blob: 9f4f5fecc0963c12db1f806be61999a26a9ba7ea [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 */
Al Viroe59cc472011-12-07 13:17:19 -05005#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07006#include <linux/fs.h>
7#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/backing-dev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/hash.h>
10#include <linux/swap.h>
11#include <linux/security.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/cdev.h>
13#include <linux/bootmem.h>
Eric Paris3be25f42009-05-21 17:01:26 -040014#include <linux/fsnotify.h>
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -080015#include <linux/mount.h>
Al Virof19d4a82009-06-08 19:50:45 -040016#include <linux/posix_acl.h>
Heiko Carstens9ce6e0b2011-05-22 18:54:21 +020017#include <linux/prefetch.h>
Christoph Hellwig4b4563d2011-05-27 09:28:01 -040018#include <linux/buffer_head.h> /* for inode_has_buffers */
Miklos Szeredi7ada4db2011-11-21 12:11:32 +010019#include <linux/ratelimit.h>
Dave Chinnera66979a2011-03-22 22:23:41 +110020#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
22/*
Christoph Hellwig4b4563d2011-05-27 09:28:01 -040023 * Inode locking rules:
Dave Chinner250df6e2011-03-22 22:23:36 +110024 *
25 * inode->i_lock protects:
26 * inode->i_state, inode->i_hash, __iget()
Dave Chinner09cc9fc2011-07-08 14:14:40 +100027 * inode->i_sb->s_inode_lru_lock protects:
Dave Chinner98b745c2011-07-08 14:14:39 +100028 * inode->i_sb->s_inode_lru, inode->i_lru
Dave Chinner55fa6092011-03-22 22:23:40 +110029 * inode_sb_list_lock protects:
30 * sb->s_inodes, inode->i_sb_list
Christoph Hellwigf758eea2011-04-21 18:19:44 -060031 * bdi->wb.list_lock protects:
Dave Chinnera66979a2011-03-22 22:23:41 +110032 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
Dave Chinner67a23c42011-03-22 22:23:42 +110033 * inode_hash_lock protects:
34 * inode_hashtable, inode->i_hash
Dave Chinner250df6e2011-03-22 22:23:36 +110035 *
36 * Lock ordering:
Dave Chinner55fa6092011-03-22 22:23:40 +110037 *
38 * inode_sb_list_lock
39 * inode->i_lock
Dave Chinner09cc9fc2011-07-08 14:14:40 +100040 * inode->i_sb->s_inode_lru_lock
Dave Chinnera66979a2011-03-22 22:23:41 +110041 *
Christoph Hellwigf758eea2011-04-21 18:19:44 -060042 * bdi->wb.list_lock
Dave Chinnera66979a2011-03-22 22:23:41 +110043 * inode->i_lock
Dave Chinner67a23c42011-03-22 22:23:42 +110044 *
45 * inode_hash_lock
46 * inode_sb_list_lock
47 * inode->i_lock
48 *
49 * iunique_lock
50 * inode_hash_lock
Dave Chinner250df6e2011-03-22 22:23:36 +110051 */
52
Eric Dumazetfa3536c2006-03-26 01:37:24 -080053static unsigned int i_hash_mask __read_mostly;
54static unsigned int i_hash_shift __read_mostly;
Dave Chinner67a23c42011-03-22 22:23:42 +110055static struct hlist_head *inode_hashtable __read_mostly;
56static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070057
Dave Chinner55fa6092011-03-22 22:23:40 +110058__cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
59
Linus Torvalds1da177e2005-04-16 15:20:36 -070060/*
Jens Axboe7dcda1c2011-04-05 23:51:48 +020061 * Empty aops. Can be used for the cases where the user does not
62 * define any of the address_space operations.
63 */
64const struct address_space_operations empty_aops = {
65};
66EXPORT_SYMBOL(empty_aops);
67
68/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 * Statistics gathering..
70 */
71struct inodes_stat_t inodes_stat;
72
Nick Piggin3e880fb2011-01-07 17:49:19 +110073static DEFINE_PER_CPU(unsigned int, nr_inodes);
Dave Chinnerfcb94f72011-07-08 14:14:38 +100074static DEFINE_PER_CPU(unsigned int, nr_unused);
Dave Chinnercffbc8a2010-10-23 05:03:02 -040075
Manish Katiyar6b3304b2009-03-31 19:35:54 +053076static struct kmem_cache *inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070077
Nick Piggin3e880fb2011-01-07 17:49:19 +110078static int get_nr_inodes(void)
Dave Chinnercffbc8a2010-10-23 05:03:02 -040079{
Nick Piggin3e880fb2011-01-07 17:49:19 +110080 int i;
81 int sum = 0;
82 for_each_possible_cpu(i)
83 sum += per_cpu(nr_inodes, i);
84 return sum < 0 ? 0 : sum;
Dave Chinnercffbc8a2010-10-23 05:03:02 -040085}
86
87static inline int get_nr_inodes_unused(void)
88{
Dave Chinnerfcb94f72011-07-08 14:14:38 +100089 int i;
90 int sum = 0;
91 for_each_possible_cpu(i)
92 sum += per_cpu(nr_unused, i);
93 return sum < 0 ? 0 : sum;
Dave Chinnercffbc8a2010-10-23 05:03:02 -040094}
95
96int get_nr_dirty_inodes(void)
97{
Nick Piggin3e880fb2011-01-07 17:49:19 +110098 /* not actually dirty inodes, but a wild approximation */
Dave Chinnercffbc8a2010-10-23 05:03:02 -040099 int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
100 return nr_dirty > 0 ? nr_dirty : 0;
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400101}
102
103/*
104 * Handle nr_inode sysctl
105 */
106#ifdef CONFIG_SYSCTL
107int proc_nr_inodes(ctl_table *table, int write,
108 void __user *buffer, size_t *lenp, loff_t *ppos)
109{
110 inodes_stat.nr_inodes = get_nr_inodes();
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000111 inodes_stat.nr_unused = get_nr_inodes_unused();
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400112 return proc_dointvec(table, write, buffer, lenp, ppos);
113}
114#endif
115
David Chinner2cb15992008-10-30 17:32:23 +1100116/**
117 * inode_init_always - perform inode structure intialisation
Randy Dunlap0bc02f32009-01-06 14:41:13 -0800118 * @sb: superblock inode belongs to
119 * @inode: inode to initialise
David Chinner2cb15992008-10-30 17:32:23 +1100120 *
121 * These are initializations that need to be done on every inode
122 * allocation as the fields are not initialised by slab allocation.
123 */
Christoph Hellwig54e34622009-08-07 14:38:25 -0300124int inode_init_always(struct super_block *sb, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125{
Alexey Dobriyan6e1d5dc2009-09-21 17:01:11 -0700126 static const struct inode_operations empty_iops;
Arjan van de Ven99ac48f2006-03-28 01:56:41 -0800127 static const struct file_operations empty_fops;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530128 struct address_space *const mapping = &inode->i_data;
David Chinner2cb15992008-10-30 17:32:23 +1100129
130 inode->i_sb = sb;
131 inode->i_blkbits = sb->s_blocksize_bits;
132 inode->i_flags = 0;
133 atomic_set(&inode->i_count, 1);
134 inode->i_op = &empty_iops;
135 inode->i_fop = &empty_fops;
Miklos Szeredia78ef702011-10-28 14:13:30 +0200136 inode->__i_nlink = 1;
Linus Torvalds3ddcd052011-08-06 22:45:50 -0700137 inode->i_opflags = 0;
Al Viro56ff5ef2008-12-09 09:34:39 -0500138 inode->i_uid = 0;
139 inode->i_gid = 0;
David Chinner2cb15992008-10-30 17:32:23 +1100140 atomic_set(&inode->i_writecount, 0);
141 inode->i_size = 0;
142 inode->i_blocks = 0;
143 inode->i_bytes = 0;
144 inode->i_generation = 0;
145#ifdef CONFIG_QUOTA
146 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
147#endif
148 inode->i_pipe = NULL;
149 inode->i_bdev = NULL;
150 inode->i_cdev = NULL;
151 inode->i_rdev = 0;
152 inode->dirtied_when = 0;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500153
154 if (security_inode_alloc(inode))
Christoph Hellwig54e34622009-08-07 14:38:25 -0300155 goto out;
David Chinner2cb15992008-10-30 17:32:23 +1100156 spin_lock_init(&inode->i_lock);
157 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
158
159 mutex_init(&inode->i_mutex);
160 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
161
Christoph Hellwigbd5fe6c2011-06-24 14:29:43 -0400162 atomic_set(&inode->i_dio_count, 0);
David Chinner2cb15992008-10-30 17:32:23 +1100163
164 mapping->a_ops = &empty_aops;
165 mapping->host = inode;
166 mapping->flags = 0;
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800167 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
David Chinner2cb15992008-10-30 17:32:23 +1100168 mapping->assoc_mapping = NULL;
169 mapping->backing_dev_info = &default_backing_dev_info;
170 mapping->writeback_index = 0;
171
172 /*
173 * If the block_device provides a backing_dev_info for client
174 * inodes then use that. Otherwise the inode share the bdev's
175 * backing_dev_info.
