Commit 2f11e3a821aff1a50c8ebc2fadbeb22c1ea1adbd
Committed by
Jiri Slaby
1 parent
94109cd2a5
Exists in
ti-linux-3.12.y
and in
2 other branches
fs/superblock: unregister sb shrinker before ->kill_sb()
commit 28f2cd4f6da24a1aa06c226618ed5ad69e13df64 upstream. This series is aimed at regressions noticed during reclaim activity. The first two patches are shrinker patches that were posted ages ago but never merged for reasons that are unclear to me. I'm posting them again to see if there was a reason they were dropped or if they just got lost. Dave? Time? The last patch adjusts proportional reclaim. Yuanhan Liu, can you retest the vm scalability test cases on a larger machine? Hugh, does this work for you on the memcg test cases? Based on ext4, I get the following results but unfortunately my larger test machines are all unavailable so this is based on a relatively small machine. postmark 3.15.0-rc5 3.15.0-rc5 vanilla proportion-v1r4 Ops/sec Transactions 21.00 ( 0.00%) 25.00 ( 19.05%) Ops/sec FilesCreate 39.00 ( 0.00%) 45.00 ( 15.38%) Ops/sec CreateTransact 10.00 ( 0.00%) 12.00 ( 20.00%) Ops/sec FilesDeleted 6202.00 ( 0.00%) 6202.00 ( 0.00%) Ops/sec DeleteTransact 11.00 ( 0.00%) 12.00 ( 9.09%) Ops/sec DataRead/MB 25.97 ( 0.00%) 30.02 ( 15.59%) Ops/sec DataWrite/MB 49.99 ( 0.00%) 57.78 ( 15.58%) ffsb (mail server simulator) 3.15.0-rc5 3.15.0-rc5 vanilla proportion-v1r4 Ops/sec readall 9402.63 ( 0.00%) 9805.74 ( 4.29%) Ops/sec create 4695.45 ( 0.00%) 4781.39 ( 1.83%) Ops/sec delete 173.72 ( 0.00%) 177.23 ( 2.02%) Ops/sec Transactions 14271.80 ( 0.00%) 14764.37 ( 3.45%) Ops/sec Read 37.00 ( 0.00%) 38.50 ( 4.05%) Ops/sec Write 18.20 ( 0.00%) 18.50 ( 1.65%) dd of a large file 3.15.0-rc5 3.15.0-rc5 vanilla proportion-v1r4 WallTime DownloadTar 75.00 ( 0.00%) 61.00 ( 18.67%) WallTime DD 423.00 ( 0.00%) 401.00 ( 5.20%) WallTime Delete 2.00 ( 0.00%) 5.00 (-150.00%) stutter (times mmap latency during large amounts of IO) 3.15.0-rc5 3.15.0-rc5 vanilla proportion-v1r4 Unit >5ms Delays 80252.0000 ( 0.00%) 81523.0000 ( -1.58%) Unit Mmap min 8.2118 ( 0.00%) 8.3206 ( -1.33%) Unit Mmap mean 17.4614 ( 0.00%) 17.2868 ( 1.00%) Unit Mmap stddev 24.9059 ( 0.00%) 34.6771 (-39.23%) Unit Mmap max 2811.6433 ( 0.00%) 2645.1398 ( 5.92%) Unit Mmap 90% 20.5098 ( 0.00%) 18.3105 ( 10.72%) Unit Mmap 93% 22.9180 ( 0.00%) 20.1751 ( 11.97%) Unit Mmap 95% 25.2114 ( 0.00%) 22.4988 ( 10.76%) Unit Mmap 99% 46.1430 ( 0.00%) 43.5952 ( 5.52%) Unit Ideal Tput 85.2623 ( 0.00%) 78.8906 ( 7.47%) Unit Tput min 44.0666 ( 0.00%) 43.9609 ( 0.24%) Unit Tput mean 45.5646 ( 0.00%) 45.2009 ( 0.80%) Unit Tput stddev 0.9318 ( 0.00%) 1.1084 (-18.95%) Unit Tput max 46.7375 ( 0.00%) 46.7539 ( -0.04%) This patch (of 3): We will like to unregister the sb shrinker before ->kill_sb(). This will allow cached objects to be counted without call to grab_super_passive() to update ref count on sb. We want to avoid locking during memory reclamation especially when we are skipping the memory reclaim when we are out of cached objects. This is safe because grab_super_passive does a try-lock on the sb->s_umount now, and so if we are in the unmount process, it won't ever block. That means what used to be a deadlock and races we were avoiding by using grab_super_passive() is now: shrinker umount down_read(shrinker_rwsem) down_write(sb->s_umount) shrinker_unregister down_write(shrinker_rwsem) <blocks> grab_super_passive(sb) down_read_trylock(sb->s_umount) <fails> <shrinker aborts> .... <shrinkers finish running> up_read(shrinker_rwsem) <unblocks> <removes shrinker> up_write(shrinker_rwsem) ->kill_sb() .... So it is safe to deregister the shrinker before ->kill_sb(). Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Mel Gorman <mgorman@suse.de> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Hugh Dickins <hughd@google.com> Cc: Dave Chinner <david@fromorbit.com> Tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Cc: Bob Liu <bob.liu@oracle.com> Cc: Jan Kara <jack@suse.cz> Acked-by: Rik van Riel <riel@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Mel Gorman <mgorman@suse.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Showing 1 changed file with 1 additions and 3 deletions Inline Diff
fs/super.c
1 | /* | 1 | /* |
2 | * linux/fs/super.c | 2 | * linux/fs/super.c |
3 | * | 3 | * |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | 4 | * Copyright (C) 1991, 1992 Linus Torvalds |
5 | * | 5 | * |
6 | * super.c contains code to handle: - mount structures | 6 | * super.c contains code to handle: - mount structures |
7 | * - super-block tables | 7 | * - super-block tables |
8 | * - filesystem drivers list | 8 | * - filesystem drivers list |
9 | * - mount system call | 9 | * - mount system call |
10 | * - umount system call | 10 | * - umount system call |
11 | * - ustat system call | 11 | * - ustat system call |
12 | * | 12 | * |
13 | * GK 2/5/95 - Changed to support mounting the root fs via NFS | 13 | * GK 2/5/95 - Changed to support mounting the root fs via NFS |
14 | * | 14 | * |
15 | * Added kerneld support: Jacques Gelinas and Bjorn Ekwall | 15 | * Added kerneld support: Jacques Gelinas and Bjorn Ekwall |
16 | * Added change_root: Werner Almesberger & Hans Lermen, Feb '96 | 16 | * Added change_root: Werner Almesberger & Hans Lermen, Feb '96 |
17 | * Added options to /proc/mounts: | 17 | * Added options to /proc/mounts: |
18 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. | 18 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. |
19 | * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998 | 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 | 20 | * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000 |
21 | */ | 21 | */ |
22 | 22 | ||
23 | #include <linux/export.h> | 23 | #include <linux/export.h> |
24 | #include <linux/slab.h> | 24 | #include <linux/slab.h> |
25 | #include <linux/acct.h> | 25 | #include <linux/acct.h> |
26 | #include <linux/blkdev.h> | 26 | #include <linux/blkdev.h> |
27 | #include <linux/mount.h> | 27 | #include <linux/mount.h> |
28 | #include <linux/security.h> | 28 | #include <linux/security.h> |
29 | #include <linux/writeback.h> /* for the emergency remount stuff */ | 29 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
30 | #include <linux/idr.h> | 30 | #include <linux/idr.h> |
31 | #include <linux/mutex.h> | 31 | #include <linux/mutex.h> |
32 | #include <linux/backing-dev.h> | 32 | #include <linux/backing-dev.h> |
33 | #include <linux/rculist_bl.h> | 33 | #include <linux/rculist_bl.h> |
34 | #include <linux/cleancache.h> | 34 | #include <linux/cleancache.h> |
35 | #include <linux/fsnotify.h> | 35 | #include <linux/fsnotify.h> |
36 | #include <linux/lockdep.h> | 36 | #include <linux/lockdep.h> |
37 | #include "internal.h" | 37 | #include "internal.h" |
38 | 38 | ||
39 | 39 | ||
40 | LIST_HEAD(super_blocks); | 40 | LIST_HEAD(super_blocks); |
41 | DEFINE_SPINLOCK(sb_lock); | 41 | DEFINE_SPINLOCK(sb_lock); |
42 | 42 | ||
43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { | 43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
44 | "sb_writers", | 44 | "sb_writers", |
45 | "sb_pagefaults", | 45 | "sb_pagefaults", |
46 | "sb_internal", | 46 | "sb_internal", |
47 | }; | 47 | }; |
48 | 48 | ||
49 | /* | 49 | /* |
50 | * One thing we have to be careful of with a per-sb shrinker is that we don't | 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. | 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 | 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 | 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. | 54 | * take a passive reference to the superblock to avoid this from occurring. |
55 | */ | 55 | */ |
56 | static unsigned long super_cache_scan(struct shrinker *shrink, | 56 | static unsigned long super_cache_scan(struct shrinker *shrink, |
57 | struct shrink_control *sc) | 57 | struct shrink_control *sc) |
58 | { | 58 | { |
59 | struct super_block *sb; | 59 | struct super_block *sb; |
60 | long fs_objects = 0; | 60 | long fs_objects = 0; |
61 | long total_objects; | 61 | long total_objects; |
62 | long freed = 0; | 62 | long freed = 0; |
63 | long dentries; | 63 | long dentries; |
64 | long inodes; | 64 | long inodes; |
65 | 65 | ||
66 | sb = container_of(shrink, struct super_block, s_shrink); | 66 | sb = container_of(shrink, struct super_block, s_shrink); |
67 | 67 | ||
68 | /* | 68 | /* |
69 | * Deadlock avoidance. We may hold various FS locks, and we don't want | 69 | * Deadlock avoidance. We may hold various FS locks, and we don't want |
70 | * to recurse into the FS that called us in clear_inode() and friends.. | 70 | * to recurse into the FS that called us in clear_inode() and friends.. |
71 | */ | 71 | */ |
72 | if (!(sc->gfp_mask & __GFP_FS)) | 72 | if (!(sc->gfp_mask & __GFP_FS)) |
73 | return SHRINK_STOP; | 73 | return SHRINK_STOP; |
74 | 74 | ||
75 | if (!grab_super_passive(sb)) | 75 | if (!grab_super_passive(sb)) |
76 | return SHRINK_STOP; | 76 | return SHRINK_STOP; |
77 | 77 | ||
78 | if (sb->s_op->nr_cached_objects) | 78 | if (sb->s_op->nr_cached_objects) |
79 | fs_objects = sb->s_op->nr_cached_objects(sb, sc->nid); | 79 | fs_objects = sb->s_op->nr_cached_objects(sb, sc->nid); |
80 | 80 | ||
81 | inodes = list_lru_count_node(&sb->s_inode_lru, sc->nid); | 81 | inodes = list_lru_count_node(&sb->s_inode_lru, sc->nid); |
82 | dentries = list_lru_count_node(&sb->s_dentry_lru, sc->nid); | 82 | dentries = list_lru_count_node(&sb->s_dentry_lru, sc->nid); |
83 | total_objects = dentries + inodes + fs_objects + 1; | 83 | total_objects = dentries + inodes + fs_objects + 1; |
84 | 84 | ||
85 | /* proportion the scan between the caches */ | 85 | /* proportion the scan between the caches */ |
86 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); | 86 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
87 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); | 87 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
88 | 88 | ||
89 | /* | 89 | /* |
90 | * prune the dcache first as the icache is pinned by it, then | 90 | * prune the dcache first as the icache is pinned by it, then |
91 | * prune the icache, followed by the filesystem specific caches | 91 | * prune the icache, followed by the filesystem specific caches |
92 | */ | 92 | */ |
93 | freed = prune_dcache_sb(sb, dentries, sc->nid); | 93 | freed = prune_dcache_sb(sb, dentries, sc->nid); |
94 | freed += prune_icache_sb(sb, inodes, sc->nid); | 94 | freed += prune_icache_sb(sb, inodes, sc->nid); |
95 | 95 | ||
96 | if (fs_objects) { | 96 | if (fs_objects) { |
97 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, | 97 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, |
98 | total_objects); | 98 | total_objects); |
99 | freed += sb->s_op->free_cached_objects(sb, fs_objects, | 99 | freed += sb->s_op->free_cached_objects(sb, fs_objects, |
100 | sc->nid); | 100 | sc->nid); |
101 | } | 101 | } |
102 | 102 | ||
103 | drop_super(sb); | 103 | drop_super(sb); |
104 | return freed; | 104 | return freed; |
105 | } | 105 | } |
106 | 106 | ||
107 | static unsigned long super_cache_count(struct shrinker *shrink, | 107 | static unsigned long super_cache_count(struct shrinker *shrink, |
108 | struct shrink_control *sc) | 108 | struct shrink_control *sc) |
109 | { | 109 | { |
110 | struct super_block *sb; | 110 | struct super_block *sb; |
111 | long total_objects = 0; | 111 | long total_objects = 0; |
112 | 112 | ||
113 | sb = container_of(shrink, struct super_block, s_shrink); | 113 | sb = container_of(shrink, struct super_block, s_shrink); |
114 | 114 | ||
115 | if (!grab_super_passive(sb)) | 115 | if (!grab_super_passive(sb)) |
116 | return 0; | 116 | return 0; |
117 | 117 | ||
118 | if (sb->s_op && sb->s_op->nr_cached_objects) | 118 | if (sb->s_op && sb->s_op->nr_cached_objects) |
