Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/file_table.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu) | |
6 | */ | |
7 | ||
8 | #include <linux/string.h> | |
9 | #include <linux/slab.h> | |
10 | #include <linux/file.h> | |
11 | #include <linux/init.h> | |
12 | #include <linux/module.h> | |
13 | #include <linux/smp_lock.h> | |
14 | #include <linux/fs.h> | |
15 | #include <linux/security.h> | |
16 | #include <linux/eventpoll.h> | |
17 | #include <linux/mount.h> | |
18 | #include <linux/cdev.h> | |
0eeca283 | 19 | #include <linux/fsnotify.h> |
1da177e4 LT |
20 | |
21 | /* sysctl tunables... */ | |
22 | struct files_stat_struct files_stat = { | |
23 | .max_files = NR_FILE | |
24 | }; | |
25 | ||
26 | EXPORT_SYMBOL(files_stat); /* Needed by unix.o */ | |
27 | ||
28 | /* public. Not pretty! */ | |
29 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock); | |
30 | ||
31 | static DEFINE_SPINLOCK(filp_count_lock); | |
32 | ||
33 | /* slab constructors and destructors are called from arbitrary | |
34 | * context and must be fully threaded - use a local spinlock | |
35 | * to protect files_stat.nr_files | |
36 | */ | |
37 | void filp_ctor(void * objp, struct kmem_cache_s *cachep, unsigned long cflags) | |
38 | { | |
39 | if ((cflags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
40 | SLAB_CTOR_CONSTRUCTOR) { | |
41 | unsigned long flags; | |
42 | spin_lock_irqsave(&filp_count_lock, flags); | |
43 | files_stat.nr_files++; | |
44 | spin_unlock_irqrestore(&filp_count_lock, flags); | |
45 | } | |
46 | } | |
47 | ||
48 | void filp_dtor(void * objp, struct kmem_cache_s *cachep, unsigned long dflags) | |
49 | { | |
50 | unsigned long flags; | |
51 | spin_lock_irqsave(&filp_count_lock, flags); | |
52 | files_stat.nr_files--; | |
53 | spin_unlock_irqrestore(&filp_count_lock, flags); | |
54 | } | |
55 | ||
56 | static inline void file_free(struct file *f) | |
57 | { | |
58 | kmem_cache_free(filp_cachep, f); | |
59 | } | |
60 | ||
61 | /* Find an unused file structure and return a pointer to it. | |
62 | * Returns NULL, if there are no more free file structures or | |
63 | * we run out of memory. | |
64 | */ | |
65 | struct file *get_empty_filp(void) | |
66 | { | |
af4d2ecb | 67 | static int old_max; |
1da177e4 LT |
68 | struct file * f; |
69 | ||
70 | /* | |
71 | * Privileged users can go above max_files | |
72 | */ | |
af4d2ecb KK |
73 | if (files_stat.nr_files >= files_stat.max_files && |
74 | !capable(CAP_SYS_ADMIN)) | |
75 | goto over; | |
76 | ||
77 | f = kmem_cache_alloc(filp_cachep, GFP_KERNEL); | |
78 | if (f == NULL) | |
79 | goto fail; | |
80 | ||
81 | memset(f, 0, sizeof(*f)); | |
82 | if (security_file_alloc(f)) | |
83 | goto fail_sec; | |
1da177e4 | 84 | |
af4d2ecb KK |
85 | eventpoll_init_file(f); |
86 | atomic_set(&f->f_count, 1); | |
87 | f->f_uid = current->fsuid; | |
88 | f->f_gid = current->fsgid; | |
89 | rwlock_init(&f->f_owner.lock); | |
90 | /* f->f_version: 0 */ | |
91 | INIT_LIST_HEAD(&f->f_list); | |
92 | f->f_maxcount = INT_MAX; | |
93 | return f; | |
94 | ||
95 | over: | |
1da177e4 | 96 | /* Ran out of filps - report that */ |
af4d2ecb | 97 | if (files_stat.nr_files > old_max) { |
1da177e4 LT |
98 | printk(KERN_INFO "VFS: file-max limit %d reached\n", |
99 | files_stat.max_files); | |
af4d2ecb | 100 | old_max = files_stat.nr_files; |
1da177e4 | 101 | } |
af4d2ecb KK |
102 | goto fail; |
103 | ||
104 | fail_sec: | |
105 | file_free(f); | |
1da177e4 LT |
106 | fail: |
107 | return NULL; | |
108 | } | |
109 | ||
110 | EXPORT_SYMBOL(get_empty_filp); | |
111 | ||
112 | void fastcall fput(struct file *file) | |
113 | { | |
114 | if (atomic_dec_and_test(&file->f_count)) | |
115 | __fput(file); | |
116 | } | |
117 | ||
118 | EXPORT_SYMBOL(fput); | |
119 | ||
120 | /* __fput is called from task context when aio completion releases the last | |
121 | * last use of a struct file *. Do not use otherwise. | |
122 | */ | |
123 | void fastcall __fput(struct file *file) | |
124 | { | |
125 | struct dentry *dentry = file->f_dentry; | |
126 | struct vfsmount *mnt = file->f_vfsmnt; | |
127 | struct inode *inode = dentry->d_inode; | |
128 | ||
129 | might_sleep(); | |
0eeca283 RL |