176 */
177 if (sb->s_bdev) {
178 struct backing_dev_info *bdi;
179
Jens Axboe2c96ce92009-09-15 09:43:56 +0200180 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
David Chinner2cb15992008-10-30 17:32:23 +1100181 mapping->backing_dev_info = bdi;
182 }
183 inode->i_private = NULL;
184 inode->i_mapping = mapping;
Al Viro6b520e02011-12-12 15:51:45 -0500185 INIT_LIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */
Al Virof19d4a82009-06-08 19:50:45 -0400186#ifdef CONFIG_FS_POSIX_ACL
187 inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
188#endif
David Chinner2cb15992008-10-30 17:32:23 +1100189
Eric Paris3be25f42009-05-21 17:01:26 -0400190#ifdef CONFIG_FSNOTIFY
191 inode->i_fsnotify_mask = 0;
192#endif
193
Nick Piggin3e880fb2011-01-07 17:49:19 +1100194 this_cpu_inc(nr_inodes);
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400195
Christoph Hellwig54e34622009-08-07 14:38:25 -0300196 return 0;
Christoph Hellwig54e34622009-08-07 14:38:25 -0300197out:
198 return -ENOMEM;
David Chinner2cb15992008-10-30 17:32:23 +1100199}
200EXPORT_SYMBOL(inode_init_always);
201
202static struct inode *alloc_inode(struct super_block *sb)
203{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 struct inode *inode;
205
206 if (sb->s_op->alloc_inode)
207 inode = sb->s_op->alloc_inode(sb);
208 else
David Chinner2cb15992008-10-30 17:32:23 +1100209 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210
Christoph Hellwig54e34622009-08-07 14:38:25 -0300211 if (!inode)
212 return NULL;
213
214 if (unlikely(inode_init_always(sb, inode))) {
215 if (inode->i_sb->s_op->destroy_inode)
216 inode->i_sb->s_op->destroy_inode(inode);
217 else
218 kmem_cache_free(inode_cachep, inode);
219 return NULL;
220 }
221
222 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
224
Nick Pigginff0c7d12011-01-07 17:49:50 +1100225void free_inode_nonrcu(struct inode *inode)
226{
227 kmem_cache_free(inode_cachep, inode);
228}
229EXPORT_SYMBOL(free_inode_nonrcu);
230
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300231void __destroy_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232{
Eric Sesterhennb7542f82006-04-02 13:38:18 +0200233 BUG_ON(inode_has_buffers(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 security_inode_free(inode);
Eric Paris3be25f42009-05-21 17:01:26 -0400235 fsnotify_inode_delete(inode);
Miklos Szeredi7ada4db2011-11-21 12:11:32 +0100236 if (!inode->i_nlink) {
237 WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0);
238 atomic_long_dec(&inode->i_sb->s_remove_count);
239 }
240
Al Virof19d4a82009-06-08 19:50:45 -0400241#ifdef CONFIG_FS_POSIX_ACL
242 if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED)
243 posix_acl_release(inode->i_acl);
244 if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
245 posix_acl_release(inode->i_default_acl);
246#endif
Nick Piggin3e880fb2011-01-07 17:49:19 +1100247 this_cpu_dec(nr_inodes);
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300248}
249EXPORT_SYMBOL(__destroy_inode);
250
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100251static void i_callback(struct rcu_head *head)
252{
253 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100254 kmem_cache_free(inode_cachep, inode);
255}
256
Christoph Hellwig56b0dac2010-10-06 10:48:55 +0200257static void destroy_inode(struct inode *inode)
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300258{
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100259 BUG_ON(!list_empty(&inode->i_lru));
Christoph Hellwig2e00c972009-08-07 14:38:29 -0300260 __destroy_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 if (inode->i_sb->s_op->destroy_inode)
262 inode->i_sb->s_op->destroy_inode(inode);
263 else
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100264 call_rcu(&inode->i_rcu, i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
Miklos Szeredi7ada4db2011-11-21 12:11:32 +0100267/**
268 * drop_nlink - directly drop an inode's link count
269 * @inode: inode
270 *
271 * This is a low-level filesystem helper to replace any
272 * direct filesystem manipulation of i_nlink. In cases
273 * where we are attempting to track writes to the
274 * filesystem, a decrement to zero means an imminent
275 * write when the file is truncated and actually unlinked
276 * on the filesystem.
277 */
278void drop_nlink(struct inode *inode)
279{
280 WARN_ON(inode->i_nlink == 0);
281 inode->__i_nlink--;
282 if (!inode->i_nlink)
283 atomic_long_inc(&inode->i_sb->s_remove_count);
284}
285EXPORT_SYMBOL(drop_nlink);
286
287/**
288 * clear_nlink - directly zero an inode's link count
289 * @inode: inode
290 *
291 * This is a low-level filesystem helper to replace any
292 * direct filesystem manipulation of i_nlink. See
293 * drop_nlink() for why we care about i_nlink hitting zero.
294 */
295void clear_nlink(struct inode *inode)
296{
297 if (inode->i_nlink) {
298 inode->__i_nlink = 0;
299 atomic_long_inc(&inode->i_sb->s_remove_count);
300 }
301}
302EXPORT_SYMBOL(clear_nlink);
303
304/**
305 * set_nlink - directly set an inode's link count
306 * @inode: inode
307 * @nlink: new nlink (should be non-zero)
308 *
309 * This is a low-level filesystem helper to replace any
310 * direct filesystem manipulation of i_nlink.
311 */
312void set_nlink(struct inode *inode, unsigned int nlink)
313{
314 if (!nlink) {
Miklos Szeredi7ada4db2011-11-21 12:11:32 +0100315 clear_nlink(inode);
316 } else {
317 /* Yes, some filesystems do change nlink from zero to one */
318 if (inode->i_nlink == 0)
319 atomic_long_dec(&inode->i_sb->s_remove_count);
320
321 inode->__i_nlink = nlink;
322 }
323}
324EXPORT_SYMBOL(set_nlink);
325
326/**
327 * inc_nlink - directly increment an inode's link count
328 * @inode: inode
329 *
330 * This is a low-level filesystem helper to replace any
331 * direct filesystem manipulation of i_nlink. Currently,
332 * it is only here for parity with dec_nlink().
333 */
334void inc_nlink(struct inode *inode)
335{
336 if (WARN_ON(inode->i_nlink == 0))
337 atomic_long_dec(&inode->i_sb->s_remove_count);
338
339 inode->__i_nlink++;
340}
341EXPORT_SYMBOL(inc_nlink);
342
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100343void address_space_init_once(struct address_space *mapping)
344{
345 memset(mapping, 0, sizeof(*mapping));
346 INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
347 spin_lock_init(&mapping->tree_lock);
Peter Zijlstra3d48ae42011-05-24 17:12:06 -0700348 mutex_init(&mapping->i_mmap_mutex);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100349 INIT_LIST_HEAD(&mapping->private_list);
350 spin_lock_init(&mapping->private_lock);
351 INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
352 INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100353}
354EXPORT_SYMBOL(address_space_init_once);
355
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356/*
357 * These are initializations that only need to be done
358 * once, because the fields are idempotent across use
359 * of the inode, so let the slab aware of that.
360 */
361void inode_init_once(struct inode *inode)
362{
363 memset(inode, 0, sizeof(*inode));
364 INIT_HLIST_NODE(&inode->i_hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 INIT_LIST_HEAD(&inode->i_devices);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100366 INIT_LIST_HEAD(&inode->i_wb_list);
367 INIT_LIST_HEAD(&inode->i_lru);
Miklos Szeredi2aa15892011-02-23 13:49:47 +0100368 address_space_init_once(&inode->i_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 i_size_ordered_init(inode);
Eric Paris3be25f42009-05-21 17:01:26 -0400370#ifdef CONFIG_FSNOTIFY
Eric Parise61ce862009-12-17 21:24:24 -0500371 INIT_HLIST_HEAD(&inode->i_fsnotify_marks);
Eric Paris3be25f42009-05-21 17:01:26 -0400372#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374EXPORT_SYMBOL(inode_init_once);
375
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700376static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530378 struct inode *inode = (struct inode *) foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379
Christoph Lametera35afb82007-05-16 22:10:57 -0700380 inode_init_once(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381}
382
383/*
Dave Chinner250df6e2011-03-22 22:23:36 +1100384 * inode->i_lock must be held
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530386void __iget(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387{
Nick Piggin9e38d862010-10-23 06:55:17 -0400388 atomic_inc(&inode->i_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389}
390
Al Viro7de9c6ee2010-10-23 11:11:40 -0400391/*
392 * get additional reference to inode; caller must already hold one.