119 | total_objects = sb->s_op->nr_cached_objects(sb, | 119 | total_objects = sb->s_op->nr_cached_objects(sb, |
120 | sc->nid); | 120 | sc->nid); |
121 | 121 | ||
122 | total_objects += list_lru_count_node(&sb->s_dentry_lru, | 122 | total_objects += list_lru_count_node(&sb->s_dentry_lru, |
123 | sc->nid); | 123 | sc->nid); |
124 | total_objects += list_lru_count_node(&sb->s_inode_lru, | 124 | total_objects += list_lru_count_node(&sb->s_inode_lru, |
125 | sc->nid); | 125 | sc->nid); |
126 | 126 | ||
127 | total_objects = vfs_pressure_ratio(total_objects); | 127 | total_objects = vfs_pressure_ratio(total_objects); |
128 | drop_super(sb); | 128 | drop_super(sb); |
129 | return total_objects; | 129 | return total_objects; |
130 | } | 130 | } |
131 | 131 | ||
132 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) | 132 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
133 | { | 133 | { |
134 | int err; | 134 | int err; |
135 | int i; | 135 | int i; |
136 | 136 | ||
137 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { | 137 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
138 | err = percpu_counter_init(&s->s_writers.counter[i], 0); | 138 | err = percpu_counter_init(&s->s_writers.counter[i], 0); |
139 | if (err < 0) | 139 | if (err < 0) |
140 | goto err_out; | 140 | goto err_out; |
141 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | 141 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], |
142 | &type->s_writers_key[i], 0); | 142 | &type->s_writers_key[i], 0); |
143 | } | 143 | } |
144 | init_waitqueue_head(&s->s_writers.wait); | 144 | init_waitqueue_head(&s->s_writers.wait); |
145 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | 145 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
146 | return 0; | 146 | return 0; |
147 | err_out: | 147 | err_out: |
148 | while (--i >= 0) | 148 | while (--i >= 0) |
149 | percpu_counter_destroy(&s->s_writers.counter[i]); | 149 | percpu_counter_destroy(&s->s_writers.counter[i]); |
150 | return err; | 150 | return err; |
151 | } | 151 | } |
152 | 152 | ||
153 | static void destroy_sb_writers(struct super_block *s) | 153 | static void destroy_sb_writers(struct super_block *s) |
154 | { | 154 | { |
155 | int i; | 155 | int i; |
156 | 156 | ||
157 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | 157 | for (i = 0; i < SB_FREEZE_LEVELS; i++) |
158 | percpu_counter_destroy(&s->s_writers.counter[i]); | 158 | percpu_counter_destroy(&s->s_writers.counter[i]); |
159 | } | 159 | } |
160 | 160 | ||
161 | /** | 161 | /** |
162 | * alloc_super - create new superblock | 162 | * alloc_super - create new superblock |
163 | * @type: filesystem type superblock should belong to | 163 | * @type: filesystem type superblock should belong to |
164 | * @flags: the mount flags | 164 | * @flags: the mount flags |
165 | * | 165 | * |
166 | * Allocates and initializes a new &struct super_block. alloc_super() | 166 | * Allocates and initializes a new &struct super_block. alloc_super() |
167 | * returns a pointer new superblock or %NULL if allocation had failed. | 167 | * returns a pointer new superblock or %NULL if allocation had failed. |
168 | */ | 168 | */ |
169 | static struct super_block *alloc_super(struct file_system_type *type, int flags) | 169 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
170 | { | 170 | { |
171 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); | 171 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
172 | static const struct super_operations default_op; | 172 | static const struct super_operations default_op; |
173 | 173 | ||
174 | if (s) { | 174 | if (s) { |
175 | if (security_sb_alloc(s)) | 175 | if (security_sb_alloc(s)) |
176 | goto out_free_sb; | 176 | goto out_free_sb; |
177 | 177 | ||
178 | #ifdef CONFIG_SMP | 178 | #ifdef CONFIG_SMP |
179 | s->s_files = alloc_percpu(struct list_head); | 179 | s->s_files = alloc_percpu(struct list_head); |
180 | if (!s->s_files) | 180 | if (!s->s_files) |
181 | goto err_out; | 181 | goto err_out; |
182 | else { | 182 | else { |
183 | int i; | 183 | int i; |
184 | 184 | ||
185 | for_each_possible_cpu(i) | 185 | for_each_possible_cpu(i) |
186 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | 186 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); |
187 | } | 187 | } |
188 | #else | 188 | #else |
189 | INIT_LIST_HEAD(&s->s_files); | 189 | INIT_LIST_HEAD(&s->s_files); |
190 | #endif | 190 | #endif |
191 | if (init_sb_writers(s, type)) | 191 | if (init_sb_writers(s, type)) |
192 | goto err_out; | 192 | goto err_out; |
193 | s->s_flags = flags; | 193 | s->s_flags = flags; |
194 | s->s_bdi = &default_backing_dev_info; | 194 | s->s_bdi = &default_backing_dev_info; |
195 | INIT_HLIST_NODE(&s->s_instances); | 195 | INIT_HLIST_NODE(&s->s_instances); |
196 | INIT_HLIST_BL_HEAD(&s->s_anon); | 196 | INIT_HLIST_BL_HEAD(&s->s_anon); |
197 | INIT_LIST_HEAD(&s->s_inodes); | 197 | INIT_LIST_HEAD(&s->s_inodes); |
198 | 198 | ||
199 | if (list_lru_init(&s->s_dentry_lru)) | 199 | if (list_lru_init(&s->s_dentry_lru)) |
200 | goto err_out; | 200 | goto err_out; |
201 | if (list_lru_init(&s->s_inode_lru)) | 201 | if (list_lru_init(&s->s_inode_lru)) |
202 | goto err_out_dentry_lru; | 202 | goto err_out_dentry_lru; |
203 | 203 | ||
204 | INIT_LIST_HEAD(&s->s_mounts); | 204 | INIT_LIST_HEAD(&s->s_mounts); |
205 | init_rwsem(&s->s_umount); | 205 | init_rwsem(&s->s_umount); |
206 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | 206 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
207 | /* | 207 | /* |
208 | * sget() can have s_umount recursion. | 208 | * sget() can have s_umount recursion. |
209 | * | 209 | * |
210 | * When it cannot find a suitable sb, it allocates a new | 210 | * When it cannot find a suitable sb, it allocates a new |
211 | * one (this one), and tries again to find a suitable old | 211 | * one (this one), and tries again to find a suitable old |
212 | * one. | 212 | * one. |
213 | * | 213 | * |
214 | * In case that succeeds, it will acquire the s_umount | 214 | * In case that succeeds, it will acquire the s_umount |
215 | * lock of the old one. Since these are clearly distrinct | 215 | * lock of the old one. Since these are clearly distrinct |
216 | * locks, and this object isn't exposed yet, there's no | 216 | * locks, and this object isn't exposed yet, there's no |
217 | * risk of deadlocks. | 217 | * risk of deadlocks. |
218 | * | 218 | * |
219 | * Annotate this by putting this lock in a different | 219 | * Annotate this by putting this lock in a different |
220 | * subclass. | 220 | * subclass. |
221 | */ | 221 | */ |
222 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | 222 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); |
223 | s->s_count = 1; | 223 | s->s_count = 1; |
224 | atomic_set(&s->s_active, 1); | 224 | atomic_set(&s->s_active, 1); |
225 | mutex_init(&s->s_vfs_rename_mutex); | 225 | mutex_init(&s->s_vfs_rename_mutex); |
226 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | 226 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
227 | mutex_init(&s->s_dquot.dqio_mutex); | 227 | mutex_init(&s->s_dquot.dqio_mutex); |
228 | mutex_init(&s->s_dquot.dqonoff_mutex); | 228 | mutex_init(&s->s_dquot.dqonoff_mutex); |
229 | init_rwsem(&s->s_dquot.dqptr_sem); | 229 | init_rwsem(&s->s_dquot.dqptr_sem); |
230 | s->s_maxbytes = MAX_NON_LFS; | 230 | s->s_maxbytes = MAX_NON_LFS; |
231 | s->s_op = &default_op; | 231 | s->s_op = &default_op; |
232 | s->s_time_gran = 1000000000; | 232 | s->s_time_gran = 1000000000; |
233 | s->cleancache_poolid = -1; | 233 | s->cleancache_poolid = -1; |
234 | 234 | ||
235 | s->s_shrink.seeks = DEFAULT_SEEKS; | 235 | s->s_shrink.seeks = DEFAULT_SEEKS; |
236 | s->s_shrink.scan_objects = super_cache_scan; | 236 | s->s_shrink.scan_objects = super_cache_scan; |
237 | s->s_shrink.count_objects = super_cache_count; | 237 | s->s_shrink.count_objects = super_cache_count; |
238 | s->s_shrink.batch = 1024; | 238 | s->s_shrink.batch = 1024; |
239 | s->s_shrink.flags = SHRINKER_NUMA_AWARE; | 239 | s->s_shrink.flags = SHRINKER_NUMA_AWARE; |
240 | } | 240 | } |
241 | out: | 241 | out: |
242 | return s; | 242 | return s; |
243 | 243 | ||
244 | err_out_dentry_lru: | 244 | err_out_dentry_lru: |
245 | list_lru_destroy(&s->s_dentry_lru); | 245 | list_lru_destroy(&s->s_dentry_lru); |
246 | err_out: | 246 | err_out: |
247 | security_sb_free(s); | 247 | security_sb_free(s); |
248 | #ifdef CONFIG_SMP | 248 | #ifdef CONFIG_SMP |
249 | if (s->s_files) | 249 | if (s->s_files) |
250 | free_percpu(s->s_files); | 250 | free_percpu(s->s_files); |
251 | #endif | 251 | #endif |
252 | destroy_sb_writers(s); | 252 | destroy_sb_writers(s); |
253 | out_free_sb: | 253 | out_free_sb: |
254 | kfree(s); | 254 | kfree(s); |
255 | s = NULL; | 255 | s = NULL; |
256 | goto out; | 256 | goto out; |
257 | } | 257 | } |
258 | 258 | ||
259 | /** | 259 | /** |
260 | * destroy_super - frees a superblock | 260 | * destroy_super - frees a superblock |
261 | * @s: superblock to free | 261 | * @s: superblock to free |
262 | * | 262 | * |
263 | * Frees a superblock. | 263 | * Frees a superblock. |
264 | */ | 264 | */ |
265 | static inline void destroy_super(struct super_block *s) | 265 | static inline void destroy_super(struct super_block *s) |
266 | { | 266 | { |
267 | list_lru_destroy(&s->s_dentry_lru); | 267 | list_lru_destroy(&s->s_dentry_lru); |
268 | list_lru_destroy(&s->s_inode_lru); | 268 | list_lru_destroy(&s->s_inode_lru); |
269 | #ifdef CONFIG_SMP | 269 | #ifdef CONFIG_SMP |
270 | free_percpu(s->s_files); | 270 | free_percpu(s->s_files); |
271 | #endif | 271 | #endif |
272 | destroy_sb_writers(s); | 272 | destroy_sb_writers(s); |
273 | security_sb_free(s); | 273 | security_sb_free(s); |
274 | WARN_ON(!list_empty(&s->s_mounts)); | 274 | WARN_ON(!list_empty(&s->s_mounts)); |
275 | kfree(s->s_subtype); | 275 | kfree(s->s_subtype); |
276 | kfree(s->s_options); | 276 | kfree(s->s_options); |
277 | kfree(s); | 277 | kfree(s); |
278 | } | 278 | } |
279 | 279 | ||
280 | /* Superblock refcounting */ | 280 | /* Superblock refcounting */ |
281 | 281 | ||
282 | /* | 282 | /* |
283 | * Drop a superblock's refcount. The caller must hold sb_lock. | 283 | * Drop a superblock's refcount. The caller must hold sb_lock. |
284 | */ | 284 | */ |
285 | static void __put_super(struct super_block *sb) | 285 | static void __put_super(struct super_block *sb) |
286 | { | 286 | { |
287 | if (!--sb->s_count) { | 287 | if (!--sb->s_count) { |
288 | list_del_init(&sb->s_list); | 288 | list_del_init(&sb->s_list); |
289 | destroy_super(sb); | 289 | destroy_super(sb); |
290 | } | 290 | } |
291 | } | 291 | } |
292 | 292 | ||
293 | /** | 293 | /** |
294 | * put_super - drop a temporary reference to superblock | 294 | * put_super - drop a temporary reference to superblock |
295 | * @sb: superblock in question | 295 | * @sb: superblock in question |
296 | * | 296 | * |
297 | * Drops a temporary reference, frees superblock if there's no | 297 | * Drops a temporary reference, frees superblock if there's no |
298 | * references left. | 298 | * references left. |
299 | */ | 299 | */ |
300 | static void put_super(struct super_block *sb) | 300 | static void put_super(struct super_block *sb) |
301 | { | 301 | { |
302 | spin_lock(&sb_lock); | 302 | spin_lock(&sb_lock); |
303 | __put_super(sb); | 303 | __put_super(sb); |
304 | spin_unlock(&sb_lock); | 304 | spin_unlock(&sb_lock); |
305 | } | 305 | } |
306 | 306 | ||
307 | 307 | ||
308 | /** | 308 | /** |
309 | * deactivate_locked_super - drop an active reference to superblock | 309 | * deactivate_locked_super - drop an active reference to superblock |
310 | * @s: superblock to deactivate | 310 | * @s: superblock to deactivate |
