1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/fs/file_table.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
7 */
8
9 #include <linux/string.h>
10 #include <linux/slab.h>
11 #include <linux/file.h>
12 #include <linux/fdtable.h>
13 #include <linux/init.h>
14 #include <linux/module.h>
15 #include <linux/fs.h>
16 #include <linux/filelock.h>
17 #include <linux/security.h>
18 #include <linux/cred.h>
19 #include <linux/eventpoll.h>
20 #include <linux/rcupdate.h>
21 #include <linux/mount.h>
22 #include <linux/capability.h>
23 #include <linux/cdev.h>
24 #include <linux/fsnotify.h>
25 #include <linux/sysctl.h>
26 #include <linux/percpu_counter.h>
27 #include <linux/percpu.h>
28 #include <linux/task_work.h>
29 #include <linux/ima.h>
30 #include <linux/swap.h>
31 #include <linux/kmemleak.h>
32
33 #include <linux/atomic.h>
34
35 #include "internal.h"
36
37 /* sysctl tunables... */
38 static struct files_stat_struct files_stat = {
39 .max_files = NR_FILE
40 };
41
42 /* SLAB cache for file structures */
43 static struct kmem_cache *filp_cachep __read_mostly;
44
45 static struct percpu_counter nr_files __cacheline_aligned_in_smp;
46
file_free_rcu(struct rcu_head * head)47 static void file_free_rcu(struct rcu_head *head)
48 {
49 struct file *f = container_of(head, struct file, f_rcuhead);
50
51 put_cred(f->f_cred);
52 kmem_cache_free(filp_cachep, f);
53 }
54
file_free(struct file * f)55 static inline void file_free(struct file *f)
56 {
57 security_file_free(f);
58 if (!(f->f_mode & FMODE_NOACCOUNT))
59 percpu_counter_dec(&nr_files);
60 call_rcu(&f->f_rcuhead, file_free_rcu);
61 }
62
63 /*
64 * Return the total number of open files in the system
65 */
get_nr_files(void)66 static long get_nr_files(void)
67 {
68 return percpu_counter_read_positive(&nr_files);
69 }
70
71 /*
72 * Return the maximum number of open files in the system
73 */
get_max_files(void)74 unsigned long get_max_files(void)
75 {
76 return files_stat.max_files;
77 }
78 EXPORT_SYMBOL_GPL(get_max_files);
79
80 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
81
82 /*
83 * Handle nr_files sysctl
84 */
proc_nr_files(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)85 static int proc_nr_files(struct ctl_table *table, int write, void *buffer,
86 size_t *lenp, loff_t *ppos)
87 {
88 files_stat.nr_files = get_nr_files();
89 return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
90 }
91
92 static struct ctl_table fs_stat_sysctls[] = {
93 {
94 .procname = "file-nr",
95 .data = &files_stat,
96 .maxlen = sizeof(files_stat),
97 .mode = 0444,
98 .proc_handler = proc_nr_files,
99 },
100 {
101 .procname = "file-max",
102 .data = &files_stat.max_files,
103 .maxlen = sizeof(files_stat.max_files),
104 .mode = 0644,
105 .proc_handler = proc_doulongvec_minmax,
106 .extra1 = SYSCTL_LONG_ZERO,
107 .extra2 = SYSCTL_LONG_MAX,
108 },
109 {
110 .procname = "nr_open",
111 .data = &sysctl_nr_open,
112 .maxlen = sizeof(unsigned int),
113 .mode = 0644,
114 .proc_handler = proc_dointvec_minmax,
115 .extra1 = &sysctl_nr_open_min,
116 .extra2 = &sysctl_nr_open_max,
117 },
118 { }
119 };
120
init_fs_stat_sysctls(void)121 static int __init init_fs_stat_sysctls(void)
122 {
123 register_sysctl_init("fs", fs_stat_sysctls);
124 if (IS_ENABLED(CONFIG_BINFMT_MISC)) {
125 struct ctl_table_header *hdr;
126 hdr = register_sysctl_mount_point("fs/binfmt_misc");
127 kmemleak_not_leak(hdr);
128 }
129 return 0;
130 }
131 fs_initcall(init_fs_stat_sysctls);
132 #endif
133
__alloc_file(int flags,const struct cred * cred)134 static struct file *__alloc_file(int flags, const struct cred *cred)
135 {
136 struct file *f;
137 int error;
138
139 f = kmem_cache_zalloc(filp_cachep, GFP_KERNEL);
140 if (unlikely(!f))
141 return ERR_PTR(-ENOMEM);
142
143 f->f_cred = get_cred(cred);
144 error = security_file_alloc(f);
145 if (unlikely(error)) {
146 file_free_rcu(&f->f_rcuhead);
147 return ERR_PTR(error);
148 }
149
150 atomic_long_set(&f->f_count, 1);
151 rwlock_init(&f->f_owner.lock);
152 spin_lock_init(&f->f_lock);
153 mutex_init(&f->f_pos_lock);
154 f->f_flags = flags;
155 f->f_mode = OPEN_FMODE(flags);
156 /* f->f_version: 0 */
157
158 return f;
159 }
160
161 /* Find an unused file structure and return a pointer to it.