393 */
394void ihold(struct inode *inode)
395{
396 WARN_ON(atomic_inc_return(&inode->i_count) < 2);
397}
398EXPORT_SYMBOL(ihold);
399
Nick Piggin9e38d862010-10-23 06:55:17 -0400400static void inode_lru_list_add(struct inode *inode)
401{
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000402 spin_lock(&inode->i_sb->s_inode_lru_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100403 if (list_empty(&inode->i_lru)) {
Dave Chinner98b745c2011-07-08 14:14:39 +1000404 list_add(&inode->i_lru, &inode->i_sb->s_inode_lru);
405 inode->i_sb->s_nr_inodes_unused++;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000406 this_cpu_inc(nr_unused);
Nick Piggin9e38d862010-10-23 06:55:17 -0400407 }
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000408 spin_unlock(&inode->i_sb->s_inode_lru_lock);
Nick Piggin9e38d862010-10-23 06:55:17 -0400409}
410
411static void inode_lru_list_del(struct inode *inode)
412{
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000413 spin_lock(&inode->i_sb->s_inode_lru_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100414 if (!list_empty(&inode->i_lru)) {
415 list_del_init(&inode->i_lru);
Dave Chinner98b745c2011-07-08 14:14:39 +1000416 inode->i_sb->s_nr_inodes_unused--;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000417 this_cpu_dec(nr_unused);
Nick Piggin9e38d862010-10-23 06:55:17 -0400418 }
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000419 spin_unlock(&inode->i_sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420}
421
Christoph Hellwig646ec462010-10-23 07:15:32 -0400422/**
423 * inode_sb_list_add - add inode to the superblock list of inodes
424 * @inode: inode to add
425 */
426void inode_sb_list_add(struct inode *inode)
427{
Dave Chinner55fa6092011-03-22 22:23:40 +1100428 spin_lock(&inode_sb_list_lock);
429 list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
430 spin_unlock(&inode_sb_list_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400431}
432EXPORT_SYMBOL_GPL(inode_sb_list_add);
433
Dave Chinner55fa6092011-03-22 22:23:40 +1100434static inline void inode_sb_list_del(struct inode *inode)
Christoph Hellwig646ec462010-10-23 07:15:32 -0400435{
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200436 if (!list_empty(&inode->i_sb_list)) {
437 spin_lock(&inode_sb_list_lock);
438 list_del_init(&inode->i_sb_list);
439 spin_unlock(&inode_sb_list_lock);
440 }
Christoph Hellwig646ec462010-10-23 07:15:32 -0400441}
442
Dave Chinner4c51acb2010-10-23 06:58:09 -0400443static unsigned long hash(struct super_block *sb, unsigned long hashval)
444{
445 unsigned long tmp;
446
447 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
448 L1_CACHE_BYTES;
Christoph Hellwig4b4563d2011-05-27 09:28:01 -0400449 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift);
450 return tmp & i_hash_mask;
Dave Chinner4c51acb2010-10-23 06:58:09 -0400451}
452
453/**
454 * __insert_inode_hash - hash an inode
455 * @inode: unhashed inode
456 * @hashval: unsigned long value used to locate this object in the
457 * inode_hashtable.
458 *
459 * Add an inode to the inode hash for this superblock.
460 */
461void __insert_inode_hash(struct inode *inode, unsigned long hashval)
462{
Christoph Hellwig646ec462010-10-23 07:15:32 -0400463 struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
464
Dave Chinner67a23c42011-03-22 22:23:42 +1100465 spin_lock(&inode_hash_lock);
Dave Chinner250df6e2011-03-22 22:23:36 +1100466 spin_lock(&inode->i_lock);
Christoph Hellwig646ec462010-10-23 07:15:32 -0400467 hlist_add_head(&inode->i_hash, b);
Dave Chinner250df6e2011-03-22 22:23:36 +1100468 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100469 spin_unlock(&inode_hash_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400470}
471EXPORT_SYMBOL(__insert_inode_hash);
472
473/**
Eric Dumazetf2ee7ab2011-07-28 06:41:09 +0200474 * __remove_inode_hash - remove an inode from the hash
Dave Chinner4c51acb2010-10-23 06:58:09 -0400475 * @inode: inode to unhash
476 *
477 * Remove an inode from the superblock.
478 */
Eric Dumazetf2ee7ab2011-07-28 06:41:09 +0200479void __remove_inode_hash(struct inode *inode)
Dave Chinner4c51acb2010-10-23 06:58:09 -0400480{
Dave Chinner67a23c42011-03-22 22:23:42 +1100481 spin_lock(&inode_hash_lock);
Dave Chinner250df6e2011-03-22 22:23:36 +1100482 spin_lock(&inode->i_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400483 hlist_del_init(&inode->i_hash);
Dave Chinner250df6e2011-03-22 22:23:36 +1100484 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100485 spin_unlock(&inode_hash_lock);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400486}
Eric Dumazetf2ee7ab2011-07-28 06:41:09 +0200487EXPORT_SYMBOL(__remove_inode_hash);
Dave Chinner4c51acb2010-10-23 06:58:09 -0400488
Al Virob0683aa2010-06-04 20:55:25 -0400489void end_writeback(struct inode *inode)
490{
491 might_sleep();
Jan Kara08142572011-06-27 16:18:10 -0700492 /*
493 * We have to cycle tree_lock here because reclaim can be still in the
494 * process of removing the last page (in __delete_from_page_cache())
495 * and we must not free mapping under it.
496 */
497 spin_lock_irq(&inode->i_data.tree_lock);
Al Virob0683aa2010-06-04 20:55:25 -0400498 BUG_ON(inode->i_data.nrpages);
Jan Kara08142572011-06-27 16:18:10 -0700499 spin_unlock_irq(&inode->i_data.tree_lock);
Al Virob0683aa2010-06-04 20:55:25 -0400500 BUG_ON(!list_empty(&inode->i_data.private_list));
501 BUG_ON(!(inode->i_state & I_FREEING));
502 BUG_ON(inode->i_state & I_CLEAR);
503 inode_sync_wait(inode);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100504 /* don't need i_lock here, no concurrent mods to i_state */
Al Virob0683aa2010-06-04 20:55:25 -0400505 inode->i_state = I_FREEING | I_CLEAR;
506}
507EXPORT_SYMBOL(end_writeback);
508
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100509/*
510 * Free the inode passed in, removing it from the lists it is still connected
511 * to. We remove any pages still attached to the inode and wait for any IO that
512 * is still in progress before finally destroying the inode.
513 *
514 * An inode must already be marked I_FREEING so that we avoid the inode being
515 * moved back onto lists if we race with other code that manipulates the lists
516 * (e.g. writeback_single_inode). The caller is responsible for setting this.
517 *
518 * An inode must already be removed from the LRU list before being evicted from
519 * the cache. This should occur atomically with setting the I_FREEING state
520 * flag, so no inodes here should ever be on the LRU when being evicted.
521 */
Al Viro644da592010-06-07 13:21:05 -0400522static void evict(struct inode *inode)
Al Virob4272d42010-06-04 19:33:20 -0400523{
524 const struct super_operations *op = inode->i_sb->s_op;
525
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100526 BUG_ON(!(inode->i_state & I_FREEING));
527 BUG_ON(!list_empty(&inode->i_lru));
528
Eric Dumazetb12362b2011-07-28 06:11:47 +0200529 if (!list_empty(&inode->i_wb_list))
530 inode_wb_list_del(inode);
531
Dave Chinner55fa6092011-03-22 22:23:40 +1100532 inode_sb_list_del(inode);
533
Al Virobe7ce412010-06-04 19:40:39 -0400534 if (op->evict_inode) {
535 op->evict_inode(inode);
Al Virob4272d42010-06-04 19:33:20 -0400536 } else {
537 if (inode->i_data.nrpages)
538 truncate_inode_pages(&inode->i_data, 0);
Al Viro30140832010-06-07 13:23:20 -0400539 end_writeback(inode);
Al Virob4272d42010-06-04 19:33:20 -0400540 }
Al Viro661074e2010-06-04 20:19:55 -0400541 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
542 bd_forget(inode);
543 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
544 cd_forget(inode);
Dave Chinnerb2b2af82011-03-22 22:23:37 +1100545
546 remove_inode_hash(inode);
547
548 spin_lock(&inode->i_lock);
549 wake_up_bit(&inode->i_state, __I_NEW);
550 BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
551 spin_unlock(&inode->i_lock);
552
553 destroy_inode(inode);
Al Virob4272d42010-06-04 19:33:20 -0400554}
555
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556/*
557 * dispose_list - dispose of the contents of a local list
558 * @head: the head of the list to free
559 *
560 * Dispose-list gets a local list with local inodes in it, so it doesn't
561 * need to worry about list corruption and SMP locks.
562 */
563static void dispose_list(struct list_head *head)
564{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 while (!list_empty(head)) {
566 struct inode *inode;
567
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100568 inode = list_first_entry(head, struct inode, i_lru);
569 list_del_init(&inode->i_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
Al Viro644da592010-06-07 13:21:05 -0400571 evict(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573}
574
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575/**
Al Viro63997e92010-10-25 20:49:35 -0400576 * evict_inodes - evict all evictable inodes for a superblock
577 * @sb: superblock to operate on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 *
Al Viro63997e92010-10-25 20:49:35 -0400579 * Make sure that no inodes with zero refcount are retained. This is
580 * called by superblock shutdown after having MS_ACTIVE flag removed,
581 * so any inode reaching zero refcount during or after that call will
582 * be immediately evicted.
583 */
584void evict_inodes(struct super_block *sb)
585{
586 struct inode *inode, *next;
587 LIST_HEAD(dispose);
588
Dave Chinner55fa6092011-03-22 22:23:40 +1100589 spin_lock(&inode_sb_list_lock);
Al Viro63997e92010-10-25 20:49:35 -0400590 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
591 if (atomic_read(&inode->i_count))
592 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100593
594 spin_lock(&inode->i_lock);
595 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
596 spin_unlock(&inode->i_lock);
Al Viro63997e92010-10-25 20:49:35 -0400597 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100598 }
Al Viro63997e92010-10-25 20:49:35 -0400599
600 inode->i_state |= I_FREEING;
Dave Chinner02afc412011-03-22 22:23:38 +1100601 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100602 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100603 list_add(&inode->i_lru, &dispose);
Al Viro63997e92010-10-25 20:49:35 -0400604 }
Dave Chinner55fa6092011-03-22 22:23:40 +1100605 spin_unlock(&inode_sb_list_lock);
Al Viro63997e92010-10-25 20:49:35 -0400606
607 dispose_list(&dispose);
Al Viro63997e92010-10-25 20:49:35 -0400608}
609
610/**
Christoph Hellwiga0318782010-10-24 19:40:33 +0200611 * invalidate_inodes - attempt to free all inodes on a superblock
612 * @sb: superblock to operate on
NeilBrown93b270f2011-02-24 17:25:47 +1100613 * @kill_dirty: flag to guide handling of dirty inodes
Christoph Hellwiga0318782010-10-24 19:40:33 +0200614 *
615 * Attempts to free all inodes for a given superblock. If there were any
616 * busy inodes return a non-zero value, else zero.
