]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - fs/file_table.c
ARM: dts: bcm283x: Enable the VEC IP on all RaspberryPi boards
[mirror_ubuntu-zesty-kernel.git] / fs / file_table.c
CommitLineData
1da177e4
LT
1/*
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>
9f3acc31 11#include <linux/fdtable.h>
1da177e4
LT
12#include <linux/init.h>
13#include <linux/module.h>
1da177e4
LT
14#include <linux/fs.h>
15#include <linux/security.h>
16#include <linux/eventpoll.h>
ab2af1f5 17#include <linux/rcupdate.h>
1da177e4 18#include <linux/mount.h>
16f7e0fe 19#include <linux/capability.h>
1da177e4 20#include <linux/cdev.h>
0eeca283 21#include <linux/fsnotify.h>
529bf6be
DS
22#include <linux/sysctl.h>
23#include <linux/percpu_counter.h>
6416ccb7 24#include <linux/percpu.h>
4a9d4b02
AV
25#include <linux/hardirq.h>
26#include <linux/task_work.h>
0552f879 27#include <linux/ima.h>
4248b0da 28#include <linux/swap.h>
529bf6be 29
60063497 30#include <linux/atomic.h>
1da177e4 31
e81e3f4d
EP
32#include "internal.h"
33
1da177e4
LT
34/* sysctl tunables... */
35struct files_stat_struct files_stat = {
36 .max_files = NR_FILE
37};
38
b6b3fdea
ED
39/* SLAB cache for file structures */
40static struct kmem_cache *filp_cachep __read_mostly;
41
529bf6be 42static struct percpu_counter nr_files __cacheline_aligned_in_smp;
1da177e4 43
5c33b183 44static void file_free_rcu(struct rcu_head *head)
1da177e4 45{
d76b0d9b
DH
46 struct file *f = container_of(head, struct file, f_u.fu_rcuhead);
47
48 put_cred(f->f_cred);
529bf6be 49 kmem_cache_free(filp_cachep, f);
1da177e4
LT
50}
51
529bf6be 52static inline void file_free(struct file *f)
1da177e4 53{
529bf6be
DS
54 percpu_counter_dec(&nr_files);
55 call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
1da177e4
LT
56}
57
529bf6be
DS
58/*
59 * Return the total number of open files in the system
60 */
518de9b3 61static long get_nr_files(void)
1da177e4 62{
529bf6be 63 return percpu_counter_read_positive(&nr_files);
1da177e4
LT
64}
65
529bf6be
DS
66/*
67 * Return the maximum number of open files in the system
68 */
518de9b3 69unsigned long get_max_files(void)
ab2af1f5 70{
529bf6be 71 return files_stat.max_files;
ab2af1f5 72}
529bf6be
DS
73EXPORT_SYMBOL_GPL(get_max_files);
74
75/*
76 * Handle nr_files sysctl
77 */
78#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
1f7e0616 79int proc_nr_files(struct ctl_table *table, int write,
529bf6be
DS
80 void __user *buffer, size_t *lenp, loff_t *ppos)
81{
82 files_stat.nr_files = get_nr_files();
518de9b3 83 return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
529bf6be
DS
84}
85#else
1f7e0616 86int proc_nr_files(struct ctl_table *table, int write,
529bf6be
DS
87 void __user *buffer, size_t *lenp, loff_t *ppos)
88{
89 return -ENOSYS;
90}
91#endif
ab2af1f5 92
1da177e4 93/* Find an unused file structure and return a pointer to it.
1afc99be
AV
94 * Returns an error pointer if some error happend e.g. we over file
95 * structures limit, run out of memory or operation is not permitted.
430e285e
DH
96 *
97 * Be very careful using this. You are responsible for
98 * getting write access to any mount that you might assign
99 * to this filp, if it is opened for write. If this is not
100 * done, you will imbalance int the mount's writer count
101 * and a warning at __fput() time.
1da177e4
LT
102 */
103struct file *get_empty_filp(void)
104{
86a264ab 105 const struct cred *cred = current_cred();
518de9b3 106 static long old_max;
1afc99be
AV
107 struct file *f;
108 int error;
1da177e4
LT
109
110 /*
111 * Privileged users can go above max_files
112 */
529bf6be
DS
113 if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
114 /*
115 * percpu_counters are inaccurate. Do an expensive check before
116 * we go and fail.
