1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/proc/inode.c
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/cache.h>
9 #include <linux/time.h>
10 #include <linux/proc_fs.h>
11 #include <linux/kernel.h>
12 #include <linux/pid_namespace.h>
14 #include <linux/string.h>
15 #include <linux/stat.h>
16 #include <linux/completion.h>
17 #include <linux/poll.h>
18 #include <linux/printk.h>
19 #include <linux/file.h>
20 #include <linux/limits.h>
21 #include <linux/init.h>
22 #include <linux/module.h>
23 #include <linux/sysctl.h>
24 #include <linux/seq_file.h>
25 #include <linux/slab.h>
26 #include <linux/mount.h>
28 #include <linux/uaccess.h>
32 static void proc_evict_inode(struct inode *inode)
34 struct proc_dir_entry *de;
35 struct ctl_table_header *head;
37 truncate_inode_pages_final(&inode->i_data);
40 /* Stop tracking associated processes */
41 put_pid(PROC_I(inode)->pid);
43 /* Let go of any associated proc directory entry */
48 head = PROC_I(inode)->sysctl;
50 RCU_INIT_POINTER(PROC_I(inode)->sysctl, NULL);
51 proc_sys_evict_inode(inode, head);
55 static struct kmem_cache *proc_inode_cachep __ro_after_init;
56 static struct kmem_cache *pde_opener_cache __ro_after_init;
58 static struct inode *proc_alloc_inode(struct super_block *sb)
60 struct proc_inode *ei;
62 ei = kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
67 ei->op.proc_get_link = NULL;
70 ei->sysctl_entry = NULL;
72 return &ei->vfs_inode;
75 static void proc_i_callback(struct rcu_head *head)
77 struct inode *inode = container_of(head, struct inode, i_rcu);
78 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
81 static void proc_destroy_inode(struct inode *inode)
83 call_rcu(&inode->i_rcu, proc_i_callback);
86 static void init_once(void *foo)
88 struct proc_inode *ei = (struct proc_inode *) foo;
90 inode_init_once(&ei->vfs_inode);
93 void __init proc_init_kmemcache(void)
95 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
96 sizeof(struct proc_inode),
97 0, (SLAB_RECLAIM_ACCOUNT|
98 SLAB_MEM_SPREAD|SLAB_ACCOUNT|
102 kmem_cache_create("pde_opener", sizeof(struct pde_opener), 0,
103 SLAB_ACCOUNT|SLAB_PANIC, NULL);
104 proc_dir_entry_cache = kmem_cache_create_usercopy(
105 "proc_dir_entry", SIZEOF_PDE, 0, SLAB_PANIC,
106 offsetof(struct proc_dir_entry, inline_name),
107 SIZEOF_PDE_INLINE_NAME, NULL);
108 BUILD_BUG_ON(sizeof(struct proc_dir_entry) >= SIZEOF_PDE);
111 static int proc_show_options(struct seq_file *seq, struct dentry *root)
113 struct super_block *sb = root->d_sb;
114 struct pid_namespace *pid = sb->s_fs_info;
116 if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
117 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
118 if (pid->hide_pid != HIDEPID_OFF)
119 seq_printf(seq, ",hidepid=%u", pid->hide_pid);
124 const struct super_operations proc_sops = {
125 .alloc_inode = proc_alloc_inode,
126 .destroy_inode = proc_destroy_inode,
127 .drop_inode = generic_delete_inode,
128 .evict_inode = proc_evict_inode,
129 .statfs = simple_statfs,
130 .show_options = proc_show_options,
133 enum {BIAS = -1U<<31};
135 static inline int use_pde(struct proc_dir_entry *pde)
137 return likely(atomic_inc_unless_negative(&pde->in_use));
140 static void unuse_pde(struct proc_dir_entry *pde)
142 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
143 complete(pde->pde_unload_completion);
146 /* pde is locked on entry, unlocked on exit */
147 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
150 * close() (proc_reg_release()) can't delete an entry and proceed:
151 * ->release hook needs to be available at the right moment.
153 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
154 * "struct file" needs to be available at the right moment.
156 * Therefore, first process to enter this function does ->release() and
157 * signals its completion to the other process which does nothing.
