blob: ee4e66b998f40d170b1822db68875df73d445db6 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * (C) 1997 Linus Torvalds
Christoph Hellwig4b4563d2011-05-27 09:28:01 -04003 * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/fs.h>
6#include <linux/mm.h>
7#include <linux/dcache.h>
8#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/slab.h>
10#include <linux/writeback.h>
11#include <linux/module.h>
12#include <linux/backing-dev.h>
13#include <linux/wait.h>
Nick Piggin88e0fbc2009-09-22 16:43:50 -070014#include <linux/rwsem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/hash.h>
16#include <linux/swap.h>
17#include <linux/security.h>
18#include <linux/pagemap.h>
19#include <linux/cdev.h>
20#include <linux/bootmem.h>
Eric Paris3be25f42009-05-21 17:01:26 -040021#include <linux/fsnotify.h>
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -080022#include <linux/mount.h>
Arjan van de Venefaee192009-01-06 07:20:54 -080023#include <linux/async.h>
Al Virof19d4a82009-06-08 19:50:45 -040024#include <linux/posix_acl.h>
Heiko Carstens9ce6e0b2011-05-22 18:54:21 +020025#include <linux/prefetch.h>
Eric Parisa178d202010-10-25 14:41:59 -040026#include <linux/ima.h>
Serge E. Hallyne795b712011-03-23 16:43:25 -070027#include <linux/cred.h>
Christoph Hellwig4b4563d2011-05-27 09:28:01 -040028#include <linux/buffer_head.h> /* for inode_has_buffers */
Dave Chinnera66979a2011-03-22 22:23:41 +110029#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
31/*
Christoph Hellwig4b4563d2011-05-27 09:28:01 -040032 * Inode locking rules:
Dave Chinner250df6e2011-03-22 22:23:36 +110033 *
34 * inode->i_lock protects:
35 * inode->i_state, inode->i_hash, __iget()
Dave Chinner09cc9fc2011-07-08 14:14:40 +100036 * inode->i_sb->s_inode_lru_lock protects:
Dave Chinner98b745c2011-07-08 14:14:39 +100037 * inode->i_sb->s_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
Christoph Hellwigf758eea2011-04-21 18:19:44 -060040 * bdi->wb.list_lock protects:
Dave Chinnera66979a2011-03-22 22:23:41 +110041 * 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 Chinner09cc9fc2011-07-08 14:14:40 +100049 * inode->i_sb->s_inode_lru_lock
Dave Chinnera66979a2011-03-22 22:23:41 +110050 *
Christoph Hellwigf758eea2011-04-21 18:19:44 -060051 * bdi->wb.list_lock
Dave Chinnera66979a2011-03-22 22:23:41 +110052 * 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
Eric Dumazetfa3536c2006-03-26 01:37:24 -080062static unsigned int i_hash_mask __read_mostly;
63static unsigned int i_hash_shift __read_mostly;
Dave Chinner67a23c42011-03-22 22:23:42 +110064static struct hlist_head *inode_hashtable __read_mostly;
65static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
Dave Chinner55fa6092011-03-22 22:23:40 +110067__cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
68
Linus Torvalds1da177e2005-04-16 15:20:36 -070069/*
Jens Axboe7dcda1c2011-04-05 23:51:48 +020070 * Empty aops. Can be used for the cases where the user does not
71 * define any of the address_space operations.
72 */
73const struct address_space_operations empty_aops = {
74};
75EXPORT_SYMBOL(empty_aops);
76
77/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 * Statistics gathering..
79 */
80struct inodes_stat_t inodes_stat;
81
Nick Piggin3e880fb2011-01-07 17:49:19 +110082static DEFINE_PER_CPU(unsigned int, nr_inodes);
Dave Chinnerfcb94f72011-07-08 14:14:38 +100083static DEFINE_PER_CPU(unsigned int, nr_unused);
Dave Chinnercffbc8a2010-10-23 05:03:02 -040084
Manish Katiyar6b3304b2009-03-31 19:35:54 +053085static struct kmem_cache *inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
Nick Piggin3e880fb2011-01-07 17:49:19 +110087static int get_nr_inodes(void)
Dave Chinnercffbc8a2010-10-23 05:03:02 -040088{
Nick Piggin3e880fb2011-01-07 17:49:19 +110089 int i;
90 int sum = 0;
91 for_each_possible_cpu(i)
92 sum += per_cpu(nr_inodes, i);
93 return sum < 0 ? 0 : sum;
Dave Chinnercffbc8a2010-10-23 05:03:02 -040094}
95
96static inline int get_nr_inodes_unused(void)
97{
Dave Chinnerfcb94f72011-07-08 14:14:38 +100098 int i;
99 int sum = 0;
100 for_each_possible_cpu(i)
101 sum += per_cpu(nr_unused, i);
102 return sum < 0 ? 0 : sum;
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400103}
104
105int get_nr_dirty_inodes(void)
106{
Nick Piggin3e880fb2011-01-07 17:49:19 +1100107 /* not actually dirty inodes, but a wild approximation */
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400108 int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
109 return nr_dirty > 0 ? nr_dirty : 0;
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400110}
111
112/*
113 * Handle nr_inode sysctl
114 */
115#ifdef CONFIG_SYSCTL
116int proc_nr_inodes(ctl_table *table, int write,
117 void __user *buffer, size_t *lenp, loff_t *ppos)
118{
119 inodes_stat.nr_inodes = get_nr_inodes();
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000120 inodes_stat.nr_unused = get_nr_inodes_unused();
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400121 return proc_dointvec(table, write, buffer, lenp, ppos);
122}
123#endif
124
David Chinner2cb15992008-10-30 17:32:23 +1100125/**
126 * inode_init_always - perform inode structure intialisation
Randy Dunlap0bc02f32009-01-06 14:41:13 -0800127 * @sb: superblock inode belongs to
128 * @inode: inode to initialise
David Chinner2cb15992008-10-30 17:32:23 +1100129 *
130 * These are initializations that need to be done on every inode
131 * allocation as the fields are not initialised by slab allocation.
132 */
Christoph Hellwig54e34622009-08-07 14:38:25 -0300133int inode_init_always(struct super_block *sb, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134{
Alexey Dobriyan6e1d5dc2009-09-21 17:01:11 -0700135 static const struct inode_operations empty_iops;
Arjan van de Ven99ac48f2006-03-28 01:56:41 -0800136 static const struct file_operations empty_fops;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530137 struct address_space *const mapping = &inode->i_data;
David Chinner2cb15992008-10-30 17:32:23 +1100138
139 inode->i_sb = sb;
140 inode->i_blkbits = sb->s_blocksize_bits;
141 inode->i_flags = 0;
142 atomic_set(&inode->i_count, 1);
143 inode->i_op = &empty_iops;
144 inode->i_fop = &empty_fops;
Miklos Szeredia78ef702011-10-28 14:13:30 +0200145 inode->__i_nlink = 1;
Linus Torvalds3ddcd052011-08-06 22:45:50 -0700146 inode->i_opflags = 0;
Al Viro56ff5ef2008-12-09 09:34:39 -0500147 inode->i_uid = 0;
148 inode->i_gid = 0;
David Chinner2cb15992008-10-30 17:32:23 +1100149 atomic_set(&inode->i_writecount, 0);
150 inode->i_size = 0;
151 inode->i_blocks = 0;
152 inode->i_bytes = 0;
153 inode->i_generation = 0;
154#ifdef CONFIG_QUOTA
155 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
156#endif
157 inode->i_pipe = NULL;
158 inode->i_bdev = NULL;
159 inode->i_cdev = NULL;
160 inode->i_rdev = 0;
161 inode->dirtied_when = 0;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500162
163 if (security_inode_alloc(inode))
Christoph Hellwig54e34622009-08-07 14:38:25 -0300164 goto out;
David Chinner2cb15992008-10-30 17:32:23 +1100165 spin_lock_init(&inode->i_lock);
166 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
167
168 mutex_init(&inode->i_mutex);
169 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
170
Christoph Hellwigbd5fe6c2011-06-24 14:29:43 -0400171 atomic_set(&inode->i_dio_count, 0);
David Chinner2cb15992008-10-30 17:32:23 +1100172
173 mapping->a_ops = &empty_aops;
174 mapping->host = inode;
175 mapping->flags = 0;
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800176 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
David Chinner2cb15992008-10-30 17:32:23 +1100177 mapping->assoc_mapping = NULL;
178 mapping->backing_dev_info = &default_backing_dev_info;
179 mapping->writeback_index = 0;
180
181 /*
182 * If the block_device provides a backing_dev_info for client
183 * inodes then use that. Otherwise the inode share the bdev's
184 * backing_dev_info.
