Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/fs/super.c |
| 3 | * |
| 4 | * Copyright (C) 1991, 1992 Linus Torvalds |
| 5 | * |
| 6 | * super.c contains code to handle: - mount structures |
| 7 | * - super-block tables |
| 8 | * - filesystem drivers list |
| 9 | * - mount system call |
| 10 | * - umount system call |
| 11 | * - ustat system call |
| 12 | * |
| 13 | * GK 2/5/95 - Changed to support mounting the root fs via NFS |
| 14 | * |
| 15 | * Added kerneld support: Jacques Gelinas and Bjorn Ekwall |
| 16 | * Added change_root: Werner Almesberger & Hans Lermen, Feb '96 |
| 17 | * Added options to /proc/mounts: |
Jan Engelhardt | 96de0e2 | 2007-10-19 23:21:04 +0200 | [diff] [blame] | 18 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 19 | * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998 |
| 20 | * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000 |
| 21 | */ |
| 22 | |
Paul Gortmaker | 630d9c4 | 2011-11-16 23:57:37 -0500 | [diff] [blame] | 23 | #include <linux/export.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/slab.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | #include <linux/acct.h> |
| 26 | #include <linux/blkdev.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | #include <linux/mount.h> |
| 28 | #include <linux/security.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 29 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
| 30 | #include <linux/idr.h> |
Ingo Molnar | 353ab6e | 2006-03-26 01:37:12 -0800 | [diff] [blame] | 31 | #include <linux/mutex.h> |
Jens Axboe | 5477d0f | 2010-04-29 20:33:35 +0200 | [diff] [blame] | 32 | #include <linux/backing-dev.h> |
Nick Piggin | ceb5bdc | 2011-01-07 17:50:05 +1100 | [diff] [blame] | 33 | #include <linux/rculist_bl.h> |
Dan Magenheimer | c515e1f | 2011-05-26 10:01:43 -0600 | [diff] [blame] | 34 | #include <linux/cleancache.h> |
Al Viro | 4040153 | 2012-02-13 03:58:52 +0000 | [diff] [blame] | 35 | #include <linux/fsnotify.h> |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 36 | #include <linux/lockdep.h> |
Al Viro | 6d59e7f | 2008-03-22 15:48:17 -0400 | [diff] [blame] | 37 | #include "internal.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 38 | |
| 39 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | LIST_HEAD(super_blocks); |
| 41 | DEFINE_SPINLOCK(sb_lock); |
| 42 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
| 44 | "sb_writers", |
| 45 | "sb_pagefaults", |
| 46 | "sb_internal", |
| 47 | }; |
| 48 | |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 49 | /* |
| 50 | * One thing we have to be careful of with a per-sb shrinker is that we don't |
| 51 | * drop the last active reference to the superblock from within the shrinker. |
| 52 | * If that happens we could trigger unregistering the shrinker from within the |
| 53 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we |
| 54 | * take a passive reference to the superblock to avoid this from occurring. |
| 55 | */ |
| 56 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) |
| 57 | { |
| 58 | struct super_block *sb; |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 59 | int fs_objects = 0; |
| 60 | int total_objects; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 61 | |
| 62 | sb = container_of(shrink, struct super_block, s_shrink); |
| 63 | |
| 64 | /* |
| 65 | * Deadlock avoidance. We may hold various FS locks, and we don't want |
| 66 | * to recurse into the FS that called us in clear_inode() and friends.. |
| 67 | */ |
| 68 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) |
| 69 | return -1; |
| 70 | |
| 71 | if (!grab_super_passive(sb)) |
Minchan Kim | 8e125cd | 2012-07-31 16:43:16 -0700 | [diff] [blame] | 72 | return -1; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 73 | |
Al Viro | d040790 | 2013-07-19 21:12:31 +0400 | [diff] [blame^] | 74 | if (sb->s_op->nr_cached_objects) |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 75 | fs_objects = sb->s_op->nr_cached_objects(sb); |
| 76 | |
| 77 | total_objects = sb->s_nr_dentry_unused + |
| 78 | sb->s_nr_inodes_unused + fs_objects + 1; |
| 79 | |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 80 | if (sc->nr_to_scan) { |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 81 | int dentries; |
| 82 | int inodes; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 83 | |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 84 | /* proportion the scan between the caches */ |
| 85 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / |
| 86 | total_objects; |
| 87 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / |
| 88 | total_objects; |
| 89 | if (fs_objects) |
| 90 | fs_objects = (sc->nr_to_scan * fs_objects) / |
| 91 | total_objects; |
| 92 | /* |
| 93 | * prune the dcache first as the icache is pinned by it, then |
| 94 | * prune the icache, followed by the filesystem specific caches |
| 95 | */ |
| 96 | prune_dcache_sb(sb, dentries); |
| 97 | prune_icache_sb(sb, inodes); |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 98 | |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 99 | if (fs_objects && sb->s_op->free_cached_objects) { |
| 100 | sb->s_op->free_cached_objects(sb, fs_objects); |
| 101 | fs_objects = sb->s_op->nr_cached_objects(sb); |
| 102 | } |
| 103 | total_objects = sb->s_nr_dentry_unused + |
| 104 | sb->s_nr_inodes_unused + fs_objects; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 105 | } |
| 106 | |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 107 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 108 | drop_super(sb); |
Dave Chinner | 0e1fdaf | 2011-07-08 14:14:44 +1000 | [diff] [blame] | 109 | return total_objects; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 110 | } |
| 111 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 112 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
| 113 | { |
| 114 | int err; |
| 115 | int i; |
| 116 | |
| 117 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
| 118 | err = percpu_counter_init(&s->s_writers.counter[i], 0); |
| 119 | if (err < 0) |
| 120 | goto err_out; |
| 121 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], |
| 122 | &type->s_writers_key[i], 0); |
| 123 | } |
| 124 | init_waitqueue_head(&s->s_writers.wait); |
| 125 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
| 126 | return 0; |
| 127 | err_out: |
| 128 | while (--i >= 0) |
| 129 | percpu_counter_destroy(&s->s_writers.counter[i]); |
| 130 | return err; |
| 131 | } |
| 132 | |
| 133 | static void destroy_sb_writers(struct super_block *s) |
| 134 | { |
| 135 | int i; |
| 136 | |
| 137 | for (i = 0; i < SB_FREEZE_LEVELS; i++) |
| 138 | percpu_counter_destroy(&s->s_writers.counter[i]); |
| 139 | } |
| 140 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 141 | /** |
| 142 | * alloc_super - create new superblock |
Henrik Kretzschmar | fe2bbc4 | 2006-09-06 00:03:41 -0700 | [diff] [blame] | 143 | * @type: filesystem type superblock should belong to |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 144 | * @flags: the mount flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 145 | * |
| 146 | * Allocates and initializes a new &struct super_block. alloc_super() |
| 147 | * returns a pointer new superblock or %NULL if allocation had failed. |
| 148 | */ |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 149 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 150 | { |
Oliver Neukum | 11b0b5a | 2006-03-25 03:08:13 -0800 | [diff] [blame] | 151 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
Alexey Dobriyan | b87221d | 2009-09-21 17:01:09 -0700 | [diff] [blame] | 152 | static const struct super_operations default_op; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 153 | |
| 154 | if (s) { |
Christoph Hellwig | 7b7a866 | 2013-09-04 15:04:39 +0200 | [diff] [blame] | 155 | if (security_sb_alloc(s)) |
| 156 | goto out_free_sb; |
| 157 | |
Nick Piggin | 6416ccb | 2010-08-18 04:37:38 +1000 | [diff] [blame] | 158 | #ifdef CONFIG_SMP |
| 159 | s->s_files = alloc_percpu(struct list_head); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 160 | if (!s->s_files) |
| 161 | goto err_out; |
| 162 | else { |
Nick Piggin | 6416ccb | 2010-08-18 04:37:38 +1000 | [diff] [blame] | 163 | int i; |
| 164 | |
| 165 | for_each_possible_cpu(i) |
| 166 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); |
| 167 | } |
| 168 | #else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 169 | INIT_LIST_HEAD(&s->s_files); |
Nick Piggin | 6416ccb | 2010-08-18 04:37:38 +1000 | [diff] [blame] | 170 | #endif |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 171 | if (init_sb_writers(s, type)) |
| 172 | goto err_out; |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 173 | s->s_flags = flags; |
Jens Axboe | 95f2860 | 2011-03-17 11:13:12 +0100 | [diff] [blame] | 174 | s->s_bdi = &default_backing_dev_info; |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 175 | INIT_HLIST_NODE(&s->s_instances); |
Nick Piggin | ceb5bdc | 2011-01-07 17:50:05 +1100 | [diff] [blame] | 176 | INIT_HLIST_BL_HEAD(&s->s_anon); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 177 | INIT_LIST_HEAD(&s->s_inodes); |