311 | * | 311 | * |
312 | * Drops an active reference to superblock, converting it into a temprory | 312 | * Drops an active reference to superblock, converting it into a temprory |
313 | * one if there is no other active references left. In that case we | 313 | * one if there is no other active references left. In that case we |
314 | * tell fs driver to shut it down and drop the temporary reference we | 314 | * tell fs driver to shut it down and drop the temporary reference we |
315 | * had just acquired. | 315 | * had just acquired. |
316 | * | 316 | * |
317 | * Caller holds exclusive lock on superblock; that lock is released. | 317 | * Caller holds exclusive lock on superblock; that lock is released. |
318 | */ | 318 | */ |
319 | void deactivate_locked_super(struct super_block *s) | 319 | void deactivate_locked_super(struct super_block *s) |
320 | { | 320 | { |
321 | struct file_system_type *fs = s->s_type; | 321 | struct file_system_type *fs = s->s_type; |
322 | if (atomic_dec_and_test(&s->s_active)) { | 322 | if (atomic_dec_and_test(&s->s_active)) { |
323 | cleancache_invalidate_fs(s); | 323 | cleancache_invalidate_fs(s); |
324 | fs->kill_sb(s); | ||
325 | |||
326 | /* caches are now gone, we can safely kill the shrinker now */ | ||
327 | unregister_shrinker(&s->s_shrink); | 324 | unregister_shrinker(&s->s_shrink); |
325 | fs->kill_sb(s); | ||
328 | 326 | ||
329 | put_filesystem(fs); | 327 | put_filesystem(fs); |
330 | put_super(s); | 328 | put_super(s); |
331 | } else { | 329 | } else { |
332 | up_write(&s->s_umount); | 330 | up_write(&s->s_umount); |
333 | } | 331 | } |
334 | } | 332 | } |
335 | 333 | ||
336 | EXPORT_SYMBOL(deactivate_locked_super); | 334 | EXPORT_SYMBOL(deactivate_locked_super); |
337 | 335 | ||
338 | /** | 336 | /** |
339 | * deactivate_super - drop an active reference to superblock | 337 | * deactivate_super - drop an active reference to superblock |
340 | * @s: superblock to deactivate | 338 | * @s: superblock to deactivate |
341 | * | 339 | * |
342 | * Variant of deactivate_locked_super(), except that superblock is *not* | 340 | * Variant of deactivate_locked_super(), except that superblock is *not* |
343 | * locked by caller. If we are going to drop the final active reference, | 341 | * locked by caller. If we are going to drop the final active reference, |
344 | * lock will be acquired prior to that. | 342 | * lock will be acquired prior to that. |
345 | */ | 343 | */ |
346 | void deactivate_super(struct super_block *s) | 344 | void deactivate_super(struct super_block *s) |
347 | { | 345 | { |
348 | if (!atomic_add_unless(&s->s_active, -1, 1)) { | 346 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
349 | down_write(&s->s_umount); | 347 | down_write(&s->s_umount); |
350 | deactivate_locked_super(s); | 348 | deactivate_locked_super(s); |
351 | } | 349 | } |
352 | } | 350 | } |
353 | 351 | ||
354 | EXPORT_SYMBOL(deactivate_super); | 352 | EXPORT_SYMBOL(deactivate_super); |
355 | 353 | ||
356 | /** | 354 | /** |
357 | * grab_super - acquire an active reference | 355 | * grab_super - acquire an active reference |
358 | * @s: reference we are trying to make active | 356 | * @s: reference we are trying to make active |
359 | * | 357 | * |
360 | * Tries to acquire an active reference. grab_super() is used when we | 358 | * Tries to acquire an active reference. grab_super() is used when we |
361 | * had just found a superblock in super_blocks or fs_type->fs_supers | 359 | * had just found a superblock in super_blocks or fs_type->fs_supers |
362 | * and want to turn it into a full-blown active reference. grab_super() | 360 | * and want to turn it into a full-blown active reference. grab_super() |
363 | * is called with sb_lock held and drops it. Returns 1 in case of | 361 | * is called with sb_lock held and drops it. Returns 1 in case of |
364 | * success, 0 if we had failed (superblock contents was already dead or | 362 | * success, 0 if we had failed (superblock contents was already dead or |
365 | * dying when grab_super() had been called). Note that this is only | 363 | * dying when grab_super() had been called). Note that this is only |
366 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | 364 | * called for superblocks not in rundown mode (== ones still on ->fs_supers |
367 | * of their type), so increment of ->s_count is OK here. | 365 | * of their type), so increment of ->s_count is OK here. |
368 | */ | 366 | */ |
369 | static int grab_super(struct super_block *s) __releases(sb_lock) | 367 | static int grab_super(struct super_block *s) __releases(sb_lock) |
370 | { | 368 | { |
371 | s->s_count++; | 369 | s->s_count++; |
372 | spin_unlock(&sb_lock); | 370 | spin_unlock(&sb_lock); |
373 | down_write(&s->s_umount); | 371 | down_write(&s->s_umount); |
374 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { | 372 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
375 | put_super(s); | 373 | put_super(s); |
376 | return 1; | 374 | return 1; |
377 | } | 375 | } |
378 | up_write(&s->s_umount); | 376 | up_write(&s->s_umount); |
379 | put_super(s); | 377 | put_super(s); |
380 | return 0; | 378 | return 0; |
381 | } | 379 | } |
382 | 380 | ||
383 | /* | 381 | /* |
384 | * grab_super_passive - acquire a passive reference | 382 | * grab_super_passive - acquire a passive reference |
385 | * @sb: reference we are trying to grab | 383 | * @sb: reference we are trying to grab |
386 | * | 384 | * |
387 | * Tries to acquire a passive reference. This is used in places where we | 385 | * Tries to acquire a passive reference. This is used in places where we |
388 | * cannot take an active reference but we need to ensure that the | 386 | * cannot take an active reference but we need to ensure that the |
389 | * superblock does not go away while we are working on it. It returns | 387 | * superblock does not go away while we are working on it. It returns |
390 | * false if a reference was not gained, and returns true with the s_umount | 388 | * false if a reference was not gained, and returns true with the s_umount |
391 | * lock held in read mode if a reference is gained. On successful return, | 389 | * lock held in read mode if a reference is gained. On successful return, |
392 | * the caller must drop the s_umount lock and the passive reference when | 390 | * the caller must drop the s_umount lock and the passive reference when |
393 | * done. | 391 | * done. |
394 | */ | 392 | */ |
395 | bool grab_super_passive(struct super_block *sb) | 393 | bool grab_super_passive(struct super_block *sb) |
396 | { | 394 | { |
397 | spin_lock(&sb_lock); | 395 | spin_lock(&sb_lock); |
398 | if (hlist_unhashed(&sb->s_instances)) { | 396 | if (hlist_unhashed(&sb->s_instances)) { |
399 | spin_unlock(&sb_lock); | 397 | spin_unlock(&sb_lock); |
400 | return false; | 398 | return false; |
401 | } | 399 | } |
402 | 400 | ||
403 | sb->s_count++; | 401 | sb->s_count++; |
404 | spin_unlock(&sb_lock); | 402 | spin_unlock(&sb_lock); |
405 | 403 | ||
406 | if (down_read_trylock(&sb->s_umount)) { | 404 | if (down_read_trylock(&sb->s_umount)) { |
407 | if (sb->s_root && (sb->s_flags & MS_BORN)) | 405 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
408 | return true; | 406 | return true; |
409 | up_read(&sb->s_umount); | 407 | up_read(&sb->s_umount); |
410 | } | 408 | } |
411 | 409 | ||
412 | put_super(sb); | 410 | put_super(sb); |
413 | return false; | 411 | return false; |
414 | } | 412 | } |
415 | 413 | ||
416 | /** | 414 | /** |
417 | * generic_shutdown_super - common helper for ->kill_sb() | 415 | * generic_shutdown_super - common helper for ->kill_sb() |
418 | * @sb: superblock to kill | 416 | * @sb: superblock to kill |
419 | * | 417 | * |
420 | * generic_shutdown_super() does all fs-independent work on superblock | 418 | * generic_shutdown_super() does all fs-independent work on superblock |
421 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | 419 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects |
422 | * that need destruction out of superblock, call generic_shutdown_super() | 420 | * that need destruction out of superblock, call generic_shutdown_super() |
423 | * and release aforementioned objects. Note: dentries and inodes _are_ | 421 | * and release aforementioned objects. Note: dentries and inodes _are_ |
424 | * taken care of and do not need specific handling. | 422 | * taken care of and do not need specific handling. |
425 | * | 423 | * |
426 | * Upon calling this function, the filesystem may no longer alter or | 424 | * Upon calling this function, the filesystem may no longer alter or |
427 | * rearrange the set of dentries belonging to this super_block, nor may it | 425 | * rearrange the set of dentries belonging to this super_block, nor may it |
428 | * change the attachments of dentries to inodes. | 426 | * change the attachments of dentries to inodes. |
429 | */ | 427 | */ |
430 | void generic_shutdown_super(struct super_block *sb) | 428 | void generic_shutdown_super(struct super_block *sb) |
431 | { | 429 | { |
432 | const struct super_operations *sop = sb->s_op; | 430 | const struct super_operations *sop = sb->s_op; |
433 | 431 | ||
434 | if (sb->s_root) { | 432 | if (sb->s_root) { |
435 | shrink_dcache_for_umount(sb); | 433 | shrink_dcache_for_umount(sb); |
436 | sync_filesystem(sb); | 434 | sync_filesystem(sb); |
437 | sb->s_flags &= ~MS_ACTIVE; | 435 | sb->s_flags &= ~MS_ACTIVE; |
438 | 436 | ||
439 | fsnotify_unmount_inodes(&sb->s_inodes); | 437 | fsnotify_unmount_inodes(&sb->s_inodes); |
440 | 438 | ||
441 | evict_inodes(sb); | 439 | evict_inodes(sb); |
442 | 440 | ||
443 | if (sb->s_dio_done_wq) { | 441 | if (sb->s_dio_done_wq) { |
444 | destroy_workqueue(sb->s_dio_done_wq); | 442 | destroy_workqueue(sb->s_dio_done_wq); |
445 | sb->s_dio_done_wq = NULL; | 443 | sb->s_dio_done_wq = NULL; |
446 | } | 444 | } |
447 | 445 | ||
448 | if (sop->put_super) | 446 | if (sop->put_super) |
449 | sop->put_super(sb); | 447 | sop->put_super(sb); |
450 | 448 | ||
451 | if (!list_empty(&sb->s_inodes)) { | 449 | if (!list_empty(&sb->s_inodes)) { |
452 | printk("VFS: Busy inodes after unmount of %s. " | 450 | printk("VFS: Busy inodes after unmount of %s. " |
453 | "Self-destruct in 5 seconds. Have a nice day...\n", | 451 | "Self-destruct in 5 seconds. Have a nice day...\n", |
454 | sb->s_id); | 452 | sb->s_id); |
455 | } | 453 | } |
456 | } | 454 | } |
457 | spin_lock(&sb_lock); | 455 | spin_lock(&sb_lock); |
458 | /* should be initialized for __put_super_and_need_restart() */ | 456 | /* should be initialized for __put_super_and_need_restart() */ |
459 | hlist_del_init(&sb->s_instances); | 457 | hlist_del_init(&sb->s_instances); |
460 | spin_unlock(&sb_lock); | 458 | spin_unlock(&sb_lock); |
461 | up_write(&sb->s_umount); | 459 | up_write(&sb->s_umount); |
462 | } | 460 | } |
463 | 461 | ||
464 | EXPORT_SYMBOL(generic_shutdown_super); | 462 | EXPORT_SYMBOL(generic_shutdown_super); |
465 | 463 | ||
466 | /** | 464 | /** |
467 | * sget - find or create a superblock | 465 | * sget - find or create a superblock |
468 | * @type: filesystem type superblock should belong to | 466 | * @type: filesystem type superblock should belong to |
469 | * @test: comparison callback | 467 | * @test: comparison callback |
470 | * @set: setup callback | 468 | * @set: setup callback |
471 | * @flags: mount flags | 469 | * @flags: mount flags |
472 | * @data: argument to each of them | 470 | * @data: argument to each of them |
473 | */ | 471 | */ |
474 | struct super_block *sget(struct file_system_type *type, | 472 | struct super_block *sget(struct file_system_type *type, |
475 | int (*test)(struct super_block *,void *), | 473 | int (*test)(struct super_block *,void *), |