185 */
186 if (sb->s_bdev) {
187 struct backing_dev_info *bdi;
188
Jens Axboe2c96ce92009-09-15 09:43:56 +0200189 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
David Chinner2cb15992008-10-30 17:32:23 +1100190 mapping->backing_dev_info = bdi;
191 }
192 inode->i_private = NULL;
193 inode->i_mapping = mapping;
Al Virof19d4a82009-06-08 19:50:45 -0400194#ifdef CONFIG_FS_POSIX_ACL
195 inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
196#endif
David Chinner2cb15992008-10-30 17:32:23 +1100197
Eric Paris3be25f42009-05-21 17:01:26 -0400198#ifdef CONFIG_FSNOTIFY
199 inode->i_fsnotify_mask = 0;
200#endif
201
Nick Piggin3e880fb2011-01-07 17:49:19 +1100202 this_cpu_inc(nr_inodes);
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400203
Christoph Hellwig54e34622009-08-07 14:38:25 -0300204 return 0;
Christoph Hellwig54e34622009-08-07 14:38:25 -0300205out:
206 return -ENOMEM;
David Chinner2cb15992008-10-30 17:32:23 +1100207}
208EXPORT_SYMBOL(inode_init_always);
209
210static struct inode *alloc_inode(struct super_block *sb)
211{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 struct inode *inode;
213
214 if (sb->s_op->alloc_inode)
215 inode = sb->s_op->alloc_inode(sb);
216 else
David Chinner2cb15992008-10-30 17:32:23 +1100217 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
Christoph Hellwig54e34622009-08-07 14:38:25 -0300219 if (!inode)
220 return NULL;
221
222 if (unlikely(inode_init_always(sb, 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 return NULL;
228 }
229
230 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231}
232
Nick Pigginff0c7d12011-01-07 17:49:50 +1100233void free_inode_nonrcu(struct inode *inode)
234{
235 kmem_cache_free(inode_cachep, inode);
236}
237EXPORT_SYMBOL(free_inode_nonrcu);
238
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300239void __destroy_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240{
Eric Sesterhennb7542f82006-04-02 13:38:18 +0200241 BUG_ON(inode_has_buffers(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 security_inode_free(inode);
Eric Paris3be25f42009-05-21 17:01:26 -0400243 fsnotify_inode_delete(inode);
Al Virof19d4a82009-06-08 19:50:45 -0400244#ifdef CONFIG_FS_POSIX_ACL
245 if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED)
246 posix_acl_release(inode->i_acl);
247 if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
248 posix_acl_release(inode->i_default_acl);
249#endif
Nick Piggin3e880fb2011-01-07 17:49:19 +1100250 this_cpu_dec(nr_inodes);
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300251}
252EXPORT_SYMBOL(__destroy_inode);
253
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100254static void i_callback(struct rcu_head *head)
255{
256 struct inode *inode = container_of(head, struct inode, i_rcu);
257 INIT_LIST_HEAD(&inode->i_dentry);
258 kmem_cache_free(inode_cachep, inode);
259}
260
Christoph Hellwig56b0dac2010-10-06 10:48:55 +0200261static void destroy_inode(struct inode *inode)
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300262{
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100263 BUG_ON(!list_empty(&inode->i_lru));
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300264 __destroy_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 if (inode->i_sb->s_op->destroy_inode)
266 inode->i_sb->s_op->destroy_inode(inode);
267 else
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100268 call_rcu(&inode->i_rcu, i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100271void address_space_init_once(struct address_space *mapping)
272{
273 memset(mapping, 0, sizeof(*mapping));
274 INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
275 spin_lock_init(&mapping->tree_lock);
Peter Zijlstra3d48ae42011-05-24 17:12:06 -0700276 mutex_init(&mapping->i_mmap_mutex);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100277 INIT_LIST_HEAD(&mapping->private_list);
278 spin_lock_init(&mapping->private_lock);
279 INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
280 INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100281}
282EXPORT_SYMBOL(address_space_init_once);
283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284/*
285 * These are initializations that only need to be done
286 * once, because the fields are idempotent across use
287 * of the inode, so let the slab aware of that.
288 */
289void inode_init_once(struct inode *inode)
290{
291 memset(inode, 0, sizeof(*inode));
292 INIT_HLIST_NODE(&inode->i_hash);
293 INIT_LIST_HEAD(&inode->i_dentry);
294 INIT_LIST_HEAD(&inode->i_devices);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100295 INIT_LIST_HEAD(&inode->i_wb_list);
296 INIT_LIST_HEAD(&inode->i_lru);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100297 address_space_init_once(&inode->i_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 i_size_ordered_init(inode);
Eric Paris3be25f42009-05-21 17:01:26 -0400299#ifdef CONFIG_FSNOTIFY
Eric Parise61ce862009-12-17 21:24:24 -0500300 INIT_HLIST_HEAD(&inode->i_fsnotify_marks);
Eric Paris3be25f42009-05-21 17:01:26 -0400301#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303EXPORT_SYMBOL(inode_init_once);
304
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700305static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530307 struct inode *inode = (struct inode *) foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Christoph Lametera35afb82007-05-16 22:10:57 -0700309 inode_init_once(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310}
311
312/*
Dave Chinner250df6e2011-03-22 22:23:36 +1100313 * inode->i_lock must be held
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530315void __iget(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316{
Nick Piggin9e38d862010-10-23 06:55:17 -0400317 atomic_inc(&inode->i_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318}
319
Al Viro7de9c6ee2010-10-23 11:11:40 -0400320/*
321 * get additional reference to inode; caller must already hold one.
322 */
323void ihold(struct inode *inode)
324{
325 WARN_ON(atomic_inc_return(&inode->i_count) < 2);
326}
327EXPORT_SYMBOL(ihold);
328
Nick Piggin9e38d862010-10-23 06:55:17 -0400329static void inode_lru_list_add(struct inode *inode)
330{
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000331 spin_lock(&inode->i_sb->s_inode_lru_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100332 if (list_empty(&inode->i_lru)) {
Dave Chinner98b745c2011-07-08 14:14:39 +1000333 list_add(&inode->i_lru, &inode->i_sb->s_inode_lru);
334 inode->i_sb->s_nr_inodes_unused++;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000335 this_cpu_inc(nr_unused);
Nick Piggin9e38d862010-10-23 06:55:17 -0400336 }
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000337 spin_unlock(&inode->i_sb->s_inode_lru_lock);
Nick Piggin9e38d862010-10-23 06:55:17 -0400338}
339
340static void inode_lru_list_del(struct inode *inode)
341{
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000342 spin_lock(&inode->i_sb->s_inode_lru_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100343 if (!list_empty(&inode->i_lru)) {
344 list_del_init(&inode->i_lru);
Dave Chinner98b745c2011-07-08 14:14:39 +1000345 inode->i_sb->s_nr_inodes_unused--;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000346 this_cpu_dec(nr_unused);
Nick Piggin9e38d862010-10-23 06:55:17 -0400347 }
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000348 spin_unlock(&inode->i_sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349}
350
Christoph Hellwig646ec462010-10-23 07:15:32 -0400351/**
352 * inode_sb_list_add - add inode to the superblock list of inodes
353 * @inode: inode to add
354 */
355void inode_sb_list_add(struct inode *inode)
356{
Dave Chinner55fa6092011-03-22 22:23:40 +1100357 spin_lock(&inode_sb_list_lock);
358 list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
359 spin_unlock(&inode_sb_list_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400360}
361EXPORT_SYMBOL_GPL(inode_sb_list_add);
362
Dave Chinner55fa6092011-03-22 22:23:40 +1100363static inline void inode_sb_list_del(struct inode *inode)
Christoph Hellwig646ec462010-10-23 07:15:32 -0400364{
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200365 if (!list_empty(&inode->i_sb_list)) {
366 spin_lock(&inode_sb_list_lock);
367 list_del_init(&inode->i_sb_list);
368 spin_unlock(&inode_sb_list_lock);
369 }
Christoph Hellwig646ec462010-10-23 07:15:32 -0400370}
371
Dave Chinner4c51acb2010-10-23 06:58:09 -0400372static unsigned long hash(struct super_block *sb, unsigned long hashval)
373{
374 unsigned long tmp;
375
376 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
377 L1_CACHE_BYTES;
Christoph Hellwig4b4563d2011-05-27 09:28:01 -0400378 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift);
379 return tmp & i_hash_mask;
Dave Chinner4c51acb2010-10-23 06:58:09 -0400380}
381
382/**
383 * __insert_inode_hash - hash an inode
384 * @inode: unhashed inode
385 * @hashval: unsigned long value used to locate this object in the
386 * inode_hashtable.