162 * Returns an error pointer if some error happend e.g. we over file
163 * structures limit, run out of memory or operation is not permitted.
164 *
165 * Be very careful using this. You are responsible for
166 * getting write access to any mount that you might assign
167 * to this filp, if it is opened for write. If this is not
168 * done, you will imbalance int the mount's writer count
169 * and a warning at __fput() time.
170 */
alloc_empty_file(int flags,const struct cred * cred)171 struct file *alloc_empty_file(int flags, const struct cred *cred)
172 {
173 static long old_max;
174 struct file *f;
175
176 /*
177 * Privileged users can go above max_files
178 */
179 if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
180 /*
181 * percpu_counters are inaccurate. Do an expensive check before
182 * we go and fail.
183 */
184 if (percpu_counter_sum_positive(&nr_files) >= files_stat.max_files)
185 goto over;
186 }
187
188 f = __alloc_file(flags, cred);
189 if (!IS_ERR(f))
190 percpu_counter_inc(&nr_files);
191
192 return f;
193
194 over:
195 /* Ran out of filps - report that */
196 if (get_nr_files() > old_max) {
197 pr_info("VFS: file-max limit %lu reached\n", get_max_files());
198 old_max = get_nr_files();
199 }
200 return ERR_PTR(-ENFILE);
201 }
202
203 /*
204 * Variant of alloc_empty_file() that doesn't check and modify nr_files.
205 *
206 * Should not be used unless there's a very good reason to do so.
207 */
alloc_empty_file_noaccount(int flags,const struct cred * cred)208 struct file *alloc_empty_file_noaccount(int flags, const struct cred *cred)
209 {
210 struct file *f = __alloc_file(flags, cred);
211
212 if (!IS_ERR(f))
213 f->f_mode |= FMODE_NOACCOUNT;
214
215 return f;
216 }
217
218 /**
219 * alloc_file - allocate and initialize a 'struct file'
220 *
221 * @path: the (dentry, vfsmount) pair for the new file
222 * @flags: O_... flags with which the new file will be opened
223 * @fop: the 'struct file_operations' for the new file
224 */
alloc_file(const struct path * path,int flags,const struct file_operations * fop)225 static struct file *alloc_file(const struct path *path, int flags,
226 const struct file_operations *fop)
227 {
228 struct file *file;
229
230 file = alloc_empty_file(flags, current_cred());
231 if (IS_ERR(file))
232 return file;
233
234 file->f_path = *path;
235 file->f_inode = path->dentry->d_inode;
236 file->f_mapping = path->dentry->d_inode->i_mapping;
237 file->f_wb_err = filemap_sample_wb_err(file->f_mapping);
238 file->f_sb_err = file_sample_sb_err(file);
239 if (fop->llseek)
240 file->f_mode |= FMODE_LSEEK;
241 if ((file->f_mode & FMODE_READ) &&
242 likely(fop->read || fop->read_iter))
243 file->f_mode |= FMODE_CAN_READ;
244 if ((file->f_mode & FMODE_WRITE) &&
245 likely(fop->write || fop->write_iter))
246 file->f_mode |= FMODE_CAN_WRITE;
247 file->f_iocb_flags = iocb_flags(file);
248 file->f_mode |= FMODE_OPENED;
249 file->f_op = fop;
250 if ((file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
251 i_readcount_inc(path->dentry->d_inode);
252 return file;
253 }
254
alloc_file_pseudo(struct inode * inode,struct vfsmount * mnt,const char * name,int flags,const struct file_operations * fops)255 struct file *alloc_file_pseudo(struct inode *inode, struct vfsmount *mnt,
256 const char *name, int flags,
257 const struct file_operations *fops)
258 {
259 static const struct dentry_operations anon_ops = {
260 .d_dname = simple_dname
261 };
262 struct qstr this = QSTR_INIT(name, strlen(name));
263 struct path path;
264 struct file *file;
265
266 path.dentry = d_alloc_pseudo(mnt->mnt_sb, &this);
267 if (!path.dentry)
268 return ERR_PTR(-ENOMEM);
269 if (!mnt->mnt_sb->s_d_op)
270 d_set_d_op(path.dentry, &anon_ops);
271 path.mnt = mntget(mnt);
272 d_instantiate(path.dentry, inode);
273 file = alloc_file(&path, flags, fops);
274 if (IS_ERR(file)) {
275 ihold(inode);
276 path_put(&path);
277 }
278 return file;
279 }
280 EXPORT_SYMBOL(alloc_file_pseudo);
281
alloc_file_clone(struct file * base,int flags,const struct file_operations * fops)282 struct file *alloc_file_clone(struct file *base, int flags,
283 const struct file_operations *fops)
284 {
285 struct file *f = alloc_file(&base->f_path, flags, fops);
286 if (!IS_ERR(f)) {
287 path_get(&f->f_path);
288 f->f_mapping = base->f_mapping;
289 }
290 return f;
291 }
292
293 /* the real guts of fput() - releasing the last reference to file
294 */
__fput(struct file * file)295 static void __fput(struct file *file)
296 {
297 struct dentry *dentry = file->f_path.dentry;
298 struct vfsmount *mnt = file->f_path.mnt;
299 struct inode *inode = file->f_inode;
300 fmode_t mode = file->f_mode;
301
302 if (unlikely(!(file->f_mode & FMODE_OPENED)))
303 goto out;
304
305 might_sleep();
306
307 fsnotify_close(file);
308 /*
309 * The function eventpoll_release() should be the first called
310 * in the file cleanup chain.