476 | int (*set)(struct super_block *,void *), | 474 | int (*set)(struct super_block *,void *), |
477 | int flags, | 475 | int flags, |
478 | void *data) | 476 | void *data) |
479 | { | 477 | { |
480 | struct super_block *s = NULL; | 478 | struct super_block *s = NULL; |
481 | struct super_block *old; | 479 | struct super_block *old; |
482 | int err; | 480 | int err; |
483 | 481 | ||
484 | retry: | 482 | retry: |
485 | spin_lock(&sb_lock); | 483 | spin_lock(&sb_lock); |
486 | if (test) { | 484 | if (test) { |
487 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { | 485 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
488 | if (!test(old, data)) | 486 | if (!test(old, data)) |
489 | continue; | 487 | continue; |
490 | if (!grab_super(old)) | 488 | if (!grab_super(old)) |
491 | goto retry; | 489 | goto retry; |
492 | if (s) { | 490 | if (s) { |
493 | up_write(&s->s_umount); | 491 | up_write(&s->s_umount); |
494 | destroy_super(s); | 492 | destroy_super(s); |
495 | s = NULL; | 493 | s = NULL; |
496 | } | 494 | } |
497 | return old; | 495 | return old; |
498 | } | 496 | } |
499 | } | 497 | } |
500 | if (!s) { | 498 | if (!s) { |
501 | spin_unlock(&sb_lock); | 499 | spin_unlock(&sb_lock); |
502 | s = alloc_super(type, flags); | 500 | s = alloc_super(type, flags); |
503 | if (!s) | 501 | if (!s) |
504 | return ERR_PTR(-ENOMEM); | 502 | return ERR_PTR(-ENOMEM); |
505 | goto retry; | 503 | goto retry; |
506 | } | 504 | } |
507 | 505 | ||
508 | err = set(s, data); | 506 | err = set(s, data); |
509 | if (err) { | 507 | if (err) { |
510 | spin_unlock(&sb_lock); | 508 | spin_unlock(&sb_lock); |
511 | up_write(&s->s_umount); | 509 | up_write(&s->s_umount); |
512 | destroy_super(s); | 510 | destroy_super(s); |
513 | return ERR_PTR(err); | 511 | return ERR_PTR(err); |
514 | } | 512 | } |
515 | s->s_type = type; | 513 | s->s_type = type; |
516 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | 514 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); |
517 | list_add_tail(&s->s_list, &super_blocks); | 515 | list_add_tail(&s->s_list, &super_blocks); |
518 | hlist_add_head(&s->s_instances, &type->fs_supers); | 516 | hlist_add_head(&s->s_instances, &type->fs_supers); |
519 | spin_unlock(&sb_lock); | 517 | spin_unlock(&sb_lock); |
520 | get_filesystem(type); | 518 | get_filesystem(type); |
521 | register_shrinker(&s->s_shrink); | 519 | register_shrinker(&s->s_shrink); |
522 | return s; | 520 | return s; |
523 | } | 521 | } |
524 | 522 | ||
525 | EXPORT_SYMBOL(sget); | 523 | EXPORT_SYMBOL(sget); |
526 | 524 | ||
527 | void drop_super(struct super_block *sb) | 525 | void drop_super(struct super_block *sb) |
528 | { | 526 | { |
529 | up_read(&sb->s_umount); | 527 | up_read(&sb->s_umount); |
530 | put_super(sb); | 528 | put_super(sb); |
531 | } | 529 | } |
532 | 530 | ||
533 | EXPORT_SYMBOL(drop_super); | 531 | EXPORT_SYMBOL(drop_super); |
534 | 532 | ||
535 | /** | 533 | /** |
536 | * iterate_supers - call function for all active superblocks | 534 | * iterate_supers - call function for all active superblocks |
537 | * @f: function to call | 535 | * @f: function to call |
538 | * @arg: argument to pass to it | 536 | * @arg: argument to pass to it |
539 | * | 537 | * |
540 | * Scans the superblock list and calls given function, passing it | 538 | * Scans the superblock list and calls given function, passing it |
541 | * locked superblock and given argument. | 539 | * locked superblock and given argument. |
542 | */ | 540 | */ |
543 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | 541 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) |
544 | { | 542 | { |
545 | struct super_block *sb, *p = NULL; | 543 | struct super_block *sb, *p = NULL; |
546 | 544 | ||
547 | spin_lock(&sb_lock); | 545 | spin_lock(&sb_lock); |
548 | list_for_each_entry(sb, &super_blocks, s_list) { | 546 | list_for_each_entry(sb, &super_blocks, s_list) { |
549 | if (hlist_unhashed(&sb->s_instances)) | 547 | if (hlist_unhashed(&sb->s_instances)) |
550 | continue; | 548 | continue; |
551 | sb->s_count++; | 549 | sb->s_count++; |
552 | spin_unlock(&sb_lock); | 550 | spin_unlock(&sb_lock); |
553 | 551 | ||
554 | down_read(&sb->s_umount); | 552 | down_read(&sb->s_umount); |
555 | if (sb->s_root && (sb->s_flags & MS_BORN)) | 553 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
556 | f(sb, arg); | 554 | f(sb, arg); |
557 | up_read(&sb->s_umount); | 555 | up_read(&sb->s_umount); |
558 | 556 | ||
559 | spin_lock(&sb_lock); | 557 | spin_lock(&sb_lock); |
560 | if (p) | 558 | if (p) |
561 | __put_super(p); | 559 | __put_super(p); |
562 | p = sb; | 560 | p = sb; |
563 | } | 561 | } |
564 | if (p) | 562 | if (p) |
565 | __put_super(p); | 563 | __put_super(p); |
566 | spin_unlock(&sb_lock); | 564 | spin_unlock(&sb_lock); |
567 | } | 565 | } |
568 | 566 | ||
569 | /** | 567 | /** |
570 | * iterate_supers_type - call function for superblocks of given type | 568 | * iterate_supers_type - call function for superblocks of given type |
571 | * @type: fs type | 569 | * @type: fs type |
572 | * @f: function to call | 570 | * @f: function to call |
573 | * @arg: argument to pass to it | 571 | * @arg: argument to pass to it |
574 | * | 572 | * |
575 | * Scans the superblock list and calls given function, passing it | 573 | * Scans the superblock list and calls given function, passing it |
576 | * locked superblock and given argument. | 574 | * locked superblock and given argument. |
577 | */ | 575 | */ |
578 | void iterate_supers_type(struct file_system_type *type, | 576 | void iterate_supers_type(struct file_system_type *type, |
579 | void (*f)(struct super_block *, void *), void *arg) | 577 | void (*f)(struct super_block *, void *), void *arg) |
580 | { | 578 | { |
581 | struct super_block *sb, *p = NULL; | 579 | struct super_block *sb, *p = NULL; |
582 | 580 | ||
583 | spin_lock(&sb_lock); | 581 | spin_lock(&sb_lock); |
584 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { | 582 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
585 | sb->s_count++; | 583 | sb->s_count++; |
586 | spin_unlock(&sb_lock); | 584 | spin_unlock(&sb_lock); |
587 | 585 | ||
588 | down_read(&sb->s_umount); | 586 | down_read(&sb->s_umount); |
589 | if (sb->s_root && (sb->s_flags & MS_BORN)) | 587 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
590 | f(sb, arg); | 588 | f(sb, arg); |
591 | up_read(&sb->s_umount); | 589 | up_read(&sb->s_umount); |
592 | 590 | ||
593 | spin_lock(&sb_lock); | 591 | spin_lock(&sb_lock); |
594 | if (p) | 592 | if (p) |
595 | __put_super(p); | 593 | __put_super(p); |
596 | p = sb; | 594 | p = sb; |
597 | } | 595 | } |
598 | if (p) | 596 | if (p) |
599 | __put_super(p); | 597 | __put_super(p); |
600 | spin_unlock(&sb_lock); | 598 | spin_unlock(&sb_lock); |
601 | } | 599 | } |
602 | 600 | ||
603 | EXPORT_SYMBOL(iterate_supers_type); | 601 | EXPORT_SYMBOL(iterate_supers_type); |
604 | 602 | ||
605 | /** | 603 | /** |
606 | * get_super - get the superblock of a device | 604 | * get_super - get the superblock of a device |
607 | * @bdev: device to get the superblock for | 605 | * @bdev: device to get the superblock for |
608 | * | 606 | * |
609 | * Scans the superblock list and finds the superblock of the file system | 607 | * Scans the superblock list and finds the superblock of the file system |
610 | * mounted on the device given. %NULL is returned if no match is found. | 608 | * mounted on the device given. %NULL is returned if no match is found. |
611 | */ | 609 | */ |
612 | 610 | ||
613 | struct super_block *get_super(struct block_device *bdev) | 611 | struct super_block *get_super(struct block_device *bdev) |
614 | { | 612 | { |
615 | struct super_block *sb; | 613 | struct super_block *sb; |
616 | 614 | ||
617 | if (!bdev) | 615 | if (!bdev) |
618 | return NULL; | 616 | return NULL; |
619 | 617 | ||
620 | spin_lock(&sb_lock); | 618 | spin_lock(&sb_lock); |
621 | rescan: | 619 | rescan: |
622 | list_for_each_entry(sb, &super_blocks, s_list) { | 620 | list_for_each_entry(sb, &super_blocks, s_list) { |
623 | if (hlist_unhashed(&sb->s_instances)) | 621 | if (hlist_unhashed(&sb->s_instances)) |
624 | continue; | 622 | continue; |
625 | if (sb->s_bdev == bdev) { | 623 | if (sb->s_bdev == bdev) { |
626 | sb->s_count++; | 624 | sb->s_count++; |
627 | spin_unlock(&sb_lock); | 625 | spin_unlock(&sb_lock); |
628 | down_read(&sb->s_umount); | 626 | down_read(&sb->s_umount); |
629 | /* still alive? */ | 627 | /* still alive? */ |
630 | if (sb->s_root && (sb->s_flags & MS_BORN)) | 628 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
631 | return sb; | 629 | return sb; |
632 | up_read(&sb->s_umount); | 630 | up_read(&sb->s_umount); |
633 | /* nope, got unmounted */ | 631 | /* nope, got unmounted */ |
634 | spin_lock(&sb_lock); | 632 | spin_lock(&sb_lock); |
635 | __put_super(sb); | 633 | __put_super(sb); |
636 | goto rescan; | 634 | goto rescan; |
637 | } | 635 | } |
638 | } | 636 | } |
639 | spin_unlock(&sb_lock); | 637 | spin_unlock(&sb_lock); |
640 | return NULL; | 638 | return NULL; |
641 | } | 639 | } |
642 | 640 | ||
643 | EXPORT_SYMBOL(get_super); | 641 | EXPORT_SYMBOL(get_super); |
644 | 642 | ||
645 | /** | 643 | /** |
646 | * get_super_thawed - get thawed superblock of a device | 644 | * get_super_thawed - get thawed superblock of a device |
647 | * @bdev: device to get the superblock for | 645 | * @bdev: device to get the superblock for |
648 | * | 646 | * |
649 | * Scans the superblock list and finds the superblock of the file system | 647 | * Scans the superblock list and finds the superblock of the file system |
650 | * mounted on the device. The superblock is returned once it is thawed | 648 | * mounted on the device. The superblock is returned once it is thawed |
651 | * (or immediately if it was not frozen). %NULL is returned if no match | 649 | * (or immediately if it was not frozen). %NULL is returned if no match |
652 | * is found. | 650 | * is found. |
653 | */ | 651 | */ |
654 | struct super_block *get_super_thawed(struct block_device *bdev) | 652 | struct super_block *get_super_thawed(struct block_device *bdev) |
655 | { | 653 | { |
656 | while (1) { | 654 | while (1) { |
657 | struct super_block *s = get_super(bdev); | 655 | struct super_block *s = get_super(bdev); |
658 | if (!s || s->s_writers.frozen == SB_UNFROZEN) | 656 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
659 | return s; | 657 | return s; |
660 | up_read(&s->s_umount); | 658 | up_read(&s->s_umount); |
661 | wait_event(s->s_writers.wait_unfrozen, | 659 | wait_event(s->s_writers.wait_unfrozen, |
662 | s->s_writers.frozen == SB_UNFROZEN); | 660 | s->s_writers.frozen == SB_UNFROZEN); |
663 | put_super(s); | 661 | put_super(s); |
664 | } | 662 | } |
665 | } | 663 | } |
666 | EXPORT_SYMBOL(get_super_thawed); | 664 | EXPORT_SYMBOL(get_super_thawed); |
667 | 665 | ||
668 | /** | 666 | /** |
669 | * get_active_super - get an active reference to the superblock of a device | 667 | * get_active_super - get an active reference to the superblock of a device |
670 | * @bdev: device to get the superblock for | 668 | * @bdev: device to get the superblock for |
671 | * | 669 | * |
672 | * Scans the superblock list and finds the superblock of the file system | 670 | * Scans the superblock list and finds the superblock of the file system |
673 | * mounted on the device given. Returns the superblock with an active | 671 | * mounted on the device given. Returns the superblock with an active |
674 | * reference or %NULL if none was found. | 672 | * reference or %NULL if none was found. |
675 | */ | 673 | */ |
676 | struct super_block *get_active_super(struct block_device *bdev) | 674 | struct super_block *get_active_super(struct block_device *bdev) |
677 | { | 675 | { |
