blob: 287fdce7f767fe3b3b61be8e95905e379086df81 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/fs.h>
14#include <linux/security.h>
15#include <linux/eventpoll.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070016#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/mount.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080018#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/cdev.h>
Robert Love0eeca282005-07-12 17:06:03 -040020#include <linux/fsnotify.h>
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080021#include <linux/sysctl.h>
22#include <linux/percpu_counter.h>
23
24#include <asm/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
26/* sysctl tunables... */
27struct files_stat_struct files_stat = {
28 .max_files = NR_FILE
29};
30
Linus Torvalds1da177e2005-04-16 15:20:36 -070031/* public. Not pretty! */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080032__cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080034static struct percpu_counter nr_files __cacheline_aligned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070036static inline void file_free_rcu(struct rcu_head *head)
37{
Eric Dumazet2f512012005-10-30 15:02:16 -080038 struct file *f = container_of(head, struct file, f_u.fu_rcuhead);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070039 kmem_cache_free(filp_cachep, f);
40}
41
Linus Torvalds1da177e2005-04-16 15:20:36 -070042static inline void file_free(struct file *f)
43{
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080044 percpu_counter_dec(&nr_files);
Eric Dumazet2f512012005-10-30 15:02:16 -080045 call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
Linus Torvalds1da177e2005-04-16 15:20:36 -070046}
47
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080048/*
49 * Return the total number of open files in the system
50 */
51static int get_nr_files(void)
52{
53 return percpu_counter_read_positive(&nr_files);
54}
55
56/*
57 * Return the maximum number of open files in the system
58 */
59int get_max_files(void)
60{
61 return files_stat.max_files;
62}
63EXPORT_SYMBOL_GPL(get_max_files);
64
65/*
66 * Handle nr_files sysctl
67 */
68#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
69int proc_nr_files(ctl_table *table, int write, struct file *filp,
70 void __user *buffer, size_t *lenp, loff_t *ppos)
71{
72 files_stat.nr_files = get_nr_files();
73 return proc_dointvec(table, write, filp, buffer, lenp, ppos);
74}
75#else
76int proc_nr_files(ctl_table *table, int write, struct file *filp,
77 void __user *buffer, size_t *lenp, loff_t *ppos)
78{
79 return -ENOSYS;
80}
81#endif
82
Linus Torvalds1da177e2005-04-16 15:20:36 -070083/* Find an unused file structure and return a pointer to it.
84 * Returns NULL, if there are no more free file structures or
85 * we run out of memory.
86 */
87struct file *get_empty_filp(void)
88{
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -080089 struct task_struct *tsk;
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -070090 static int old_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 struct file * f;
92
93 /*
94 * Privileged users can go above max_files
95 */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -080096 if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
97 /*
98 * percpu_counters are inaccurate. Do an expensive check before
99 * we go and fail.
100 */
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -0700101 if (percpu_counter_sum_positive(&nr_files) >= files_stat.max_files)
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800102 goto over;
103 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700105 f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
106 if (f == NULL)
107 goto fail;
108
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800109 percpu_counter_inc(&nr_files);
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700110 memset(f, 0, sizeof(*f));
111 if (security_file_alloc(f))
112 goto fail_sec;
113
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -0800114 tsk = current;
Eric Dumazet2f512012005-10-30 15:02:16 -0800115 INIT_LIST_HEAD(&f->f_u.fu_list);
Benjamin LaHaise5a6b7952006-03-23 03:01:03 -0800116 atomic_set(&f->f_count, 1);
117 rwlock_init(&f->f_owner.lock);
118 f->f_uid = tsk->fsuid;
119 f->f_gid = tsk->fsgid;
120 eventpoll_init_file(f);
121 /* f->f_version: 0 */
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700122 return f;
123
124over:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 /* Ran out of filps - report that */
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800126 if (get_nr_files() > old_max) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 printk(KERN_INFO "VFS: file-max limit %d reached\n",
Dipankar Sarma529bf6b2006-03-07 21:55:35 -0800128 get_max_files());
129 old_max = get_nr_files();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 }
Kirill Korotaevaf4d2ec2005-06-23 00:09:50 -0700131 goto fail;
132
133fail_sec:
134 file_free(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135fail:
136 return NULL;
137}
138
139EXPORT_SYMBOL(get_empty_filp);
140
141void fastcall fput(struct file *file)
142{
Nick Piggin095975d2006-01-08 01:02:19 -0800143 if (atomic_dec_and_test(&file->f_count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 __fput(file);
145}
146
147EXPORT_SYMBOL(fput);
148
149/* __fput is called from task context when aio completion releases the last
150 * last use of a struct file *. Do not use otherwise.
151 */
152void fastcall __fput(struct file *file)
153{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800154 struct dentry *dentry = file->f_path.dentry;
155 struct vfsmount *mnt = file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 struct inode *inode = dentry->d_inode;
157
158 might_sleep();
Robert Love0eeca282005-07-12 17:06:03 -0400159
160 fsnotify_close(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 /*
162 * The function eventpoll_release() should be the first called
163 * in the file cleanup chain.
164 */
165 eventpoll_release(file);
166 locks_remove_flock(file);
167
168 if (file->f_op && file->f_op->release)
169 file->f_op->release(inode, file);
170 security_file_free(file);
Theodore Ts'o577c4eb2006-09-27 01:50:49 -0700171 if (unlikely(S_ISCHR(inode->i_mode) && inode->i_cdev != NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 cdev_put(inode->i_cdev);
173 fops_put(file->f_op);
174 if (file->f_mode & FMODE_WRITE)
175 put_write_access(inode);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700176 put_pid(file->f_owner.pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 file_kill(file);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800178 file->f_path.dentry = NULL;
179 file->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 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);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800273 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
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}