448 * Is mtime younger than atime? If yes, update atime:
1449 */
1450 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1451 return 1;
1452 /*
1453 * Is ctime younger than atime? If yes, update atime:
1454 */
1455 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1456 return 1;
1457
1458 /*
1459 * Is the previous atime value older than a day? If yes,
1460 * update atime:
1461 */
1462 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1463 return 1;
1464 /*
1465 * Good, we can skip the atime update:
1466 */
1467 return 0;
1468}
1469
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470/**
Christoph Hellwig869243a2006-01-09 20:52:03 -08001471 * touch_atime - update the access time
1472 * @mnt: mount the inode is accessed on
Martin Waitz7045f372006-02-01 03:06:57 -08001473 * @dentry: dentry accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 *
1475 * Update the accessed time on an inode and mark it for writeback.
1476 * This function automatically handles read only file systems and media,
1477 * as well as the "noatime" flag and inode specific "noatime" markers.
1478 */
Christoph Hellwig869243a2006-01-09 20:52:03 -08001479void touch_atime(struct vfsmount *mnt, struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480{
Christoph Hellwig869243a2006-01-09 20:52:03 -08001481 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 struct timespec now;
1483
Andrew Mortonb2276132006-12-13 00:34:33 -08001484 if (inode->i_flags & S_NOATIME)
Andi Kleenb12536c2009-09-18 13:05:47 -07001485 return;
Eric Dumazet37756ce2007-02-10 01:44:49 -08001486 if (IS_NOATIME(inode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001487 return;
Andrew Mortonb2276132006-12-13 00:34:33 -08001488 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001489 return;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001490
Dave Hansencdb70f32008-02-15 14:37:41 -08001491 if (mnt->mnt_flags & MNT_NOATIME)
Andi Kleenb12536c2009-09-18 13:05:47 -07001492 return;
Dave Hansencdb70f32008-02-15 14:37:41 -08001493 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001494 return;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001495
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 now = current_fs_time(inode->i_sb);
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001497
1498 if (!relatime_need_update(mnt, inode, now))
Andi Kleenb12536c2009-09-18 13:05:47 -07001499 return;
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001500
Valerie Henson47ae32d2006-12-13 00:34:34 -08001501 if (timespec_equal(&inode->i_atime, &now))
Andi Kleenb12536c2009-09-18 13:05:47 -07001502 return;
1503
1504 if (mnt_want_write(mnt))
1505 return;
Valerie Henson47ae32d2006-12-13 00:34:34 -08001506
1507 inode->i_atime = now;
1508 mark_inode_dirty_sync(inode);
Dave Hansencdb70f32008-02-15 14:37:41 -08001509 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510}
Christoph Hellwig869243a2006-01-09 20:52:03 -08001511EXPORT_SYMBOL(touch_atime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512
1513/**
Christoph Hellwig870f4812006-01-09 20:52:01 -08001514 * file_update_time - update mtime and ctime time
1515 * @file: file accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 *
Christoph Hellwig870f4812006-01-09 20:52:01 -08001517 * Update the mtime and ctime members of an inode and mark the inode
1518 * for writeback. Note that this function is meant exclusively for
1519 * usage in the file write path of filesystems, and filesystems may
1520 * choose to explicitly ignore update via this function with the
Wolfram Sang2eadfc02009-04-02 15:23:37 +02001521 * S_NOCMTIME inode flag, e.g. for network filesystem where these
Christoph Hellwig870f4812006-01-09 20:52:01 -08001522 * timestamps are handled by the server.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 */
1524
Christoph Hellwig870f4812006-01-09 20:52:01 -08001525void file_update_time(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001527 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 struct timespec now;
Andi Kleence06e0b2009-09-18 13:05:48 -07001529 enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530
Andi Kleence06e0b2009-09-18 13:05:48 -07001531 /* First try to exhaust all avenues to not sync */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 if (IS_NOCMTIME(inode))
1533 return;
Dave Hansen20ddee22008-02-15 14:37:43 -08001534
Andi Kleence06e0b2009-09-18 13:05:48 -07001535 now = current_fs_time(inode->i_sb);
1536 if (!timespec_equal(&inode->i_mtime, &now))
1537 sync_it = S_MTIME;
1538
1539 if (!timespec_equal(&inode->i_ctime, &now))
1540 sync_it |= S_CTIME;
1541
1542 if (IS_I_VERSION(inode))
1543 sync_it |= S_VERSION;
1544
1545 if (!sync_it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 return;
1547
Andi Kleence06e0b2009-09-18 13:05:48 -07001548 /* Finally allowed to write? Takes lock. */
1549 if (mnt_want_write_file(file))
1550 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551
Andi Kleence06e0b2009-09-18 13:05:48 -07001552 /* Only change inode inside the lock region */
1553 if (sync_it & S_VERSION)
Jean Noel Cordenner7a224222008-01-28 23:58:27 -05001554 inode_inc_iversion(inode);
Andi Kleence06e0b2009-09-18 13:05:48 -07001555 if (sync_it & S_CTIME)
1556 inode->i_ctime = now;
1557 if (sync_it & S_MTIME)
1558 inode->i_mtime = now;
1559 mark_inode_dirty_sync(inode);
Dave Hansen20ddee22008-02-15 14:37:43 -08001560 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561}
Christoph Hellwig870f4812006-01-09 20:52:01 -08001562EXPORT_SYMBOL(file_update_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
1564int inode_needs_sync(struct inode *inode)
1565{
1566 if (IS_SYNC(inode))
1567 return 1;
1568 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1569 return 1;
1570 return 0;
1571}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572EXPORT_SYMBOL(inode_needs_sync);
1573
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574int inode_wait(void *word)
1575{
1576 schedule();
1577 return 0;
1578}
Stephen Rothwelld44dab82008-11-10 17:06:05 +11001579EXPORT_SYMBOL(inode_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580
1581/*
Miklos Szeredi168a9fd2005-07-12 13:58:10 -07001582 * If we try to find an inode in the inode hash while it is being
1583 * deleted, we have to wait until the filesystem completes its
1584 * deletion before reporting that it isn't found. This function waits
1585 * until the deletion _might_ have completed. Callers are responsible
1586 * to recheck inode state.