387 *
388 * Add an inode to the inode hash for this superblock.
389 */
390void __insert_inode_hash(struct inode *inode, unsigned long hashval)
391{
Christoph Hellwig646ec462010-10-23 07:15:32 -0400392 struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
393
Dave Chinner67a23c42011-03-22 22:23:42 +1100394 spin_lock(&inode_hash_lock);
Dave Chinner250df6e2011-03-22 22:23:36 +1100395 spin_lock(&inode->i_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400396 hlist_add_head(&inode->i_hash, b);
Dave Chinner250df6e2011-03-22 22:23:36 +1100397 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100398 spin_unlock(&inode_hash_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400399}
400EXPORT_SYMBOL(__insert_inode_hash);
401
402/**
Eric Dumazetf2ee7ab2011-07-28 06:41:09 +0200403 * __remove_inode_hash - remove an inode from the hash
Dave Chinner4c51acb2010-10-23 06:58:09 -0400404 * @inode: inode to unhash
405 *
406 * Remove an inode from the superblock.
407 */
Eric Dumazetf2ee7ab2011-07-28 06:41:09 +0200408void __remove_inode_hash(struct inode *inode)
Dave Chinner4c51acb2010-10-23 06:58:09 -0400409{
Dave Chinner67a23c42011-03-22 22:23:42 +1100410 spin_lock(&inode_hash_lock);
Dave Chinner250df6e2011-03-22 22:23:36 +1100411 spin_lock(&inode->i_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400412 hlist_del_init(&inode->i_hash);
Dave Chinner250df6e2011-03-22 22:23:36 +1100413 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100414 spin_unlock(&inode_hash_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400415}
Eric Dumazetf2ee7ab2011-07-28 06:41:09 +0200416EXPORT_SYMBOL(__remove_inode_hash);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400417
Al Virob0683aa2010-06-04 20:55:25 -0400418void end_writeback(struct inode *inode)
419{
420 might_sleep();
Jan Kara08142572011-06-27 16:18:10 -0700421 /*
422 * We have to cycle tree_lock here because reclaim can be still in the
423 * process of removing the last page (in __delete_from_page_cache())
424 * and we must not free mapping under it.
425 */
426 spin_lock_irq(&inode->i_data.tree_lock);
Al Virob0683aa2010-06-04 20:55:25 -0400427 BUG_ON(inode->i_data.nrpages);
Jan Kara08142572011-06-27 16:18:10 -0700428 spin_unlock_irq(&inode->i_data.tree_lock);
Al Virob0683aa2010-06-04 20:55:25 -0400429 BUG_ON(!list_empty(&inode->i_data.private_list));
430 BUG_ON(!(inode->i_state & I_FREEING));
431 BUG_ON(inode->i_state & I_CLEAR);
432 inode_sync_wait(inode);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100433 /* don't need i_lock here, no concurrent mods to i_state */
Al Virob0683aa2010-06-04 20:55:25 -0400434 inode->i_state = I_FREEING | I_CLEAR;
435}
436EXPORT_SYMBOL(end_writeback);
437
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100438/*
439 * Free the inode passed in, removing it from the lists it is still connected
440 * to. We remove any pages still attached to the inode and wait for any IO that
441 * is still in progress before finally destroying the inode.
442 *
443 * An inode must already be marked I_FREEING so that we avoid the inode being
444 * moved back onto lists if we race with other code that manipulates the lists
445 * (e.g. writeback_single_inode). The caller is responsible for setting this.
446 *
447 * An inode must already be removed from the LRU list before being evicted from
448 * the cache. This should occur atomically with setting the I_FREEING state
449 * flag, so no inodes here should ever be on the LRU when being evicted.
450 */
Al Viro644da592010-06-07 13:21:05 -0400451static void evict(struct inode *inode)
Al Virob4272d42010-06-04 19:33:20 -0400452{
453 const struct super_operations *op = inode->i_sb->s_op;
454
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100455 BUG_ON(!(inode->i_state & I_FREEING));
456 BUG_ON(!list_empty(&inode->i_lru));
457
Eric Dumazetb12362b2011-07-28 06:11:47 +0200458 if (!list_empty(&inode->i_wb_list))
459 inode_wb_list_del(inode);
460
Dave Chinner55fa6092011-03-22 22:23:40 +1100461 inode_sb_list_del(inode);
462
Al Virobe7ce412010-06-04 19:40:39 -0400463 if (op->evict_inode) {
464 op->evict_inode(inode);
Al Virob4272d42010-06-04 19:33:20 -0400465 } else {
466 if (inode->i_data.nrpages)
467 truncate_inode_pages(&inode->i_data, 0);
Al Viro30140832010-06-07 13:23:20 -0400468 end_writeback(inode);
Al Virob4272d42010-06-04 19:33:20 -0400469 }
Al Viro661074e2010-06-04 20:19:55 -0400470 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
471 bd_forget(inode);
472 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
473 cd_forget(inode);
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100474
475 remove_inode_hash(inode);
476
477 spin_lock(&inode->i_lock);
478 wake_up_bit(&inode->i_state, __I_NEW);
479 BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
480 spin_unlock(&inode->i_lock);
481
482 destroy_inode(inode);
Al Virob4272d42010-06-04 19:33:20 -0400483}
484
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485/*
486 * dispose_list - dispose of the contents of a local list
487 * @head: the head of the list to free
488 *
489 * Dispose-list gets a local list with local inodes in it, so it doesn't
490 * need to worry about list corruption and SMP locks.
491 */
492static void dispose_list(struct list_head *head)
493{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 while (!list_empty(head)) {
495 struct inode *inode;
496
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100497 inode = list_first_entry(head, struct inode, i_lru);
498 list_del_init(&inode->i_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Al Viro644da592010-06-07 13:21:05 -0400500 evict(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502}
503
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504/**
Al Viro63997e92010-10-25 20:49:35 -0400505 * evict_inodes - evict all evictable inodes for a superblock
506 * @sb: superblock to operate on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 *
Al Viro63997e92010-10-25 20:49:35 -0400508 * Make sure that no inodes with zero refcount are retained. This is
509 * called by superblock shutdown after having MS_ACTIVE flag removed,
510 * so any inode reaching zero refcount during or after that call will
511 * be immediately evicted.
512 */
513void evict_inodes(struct super_block *sb)
514{
515 struct inode *inode, *next;
516 LIST_HEAD(dispose);
517
Dave Chinner55fa6092011-03-22 22:23:40 +1100518 spin_lock(&inode_sb_list_lock);
Al Viro63997e92010-10-25 20:49:35 -0400519 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
520 if (atomic_read(&inode->i_count))
521 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100522
523 spin_lock(&inode->i_lock);
524 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
525 spin_unlock(&inode->i_lock);
Al Viro63997e92010-10-25 20:49:35 -0400526 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100527 }
Al Viro63997e92010-10-25 20:49:35 -0400528
529 inode->i_state |= I_FREEING;
Dave Chinner02afc412011-03-22 22:23:38 +1100530 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100531 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100532 list_add(&inode->i_lru, &dispose);
Al Viro63997e92010-10-25 20:49:35 -0400533 }
Dave Chinner55fa6092011-03-22 22:23:40 +1100534 spin_unlock(&inode_sb_list_lock);
Al Viro63997e92010-10-25 20:49:35 -0400535
536 dispose_list(&dispose);
Al Viro63997e92010-10-25 20:49:35 -0400537}
538
539/**
Christoph Hellwiga0318782010-10-24 19:40:33 +0200540 * invalidate_inodes - attempt to free all inodes on a superblock
541 * @sb: superblock to operate on
NeilBrown93b270f2011-02-24 17:25:47 +1100542 * @kill_dirty: flag to guide handling of dirty inodes
Christoph Hellwiga0318782010-10-24 19:40:33 +0200543 *
544 * Attempts to free all inodes for a given superblock. If there were any
545 * busy inodes return a non-zero value, else zero.