Kentaro Makita | da3bbdd | 2008-07-23 21:27:13 -0700 | [diff] [blame] | 178 | INIT_LIST_HEAD(&s->s_dentry_lru); |
Dave Chinner | 98b745c | 2011-07-08 14:14:39 +1000 | [diff] [blame] | 179 | INIT_LIST_HEAD(&s->s_inode_lru); |
Dave Chinner | 09cc9fc | 2011-07-08 14:14:40 +1000 | [diff] [blame] | 180 | spin_lock_init(&s->s_inode_lru_lock); |
Miklos Szeredi | 39f7c4d | 2011-11-21 12:11:30 +0100 | [diff] [blame] | 181 | INIT_LIST_HEAD(&s->s_mounts); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | init_rwsem(&s->s_umount); |
Arjan van de Ven | 897c6ff | 2006-07-03 00:25:28 -0700 | [diff] [blame] | 183 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
Ingo Molnar | cf51624 | 2006-07-03 00:25:27 -0700 | [diff] [blame] | 184 | /* |
Peter Zijlstra | ada723d | 2009-02-18 14:48:30 -0800 | [diff] [blame] | 185 | * sget() can have s_umount recursion. |
| 186 | * |
| 187 | * When it cannot find a suitable sb, it allocates a new |
| 188 | * one (this one), and tries again to find a suitable old |
| 189 | * one. |
| 190 | * |
| 191 | * In case that succeeds, it will acquire the s_umount |
| 192 | * lock of the old one. Since these are clearly distrinct |
| 193 | * locks, and this object isn't exposed yet, there's no |
| 194 | * risk of deadlocks. |
| 195 | * |
| 196 | * Annotate this by putting this lock in a different |
| 197 | * subclass. |
| 198 | */ |
| 199 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); |
Al Viro | b20bd1a | 2010-03-22 08:53:19 -0400 | [diff] [blame] | 200 | s->s_count = 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 201 | atomic_set(&s->s_active, 1); |
Arjan van de Ven | a11f3a0 | 2006-03-23 03:00:33 -0800 | [diff] [blame] | 202 | mutex_init(&s->s_vfs_rename_mutex); |
Roland Dreier | 51ee049 | 2010-04-27 14:23:57 -0700 | [diff] [blame] | 203 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
Ingo Molnar | d3be915 | 2006-03-23 03:00:29 -0800 | [diff] [blame] | 204 | mutex_init(&s->s_dquot.dqio_mutex); |
| 205 | mutex_init(&s->s_dquot.dqonoff_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 206 | init_rwsem(&s->s_dquot.dqptr_sem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 207 | s->s_maxbytes = MAX_NON_LFS; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 208 | s->s_op = &default_op; |
| 209 | s->s_time_gran = 1000000000; |
Dan Magenheimer | c515e1f | 2011-05-26 10:01:43 -0600 | [diff] [blame] | 210 | s->cleancache_poolid = -1; |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 211 | |
| 212 | s->s_shrink.seeks = DEFAULT_SEEKS; |
| 213 | s->s_shrink.shrink = prune_super; |
Dave Chinner | 8ab4766 | 2011-07-08 14:14:45 +1000 | [diff] [blame] | 214 | s->s_shrink.batch = 1024; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | } |
| 216 | out: |
| 217 | return s; |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 218 | err_out: |
| 219 | security_sb_free(s); |
| 220 | #ifdef CONFIG_SMP |
| 221 | if (s->s_files) |
| 222 | free_percpu(s->s_files); |
| 223 | #endif |
| 224 | destroy_sb_writers(s); |
Christoph Hellwig | 7b7a866 | 2013-09-04 15:04:39 +0200 | [diff] [blame] | 225 | out_free_sb: |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 226 | kfree(s); |
| 227 | s = NULL; |
| 228 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 229 | } |
| 230 | |
| 231 | /** |
| 232 | * destroy_super - frees a superblock |
| 233 | * @s: superblock to free |
| 234 | * |
| 235 | * Frees a superblock. |
| 236 | */ |
| 237 | static inline void destroy_super(struct super_block *s) |
| 238 | { |
Nick Piggin | 6416ccb | 2010-08-18 04:37:38 +1000 | [diff] [blame] | 239 | #ifdef CONFIG_SMP |
| 240 | free_percpu(s->s_files); |
| 241 | #endif |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 242 | destroy_sb_writers(s); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 243 | security_sb_free(s); |
Miklos Szeredi | 39f7c4d | 2011-11-21 12:11:30 +0100 | [diff] [blame] | 244 | WARN_ON(!list_empty(&s->s_mounts)); |
Miklos Szeredi | 79c0b2d | 2007-05-08 00:25:43 -0700 | [diff] [blame] | 245 | kfree(s->s_subtype); |
Miklos Szeredi | b3b304a | 2008-02-08 04:21:35 -0800 | [diff] [blame] | 246 | kfree(s->s_options); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 247 | kfree(s); |
| 248 | } |
| 249 | |
| 250 | /* Superblock refcounting */ |
| 251 | |
| 252 | /* |
Al Viro | 35cf7ba | 2010-03-22 21:13:53 -0400 | [diff] [blame] | 253 | * Drop a superblock's refcount. The caller must hold sb_lock. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | */ |
Al Viro | f47ec3f | 2011-11-21 21:15:42 -0500 | [diff] [blame] | 255 | static void __put_super(struct super_block *sb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 256 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 257 | if (!--sb->s_count) { |
Al Viro | 551de6f | 2010-03-22 19:36:35 -0400 | [diff] [blame] | 258 | list_del_init(&sb->s_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | destroy_super(sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 260 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 261 | } |
| 262 | |
| 263 | /** |
| 264 | * put_super - drop a temporary reference to superblock |
| 265 | * @sb: superblock in question |
| 266 | * |
| 267 | * Drops a temporary reference, frees superblock if there's no |
| 268 | * references left. |
| 269 | */ |
Al Viro | f47ec3f | 2011-11-21 21:15:42 -0500 | [diff] [blame] | 270 | static void put_super(struct super_block *sb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 271 | { |
| 272 | spin_lock(&sb_lock); |
| 273 | __put_super(sb); |
| 274 | spin_unlock(&sb_lock); |
| 275 | } |
| 276 | |
| 277 | |
| 278 | /** |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 279 | * deactivate_locked_super - drop an active reference to superblock |
| 280 | * @s: superblock to deactivate |
| 281 | * |
Al Viro | 1712ac8 | 2010-03-22 15:22:31 -0400 | [diff] [blame] | 282 | * Drops an active reference to superblock, converting it into a temprory |
| 283 | * one if there is no other active references left. In that case we |
| 284 | * tell fs driver to shut it down and drop the temporary reference we |
| 285 | * had just acquired. |
| 286 | * |
| 287 | * Caller holds exclusive lock on superblock; that lock is released. |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 288 | */ |
| 289 | void deactivate_locked_super(struct super_block *s) |
| 290 | { |
| 291 | struct file_system_type *fs = s->s_type; |
Al Viro | b20bd1a | 2010-03-22 08:53:19 -0400 | [diff] [blame] | 292 | if (atomic_dec_and_test(&s->s_active)) { |
Dan Magenheimer | 3167760 | 2011-09-21 11:56:28 -0400 | [diff] [blame] | 293 | cleancache_invalidate_fs(s); |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 294 | fs->kill_sb(s); |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 295 | |
| 296 | /* caches are now gone, we can safely kill the shrinker now */ |
| 297 | unregister_shrinker(&s->s_shrink); |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 298 | put_filesystem(fs); |
| 299 | put_super(s); |
| 300 | } else { |
| 301 | up_write(&s->s_umount); |
| 302 | } |
| 303 | } |
| 304 | |
| 305 | EXPORT_SYMBOL(deactivate_locked_super); |
| 306 | |
| 307 | /** |
Al Viro | 1712ac8 | 2010-03-22 15:22:31 -0400 | [diff] [blame] | 308 | * deactivate_super - drop an active reference to superblock |
| 309 | * @s: superblock to deactivate |
| 310 | * |
| 311 | * Variant of deactivate_locked_super(), except that superblock is *not* |
| 312 | * locked by caller. If we are going to drop the final active reference, |
| 313 | * lock will be acquired prior to that. |
| 314 | */ |
| 315 | void deactivate_super(struct super_block *s) |
| 316 | { |
| 317 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
| 318 | down_write(&s->s_umount); |
| 319 | deactivate_locked_super(s); |
| 320 | } |
| 321 | } |
| 322 | |
| 323 | EXPORT_SYMBOL(deactivate_super); |
| 324 | |
| 325 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 326 | * grab_super - acquire an active reference |
| 327 | * @s: reference we are trying to make active |
| 328 | * |
| 329 | * Tries to acquire an active reference. grab_super() is used when we |
| 330 | * had just found a superblock in super_blocks or fs_type->fs_supers |
| 331 | * and want to turn it into a full-blown active reference. grab_super() |
| 332 | * is called with sb_lock held and drops it. Returns 1 in case of |
| 333 | * success, 0 if we had failed (superblock contents was already dead or |
Al Viro | acfec9a | 2013-07-20 03:13:55 +0400 | [diff] [blame] | 334 | * dying when grab_super() had been called). Note that this is only |
| 335 | * called for superblocks not in rundown mode (== ones still on ->fs_supers |
| 336 | * of their type), so increment of ->s_count is OK here. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 337 | */ |
Josh Triplett | 9c4dbee | 2006-09-29 01:59:29 -0700 | [diff] [blame] | 338 | static int grab_super(struct super_block *s) __releases(sb_lock) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 339 | { |
| 340 | s->s_count++; |
| 341 | spin_unlock(&sb_lock); |
| 342 | down_write(&s->s_umount); |
Al Viro | acfec9a | 2013-07-20 03:13:55 +0400 | [diff] [blame] | 343 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
| 344 | put_super(s); |
| 345 | return 1; |