160 /* somebody else is doing that, just wait */
161 DECLARE_COMPLETION_ONSTACK(c);
163 spin_unlock(&pde->pde_unload_lock);
164 wait_for_completion(&c);
167 struct completion *c;
169 pdeo->closing = true;
170 spin_unlock(&pde->pde_unload_lock);
172 pde->proc_fops->release(file_inode(file), file);
173 spin_lock(&pde->pde_unload_lock);
174 /* After ->release. */
177 spin_unlock(&pde->pde_unload_lock);
180 kmem_cache_free(pde_opener_cache, pdeo);
184 void proc_entry_rundown(struct proc_dir_entry *de)
186 DECLARE_COMPLETION_ONSTACK(c);
187 /* Wait until all existing callers into module are done. */
188 de->pde_unload_completion = &c;
189 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
190 wait_for_completion(&c);
192 /* ->pde_openers list can't grow from now on. */
194 spin_lock(&de->pde_unload_lock);
195 while (!list_empty(&de->pde_openers)) {
196 struct pde_opener *pdeo;
197 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
198 close_pdeo(de, pdeo);
199 spin_lock(&de->pde_unload_lock);
201 spin_unlock(&de->pde_unload_lock);
204 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
206 struct proc_dir_entry *pde = PDE(file_inode(file));
209 loff_t (*llseek)(struct file *, loff_t, int);
210 llseek = pde->proc_fops->llseek;
212 llseek = default_llseek;
213 rv = llseek(file, offset, whence);
219 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
221 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
222 struct proc_dir_entry *pde = PDE(file_inode(file));
225 read = pde->proc_fops->read;
227 rv = read(file, buf, count, ppos);
233 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
235 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
236 struct proc_dir_entry *pde = PDE(file_inode(file));
239 write = pde->proc_fops->write;
241 rv = write(file, buf, count, ppos);
247 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
249 struct proc_dir_entry *pde = PDE(file_inode(file));
250 __poll_t rv = DEFAULT_POLLMASK;
251 __poll_t (*poll)(struct file *, struct poll_table_struct *);
253 poll = pde->proc_fops->poll;
255 rv = poll(file, pts);
261 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
263 struct proc_dir_entry *pde = PDE(file_inode(file));
265 long (*ioctl)(struct file *, unsigned int, unsigned long);
267 ioctl = pde->proc_fops->unlocked_ioctl;
269 rv = ioctl(file, cmd, arg);
276 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
278 struct proc_dir_entry *pde = PDE(file_inode(file));
280 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
282 compat_ioctl = pde->proc_fops->compat_ioctl;
284 rv = compat_ioctl(file, cmd, arg);
291 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
293 struct proc_dir_entry *pde = PDE(file_inode(file));
295 int (*mmap)(struct file *, struct vm_area_struct *);
297 mmap = pde->proc_fops->mmap;
299 rv = mmap(file, vma);
306 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
307 unsigned long len, unsigned long pgoff,
310 struct proc_dir_entry *pde = PDE(file_inode(file));
311 unsigned long rv = -EIO;
314 typeof(proc_reg_get_unmapped_area) *get_area;
316 get_area = pde->proc_fops->get_unmapped_area;
319 get_area = current->mm->get_unmapped_area;
323 rv = get_area(file, orig_addr, len, pgoff, flags);
331 static int proc_reg_open(struct inode *inode, struct file *file)
333 struct proc_dir_entry *pde = PDE(inode);
335 int (*open)(struct inode *, struct file *);
336 int (*release)(struct inode *, struct file *);
337 struct pde_opener *pdeo;
341 * 1) PDE's ->release hook will be called no matter what
342 * either normally by close()/->release, or forcefully by
343 * rmmod/remove_proc_entry.
345 * 2) rmmod isn't blocked by opening file in /proc and sitting on
346 * the descriptor (including "rmmod foo </proc/foo" scenario).
348 * Save every "struct file" with custom ->release hook.
353 release = pde->proc_fops->release;
355 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
362 open = pde->proc_fops->open;
364 rv = open(inode, file);
368 /* To know what to release. */
370 pdeo->closing = false;
372 spin_lock(&pde->pde_unload_lock);
373 list_add(&pdeo->lh, &pde->pde_openers);
374 spin_unlock(&pde->pde_unload_lock);
376 kmem_cache_free(pde_opener_cache, pdeo);
384 static int proc_reg_release(struct inode *inode, struct file *file)
386 struct proc_dir_entry *pde = PDE(inode);
387 struct pde_opener *pdeo;
388 spin_lock(&pde->pde_unload_lock);
389 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
390 if (pdeo->file == file) {
391 close_pdeo(pde, pdeo);
395 spin_unlock(&pde->pde_unload_lock);
399 static const struct file_operations proc_reg_file_ops = {
400 .llseek = proc_reg_llseek,
401 .read = proc_reg_read,
402 .write = proc_reg_write,
403 .poll = proc_reg_poll,
404 .unlocked_ioctl = proc_reg_unlocked_ioctl,
406 .compat_ioctl = proc_reg_compat_ioctl,
408 .mmap = proc_reg_mmap,
409 .get_unmapped_area = proc_reg_get_unmapped_area,
410 .open = proc_reg_open,
411 .release = proc_reg_release,
415 static const struct file_operations proc_reg_file_ops_no_compat = {
416 .llseek = proc_reg_llseek,
417 .read = proc_reg_read,
418 .write = proc_reg_write,
419 .poll = proc_reg_poll,
420 .unlocked_ioctl = proc_reg_unlocked_ioctl,
421 .mmap = proc_reg_mmap,
422 .get_unmapped_area = proc_reg_get_unmapped_area,
423 .open = proc_reg_open,
424 .release = proc_reg_release,
428 static void proc_put_link(void *p)
433 static const char *proc_get_link(struct dentry *dentry,
435 struct delayed_call *done)
437 struct proc_dir_entry *pde = PDE(inode);
439 return ERR_PTR(-EINVAL);
440 set_delayed_call(done, proc_put_link, pde);
444 const struct inode_operations proc_link_inode_operations = {
445 .get_link = proc_get_link,
448 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
450 struct inode *inode = new_inode_pseudo(sb);
453 inode->i_ino = de->low_ino;
454 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
455 PROC_I(inode)->pde = de;
457 if (is_empty_pde(de)) {
458 make_empty_dir_inode(inode);
462 inode->i_mode = de->mode;
463 inode->i_uid = de->uid;
464 inode->i_gid = de->gid;
467 inode->i_size = de->size;
469 set_nlink(inode, de->nlink);
470 WARN_ON(!de->proc_iops);
471 inode->i_op = de->proc_iops;
473 if (S_ISREG(inode->i_mode)) {
475 if (!de->proc_fops->compat_ioctl)
477 &proc_reg_file_ops_no_compat;
480 inode->i_fop = &proc_reg_file_ops;
482 inode->i_fop = de->proc_fops;