117 */
52d9f3b4 118 if (percpu_counter_sum_positive(&nr_files) >= files_stat.max_files)
529bf6be
DS
119 goto over;
120 }
af4d2ecb 121
4975e45f 122 f = kmem_cache_zalloc(filp_cachep, GFP_KERNEL);
1afc99be
AV
123 if (unlikely(!f))
124 return ERR_PTR(-ENOMEM);
af4d2ecb 125
529bf6be 126 percpu_counter_inc(&nr_files);
78d29788 127 f->f_cred = get_cred(cred);
1afc99be
AV
128 error = security_file_alloc(f);
129 if (unlikely(error)) {
130 file_free(f);
131 return ERR_PTR(error);
132 }
1da177e4 133
516e0cc5 134 atomic_long_set(&f->f_count, 1);
af4d2ecb 135 rwlock_init(&f->f_owner.lock);
68499914 136 spin_lock_init(&f->f_lock);
9c225f26 137 mutex_init(&f->f_pos_lock);
5a6b7951 138 eventpoll_init_file(f);
af4d2ecb 139 /* f->f_version: 0 */
af4d2ecb
KK
140 return f;
141
142over:
1da177e4 143 /* Ran out of filps - report that */
529bf6be 144 if (get_nr_files() > old_max) {
518de9b3 145 pr_info("VFS: file-max limit %lu reached\n", get_max_files());
529bf6be 146 old_max = get_nr_files();
1da177e4 147 }
1afc99be 148 return ERR_PTR(-ENFILE);
1da177e4 149}
e14748e8 150EXPORT_SYMBOL_GPL(get_empty_filp);
1da177e4 151
ce8d2cdf
DH
152/**
153 * alloc_file - allocate and initialize a 'struct file'
a457606a
EB
154 *
155 * @path: the (dentry, vfsmount) pair for the new file
ce8d2cdf
DH
156 * @mode: the mode with which the new file will be opened
157 * @fop: the 'struct file_operations' for the new file
ce8d2cdf 158 */
a4141d7c 159struct file *alloc_file(const struct path *path, fmode_t mode,
2c48b9c4 160 const struct file_operations *fop)
ce8d2cdf
DH
161{
162 struct file *file;
ce8d2cdf
DH
163
164 file = get_empty_filp();
1afc99be 165 if (IS_ERR(file))
39b65252 166 return file;
ce8d2cdf 167
2c48b9c4 168 file->f_path = *path;
dd37978c 169 file->f_inode = path->dentry->d_inode;
2c48b9c4 170 file->f_mapping = path->dentry->d_inode->i_mapping;
293bc982 171 if ((mode & FMODE_READ) &&
84363182 172 likely(fop->read || fop->read_iter))
7f7f25e8 173 mode |= FMODE_CAN_READ;
293bc982 174 if ((mode & FMODE_WRITE) &&
84363182 175 likely(fop->write || fop->write_iter))
7f7f25e8 176 mode |= FMODE_CAN_WRITE;
ce8d2cdf
DH
177 file->f_mode = mode;
178 file->f_op = fop;
890275b5
MZ
179 if ((mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
180 i_readcount_inc(path->dentry->d_inode);
3d1e4631 181 return file;
ce8d2cdf 182}
73efc468 183EXPORT_SYMBOL(alloc_file);
ce8d2cdf 184
d7065da0 185/* the real guts of fput() - releasing the last reference to file
1da177e4 186 */
d7065da0 187static void __fput(struct file *file)
1da177e4 188{
0f7fc9e4
JJS
189 struct dentry *dentry = file->f_path.dentry;
190 struct vfsmount *mnt = file->f_path.mnt;
c77cecee 191 struct inode *inode = file->f_inode;
1da177e4
LT
192
193 might_sleep();
0eeca283
RL
194
195 fsnotify_close(file);
1da177e4
LT
196 /*
197 * The function eventpoll_release() should be the first called
198 * in the file cleanup chain.
