blob: 506d5307108d71be668f4c00d25960c9d7528307 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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
Dipankar Sarma529bf6b2006-03-07 21:55:35 -08008#include <linux/config.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/string.h>
10#include <linux/slab.h>
11#include <linux/file.h>
12#include <linux/init.h>
13#include <linux/module.h>
14#include <linux/smp_lock.h>
15#include <linux/fs.h>
16#include <linux/security.h>
17#include <linux/eventpoll.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070018#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/mount.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080020#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/cdev.h>
Robert Love0eeca282005-07-12 17:06:03 -040022#include <linux/fsnotify.h>
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080023#include <linux/sysctl.h>
24#include <linux/percpu_counter.h>
25
26#include <asm/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
28/* sysctl tunables... */
29struct files_stat_struct files_stat = {
30 .max_files = NR_FILE
31};
32
Linus Torvalds1da177e2005-04-16 15:20:36 -070033/* public. Not pretty! */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080034__cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080036static struct percpu_counter nr_files __cacheline_aligned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070038static inline void file_free_rcu(struct rcu_head *head)
39{
Eric Dumazet2f512012005-10-30 15:02:16 -080040 struct file *f = container_of(head, struct file, f_u.fu_rcuhead);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070041 kmem_cache_free(filp_cachep, f);
42}
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044static inline void file_free(struct file *f)
45{
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080046 percpu_counter_dec(&nr_files);
Eric Dumazet2f512012005-10-30 15:02:16 -080047 call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
Linus Torvalds1da177e2005-04-16 15:20:36 -070048}
49
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080050/*
51 * Return the total number of open files in the system
52 */
53static int get_nr_files(void)
54{
55 return percpu_counter_read_positive(&nr_files);
56}
57
58/*
59 * Return the maximum number of open files in the system
60 */
61int get_max_files(void)
62{
63 return files_stat.max_files;
64}
65EXPORT_SYMBOL_GPL(get_max_files);
66
67/*
68 * Handle nr_files sysctl
69 */
70#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
71int proc_nr_files(ctl_table *table, int write, struct file *filp,
72 void __user *buffer, size_t *lenp, loff_t *ppos)
73{
74 files_stat.nr_files = get_nr_files();
75 return proc_dointvec(table, write, filp, buffer, lenp, ppos);
76}
77#else
78int proc_nr_files(ctl_table *table, int write, struct file *filp,
79 void __user *buffer, size_t *lenp, loff_t *ppos)
80{
81 return -ENOSYS;
82}
83#endif
84
Linus Torvalds1da177e2005-04-16 15:20:36 -070085/* Find an unused file structure and return a pointer to it.
86 * Returns NULL, if there are no more free file structures or
87 * we run out of memory.
88 */
89struct file *get_empty_filp(void)
90{
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -080091 struct task_struct *tsk;
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -070092 static int old_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 struct file * f;
94
95 /*
96 * Privileged users can go above max_files
97 */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080098 if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
99 /*
100 * percpu_counters are inaccurate. Do an expensive check before
101 * we go and fail.
102 */
103 if (percpu_counter_sum(&nr_files) >= files_stat.max_files)
104 goto over;
105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700107 f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
108 if (f == NULL)
109 goto fail;
110
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800111 percpu_counter_inc(&nr_files);
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700112 memset(f, 0, sizeof(*f));
113 if (security_file_alloc(f))
114 goto fail_sec;
115
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -0800116 tsk = current;
Eric Dumazet2f512012005-10-30 15:02:16 -0800117 INIT_LIST_HEAD(&f->f_u.fu_list);
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -0800118 atomic_set(&f->f_count, 1);
119 rwlock_init(&f->f_owner.lock);
120 f->f_uid = tsk->fsuid;
121 f->f_gid = tsk->fsgid;
122 eventpoll_init_file(f);
123 /* f->f_version: 0 */
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700124 return f;
125
126over:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 /* Ran out of filps - report that */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800128 if (get_nr_files() > old_max) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 printk(KERN_INFO "VFS: file-max limit %d reached\n",
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800130 get_max_files());
131 old_max = get_nr_files();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 }
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700133 goto fail;
134
135fail_sec:
136 file_free(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137fail:
138 return NULL;
139}
140
141EXPORT_SYMBOL(get_empty_filp);
142
143void fastcall fput(struct file *file)
144{
Nick Piggin095975d2006-01-08 01:02:19 -0800145 if (atomic_dec_and_test(&file->f_count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 __fput(file);
147}
148
149EXPORT_SYMBOL(fput);
150
151/* __fput is called from task context when aio completion releases the last
152 * last use of a struct file *. Do not use otherwise.