NeilBrown93b270f2011-02-24 17:25:47 +1100546 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
547 * them as busy.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 */
NeilBrown93b270f2011-02-24 17:25:47 +1100549int invalidate_inodes(struct super_block *sb, bool kill_dirty)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400551 int busy = 0;
Christoph Hellwiga0318782010-10-24 19:40:33 +0200552 struct inode *inode, *next;
553 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
Dave Chinner55fa6092011-03-22 22:23:40 +1100555 spin_lock(&inode_sb_list_lock);
Christoph Hellwiga0318782010-10-24 19:40:33 +0200556 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100557 spin_lock(&inode->i_lock);
558 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
559 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100561 }
NeilBrown93b270f2011-02-24 17:25:47 +1100562 if (inode->i_state & I_DIRTY && !kill_dirty) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100563 spin_unlock(&inode->i_lock);
NeilBrown93b270f2011-02-24 17:25:47 +1100564 busy = 1;
565 continue;
566 }
Christoph Hellwig99a38912010-10-23 19:07:20 +0200567 if (atomic_read(&inode->i_count)) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100568 spin_unlock(&inode->i_lock);
Christoph Hellwig99a38912010-10-23 19:07:20 +0200569 busy = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 continue;
571 }
Christoph Hellwig99a38912010-10-23 19:07:20 +0200572
Christoph Hellwig99a38912010-10-23 19:07:20 +0200573 inode->i_state |= I_FREEING;
Dave Chinner02afc412011-03-22 22:23:38 +1100574 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100575 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100576 list_add(&inode->i_lru, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 }
Dave Chinner55fa6092011-03-22 22:23:40 +1100578 spin_unlock(&inode_sb_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579
Christoph Hellwiga0318782010-10-24 19:40:33 +0200580 dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
582 return busy;
583}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
585static int can_unuse(struct inode *inode)
586{
Nick Piggin9e38d862010-10-23 06:55:17 -0400587 if (inode->i_state & ~I_REFERENCED)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 return 0;
589 if (inode_has_buffers(inode))
590 return 0;
591 if (atomic_read(&inode->i_count))
592 return 0;
593 if (inode->i_data.nrpages)
594 return 0;
595 return 1;
596}
597
598/*
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000599 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
600 * This is called from the superblock shrinker function with a number of inodes
601 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
602 * then are freed outside inode_lock by dispose_list().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 *
604 * Any inodes which are pinned purely because of attached pagecache have their
Nick Piggin9e38d862010-10-23 06:55:17 -0400605 * pagecache removed. If the inode has metadata buffers attached to
606 * mapping->private_list then try to remove them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 *
Nick Piggin9e38d862010-10-23 06:55:17 -0400608 * If the inode has the I_REFERENCED flag set, then it means that it has been
609 * used recently - the flag is set in iput_final(). When we encounter such an
610 * inode, clear the flag and move it to the back of the LRU so it gets another
611 * pass through the LRU before it gets reclaimed. This is necessary because of
612 * the fact we are doing lazy LRU updates to minimise lock contention so the
613 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
614 * with this flag set because they are the inodes that are out of order.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 */
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000616void prune_icache_sb(struct super_block *sb, int nr_to_scan)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617{
618 LIST_HEAD(freeable);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 int nr_scanned;
620 unsigned long reap = 0;
621
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000622 spin_lock(&sb->s_inode_lru_lock);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000623 for (nr_scanned = nr_to_scan; nr_scanned >= 0; nr_scanned--) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 struct inode *inode;
625
Dave Chinner98b745c2011-07-08 14:14:39 +1000626 if (list_empty(&sb->s_inode_lru))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 break;
628
Dave Chinner98b745c2011-07-08 14:14:39 +1000629 inode = list_entry(sb->s_inode_lru.prev, struct inode, i_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630
Nick Piggin9e38d862010-10-23 06:55:17 -0400631 /*
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000632 * we are inverting the sb->s_inode_lru_lock/inode->i_lock here,
Dave Chinner02afc412011-03-22 22:23:38 +1100633 * so use a trylock. If we fail to get the lock, just move the
634 * inode to the back of the list so we don't spin on it.
635 */
636 if (!spin_trylock(&inode->i_lock)) {
Christoph Hellwig62a3dde2011-10-28 10:03:41 +0200637 list_move_tail(&inode->i_lru, &sb->s_inode_lru);
Dave Chinner02afc412011-03-22 22:23:38 +1100638 continue;
639 }
640
641 /*
Nick Piggin9e38d862010-10-23 06:55:17 -0400642 * Referenced or dirty inodes are still in use. Give them
643 * another pass through the LRU as we canot reclaim them now.
644 */
645 if (atomic_read(&inode->i_count) ||
646 (inode->i_state & ~I_REFERENCED)) {
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100647 list_del_init(&inode->i_lru);
Dave Chinnerf283c862011-03-22 22:23:39 +1100648 spin_unlock(&inode->i_lock);
Dave Chinner98b745c2011-07-08 14:14:39 +1000649 sb->s_nr_inodes_unused--;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000650 this_cpu_dec(nr_unused);
Nick Piggin9e38d862010-10-23 06:55:17 -0400651 continue;
652 }
653
654 /* recently referenced inodes get one more pass */
655 if (inode->i_state & I_REFERENCED) {
Nick Piggin9e38d862010-10-23 06:55:17 -0400656 inode->i_state &= ~I_REFERENCED;
Dave Chinner98b745c2011-07-08 14:14:39 +1000657 list_move(&inode->i_lru, &sb->s_inode_lru);
Dave Chinnerf283c862011-03-22 22:23:39 +1100658 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 continue;
660 }
661 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
662 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100663 spin_unlock(&inode->i_lock);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000664 spin_unlock(&sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 if (remove_inode_buffers(inode))
Andrew Mortonfc0ecff2007-02-10 01:45:39 -0800666 reap += invalidate_mapping_pages(&inode->i_data,
667 0, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 iput(inode);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000669 spin_lock(&sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
Dave Chinner98b745c2011-07-08 14:14:39 +1000671 if (inode != list_entry(sb->s_inode_lru.next,
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100672 struct inode, i_lru))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 continue; /* wrong inode or list_empty */
Dave Chinner02afc412011-03-22 22:23:38 +1100674 /* avoid lock inversions with trylock */
675 if (!spin_trylock(&inode->i_lock))
676 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100677 if (!can_unuse(inode)) {
678 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100680 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 }
Nick Piggin7ef0d732009-03-12 14:31:38 -0700682 WARN_ON(inode->i_state & I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 inode->i_state |= I_FREEING;
Dave Chinner250df6e2011-03-22 22:23:36 +1100684 spin_unlock(&inode->i_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100685
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100686 list_move(&inode->i_lru, &freeable);
Dave Chinner98b745c2011-07-08 14:14:39 +1000687 sb->s_nr_inodes_unused--;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000688 this_cpu_dec(nr_unused);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 }
Christoph Lameterf8891e52006-06-30 01:55:45 -0700690 if (current_is_kswapd())
691 __count_vm_events(KSWAPD_INODESTEAL, reap);
692 else
693 __count_vm_events(PGINODESTEAL, reap);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000694 spin_unlock(&sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
696 dispose_list(&freeable);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697}
698
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699static void __wait_on_freeing_inode(struct inode *inode);
700/*
701 * Called with the inode lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530703static struct inode *find_inode(struct super_block *sb,
704 struct hlist_head *head,
705 int (*test)(struct inode *, void *),
706 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707{
708 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530709 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710
711repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700712 hlist_for_each_entry(inode, node, head, i_hash) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100713 spin_lock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100714 if (inode->i_sb != sb) {
715 spin_unlock(&inode->i_lock);
716 continue;
717 }
718 if (!test(inode, data)) {
719 spin_unlock(&inode->i_lock);
720 continue;
721 }
Al Viroa4ffdde2010-06-02 17:38:30 -0400722 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 __wait_on_freeing_inode(inode);
724 goto repeat;
725 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400726 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100727 spin_unlock(&inode->i_lock);
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400728 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400730 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731}
732
733/*
734 * find_inode_fast is the fast path version of find_inode, see the comment at
735 * iget_locked for details.
736 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530737static struct inode *find_inode_fast(struct super_block *sb,
738 struct hlist_head *head, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
740 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530741 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742
743repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700744 hlist_for_each_entry(inode, node, head, i_hash) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100745 spin_lock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100746 if (inode->i_ino != ino) {
747 spin_unlock(&inode->i_lock);
748 continue;
749 }
750 if (inode->i_sb != sb) {
751 spin_unlock(&inode->i_lock);
752 continue;
753 }
Al Viroa4ffdde2010-06-02 17:38:30 -0400754 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 __wait_on_freeing_inode(inode);
756 goto repeat;
757 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400758 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100759 spin_unlock(&inode->i_lock);
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400760 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400762 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763}
764
Eric Dumazetf991bd22010-10-23 11:18:01 -0400765/*
766 * Each cpu owns a range of LAST_INO_BATCH numbers.
767 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
768 * to renew the exhausted range.
David Chinner8290c352008-10-30 17:35:24 +1100769 *
Eric Dumazetf991bd22010-10-23 11:18:01 -0400770 * This does not significantly increase overflow rate because every CPU can
771 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
772 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
773 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
774 * overflow rate by 2x, which does not seem too significant.