| 346 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 347 | up_write(&s->s_umount); |
| 348 | put_super(s); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 349 | return 0; |
| 350 | } |
| 351 | |
David Howells | cf9a2ae | 2006-08-29 19:05:54 +0100 | [diff] [blame] | 352 | /* |
Dave Chinner | 12ad3ab | 2011-07-08 14:14:41 +1000 | [diff] [blame] | 353 | * grab_super_passive - acquire a passive reference |
Wanpeng Li | 331cbde | 2012-06-09 11:10:55 +0800 | [diff] [blame] | 354 | * @sb: reference we are trying to grab |
Dave Chinner | 12ad3ab | 2011-07-08 14:14:41 +1000 | [diff] [blame] | 355 | * |
| 356 | * Tries to acquire a passive reference. This is used in places where we |
| 357 | * cannot take an active reference but we need to ensure that the |
| 358 | * superblock does not go away while we are working on it. It returns |
| 359 | * false if a reference was not gained, and returns true with the s_umount |
| 360 | * lock held in read mode if a reference is gained. On successful return, |
| 361 | * the caller must drop the s_umount lock and the passive reference when |
| 362 | * done. |
| 363 | */ |
| 364 | bool grab_super_passive(struct super_block *sb) |
| 365 | { |
| 366 | spin_lock(&sb_lock); |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 367 | if (hlist_unhashed(&sb->s_instances)) { |
Dave Chinner | 12ad3ab | 2011-07-08 14:14:41 +1000 | [diff] [blame] | 368 | spin_unlock(&sb_lock); |
| 369 | return false; |
| 370 | } |
| 371 | |
| 372 | sb->s_count++; |
| 373 | spin_unlock(&sb_lock); |
| 374 | |
| 375 | if (down_read_trylock(&sb->s_umount)) { |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 376 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
Dave Chinner | 12ad3ab | 2011-07-08 14:14:41 +1000 | [diff] [blame] | 377 | return true; |
| 378 | up_read(&sb->s_umount); |
| 379 | } |
| 380 | |
| 381 | put_super(sb); |
| 382 | return false; |
| 383 | } |
| 384 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 385 | /** |
| 386 | * generic_shutdown_super - common helper for ->kill_sb() |
| 387 | * @sb: superblock to kill |
| 388 | * |
| 389 | * generic_shutdown_super() does all fs-independent work on superblock |
| 390 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects |
| 391 | * that need destruction out of superblock, call generic_shutdown_super() |
| 392 | * and release aforementioned objects. Note: dentries and inodes _are_ |
| 393 | * taken care of and do not need specific handling. |
David Howells | c636ebd | 2006-10-11 01:22:19 -0700 | [diff] [blame] | 394 | * |
| 395 | * Upon calling this function, the filesystem may no longer alter or |
| 396 | * rearrange the set of dentries belonging to this super_block, nor may it |
| 397 | * change the attachments of dentries to inodes. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 398 | */ |
| 399 | void generic_shutdown_super(struct super_block *sb) |
| 400 | { |
Josef 'Jeff' Sipek | ee9b6d6 | 2007-02-12 00:55:41 -0800 | [diff] [blame] | 401 | const struct super_operations *sop = sb->s_op; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 402 | |
David Howells | c636ebd | 2006-10-11 01:22:19 -0700 | [diff] [blame] | 403 | if (sb->s_root) { |
| 404 | shrink_dcache_for_umount(sb); |
Jan Kara | 60b0680 | 2009-04-27 16:43:53 +0200 | [diff] [blame] | 405 | sync_filesystem(sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 406 | sb->s_flags &= ~MS_ACTIVE; |
Arjan van de Ven | efaee19 | 2009-01-06 07:20:54 -0800 | [diff] [blame] | 407 | |
Al Viro | 63997e9 | 2010-10-25 20:49:35 -0400 | [diff] [blame] | 408 | fsnotify_unmount_inodes(&sb->s_inodes); |
| 409 | |
| 410 | evict_inodes(sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 411 | |
Christoph Hellwig | 7b7a866 | 2013-09-04 15:04:39 +0200 | [diff] [blame] | 412 | if (sb->s_dio_done_wq) { |
| 413 | destroy_workqueue(sb->s_dio_done_wq); |
| 414 | sb->s_dio_done_wq = NULL; |
| 415 | } |
| 416 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 417 | if (sop->put_super) |
| 418 | sop->put_super(sb); |
| 419 | |
Al Viro | 63997e9 | 2010-10-25 20:49:35 -0400 | [diff] [blame] | 420 | if (!list_empty(&sb->s_inodes)) { |
Dave Jones | 7b4fe29 | 2006-02-07 12:58:48 -0800 | [diff] [blame] | 421 | printk("VFS: Busy inodes after unmount of %s. " |
| 422 | "Self-destruct in 5 seconds. Have a nice day...\n", |
| 423 | sb->s_id); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 424 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 425 | } |
| 426 | spin_lock(&sb_lock); |
| 427 | /* should be initialized for __put_super_and_need_restart() */ |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 428 | hlist_del_init(&sb->s_instances); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 429 | spin_unlock(&sb_lock); |
| 430 | up_write(&sb->s_umount); |
| 431 | } |
| 432 | |
| 433 | EXPORT_SYMBOL(generic_shutdown_super); |
| 434 | |
| 435 | /** |
| 436 | * sget - find or create a superblock |
| 437 | * @type: filesystem type superblock should belong to |
| 438 | * @test: comparison callback |
| 439 | * @set: setup callback |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 440 | * @flags: mount flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 441 | * @data: argument to each of them |
| 442 | */ |
| 443 | struct super_block *sget(struct file_system_type *type, |
| 444 | int (*test)(struct super_block *,void *), |
| 445 | int (*set)(struct super_block *,void *), |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 446 | int flags, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 447 | void *data) |
| 448 | { |
| 449 | struct super_block *s = NULL; |
Matthias Kaehlcke | d473012 | 2007-10-18 23:39:57 -0700 | [diff] [blame] | 450 | struct super_block *old; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 451 | int err; |
| 452 | |
| 453 | retry: |
| 454 | spin_lock(&sb_lock); |
Matthias Kaehlcke | d473012 | 2007-10-18 23:39:57 -0700 | [diff] [blame] | 455 | if (test) { |
Sasha Levin | b67bfe0 | 2013-02-27 17:06:00 -0800 | [diff] [blame] | 456 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
Matthias Kaehlcke | d473012 | 2007-10-18 23:39:57 -0700 | [diff] [blame] | 457 | if (!test(old, data)) |
| 458 | continue; |
| 459 | if (!grab_super(old)) |
| 460 | goto retry; |
Li Zefan | a3cfbb5 | 2009-03-12 14:31:29 -0700 | [diff] [blame] | 461 | if (s) { |
| 462 | up_write(&s->s_umount); |
Matthias Kaehlcke | d473012 | 2007-10-18 23:39:57 -0700 | [diff] [blame] | 463 | destroy_super(s); |
Al Viro | 7a4dec5 | 2010-08-09 12:05:43 -0400 | [diff] [blame] | 464 | s = NULL; |
Li Zefan | a3cfbb5 | 2009-03-12 14:31:29 -0700 | [diff] [blame] | 465 | } |
Matthias Kaehlcke | d473012 | 2007-10-18 23:39:57 -0700 | [diff] [blame] | 466 | return old; |
| 467 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 468 | } |
| 469 | if (!s) { |
| 470 | spin_unlock(&sb_lock); |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 471 | s = alloc_super(type, flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 472 | if (!s) |
| 473 | return ERR_PTR(-ENOMEM); |
| 474 | goto retry; |
| 475 | } |
| 476 | |
| 477 | err = set(s, data); |
| 478 | if (err) { |
| 479 | spin_unlock(&sb_lock); |
Li Zefan | a3cfbb5 | 2009-03-12 14:31:29 -0700 | [diff] [blame] | 480 | up_write(&s->s_umount); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 481 | destroy_super(s); |
| 482 | return ERR_PTR(err); |
| 483 | } |
| 484 | s->s_type = type; |
| 485 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); |
| 486 | list_add_tail(&s->s_list, &super_blocks); |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 487 | hlist_add_head(&s->s_instances, &type->fs_supers); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 488 | spin_unlock(&sb_lock); |
| 489 | get_filesystem(type); |
Dave Chinner | b0d40c9 | 2011-07-08 14:14:42 +1000 | [diff] [blame] | 490 | register_shrinker(&s->s_shrink); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 491 | return s; |
| 492 | } |
| 493 | |
| 494 | EXPORT_SYMBOL(sget); |
| 495 | |
| 496 | void drop_super(struct super_block *sb) |
| 497 | { |
| 498 | up_read(&sb->s_umount); |
| 499 | put_super(sb); |
| 500 | } |
| 501 | |
| 502 | EXPORT_SYMBOL(drop_super); |
| 503 | |
Christoph Hellwig | e500475 | 2009-05-05 16:08:56 +0200 | [diff] [blame] | 504 | /** |
Al Viro | 01a05b3 | 2010-03-23 06:06:58 -0400 | [diff] [blame] | 505 | * iterate_supers - call function for all active superblocks |
| 506 | * @f: function to call |
| 507 | * @arg: argument to pass to it |
| 508 | * |
| 509 | * Scans the superblock list and calls given function, passing it |
| 510 | * locked superblock and given argument. |
| 511 | */ |
| 512 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) |
| 513 | { |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 514 | struct super_block *sb, *p = NULL; |
Al Viro | 01a05b3 | 2010-03-23 06:06:58 -0400 | [diff] [blame] | 515 | |
| 516 | spin_lock(&sb_lock); |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 517 | list_for_each_entry(sb, &super_blocks, s_list) { |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 518 | if (hlist_unhashed(&sb->s_instances)) |
Al Viro | 01a05b3 | 2010-03-23 06:06:58 -0400 | [diff] [blame] | 519 | continue; |