NeilBrown93b270f2011-02-24 17:25:47 +1100617 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
618 * them as busy.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 */
NeilBrown93b270f2011-02-24 17:25:47 +1100620int invalidate_inodes(struct super_block *sb, bool kill_dirty)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621{
Dave Chinnercffbc8a2010-10-23 05:03:02 -0400622 int busy = 0;
Christoph Hellwiga0318782010-10-24 19:40:33 +0200623 struct inode *inode, *next;
624 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
Dave Chinner55fa6092011-03-22 22:23:40 +1100626 spin_lock(&inode_sb_list_lock);
Christoph Hellwiga0318782010-10-24 19:40:33 +0200627 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100628 spin_lock(&inode->i_lock);
629 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
630 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100632 }
NeilBrown93b270f2011-02-24 17:25:47 +1100633 if (inode->i_state & I_DIRTY && !kill_dirty) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100634 spin_unlock(&inode->i_lock);
NeilBrown93b270f2011-02-24 17:25:47 +1100635 busy = 1;
636 continue;
637 }
Christoph Hellwig99a38912010-10-23 19:07:20 +0200638 if (atomic_read(&inode->i_count)) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100639 spin_unlock(&inode->i_lock);
Christoph Hellwig99a38912010-10-23 19:07:20 +0200640 busy = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 continue;
642 }
Christoph Hellwig99a38912010-10-23 19:07:20 +0200643
Christoph Hellwig99a38912010-10-23 19:07:20 +0200644 inode->i_state |= I_FREEING;
Dave Chinner02afc412011-03-22 22:23:38 +1100645 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100646 spin_unlock(&inode->i_lock);
Dave Chinner02afc412011-03-22 22:23:38 +1100647 list_add(&inode->i_lru, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 }
Dave Chinner55fa6092011-03-22 22:23:40 +1100649 spin_unlock(&inode_sb_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Christoph Hellwiga0318782010-10-24 19:40:33 +0200651 dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653 return busy;
654}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656static int can_unuse(struct inode *inode)
657{
Nick Piggin9e38d862010-10-23 06:55:17 -0400658 if (inode->i_state & ~I_REFERENCED)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 return 0;
660 if (inode_has_buffers(inode))
661 return 0;
662 if (atomic_read(&inode->i_count))
663 return 0;
664 if (inode->i_data.nrpages)
665 return 0;
666 return 1;
667}
668
669/*
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000670 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
671 * This is called from the superblock shrinker function with a number of inodes
672 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
673 * then are freed outside inode_lock by dispose_list().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 *
675 * Any inodes which are pinned purely because of attached pagecache have their
Nick Piggin9e38d862010-10-23 06:55:17 -0400676 * pagecache removed. If the inode has metadata buffers attached to
677 * mapping->private_list then try to remove them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 *
Nick Piggin9e38d862010-10-23 06:55:17 -0400679 * If the inode has the I_REFERENCED flag set, then it means that it has been
680 * used recently - the flag is set in iput_final(). When we encounter such an
681 * inode, clear the flag and move it to the back of the LRU so it gets another
682 * pass through the LRU before it gets reclaimed. This is necessary because of
683 * the fact we are doing lazy LRU updates to minimise lock contention so the
684 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
685 * with this flag set because they are the inodes that are out of order.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 */
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000687void prune_icache_sb(struct super_block *sb, int nr_to_scan)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688{
689 LIST_HEAD(freeable);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 int nr_scanned;
691 unsigned long reap = 0;
692
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000693 spin_lock(&sb->s_inode_lru_lock);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000694 for (nr_scanned = nr_to_scan; nr_scanned >= 0; nr_scanned--) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 struct inode *inode;
696
Dave Chinner98b745c2011-07-08 14:14:39 +1000697 if (list_empty(&sb->s_inode_lru))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 break;
699
Dave Chinner98b745c2011-07-08 14:14:39 +1000700 inode = list_entry(sb->s_inode_lru.prev, struct inode, i_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
Nick Piggin9e38d862010-10-23 06:55:17 -0400702 /*
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000703 * we are inverting the sb->s_inode_lru_lock/inode->i_lock here,
Dave Chinner02afc412011-03-22 22:23:38 +1100704 * so use a trylock. If we fail to get the lock, just move the
705 * inode to the back of the list so we don't spin on it.
706 */
707 if (!spin_trylock(&inode->i_lock)) {
Christoph Hellwig62a3dde2011-10-28 10:03:41 +0200708 list_move_tail(&inode->i_lru, &sb->s_inode_lru);
Dave Chinner02afc412011-03-22 22:23:38 +1100709 continue;
710 }
711
712 /*
Nick Piggin9e38d862010-10-23 06:55:17 -0400713 * Referenced or dirty inodes are still in use. Give them
714 * another pass through the LRU as we canot reclaim them now.
715 */
716 if (atomic_read(&inode->i_count) ||
717 (inode->i_state & ~I_REFERENCED)) {
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100718 list_del_init(&inode->i_lru);
Dave Chinnerf283c862011-03-22 22:23:39 +1100719 spin_unlock(&inode->i_lock);
Dave Chinner98b745c2011-07-08 14:14:39 +1000720 sb->s_nr_inodes_unused--;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000721 this_cpu_dec(nr_unused);
Nick Piggin9e38d862010-10-23 06:55:17 -0400722 continue;
723 }
724
725 /* recently referenced inodes get one more pass */
726 if (inode->i_state & I_REFERENCED) {
Nick Piggin9e38d862010-10-23 06:55:17 -0400727 inode->i_state &= ~I_REFERENCED;
Dave Chinner98b745c2011-07-08 14:14:39 +1000728 list_move(&inode->i_lru, &sb->s_inode_lru);
Dave Chinnerf283c862011-03-22 22:23:39 +1100729 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 continue;
731 }
732 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
733 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100734 spin_unlock(&inode->i_lock);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000735 spin_unlock(&sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 if (remove_inode_buffers(inode))
Andrew Mortonfc0ecff2007-02-10 01:45:39 -0800737 reap += invalidate_mapping_pages(&inode->i_data,
738 0, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 iput(inode);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000740 spin_lock(&sb->s_inode_lru_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
Dave Chinner98b745c2011-07-08 14:14:39 +1000742 if (inode != list_entry(sb->s_inode_lru.next,
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100743 struct inode, i_lru))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 continue; /* wrong inode or list_empty */
Dave Chinner02afc412011-03-22 22:23:38 +1100745 /* avoid lock inversions with trylock */
746 if (!spin_trylock(&inode->i_lock))
747 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100748 if (!can_unuse(inode)) {
749 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +1100751 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 }
Nick Piggin7ef0d732009-03-12 14:31:38 -0700753 WARN_ON(inode->i_state & I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 inode->i_state |= I_FREEING;
Dave Chinner250df6e2011-03-22 22:23:36 +1100755 spin_unlock(&inode->i_lock);
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100756
Nick Piggin7ccf19a2010-10-21 11:49:30 +1100757 list_move(&inode->i_lru, &freeable);
Dave Chinner98b745c2011-07-08 14:14:39 +1000758 sb->s_nr_inodes_unused--;
Dave Chinnerfcb94f72011-07-08 14:14:38 +1000759 this_cpu_dec(nr_unused);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 }
Christoph Lameterf8891e52006-06-30 01:55:45 -0700761 if (current_is_kswapd())
762 __count_vm_events(KSWAPD_INODESTEAL, reap);
763 else
764 __count_vm_events(PGINODESTEAL, reap);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000765 spin_unlock(&sb->s_inode_lru_lock);
Konstantin Khlebnikov5f8aefd2012-01-10 15:07:18 -0800766 if (current->reclaim_state)
767 current->reclaim_state->reclaimed_slab += reap;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769 dispose_list(&freeable);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770}
771
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772static void __wait_on_freeing_inode(struct inode *inode);
773/*
774 * Called with the inode lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530776static struct inode *find_inode(struct super_block *sb,
777 struct hlist_head *head,
778 int (*test)(struct inode *, void *),
779 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780{
781 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530782 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
784repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700785 hlist_for_each_entry(inode, node, head, i_hash) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100786 spin_lock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100787 if (inode->i_sb != sb) {
788 spin_unlock(&inode->i_lock);
789 continue;
790 }
791 if (!test(inode, data)) {
792 spin_unlock(&inode->i_lock);
793 continue;
794 }
Al Viroa4ffdde2010-06-02 17:38:30 -0400795 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 __wait_on_freeing_inode(inode);
797 goto repeat;
798 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400799 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100800 spin_unlock(&inode->i_lock);
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400801 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400803 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804}
805
806/*
807 * find_inode_fast is the fast path version of find_inode, see the comment at
808 * iget_locked for details.