775 *
776 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
777 * error if st_ino won't fit in target struct field. Use 32bit counter
778 * here to attempt to avoid that.
David Chinner8290c352008-10-30 17:35:24 +1100779 */
Eric Dumazetf991bd22010-10-23 11:18:01 -0400780#define LAST_INO_BATCH 1024
781static DEFINE_PER_CPU(unsigned int, last_ino);
David Chinner8290c352008-10-30 17:35:24 +1100782
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400783unsigned int get_next_ino(void)
Eric Dumazetf991bd22010-10-23 11:18:01 -0400784{
785 unsigned int *p = &get_cpu_var(last_ino);
786 unsigned int res = *p;
787
788#ifdef CONFIG_SMP
789 if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
790 static atomic_t shared_last_ino;
791 int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
792
793 res = next - LAST_INO_BATCH;
794 }
795#endif
796
797 *p = ++res;
798 put_cpu_var(last_ino);
799 return res;
David Chinner8290c352008-10-30 17:35:24 +1100800}
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400801EXPORT_SYMBOL(get_next_ino);
David Chinner8290c352008-10-30 17:35:24 +1100802
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803/**
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200804 * new_inode_pseudo - obtain an inode
805 * @sb: superblock
806 *
807 * Allocates a new inode for given superblock.
808 * Inode wont be chained in superblock s_inodes list
809 * This means :
810 * - fs can't be unmount
811 * - quotas, fsnotify, writeback can't work
812 */
813struct inode *new_inode_pseudo(struct super_block *sb)
814{
815 struct inode *inode = alloc_inode(sb);
816
817 if (inode) {
818 spin_lock(&inode->i_lock);
819 inode->i_state = 0;
820 spin_unlock(&inode->i_lock);
821 INIT_LIST_HEAD(&inode->i_sb_list);
822 }
823 return inode;
824}
825
826/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 * new_inode - obtain an inode
828 * @sb: superblock
829 *
Mel Gorman769848c2007-07-17 04:03:05 -0700830 * Allocates a new inode for given superblock. The default gfp_mask
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800831 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
Mel Gorman769848c2007-07-17 04:03:05 -0700832 * If HIGHMEM pages are unsuitable or it is known that pages allocated
833 * for the page cache are not reclaimable or migratable,
834 * mapping_set_gfp_mask() must be called with suitable flags on the
835 * newly created inode's mapping
836 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 */
838struct inode *new_inode(struct super_block *sb)
839{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530840 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
Dave Chinner55fa6092011-03-22 22:23:40 +1100842 spin_lock_prefetch(&inode_sb_list_lock);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530843
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200844 inode = new_inode_pseudo(sb);
845 if (inode)
Dave Chinner55fa6092011-03-22 22:23:40 +1100846 inode_sb_list_add(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 return inode;
848}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849EXPORT_SYMBOL(new_inode);
850
Peter Zijlstra14358e62007-10-14 01:38:33 +0200851#ifdef CONFIG_DEBUG_LOCK_ALLOC
Josh Boyere096d0c2011-08-25 07:48:12 -0400852void lockdep_annotate_inode_mutex_key(struct inode *inode)
853{
Namhyung Kima3314a02010-10-11 22:38:00 +0900854 if (S_ISDIR(inode->i_mode)) {
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200855 struct file_system_type *type = inode->i_sb->s_type;
856
Jan Kara9a7aa122009-06-04 15:26:49 +0200857 /* Set new key only if filesystem hasn't already changed it */
858 if (!lockdep_match_class(&inode->i_mutex,
859 &type->i_mutex_key)) {
860 /*
861 * ensure nobody is actually holding i_mutex
862 */
863 mutex_destroy(&inode->i_mutex);
864 mutex_init(&inode->i_mutex);
865 lockdep_set_class(&inode->i_mutex,
866 &type->i_mutex_dir_key);
867 }
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200868 }
Josh Boyere096d0c2011-08-25 07:48:12 -0400869}
870EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key);
Peter Zijlstra14358e62007-10-14 01:38:33 +0200871#endif
Josh Boyere096d0c2011-08-25 07:48:12 -0400872
873/**
874 * unlock_new_inode - clear the I_NEW state and wake up any waiters
875 * @inode: new inode to unlock
876 *
877 * Called when the inode is fully initialised to clear the new state of the
878 * inode and wake up anyone waiting for the inode to finish initialisation.
879 */
880void unlock_new_inode(struct inode *inode)
881{
882 lockdep_annotate_inode_mutex_key(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100883 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +0100884 WARN_ON(!(inode->i_state & I_NEW));
885 inode->i_state &= ~I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +1100886 wake_up_bit(&inode->i_state, __I_NEW);
887 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889EXPORT_SYMBOL(unlock_new_inode);
890
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400891/**
892 * iget5_locked - obtain an inode from a mounted file system
893 * @sb: super block of file system
894 * @hashval: hash value (usually inode number) to get
895 * @test: callback used for comparisons between inodes
896 * @set: callback used to initialize a new struct inode
897 * @data: opaque data pointer to pass to @test and @set
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400899 * Search for the inode specified by @hashval and @data in the inode cache,
900 * and if present it is return it with an increased reference count. This is
901 * a generalized version of iget_locked() for file systems where the inode
902 * number is not sufficient for unique identification of an inode.
903 *
904 * If the inode is not in cache, allocate a new inode and return it locked,
905 * hashed, and with the I_NEW flag set. The file system gets to fill it in
906 * before unlocking it via unlock_new_inode().
907 *
908 * Note both @test and @set are called with the inode_hash_lock held, so can't
909 * sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 */
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400911struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
912 int (*test)(struct inode *, void *),
913 int (*set)(struct inode *, void *), void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400915 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530916 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400918 spin_lock(&inode_hash_lock);
919 inode = find_inode(sb, head, test, data);
920 spin_unlock(&inode_hash_lock);
921
922 if (inode) {
923 wait_on_inode(inode);
924 return inode;
925 }
926
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 inode = alloc_inode(sb);
928 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530929 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
Dave Chinner67a23c42011-03-22 22:23:42 +1100931 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 /* We released the lock, so.. */
933 old = find_inode(sb, head, test, data);
934 if (!old) {
935 if (set(inode, data))
936 goto set_failed;
937
Dave Chinner250df6e2011-03-22 22:23:36 +1100938 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +0100939 inode->i_state = I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +1100940 hlist_add_head(&inode->i_hash, head);
941 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +1100942 inode_sb_list_add(inode);
Dave Chinner67a23c42011-03-22 22:23:42 +1100943 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944
945 /* Return the locked inode with I_NEW set, the
946 * caller is responsible for filling in the contents
947 */
948 return inode;
949 }
950
951 /*
952 * Uhhuh, somebody else created the same inode under
953 * us. Use the old inode instead of the one we just
954 * allocated.
955 */
Dave Chinner67a23c42011-03-22 22:23:42 +1100956 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 destroy_inode(inode);
958 inode = old;
959 wait_on_inode(inode);
960 }
961 return inode;
962
963set_failed:
Dave Chinner67a23c42011-03-22 22:23:42 +1100964 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 destroy_inode(inode);
966 return NULL;
967}
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400968EXPORT_SYMBOL(iget5_locked);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400970/**
971 * iget_locked - obtain an inode from a mounted file system
972 * @sb: super block of file system
973 * @ino: inode number to get
974 *
975 * Search for the inode specified by @ino in the inode cache and if present
976 * return it with an increased reference count. This is for file systems
977 * where the inode number is sufficient for unique identification of an inode.
978 *
979 * If the inode is not in cache, allocate a new inode and return it locked,
980 * hashed, and with the I_NEW flag set. The file system gets to fill it in
981 * before unlocking it via unlock_new_inode().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 */
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400983struct inode *iget_locked(struct super_block *sb, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400985 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530986 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400988 spin_lock(&inode_hash_lock);
989 inode = find_inode_fast(sb, head, ino);
990 spin_unlock(&inode_hash_lock);
991 if (inode) {
992 wait_on_inode(inode);
993 return inode;
994 }
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 inode = alloc_inode(sb);
997 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530998 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
Dave Chinner67a23c42011-03-22 22:23:42 +11001000 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 /* We released the lock, so.. */
1002 old = find_inode_fast(sb, head, ino);
1003 if (!old) {
1004 inode->i_ino = ino;
Dave Chinner250df6e2011-03-22 22:23:36 +11001005 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001006 inode->i_state = I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +11001007 hlist_add_head(&inode->i_hash, head);
1008 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +11001009 inode_sb_list_add(inode);
Dave Chinner67a23c42011-03-22 22:23:42 +11001010 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
1012 /* Return the locked inode with I_NEW set, the
1013 * caller is responsible for filling in the contents
1014 */
1015 return inode;
1016 }
1017
1018 /*
1019 * Uhhuh, somebody else created the same inode under
1020 * us. Use the old inode instead of the one we just
1021 * allocated.