| 520 | sb->s_count++; |
| 521 | spin_unlock(&sb_lock); |
| 522 | |
| 523 | down_read(&sb->s_umount); |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 524 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
Al Viro | 01a05b3 | 2010-03-23 06:06:58 -0400 | [diff] [blame] | 525 | f(sb, arg); |
| 526 | up_read(&sb->s_umount); |
| 527 | |
| 528 | spin_lock(&sb_lock); |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 529 | if (p) |
| 530 | __put_super(p); |
| 531 | p = sb; |
Al Viro | 01a05b3 | 2010-03-23 06:06:58 -0400 | [diff] [blame] | 532 | } |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 533 | if (p) |
| 534 | __put_super(p); |
Al Viro | 01a05b3 | 2010-03-23 06:06:58 -0400 | [diff] [blame] | 535 | spin_unlock(&sb_lock); |
| 536 | } |
| 537 | |
| 538 | /** |
Al Viro | 43e15cd | 2011-06-03 20:16:57 -0400 | [diff] [blame] | 539 | * iterate_supers_type - call function for superblocks of given type |
| 540 | * @type: fs type |
| 541 | * @f: function to call |
| 542 | * @arg: argument to pass to it |
| 543 | * |
| 544 | * Scans the superblock list and calls given function, passing it |
| 545 | * locked superblock and given argument. |
| 546 | */ |
| 547 | void iterate_supers_type(struct file_system_type *type, |
| 548 | void (*f)(struct super_block *, void *), void *arg) |
| 549 | { |
| 550 | struct super_block *sb, *p = NULL; |
| 551 | |
| 552 | spin_lock(&sb_lock); |
Sasha Levin | b67bfe0 | 2013-02-27 17:06:00 -0800 | [diff] [blame] | 553 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
Al Viro | 43e15cd | 2011-06-03 20:16:57 -0400 | [diff] [blame] | 554 | sb->s_count++; |
| 555 | spin_unlock(&sb_lock); |
| 556 | |
| 557 | down_read(&sb->s_umount); |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 558 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
Al Viro | 43e15cd | 2011-06-03 20:16:57 -0400 | [diff] [blame] | 559 | f(sb, arg); |
| 560 | up_read(&sb->s_umount); |
| 561 | |
| 562 | spin_lock(&sb_lock); |
| 563 | if (p) |
| 564 | __put_super(p); |
| 565 | p = sb; |
| 566 | } |
| 567 | if (p) |
| 568 | __put_super(p); |
| 569 | spin_unlock(&sb_lock); |
| 570 | } |
| 571 | |
| 572 | EXPORT_SYMBOL(iterate_supers_type); |
| 573 | |
| 574 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 575 | * get_super - get the superblock of a device |
| 576 | * @bdev: device to get the superblock for |
| 577 | * |
| 578 | * Scans the superblock list and finds the superblock of the file system |
| 579 | * mounted on the device given. %NULL is returned if no match is found. |
| 580 | */ |
| 581 | |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 582 | struct super_block *get_super(struct block_device *bdev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 583 | { |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 584 | struct super_block *sb; |
| 585 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 586 | if (!bdev) |
| 587 | return NULL; |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 588 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 589 | spin_lock(&sb_lock); |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 590 | rescan: |
| 591 | list_for_each_entry(sb, &super_blocks, s_list) { |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 592 | if (hlist_unhashed(&sb->s_instances)) |
Al Viro | 551de6f | 2010-03-22 19:36:35 -0400 | [diff] [blame] | 593 | continue; |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 594 | if (sb->s_bdev == bdev) { |
| 595 | sb->s_count++; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 596 | spin_unlock(&sb_lock); |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 597 | down_read(&sb->s_umount); |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 598 | /* still alive? */ |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 599 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 600 | return sb; |
| 601 | up_read(&sb->s_umount); |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 602 | /* nope, got unmounted */ |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 603 | spin_lock(&sb_lock); |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 604 | __put_super(sb); |
| 605 | goto rescan; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 606 | } |
| 607 | } |
| 608 | spin_unlock(&sb_lock); |
| 609 | return NULL; |
| 610 | } |
| 611 | |
| 612 | EXPORT_SYMBOL(get_super); |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 613 | |
| 614 | /** |
Jan Kara | 6b6dc83 | 2012-02-10 11:03:00 +0100 | [diff] [blame] | 615 | * get_super_thawed - get thawed superblock of a device |
| 616 | * @bdev: device to get the superblock for |
| 617 | * |
| 618 | * Scans the superblock list and finds the superblock of the file system |
| 619 | * mounted on the device. The superblock is returned once it is thawed |
| 620 | * (or immediately if it was not frozen). %NULL is returned if no match |
| 621 | * is found. |
| 622 | */ |
| 623 | struct super_block *get_super_thawed(struct block_device *bdev) |
| 624 | { |
| 625 | while (1) { |
| 626 | struct super_block *s = get_super(bdev); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 627 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
Jan Kara | 6b6dc83 | 2012-02-10 11:03:00 +0100 | [diff] [blame] | 628 | return s; |
| 629 | up_read(&s->s_umount); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 630 | wait_event(s->s_writers.wait_unfrozen, |
| 631 | s->s_writers.frozen == SB_UNFROZEN); |
Jan Kara | 6b6dc83 | 2012-02-10 11:03:00 +0100 | [diff] [blame] | 632 | put_super(s); |
| 633 | } |
| 634 | } |
| 635 | EXPORT_SYMBOL(get_super_thawed); |
| 636 | |
| 637 | /** |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 638 | * get_active_super - get an active reference to the superblock of a device |
| 639 | * @bdev: device to get the superblock for |
| 640 | * |
| 641 | * Scans the superblock list and finds the superblock of the file system |
| 642 | * mounted on the device given. Returns the superblock with an active |
Al Viro | d3f2147 | 2010-03-23 11:11:05 -0400 | [diff] [blame] | 643 | * reference or %NULL if none was found. |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 644 | */ |
| 645 | struct super_block *get_active_super(struct block_device *bdev) |
| 646 | { |
| 647 | struct super_block *sb; |
| 648 | |
| 649 | if (!bdev) |
| 650 | return NULL; |
| 651 | |
Al Viro | 1494583 | 2010-03-22 20:15:33 -0400 | [diff] [blame] | 652 | restart: |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 653 | spin_lock(&sb_lock); |
| 654 | list_for_each_entry(sb, &super_blocks, s_list) { |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 655 | if (hlist_unhashed(&sb->s_instances)) |
Al Viro | 551de6f | 2010-03-22 19:36:35 -0400 | [diff] [blame] | 656 | continue; |
Al Viro | 1494583 | 2010-03-22 20:15:33 -0400 | [diff] [blame] | 657 | if (sb->s_bdev == bdev) { |
Al Viro | acfec9a | 2013-07-20 03:13:55 +0400 | [diff] [blame] | 658 | if (!grab_super(sb)) |
Al Viro | 1494583 | 2010-03-22 20:15:33 -0400 | [diff] [blame] | 659 | goto restart; |
Al Viro | acfec9a | 2013-07-20 03:13:55 +0400 | [diff] [blame] | 660 | up_write(&sb->s_umount); |
| 661 | return sb; |
Al Viro | 1494583 | 2010-03-22 20:15:33 -0400 | [diff] [blame] | 662 | } |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 663 | } |
| 664 | spin_unlock(&sb_lock); |
| 665 | return NULL; |
| 666 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 667 | |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 668 | struct super_block *user_get_super(dev_t dev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 669 | { |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 670 | struct super_block *sb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 671 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 672 | spin_lock(&sb_lock); |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 673 | rescan: |
| 674 | list_for_each_entry(sb, &super_blocks, s_list) { |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 675 | if (hlist_unhashed(&sb->s_instances)) |
Al Viro | 551de6f | 2010-03-22 19:36:35 -0400 | [diff] [blame] | 676 | continue; |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 677 | if (sb->s_dev == dev) { |
| 678 | sb->s_count++; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 679 | spin_unlock(&sb_lock); |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 680 | down_read(&sb->s_umount); |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 681 | /* still alive? */ |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 682 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 683 | return sb; |
| 684 | up_read(&sb->s_umount); |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 685 | /* nope, got unmounted */ |
Kirill Korotaev | 618f063 | 2005-06-23 00:09:54 -0700 | [diff] [blame] | 686 | spin_lock(&sb_lock); |
Al Viro | df40c01 | 2010-03-22 20:23:25 -0400 | [diff] [blame] | 687 | __put_super(sb); |
| 688 | goto rescan; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 689 | } |
| 690 | } |
| 691 | spin_unlock(&sb_lock); |
| 692 | return NULL; |
| 693 | } |
