2 * linux/fs/file_table.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
8 #include <linux/config.h>
9 #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>
16 #include <linux/security.h>
17 #include <linux/eventpoll.h>
18 #include <linux/rcupdate.h>
19 #include <linux/mount.h>
20 #include <linux/capability.h>
21 #include <linux/cdev.h>
22 #include <linux/fsnotify.h>
23 #include <linux/sysctl.h>
24 #include <linux/percpu_counter.h>
25 #include <linux/vs_limit.h>
26 #include <linux/vs_context.h>
28 #include <asm/atomic.h>
30 /* sysctl tunables... */
31 struct files_stat_struct files_stat = {
35 /* public. Not pretty! */
36 __cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock);
38 static struct percpu_counter nr_files __cacheline_aligned_in_smp;
40 static inline void file_free_rcu(struct rcu_head *head)
42 struct file *f = container_of(head, struct file, f_u.fu_rcuhead);
43 kmem_cache_free(filp_cachep, f);
46 static inline void file_free(struct file *f)
48 percpu_counter_dec(&nr_files);
49 call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
53 * Return the total number of open files in the system
55 static int get_nr_files(void)
57 return percpu_counter_read_positive(&nr_files);
61 * Return the maximum number of open files in the system
63 int get_max_files(void)
65 return files_stat.max_files;
67 EXPORT_SYMBOL_GPL(get_max_files);
70 * Handle nr_files sysctl
72 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
73 int proc_nr_files(ctl_table *table, int write, struct file *filp,
74 void __user *buffer, size_t *lenp, loff_t *ppos)
76 files_stat.nr_files = get_nr_files();
77 return proc_dointvec(table, write, filp, buffer, lenp, ppos);
80 int proc_nr_files(ctl_table *table, int write, struct file *filp,
81 void __user *buffer, size_t *lenp, loff_t *ppos)
87 /* Find an unused file structure and return a pointer to it.
88 * Returns NULL, if there are no more free file structures or
89 * we run out of memory.
91 struct file *get_empty_filp(void)
97 * Privileged users can go above max_files
99 if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
101 * percpu_counters are inaccurate. Do an expensive check before
104 if (percpu_counter_sum(&nr_files) >= files_stat.max_files)
108 f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
112 percpu_counter_inc(&nr_files);
113 memset(f, 0, sizeof(*f));
114 if (security_file_alloc(f))
117 eventpoll_init_file(f);
118 atomic_set(&f->f_count, 1);
119 f->f_uid = current->fsuid;
120 f->f_gid = current->fsgid;
121 rwlock_init(&f->f_owner.lock);
122 /* f->f_version: 0 */
123 INIT_LIST_HEAD(&f->f_u.fu_list);
124 f->f_xid = vx_current_xid();
129 /* Ran out of filps - report that */
130 if (get_nr_files() > old_max) {
131 printk(KERN_INFO "VFS: file-max limit %d reached\n",
133 old_max = get_nr_files();
143 EXPORT_SYMBOL(get_empty_filp);
145 void fastcall fput(struct file *file)
147 if (atomic_dec_and_test(&file->f_count))
153 /* __fput is called from task context when aio completion releases the last
154 * last use of a struct file *. Do not use otherwise.
156 void fastcall __fput(struct file *file)
158 struct dentry *dentry = file->f_dentry;
159 struct vfsmount *mnt = file->f_vfsmnt;
160 struct inode *inode = dentry->d_inode;
164 fsnotify_close(file);
166 * The function eventpoll_release() should be the first called
167 * in the file cleanup chain.
169 eventpoll_release(file);
170 locks_remove_flock(file);
172 if (file->f_op && file->f_op->release)
173 file->f_op->release(inode, file);
174 security_file_free(file);
175 if (unlikely(inode->i_cdev != NULL))
176 cdev_put(inode->i_cdev);
177 fops_put(file->f_op);
178 if (file->f_mode & FMODE_WRITE)
179 put_write_access(inode);
183 file->f_dentry = NULL;
184 file->f_vfsmnt = NULL;
190 struct file fastcall *fget(unsigned int fd)
193 struct files_struct *files = current->files;
196 file = fcheck_files(files, fd);
198 if (!atomic_inc_not_zero(&file->f_count)) {
199 /* File object ref couldn't be taken */
212 * Lightweight file lookup - no refcnt increment if fd table isn't shared.
213 * You can use this only if it is guranteed that the current task already
214 * holds a refcnt to that file. That check has to be done at fget() only
215 * and a flag is returned to be passed to the corresponding fput_light().
216 * There must not be a cloning between an fget_light/fput_light pair.
218 struct file fastcall *fget_light(unsigned int fd, int *fput_needed)
221 struct files_struct *files = current->files;
224 if (likely((atomic_read(&files->count) == 1))) {
225 file = fcheck_files(files, fd);
228 file = fcheck_files(files, fd);
230 if (atomic_inc_not_zero(&file->f_count))
233 /* Didn't get the reference, someone's freed */
243 void put_filp(struct file *file)
245 if (atomic_dec_and_test(&file->f_count)) {
246 security_file_free(file);
254 void file_move(struct file *file, struct list_head *list)
259 list_move(&file->f_u.fu_list, list);
263 void file_kill(struct file *file)
265 if (!list_empty(&file->f_u.fu_list)) {
267 list_del_init(&file->f_u.fu_list);
272 int fs_may_remount_ro(struct super_block *sb)
276 /* Check that no files are currently opened for writing. */
278 list_for_each(p, &sb->s_files) {
279 struct file *file = list_entry(p, struct file, f_u.fu_list);
280 struct inode *inode = file->f_dentry->d_inode;
282 /* File with pending delete? */
283 if (inode->i_nlink == 0)
286 /* Writeable file? */
287 if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE))
291 return 1; /* Tis' cool bro. */
297 void __init files_init(unsigned long mempages)
300 /* One file with associated inode and dcache is very roughly 1K.
301 * Per default don't use more than 10% of our memory for files.
304 n = (mempages * (PAGE_SIZE / 1024)) / 10;
305 files_stat.max_files = n;
306 if (files_stat.max_files < NR_FILE)
307 files_stat.max_files = NR_FILE;
309 percpu_counter_init(&nr_files);