199 */
200 eventpoll_release(file);
78ed8a13 201 locks_remove_file(file);
1da177e4 202
233e70f4 203 if (unlikely(file->f_flags & FASYNC)) {
72c2d531 204 if (file->f_op->fasync)
233e70f4
AV
205 file->f_op->fasync(-1, file, 0);
206 }
4199d35c 207 ima_file_free(file);
72c2d531 208 if (file->f_op->release)
1da177e4
LT
209 file->f_op->release(inode, file);
210 security_file_free(file);
60ed8cf7
MS
211 if (unlikely(S_ISCHR(inode->i_mode) && inode->i_cdev != NULL &&
212 !(file->f_mode & FMODE_PATH))) {
1da177e4 213 cdev_put(inode->i_cdev);
60ed8cf7 214 }
1da177e4 215 fops_put(file->f_op);
609d7fa9 216 put_pid(file->f_owner.pid);
890275b5
MZ
217 if ((file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
218 i_readcount_dec(inode);
83f936c7
AV
219 if (file->f_mode & FMODE_WRITER) {
220 put_write_access(inode);
221 __mnt_drop_write(mnt);
222 }
0f7fc9e4
JJS
223 file->f_path.dentry = NULL;
224 file->f_path.mnt = NULL;
dd37978c 225 file->f_inode = NULL;
1da177e4
LT
226 file_free(file);
227 dput(dentry);
228 mntput(mnt);
229}
230
4f5e65a1 231static LLIST_HEAD(delayed_fput_list);
4a9d4b02
AV
232static void delayed_fput(struct work_struct *unused)
233{
4f5e65a1
ON
234 struct llist_node *node = llist_del_all(&delayed_fput_list);
235 struct llist_node *next;
236
237 for (; node; node = next) {
238 next = llist_next(node);
239 __fput(llist_entry(node, struct file, f_u.fu_llist));
4a9d4b02
AV
240 }
241}
242
243static void ____fput(struct callback_head *work)
244{
245 __fput(container_of(work, struct file, f_u.fu_rcuhead));
246}
247
248/*
249 * If kernel thread really needs to have the final fput() it has done
250 * to complete, call this. The only user right now is the boot - we
251 * *do* need to make sure our writes to binaries on initramfs has
252 * not left us with opened struct file waiting for __fput() - execve()
253 * won't work without that. Please, don't add more callers without
254 * very good reasons; in particular, never call that with locks
255 * held and never call that from a thread that might need to do
256 * some work on any kind of umount.
257 */
258void flush_delayed_fput(void)
259{
260 delayed_fput(NULL);
261}
e14748e8 262EXPORT_SYMBOL_GPL(flush_delayed_fput);
4a9d4b02 263
c7314d74 264static DECLARE_DELAYED_WORK(delayed_fput_work, delayed_fput);
4a9d4b02 265
d7065da0
AV
266void fput(struct file *file)
267{
4a9d4b02
AV
268 if (atomic_long_dec_and_test(&file->f_count)) {
269 struct task_struct *task = current;
e7b2c406 270
e7b2c406
ON
271 if (likely(!in_interrupt() && !(task->flags & PF_KTHREAD))) {
272 init_task_work(&file->f_u.fu_rcuhead, ____fput);
273 if (!task_work_add(task, &file->f_u.fu_rcuhead, true))
274 return;
64372501
AM
275 /*
276 * After this task has run exit_task_work(),
be49b30a 277 * task_work_add() will fail. Fall through to delayed
64372501
AM
278 * fput to avoid leaking *file.
279 */
4a9d4b02 280 }
4f5e65a1
ON
281
282 if (llist_add(&file->f_u.fu_llist, &delayed_fput_list))
c7314d74 283 schedule_delayed_work(&delayed_fput_work, 1);
4a9d4b02
AV
284 }
285}
286
287/*
288 * synchronous analog of fput(); for kernel threads that might be needed
289 * in some umount() (and thus can't use flush_delayed_fput() without
290 * risking deadlocks), need to wait for completion of __fput() and know
291 * for this specific struct file it won't involve anything that would
292 * need them. Use only if you really need it - at the very least,
293 * don't blindly convert fput() by kernel thread to that.
294 */
295void __fput_sync(struct file *file)
296{
297 if (atomic_long_dec_and_test(&file->f_count)) {
298 struct task_struct *task = current;
4a9d4b02 299 BUG_ON(!(task->flags & PF_KTHREAD));
d7065da0 300 __fput(file);
4a9d4b02 301 }
d7065da0
AV
302}
303
304EXPORT_SYMBOL(fput);
e14748e8 305EXPORT_SYMBOL_GPL(__fput_sync);
d7065da0 306
1da177e4
LT
307void put_filp(struct file *file)
308{
516e0cc5 309 if (atomic_long_dec_and_test(&file->f_count)) {
1da177e4 310 security_file_free(file);
1da177e4
LT
311 file_free(file);
312 }
313}
e14748e8 314EXPORT_SYMBOL_GPL(put_filp);
1da177e4 315
4248b0da 316void __init files_init(void)
1da177e4 317{
b6b3fdea
ED
318 filp_cachep = kmem_cache_create("filp", sizeof(struct file), 0,
319 SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
4248b0da
MG
320 percpu_counter_init(&nr_files, 0, GFP_KERNEL);
321}
b6b3fdea 322
4248b0da
MG
323/*
324 * One file with associated inode and dcache is very roughly 1K. Per default
325 * do not use more than 10% of our memory for files.
326 */
327void __init files_maxfiles_init(void)
328{
329 unsigned long n;
330 unsigned long memreserve = (totalram_pages - nr_free_pages()) * 3/2;
331
332 memreserve = min(memreserve, totalram_pages - 1);
333 n = ((totalram_pages - memreserve) * (PAGE_SIZE / 1024)) / 10;
1da177e4 334
518de9b3 335 files_stat.max_files = max_t(unsigned long, n, NR_FILE);
1da177e4 336}