809 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530810static struct inode *find_inode_fast(struct super_block *sb,
811 struct hlist_head *head, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
813 struct hlist_node *node;
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530814 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815
816repeat:
Matthias Kaehlckec5c8be32008-04-29 00:59:40 -0700817 hlist_for_each_entry(inode, node, head, i_hash) {
Dave Chinner250df6e2011-03-22 22:23:36 +1100818 spin_lock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +1100819 if (inode->i_ino != ino) {
820 spin_unlock(&inode->i_lock);
821 continue;
822 }
823 if (inode->i_sb != sb) {
824 spin_unlock(&inode->i_lock);
825 continue;
826 }
Al Viroa4ffdde2010-06-02 17:38:30 -0400827 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 __wait_on_freeing_inode(inode);
829 goto repeat;
830 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400831 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100832 spin_unlock(&inode->i_lock);
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400833 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 }
Christoph Hellwigf7899bd2010-10-23 07:09:06 -0400835 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836}
837
Eric Dumazetf991bd22010-10-23 11:18:01 -0400838/*
839 * Each cpu owns a range of LAST_INO_BATCH numbers.
840 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
841 * to renew the exhausted range.
David Chinner8290c352008-10-30 17:35:24 +1100842 *
Eric Dumazetf991bd22010-10-23 11:18:01 -0400843 * This does not significantly increase overflow rate because every CPU can
844 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
845 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
846 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
847 * overflow rate by 2x, which does not seem too significant.
848 *
849 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
850 * error if st_ino won't fit in target struct field. Use 32bit counter
851 * here to attempt to avoid that.
David Chinner8290c352008-10-30 17:35:24 +1100852 */
Eric Dumazetf991bd22010-10-23 11:18:01 -0400853#define LAST_INO_BATCH 1024
854static DEFINE_PER_CPU(unsigned int, last_ino);
David Chinner8290c352008-10-30 17:35:24 +1100855
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400856unsigned int get_next_ino(void)
Eric Dumazetf991bd22010-10-23 11:18:01 -0400857{
858 unsigned int *p = &get_cpu_var(last_ino);
859 unsigned int res = *p;
860
861#ifdef CONFIG_SMP
862 if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
863 static atomic_t shared_last_ino;
864 int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
865
866 res = next - LAST_INO_BATCH;
867 }
868#endif
869
870 *p = ++res;
871 put_cpu_var(last_ino);
872 return res;
David Chinner8290c352008-10-30 17:35:24 +1100873}
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400874EXPORT_SYMBOL(get_next_ino);
David Chinner8290c352008-10-30 17:35:24 +1100875
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876/**
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200877 * new_inode_pseudo - obtain an inode
878 * @sb: superblock
879 *
880 * Allocates a new inode for given superblock.
881 * Inode wont be chained in superblock s_inodes list
882 * This means :
883 * - fs can't be unmount
884 * - quotas, fsnotify, writeback can't work
885 */
886struct inode *new_inode_pseudo(struct super_block *sb)
887{
888 struct inode *inode = alloc_inode(sb);
889
890 if (inode) {
891 spin_lock(&inode->i_lock);
892 inode->i_state = 0;
893 spin_unlock(&inode->i_lock);
894 INIT_LIST_HEAD(&inode->i_sb_list);
895 }
896 return inode;
897}
898
899/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 * new_inode - obtain an inode
901 * @sb: superblock
902 *
Mel Gorman769848c2007-07-17 04:03:05 -0700903 * Allocates a new inode for given superblock. The default gfp_mask
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800904 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
Mel Gorman769848c2007-07-17 04:03:05 -0700905 * If HIGHMEM pages are unsuitable or it is known that pages allocated
906 * for the page cache are not reclaimable or migratable,
907 * mapping_set_gfp_mask() must be called with suitable flags on the
908 * newly created inode's mapping
909 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 */
911struct inode *new_inode(struct super_block *sb)
912{
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530913 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
Dave Chinner55fa6092011-03-22 22:23:40 +1100915 spin_lock_prefetch(&inode_sb_list_lock);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530916
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200917 inode = new_inode_pseudo(sb);
918 if (inode)
Dave Chinner55fa6092011-03-22 22:23:40 +1100919 inode_sb_list_add(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 return inode;
921}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922EXPORT_SYMBOL(new_inode);
923
Peter Zijlstra14358e62007-10-14 01:38:33 +0200924#ifdef CONFIG_DEBUG_LOCK_ALLOC
Josh Boyere096d0c2011-08-25 07:48:12 -0400925void lockdep_annotate_inode_mutex_key(struct inode *inode)
926{
Namhyung Kima3314a02010-10-11 22:38:00 +0900927 if (S_ISDIR(inode->i_mode)) {
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200928 struct file_system_type *type = inode->i_sb->s_type;
929
Jan Kara9a7aa122009-06-04 15:26:49 +0200930 /* Set new key only if filesystem hasn't already changed it */
Tyler Hicks978d6d82011-12-12 10:02:30 -0600931 if (lockdep_match_class(&inode->i_mutex, &type->i_mutex_key)) {
Jan Kara9a7aa122009-06-04 15:26:49 +0200932 /*
933 * ensure nobody is actually holding i_mutex
934 */
935 mutex_destroy(&inode->i_mutex);
936 mutex_init(&inode->i_mutex);
937 lockdep_set_class(&inode->i_mutex,
938 &type->i_mutex_dir_key);
939 }
Peter Zijlstra1e89a5e2007-10-16 06:47:54 +0200940 }
Josh Boyere096d0c2011-08-25 07:48:12 -0400941}
942EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key);
Peter Zijlstra14358e62007-10-14 01:38:33 +0200943#endif
Josh Boyere096d0c2011-08-25 07:48:12 -0400944
945/**
946 * unlock_new_inode - clear the I_NEW state and wake up any waiters
947 * @inode: new inode to unlock
948 *
949 * Called when the inode is fully initialised to clear the new state of the
950 * inode and wake up anyone waiting for the inode to finish initialisation.
951 */
952void unlock_new_inode(struct inode *inode)
953{
954 lockdep_annotate_inode_mutex_key(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +1100955 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +0100956 WARN_ON(!(inode->i_state & I_NEW));
957 inode->i_state &= ~I_NEW;
Al Viro310fa7a2012-03-10 17:07:28 -0500958 smp_mb();
Dave Chinner250df6e2011-03-22 22:23:36 +1100959 wake_up_bit(&inode->i_state, __I_NEW);
960 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962EXPORT_SYMBOL(unlock_new_inode);
963
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400964/**
965 * iget5_locked - obtain an inode from a mounted file system
966 * @sb: super block of file system
967 * @hashval: hash value (usually inode number) to get
968 * @test: callback used for comparisons between inodes
969 * @set: callback used to initialize a new struct inode
970 * @data: opaque data pointer to pass to @test and @set
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400972 * Search for the inode specified by @hashval and @data in the inode cache,
973 * and if present it is return it with an increased reference count. This is
974 * a generalized version of iget_locked() for file systems where the inode
975 * number is not sufficient for unique identification of an inode.
976 *
977 * If the inode is not in cache, allocate a new inode and return it locked,
978 * hashed, and with the I_NEW flag set. The file system gets to fill it in
979 * before unlocking it via unlock_new_inode().
980 *
981 * Note both @test and @set are called with the inode_hash_lock held, so can't
982 * sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 */
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400984struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
985 int (*test)(struct inode *, void *),
986 int (*set)(struct inode *, void *), void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400988 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Manish Katiyar6b3304b2009-03-31 19:35:54 +0530989 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Christoph Hellwig0b2d0722011-03-23 15:03:28 -0400991 spin_lock(&inode_hash_lock);
992 inode = find_inode(sb, head, test, data);
993 spin_unlock(&inode_hash_lock);
994
995 if (inode) {
996 wait_on_inode(inode);
997 return inode;
998 }
999
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 inode = alloc_inode(sb);
1001 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301002 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Dave Chinner67a23c42011-03-22 22:23:42 +11001004 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 /* We released the lock, so.. */
1006 old = find_inode(sb, head, test, data);
1007 if (!old) {
1008 if (set(inode, data))
1009 goto set_failed;
1010
Dave Chinner250df6e2011-03-22 22:23:36 +11001011 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001012 inode->i_state = I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +11001013 hlist_add_head(&inode->i_hash, head);
1014 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +11001015 inode_sb_list_add(inode);
Dave Chinner67a23c42011-03-22 22:23:42 +11001016 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
1018 /* Return the locked inode with I_NEW set, the
1019 * caller is responsible for filling in the contents
1020 */
1021 return inode;
1022 }
1023
1024 /*
1025 * Uhhuh, somebody else created the same inode under
1026 * us. Use the old inode instead of the one we just
1027 * allocated.
1028 */
Dave Chinner67a23c42011-03-22 22:23:42 +11001029 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 destroy_inode(inode);
1031 inode = old;
1032 wait_on_inode(inode);
1033 }
1034 return inode;
1035
1036set_failed:
Dave Chinner67a23c42011-03-22 22:23:42 +11001037 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 destroy_inode(inode);
1039 return NULL;
1040}
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001041EXPORT_SYMBOL(iget5_locked);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001043/**
1044 * iget_locked - obtain an inode from a mounted file system
1045 * @sb: super block of file system
1046 * @ino: inode number to get
1047 *
1048 * Search for the inode specified by @ino in the inode cache and if present
1049 * return it with an increased reference count. This is for file systems
1050 * where the inode number is sufficient for unique identification of an inode.