| 694 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 695 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 696 | * do_remount_sb - asks filesystem to change mount options. |
| 697 | * @sb: superblock in question |
| 698 | * @flags: numeric part of options |
| 699 | * @data: the rest of options |
| 700 | * @force: whether or not to force the change |
| 701 | * |
| 702 | * Alters the mount options of a mounted file system. |
| 703 | */ |
| 704 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) |
| 705 | { |
| 706 | int retval; |
Christoph Hellwig | c79d967 | 2010-05-19 07:16:40 -0400 | [diff] [blame] | 707 | int remount_ro; |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 708 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 709 | if (sb->s_writers.frozen != SB_UNFROZEN) |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 710 | return -EBUSY; |
| 711 | |
David Howells | 9361401 | 2006-09-30 20:45:40 +0200 | [diff] [blame] | 712 | #ifdef CONFIG_BLOCK |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 713 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
| 714 | return -EACCES; |
David Howells | 9361401 | 2006-09-30 20:45:40 +0200 | [diff] [blame] | 715 | #endif |
Christoph Hellwig | 4504230 | 2009-08-03 23:28:35 +0200 | [diff] [blame] | 716 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 717 | if (flags & MS_RDONLY) |
| 718 | acct_auto_close(sb); |
| 719 | shrink_dcache_sb(sb); |
Jan Kara | 60b0680 | 2009-04-27 16:43:53 +0200 | [diff] [blame] | 720 | sync_filesystem(sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 721 | |
Nick Piggin | d208bbd | 2009-12-21 16:28:53 -0800 | [diff] [blame] | 722 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
Nick Piggin | d208bbd | 2009-12-21 16:28:53 -0800 | [diff] [blame] | 723 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 724 | /* If we are remounting RDONLY and current sb is read/write, |
| 725 | make sure there are no rw files opened */ |
Nick Piggin | d208bbd | 2009-12-21 16:28:53 -0800 | [diff] [blame] | 726 | if (remount_ro) { |
Miklos Szeredi | 4ed5e82 | 2011-11-21 12:11:31 +0100 | [diff] [blame] | 727 | if (force) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 728 | mark_files_ro(sb); |
Miklos Szeredi | 4ed5e82 | 2011-11-21 12:11:31 +0100 | [diff] [blame] | 729 | } else { |
| 730 | retval = sb_prepare_remount_readonly(sb); |
| 731 | if (retval) |
| 732 | return retval; |
Miklos Szeredi | 4ed5e82 | 2011-11-21 12:11:31 +0100 | [diff] [blame] | 733 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 734 | } |
| 735 | |
| 736 | if (sb->s_op->remount_fs) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 737 | retval = sb->s_op->remount_fs(sb, &flags, data); |
Miklos Szeredi | 2833eb2 | 2011-10-28 14:13:20 +0200 | [diff] [blame] | 738 | if (retval) { |
| 739 | if (!force) |
Miklos Szeredi | 4ed5e82 | 2011-11-21 12:11:31 +0100 | [diff] [blame] | 740 | goto cancel_readonly; |
Miklos Szeredi | 2833eb2 | 2011-10-28 14:13:20 +0200 | [diff] [blame] | 741 | /* If forced remount, go ahead despite any errors */ |
| 742 | WARN(1, "forced remount of a %s fs returned %i\n", |
| 743 | sb->s_type->name, retval); |
| 744 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 745 | } |
| 746 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); |
Miklos Szeredi | 4ed5e82 | 2011-11-21 12:11:31 +0100 | [diff] [blame] | 747 | /* Needs to be ordered wrt mnt_is_readonly() */ |
| 748 | smp_wmb(); |
| 749 | sb->s_readonly_remount = 0; |
Christoph Hellwig | c79d967 | 2010-05-19 07:16:40 -0400 | [diff] [blame] | 750 | |
Nick Piggin | d208bbd | 2009-12-21 16:28:53 -0800 | [diff] [blame] | 751 | /* |
| 752 | * Some filesystems modify their metadata via some other path than the |
| 753 | * bdev buffer cache (eg. use a private mapping, or directories in |
| 754 | * pagecache, etc). Also file data modifications go via their own |
| 755 | * mappings. So If we try to mount readonly then copy the filesystem |
| 756 | * from bdev, we could get stale data, so invalidate it to give a best |
| 757 | * effort at coherency. |
| 758 | */ |
| 759 | if (remount_ro && sb->s_bdev) |
| 760 | invalidate_bdev(sb->s_bdev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 761 | return 0; |
Miklos Szeredi | 4ed5e82 | 2011-11-21 12:11:31 +0100 | [diff] [blame] | 762 | |
| 763 | cancel_readonly: |
| 764 | sb->s_readonly_remount = 0; |
| 765 | return retval; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 766 | } |
| 767 | |
Jens Axboe | a2a9537 | 2009-03-17 09:38:40 +0100 | [diff] [blame] | 768 | static void do_emergency_remount(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 769 | { |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 770 | struct super_block *sb, *p = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 771 | |
| 772 | spin_lock(&sb_lock); |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 773 | list_for_each_entry(sb, &super_blocks, s_list) { |
Al Viro | a516616 | 2011-12-12 22:53:00 -0500 | [diff] [blame] | 774 | if (hlist_unhashed(&sb->s_instances)) |
Al Viro | 551de6f | 2010-03-22 19:36:35 -0400 | [diff] [blame] | 775 | continue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 776 | sb->s_count++; |
| 777 | spin_unlock(&sb_lock); |
Al Viro | 443b94b | 2009-05-05 23:48:50 -0400 | [diff] [blame] | 778 | down_write(&sb->s_umount); |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 779 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
| 780 | !(sb->s_flags & MS_RDONLY)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 781 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 782 | * What lock protects sb->s_flags?? |
| 783 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 785 | } |
Al Viro | 443b94b | 2009-05-05 23:48:50 -0400 | [diff] [blame] | 786 | up_write(&sb->s_umount); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 787 | spin_lock(&sb_lock); |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 788 | if (p) |
| 789 | __put_super(p); |
| 790 | p = sb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 791 | } |
Al Viro | dca3325 | 2010-07-25 02:31:46 +0400 | [diff] [blame] | 792 | if (p) |
| 793 | __put_super(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 794 | spin_unlock(&sb_lock); |
Jens Axboe | a2a9537 | 2009-03-17 09:38:40 +0100 | [diff] [blame] | 795 | kfree(work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 796 | printk("Emergency Remount complete\n"); |
| 797 | } |
| 798 | |
| 799 | void emergency_remount(void) |
| 800 | { |
Jens Axboe | a2a9537 | 2009-03-17 09:38:40 +0100 | [diff] [blame] | 801 | struct work_struct *work; |
| 802 | |
| 803 | work = kmalloc(sizeof(*work), GFP_ATOMIC); |
| 804 | if (work) { |
| 805 | INIT_WORK(work, do_emergency_remount); |
| 806 | schedule_work(work); |
| 807 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 808 | } |
| 809 | |
| 810 | /* |
| 811 | * Unnamed block devices are dummy devices used by virtual |
| 812 | * filesystems which don't use real block-devices. -- jrs |
| 813 | */ |
| 814 | |
Alexey Dobriyan | ad76cbc | 2008-08-28 06:26:23 +0400 | [diff] [blame] | 815 | static DEFINE_IDA(unnamed_dev_ida); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 816 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
Al Viro | c63e09e | 2009-06-24 02:05:18 -0400 | [diff] [blame] | 817 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 818 | |
Al Viro | 0ee5dc6 | 2011-07-07 15:44:25 -0400 | [diff] [blame] | 819 | int get_anon_bdev(dev_t *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 820 | { |
| 821 | int dev; |
| 822 | int error; |
| 823 | |
| 824 | retry: |
Alexey Dobriyan | ad76cbc | 2008-08-28 06:26:23 +0400 | [diff] [blame] | 825 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 826 | return -ENOMEM; |
| 827 | spin_lock(&unnamed_dev_lock); |
Al Viro | c63e09e | 2009-06-24 02:05:18 -0400 | [diff] [blame] | 828 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
Al Viro | f21f622 | 2009-06-24 03:12:00 -0400 | [diff] [blame] | 829 | if (!error) |
| 830 | unnamed_dev_start = dev + 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 831 | spin_unlock(&unnamed_dev_lock); |
| 832 | if (error == -EAGAIN) |
| 833 | /* We raced and lost with another CPU. */ |
| 834 | goto retry; |
| 835 | else if (error) |
| 836 | return -EAGAIN; |
| 837 | |
Tejun Heo | e8c8d1b | 2013-02-27 17:05:04 -0800 | [diff] [blame] | 838 | if (dev == (1 << MINORBITS)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 839 | spin_lock(&unnamed_dev_lock); |
Alexey Dobriyan | ad76cbc | 2008-08-28 06:26:23 +0400 | [diff] [blame] | 840 | ida_remove(&unnamed_dev_ida, dev); |
Al Viro | f21f622 | 2009-06-24 03:12:00 -0400 | [diff] [blame] | 841 | if (unnamed_dev_start > dev) |
| 842 | unnamed_dev_start = dev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 843 | spin_unlock(&unnamed_dev_lock); |
| 844 | return -EMFILE; |
| 845 | } |
Al Viro | 0ee5dc6 | 2011-07-07 15:44:25 -0400 | [diff] [blame] | 846 | *p = MKDEV(0, dev & MINORMASK); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 847 | return 0; |
| 848 | } |
Al Viro | 0ee5dc6 | 2011-07-07 15:44:25 -0400 | [diff] [blame] | 849 | EXPORT_SYMBOL(get_anon_bdev); |
| 850 | |
| 851 | void free_anon_bdev(dev_t dev) |
| 852 | { |
| 853 | int slot = MINOR(dev); |