678 | struct super_block *sb; | 676 | struct super_block *sb; |
679 | 677 | ||
680 | if (!bdev) | 678 | if (!bdev) |
681 | return NULL; | 679 | return NULL; |
682 | 680 | ||
683 | restart: | 681 | restart: |
684 | spin_lock(&sb_lock); | 682 | spin_lock(&sb_lock); |
685 | list_for_each_entry(sb, &super_blocks, s_list) { | 683 | list_for_each_entry(sb, &super_blocks, s_list) { |
686 | if (hlist_unhashed(&sb->s_instances)) | 684 | if (hlist_unhashed(&sb->s_instances)) |
687 | continue; | 685 | continue; |
688 | if (sb->s_bdev == bdev) { | 686 | if (sb->s_bdev == bdev) { |
689 | if (!grab_super(sb)) | 687 | if (!grab_super(sb)) |
690 | goto restart; | 688 | goto restart; |
691 | up_write(&sb->s_umount); | 689 | up_write(&sb->s_umount); |
692 | return sb; | 690 | return sb; |
693 | } | 691 | } |
694 | } | 692 | } |
695 | spin_unlock(&sb_lock); | 693 | spin_unlock(&sb_lock); |
696 | return NULL; | 694 | return NULL; |
697 | } | 695 | } |
698 | 696 | ||
699 | struct super_block *user_get_super(dev_t dev) | 697 | struct super_block *user_get_super(dev_t dev) |
700 | { | 698 | { |
701 | struct super_block *sb; | 699 | struct super_block *sb; |
702 | 700 | ||
703 | spin_lock(&sb_lock); | 701 | spin_lock(&sb_lock); |
704 | rescan: | 702 | rescan: |
705 | list_for_each_entry(sb, &super_blocks, s_list) { | 703 | list_for_each_entry(sb, &super_blocks, s_list) { |
706 | if (hlist_unhashed(&sb->s_instances)) | 704 | if (hlist_unhashed(&sb->s_instances)) |
707 | continue; | 705 | continue; |
708 | if (sb->s_dev == dev) { | 706 | if (sb->s_dev == dev) { |
709 | sb->s_count++; | 707 | sb->s_count++; |
710 | spin_unlock(&sb_lock); | 708 | spin_unlock(&sb_lock); |
711 | down_read(&sb->s_umount); | 709 | down_read(&sb->s_umount); |
712 | /* still alive? */ | 710 | /* still alive? */ |
713 | if (sb->s_root && (sb->s_flags & MS_BORN)) | 711 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
714 | return sb; | 712 | return sb; |
715 | up_read(&sb->s_umount); | 713 | up_read(&sb->s_umount); |
716 | /* nope, got unmounted */ | 714 | /* nope, got unmounted */ |
717 | spin_lock(&sb_lock); | 715 | spin_lock(&sb_lock); |
718 | __put_super(sb); | 716 | __put_super(sb); |
719 | goto rescan; | 717 | goto rescan; |
720 | } | 718 | } |
721 | } | 719 | } |
722 | spin_unlock(&sb_lock); | 720 | spin_unlock(&sb_lock); |
723 | return NULL; | 721 | return NULL; |
724 | } | 722 | } |
725 | 723 | ||
726 | /** | 724 | /** |
727 | * do_remount_sb - asks filesystem to change mount options. | 725 | * do_remount_sb - asks filesystem to change mount options. |
728 | * @sb: superblock in question | 726 | * @sb: superblock in question |
729 | * @flags: numeric part of options | 727 | * @flags: numeric part of options |
730 | * @data: the rest of options | 728 | * @data: the rest of options |
731 | * @force: whether or not to force the change | 729 | * @force: whether or not to force the change |
732 | * | 730 | * |
733 | * Alters the mount options of a mounted file system. | 731 | * Alters the mount options of a mounted file system. |
734 | */ | 732 | */ |
735 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | 733 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) |
736 | { | 734 | { |
737 | int retval; | 735 | int retval; |
738 | int remount_ro; | 736 | int remount_ro; |
739 | 737 | ||
740 | if (sb->s_writers.frozen != SB_UNFROZEN) | 738 | if (sb->s_writers.frozen != SB_UNFROZEN) |
741 | return -EBUSY; | 739 | return -EBUSY; |
742 | 740 | ||
743 | #ifdef CONFIG_BLOCK | 741 | #ifdef CONFIG_BLOCK |
744 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) | 742 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
745 | return -EACCES; | 743 | return -EACCES; |
746 | #endif | 744 | #endif |
747 | 745 | ||
748 | if (flags & MS_RDONLY) | 746 | if (flags & MS_RDONLY) |
749 | acct_auto_close(sb); | 747 | acct_auto_close(sb); |
750 | shrink_dcache_sb(sb); | 748 | shrink_dcache_sb(sb); |
751 | sync_filesystem(sb); | 749 | sync_filesystem(sb); |
752 | 750 | ||
753 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); | 751 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
754 | 752 | ||
755 | /* If we are remounting RDONLY and current sb is read/write, | 753 | /* If we are remounting RDONLY and current sb is read/write, |
756 | make sure there are no rw files opened */ | 754 | make sure there are no rw files opened */ |
757 | if (remount_ro) { | 755 | if (remount_ro) { |
758 | if (force) { | 756 | if (force) { |
759 | mark_files_ro(sb); | 757 | mark_files_ro(sb); |
760 | } else { | 758 | } else { |
761 | retval = sb_prepare_remount_readonly(sb); | 759 | retval = sb_prepare_remount_readonly(sb); |
762 | if (retval) | 760 | if (retval) |
763 | return retval; | 761 | return retval; |
764 | } | 762 | } |
765 | } | 763 | } |
766 | 764 | ||
767 | if (sb->s_op->remount_fs) { | 765 | if (sb->s_op->remount_fs) { |
768 | retval = sb->s_op->remount_fs(sb, &flags, data); | 766 | retval = sb->s_op->remount_fs(sb, &flags, data); |
769 | if (retval) { | 767 | if (retval) { |
770 | if (!force) | 768 | if (!force) |
771 | goto cancel_readonly; | 769 | goto cancel_readonly; |
772 | /* If forced remount, go ahead despite any errors */ | 770 | /* If forced remount, go ahead despite any errors */ |
773 | WARN(1, "forced remount of a %s fs returned %i\n", | 771 | WARN(1, "forced remount of a %s fs returned %i\n", |
774 | sb->s_type->name, retval); | 772 | sb->s_type->name, retval); |
775 | } | 773 | } |
776 | } | 774 | } |
777 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | 775 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); |
778 | /* Needs to be ordered wrt mnt_is_readonly() */ | 776 | /* Needs to be ordered wrt mnt_is_readonly() */ |
779 | smp_wmb(); | 777 | smp_wmb(); |
780 | sb->s_readonly_remount = 0; | 778 | sb->s_readonly_remount = 0; |
781 | 779 | ||
782 | /* | 780 | /* |
783 | * Some filesystems modify their metadata via some other path than the | 781 | * Some filesystems modify their metadata via some other path than the |
784 | * bdev buffer cache (eg. use a private mapping, or directories in | 782 | * bdev buffer cache (eg. use a private mapping, or directories in |
785 | * pagecache, etc). Also file data modifications go via their own | 783 | * pagecache, etc). Also file data modifications go via their own |
786 | * mappings. So If we try to mount readonly then copy the filesystem | 784 | * mappings. So If we try to mount readonly then copy the filesystem |
787 | * from bdev, we could get stale data, so invalidate it to give a best | 785 | * from bdev, we could get stale data, so invalidate it to give a best |
788 | * effort at coherency. | 786 | * effort at coherency. |
789 | */ | 787 | */ |
790 | if (remount_ro && sb->s_bdev) | 788 | if (remount_ro && sb->s_bdev) |
791 | invalidate_bdev(sb->s_bdev); | 789 | invalidate_bdev(sb->s_bdev); |
792 | return 0; | 790 | return 0; |
793 | 791 | ||
794 | cancel_readonly: | 792 | cancel_readonly: |
795 | sb->s_readonly_remount = 0; | 793 | sb->s_readonly_remount = 0; |
796 | return retval; | 794 | return retval; |
797 | } | 795 | } |
798 | 796 | ||
799 | static void do_emergency_remount(struct work_struct *work) | 797 | static void do_emergency_remount(struct work_struct *work) |
800 | { | 798 | { |
801 | struct super_block *sb, *p = NULL; | 799 | struct super_block *sb, *p = NULL; |
802 | 800 | ||
803 | spin_lock(&sb_lock); | 801 | spin_lock(&sb_lock); |
804 | list_for_each_entry(sb, &super_blocks, s_list) { | 802 | list_for_each_entry(sb, &super_blocks, s_list) { |
805 | if (hlist_unhashed(&sb->s_instances)) | 803 | if (hlist_unhashed(&sb->s_instances)) |
806 | continue; | 804 | continue; |
807 | sb->s_count++; | 805 | sb->s_count++; |
808 | spin_unlock(&sb_lock); | 806 | spin_unlock(&sb_lock); |
809 | down_write(&sb->s_umount); | 807 | down_write(&sb->s_umount); |
810 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && | 808 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
811 | !(sb->s_flags & MS_RDONLY)) { | 809 | !(sb->s_flags & MS_RDONLY)) { |
812 | /* | 810 | /* |
813 | * What lock protects sb->s_flags?? | 811 | * What lock protects sb->s_flags?? |
814 | */ | 812 | */ |
815 | do_remount_sb(sb, MS_RDONLY, NULL, 1); | 813 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
816 | } | 814 | } |
817 | up_write(&sb->s_umount); | 815 | up_write(&sb->s_umount); |
818 | spin_lock(&sb_lock); | 816 | spin_lock(&sb_lock); |
819 | if (p) | 817 | if (p) |
820 | __put_super(p); | 818 | __put_super(p); |
821 | p = sb; | 819 | p = sb; |
822 | } | 820 | } |
823 | if (p) | 821 | if (p) |
824 | __put_super(p); | 822 | __put_super(p); |
825 | spin_unlock(&sb_lock); | 823 | spin_unlock(&sb_lock); |
826 | kfree(work); | 824 | kfree(work); |
827 | printk("Emergency Remount complete\n"); | 825 | printk("Emergency Remount complete\n"); |
828 | } | 826 | } |
829 | 827 | ||
830 | void emergency_remount(void) | 828 | void emergency_remount(void) |
831 | { | 829 | { |
832 | struct work_struct *work; | 830 | struct work_struct *work; |
833 | 831 | ||
834 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | 832 | work = kmalloc(sizeof(*work), GFP_ATOMIC); |
835 | if (work) { | 833 | if (work) { |
836 | INIT_WORK(work, do_emergency_remount); | 834 | INIT_WORK(work, do_emergency_remount); |
837 | schedule_work(work); | 835 | schedule_work(work); |
838 | } | 836 | } |
839 | } | 837 | } |
840 | 838 | ||
841 | /* | 839 | /* |
842 | * Unnamed block devices are dummy devices used by virtual | 840 | * Unnamed block devices are dummy devices used by virtual |
843 | * filesystems which don't use real block-devices. -- jrs | 841 | * filesystems which don't use real block-devices. -- jrs |
844 | */ | 842 | */ |
845 | 843 | ||
846 | static DEFINE_IDA(unnamed_dev_ida); | 844 | static DEFINE_IDA(unnamed_dev_ida); |
847 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ | 845 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
848 | /* Many userspace utilities consider an FSID of 0 invalid. | 846 | /* Many userspace utilities consider an FSID of 0 invalid. |
849 | * Always return at least 1 from get_anon_bdev. | 847 | * Always return at least 1 from get_anon_bdev. |
850 | */ | 848 | */ |
851 | static int unnamed_dev_start = 1; | 849 | static int unnamed_dev_start = 1; |
852 | 850 | ||
853 | int get_anon_bdev(dev_t *p) | 851 | int get_anon_bdev(dev_t *p) |
854 | { | 852 | { |
855 | int dev; | 853 | int dev; |
856 | int error; | 854 | int error; |
857 | 855 | ||
858 | retry: | 856 | retry: |
859 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) | 857 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
860 | return -ENOMEM; | 858 | return -ENOMEM; |
861 | spin_lock(&unnamed_dev_lock); | 859 | spin_lock(&unnamed_dev_lock); |
862 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); | 860 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
863 | if (!error) | 861 | if (!error) |
864 | unnamed_dev_start = dev + 1; | 862 | unnamed_dev_start = dev + 1; |
865 | spin_unlock(&unnamed_dev_lock); | 863 | spin_unlock(&unnamed_dev_lock); |
866 | if (error == -EAGAIN) | 864 | if (error == -EAGAIN) |
867 | /* We raced and lost with another CPU. */ | 865 | /* We raced and lost with another CPU. */ |
868 | goto retry; | 866 | goto retry; |
869 | else if (error) | 867 | else if (error) |
870 | return -EAGAIN; | 868 | return -EAGAIN; |
871 | 869 | ||
872 | if (dev == (1 << MINORBITS)) { | 870 | if (dev == (1 << MINORBITS)) { |
873 | spin_lock(&unnamed_dev_lock); | 871 | spin_lock(&unnamed_dev_lock); |
874 | ida_remove(&unnamed_dev_ida, dev); | 872 | ida_remove(&unnamed_dev_ida, dev); |
875 | if (unnamed_dev_start > dev) | 873 | if (unnamed_dev_start > dev) |
876 | unnamed_dev_start = dev; | 874 | unnamed_dev_start = dev; |
