blob: bc35a40417d7ada53223f5c93ed26875af97a646 [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
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>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070017#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/mount.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080019#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/cdev.h>
Robert Love0eeca282005-07-12 17:06:03 -040021#include <linux/fsnotify.h>
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080022#include <linux/sysctl.h>
23#include <linux/percpu_counter.h>
24
25#include <asm/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
27/* sysctl tunables... */
28struct files_stat_struct files_stat = {
29 .max_files = NR_FILE
30};
31
Linus Torvalds1da177e2005-04-16 15:20:36 -070032/* public. Not pretty! */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080033__cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080035static struct percpu_counter nr_files __cacheline_aligned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070037static inline void file_free_rcu(struct rcu_head *head)
38{
Eric Dumazet2f512012005-10-30 15:02:16 -080039 struct file *f = container_of(head, struct file, f_u.fu_rcuhead);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070040 kmem_cache_free(filp_cachep, f);
41}
42
Linus Torvalds1da177e2005-04-16 15:20:36 -070043static inline void file_free(struct file *f)
44{
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080045 percpu_counter_dec(&nr_files);
Eric Dumazet2f512012005-10-30 15:02:16 -080046 call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
Linus Torvalds1da177e2005-04-16 15:20:36 -070047}
48
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080049/*
50 * Return the total number of open files in the system
51 */
52static int get_nr_files(void)
53{
54 return percpu_counter_read_positive(&nr_files);
55}
56
57/*
58 * Return the maximum number of open files in the system
59 */
60int get_max_files(void)
61{
62 return files_stat.max_files;
63}
64EXPORT_SYMBOL_GPL(get_max_files);
65
66/*
67 * Handle nr_files sysctl
68 */
69#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
70int proc_nr_files(ctl_table *table, int write, struct file *filp,
71 void __user *buffer, size_t *lenp, loff_t *ppos)
72{
73 files_stat.nr_files = get_nr_files();
74 return proc_dointvec(table, write, filp, buffer, lenp, ppos);
75}
76#else
77int proc_nr_files(ctl_table *table, int write, struct file *filp,
78 void __user *buffer, size_t *lenp, loff_t *ppos)
79{
80 return -ENOSYS;
81}
82#endif
83
Linus Torvalds1da177e2005-04-16 15:20:36 -070084/* Find an unused file structure and return a pointer to it.
85 * Returns NULL, if there are no more free file structures or
86 * we run out of memory.
87 */
88struct file *get_empty_filp(void)
89{
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -080090 struct task_struct *tsk;
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -070091 static int old_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 struct file * f;
93
94 /*
95 * Privileged users can go above max_files
96 */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080097 if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
98 /*
99 * percpu_counters are inaccurate. Do an expensive check before
100 * we go and fail.
101 */
102 if (percpu_counter_sum(&nr_files) >= files_stat.max_files)
103 goto over;
104 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700106 f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
107 if (f == NULL)
108 goto fail;
109
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800110 percpu_counter_inc(&nr_files);
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700111 memset(f, 0, sizeof(*f));
112 if (security_file_alloc(f))
113 goto fail_sec;
114
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -0800115 tsk = current;
Eric Dumazet2f512012005-10-30 15:02:16 -0800116 INIT_LIST_HEAD(&f->f_u.fu_list);
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -0800117 atomic_set(&f->f_count, 1);
118 rwlock_init(&f->f_owner.lock);
119 f->f_uid = tsk->fsuid;
120 f->f_gid = tsk->fsgid;
121 eventpoll_init_file(f);
122 /* f->f_version: 0 */
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700123 return f;
124
125over:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 /* Ran out of filps - report that */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800127 if (get_nr_files() > old_max) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 printk(KERN_INFO "VFS: file-max limit %d reached\n",
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800129 get_max_files());
130 old_max = get_nr_files();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 }
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700132 goto fail;
133
134fail_sec:
135 file_free(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136fail:
137 return NULL;
138}
139
140EXPORT_SYMBOL(get_empty_filp);
141
142void fastcall fput(struct file *file)
143{
Nick Piggin095975d2006-01-08 01:02:19 -0800144 if (atomic_dec_and_test(&file->f_count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 __fput(file);
146}
147
148EXPORT_SYMBOL(fput);
149
150/* __fput is called from task context when aio completion releases the last
151 * last use of a struct file *. Do not use otherwise.
152 */
153void fastcall __fput(struct file *file)
154{
155 struct dentry *dentry = file->f_dentry;
156 struct vfsmount *mnt = file->f_vfsmnt;
157 struct inode *inode = dentry->d_inode;
158
159 might_sleep();
Robert Love0eeca282005-07-12 17:06:03 -0400160
161 fsnotify_close(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 /*
163 * The function eventpoll_release() should be the first called
164 * in the file cleanup chain.