1051 *
1052 * If the inode is not in cache, allocate a new inode and return it locked,
1053 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1054 * before unlocking it via unlock_new_inode().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 */
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001056struct inode *iget_locked(struct super_block *sb, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001058 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301059 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001061 spin_lock(&inode_hash_lock);
1062 inode = find_inode_fast(sb, head, ino);
1063 spin_unlock(&inode_hash_lock);
1064 if (inode) {
1065 wait_on_inode(inode);
1066 return inode;
1067 }
1068
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 inode = alloc_inode(sb);
1070 if (inode) {
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301071 struct inode *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072
Dave Chinner67a23c42011-03-22 22:23:42 +11001073 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 /* We released the lock, so.. */
1075 old = find_inode_fast(sb, head, ino);
1076 if (!old) {
1077 inode->i_ino = ino;
Dave Chinner250df6e2011-03-22 22:23:36 +11001078 spin_lock(&inode->i_lock);
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001079 inode->i_state = I_NEW;
Dave Chinner250df6e2011-03-22 22:23:36 +11001080 hlist_add_head(&inode->i_hash, head);
1081 spin_unlock(&inode->i_lock);
Dave Chinner55fa6092011-03-22 22:23:40 +11001082 inode_sb_list_add(inode);
Dave Chinner67a23c42011-03-22 22:23:42 +11001083 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084
1085 /* Return the locked inode with I_NEW set, the
1086 * caller is responsible for filling in the contents
1087 */
1088 return inode;
1089 }
1090
1091 /*
1092 * Uhhuh, somebody else created the same inode under
1093 * us. Use the old inode instead of the one we just
1094 * allocated.
1095 */
Dave Chinner67a23c42011-03-22 22:23:42 +11001096 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 destroy_inode(inode);
1098 inode = old;
1099 wait_on_inode(inode);
1100 }
1101 return inode;
1102}
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001103EXPORT_SYMBOL(iget_locked);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001105/*
1106 * search the inode cache for a matching inode number.
1107 * If we find one, then the inode number we are trying to
1108 * allocate is not unique and so we should not use it.
1109 *
1110 * Returns 1 if the inode number is unique, 0 if it is not.
1111 */
1112static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1113{
1114 struct hlist_head *b = inode_hashtable + hash(sb, ino);
1115 struct hlist_node *node;
1116 struct inode *inode;
1117
Dave Chinner67a23c42011-03-22 22:23:42 +11001118 spin_lock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001119 hlist_for_each_entry(inode, node, b, i_hash) {
Dave Chinner67a23c42011-03-22 22:23:42 +11001120 if (inode->i_ino == ino && inode->i_sb == sb) {
1121 spin_unlock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001122 return 0;
Dave Chinner67a23c42011-03-22 22:23:42 +11001123 }
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001124 }
Dave Chinner67a23c42011-03-22 22:23:42 +11001125 spin_unlock(&inode_hash_lock);
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001126
1127 return 1;
1128}
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130/**
1131 * iunique - get a unique inode number
1132 * @sb: superblock
1133 * @max_reserved: highest reserved inode number
1134 *
1135 * Obtain an inode number that is unique on the system for a given
1136 * superblock. This is used by file systems that have no natural
1137 * permanent inode numbering system. An inode number is returned that
1138 * is higher than the reserved limit but unique.
1139 *
1140 * BUGS:
1141 * With a large number of inodes live on the file system this function
1142 * currently becomes quite slow.
1143 */
1144ino_t iunique(struct super_block *sb, ino_t max_reserved)
1145{
Jeff Layton866b04f2007-05-08 00:32:29 -07001146 /*
1147 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1148 * error if st_ino won't fit in target struct field. Use 32bit counter
1149 * here to attempt to avoid that.
1150 */
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001151 static DEFINE_SPINLOCK(iunique_lock);
Jeff Layton866b04f2007-05-08 00:32:29 -07001152 static unsigned int counter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 ino_t res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001155 spin_lock(&iunique_lock);
Jeffrey Layton3361c7b2007-05-08 00:29:48 -07001156 do {
1157 if (counter <= max_reserved)
1158 counter = max_reserved + 1;
1159 res = counter++;
Christoph Hellwigad5e1952010-10-23 07:00:16 -04001160 } while (!test_inode_iunique(sb, res));
1161 spin_unlock(&iunique_lock);
Jeffrey Layton3361c7b2007-05-08 00:29:48 -07001162
1163 return res;
1164}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165EXPORT_SYMBOL(iunique);
1166
1167struct inode *igrab(struct inode *inode)
1168{
Dave Chinner250df6e2011-03-22 22:23:36 +11001169 spin_lock(&inode->i_lock);
1170 if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 __iget(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +11001172 spin_unlock(&inode->i_lock);
1173 } else {
1174 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 /*
1176 * Handle the case where s_op->clear_inode is not been
1177 * called yet, and somebody is calling igrab
1178 * while the inode is getting freed.
1179 */
1180 inode = NULL;
Dave Chinner250df6e2011-03-22 22:23:36 +11001181 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 return inode;
1183}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184EXPORT_SYMBOL(igrab);
1185
1186/**
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001187 * ilookup5_nowait - search for an inode in the inode cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 * @sb: super block of file system to search
1189 * @hashval: hash value (usually inode number) to search for
1190 * @test: callback used for comparisons between inodes
1191 * @data: opaque data pointer to pass to @test
1192 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001193 * Search for the inode specified by @hashval and @data in the inode cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 * If the inode is in the cache, the inode is returned with an incremented
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001195 * reference count.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001196 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001197 * Note: I_NEW is not waited upon so you have to be very careful what you do
1198 * with the returned inode. You probably should be using ilookup5() instead.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001199 *
Randy Dunlapb6d0ad62011-03-26 13:27:47 -07001200 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001201 */
1202struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1203 int (*test)(struct inode *, void *), void *data)
1204{
1205 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001206 struct inode *inode;
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001207
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001208 spin_lock(&inode_hash_lock);
1209 inode = find_inode(sb, head, test, data);
1210 spin_unlock(&inode_hash_lock);
1211
1212 return inode;
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001213}
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001214EXPORT_SYMBOL(ilookup5_nowait);
1215
1216/**
1217 * ilookup5 - search for an inode in the inode cache
1218 * @sb: super block of file system to search
1219 * @hashval: hash value (usually inode number) to search for
1220 * @test: callback used for comparisons between inodes
1221 * @data: opaque data pointer to pass to @test
1222 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001223 * Search for the inode specified by @hashval and @data in the inode cache,
1224 * and if the inode is in the cache, return the inode with an incremented
1225 * reference count. Waits on I_NEW before returning the inode.
Anton Altaparmakov88bd5122005-07-13 01:10:44 -07001226 * returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001228 * This is a generalized version of ilookup() for file systems where the
1229 * inode number is not sufficient for unique identification of an inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001231 * Note: @test is called with the inode_hash_lock held, so can't sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 */
1233struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1234 int (*test)(struct inode *, void *), void *data)
1235{
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001236 struct inode *inode = ilookup5_nowait(sb, hashval, test, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001238 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(ilookup5);
1243
1244/**
1245 * ilookup - search for an inode in the inode cache
1246 * @sb: super block of file system to search
1247 * @ino: inode number to search for
1248 *
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001249 * Search for the inode @ino in the inode cache, and if the inode is in the
1250 * cache, the inode is returned with an incremented reference count.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 */
1252struct inode *ilookup(struct super_block *sb, unsigned long ino)
1253{
1254 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001255 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256
Christoph Hellwig0b2d0722011-03-23 15:03:28 -04001257 spin_lock(&inode_hash_lock);
1258 inode = find_inode_fast(sb, head, ino);
1259 spin_unlock(&inode_hash_lock);
1260
1261 if (inode)
1262 wait_on_inode(inode);