877 | spin_unlock(&unnamed_dev_lock); | 875 | spin_unlock(&unnamed_dev_lock); |
878 | return -EMFILE; | 876 | return -EMFILE; |
879 | } | 877 | } |
880 | *p = MKDEV(0, dev & MINORMASK); | 878 | *p = MKDEV(0, dev & MINORMASK); |
881 | return 0; | 879 | return 0; |
882 | } | 880 | } |
883 | EXPORT_SYMBOL(get_anon_bdev); | 881 | EXPORT_SYMBOL(get_anon_bdev); |
884 | 882 | ||
885 | void free_anon_bdev(dev_t dev) | 883 | void free_anon_bdev(dev_t dev) |
886 | { | 884 | { |
887 | int slot = MINOR(dev); | 885 | int slot = MINOR(dev); |
888 | spin_lock(&unnamed_dev_lock); | 886 | spin_lock(&unnamed_dev_lock); |
889 | ida_remove(&unnamed_dev_ida, slot); | 887 | ida_remove(&unnamed_dev_ida, slot); |
890 | if (slot < unnamed_dev_start) | 888 | if (slot < unnamed_dev_start) |
891 | unnamed_dev_start = slot; | 889 | unnamed_dev_start = slot; |
892 | spin_unlock(&unnamed_dev_lock); | 890 | spin_unlock(&unnamed_dev_lock); |
893 | } | 891 | } |
894 | EXPORT_SYMBOL(free_anon_bdev); | 892 | EXPORT_SYMBOL(free_anon_bdev); |
895 | 893 | ||
896 | int set_anon_super(struct super_block *s, void *data) | 894 | int set_anon_super(struct super_block *s, void *data) |
897 | { | 895 | { |
898 | int error = get_anon_bdev(&s->s_dev); | 896 | int error = get_anon_bdev(&s->s_dev); |
899 | if (!error) | 897 | if (!error) |
900 | s->s_bdi = &noop_backing_dev_info; | 898 | s->s_bdi = &noop_backing_dev_info; |
901 | return error; | 899 | return error; |
902 | } | 900 | } |
903 | 901 | ||
904 | EXPORT_SYMBOL(set_anon_super); | 902 | EXPORT_SYMBOL(set_anon_super); |
905 | 903 | ||
906 | void kill_anon_super(struct super_block *sb) | 904 | void kill_anon_super(struct super_block *sb) |
907 | { | 905 | { |
908 | dev_t dev = sb->s_dev; | 906 | dev_t dev = sb->s_dev; |
909 | generic_shutdown_super(sb); | 907 | generic_shutdown_super(sb); |
910 | free_anon_bdev(dev); | 908 | free_anon_bdev(dev); |
911 | } | 909 | } |
912 | 910 | ||
913 | EXPORT_SYMBOL(kill_anon_super); | 911 | EXPORT_SYMBOL(kill_anon_super); |
914 | 912 | ||
915 | void kill_litter_super(struct super_block *sb) | 913 | void kill_litter_super(struct super_block *sb) |
916 | { | 914 | { |
917 | if (sb->s_root) | 915 | if (sb->s_root) |
918 | d_genocide(sb->s_root); | 916 | d_genocide(sb->s_root); |
919 | kill_anon_super(sb); | 917 | kill_anon_super(sb); |
920 | } | 918 | } |
921 | 919 | ||
922 | EXPORT_SYMBOL(kill_litter_super); | 920 | EXPORT_SYMBOL(kill_litter_super); |
923 | 921 | ||
924 | static int ns_test_super(struct super_block *sb, void *data) | 922 | static int ns_test_super(struct super_block *sb, void *data) |
925 | { | 923 | { |
926 | return sb->s_fs_info == data; | 924 | return sb->s_fs_info == data; |
927 | } | 925 | } |
928 | 926 | ||
929 | static int ns_set_super(struct super_block *sb, void *data) | 927 | static int ns_set_super(struct super_block *sb, void *data) |
930 | { | 928 | { |
931 | sb->s_fs_info = data; | 929 | sb->s_fs_info = data; |
932 | return set_anon_super(sb, NULL); | 930 | return set_anon_super(sb, NULL); |
933 | } | 931 | } |
934 | 932 | ||
935 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, | 933 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
936 | void *data, int (*fill_super)(struct super_block *, void *, int)) | 934 | void *data, int (*fill_super)(struct super_block *, void *, int)) |
937 | { | 935 | { |
938 | struct super_block *sb; | 936 | struct super_block *sb; |
939 | 937 | ||
940 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); | 938 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
941 | if (IS_ERR(sb)) | 939 | if (IS_ERR(sb)) |
942 | return ERR_CAST(sb); | 940 | return ERR_CAST(sb); |
943 | 941 | ||
944 | if (!sb->s_root) { | 942 | if (!sb->s_root) { |
945 | int err; | 943 | int err; |
946 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | 944 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
947 | if (err) { | 945 | if (err) { |
948 | deactivate_locked_super(sb); | 946 | deactivate_locked_super(sb); |
949 | return ERR_PTR(err); | 947 | return ERR_PTR(err); |
950 | } | 948 | } |
951 | 949 | ||
952 | sb->s_flags |= MS_ACTIVE; | 950 | sb->s_flags |= MS_ACTIVE; |
953 | } | 951 | } |
954 | 952 | ||
955 | return dget(sb->s_root); | 953 | return dget(sb->s_root); |
956 | } | 954 | } |
957 | 955 | ||
958 | EXPORT_SYMBOL(mount_ns); | 956 | EXPORT_SYMBOL(mount_ns); |
959 | 957 | ||
960 | #ifdef CONFIG_BLOCK | 958 | #ifdef CONFIG_BLOCK |
961 | static int set_bdev_super(struct super_block *s, void *data) | 959 | static int set_bdev_super(struct super_block *s, void *data) |
962 | { | 960 | { |
963 | s->s_bdev = data; | 961 | s->s_bdev = data; |
964 | s->s_dev = s->s_bdev->bd_dev; | 962 | s->s_dev = s->s_bdev->bd_dev; |
965 | 963 | ||
966 | /* | 964 | /* |
967 | * We set the bdi here to the queue backing, file systems can | 965 | * We set the bdi here to the queue backing, file systems can |
968 | * overwrite this in ->fill_super() | 966 | * overwrite this in ->fill_super() |
969 | */ | 967 | */ |
970 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | 968 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; |
971 | return 0; | 969 | return 0; |
972 | } | 970 | } |
973 | 971 | ||
974 | static int test_bdev_super(struct super_block *s, void *data) | 972 | static int test_bdev_super(struct super_block *s, void *data) |
975 | { | 973 | { |
976 | return (void *)s->s_bdev == data; | 974 | return (void *)s->s_bdev == data; |
977 | } | 975 | } |
978 | 976 | ||
979 | struct dentry *mount_bdev(struct file_system_type *fs_type, | 977 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
980 | int flags, const char *dev_name, void *data, | 978 | int flags, const char *dev_name, void *data, |
981 | int (*fill_super)(struct super_block *, void *, int)) | 979 | int (*fill_super)(struct super_block *, void *, int)) |
982 | { | 980 | { |
983 | struct block_device *bdev; | 981 | struct block_device *bdev; |
984 | struct super_block *s; | 982 | struct super_block *s; |
985 | fmode_t mode = FMODE_READ | FMODE_EXCL; | 983 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
986 | int error = 0; | 984 | int error = 0; |
987 | 985 | ||
988 | if (!(flags & MS_RDONLY)) | 986 | if (!(flags & MS_RDONLY)) |
989 | mode |= FMODE_WRITE; | 987 | mode |= FMODE_WRITE; |
990 | 988 | ||
991 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); | 989 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
992 | if (IS_ERR(bdev)) | 990 | if (IS_ERR(bdev)) |
993 | return ERR_CAST(bdev); | 991 | return ERR_CAST(bdev); |
994 | 992 | ||
995 | /* | 993 | /* |
996 | * once the super is inserted into the list by sget, s_umount | 994 | * once the super is inserted into the list by sget, s_umount |
997 | * will protect the lockfs code from trying to start a snapshot | 995 | * will protect the lockfs code from trying to start a snapshot |
998 | * while we are mounting | 996 | * while we are mounting |
999 | */ | 997 | */ |
1000 | mutex_lock(&bdev->bd_fsfreeze_mutex); | 998 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1001 | if (bdev->bd_fsfreeze_count > 0) { | 999 | if (bdev->bd_fsfreeze_count > 0) { |
1002 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | 1000 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1003 | error = -EBUSY; | 1001 | error = -EBUSY; |
1004 | goto error_bdev; | 1002 | goto error_bdev; |
1005 | } | 1003 | } |
1006 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, | 1004 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
1007 | bdev); | 1005 | bdev); |
1008 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | 1006 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1009 | if (IS_ERR(s)) | 1007 | if (IS_ERR(s)) |
1010 | goto error_s; | 1008 | goto error_s; |
1011 | 1009 | ||
1012 | if (s->s_root) { | 1010 | if (s->s_root) { |
1013 | if ((flags ^ s->s_flags) & MS_RDONLY) { | 1011 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
1014 | deactivate_locked_super(s); | 1012 | deactivate_locked_super(s); |
1015 | error = -EBUSY; | 1013 | error = -EBUSY; |
1016 | goto error_bdev; | 1014 | goto error_bdev; |
1017 | } | 1015 | } |
1018 | 1016 | ||
1019 | /* | 1017 | /* |
1020 | * s_umount nests inside bd_mutex during | 1018 | * s_umount nests inside bd_mutex during |
1021 | * __invalidate_device(). blkdev_put() acquires | 1019 | * __invalidate_device(). blkdev_put() acquires |
1022 | * bd_mutex and can't be called under s_umount. Drop | 1020 | * bd_mutex and can't be called under s_umount. Drop |
1023 | * s_umount temporarily. This is safe as we're | 1021 | * s_umount temporarily. This is safe as we're |
1024 | * holding an active reference. | 1022 | * holding an active reference. |
1025 | */ | 1023 | */ |
1026 | up_write(&s->s_umount); | 1024 | up_write(&s->s_umount); |
1027 | blkdev_put(bdev, mode); | 1025 | blkdev_put(bdev, mode); |
1028 | down_write(&s->s_umount); | 1026 | down_write(&s->s_umount); |
1029 | } else { | 1027 | } else { |
1030 | char b[BDEVNAME_SIZE]; | 1028 | char b[BDEVNAME_SIZE]; |
1031 | 1029 | ||
1032 | s->s_mode = mode; | 1030 | s->s_mode = mode; |
1033 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | 1031 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
1034 | sb_set_blocksize(s, block_size(bdev)); | 1032 | sb_set_blocksize(s, block_size(bdev)); |
1035 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); | 1033 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1036 | if (error) { | 1034 | if (error) { |
1037 | deactivate_locked_super(s); | 1035 | deactivate_locked_super(s); |
1038 | goto error; | 1036 | goto error; |
1039 | } | 1037 | } |
1040 | 1038 | ||
1041 | s->s_flags |= MS_ACTIVE; | 1039 | s->s_flags |= MS_ACTIVE; |
1042 | bdev->bd_super = s; | 1040 | bdev->bd_super = s; |
1043 | } | 1041 | } |
1044 | 1042 | ||
1045 | return dget(s->s_root); | 1043 | return dget(s->s_root); |
1046 | 1044 | ||
1047 | error_s: | 1045 | error_s: |
1048 | error = PTR_ERR(s); | 1046 | error = PTR_ERR(s); |
1049 | error_bdev: | 1047 | error_bdev: |
1050 | blkdev_put(bdev, mode); | 1048 | blkdev_put(bdev, mode); |
1051 | error: | 1049 | error: |
1052 | return ERR_PTR(error); | 1050 | return ERR_PTR(error); |
1053 | } | 1051 | } |
1054 | EXPORT_SYMBOL(mount_bdev); | 1052 | EXPORT_SYMBOL(mount_bdev); |
1055 | 1053 | ||
1056 | void kill_block_super(struct super_block *sb) | 1054 | void kill_block_super(struct super_block *sb) |
1057 | { | 1055 | { |
1058 | struct block_device *bdev = sb->s_bdev; | 1056 | struct block_device *bdev = sb->s_bdev; |
1059 | fmode_t mode = sb->s_mode; | 1057 | fmode_t mode = sb->s_mode; |
1060 | 1058 | ||
1061 | bdev->bd_super = NULL; | 1059 | bdev->bd_super = NULL; |
1062 | generic_shutdown_super(sb); | 1060 | generic_shutdown_super(sb); |
1063 | sync_blockdev(bdev); | 1061 | sync_blockdev(bdev); |
1064 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); | 1062 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
1065 | blkdev_put(bdev, mode | FMODE_EXCL); | 1063 | blkdev_put(bdev, mode | FMODE_EXCL); |
1066 | } | 1064 | } |
1067 | 1065 | ||
1068 | EXPORT_SYMBOL(kill_block_super); | 1066 | EXPORT_SYMBOL(kill_block_super); |
1069 | #endif | 1067 | #endif |
1070 | 1068 | ||
1071 | struct dentry *mount_nodev(struct file_system_type *fs_type, | 1069 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1072 | int flags, void *data, | 1070 | int flags, void *data, |
1073 | int (*fill_super)(struct super_block *, void *, int)) | 1071 | int (*fill_super)(struct super_block *, void *, int)) |
1074 | { | 1072 | { |
1075 | int error; | 1073 | int error; |
1076 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); | 1074 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1077 | 1075 | ||
1078 | if (IS_ERR(s)) | 1076 | if (IS_ERR(s)) |
1079 | return ERR_CAST(s); | 1077 | return ERR_CAST(s); |