| 854 | spin_lock(&unnamed_dev_lock); |
| 855 | ida_remove(&unnamed_dev_ida, slot); |
| 856 | if (slot < unnamed_dev_start) |
| 857 | unnamed_dev_start = slot; |
| 858 | spin_unlock(&unnamed_dev_lock); |
| 859 | } |
| 860 | EXPORT_SYMBOL(free_anon_bdev); |
| 861 | |
| 862 | int set_anon_super(struct super_block *s, void *data) |
| 863 | { |
| 864 | int error = get_anon_bdev(&s->s_dev); |
| 865 | if (!error) |
| 866 | s->s_bdi = &noop_backing_dev_info; |
| 867 | return error; |
| 868 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 869 | |
| 870 | EXPORT_SYMBOL(set_anon_super); |
| 871 | |
| 872 | void kill_anon_super(struct super_block *sb) |
| 873 | { |
Al Viro | 0ee5dc6 | 2011-07-07 15:44:25 -0400 | [diff] [blame] | 874 | dev_t dev = sb->s_dev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 875 | generic_shutdown_super(sb); |
Al Viro | 0ee5dc6 | 2011-07-07 15:44:25 -0400 | [diff] [blame] | 876 | free_anon_bdev(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 877 | } |
| 878 | |
| 879 | EXPORT_SYMBOL(kill_anon_super); |
| 880 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 881 | void kill_litter_super(struct super_block *sb) |
| 882 | { |
| 883 | if (sb->s_root) |
| 884 | d_genocide(sb->s_root); |
| 885 | kill_anon_super(sb); |
| 886 | } |
| 887 | |
| 888 | EXPORT_SYMBOL(kill_litter_super); |
| 889 | |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 890 | static int ns_test_super(struct super_block *sb, void *data) |
| 891 | { |
| 892 | return sb->s_fs_info == data; |
| 893 | } |
| 894 | |
| 895 | static int ns_set_super(struct super_block *sb, void *data) |
| 896 | { |
| 897 | sb->s_fs_info = data; |
| 898 | return set_anon_super(sb, NULL); |
| 899 | } |
| 900 | |
Al Viro | ceefda6 | 2010-07-26 13:16:50 +0400 | [diff] [blame] | 901 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
| 902 | void *data, int (*fill_super)(struct super_block *, void *, int)) |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 903 | { |
| 904 | struct super_block *sb; |
| 905 | |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 906 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 907 | if (IS_ERR(sb)) |
Al Viro | ceefda6 | 2010-07-26 13:16:50 +0400 | [diff] [blame] | 908 | return ERR_CAST(sb); |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 909 | |
| 910 | if (!sb->s_root) { |
| 911 | int err; |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 912 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
| 913 | if (err) { |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 914 | deactivate_locked_super(sb); |
Al Viro | ceefda6 | 2010-07-26 13:16:50 +0400 | [diff] [blame] | 915 | return ERR_PTR(err); |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 916 | } |
| 917 | |
| 918 | sb->s_flags |= MS_ACTIVE; |
| 919 | } |
| 920 | |
Al Viro | ceefda6 | 2010-07-26 13:16:50 +0400 | [diff] [blame] | 921 | return dget(sb->s_root); |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 922 | } |
| 923 | |
Al Viro | ceefda6 | 2010-07-26 13:16:50 +0400 | [diff] [blame] | 924 | EXPORT_SYMBOL(mount_ns); |
Serge E. Hallyn | 909e6d9 | 2009-04-06 19:01:07 -0700 | [diff] [blame] | 925 | |
David Howells | 9361401 | 2006-09-30 20:45:40 +0200 | [diff] [blame] | 926 | #ifdef CONFIG_BLOCK |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 927 | static int set_bdev_super(struct super_block *s, void *data) |
| 928 | { |
| 929 | s->s_bdev = data; |
| 930 | s->s_dev = s->s_bdev->bd_dev; |
Jens Axboe | 32a88aa | 2009-09-16 15:02:33 +0200 | [diff] [blame] | 931 | |
| 932 | /* |
| 933 | * We set the bdi here to the queue backing, file systems can |
| 934 | * overwrite this in ->fill_super() |
| 935 | */ |
| 936 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 937 | return 0; |
| 938 | } |
| 939 | |
| 940 | static int test_bdev_super(struct super_block *s, void *data) |
| 941 | { |
| 942 | return (void *)s->s_bdev == data; |
| 943 | } |
| 944 | |
Al Viro | 152a083 | 2010-07-25 00:46:55 +0400 | [diff] [blame] | 945 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 946 | int flags, const char *dev_name, void *data, |
Al Viro | 152a083 | 2010-07-25 00:46:55 +0400 | [diff] [blame] | 947 | int (*fill_super)(struct super_block *, void *, int)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 948 | { |
| 949 | struct block_device *bdev; |
| 950 | struct super_block *s; |
Tejun Heo | d4d7762 | 2010-11-13 11:55:18 +0100 | [diff] [blame] | 951 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 952 | int error = 0; |
| 953 | |
Al Viro | 30c40d2 | 2008-02-22 19:50:45 -0500 | [diff] [blame] | 954 | if (!(flags & MS_RDONLY)) |
| 955 | mode |= FMODE_WRITE; |
| 956 | |
Tejun Heo | d4d7762 | 2010-11-13 11:55:18 +0100 | [diff] [blame] | 957 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 958 | if (IS_ERR(bdev)) |
Al Viro | 152a083 | 2010-07-25 00:46:55 +0400 | [diff] [blame] | 959 | return ERR_CAST(bdev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 960 | |
| 961 | /* |
| 962 | * once the super is inserted into the list by sget, s_umount |
| 963 | * will protect the lockfs code from trying to start a snapshot |
| 964 | * while we are mounting |
| 965 | */ |
Christoph Hellwig | 4fadd7b | 2009-08-03 23:28:06 +0200 | [diff] [blame] | 966 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
| 967 | if (bdev->bd_fsfreeze_count > 0) { |
| 968 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
| 969 | error = -EBUSY; |
| 970 | goto error_bdev; |
| 971 | } |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 972 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
| 973 | bdev); |
Christoph Hellwig | 4fadd7b | 2009-08-03 23:28:06 +0200 | [diff] [blame] | 974 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 975 | if (IS_ERR(s)) |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 976 | goto error_s; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 977 | |
| 978 | if (s->s_root) { |
| 979 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 980 | deactivate_locked_super(s); |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 981 | error = -EBUSY; |
| 982 | goto error_bdev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 983 | } |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 984 | |
Tejun Heo | 4f331f0 | 2010-07-20 15:18:07 -0700 | [diff] [blame] | 985 | /* |
| 986 | * s_umount nests inside bd_mutex during |
Tejun Heo | e525fd8 | 2010-11-13 11:55:17 +0100 | [diff] [blame] | 987 | * __invalidate_device(). blkdev_put() acquires |
| 988 | * bd_mutex and can't be called under s_umount. Drop |
| 989 | * s_umount temporarily. This is safe as we're |
| 990 | * holding an active reference. |
Tejun Heo | 4f331f0 | 2010-07-20 15:18:07 -0700 | [diff] [blame] | 991 | */ |
| 992 | up_write(&s->s_umount); |
Tejun Heo | d4d7762 | 2010-11-13 11:55:18 +0100 | [diff] [blame] | 993 | blkdev_put(bdev, mode); |
Tejun Heo | 4f331f0 | 2010-07-20 15:18:07 -0700 | [diff] [blame] | 994 | down_write(&s->s_umount); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 995 | } else { |
| 996 | char b[BDEVNAME_SIZE]; |
| 997 | |
Al Viro | 30c40d2 | 2008-02-22 19:50:45 -0500 | [diff] [blame] | 998 | s->s_mode = mode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 999 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
Pekka Enberg | e78c9a0 | 2006-01-08 01:03:39 -0800 | [diff] [blame] | 1000 | sb_set_blocksize(s, block_size(bdev)); |
Theodore Ts'o | 9b04c99 | 2006-03-24 03:15:10 -0800 | [diff] [blame] | 1001 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1002 | if (error) { |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 1003 | deactivate_locked_super(s); |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1004 | goto error; |
Greg Kroah-Hartman | fa67576 | 2006-02-22 09:39:02 -0800 | [diff] [blame] | 1005 | } |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1006 | |
| 1007 | s->s_flags |= MS_ACTIVE; |
Theodore Ts'o | 87d8fe1 | 2009-01-03 09:47:09 -0500 | [diff] [blame] | 1008 | bdev->bd_super = s; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1009 | } |
| 1010 | |
Al Viro | 152a083 | 2010-07-25 00:46:55 +0400 | [diff] [blame] | 1011 | return dget(s->s_root); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1012 | |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1013 | error_s: |
| 1014 | error = PTR_ERR(s); |
| 1015 | error_bdev: |
Tejun Heo | d4d7762 | 2010-11-13 11:55:18 +0100 | [diff] [blame] | 1016 | blkdev_put(bdev, mode); |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1017 | error: |
Al Viro | 152a083 | 2010-07-25 00:46:55 +0400 | [diff] [blame] | 1018 | return ERR_PTR(error); |
| 1019 | } |
| 1020 | EXPORT_SYMBOL(mount_bdev); |
| 1021 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1022 | void kill_block_super(struct super_block *sb) |
| 1023 | { |
| 1024 | struct block_device *bdev = sb->s_bdev; |
Al Viro | 30c40d2 | 2008-02-22 19:50:45 -0500 | [diff] [blame] | 1025 | fmode_t mode = sb->s_mode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1026 | |
H Hartley Sweeten | ddbaaf3 | 2009-04-29 20:14:57 -0400 | [diff] [blame] | 1027 | bdev->bd_super = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1028 | generic_shutdown_super(sb); |