1022 */
Dave Chinner67a23c42011-03-22 22:23:42 +11001023 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 destroy_inode(inode);
1025 inode = old;
1026 wait_on_inode(inode);
1027 }
1028 return inode;
1029}
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001030EXPORT_SYMBOL(iget_locked);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001032/*
1033 * search the inode cache for a matching inode number.
1034 * If we find one, then the inode number we are trying to
1035 * allocate is not unique and so we should not use it.
1036 *
1037 * Returns 1 if the inode number is unique, 0 if it is not.
1038 */
1039static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1040{
1041 struct hlist_head *b = inode_hashtable + hash(sb, ino);
1042 struct hlist_node *node;
1043 struct inode *inode;
1044
Dave Chinner67a23c42011-03-22 22:23:42 +11001045 spin_lock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001046 hlist_for_each_entry(inode, node, b, i_hash) {
Dave Chinner67a23c42011-03-22 22:23:42 +11001047 if (inode->i_ino == ino && inode->i_sb == sb) {
1048 spin_unlock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001049 return 0;
Dave Chinner67a23c42011-03-22 22:23:42 +11001050 }
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001051 }
Dave Chinner67a23c42011-03-22 22:23:42 +11001052 spin_unlock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001053
1054 return 1;
1055}
1056
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057/**
1058 * iunique - get a unique inode number
1059 * @sb: superblock
1060 * @max_reserved: highest reserved inode number
1061 *
1062 * Obtain an inode number that is unique on the system for a given
1063 * superblock. This is used by file systems that have no natural
1064 * permanent inode numbering system. An inode number is returned that
1065 * is higher than the reserved limit but unique.
1066 *
1067 * BUGS:
1068 * With a large number of inodes live on the file system this function
1069 * currently becomes quite slow.
1070 */
1071ino_t iunique(struct super_block *sb, ino_t max_reserved)
1072{
Jeff Layton866b04f2007-05-08 00:32:29 -07001073 /*
1074 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1075 * error if st_ino won't fit in target struct field. Use 32bit counter
1076 * here to attempt to avoid that.
1077 */
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001078 static DEFINE_SPINLOCK(iunique_lock);
Jeff Layton866b04f2007-05-08 00:32:29 -07001079 static unsigned int counter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 ino_t res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001082 spin_lock(&iunique_lock);
Jeffrey Layton3361c7b2007-05-08 00:29:48 -07001083 do {
1084 if (counter <= max_reserved)
1085 counter = max_reserved + 1;
1086 res = counter++;
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001087 } while (!test_inode_iunique(sb, res));
1088 spin_unlock(&iunique_lock);
Jeffrey Layton3361c7b2007-05-08 00:29:48 -07001089
1090 return res;
1091}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092EXPORT_SYMBOL(iunique);
1093
1094struct inode *igrab(struct inode *inode)
1095{
Dave Chinner250df6e2011-03-22 22:23:36 +11001096 spin_lock(&inode->i_lock);
1097 if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +11001099 spin_unlock(&inode->i_lock);
1100 } else {
1101 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 /*
1103 * Handle the case where s_op->clear_inode is not been
1104 * called yet, and somebody is calling igrab
1105 * while the inode is getting freed.
1106 */
1107 inode = NULL;
Dave Chinner250df6e2011-03-22 22:23:36 +11001108 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 return inode;
1110}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111EXPORT_SYMBOL(igrab);
1112
1113/**
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001114 * ilookup5_nowait - search for an inode in the inode cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 * @sb: super block of file system to search
1116 * @hashval: hash value (usually inode number) to search for
1117 * @test: callback used for comparisons between inodes
1118 * @data: opaque data pointer to pass to @test
1119 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001120 * Search for the inode specified by @hashval and @data in the inode cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 * If the inode is in the cache, the inode is returned with an incremented
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001122 * reference count.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001123 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001124 * Note: I_NEW is not waited upon so you have to be very careful what you do
1125 * with the returned inode. You probably should be using ilookup5() instead.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001126 *
Randy Dunlapb6d0ad62011-03-26 13:27:47 -07001127 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001128 */
1129struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1130 int (*test)(struct inode *, void *), void *data)
1131{
1132 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001133 struct inode *inode;
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001134
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001135 spin_lock(&inode_hash_lock);
1136 inode = find_inode(sb, head, test, data);
1137 spin_unlock(&inode_hash_lock);
1138
1139 return inode;
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001140}
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001141EXPORT_SYMBOL(ilookup5_nowait);
1142
1143/**
1144 * ilookup5 - search for an inode in the inode cache
1145 * @sb: super block of file system to search
1146 * @hashval: hash value (usually inode number) to search for
1147 * @test: callback used for comparisons between inodes
1148 * @data: opaque data pointer to pass to @test
1149 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001150 * Search for the inode specified by @hashval and @data in the inode cache,
1151 * and if the inode is in the cache, return the inode with an incremented
1152 * reference count. Waits on I_NEW before returning the inode.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001153 * returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001155 * This is a generalized version of ilookup() for file systems where the
1156 * inode number is not sufficient for unique identification of an inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001158 * Note: @test is called with the inode_hash_lock held, so can't sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 */
1160struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1161 int (*test)(struct inode *, void *), void *data)
1162{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001163 struct inode *inode = ilookup5_nowait(sb, hashval, test, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001165 if (inode)
1166 wait_on_inode(inode);
1167 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169EXPORT_SYMBOL(ilookup5);
1170
1171/**
1172 * ilookup - search for an inode in the inode cache
1173 * @sb: super block of file system to search
1174 * @ino: inode number to search for
1175 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001176 * Search for the inode @ino in the inode cache, and if the inode is in the
1177 * cache, the inode is returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 */
1179struct inode *ilookup(struct super_block *sb, unsigned long ino)
1180{
1181 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001182 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001184 spin_lock(&inode_hash_lock);
1185 inode = find_inode_fast(sb, head, ino);
1186 spin_unlock(&inode_hash_lock);
1187
1188 if (inode)
1189 wait_on_inode(inode);
1190 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192EXPORT_SYMBOL(ilookup);
1193
Al Viro261bca82008-12-30 01:48:21 -05001194int insert_inode_locked(struct inode *inode)
1195{
1196 struct super_block *sb = inode->i_sb;
1197 ino_t ino = inode->i_ino;
1198 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Al Viro261bca82008-12-30 01:48:21 -05001199
Al Viro261bca82008-12-30 01:48:21 -05001200 while (1) {
Al Viro72a43d62009-05-13 19:13:40 +01001201 struct hlist_node *node;
1202 struct inode *old = NULL;
Dave Chinner67a23c42011-03-22 22:23:42 +11001203 spin_lock(&inode_hash_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001204 hlist_for_each_entry(old, node, head, i_hash) {
1205 if (old->i_ino != ino)
1206 continue;
1207 if (old->i_sb != sb)
1208 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001209 spin_lock(&old->i_lock);
1210 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1211 spin_unlock(&old->i_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001212 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001213 }
Al Viro72a43d62009-05-13 19:13:40 +01001214 break;
1215 }
1216 if (likely(!node)) {
Dave Chinner250df6e2011-03-22 22:23:36 +11001217 spin_lock(&inode->i_lock);
1218 inode->i_state |= I_NEW;
Al Viro261bca82008-12-30 01:48:21 -05001219 hlist_add_head(&inode->i_hash, head);
Dave Chinner250df6e2011-03-22 22:23:36 +11001220 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001221 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001222 return 0;
1223 }
1224 __iget(old);
Dave Chinner250df6e2011-03-22 22:23:36 +11001225 spin_unlock(&old->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001226 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001227 wait_on_inode(old);
Al Viro1d3382cb2010-10-23 15:19:20 -04001228 if (unlikely(!inode_unhashed(old))) {
Al Viro261bca82008-12-30 01:48:21 -05001229 iput(old);
1230 return -EBUSY;
1231 }
1232 iput(old);
1233 }
1234}
Al Viro261bca82008-12-30 01:48:21 -05001235EXPORT_SYMBOL(insert_inode_locked);
1236
1237int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1238 int (*test)(struct inode *, void *), void *data)
1239{
1240 struct super_block *sb = inode->i_sb;
1241 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Al Viro261bca82008-12-30 01:48:21 -05001242
Al Viro261bca82008-12-30 01:48:21 -05001243 while (1) {
Al Viro72a43d62009-05-13 19:13:40 +01001244 struct hlist_node *node;
1245 struct inode *old = NULL;
1246
Dave Chinner67a23c42011-03-22 22:23:42 +11001247 spin_lock(&inode_hash_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001248 hlist_for_each_entry(old, node, head, i_hash) {
1249 if (old->i_sb != sb)
1250 continue;
1251 if (!test(old, data))
1252 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001253 spin_lock(&old->i_lock);
1254 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1255 spin_unlock(&old->i_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001256 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001257 }
Al Viro72a43d62009-05-13 19:13:40 +01001258 break;
1259 }
1260 if (likely(!node)) {
Dave Chinner250df6e2011-03-22 22:23:36 +11001261 spin_lock(&inode->i_lock);
1262 inode->i_state |= I_NEW;
Al Viro261bca82008-12-30 01:48:21 -05001263 hlist_add_head(&inode->i_hash, head);
Dave Chinner250df6e2011-03-22 22:23:36 +11001264 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001265 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001266 return 0;
1267 }
1268 __iget(old);
Dave Chinner250df6e2011-03-22 22:23:36 +11001269 spin_unlock(&old->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001270 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001271 wait_on_inode(old);
Al Viro1d3382cb2010-10-23 15:19:20 -04001272 if (unlikely(!inode_unhashed(old))) {
Al Viro261bca82008-12-30 01:48:21 -05001273 iput(old);
1274 return -EBUSY;
1275 }
1276 iput(old);
1277 }
1278}
Al Viro261bca82008-12-30 01:48:21 -05001279EXPORT_SYMBOL(insert_inode_locked4);
1280
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281
Al Viro45321ac2010-06-07 13:43:19 -04001282int generic_delete_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283{
Al Viro45321ac2010-06-07 13:43:19 -04001284 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286EXPORT_SYMBOL(generic_delete_inode);
1287
Al Viro45321ac2010-06-07 13:43:19 -04001288/*
1289 * Normal UNIX filesystem behaviour: delete the
1290 * inode when the usage count drops to zero, and
1291 * i_nlink is zero.
