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/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>
15 #include <linux/security.h>
16 #include <linux/eventpoll.h>
17 #include <linux/mount.h>
18 #include <linux/cdev.h>
19 #include <linux/vs_limit.h>
21 /* sysctl tunables... */
22 struct files_stat_struct files_stat = {
26 EXPORT_SYMBOL(files_stat); /* Needed by unix.o */
28 /* public *and* exported. Not pretty! */
29 spinlock_t __cacheline_aligned_in_smp files_lock = SPIN_LOCK_UNLOCKED;
31 EXPORT_SYMBOL(files_lock);
33 static spinlock_t filp_count_lock = SPIN_LOCK_UNLOCKED;
35 /* slab constructors and destructors are called from arbitrary
36 * context and must be fully threaded - use a local spinlock
37 * to protect files_stat.nr_files
39 void filp_ctor(void * objp, struct kmem_cache_s *cachep, unsigned long cflags)
41 if ((cflags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
42 SLAB_CTOR_CONSTRUCTOR) {
44 spin_lock_irqsave(&filp_count_lock, flags);
45 files_stat.nr_files++;
46 spin_unlock_irqrestore(&filp_count_lock, flags);
50 void filp_dtor(void * objp, struct kmem_cache_s *cachep, unsigned long dflags)
53 spin_lock_irqsave(&filp_count_lock, flags);
54 files_stat.nr_files--;
55 spin_unlock_irqrestore(&filp_count_lock, flags);
58 static inline void file_free(struct file *f)
60 kmem_cache_free(filp_cachep, f);
63 /* Find an unused file structure and return a pointer to it.
64 * Returns NULL, if there are no more free file structures or
65 * we run out of memory.
67 struct file *get_empty_filp(void)
73 * Privileged users can go above max_files
75 if (files_stat.nr_files < files_stat.max_files ||
76 capable(CAP_SYS_ADMIN)) {
77 f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
79 memset(f, 0, sizeof(*f));
80 if (security_file_alloc(f)) {
84 eventpoll_init_file(f);
85 atomic_set(&f->f_count, 1);
86 f->f_uid = current->fsuid;
87 f->f_gid = current->fsgid;
88 f->f_owner.lock = RW_LOCK_UNLOCKED;
90 INIT_LIST_HEAD(&f->f_list);
96 /* Ran out of filps - report that */
97 if (files_stat.max_files >= old_max) {
98 printk(KERN_INFO "VFS: file-max limit %d reached\n",
99 files_stat.max_files);
100 old_max = files_stat.max_files;
102 /* Big problems... */
103 printk(KERN_WARNING "VFS: filp allocation failed\n");
109 EXPORT_SYMBOL(get_empty_filp);
112 * Clear and initialize a (private) struct file for the given dentry,
113 * allocate the security structure, and call the open function (if any).
114 * The file should be released using close_private_file.
116 int open_private_file(struct file *filp, struct dentry *dentry, int flags)
119 memset(filp, 0, sizeof(*filp));
120 eventpoll_init_file(filp);
121 filp->f_flags = flags;
122 filp->f_mode = (flags+1) & O_ACCMODE;
123 atomic_set(&filp->f_count, 1);
124 filp->f_dentry = dentry;
125 filp->f_mapping = dentry->d_inode->i_mapping;
126 filp->f_uid = current->fsuid;
127 filp->f_gid = current->fsgid;
128 filp->f_op = dentry->d_inode->i_fop;
129 INIT_LIST_HEAD(&filp->f_list);
130 error = security_file_alloc(filp);
132 if (filp->f_op && filp->f_op->open) {
133 error = filp->f_op->open(dentry->d_inode, filp);
135 security_file_free(filp);
140 EXPORT_SYMBOL(open_private_file);
143 * Release a private file by calling the release function (if any) and
144 * freeing the security structure.
146 void close_private_file(struct file *file)
148 struct inode * inode = file->f_dentry->d_inode;
150 if (file->f_op && file->f_op->release)
151 file->f_op->release(inode, file);
152 security_file_free(file);
155 EXPORT_SYMBOL(close_private_file);
157 void fastcall fput(struct file *file)
159 if (atomic_dec_and_test(&file->f_count))
165 /* __fput is called from task context when aio completion releases the last
166 * last use of a struct file *. Do not use otherwise.
168 void fastcall __fput(struct file *file)
170 struct dentry *dentry = file->f_dentry;
171 struct vfsmount *mnt = file->f_vfsmnt;
172 struct inode *inode = dentry->d_inode;
175 * The function eventpoll_release() should be the first called
176 * in the file cleanup chain.
178 eventpoll_release(file);
179 locks_remove_flock(file);
181 if (file->f_op && file->f_op->release)
182 file->f_op->release(inode, file);
183 security_file_free(file);
184 if (unlikely(inode->i_cdev != NULL))
185 cdev_put(inode->i_cdev);
186 fops_put(file->f_op);
187 if (file->f_mode & FMODE_WRITE)
188 put_write_access(inode);
191 file->f_dentry = NULL;
192 file->f_vfsmnt = NULL;
198 struct file fastcall *fget(unsigned int fd)
201 struct files_struct *files = current->files;
203 spin_lock(&files->file_lock);
204 file = fcheck_files(files, fd);
207 spin_unlock(&files->file_lock);
214 * Lightweight file lookup - no refcnt increment if fd table isn't shared.
215 * You can use this only if it is guranteed that the current task already
216 * holds a refcnt to that file. That check has to be done at fget() only
217 * and a flag is returned to be passed to the corresponding fput_light().
218 * There must not be a cloning between an fget_light/fput_light pair.
220 struct file fastcall *fget_light(unsigned int fd, int *fput_needed)
223 struct files_struct *files = current->files;
226 if (likely((atomic_read(&files->count) == 1))) {
227 file = fcheck_files(files, fd);
229 spin_lock(&files->file_lock);
230 file = fcheck_files(files, fd);
235 spin_unlock(&files->file_lock);
241 void put_filp(struct file *file)
243 if (atomic_dec_and_test(&file->f_count)) {
244 security_file_free(file);
250 EXPORT_SYMBOL(put_filp);
252 void file_move(struct file *file, struct list_head *list)
257 list_move(&file->f_list, list);
261 void file_kill(struct file *file)
263 if (!list_empty(&file->f_list)) {
265 list_del_init(&file->f_list);
270 int fs_may_remount_ro(struct super_block *sb)
274 /* Check that no files are currently opened for writing. */
276 list_for_each(p, &sb->s_files) {
277 struct file *file = list_entry(p, struct file, f_list);
278 struct inode *inode = file->f_dentry->d_inode;
280 /* File with pending delete? */
281 if (inode->i_nlink == 0)
284 /* Writeable file? */
285 if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE))
289 return 1; /* Tis' cool bro. */
295 void __init files_init(unsigned long mempages)
298 /* One file with associated inode and dcache is very roughly 1K.
299 * Per default don't use more than 10% of our memory for files.
302 n = (mempages * (PAGE_SIZE / 1024)) / 10;
303 files_stat.max_files = n;
304 if (files_stat.max_files < NR_FILE)
305 files_stat.max_files = NR_FILE;