file_table.c 8.55 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9 10
/*
 *  linux/fs/file_table.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *  Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
 */

#include <linux/string.h>
#include <linux/slab.h>
#include <linux/file.h>
Al Viro's avatar
Al Viro committed
11
#include <linux/fdtable.h>
Linus Torvalds's avatar
Linus Torvalds committed
12 13 14 15
#include <linux/init.h>
#include <linux/module.h>
#include <linux/fs.h>
#include <linux/security.h>
16
#include <linux/cred.h>
Linus Torvalds's avatar
Linus Torvalds committed
17
#include <linux/eventpoll.h>
18
#include <linux/rcupdate.h>
Linus Torvalds's avatar
Linus Torvalds committed
19
#include <linux/mount.h>
20
#include <linux/capability.h>
Linus Torvalds's avatar
Linus Torvalds committed
21
#include <linux/cdev.h>
Robert Love's avatar
Robert Love committed
22
#include <linux/fsnotify.h>
23 24
#include <linux/sysctl.h>
#include <linux/percpu_counter.h>
Nick Piggin's avatar
Nick Piggin committed
25
#include <linux/percpu.h>
Al Viro's avatar
Al Viro committed
26
#include <linux/task_work.h>
27
#include <linux/ima.h>
28
#include <linux/swap.h>
29

30
#include <linux/atomic.h>
Linus Torvalds's avatar
Linus Torvalds committed
31

32 33
#include "internal.h"

Linus Torvalds's avatar
Linus Torvalds committed
34 35 36 37 38
/* sysctl tunables... */
struct files_stat_struct files_stat = {
	.max_files = NR_FILE
};

39 40 41
/* SLAB cache for file structures */
static struct kmem_cache *filp_cachep __read_mostly;

42
static struct percpu_counter nr_files __cacheline_aligned_in_smp;
Linus Torvalds's avatar
Linus Torvalds committed
43

Al Viro's avatar
Al Viro committed
44
static void file_free_rcu(struct rcu_head *head)
Linus Torvalds's avatar
Linus Torvalds committed
45
{
46 47 48
	struct file *f = container_of(head, struct file, f_u.fu_rcuhead);

	put_cred(f->f_cred);
49
	kmem_cache_free(filp_cachep, f);
Linus Torvalds's avatar
Linus Torvalds committed
50 51
}

52
static inline void file_free(struct file *f)
Linus Torvalds's avatar
Linus Torvalds committed
53
{
54 55
	percpu_counter_dec(&nr_files);
	call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
Linus Torvalds's avatar
Linus Torvalds committed
56 57
}

58 59 60
/*
 * Return the total number of open files in the system
 */
61
static long get_nr_files(void)
Linus Torvalds's avatar
Linus Torvalds committed
62
{
63
	return percpu_counter_read_positive(&nr_files);
Linus Torvalds's avatar
Linus Torvalds committed
64 65
}

66 67 68
/*
 * Return the maximum number of open files in the system
 */
69
unsigned long get_max_files(void)
70
{
71
	return files_stat.max_files;
72
}
73 74 75 76 77 78
EXPORT_SYMBOL_GPL(get_max_files);

/*
 * Handle nr_files sysctl
 */
#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
79
int proc_nr_files(struct ctl_table *table, int write,
80 81 82
                     void __user *buffer, size_t *lenp, loff_t *ppos)
{
	files_stat.nr_files = get_nr_files();
83
	return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
84 85
}
#else
86
int proc_nr_files(struct ctl_table *table, int write,
87 88 89 90 91
                     void __user *buffer, size_t *lenp, loff_t *ppos)
{
	return -ENOSYS;
}
#endif
92

Linus Torvalds's avatar
Linus Torvalds committed
93
/* Find an unused file structure and return a pointer to it.
94 95
 * Returns an error pointer if some error happend e.g. we over file
 * structures limit, run out of memory or operation is not permitted.
96 97 98 99 100 101
 *
 * Be very careful using this.  You are responsible for
 * getting write access to any mount that you might assign
 * to this filp, if it is opened for write.  If this is not
 * done, you will imbalance int the mount's writer count
 * and a warning at __fput() time.
Linus Torvalds's avatar
Linus Torvalds committed
102 103 104
 */
struct file *get_empty_filp(void)
{
105
	const struct cred *cred = current_cred();
106
	static long old_max;
107 108
	struct file *f;
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
109 110 111 112

	/*
	 * Privileged users can go above max_files
	 */
113 114 115 116 117
	if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
		/*
		 * percpu_counters are inaccurate.  Do an expensive check before
		 * we go and fail.
		 */
118
		if (percpu_counter_sum_positive(&nr_files) >= files_stat.max_files)
119 120
			goto over;
	}
121