| 1029 | sync_blockdev(bdev); |
Tejun Heo | d4d7762 | 2010-11-13 11:55:18 +0100 | [diff] [blame] | 1030 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
Tejun Heo | e525fd8 | 2010-11-13 11:55:17 +0100 | [diff] [blame] | 1031 | blkdev_put(bdev, mode | FMODE_EXCL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1032 | } |
| 1033 | |
| 1034 | EXPORT_SYMBOL(kill_block_super); |
David Howells | 9361401 | 2006-09-30 20:45:40 +0200 | [diff] [blame] | 1035 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1036 | |
Al Viro | 3c26ff6 | 2010-07-25 11:46:36 +0400 | [diff] [blame] | 1037 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1038 | int flags, void *data, |
Al Viro | 3c26ff6 | 2010-07-25 11:46:36 +0400 | [diff] [blame] | 1039 | int (*fill_super)(struct super_block *, void *, int)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1040 | { |
| 1041 | int error; |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 1042 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1043 | |
| 1044 | if (IS_ERR(s)) |
Al Viro | 3c26ff6 | 2010-07-25 11:46:36 +0400 | [diff] [blame] | 1045 | return ERR_CAST(s); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1046 | |
Theodore Ts'o | 9b04c99 | 2006-03-24 03:15:10 -0800 | [diff] [blame] | 1047 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1048 | if (error) { |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 1049 | deactivate_locked_super(s); |
Al Viro | 3c26ff6 | 2010-07-25 11:46:36 +0400 | [diff] [blame] | 1050 | return ERR_PTR(error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1051 | } |
| 1052 | s->s_flags |= MS_ACTIVE; |
Al Viro | 3c26ff6 | 2010-07-25 11:46:36 +0400 | [diff] [blame] | 1053 | return dget(s->s_root); |
| 1054 | } |
| 1055 | EXPORT_SYMBOL(mount_nodev); |
| 1056 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1057 | static int compare_single(struct super_block *s, void *p) |
| 1058 | { |
| 1059 | return 1; |
| 1060 | } |
| 1061 | |
Al Viro | fc14f2f | 2010-07-25 01:48:30 +0400 | [diff] [blame] | 1062 | struct dentry *mount_single(struct file_system_type *fs_type, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1063 | int flags, void *data, |
Al Viro | fc14f2f | 2010-07-25 01:48:30 +0400 | [diff] [blame] | 1064 | int (*fill_super)(struct super_block *, void *, int)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1065 | { |
| 1066 | struct super_block *s; |
| 1067 | int error; |
| 1068 | |
David Howells | 9249e17 | 2012-06-25 12:55:37 +0100 | [diff] [blame] | 1069 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1070 | if (IS_ERR(s)) |
Al Viro | fc14f2f | 2010-07-25 01:48:30 +0400 | [diff] [blame] | 1071 | return ERR_CAST(s); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1072 | if (!s->s_root) { |
Theodore Ts'o | 9b04c99 | 2006-03-24 03:15:10 -0800 | [diff] [blame] | 1073 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1074 | if (error) { |
Al Viro | 74dbbdd | 2009-05-06 01:07:50 -0400 | [diff] [blame] | 1075 | deactivate_locked_super(s); |
Al Viro | fc14f2f | 2010-07-25 01:48:30 +0400 | [diff] [blame] | 1076 | return ERR_PTR(error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1077 | } |
| 1078 | s->s_flags |= MS_ACTIVE; |
Kay Sievers | 9329d1b | 2009-12-18 21:18:15 +0100 | [diff] [blame] | 1079 | } else { |
| 1080 | do_remount_sb(s, flags, data, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1081 | } |
Al Viro | fc14f2f | 2010-07-25 01:48:30 +0400 | [diff] [blame] | 1082 | return dget(s->s_root); |
| 1083 | } |
| 1084 | EXPORT_SYMBOL(mount_single); |
| 1085 | |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1086 | struct dentry * |
| 1087 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1088 | { |
Al Viro | c96e41e | 2010-07-25 00:17:56 +0400 | [diff] [blame] | 1089 | struct dentry *root; |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1090 | struct super_block *sb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1091 | char *secdata = NULL; |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1092 | int error = -ENOMEM; |
Al Viro | 8089352 | 2010-02-05 09:30:46 -0500 | [diff] [blame] | 1093 | |
Eric Paris | e000752 | 2008-03-05 10:31:54 -0500 | [diff] [blame] | 1094 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1095 | secdata = alloc_secdata(); |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1096 | if (!secdata) |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1097 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1098 | |
Eric Paris | e000752 | 2008-03-05 10:31:54 -0500 | [diff] [blame] | 1099 | error = security_sb_copy_data(data, secdata); |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1100 | if (error) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1101 | goto out_free_secdata; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1102 | } |
| 1103 | |
Al Viro | 1a102ff | 2011-03-16 09:07:58 -0400 | [diff] [blame] | 1104 | root = type->mount(type, flags, name, data); |
| 1105 | if (IS_ERR(root)) { |
| 1106 | error = PTR_ERR(root); |
| 1107 | goto out_free_secdata; |
Al Viro | c96e41e | 2010-07-25 00:17:56 +0400 | [diff] [blame] | 1108 | } |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1109 | sb = root->d_sb; |
| 1110 | BUG_ON(!sb); |
| 1111 | WARN_ON(!sb->s_bdi); |
Linus Torvalds | 6c51038 | 2011-03-24 10:16:26 -0700 | [diff] [blame] | 1112 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1113 | sb->s_flags |= MS_BORN; |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1114 | |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1115 | error = security_sb_kern_mount(sb, flags, secdata); |
Jörn Engel | 5129a46 | 2010-04-25 08:54:42 +0200 | [diff] [blame] | 1116 | if (error) |
| 1117 | goto out_sb; |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1118 | |
Jeff Layton | 42cb56a | 2009-09-18 13:05:53 -0700 | [diff] [blame] | 1119 | /* |
| 1120 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE |
| 1121 | * but s_maxbytes was an unsigned long long for many releases. Throw |
| 1122 | * this warning for a little while to try and catch filesystems that |
Jeff Layton | 4358b56 | 2011-03-29 09:33:31 -0400 | [diff] [blame] | 1123 | * violate this rule. |
Jeff Layton | 42cb56a | 2009-09-18 13:05:53 -0700 | [diff] [blame] | 1124 | */ |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1125 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
| 1126 | "negative value (%lld)\n", type->name, sb->s_maxbytes); |
Jeff Layton | 42cb56a | 2009-09-18 13:05:53 -0700 | [diff] [blame] | 1127 | |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1128 | up_write(&sb->s_umount); |
Gerald Schaefer | 8680e22 | 2005-06-21 17:15:16 -0700 | [diff] [blame] | 1129 | free_secdata(secdata); |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1130 | return root; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1131 | out_sb: |
Al Viro | 9d412a4 | 2011-03-17 22:08:28 -0400 | [diff] [blame] | 1132 | dput(root); |
| 1133 | deactivate_locked_super(sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1134 | out_free_secdata: |
| 1135 | free_secdata(secdata); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1136 | out: |
David Howells | 454e239 | 2006-06-23 02:02:57 -0700 | [diff] [blame] | 1137 | return ERR_PTR(error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1138 | } |
| 1139 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1140 | /* |
| 1141 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} |
| 1142 | * instead. |
| 1143 | */ |
| 1144 | void __sb_end_write(struct super_block *sb, int level) |
| 1145 | { |
| 1146 | percpu_counter_dec(&sb->s_writers.counter[level-1]); |
| 1147 | /* |
| 1148 | * Make sure s_writers are updated before we wake up waiters in |
| 1149 | * freeze_super(). |
| 1150 | */ |
| 1151 | smp_mb(); |
| 1152 | if (waitqueue_active(&sb->s_writers.wait)) |
| 1153 | wake_up(&sb->s_writers.wait); |
| 1154 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); |
| 1155 | } |
| 1156 | EXPORT_SYMBOL(__sb_end_write); |
| 1157 | |
| 1158 | #ifdef CONFIG_LOCKDEP |
| 1159 | /* |
| 1160 | * We want lockdep to tell us about possible deadlocks with freezing but |
| 1161 | * it's it bit tricky to properly instrument it. Getting a freeze protection |
| 1162 | * works as getting a read lock but there are subtle problems. XFS for example |
| 1163 | * gets freeze protection on internal level twice in some cases, which is OK |
| 1164 | * only because we already hold a freeze protection also on higher level. Due |
| 1165 | * to these cases we have to tell lockdep we are doing trylock when we |
| 1166 | * already hold a freeze protection for a higher freeze level. |
| 1167 | */ |
| 1168 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, |
| 1169 | unsigned long ip) |
| 1170 | { |
| 1171 | int i; |
| 1172 | |
| 1173 | if (!trylock) { |
| 1174 | for (i = 0; i < level - 1; i++) |
| 1175 | if (lock_is_held(&sb->s_writers.lock_map[i])) { |
| 1176 | trylock = true; |
| 1177 | break; |
| 1178 | } |
| 1179 | } |
| 1180 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); |
| 1181 | } |
| 1182 | #endif |
| 1183 | |
| 1184 | /* |
| 1185 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} |