1080 | 1078 | ||
1081 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); | 1079 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1082 | if (error) { | 1080 | if (error) { |
1083 | deactivate_locked_super(s); | 1081 | deactivate_locked_super(s); |
1084 | return ERR_PTR(error); | 1082 | return ERR_PTR(error); |
1085 | } | 1083 | } |
1086 | s->s_flags |= MS_ACTIVE; | 1084 | s->s_flags |= MS_ACTIVE; |
1087 | return dget(s->s_root); | 1085 | return dget(s->s_root); |
1088 | } | 1086 | } |
1089 | EXPORT_SYMBOL(mount_nodev); | 1087 | EXPORT_SYMBOL(mount_nodev); |
1090 | 1088 | ||
1091 | static int compare_single(struct super_block *s, void *p) | 1089 | static int compare_single(struct super_block *s, void *p) |
1092 | { | 1090 | { |
1093 | return 1; | 1091 | return 1; |
1094 | } | 1092 | } |
1095 | 1093 | ||
1096 | struct dentry *mount_single(struct file_system_type *fs_type, | 1094 | struct dentry *mount_single(struct file_system_type *fs_type, |
1097 | int flags, void *data, | 1095 | int flags, void *data, |
1098 | int (*fill_super)(struct super_block *, void *, int)) | 1096 | int (*fill_super)(struct super_block *, void *, int)) |
1099 | { | 1097 | { |
1100 | struct super_block *s; | 1098 | struct super_block *s; |
1101 | int error; | 1099 | int error; |
1102 | 1100 | ||
1103 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); | 1101 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1104 | if (IS_ERR(s)) | 1102 | if (IS_ERR(s)) |
1105 | return ERR_CAST(s); | 1103 | return ERR_CAST(s); |
1106 | if (!s->s_root) { | 1104 | if (!s->s_root) { |
1107 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); | 1105 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1108 | if (error) { | 1106 | if (error) { |
1109 | deactivate_locked_super(s); | 1107 | deactivate_locked_super(s); |
1110 | return ERR_PTR(error); | 1108 | return ERR_PTR(error); |
1111 | } | 1109 | } |
1112 | s->s_flags |= MS_ACTIVE; | 1110 | s->s_flags |= MS_ACTIVE; |
1113 | } else { | 1111 | } else { |
1114 | do_remount_sb(s, flags, data, 0); | 1112 | do_remount_sb(s, flags, data, 0); |
1115 | } | 1113 | } |
1116 | return dget(s->s_root); | 1114 | return dget(s->s_root); |
1117 | } | 1115 | } |
1118 | EXPORT_SYMBOL(mount_single); | 1116 | EXPORT_SYMBOL(mount_single); |
1119 | 1117 | ||
1120 | struct dentry * | 1118 | struct dentry * |
1121 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | 1119 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) |
1122 | { | 1120 | { |
1123 | struct dentry *root; | 1121 | struct dentry *root; |
1124 | struct super_block *sb; | 1122 | struct super_block *sb; |
1125 | char *secdata = NULL; | 1123 | char *secdata = NULL; |
1126 | int error = -ENOMEM; | 1124 | int error = -ENOMEM; |
1127 | 1125 | ||
1128 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { | 1126 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1129 | secdata = alloc_secdata(); | 1127 | secdata = alloc_secdata(); |
1130 | if (!secdata) | 1128 | if (!secdata) |
1131 | goto out; | 1129 | goto out; |
1132 | 1130 | ||
1133 | error = security_sb_copy_data(data, secdata); | 1131 | error = security_sb_copy_data(data, secdata); |
1134 | if (error) | 1132 | if (error) |
1135 | goto out_free_secdata; | 1133 | goto out_free_secdata; |
1136 | } | 1134 | } |
1137 | 1135 | ||
1138 | root = type->mount(type, flags, name, data); | 1136 | root = type->mount(type, flags, name, data); |
1139 | if (IS_ERR(root)) { | 1137 | if (IS_ERR(root)) { |
1140 | error = PTR_ERR(root); | 1138 | error = PTR_ERR(root); |
1141 | goto out_free_secdata; | 1139 | goto out_free_secdata; |
1142 | } | 1140 | } |
1143 | sb = root->d_sb; | 1141 | sb = root->d_sb; |
1144 | BUG_ON(!sb); | 1142 | BUG_ON(!sb); |
1145 | WARN_ON(!sb->s_bdi); | 1143 | WARN_ON(!sb->s_bdi); |
1146 | WARN_ON(sb->s_bdi == &default_backing_dev_info); | 1144 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
1147 | sb->s_flags |= MS_BORN; | 1145 | sb->s_flags |= MS_BORN; |
1148 | 1146 | ||
1149 | error = security_sb_kern_mount(sb, flags, secdata); | 1147 | error = security_sb_kern_mount(sb, flags, secdata); |
1150 | if (error) | 1148 | if (error) |
1151 | goto out_sb; | 1149 | goto out_sb; |
1152 | 1150 | ||
1153 | /* | 1151 | /* |
1154 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | 1152 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE |
1155 | * but s_maxbytes was an unsigned long long for many releases. Throw | 1153 | * but s_maxbytes was an unsigned long long for many releases. Throw |
1156 | * this warning for a little while to try and catch filesystems that | 1154 | * this warning for a little while to try and catch filesystems that |
1157 | * violate this rule. | 1155 | * violate this rule. |
1158 | */ | 1156 | */ |
1159 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " | 1157 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1160 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | 1158 | "negative value (%lld)\n", type->name, sb->s_maxbytes); |
1161 | 1159 | ||
1162 | up_write(&sb->s_umount); | 1160 | up_write(&sb->s_umount); |
1163 | free_secdata(secdata); | 1161 | free_secdata(secdata); |
1164 | return root; | 1162 | return root; |
1165 | out_sb: | 1163 | out_sb: |
1166 | dput(root); | 1164 | dput(root); |
1167 | deactivate_locked_super(sb); | 1165 | deactivate_locked_super(sb); |
1168 | out_free_secdata: | 1166 | out_free_secdata: |
1169 | free_secdata(secdata); | 1167 | free_secdata(secdata); |
1170 | out: | 1168 | out: |
1171 | return ERR_PTR(error); | 1169 | return ERR_PTR(error); |
1172 | } | 1170 | } |
1173 | 1171 | ||
1174 | /* | 1172 | /* |
1175 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | 1173 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} |
1176 | * instead. | 1174 | * instead. |
1177 | */ | 1175 | */ |
1178 | void __sb_end_write(struct super_block *sb, int level) | 1176 | void __sb_end_write(struct super_block *sb, int level) |
1179 | { | 1177 | { |
1180 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | 1178 | percpu_counter_dec(&sb->s_writers.counter[level-1]); |
1181 | /* | 1179 | /* |
1182 | * Make sure s_writers are updated before we wake up waiters in | 1180 | * Make sure s_writers are updated before we wake up waiters in |
1183 | * freeze_super(). | 1181 | * freeze_super(). |
1184 | */ | 1182 | */ |
1185 | smp_mb(); | 1183 | smp_mb(); |
1186 | if (waitqueue_active(&sb->s_writers.wait)) | 1184 | if (waitqueue_active(&sb->s_writers.wait)) |
1187 | wake_up(&sb->s_writers.wait); | 1185 | wake_up(&sb->s_writers.wait); |
1188 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | 1186 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); |
1189 | } | 1187 | } |
1190 | EXPORT_SYMBOL(__sb_end_write); | 1188 | EXPORT_SYMBOL(__sb_end_write); |
1191 | 1189 | ||
1192 | #ifdef CONFIG_LOCKDEP | 1190 | #ifdef CONFIG_LOCKDEP |
1193 | /* | 1191 | /* |
1194 | * We want lockdep to tell us about possible deadlocks with freezing but | 1192 | * We want lockdep to tell us about possible deadlocks with freezing but |
1195 | * it's it bit tricky to properly instrument it. Getting a freeze protection | 1193 | * it's it bit tricky to properly instrument it. Getting a freeze protection |
1196 | * works as getting a read lock but there are subtle problems. XFS for example | 1194 | * works as getting a read lock but there are subtle problems. XFS for example |
1197 | * gets freeze protection on internal level twice in some cases, which is OK | 1195 | * gets freeze protection on internal level twice in some cases, which is OK |
1198 | * only because we already hold a freeze protection also on higher level. Due | 1196 | * only because we already hold a freeze protection also on higher level. Due |
1199 | * to these cases we have to tell lockdep we are doing trylock when we | 1197 | * to these cases we have to tell lockdep we are doing trylock when we |
1200 | * already hold a freeze protection for a higher freeze level. | 1198 | * already hold a freeze protection for a higher freeze level. |
1201 | */ | 1199 | */ |
1202 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | 1200 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, |
1203 | unsigned long ip) | 1201 | unsigned long ip) |
1204 | { | 1202 | { |
1205 | int i; | 1203 | int i; |
1206 | 1204 | ||
1207 | if (!trylock) { | 1205 | if (!trylock) { |
1208 | for (i = 0; i < level - 1; i++) | 1206 | for (i = 0; i < level - 1; i++) |
1209 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | 1207 | if (lock_is_held(&sb->s_writers.lock_map[i])) { |
1210 | trylock = true; | 1208 | trylock = true; |
1211 | break; | 1209 | break; |
1212 | } | 1210 | } |
1213 | } | 1211 | } |
1214 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | 1212 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); |
1215 | } | 1213 | } |
1216 | #endif | 1214 | #endif |
1217 | 1215 | ||
1218 | /* | 1216 | /* |
1219 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | 1217 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} |
1220 | * instead. | 1218 | * instead. |
1221 | */ | 1219 | */ |
1222 | int __sb_start_write(struct super_block *sb, int level, bool wait) | 1220 | int __sb_start_write(struct super_block *sb, int level, bool wait) |
1223 | { | 1221 | { |
1224 | retry: | 1222 | retry: |
1225 | if (unlikely(sb->s_writers.frozen >= level)) { | 1223 | if (unlikely(sb->s_writers.frozen >= level)) { |
1226 | if (!wait) | 1224 | if (!wait) |
1227 | return 0; | 1225 | return 0; |
1228 | wait_event(sb->s_writers.wait_unfrozen, | 1226 | wait_event(sb->s_writers.wait_unfrozen, |
1229 | sb->s_writers.frozen < level); | 1227 | sb->s_writers.frozen < level); |
1230 | } | 1228 | } |
1231 | 1229 | ||
1232 | #ifdef CONFIG_LOCKDEP | 1230 | #ifdef CONFIG_LOCKDEP |
1233 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | 1231 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); |
1234 | #endif | 1232 | #endif |
1235 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | 1233 | percpu_counter_inc(&sb->s_writers.counter[level-1]); |
1236 | /* | 1234 | /* |
1237 | * Make sure counter is updated before we check for frozen. | 1235 | * Make sure counter is updated before we check for frozen. |
1238 | * freeze_super() first sets frozen and then checks the counter. | 1236 | * freeze_super() first sets frozen and then checks the counter. |
1239 | */ | 1237 | */ |
1240 | smp_mb(); | 1238 | smp_mb(); |
1241 | if (unlikely(sb->s_writers.frozen >= level)) { | 1239 | if (unlikely(sb->s_writers.frozen >= level)) { |
1242 | __sb_end_write(sb, level); | 1240 | __sb_end_write(sb, level); |
1243 | goto retry; | 1241 | goto retry; |
1244 | } | 1242 | } |
1245 | return 1; | 1243 | return 1; |
1246 | } | 1244 | } |
1247 | EXPORT_SYMBOL(__sb_start_write); | 1245 | EXPORT_SYMBOL(__sb_start_write); |
1248 | 1246 | ||
1249 | /** | 1247 | /** |
1250 | * sb_wait_write - wait until all writers to given file system finish | 1248 | * sb_wait_write - wait until all writers to given file system finish |
1251 | * @sb: the super for which we wait | 1249 | * @sb: the super for which we wait |
1252 | * @level: type of writers we wait for (normal vs page fault) | 1250 | * @level: type of writers we wait for (normal vs page fault) |
1253 | * | 1251 | * |
1254 | * This function waits until there are no writers of given type to given file | 1252 | * This function waits until there are no writers of given type to given file |
1255 | * system. Caller of this function should make sure there can be no new writers | 1253 | * system. Caller of this function should make sure there can be no new writers |
1256 | * of type @level before calling this function. Otherwise this function can | 1254 | * of type @level before calling this function. Otherwise this function can |