165 */
166 eventpoll_release(file);
167 locks_remove_flock(file);
168
169 if (file->f_op && file->f_op->release)
170 file->f_op->release(inode, file);
171 security_file_free(file);
Theodore Ts'o577c4eb2006-09-27 01:50:49 -0700172 if (unlikely(S_ISCHR(inode->i_mode) && inode->i_cdev != NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 cdev_put(inode->i_cdev);
174 fops_put(file->f_op);
175 if (file->f_mode & FMODE_WRITE)
176 put_write_access(inode);
177 file_kill(file);
178 file->f_dentry = NULL;
179 file->f_vfsmnt = NULL;
180 file_free(file);
181 dput(dentry);
182 mntput(mnt);
183}
184
185struct file fastcall *fget(unsigned int fd)
186{
187 struct file *file;
188 struct files_struct *files = current->files;
189
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700190 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 file = fcheck_files(files, fd);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700192 if (file) {
Nick Piggin095975d2006-01-08 01:02:19 -0800193 if (!atomic_inc_not_zero(&file->f_count)) {
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700194 /* File object ref couldn't be taken */
195 rcu_read_unlock();
196 return NULL;
197 }
198 }
199 rcu_read_unlock();
200
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 return file;
202}
203
204EXPORT_SYMBOL(fget);
205
206/*
207 * Lightweight file lookup - no refcnt increment if fd table isn't shared.
208 * You can use this only if it is guranteed that the current task already
209 * holds a refcnt to that file. That check has to be done at fget() only
210 * and a flag is returned to be passed to the corresponding fput_light().
211 * There must not be a cloning between an fget_light/fput_light pair.
212 */
213struct file fastcall *fget_light(unsigned int fd, int *fput_needed)
214{
215 struct file *file;
216 struct files_struct *files = current->files;
217
218 *fput_needed = 0;
219 if (likely((atomic_read(&files->count) == 1))) {
220 file = fcheck_files(files, fd);
221 } else {
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700222 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 file = fcheck_files(files, fd);
224 if (file) {
Nick Piggin095975d2006-01-08 01:02:19 -0800225 if (atomic_inc_not_zero(&file->f_count))
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700226 *fput_needed = 1;
227 else
228 /* Didn't get the reference, someone's freed */
229 file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 }
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700231 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 }
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700233
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 return file;
235}
236
237
238void put_filp(struct file *file)
239{
Nick Piggin095975d2006-01-08 01:02:19 -0800240 if (atomic_dec_and_test(&file->f_count)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 security_file_free(file);
242 file_kill(file);
243 file_free(file);
244 }
245}
246
247void file_move(struct file *file, struct list_head *list)
248{
249 if (!list)
250 return;
251 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800252 list_move(&file->f_u.fu_list, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 file_list_unlock();
254}
255
256void file_kill(struct file *file)
257{
Eric Dumazet2f512012005-10-30 15:02:16 -0800258 if (!list_empty(&file->f_u.fu_list)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800260 list_del_init(&file->f_u.fu_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 file_list_unlock();
262 }
263}
264
265int fs_may_remount_ro(struct super_block *sb)
266{
267 struct list_head *p;
268
269 /* Check that no files are currently opened for writing. */
270 file_list_lock();
271 list_for_each(p, &sb->s_files) {
Eric Dumazet2f512012005-10-30 15:02:16 -0800272 struct file *file = list_entry(p, struct file, f_u.fu_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 struct inode *inode = file->f_dentry->d_inode;
274
275 /* File with pending delete? */
276 if (inode->i_nlink == 0)
277 goto too_bad;
278
279 /* Writeable file? */
280 if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE))
281 goto too_bad;
282 }
283 file_list_unlock();
284 return 1; /* Tis' cool bro. */
285too_bad:
286 file_list_unlock();
287 return 0;
288}
289
290void __init files_init(unsigned long mempages)
291{
292 int n;
293 /* One file with associated inode and dcache is very roughly 1K.
294 * Per default don't use more than 10% of our memory for files.
295 */
296
297 n = (mempages * (PAGE_SIZE / 1024)) / 10;
298 files_stat.max_files = n;
299 if (files_stat.max_files < NR_FILE)
300 files_stat.max_files = NR_FILE;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700301 files_defer_init();
Mingming Cao0216bfc2006-06-23 02:05:41 -0700302 percpu_counter_init(&nr_files, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303}