122
	f = kmem_cache_zalloc(filp_cachep, GFP_KERNEL);
123 124
	if (unlikely(!f))
		return ERR_PTR(-ENOMEM);
125

126
	percpu_counter_inc(&nr_files);
127
	f->f_cred = get_cred(cred);
128 129 130 131 132
	error = security_file_alloc(f);
	if (unlikely(error)) {
		file_free(f);
		return ERR_PTR(error);
	}
Linus Torvalds's avatar
Linus Torvalds committed
133

134
	atomic_long_set(&f->f_count, 1);
135
	rwlock_init(&f->f_owner.lock);
136
	spin_lock_init(&f->f_lock);
137
	mutex_init(&f->f_pos_lock);
138
	eventpoll_init_file(f);
139 140 141 142
	/* f->f_version: 0 */
	return f;

over:
Linus Torvalds's avatar
Linus Torvalds committed
143
	/* Ran out of filps - report that */
144
	if (get_nr_files() > old_max) {
145
		pr_info("VFS: file-max limit %lu reached\n", get_max_files());
146
		old_max = get_nr_files();
Linus Torvalds's avatar
Linus Torvalds committed
147
	}
148
	return ERR_PTR(-ENFILE);
Linus Torvalds's avatar
Linus Torvalds committed
149 150
}

151 152
/**
 * alloc_file - allocate and initialize a 'struct file'
153 154
 *
 * @path: the (dentry, vfsmount) pair for the new file
155 156 157
 * @mode: the mode with which the new file will be opened
 * @fop: the 'struct file_operations' for the new file
 */
Al Viro's avatar
Al Viro committed
158
struct file *alloc_file(const struct path *path, fmode_t mode,
159
		const struct file_operations *fop)
160 161 162 163
{
	struct file *file;

	file = get_empty_filp();
164
	if (IS_ERR(file))
165
		return file;
166

167
	file->f_path = *path;
168
	file->f_inode = path->dentry->d_inode;
169
	file->f_mapping = path->dentry->d_inode->i_mapping;
170
	file->f_wb_err = filemap_sample_wb_err(file->f_mapping);
171
	if ((mode & FMODE_READ) &&
172
	     likely(fop->read || fop->read_iter))
173
		mode |= FMODE_CAN_READ;
174
	if ((mode & FMODE_WRITE) &&
175
	     likely(fop->write || fop->write_iter))
176
		mode |= FMODE_CAN_WRITE;
177 178
	file->f_mode = mode;
	file->f_op = fop;
179 180
	if ((mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(path->dentry->d_inode);
Al Viro's avatar
Al Viro committed
181
	return file;
182
}
Roland Dreier's avatar
Roland Dreier committed
183
EXPORT_SYMBOL(alloc_file);
184

185
/* the real guts of fput() - releasing the last reference to file
Linus Torvalds's avatar
Linus Torvalds committed
186
 */
187
static void __fput(struct file *file)
Linus Torvalds's avatar
Linus Torvalds committed
188
{
189 190
	struct dentry *dentry = file->f_path.dentry;
	struct vfsmount *mnt = file->f_path.mnt;
191
	struct inode *inode = file->f_inode;
Linus Torvalds's avatar
Linus Torvalds committed
192 193

	might_sleep();
Robert Love's avatar
Robert Love committed
194 195

	fsnotify_close(file);
Linus Torvalds's avatar
Linus Torvalds committed
196 197 198 199 200
	/*
	 * The function eventpoll_release() should be the first called
	 * in the file cleanup chain.
	 */
	eventpoll_release(file);
201
	locks_remove_file(file);
Linus Torvalds's avatar
Linus Torvalds committed
202

203
	ima_file_free(file);
204
	if (unlikely(file->f_flags & FASYNC)) {
Al Viro's avatar
Al Viro committed
205
		if (file->f_op->fasync)
206 207
			file->f_op->fasync(-1, file, 0);
	}
Al Viro's avatar
Al Viro committed
208
	if (file->f_op->release)
Linus Torvalds's avatar
Linus Torvalds committed
209 210
		file->f_op->release(inode, file);
	security_file_free(file);
211 212
	if (unlikely(S_ISCHR(inode->i_mode) && inode->i_cdev != NULL &&
		     !(file->f_mode & FMODE_PATH))) {
Linus Torvalds's avatar
Linus Torvalds committed
213
		cdev_put(inode->i_cdev);
214
	}
Linus Torvalds's avatar
Linus Torvalds committed
215
	fops_put(file->f_op);
216
	put_pid(file->f_owner.pid);
217 218
	if ((file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_dec(inode);
219 220 221 222
	if (file->f_mode & FMODE_WRITER) {
		put_write_access(inode);
		__mnt_drop_write(mnt);
	}
223 224
	file->f_path.dentry = NULL;
	file->f_path.mnt = NULL;
225
	file->f_inode = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
226 227 228 229 230
	file_free(file);
	dput(dentry);
	mntput(mnt);
}