1257 | * livelock. | 1255 | * livelock. |
1258 | */ | 1256 | */ |
1259 | static void sb_wait_write(struct super_block *sb, int level) | 1257 | static void sb_wait_write(struct super_block *sb, int level) |
1260 | { | 1258 | { |
1261 | s64 writers; | 1259 | s64 writers; |
1262 | 1260 | ||
1263 | /* | 1261 | /* |
1264 | * We just cycle-through lockdep here so that it does not complain | 1262 | * We just cycle-through lockdep here so that it does not complain |
1265 | * about returning with lock to userspace | 1263 | * about returning with lock to userspace |
1266 | */ | 1264 | */ |
1267 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | 1265 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); |
1268 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | 1266 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); |
1269 | 1267 | ||
1270 | do { | 1268 | do { |
1271 | DEFINE_WAIT(wait); | 1269 | DEFINE_WAIT(wait); |
1272 | 1270 | ||
1273 | /* | 1271 | /* |
1274 | * We use a barrier in prepare_to_wait() to separate setting | 1272 | * We use a barrier in prepare_to_wait() to separate setting |
1275 | * of frozen and checking of the counter | 1273 | * of frozen and checking of the counter |
1276 | */ | 1274 | */ |
1277 | prepare_to_wait(&sb->s_writers.wait, &wait, | 1275 | prepare_to_wait(&sb->s_writers.wait, &wait, |
1278 | TASK_UNINTERRUPTIBLE); | 1276 | TASK_UNINTERRUPTIBLE); |
1279 | 1277 | ||
1280 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | 1278 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); |
1281 | if (writers) | 1279 | if (writers) |
1282 | schedule(); | 1280 | schedule(); |
1283 | 1281 | ||
1284 | finish_wait(&sb->s_writers.wait, &wait); | 1282 | finish_wait(&sb->s_writers.wait, &wait); |
1285 | } while (writers); | 1283 | } while (writers); |
1286 | } | 1284 | } |
1287 | 1285 | ||
1288 | /** | 1286 | /** |
1289 | * freeze_super - lock the filesystem and force it into a consistent state | 1287 | * freeze_super - lock the filesystem and force it into a consistent state |
1290 | * @sb: the super to lock | 1288 | * @sb: the super to lock |
1291 | * | 1289 | * |
1292 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | 1290 | * Syncs the super to make sure the filesystem is consistent and calls the fs's |
1293 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | 1291 | * freeze_fs. Subsequent calls to this without first thawing the fs will return |
1294 | * -EBUSY. | 1292 | * -EBUSY. |
1295 | * | 1293 | * |
1296 | * During this function, sb->s_writers.frozen goes through these values: | 1294 | * During this function, sb->s_writers.frozen goes through these values: |
1297 | * | 1295 | * |
1298 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | 1296 | * SB_UNFROZEN: File system is normal, all writes progress as usual. |
1299 | * | 1297 | * |
1300 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | 1298 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New |
1301 | * writes should be blocked, though page faults are still allowed. We wait for | 1299 | * writes should be blocked, though page faults are still allowed. We wait for |
1302 | * all writes to complete and then proceed to the next stage. | 1300 | * all writes to complete and then proceed to the next stage. |
1303 | * | 1301 | * |
1304 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | 1302 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked |
1305 | * but internal fs threads can still modify the filesystem (although they | 1303 | * but internal fs threads can still modify the filesystem (although they |
1306 | * should not dirty new pages or inodes), writeback can run etc. After waiting | 1304 | * should not dirty new pages or inodes), writeback can run etc. After waiting |
1307 | * for all running page faults we sync the filesystem which will clean all | 1305 | * for all running page faults we sync the filesystem which will clean all |
1308 | * dirty pages and inodes (no new dirty pages or inodes can be created when | 1306 | * dirty pages and inodes (no new dirty pages or inodes can be created when |
1309 | * sync is running). | 1307 | * sync is running). |
1310 | * | 1308 | * |
1311 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | 1309 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs |
1312 | * modification are blocked (e.g. XFS preallocation truncation on inode | 1310 | * modification are blocked (e.g. XFS preallocation truncation on inode |
1313 | * reclaim). This is usually implemented by blocking new transactions for | 1311 | * reclaim). This is usually implemented by blocking new transactions for |
1314 | * filesystems that have them and need this additional guard. After all | 1312 | * filesystems that have them and need this additional guard. After all |
1315 | * internal writers are finished we call ->freeze_fs() to finish filesystem | 1313 | * internal writers are finished we call ->freeze_fs() to finish filesystem |
1316 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | 1314 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is |
1317 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | 1315 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. |
1318 | * | 1316 | * |
1319 | * sb->s_writers.frozen is protected by sb->s_umount. | 1317 | * sb->s_writers.frozen is protected by sb->s_umount. |
1320 | */ | 1318 | */ |
1321 | int freeze_super(struct super_block *sb) | 1319 | int freeze_super(struct super_block *sb) |
1322 | { | 1320 | { |
1323 | int ret; | 1321 | int ret; |
1324 | 1322 | ||
1325 | atomic_inc(&sb->s_active); | 1323 | atomic_inc(&sb->s_active); |
1326 | down_write(&sb->s_umount); | 1324 | down_write(&sb->s_umount); |
1327 | if (sb->s_writers.frozen != SB_UNFROZEN) { | 1325 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
1328 | deactivate_locked_super(sb); | 1326 | deactivate_locked_super(sb); |
1329 | return -EBUSY; | 1327 | return -EBUSY; |
1330 | } | 1328 | } |
1331 | 1329 | ||
1332 | if (!(sb->s_flags & MS_BORN)) { | 1330 | if (!(sb->s_flags & MS_BORN)) { |
1333 | up_write(&sb->s_umount); | 1331 | up_write(&sb->s_umount); |
1334 | return 0; /* sic - it's "nothing to do" */ | 1332 | return 0; /* sic - it's "nothing to do" */ |
1335 | } | 1333 | } |
1336 | 1334 | ||
1337 | if (sb->s_flags & MS_RDONLY) { | 1335 | if (sb->s_flags & MS_RDONLY) { |
1338 | /* Nothing to do really... */ | 1336 | /* Nothing to do really... */ |
1339 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | 1337 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
1340 | up_write(&sb->s_umount); | 1338 | up_write(&sb->s_umount); |
1341 | return 0; | 1339 | return 0; |
1342 | } | 1340 | } |
1343 | 1341 | ||
1344 | /* From now on, no new normal writers can start */ | 1342 | /* From now on, no new normal writers can start */ |
1345 | sb->s_writers.frozen = SB_FREEZE_WRITE; | 1343 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
1346 | smp_wmb(); | 1344 | smp_wmb(); |
1347 | 1345 | ||
1348 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | 1346 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1349 | up_write(&sb->s_umount); | 1347 | up_write(&sb->s_umount); |
1350 | 1348 | ||
1351 | sb_wait_write(sb, SB_FREEZE_WRITE); | 1349 | sb_wait_write(sb, SB_FREEZE_WRITE); |
1352 | 1350 | ||
1353 | /* Now we go and block page faults... */ | 1351 | /* Now we go and block page faults... */ |
1354 | down_write(&sb->s_umount); | 1352 | down_write(&sb->s_umount); |
1355 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | 1353 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
1356 | smp_wmb(); | 1354 | smp_wmb(); |
1357 | 1355 | ||
1358 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); | 1356 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1359 | 1357 | ||
1360 | /* All writers are done so after syncing there won't be dirty data */ | 1358 | /* All writers are done so after syncing there won't be dirty data */ |
1361 | sync_filesystem(sb); | 1359 | sync_filesystem(sb); |
1362 | 1360 | ||
1363 | /* Now wait for internal filesystem counter */ | 1361 | /* Now wait for internal filesystem counter */ |
1364 | sb->s_writers.frozen = SB_FREEZE_FS; | 1362 | sb->s_writers.frozen = SB_FREEZE_FS; |
1365 | smp_wmb(); | 1363 | smp_wmb(); |
1366 | sb_wait_write(sb, SB_FREEZE_FS); | 1364 | sb_wait_write(sb, SB_FREEZE_FS); |
1367 | 1365 | ||
1368 | if (sb->s_op->freeze_fs) { | 1366 | if (sb->s_op->freeze_fs) { |
1369 | ret = sb->s_op->freeze_fs(sb); | 1367 | ret = sb->s_op->freeze_fs(sb); |
1370 | if (ret) { | 1368 | if (ret) { |
1371 | printk(KERN_ERR | 1369 | printk(KERN_ERR |
1372 | "VFS:Filesystem freeze failed\n"); | 1370 | "VFS:Filesystem freeze failed\n"); |
1373 | sb->s_writers.frozen = SB_UNFROZEN; | 1371 | sb->s_writers.frozen = SB_UNFROZEN; |
1374 | smp_wmb(); | 1372 | smp_wmb(); |
1375 | wake_up(&sb->s_writers.wait_unfrozen); | 1373 | wake_up(&sb->s_writers.wait_unfrozen); |
1376 | deactivate_locked_super(sb); | 1374 | deactivate_locked_super(sb); |
1377 | return ret; | 1375 | return ret; |
1378 | } | 1376 | } |
1379 | } | 1377 | } |
1380 | /* | 1378 | /* |
1381 | * This is just for debugging purposes so that fs can warn if it | 1379 | * This is just for debugging purposes so that fs can warn if it |
1382 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | 1380 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. |
1383 | */ | 1381 | */ |
1384 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | 1382 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
1385 | up_write(&sb->s_umount); | 1383 | up_write(&sb->s_umount); |
1386 | return 0; | 1384 | return 0; |
1387 | } | 1385 | } |
1388 | EXPORT_SYMBOL(freeze_super); | 1386 | EXPORT_SYMBOL(freeze_super); |
1389 | 1387 | ||
1390 | /** | 1388 | /** |
1391 | * thaw_super -- unlock filesystem | 1389 | * thaw_super -- unlock filesystem |
1392 | * @sb: the super to thaw | 1390 | * @sb: the super to thaw |
1393 | * | 1391 | * |
1394 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | 1392 | * Unlocks the filesystem and marks it writeable again after freeze_super(). |
1395 | */ | 1393 | */ |
1396 | int thaw_super(struct super_block *sb) | 1394 | int thaw_super(struct super_block *sb) |
1397 | { | 1395 | { |
1398 | int error; | 1396 | int error; |
1399 | 1397 | ||
1400 | down_write(&sb->s_umount); | 1398 | down_write(&sb->s_umount); |
1401 | if (sb->s_writers.frozen == SB_UNFROZEN) { | 1399 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
1402 | up_write(&sb->s_umount); | 1400 | up_write(&sb->s_umount); |
1403 | return -EINVAL; | 1401 | return -EINVAL; |
1404 | } | 1402 | } |
1405 | 1403 | ||
1406 | if (sb->s_flags & MS_RDONLY) | 1404 | if (sb->s_flags & MS_RDONLY) |
1407 | goto out; | 1405 | goto out; |
1408 | 1406 | ||
1409 | if (sb->s_op->unfreeze_fs) { | 1407 | if (sb->s_op->unfreeze_fs) { |
1410 | error = sb->s_op->unfreeze_fs(sb); | 1408 | error = sb->s_op->unfreeze_fs(sb); |
1411 | if (error) { | 1409 | if (error) { |
1412 | printk(KERN_ERR | 1410 | printk(KERN_ERR |
1413 | "VFS:Filesystem thaw failed\n"); | 1411 | "VFS:Filesystem thaw failed\n"); |
1414 | up_write(&sb->s_umount); | 1412 | up_write(&sb->s_umount); |
1415 | return error; | 1413 | return error; |
1416 | } | 1414 | } |
1417 | } | 1415 | } |
1418 | 1416 | ||
1419 | out: | 1417 | out: |
1420 | sb->s_writers.frozen = SB_UNFROZEN; | 1418 | sb->s_writers.frozen = SB_UNFROZEN; |
1421 | smp_wmb(); | 1419 | smp_wmb(); |
1422 | wake_up(&sb->s_writers.wait_unfrozen); | 1420 | wake_up(&sb->s_writers.wait_unfrozen); |
1423 | deactivate_locked_super(sb); | 1421 | deactivate_locked_super(sb); |
1424 | 1422 | ||
1425 | return 0; | 1423 | return 0; |
1426 | } | 1424 | } |
1427 | EXPORT_SYMBOL(thaw_super); | 1425 | EXPORT_SYMBOL(thaw_super); |