1587 *
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001588 * It doesn't matter if I_NEW is not set initially, a call to
Dave Chinner250df6e2011-03-22 22:23:36 +11001589 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1590 * will DTRT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 */
1592static void __wait_on_freeing_inode(struct inode *inode)
1593{
1594 wait_queue_head_t *wq;
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001595 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1596 wq = bit_waitqueue(&inode->i_state, __I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
Dave Chinner250df6e2011-03-22 22:23:36 +11001598 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001599 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 schedule();
1601 finish_wait(wq, &wait.wait);
Dave Chinner67a23c42011-03-22 22:23:42 +11001602 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603}
1604
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605static __initdata unsigned long ihash_entries;
1606static int __init set_ihash_entries(char *str)
1607{
1608 if (!str)
1609 return 0;
1610 ihash_entries = simple_strtoul(str, &str, 0);
1611 return 1;
1612}
1613__setup("ihash_entries=", set_ihash_entries);
1614
1615/*
1616 * Initialize the waitqueues and inode hash table.
1617 */
1618void __init inode_init_early(void)
1619{
1620 int loop;
1621
1622 /* If hashes are distributed across NUMA nodes, defer
1623 * hash allocation until vmalloc space is available.
1624 */
1625 if (hashdist)
1626 return;
1627
1628 inode_hashtable =
1629 alloc_large_system_hash("Inode-cache",
1630 sizeof(struct hlist_head),
1631 ihash_entries,
1632 14,
1633 HASH_EARLY,
1634 &i_hash_shift,
1635 &i_hash_mask,
1636 0);
1637
1638 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1639 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1640}
1641
Denis Cheng74bf17c2007-10-16 23:26:30 -07001642void __init inode_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643{
1644 int loop;
1645
1646 /* inode slab cache */
Paul Jacksonb0196002006-03-24 03:16:09 -08001647 inode_cachep = kmem_cache_create("inode_cache",
1648 sizeof(struct inode),
1649 0,
1650 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1651 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +09001652 init_once);
Rusty Russell8e1f9362007-07-17 04:03:17 -07001653 register_shrinker(&icache_shrinker);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654
1655 /* Hash may have been set up in inode_init_early */
1656 if (!hashdist)
1657 return;
1658
1659 inode_hashtable =
1660 alloc_large_system_hash("Inode-cache",
1661 sizeof(struct hlist_head),
1662 ihash_entries,
1663 14,
1664 0,
1665 &i_hash_shift,
1666 &i_hash_mask,
1667 0);
1668
1669 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1670 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1671}
1672
1673void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1674{
1675 inode->i_mode = mode;
1676 if (S_ISCHR(mode)) {
1677 inode->i_fop = &def_chr_fops;
1678 inode->i_rdev = rdev;
1679 } else if (S_ISBLK(mode)) {
1680 inode->i_fop = &def_blk_fops;
1681 inode->i_rdev = rdev;
1682 } else if (S_ISFIFO(mode))
1683 inode->i_fop = &def_fifo_fops;
1684 else if (S_ISSOCK(mode))
1685 inode->i_fop = &bad_sock_fops;
1686 else
Manish Katiyaraf0d9ae2009-09-18 13:05:43 -07001687 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
1688 " inode %s:%lu\n", mode, inode->i_sb->s_id,
1689 inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690}
1691EXPORT_SYMBOL(init_special_inode);
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001692
1693/**
Ben Hutchingseaae6682011-02-15 12:48:09 +00001694 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001695 * @inode: New inode
1696 * @dir: Directory inode
1697 * @mode: mode of the new inode
1698 */
1699void inode_init_owner(struct inode *inode, const struct inode *dir,
1700 mode_t mode)
1701{
1702 inode->i_uid = current_fsuid();
1703 if (dir && dir->i_mode & S_ISGID) {
1704 inode->i_gid = dir->i_gid;
1705 if (S_ISDIR(mode))
1706 mode |= S_ISGID;
1707 } else
1708 inode->i_gid = current_fsgid();
1709 inode->i_mode = mode;
1710}
1711EXPORT_SYMBOL(inode_init_owner);
Serge E. Hallyne795b712011-03-23 16:43:25 -07001712
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001713/**
1714 * inode_owner_or_capable - check current task permissions to inode
1715 * @inode: inode being checked
1716 *
1717 * Return true if current either has CAP_FOWNER to the inode, or
1718 * owns the file.
Serge E. Hallyne795b712011-03-23 16:43:25 -07001719 */
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001720bool inode_owner_or_capable(const struct inode *inode)
Serge E. Hallyne795b712011-03-23 16:43:25 -07001721{
1722 struct user_namespace *ns = inode_userns(inode);
1723
1724 if (current_user_ns() == ns && current_fsuid() == inode->i_uid)
1725 return true;
1726 if (ns_capable(ns, CAP_FOWNER))
1727 return true;
1728 return false;
1729}
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001730EXPORT_SYMBOL(inode_owner_or_capable);