Jan Kara22fe40422009-09-18 13:05:44 -07001292 */
Al Viro45321ac2010-06-07 13:43:19 -04001293int generic_drop_inode(struct inode *inode)
1294{
Al Viro1d3382cb2010-10-23 15:19:20 -04001295 return !inode->i_nlink || inode_unhashed(inode);
Al Viro45321ac2010-06-07 13:43:19 -04001296}
1297EXPORT_SYMBOL_GPL(generic_drop_inode);
1298
1299/*
1300 * Called when we're dropping the last reference
1301 * to an inode.
1302 *
1303 * Call the FS "drop_inode()" function, defaulting to
1304 * the legacy UNIX filesystem behaviour. If it tells
1305 * us to evict inode, do so. Otherwise, retain inode
1306 * in cache if fs is alive, sync and evict if fs is
1307 * shutting down.
1308 */
1309static void iput_final(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310{
1311 struct super_block *sb = inode->i_sb;
Al Viro45321ac2010-06-07 13:43:19 -04001312 const struct super_operations *op = inode->i_sb->s_op;
1313 int drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314
Dave Chinner250df6e2011-03-22 22:23:36 +11001315 WARN_ON(inode->i_state & I_NEW);
1316
Al Viroe7f59092011-07-07 15:45:59 -04001317 if (op->drop_inode)
Al Viro45321ac2010-06-07 13:43:19 -04001318 drop = op->drop_inode(inode);
1319 else
1320 drop = generic_drop_inode(inode);
1321
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001322 if (!drop && (sb->s_flags & MS_ACTIVE)) {
1323 inode->i_state |= I_REFERENCED;
1324 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
1325 inode_lru_list_add(inode);
1326 spin_unlock(&inode->i_lock);
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001327 return;
1328 }
1329
Al Viro45321ac2010-06-07 13:43:19 -04001330 if (!drop) {
Alexander Viro991114c2005-06-23 00:09:01 -07001331 inode->i_state |= I_WILL_FREE;
Dave Chinner250df6e2011-03-22 22:23:36 +11001332 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 write_inode_now(inode, 1);
Dave Chinner250df6e2011-03-22 22:23:36 +11001334 spin_lock(&inode->i_lock);
Nick Piggin7ef0d732009-03-12 14:31:38 -07001335 WARN_ON(inode->i_state & I_NEW);
Alexander Viro991114c2005-06-23 00:09:01 -07001336 inode->i_state &= ~I_WILL_FREE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 }
Nick Piggin7ccf19a2010-10-21 11:49:30 +11001338
Alexander Viro991114c2005-06-23 00:09:01 -07001339 inode->i_state |= I_FREEING;
Eric Dumazetc4ae0c62011-07-28 06:55:13 +02001340 if (!list_empty(&inode->i_lru))
1341 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +11001342 spin_unlock(&inode->i_lock);
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001343
1344 evict(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345}
1346
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347/**
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301348 * iput - put an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 * @inode: inode to put
1350 *
1351 * Puts an inode, dropping its usage count. If the inode use count hits
1352 * zero, the inode is then freed and may also be destroyed.
1353 *
1354 * Consequently, iput() can sleep.
1355 */
1356void iput(struct inode *inode)
1357{
1358 if (inode) {
Al Viroa4ffdde2010-06-02 17:38:30 -04001359 BUG_ON(inode->i_state & I_CLEAR);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
Dave Chinnerf283c862011-03-22 22:23:39 +11001361 if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 iput_final(inode);
1363 }
1364}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365EXPORT_SYMBOL(iput);
1366
1367/**
1368 * bmap - find a block number in a file
1369 * @inode: inode of file
1370 * @block: block to find
1371 *
1372 * Returns the block number on the device holding the inode that
1373 * is the disk block number for the block of the file requested.
1374 * That is, asked for block 4 of inode 1 the function will return the
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301375 * disk block relative to the disk start that holds that block of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 * file.
1377 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301378sector_t bmap(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379{
1380 sector_t res = 0;
1381 if (inode->i_mapping->a_ops->bmap)
1382 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1383 return res;
1384}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385EXPORT_SYMBOL(bmap);
1386
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001387/*
1388 * With relative atime, only update atime if the previous atime is
1389 * earlier than either the ctime or mtime or if at least a day has
1390 * passed since the last atime update.
1391 */
1392static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1393 struct timespec now)
1394{
1395
1396 if (!(mnt->mnt_flags & MNT_RELATIME))
1397 return 1;
1398 /*
1399 * Is mtime younger than atime? If yes, update atime:
1400 */
1401 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1402 return 1;
1403 /*
1404 * Is ctime younger than atime? If yes, update atime:
1405 */
1406 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1407 return 1;
1408
1409 /*
1410 * Is the previous atime value older than a day? If yes,
1411 * update atime:
1412 */
1413 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1414 return 1;
1415 /*
1416 * Good, we can skip the atime update:
1417 */
1418 return 0;
1419}
1420
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421/**
Christoph Hellwig869243a2006-01-09 20:52:03 -08001422 * touch_atime - update the access time
1423 * @mnt: mount the inode is accessed on
Martin Waitz7045f372006-02-01 03:06:57 -08001424 * @dentry: dentry accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 *
1426 * Update the accessed time on an inode and mark it for writeback.
1427 * This function automatically handles read only file systems and media,
1428 * as well as the "noatime" flag and inode specific "noatime" markers.