| 1186 | * instead. |
| 1187 | */ |
| 1188 | int __sb_start_write(struct super_block *sb, int level, bool wait) |
| 1189 | { |
| 1190 | retry: |
| 1191 | if (unlikely(sb->s_writers.frozen >= level)) { |
| 1192 | if (!wait) |
| 1193 | return 0; |
| 1194 | wait_event(sb->s_writers.wait_unfrozen, |
| 1195 | sb->s_writers.frozen < level); |
| 1196 | } |
| 1197 | |
| 1198 | #ifdef CONFIG_LOCKDEP |
| 1199 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); |
| 1200 | #endif |
| 1201 | percpu_counter_inc(&sb->s_writers.counter[level-1]); |
| 1202 | /* |
| 1203 | * Make sure counter is updated before we check for frozen. |
| 1204 | * freeze_super() first sets frozen and then checks the counter. |
| 1205 | */ |
| 1206 | smp_mb(); |
| 1207 | if (unlikely(sb->s_writers.frozen >= level)) { |
| 1208 | __sb_end_write(sb, level); |
| 1209 | goto retry; |
| 1210 | } |
| 1211 | return 1; |
| 1212 | } |
| 1213 | EXPORT_SYMBOL(__sb_start_write); |
| 1214 | |
| 1215 | /** |
| 1216 | * sb_wait_write - wait until all writers to given file system finish |
| 1217 | * @sb: the super for which we wait |
| 1218 | * @level: type of writers we wait for (normal vs page fault) |
| 1219 | * |
| 1220 | * This function waits until there are no writers of given type to given file |
| 1221 | * system. Caller of this function should make sure there can be no new writers |
| 1222 | * of type @level before calling this function. Otherwise this function can |
| 1223 | * livelock. |
| 1224 | */ |
| 1225 | static void sb_wait_write(struct super_block *sb, int level) |
| 1226 | { |
| 1227 | s64 writers; |
| 1228 | |
| 1229 | /* |
| 1230 | * We just cycle-through lockdep here so that it does not complain |
| 1231 | * about returning with lock to userspace |
| 1232 | */ |
| 1233 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); |
| 1234 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); |
| 1235 | |
| 1236 | do { |
| 1237 | DEFINE_WAIT(wait); |
| 1238 | |
| 1239 | /* |
| 1240 | * We use a barrier in prepare_to_wait() to separate setting |
| 1241 | * of frozen and checking of the counter |
| 1242 | */ |
| 1243 | prepare_to_wait(&sb->s_writers.wait, &wait, |
| 1244 | TASK_UNINTERRUPTIBLE); |
| 1245 | |
| 1246 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); |
| 1247 | if (writers) |
| 1248 | schedule(); |
| 1249 | |
| 1250 | finish_wait(&sb->s_writers.wait, &wait); |
| 1251 | } while (writers); |
| 1252 | } |
| 1253 | |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1254 | /** |
Randy Dunlap | 7000d3c | 2010-05-24 22:22:34 -0700 | [diff] [blame] | 1255 | * freeze_super - lock the filesystem and force it into a consistent state |
| 1256 | * @sb: the super to lock |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1257 | * |
| 1258 | * Syncs the super to make sure the filesystem is consistent and calls the fs's |
| 1259 | * freeze_fs. Subsequent calls to this without first thawing the fs will return |
| 1260 | * -EBUSY. |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1261 | * |
| 1262 | * During this function, sb->s_writers.frozen goes through these values: |
| 1263 | * |
| 1264 | * SB_UNFROZEN: File system is normal, all writes progress as usual. |
| 1265 | * |
| 1266 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New |
| 1267 | * writes should be blocked, though page faults are still allowed. We wait for |
| 1268 | * all writes to complete and then proceed to the next stage. |
| 1269 | * |
| 1270 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked |
| 1271 | * but internal fs threads can still modify the filesystem (although they |
| 1272 | * should not dirty new pages or inodes), writeback can run etc. After waiting |
| 1273 | * for all running page faults we sync the filesystem which will clean all |
| 1274 | * dirty pages and inodes (no new dirty pages or inodes can be created when |
| 1275 | * sync is running). |
| 1276 | * |
| 1277 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs |
| 1278 | * modification are blocked (e.g. XFS preallocation truncation on inode |
| 1279 | * reclaim). This is usually implemented by blocking new transactions for |
| 1280 | * filesystems that have them and need this additional guard. After all |
| 1281 | * internal writers are finished we call ->freeze_fs() to finish filesystem |
| 1282 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is |
| 1283 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. |
| 1284 | * |
| 1285 | * sb->s_writers.frozen is protected by sb->s_umount. |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1286 | */ |
| 1287 | int freeze_super(struct super_block *sb) |
| 1288 | { |
| 1289 | int ret; |
| 1290 | |
| 1291 | atomic_inc(&sb->s_active); |
| 1292 | down_write(&sb->s_umount); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1293 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1294 | deactivate_locked_super(sb); |
| 1295 | return -EBUSY; |
| 1296 | } |
| 1297 | |
Al Viro | dabe0dc | 2012-01-03 21:01:29 -0500 | [diff] [blame] | 1298 | if (!(sb->s_flags & MS_BORN)) { |
| 1299 | up_write(&sb->s_umount); |
| 1300 | return 0; /* sic - it's "nothing to do" */ |
| 1301 | } |
| 1302 | |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1303 | if (sb->s_flags & MS_RDONLY) { |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1304 | /* Nothing to do really... */ |
| 1305 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1306 | up_write(&sb->s_umount); |
| 1307 | return 0; |
| 1308 | } |
| 1309 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1310 | /* From now on, no new normal writers can start */ |
| 1311 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1312 | smp_wmb(); |
| 1313 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1314 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
| 1315 | up_write(&sb->s_umount); |
| 1316 | |
| 1317 | sb_wait_write(sb, SB_FREEZE_WRITE); |
| 1318 | |
| 1319 | /* Now we go and block page faults... */ |
| 1320 | down_write(&sb->s_umount); |
| 1321 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
| 1322 | smp_wmb(); |
| 1323 | |
| 1324 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
| 1325 | |
| 1326 | /* All writers are done so after syncing there won't be dirty data */ |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1327 | sync_filesystem(sb); |
| 1328 | |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1329 | /* Now wait for internal filesystem counter */ |
| 1330 | sb->s_writers.frozen = SB_FREEZE_FS; |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1331 | smp_wmb(); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1332 | sb_wait_write(sb, SB_FREEZE_FS); |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1333 | |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1334 | if (sb->s_op->freeze_fs) { |
| 1335 | ret = sb->s_op->freeze_fs(sb); |
| 1336 | if (ret) { |
| 1337 | printk(KERN_ERR |
| 1338 | "VFS:Filesystem freeze failed\n"); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1339 | sb->s_writers.frozen = SB_UNFROZEN; |
Kazuya Mio | e161630 | 2011-12-01 16:51:07 +0900 | [diff] [blame] | 1340 | smp_wmb(); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1341 | wake_up(&sb->s_writers.wait_unfrozen); |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1342 | deactivate_locked_super(sb); |
| 1343 | return ret; |
| 1344 | } |
| 1345 | } |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1346 | /* |
| 1347 | * This is just for debugging purposes so that fs can warn if it |
| 1348 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. |
| 1349 | */ |
| 1350 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1351 | up_write(&sb->s_umount); |
| 1352 | return 0; |
| 1353 | } |
| 1354 | EXPORT_SYMBOL(freeze_super); |
| 1355 | |
| 1356 | /** |
| 1357 | * thaw_super -- unlock filesystem |
| 1358 | * @sb: the super to thaw |
| 1359 | * |
| 1360 | * Unlocks the filesystem and marks it writeable again after freeze_super(). |
| 1361 | */ |
| 1362 | int thaw_super(struct super_block *sb) |
| 1363 | { |
| 1364 | int error; |
| 1365 | |
| 1366 | down_write(&sb->s_umount); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1367 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1368 | up_write(&sb->s_umount); |
| 1369 | return -EINVAL; |
| 1370 | } |
| 1371 | |
| 1372 | if (sb->s_flags & MS_RDONLY) |
| 1373 | goto out; |
| 1374 | |
| 1375 | if (sb->s_op->unfreeze_fs) { |
| 1376 | error = sb->s_op->unfreeze_fs(sb); |
| 1377 | if (error) { |
| 1378 | printk(KERN_ERR |
| 1379 | "VFS:Filesystem thaw failed\n"); |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1380 | up_write(&sb->s_umount); |
| 1381 | return error; |
| 1382 | } |
| 1383 | } |
| 1384 | |
| 1385 | out: |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1386 | sb->s_writers.frozen = SB_UNFROZEN; |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1387 | smp_wmb(); |
Jan Kara | 5accdf8 | 2012-06-12 16:20:34 +0200 | [diff] [blame] | 1388 | wake_up(&sb->s_writers.wait_unfrozen); |
Josef Bacik | 18e9e51 | 2010-03-23 10:34:56 -0400 | [diff] [blame] | 1389 | deactivate_locked_super(sb); |
| 1390 | |
| 1391 | return 0; |
| 1392 | } |
| 1393 | EXPORT_SYMBOL(thaw_super); |