1263 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265EXPORT_SYMBOL(ilookup);
1266
Al Viro261bca82008-12-30 01:48:21 -05001267int insert_inode_locked(struct inode *inode)
1268{
1269 struct super_block *sb = inode->i_sb;
1270 ino_t ino = inode->i_ino;
1271 struct hlist_head *head = inode_hashtable + hash(sb, ino);
Al Viro261bca82008-12-30 01:48:21 -05001272
Al Viro261bca82008-12-30 01:48:21 -05001273 while (1) {
Al Viro72a43d62009-05-13 19:13:40 +01001274 struct hlist_node *node;
1275 struct inode *old = NULL;
Dave Chinner67a23c42011-03-22 22:23:42 +11001276 spin_lock(&inode_hash_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001277 hlist_for_each_entry(old, node, head, i_hash) {
1278 if (old->i_ino != ino)
1279 continue;
1280 if (old->i_sb != sb)
1281 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001282 spin_lock(&old->i_lock);
1283 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1284 spin_unlock(&old->i_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001285 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001286 }
Al Viro72a43d62009-05-13 19:13:40 +01001287 break;
1288 }
1289 if (likely(!node)) {
Dave Chinner250df6e2011-03-22 22:23:36 +11001290 spin_lock(&inode->i_lock);
1291 inode->i_state |= I_NEW;
Al Viro261bca82008-12-30 01:48:21 -05001292 hlist_add_head(&inode->i_hash, head);
Dave Chinner250df6e2011-03-22 22:23:36 +11001293 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001294 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001295 return 0;
1296 }
1297 __iget(old);
Dave Chinner250df6e2011-03-22 22:23:36 +11001298 spin_unlock(&old->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001299 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001300 wait_on_inode(old);
Al Viro1d3382cb2010-10-23 15:19:20 -04001301 if (unlikely(!inode_unhashed(old))) {
Al Viro261bca82008-12-30 01:48:21 -05001302 iput(old);
1303 return -EBUSY;
1304 }
1305 iput(old);
1306 }
1307}
Al Viro261bca82008-12-30 01:48:21 -05001308EXPORT_SYMBOL(insert_inode_locked);
1309
1310int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1311 int (*test)(struct inode *, void *), void *data)
1312{
1313 struct super_block *sb = inode->i_sb;
1314 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
Al Viro261bca82008-12-30 01:48:21 -05001315
Al Viro261bca82008-12-30 01:48:21 -05001316 while (1) {
Al Viro72a43d62009-05-13 19:13:40 +01001317 struct hlist_node *node;
1318 struct inode *old = NULL;
1319
Dave Chinner67a23c42011-03-22 22:23:42 +11001320 spin_lock(&inode_hash_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001321 hlist_for_each_entry(old, node, head, i_hash) {
1322 if (old->i_sb != sb)
1323 continue;
1324 if (!test(old, data))
1325 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001326 spin_lock(&old->i_lock);
1327 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1328 spin_unlock(&old->i_lock);
Al Viro72a43d62009-05-13 19:13:40 +01001329 continue;
Dave Chinner250df6e2011-03-22 22:23:36 +11001330 }
Al Viro72a43d62009-05-13 19:13:40 +01001331 break;
1332 }
1333 if (likely(!node)) {
Dave Chinner250df6e2011-03-22 22:23:36 +11001334 spin_lock(&inode->i_lock);
1335 inode->i_state |= I_NEW;
Al Viro261bca82008-12-30 01:48:21 -05001336 hlist_add_head(&inode->i_hash, head);
Dave Chinner250df6e2011-03-22 22:23:36 +11001337 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001338 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001339 return 0;
1340 }
1341 __iget(old);
Dave Chinner250df6e2011-03-22 22:23:36 +11001342 spin_unlock(&old->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001343 spin_unlock(&inode_hash_lock);
Al Viro261bca82008-12-30 01:48:21 -05001344 wait_on_inode(old);
Al Viro1d3382cb2010-10-23 15:19:20 -04001345 if (unlikely(!inode_unhashed(old))) {
Al Viro261bca82008-12-30 01:48:21 -05001346 iput(old);
1347 return -EBUSY;
1348 }
1349 iput(old);
1350 }
1351}
Al Viro261bca82008-12-30 01:48:21 -05001352EXPORT_SYMBOL(insert_inode_locked4);
1353
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
Al Viro45321ac2010-06-07 13:43:19 -04001355int generic_delete_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356{
Al Viro45321ac2010-06-07 13:43:19 -04001357 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359EXPORT_SYMBOL(generic_delete_inode);
1360
Al Viro45321ac2010-06-07 13:43:19 -04001361/*
Al Viro45321ac2010-06-07 13:43:19 -04001362 * Called when we're dropping the last reference
1363 * to an inode.
1364 *
1365 * Call the FS "drop_inode()" function, defaulting to
1366 * the legacy UNIX filesystem behaviour. If it tells
1367 * us to evict inode, do so. Otherwise, retain inode
1368 * in cache if fs is alive, sync and evict if fs is
1369 * shutting down.
1370 */
1371static void iput_final(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372{
1373 struct super_block *sb = inode->i_sb;
Al Viro45321ac2010-06-07 13:43:19 -04001374 const struct super_operations *op = inode->i_sb->s_op;
1375 int drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
Dave Chinner250df6e2011-03-22 22:23:36 +11001377 WARN_ON(inode->i_state & I_NEW);
1378
Al Viroe7f59092011-07-07 15:45:59 -04001379 if (op->drop_inode)
Al Viro45321ac2010-06-07 13:43:19 -04001380 drop = op->drop_inode(inode);
1381 else
1382 drop = generic_drop_inode(inode);
1383
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001384 if (!drop && (sb->s_flags & MS_ACTIVE)) {
1385 inode->i_state |= I_REFERENCED;
1386 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
1387 inode_lru_list_add(inode);
1388 spin_unlock(&inode->i_lock);
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001389 return;
1390 }
1391
Al Viro45321ac2010-06-07 13:43:19 -04001392 if (!drop) {
Alexander Viro991114c2005-06-23 00:09:01 -07001393 inode->i_state |= I_WILL_FREE;
Dave Chinner250df6e2011-03-22 22:23:36 +11001394 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 write_inode_now(inode, 1);
Dave Chinner250df6e2011-03-22 22:23:36 +11001396 spin_lock(&inode->i_lock);
Nick Piggin7ef0d732009-03-12 14:31:38 -07001397 WARN_ON(inode->i_state & I_NEW);
Alexander Viro991114c2005-06-23 00:09:01 -07001398 inode->i_state &= ~I_WILL_FREE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 }
Nick Piggin7ccf19a2010-10-21 11:49:30 +11001400
Alexander Viro991114c2005-06-23 00:09:01 -07001401 inode->i_state |= I_FREEING;
Eric Dumazetc4ae0c62011-07-28 06:55:13 +02001402 if (!list_empty(&inode->i_lru))
1403 inode_lru_list_del(inode);
Dave Chinner250df6e2011-03-22 22:23:36 +11001404 spin_unlock(&inode->i_lock);
Dave Chinnerb2b2af82011-03-22 22:23:37 +11001405
1406 evict(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407}
1408
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409/**
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301410 * iput - put an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 * @inode: inode to put
1412 *
1413 * Puts an inode, dropping its usage count. If the inode use count hits
1414 * zero, the inode is then freed and may also be destroyed.
1415 *
1416 * Consequently, iput() can sleep.
1417 */
1418void iput(struct inode *inode)
1419{
1420 if (inode) {
Al Viroa4ffdde2010-06-02 17:38:30 -04001421 BUG_ON(inode->i_state & I_CLEAR);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422
Dave Chinnerf283c862011-03-22 22:23:39 +11001423 if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 iput_final(inode);
1425 }
1426}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427EXPORT_SYMBOL(iput);
1428
1429/**
1430 * bmap - find a block number in a file
1431 * @inode: inode of file
1432 * @block: block to find
1433 *
1434 * Returns the block number on the device holding the inode that
1435 * is the disk block number for the block of the file requested.
1436 * That is, asked for block 4 of inode 1 the function will return the
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301437 * disk block relative to the disk start that holds that block of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 * file.
1439 */
Manish Katiyar6b3304b2009-03-31 19:35:54 +05301440sector_t bmap(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441{
1442 sector_t res = 0;
1443 if (inode->i_mapping->a_ops->bmap)
1444 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1445 return res;
1446}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447EXPORT_SYMBOL(bmap);
1448
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001449/*
1450 * With relative atime, only update atime if the previous atime is
1451 * earlier than either the ctime or mtime or if at least a day has
1452 * passed since the last atime update.
1453 */
1454static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1455 struct timespec now)
1456{
1457
1458 if (!(mnt->mnt_flags & MNT_RELATIME))
1459 return 1;
1460 /*
1461 * Is mtime younger than atime? If yes, update atime:
1462 */
1463 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1464 return 1;
1465 /*
1466 * Is ctime younger than atime? If yes, update atime:
1467 */
1468 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1469 return 1;
1470
1471 /*
1472 * Is the previous atime value older than a day? If yes,
1473 * update atime:
1474 */
1475 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1476 return 1;
1477 /*
1478 * Good, we can skip the atime update:
1479 */
1480 return 0;
1481}
1482
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483/**
Christoph Hellwig869243a2006-01-09 20:52:03 -08001484 * touch_atime - update the access time
1485 * @mnt: mount the inode is accessed on
Martin Waitz7045f372006-02-01 03:06:57 -08001486 * @dentry: dentry accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 *
1488 * Update the accessed time on an inode and mark it for writeback.
1489 * This function automatically handles read only file systems and media,
1490 * as well as the "noatime" flag and inode specific "noatime" markers.