130 | |
131 | fsnotify_close(file); | |
1da177e4 LT |
132 | /* |
133 | * The function eventpoll_release() should be the first called | |
134 | * in the file cleanup chain. | |
135 | */ | |
136 | eventpoll_release(file); | |
137 | locks_remove_flock(file); | |
138 | ||
139 | if (file->f_op && file->f_op->release) | |
140 | file->f_op->release(inode, file); | |
141 | security_file_free(file); | |
142 | if (unlikely(inode->i_cdev != NULL)) | |
143 | cdev_put(inode->i_cdev); | |
144 | fops_put(file->f_op); | |
145 | if (file->f_mode & FMODE_WRITE) | |
146 | put_write_access(inode); | |
147 | file_kill(file); | |
148 | file->f_dentry = NULL; | |
149 | file->f_vfsmnt = NULL; | |
150 | file_free(file); | |
151 | dput(dentry); | |
152 | mntput(mnt); | |
153 | } | |
154 | ||
155 | struct file fastcall *fget(unsigned int fd) | |
156 | { | |
157 | struct file *file; | |
158 | struct files_struct *files = current->files; | |
159 | ||
160 | spin_lock(&files->file_lock); | |
161 | file = fcheck_files(files, fd); | |
162 | if (file) | |
163 | get_file(file); | |
164 | spin_unlock(&files->file_lock); | |
165 | return file; | |
166 | } | |
167 | ||
168 | EXPORT_SYMBOL(fget); | |
169 | ||
170 | /* | |
171 | * Lightweight file lookup - no refcnt increment if fd table isn't shared. | |
172 | * You can use this only if it is guranteed that the current task already | |
173 | * holds a refcnt to that file. That check has to be done at fget() only | |
174 | * and a flag is returned to be passed to the corresponding fput_light(). | |
175 | * There must not be a cloning between an fget_light/fput_light pair. | |
176 | */ | |
177 | struct file fastcall *fget_light(unsigned int fd, int *fput_needed) | |
178 | { | |
179 | struct file *file; | |
180 | struct files_struct *files = current->files; | |
181 | ||
182 | *fput_needed = 0; | |
183 | if (likely((atomic_read(&files->count) == 1))) { | |
184 | file = fcheck_files(files, fd); | |
185 | } else { | |
186 | spin_lock(&files->file_lock); | |
187 | file = fcheck_files(files, fd); | |
188 | if (file) { | |
189 | get_file(file); | |
190 | *fput_needed = 1; | |
191 | } | |
192 | spin_unlock(&files->file_lock); | |
193 | } | |
194 | return file; | |
195 | } | |
196 | ||
197 | ||
198 | void put_filp(struct file *file) | |
199 | { | |
200 | if (atomic_dec_and_test(&file->f_count)) { | |
201 | security_file_free(file); | |
202 | file_kill(file); | |
203 | file_free(file); | |
204 | } | |
205 | } | |
206 | ||
207 | void file_move(struct file *file, struct list_head *list) | |
208 | { | |
209 | if (!list) | |
210 | return; | |
211 | file_list_lock(); | |
212 | list_move(&file->f_list, list); | |
213 | file_list_unlock(); | |
214 | } | |
215 | ||
216 | void file_kill(struct file *file) | |
217 | { | |
218 | if (!list_empty(&file->f_list)) { | |
219 | file_list_lock(); | |
220 | list_del_init(&file->f_list); | |
221 | file_list_unlock(); | |
222 | } | |
223 | } | |
224 | ||
225 | int fs_may_remount_ro(struct super_block *sb) | |
226 | { | |
227 | struct list_head *p; | |
228 | ||
229 | /* Check that no files are currently opened for writing. */ | |
230 | file_list_lock(); | |
231 | list_for_each(p, &sb->s_files) { | |
232 | struct file *file = list_entry(p, struct file, f_list); | |
233 | struct inode *inode = file->f_dentry->d_inode; | |
234 | ||
235 | /* File with pending delete? */ | |
236 | if (inode->i_nlink == 0) | |
237 | goto too_bad; | |
238 | ||
239 | /* Writeable file? */ | |
240 | if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE)) | |
241 | goto too_bad; | |
242 | } | |
243 | file_list_unlock(); | |
244 | return 1; /* Tis' cool bro. */ | |
245 | too_bad: | |
246 | file_list_unlock(); | |
247 | return 0; | |
248 | } | |
249 | ||
250 | void __init files_init(unsigned long mempages) | |
251 | { | |
252 | int n; | |
253 | /* One file with associated inode and dcache is very roughly 1K. | |
254 | * Per default don't use more than 10% of our memory for files. | |
255 | */ | |
256 | ||
257 | n = (mempages * (PAGE_SIZE / 1024)) / 10; | |
258 | files_stat.max_files = n; | |
259 | if (files_stat.max_files < NR_FILE) | |
260 | files_stat.max_files = NR_FILE; | |
261 | } |