1429 */
Christoph Hellwig869243a2006-01-09 20:52:03 -08001430void touch_atime(struct vfsmount *mnt, struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431{
Christoph Hellwig869243a2006-01-09 20:52:03 -08001432 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 struct timespec now;
1434
Andrew Mortonb2276132006-12-13 00:34:33 -08001435 if (inode->i_flags & S_NOATIME)
Andi Kleenb12536c2009-09-18 13:05:47 -07001436 return;
Eric Dumazet37756ce2007-02-10 01:44:49 -08001437 if (IS_NOATIME(inode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001438 return;
Andrew Mortonb2276132006-12-13 00:34:33 -08001439 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001440 return;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001441
Dave Hansencdb70f32008-02-15 14:37:41 -08001442 if (mnt->mnt_flags & MNT_NOATIME)
Andi Kleenb12536c2009-09-18 13:05:47 -07001443 return;
Dave Hansencdb70f32008-02-15 14:37:41 -08001444 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001445 return;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001446
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 now = current_fs_time(inode->i_sb);
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001448
1449 if (!relatime_need_update(mnt, inode, now))
Andi Kleenb12536c2009-09-18 13:05:47 -07001450 return;
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001451
Valerie Henson47ae32d2006-12-13 00:34:34 -08001452 if (timespec_equal(&inode->i_atime, &now))
Andi Kleenb12536c2009-09-18 13:05:47 -07001453 return;
1454
1455 if (mnt_want_write(mnt))
1456 return;
Valerie Henson47ae32d2006-12-13 00:34:34 -08001457
1458 inode->i_atime = now;
1459 mark_inode_dirty_sync(inode);
Dave Hansencdb70f32008-02-15 14:37:41 -08001460 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461}
Christoph Hellwig869243a2006-01-09 20:52:03 -08001462EXPORT_SYMBOL(touch_atime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463
1464/**
Christoph Hellwig870f4812006-01-09 20:52:01 -08001465 * file_update_time - update mtime and ctime time
1466 * @file: file accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 *
Christoph Hellwig870f4812006-01-09 20:52:01 -08001468 * Update the mtime and ctime members of an inode and mark the inode
1469 * for writeback. Note that this function is meant exclusively for
1470 * usage in the file write path of filesystems, and filesystems may
1471 * choose to explicitly ignore update via this function with the
Wolfram Sang2eadfc02009-04-02 15:23:37 +02001472 * S_NOCMTIME inode flag, e.g. for network filesystem where these
Christoph Hellwig870f4812006-01-09 20:52:01 -08001473 * timestamps are handled by the server.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 */
1475
Christoph Hellwig870f4812006-01-09 20:52:01 -08001476void file_update_time(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001478 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 struct timespec now;
Andi Kleence06e0b2009-09-18 13:05:48 -07001480 enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Andi Kleence06e0b2009-09-18 13:05:48 -07001482 /* First try to exhaust all avenues to not sync */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 if (IS_NOCMTIME(inode))
1484 return;
Dave Hansen20ddee22008-02-15 14:37:43 -08001485
Andi Kleence06e0b2009-09-18 13:05:48 -07001486 now = current_fs_time(inode->i_sb);
1487 if (!timespec_equal(&inode->i_mtime, &now))
1488 sync_it = S_MTIME;
1489
1490 if (!timespec_equal(&inode->i_ctime, &now))
1491 sync_it |= S_CTIME;
1492
1493 if (IS_I_VERSION(inode))
1494 sync_it |= S_VERSION;
1495
1496 if (!sync_it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 return;
1498
Andi Kleence06e0b2009-09-18 13:05:48 -07001499 /* Finally allowed to write? Takes lock. */
1500 if (mnt_want_write_file(file))
1501 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
Andi Kleence06e0b2009-09-18 13:05:48 -07001503 /* Only change inode inside the lock region */
1504 if (sync_it & S_VERSION)
Jean Noel Cordenner7a224222008-01-28 23:58:27 -05001505 inode_inc_iversion(inode);
Andi Kleence06e0b2009-09-18 13:05:48 -07001506 if (sync_it & S_CTIME)
1507 inode->i_ctime = now;
1508 if (sync_it & S_MTIME)
1509 inode->i_mtime = now;
1510 mark_inode_dirty_sync(inode);
Dave Hansen20ddee22008-02-15 14:37:43 -08001511 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512}
Christoph Hellwig870f4812006-01-09 20:52:01 -08001513EXPORT_SYMBOL(file_update_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514
1515int inode_needs_sync(struct inode *inode)
1516{
1517 if (IS_SYNC(inode))
1518 return 1;
1519 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1520 return 1;
1521 return 0;
1522}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523EXPORT_SYMBOL(inode_needs_sync);
1524
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525int inode_wait(void *word)
1526{
1527 schedule();
1528 return 0;
1529}
Stephen Rothwelld44dab82008-11-10 17:06:05 +11001530EXPORT_SYMBOL(inode_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
1532/*
Miklos Szeredi168a9fd2005-07-12 13:58:10 -07001533 * If we try to find an inode in the inode hash while it is being
1534 * deleted, we have to wait until the filesystem completes its
1535 * deletion before reporting that it isn't found. This function waits
1536 * until the deletion _might_ have completed. Callers are responsible
1537 * to recheck inode state.
1538 *
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001539 * It doesn't matter if I_NEW is not set initially, a call to
Dave Chinner250df6e2011-03-22 22:23:36 +11001540 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1541 * will DTRT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 */
1543static void __wait_on_freeing_inode(struct inode *inode)
1544{
1545 wait_queue_head_t *wq;
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001546 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1547 wq = bit_waitqueue(&inode->i_state, __I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
Dave Chinner250df6e2011-03-22 22:23:36 +11001549 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001550 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 schedule();
1552 finish_wait(wq, &wait.wait);
Dave Chinner67a23c42011-03-22 22:23:42 +11001553 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554}
1555
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556static __initdata unsigned long ihash_entries;
1557static int __init set_ihash_entries(char *str)
1558{
1559 if (!str)
1560 return 0;
1561 ihash_entries = simple_strtoul(str, &str, 0);
1562 return 1;
1563}
1564__setup("ihash_entries=", set_ihash_entries);
1565
1566/*
1567 * Initialize the waitqueues and inode hash table.
1568 */
1569void __init inode_init_early(void)
1570{
1571 int loop;
1572
1573 /* If hashes are distributed across NUMA nodes, defer
1574 * hash allocation until vmalloc space is available.
1575 */
1576 if (hashdist)
1577 return;
1578
1579 inode_hashtable =
1580 alloc_large_system_hash("Inode-cache",
1581 sizeof(struct hlist_head),
1582 ihash_entries,
1583 14,
1584 HASH_EARLY,
1585 &i_hash_shift,
1586 &i_hash_mask,
1587 0);
1588
1589 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1590 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1591}
1592
Denis Cheng74bf17c2007-10-16 23:26:30 -07001593void __init inode_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594{
1595 int loop;
1596
1597 /* inode slab cache */
Paul Jacksonb0196002006-03-24 03:16:09 -08001598 inode_cachep = kmem_cache_create("inode_cache",
1599 sizeof(struct inode),
1600 0,
1601 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1602 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +09001603 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
1605 /* Hash may have been set up in inode_init_early */
1606 if (!hashdist)
1607 return;
1608
1609 inode_hashtable =
1610 alloc_large_system_hash("Inode-cache",
1611 sizeof(struct hlist_head),
1612 ihash_entries,
1613 14,
1614 0,
1615 &i_hash_shift,
1616 &i_hash_mask,
1617 0);
1618
1619 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1620 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1621}
1622
1623void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1624{
1625 inode->i_mode = mode;
1626 if (S_ISCHR(mode)) {
1627 inode->i_fop = &def_chr_fops;
1628 inode->i_rdev = rdev;
1629 } else if (S_ISBLK(mode)) {
1630 inode->i_fop = &def_blk_fops;
1631 inode->i_rdev = rdev;
1632 } else if (S_ISFIFO(mode))
1633 inode->i_fop = &def_fifo_fops;
1634 else if (S_ISSOCK(mode))
1635 inode->i_fop = &bad_sock_fops;
1636 else
Manish Katiyaraf0d9ae2009-09-18 13:05:43 -07001637 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
1638 " inode %s:%lu\n", mode, inode->i_sb->s_id,
1639 inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640}
1641EXPORT_SYMBOL(init_special_inode);
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001642
1643/**
Ben Hutchingseaae668d2011-02-15 12:48:09 +00001644 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001645 * @inode: New inode
1646 * @dir: Directory inode
1647 * @mode: mode of the new inode
1648 */
1649void inode_init_owner(struct inode *inode, const struct inode *dir,
1650 mode_t mode)
1651{
1652 inode->i_uid = current_fsuid();
1653 if (dir && dir->i_mode & S_ISGID) {
1654 inode->i_gid = dir->i_gid;
1655 if (S_ISDIR(mode))
1656 mode |= S_ISGID;
1657 } else
1658 inode->i_gid = current_fsgid();
1659 inode->i_mode = mode;
1660}
1661EXPORT_SYMBOL(inode_init_owner);
Serge E. Hallyne795b712011-03-23 16:43:25 -07001662
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001663/**
1664 * inode_owner_or_capable - check current task permissions to inode
1665 * @inode: inode being checked
1666 *
1667 * Return true if current either has CAP_FOWNER to the inode, or
1668 * owns the file.
Serge E. Hallyne795b712011-03-23 16:43:25 -07001669 */
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001670bool inode_owner_or_capable(const struct inode *inode)
Serge E. Hallyne795b712011-03-23 16:43:25 -07001671{
1672 struct user_namespace *ns = inode_userns(inode);
1673
1674 if (current_user_ns() == ns && current_fsuid() == inode->i_uid)
1675 return true;
1676 if (ns_capable(ns, CAP_FOWNER))
1677 return true;
1678 return false;
1679}
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001680EXPORT_SYMBOL(inode_owner_or_capable);