1491 */
Al Viro68ac1232012-03-15 08:21:57 -04001492void touch_atime(struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493{
Al Viro68ac1232012-03-15 08:21:57 -04001494 struct vfsmount *mnt = path->mnt;
1495 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 struct timespec now;
1497
Andrew Mortonb2276132006-12-13 00:34:33 -08001498 if (inode->i_flags & S_NOATIME)
Andi Kleenb12536c2009-09-18 13:05:47 -07001499 return;
Eric Dumazet37756ce2007-02-10 01:44:49 -08001500 if (IS_NOATIME(inode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001501 return;
Andrew Mortonb2276132006-12-13 00:34:33 -08001502 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001503 return;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001504
Dave Hansencdb70f32008-02-15 14:37:41 -08001505 if (mnt->mnt_flags & MNT_NOATIME)
Andi Kleenb12536c2009-09-18 13:05:47 -07001506 return;
Dave Hansencdb70f32008-02-15 14:37:41 -08001507 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
Andi Kleenb12536c2009-09-18 13:05:47 -07001508 return;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001509
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 now = current_fs_time(inode->i_sb);
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001511
1512 if (!relatime_need_update(mnt, inode, now))
Andi Kleenb12536c2009-09-18 13:05:47 -07001513 return;
Matthew Garrett11ff6f052009-03-26 17:32:14 +00001514
Valerie Henson47ae32d2006-12-13 00:34:34 -08001515 if (timespec_equal(&inode->i_atime, &now))
Andi Kleenb12536c2009-09-18 13:05:47 -07001516 return;
1517
1518 if (mnt_want_write(mnt))
1519 return;
Valerie Henson47ae32d2006-12-13 00:34:34 -08001520
1521 inode->i_atime = now;
1522 mark_inode_dirty_sync(inode);
Dave Hansencdb70f32008-02-15 14:37:41 -08001523 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524}
Christoph Hellwig869243a2006-01-09 20:52:03 -08001525EXPORT_SYMBOL(touch_atime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
1527/**
Christoph Hellwig870f4812006-01-09 20:52:01 -08001528 * file_update_time - update mtime and ctime time
1529 * @file: file accessed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 *
Christoph Hellwig870f4812006-01-09 20:52:01 -08001531 * Update the mtime and ctime members of an inode and mark the inode
1532 * for writeback. Note that this function is meant exclusively for
1533 * usage in the file write path of filesystems, and filesystems may
1534 * choose to explicitly ignore update via this function with the
Wolfram Sang2eadfc02009-04-02 15:23:37 +02001535 * S_NOCMTIME inode flag, e.g. for network filesystem where these
Christoph Hellwig870f4812006-01-09 20:52:01 -08001536 * timestamps are handled by the server.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 */
1538
Christoph Hellwig870f4812006-01-09 20:52:01 -08001539void file_update_time(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001541 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 struct timespec now;
Andi Kleence06e0b2009-09-18 13:05:48 -07001543 enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544
Andi Kleence06e0b2009-09-18 13:05:48 -07001545 /* First try to exhaust all avenues to not sync */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 if (IS_NOCMTIME(inode))
1547 return;
Dave Hansen20ddee22008-02-15 14:37:43 -08001548
Andi Kleence06e0b2009-09-18 13:05:48 -07001549 now = current_fs_time(inode->i_sb);
1550 if (!timespec_equal(&inode->i_mtime, &now))
1551 sync_it = S_MTIME;
1552
1553 if (!timespec_equal(&inode->i_ctime, &now))
1554 sync_it |= S_CTIME;
1555
1556 if (IS_I_VERSION(inode))
1557 sync_it |= S_VERSION;
1558
1559 if (!sync_it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 return;
1561
Andi Kleence06e0b2009-09-18 13:05:48 -07001562 /* Finally allowed to write? Takes lock. */
1563 if (mnt_want_write_file(file))
1564 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Andi Kleence06e0b2009-09-18 13:05:48 -07001566 /* Only change inode inside the lock region */
1567 if (sync_it & S_VERSION)
Jean Noel Cordenner7a224222008-01-28 23:58:27 -05001568 inode_inc_iversion(inode);
Andi Kleence06e0b2009-09-18 13:05:48 -07001569 if (sync_it & S_CTIME)
1570 inode->i_ctime = now;
1571 if (sync_it & S_MTIME)
1572 inode->i_mtime = now;
1573 mark_inode_dirty_sync(inode);
Al Viro2a79f172011-12-09 08:06:57 -05001574 mnt_drop_write_file(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575}
Christoph Hellwig870f4812006-01-09 20:52:01 -08001576EXPORT_SYMBOL(file_update_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577
1578int inode_needs_sync(struct inode *inode)
1579{
1580 if (IS_SYNC(inode))
1581 return 1;
1582 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1583 return 1;
1584 return 0;
1585}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586EXPORT_SYMBOL(inode_needs_sync);
1587
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588int inode_wait(void *word)
1589{
1590 schedule();
1591 return 0;
1592}
Stephen Rothwelld44dab82008-11-10 17:06:05 +11001593EXPORT_SYMBOL(inode_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594
1595/*
Miklos Szeredi168a9fd2005-07-12 13:58:10 -07001596 * If we try to find an inode in the inode hash while it is being
1597 * deleted, we have to wait until the filesystem completes its
1598 * deletion before reporting that it isn't found. This function waits
1599 * until the deletion _might_ have completed. Callers are responsible
1600 * to recheck inode state.
1601 *
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001602 * It doesn't matter if I_NEW is not set initially, a call to
Dave Chinner250df6e2011-03-22 22:23:36 +11001603 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1604 * will DTRT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 */
1606static void __wait_on_freeing_inode(struct inode *inode)
1607{
1608 wait_queue_head_t *wq;
Christoph Hellwigeaff8072009-12-17 14:25:01 +01001609 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1610 wq = bit_waitqueue(&inode->i_state, __I_NEW);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
Dave Chinner250df6e2011-03-22 22:23:36 +11001612 spin_unlock(&inode->i_lock);
Dave Chinner67a23c42011-03-22 22:23:42 +11001613 spin_unlock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 schedule();
1615 finish_wait(wq, &wait.wait);
Dave Chinner67a23c42011-03-22 22:23:42 +11001616 spin_lock(&inode_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617}
1618
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619static __initdata unsigned long ihash_entries;
1620static int __init set_ihash_entries(char *str)
1621{
1622 if (!str)
1623 return 0;
1624 ihash_entries = simple_strtoul(str, &str, 0);
1625 return 1;
1626}
1627__setup("ihash_entries=", set_ihash_entries);
1628
1629/*
1630 * Initialize the waitqueues and inode hash table.
1631 */
1632void __init inode_init_early(void)
1633{
Dimitri Sivanich074b8512012-02-08 12:39:07 -08001634 unsigned int loop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635
1636 /* If hashes are distributed across NUMA nodes, defer
1637 * hash allocation until vmalloc space is available.
1638 */
1639 if (hashdist)
1640 return;
1641
1642 inode_hashtable =
1643 alloc_large_system_hash("Inode-cache",
1644 sizeof(struct hlist_head),
1645 ihash_entries,
1646 14,
1647 HASH_EARLY,
1648 &i_hash_shift,
1649 &i_hash_mask,
1650 0);
1651
Dimitri Sivanich074b8512012-02-08 12:39:07 -08001652 for (loop = 0; loop < (1U << i_hash_shift); loop++)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1654}
1655
Denis Cheng74bf17c2007-10-16 23:26:30 -07001656void __init inode_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657{
Dimitri Sivanich074b8512012-02-08 12:39:07 -08001658 unsigned int loop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659
1660 /* inode slab cache */
Paul Jacksonb0196002006-03-24 03:16:09 -08001661 inode_cachep = kmem_cache_create("inode_cache",
1662 sizeof(struct inode),
1663 0,
1664 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1665 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +09001666 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
1668 /* Hash may have been set up in inode_init_early */
1669 if (!hashdist)
1670 return;
1671
1672 inode_hashtable =
1673 alloc_large_system_hash("Inode-cache",
1674 sizeof(struct hlist_head),
1675 ihash_entries,
1676 14,
1677 0,
1678 &i_hash_shift,
1679 &i_hash_mask,
1680 0);
1681
Dimitri Sivanich074b8512012-02-08 12:39:07 -08001682 for (loop = 0; loop < (1U << i_hash_shift); loop++)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1684}
1685
1686void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1687{
1688 inode->i_mode = mode;
1689 if (S_ISCHR(mode)) {
1690 inode->i_fop = &def_chr_fops;
1691 inode->i_rdev = rdev;
1692 } else if (S_ISBLK(mode)) {
1693 inode->i_fop = &def_blk_fops;
1694 inode->i_rdev = rdev;
1695 } else if (S_ISFIFO(mode))
1696 inode->i_fop = &def_fifo_fops;
1697 else if (S_ISSOCK(mode))
1698 inode->i_fop = &bad_sock_fops;
1699 else
Manish Katiyaraf0d9ae2009-09-18 13:05:43 -07001700 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
1701 " inode %s:%lu\n", mode, inode->i_sb->s_id,
1702 inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703}
1704EXPORT_SYMBOL(init_special_inode);
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001705
1706/**
Ben Hutchingseaae668d2011-02-15 12:48:09 +00001707 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001708 * @inode: New inode
1709 * @dir: Directory inode
1710 * @mode: mode of the new inode
1711 */
1712void inode_init_owner(struct inode *inode, const struct inode *dir,
Al Viro62bb1092011-07-24 23:20:18 -04001713 umode_t mode)
Dmitry Monakhova1bd1202010-03-04 17:29:14 +03001714{
1715 inode->i_uid = current_fsuid();
1716 if (dir && dir->i_mode & S_ISGID) {
1717 inode->i_gid = dir->i_gid;
1718 if (S_ISDIR(mode))
1719 mode |= S_ISGID;
1720 } else
1721 inode->i_gid = current_fsgid();
1722 inode->i_mode = mode;
1723}
1724EXPORT_SYMBOL(inode_init_owner);
Serge E. Hallyne795b712011-03-23 16:43:25 -07001725
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001726/**
1727 * inode_owner_or_capable - check current task permissions to inode
1728 * @inode: inode being checked
1729 *
1730 * Return true if current either has CAP_FOWNER to the inode, or
1731 * owns the file.
Serge E. Hallyne795b712011-03-23 16:43:25 -07001732 */
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001733bool inode_owner_or_capable(const struct inode *inode)
Serge E. Hallyne795b712011-03-23 16:43:25 -07001734{
1735 struct user_namespace *ns = inode_userns(inode);
1736
1737 if (current_user_ns() == ns && current_fsuid() == inode->i_uid)
1738 return true;
1739 if (ns_capable(ns, CAP_FOWNER))
1740 return true;
1741 return false;
1742}
Serge E. Hallyn2e149672011-03-23 16:43:26 -07001743EXPORT_SYMBOL(inode_owner_or_capable);