153 */
154void fastcall __fput(struct file *file)
155{
156 struct dentry *dentry = file->f_dentry;
157 struct vfsmount *mnt = file->f_vfsmnt;
158 struct inode *inode = dentry->d_inode;
159
160 might_sleep();
Robert Love0eeca282005-07-12 17:06:03 -0400161
162 fsnotify_close(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 /*
164 * The function eventpoll_release() should be the first called
165 * in the file cleanup chain.
166 */
167 eventpoll_release(file);
168 locks_remove_flock(file);
169
170 if (file->f_op && file->f_op->release)
171 file->f_op->release(inode, file);
172 security_file_free(file);
173 if (unlikely(inode->i_cdev != NULL))
174 cdev_put(inode->i_cdev);
175 fops_put(file->f_op);
176 if (file->f_mode & FMODE_WRITE)
177 put_write_access(inode);
178 file_kill(file);
179 file->f_dentry = NULL;
180 file->f_vfsmnt = NULL;
181 file_free(file);
182 dput(dentry);
183 mntput(mnt);
184}
185
186struct file fastcall *fget(unsigned int fd)
187{
188 struct file *file;
189 struct files_struct *files = current->files;
190
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700191 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 file = fcheck_files(files, fd);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700193 if (file) {
Nick Piggin095975d2006-01-08 01:02:19 -0800194 if (!atomic_inc_not_zero(&file->f_count)) {
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700195 /* File object ref couldn't be taken */
196 rcu_read_unlock();
197 return NULL;
198 }
199 }
200 rcu_read_unlock();
201
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 return file;
203}
204
205EXPORT_SYMBOL(fget);
206
207/*
208 * Lightweight file lookup - no refcnt increment if fd table isn't shared.
209 * You can use this only if it is guranteed that the current task already
210 * holds a refcnt to that file. That check has to be done at fget() only
211 * and a flag is returned to be passed to the corresponding fput_light().
212 * There must not be a cloning between an fget_light/fput_light pair.
213 */
214struct file fastcall *fget_light(unsigned int fd, int *fput_needed)
215{
216 struct file *file;
217 struct files_struct *files = current->files;
218
219 *fput_needed = 0;
220 if (likely((atomic_read(&files->count) == 1))) {
221 file = fcheck_files(files, fd);
222 } else {
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700223 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 file = fcheck_files(files, fd);
225 if (file) {
Nick Piggin095975d2006-01-08 01:02:19 -0800226 if (atomic_inc_not_zero(&file->f_count))
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700227 *fput_needed = 1;
228 else
229 /* Didn't get the reference, someone's freed */
230 file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 }
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700232 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 }
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 return file;
236}
237
238
239void put_filp(struct file *file)
240{
Nick Piggin095975d2006-01-08 01:02:19 -0800241 if (atomic_dec_and_test(&file->f_count)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 security_file_free(file);
243 file_kill(file);
244 file_free(file);
245 }
246}
247
248void file_move(struct file *file, struct list_head *list)
249{
250 if (!list)
251 return;
252 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800253 list_move(&file->f_u.fu_list, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 file_list_unlock();
255}
256
257void file_kill(struct file *file)
258{
Eric Dumazet2f512012005-10-30 15:02:16 -0800259 if (!list_empty(&file->f_u.fu_list)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800261 list_del_init(&file->f_u.fu_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 file_list_unlock();
263 }
264}
265
266int fs_may_remount_ro(struct super_block *sb)
267{
268 struct list_head *p;
269
270 /* Check that no files are currently opened for writing. */
271 file_list_lock();
272 list_for_each(p, &sb->s_files) {
Eric Dumazet2f512012005-10-30 15:02:16 -0800273 struct file *file = list_entry(p, struct file, f_u.fu_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 struct inode *inode = file->f_dentry->d_inode;
275
276 /* File with pending delete? */
277 if (inode->i_nlink == 0)
278 goto too_bad;
279
280 /* Writeable file? */
281 if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE))
282 goto too_bad;
283 }
284 file_list_unlock();
285 return 1; /* Tis' cool bro. */
286too_bad:
287 file_list_unlock();
288 return 0;
289}
290
291void __init files_init(unsigned long mempages)
292{
293 int n;
294 /* One file with associated inode and dcache is very roughly 1K.
295 * Per default don't use more than 10% of our memory for files.
296 */
297
298 n = (mempages * (PAGE_SIZE / 1024)) / 10;
299 files_stat.max_files = n;
300 if (files_stat.max_files < NR_FILE)
301 files_stat.max_files = NR_FILE;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700302 files_defer_init();
Mingming Cao0216bfc2006-06-23 02:05:41 -0700303 percpu_counter_init(&nr_files, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304}