311 */
312 eventpoll_release(file);
313 locks_remove_file(file);
314
315 ima_file_free(file);
316 if (unlikely(file->f_flags & FASYNC)) {
317 if (file->f_op->fasync)
318 file->f_op->fasync(-1, file, 0);
319 }
320 if (file->f_op->release)
321 file->f_op->release(inode, file);
322 if (unlikely(S_ISCHR(inode->i_mode) && inode->i_cdev != NULL &&
323 !(mode & FMODE_PATH))) {
324 cdev_put(inode->i_cdev);
325 }
326 fops_put(file->f_op);
327 put_pid(file->f_owner.pid);
328 put_file_access(file);
329 dput(dentry);
330 if (unlikely(mode & FMODE_NEED_UNMOUNT))
331 dissolve_on_fput(mnt);
332 mntput(mnt);
333 out:
334 file_free(file);
335 }
336
337 static LLIST_HEAD(delayed_fput_list);
delayed_fput(struct work_struct * unused)338 static void delayed_fput(struct work_struct *unused)
339 {
340 struct llist_node *node = llist_del_all(&delayed_fput_list);
341 struct file *f, *t;
342
343 llist_for_each_entry_safe(f, t, node, f_llist)
344 __fput(f);
345 }
346
____fput(struct callback_head * work)347 static void ____fput(struct callback_head *work)
348 {
349 __fput(container_of(work, struct file, f_rcuhead));
350 }
351
352 /*
353 * If kernel thread really needs to have the final fput() it has done
354 * to complete, call this. The only user right now is the boot - we
355 * *do* need to make sure our writes to binaries on initramfs has
356 * not left us with opened struct file waiting for __fput() - execve()
357 * won't work without that. Please, don't add more callers without
358 * very good reasons; in particular, never call that with locks
359 * held and never call that from a thread that might need to do
360 * some work on any kind of umount.
361 */
flush_delayed_fput(void)362 void flush_delayed_fput(void)
363 {
364 delayed_fput(NULL);
365 }
366 EXPORT_SYMBOL_GPL(flush_delayed_fput);
367
368 static DECLARE_DELAYED_WORK(delayed_fput_work, delayed_fput);
369
fput(struct file * file)370 void fput(struct file *file)
371 {
372 if (atomic_long_dec_and_test(&file->f_count)) {
373 struct task_struct *task = current;
374
375 if (likely(!in_interrupt() && !(task->flags & PF_KTHREAD))) {
376 init_task_work(&file->f_rcuhead, ____fput);
377 if (!task_work_add(task, &file->f_rcuhead, TWA_RESUME))
378 return;
379 /*
380 * After this task has run exit_task_work(),
381 * task_work_add() will fail. Fall through to delayed
382 * fput to avoid leaking *file.
383 */
384 }
385
386 if (llist_add(&file->f_llist, &delayed_fput_list))
387 schedule_delayed_work(&delayed_fput_work, 1);
388 }
389 }
390
391 /*
392 * synchronous analog of fput(); for kernel threads that might be needed
393 * in some umount() (and thus can't use flush_delayed_fput() without
394 * risking deadlocks), need to wait for completion of __fput() and know
395 * for this specific struct file it won't involve anything that would
396 * need them. Use only if you really need it - at the very least,
397 * don't blindly convert fput() by kernel thread to that.
398 */
__fput_sync(struct file * file)399 void __fput_sync(struct file *file)
400 {
401 if (atomic_long_dec_and_test(&file->f_count)) {
402 struct task_struct *task = current;
403 BUG_ON(!(task->flags & PF_KTHREAD));
404 __fput(file);
405 }
406 }
407
408 EXPORT_SYMBOL(fput);
409 EXPORT_SYMBOL(__fput_sync);
410
files_init(void)411 void __init files_init(void)
412 {
413 filp_cachep = kmem_cache_create("filp", sizeof(struct file), 0,
414 SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT, NULL);
415 percpu_counter_init(&nr_files, 0, GFP_KERNEL);
416 }
417
418 /*
419 * One file with associated inode and dcache is very roughly 1K. Per default
420 * do not use more than 10% of our memory for files.
421 */
files_maxfiles_init(void)422 void __init files_maxfiles_init(void)
423 {
424 unsigned long n;
425 unsigned long nr_pages = totalram_pages();
426 unsigned long memreserve = (nr_pages - nr_free_pages()) * 3/2;
427
428 memreserve = min(memreserve, nr_pages - 1);
429 n = ((nr_pages - memreserve) * (PAGE_SIZE / 1024)) / 10;
430
431 files_stat.max_files = max_t(unsigned long, n, NR_FILE);
432 }
433