231
static LLIST_HEAD(delayed_fput_list);
Al Viro's avatar
Al Viro committed
232 233
static void delayed_fput(struct work_struct *unused)
{
234
	struct llist_node *node = llist_del_all(&delayed_fput_list);
235
	struct file *f, *t;
236

237 238
	llist_for_each_entry_safe(f, t, node, f_u.fu_llist)
		__fput(f);
Al Viro's avatar
Al Viro committed
239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
}

static void ____fput(struct callback_head *work)
{
	__fput(container_of(work, struct file, f_u.fu_rcuhead));
}

/*
 * If kernel thread really needs to have the final fput() it has done
 * to complete, call this.  The only user right now is the boot - we
 * *do* need to make sure our writes to binaries on initramfs has
 * not left us with opened struct file waiting for __fput() - execve()
 * won't work without that.  Please, don't add more callers without
 * very good reasons; in particular, never call that with locks
 * held and never call that from a thread that might need to do
 * some work on any kind of umount.
 */
void flush_delayed_fput(void)
{
	delayed_fput(NULL);
}

261
static DECLARE_DELAYED_WORK(delayed_fput_work, delayed_fput);
Al Viro's avatar
Al Viro committed
262

263 264
void fput(struct file *file)
{
Al Viro's avatar
Al Viro committed
265 266
	if (atomic_long_dec_and_test(&file->f_count)) {
		struct task_struct *task = current;
267 268 269 270 271

		if (likely(!in_interrupt() && !(task->flags & PF_KTHREAD))) {
			init_task_work(&file->f_u.fu_rcuhead, ____fput);
			if (!task_work_add(task, &file->f_u.fu_rcuhead, true))
				return;
272 273
			/*
			 * After this task has run exit_task_work(),
274
			 * task_work_add() will fail.  Fall through to delayed
275 276
			 * fput to avoid leaking *file.
			 */
Al Viro's avatar
Al Viro committed
277
		}
278 279

		if (llist_add(&file->f_u.fu_llist, &delayed_fput_list))
280
			schedule_delayed_work(&delayed_fput_work, 1);
Al Viro's avatar
Al Viro committed
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
	}
}

/*
 * synchronous analog of fput(); for kernel threads that might be needed
 * in some umount() (and thus can't use flush_delayed_fput() without
 * risking deadlocks), need to wait for completion of __fput() and know
 * for this specific struct file it won't involve anything that would
 * need them.  Use only if you really need it - at the very least,
 * don't blindly convert fput() by kernel thread to that.
 */
void __fput_sync(struct file *file)
{
	if (atomic_long_dec_and_test(&file->f_count)) {
		struct task_struct *task = current;
		BUG_ON(!(task->flags & PF_KTHREAD));
297
		__fput(file);
Al Viro's avatar
Al Viro committed
298
	}
299 300 301 302
}

EXPORT_SYMBOL(fput);

Linus Torvalds's avatar
Linus Torvalds committed
303 304
void put_filp(struct file *file)
{
305
	if (atomic_long_dec_and_test(&file->f_count)) {
Linus Torvalds's avatar
Linus Torvalds committed
306 307 308 309 310
		security_file_free(file);
		file_free(file);
	}
}

311
void __init files_init(void)
312
{
313
	filp_cachep = kmem_cache_create("filp", sizeof(struct file), 0,
314
			SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT, NULL);
315 316
	percpu_counter_init(&nr_files, 0, GFP_KERNEL);
}
317

318 319 320 321 322 323 324 325 326 327 328
/*
 * One file with associated inode and dcache is very roughly 1K. Per default
 * do not use more than 10% of our memory for files.
 */
void __init files_maxfiles_init(void)
{
	unsigned long n;
	unsigned long memreserve = (totalram_pages - nr_free_pages()) * 3/2;

	memreserve = min(memreserve, totalram_pages - 1);
	n = ((totalram_pages - memreserve) * (PAGE_SIZE / 1024)) / 10;
Linus Torvalds's avatar
Linus Torvalds committed
329

330
	files_stat.max_files = max_t(unsigned long, n, NR_FILE);
331
}