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>
27 #include <linux/magic.h>
29 #include <linux/uaccess.h>
33 static void proc_evict_inode(struct inode *inode)
35 struct proc_dir_entry *de;
36 struct ctl_table_header *head;
38 truncate_inode_pages_final(&inode->i_data);
41 /* Stop tracking associated processes */
42 put_pid(PROC_I(inode)->pid);
44 /* Let go of any associated proc directory entry */
49 head = PROC_I(inode)->sysctl;
51 RCU_INIT_POINTER(PROC_I(inode)->sysctl, NULL);
52 proc_sys_evict_inode(inode, head);
56 static struct kmem_cache *proc_inode_cachep __ro_after_init;
57 static struct kmem_cache *pde_opener_cache __ro_after_init;
59 static struct inode *proc_alloc_inode(struct super_block *sb)
61 struct proc_inode *ei;
64 ei = kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
69 ei->op.proc_get_link = NULL;
72 ei->sysctl_entry = NULL;
74 inode = &ei->vfs_inode;
78 static void proc_i_callback(struct rcu_head *head)
80 struct inode *inode = container_of(head, struct inode, i_rcu);
81 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
84 static void proc_destroy_inode(struct inode *inode)
86 call_rcu(&inode->i_rcu, proc_i_callback);
89 static void init_once(void *foo)
91 struct proc_inode *ei = (struct proc_inode *) foo;
93 inode_init_once(&ei->vfs_inode);
96 void __init proc_init_kmemcache(void)
98 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
99 sizeof(struct proc_inode),
100 0, (SLAB_RECLAIM_ACCOUNT|
101 SLAB_MEM_SPREAD|SLAB_ACCOUNT|
105 kmem_cache_create("pde_opener", sizeof(struct pde_opener), 0,
106 SLAB_ACCOUNT|SLAB_PANIC, NULL);
107 proc_dir_entry_cache = kmem_cache_create_usercopy(
108 "proc_dir_entry", sizeof(struct proc_dir_entry), 0, SLAB_PANIC,
109 offsetof(struct proc_dir_entry, inline_name),
110 sizeof_field(struct proc_dir_entry, inline_name), NULL);
113 static int proc_show_options(struct seq_file *seq, struct dentry *root)
115 struct super_block *sb = root->d_sb;
116 struct pid_namespace *pid = sb->s_fs_info;
118 if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
119 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
120 if (pid->hide_pid != HIDEPID_OFF)
121 seq_printf(seq, ",hidepid=%u", pid->hide_pid);
126 static const struct super_operations proc_sops = {
127 .alloc_inode = proc_alloc_inode,
128 .destroy_inode = proc_destroy_inode,
129 .drop_inode = generic_delete_inode,
130 .evict_inode = proc_evict_inode,
131 .statfs = simple_statfs,
132 .remount_fs = proc_remount,
133 .show_options = proc_show_options,
136 enum {BIAS = -1U<<31};
138 static inline int use_pde(struct proc_dir_entry *pde)
140 return likely(atomic_inc_unless_negative(&pde->in_use));
143 static void unuse_pde(struct proc_dir_entry *pde)
145 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
146 complete(pde->pde_unload_completion);
149 /* pde is locked on entry, unlocked on exit */
150 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
153 * close() (proc_reg_release()) can't delete an entry and proceed:
154 * ->release hook needs to be available at the right moment.
156 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
157 * "struct file" needs to be available at the right moment.
159 * Therefore, first process to enter this function does ->release() and
160 * signals its completion to the other process which does nothing.
163 /* somebody else is doing that, just wait */
164 DECLARE_COMPLETION_ONSTACK(c);
166 spin_unlock(&pde->pde_unload_lock);
167 wait_for_completion(&c);
170 struct completion *c;
172 pdeo->closing = true;
173 spin_unlock(&pde->pde_unload_lock);
175 pde->proc_fops->release(file_inode(file), file);
176 spin_lock(&pde->pde_unload_lock);
177 /* After ->release. */
180 spin_unlock(&pde->pde_unload_lock);
183 kmem_cache_free(pde_opener_cache, pdeo);
187 void proc_entry_rundown(struct proc_dir_entry *de)
189 DECLARE_COMPLETION_ONSTACK(c);
190 /* Wait until all existing callers into module are done. */
191 de->pde_unload_completion = &c;
192 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
193 wait_for_completion(&c);
195 /* ->pde_openers list can't grow from now on. */
197 spin_lock(&de->pde_unload_lock);
198 while (!list_empty(&de->pde_openers)) {
199 struct pde_opener *pdeo;
200 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
201 close_pdeo(de, pdeo);
202 spin_lock(&de->pde_unload_lock);
204 spin_unlock(&de->pde_unload_lock);
207 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
209 struct proc_dir_entry *pde = PDE(file_inode(file));
212 loff_t (*llseek)(struct file *, loff_t, int);
213 llseek = pde->proc_fops->llseek;
215 llseek = default_llseek;
216 rv = llseek(file, offset, whence);
222 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
224 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
225 struct proc_dir_entry *pde = PDE(file_inode(file));
228 read = pde->proc_fops->read;
230 rv = read(file, buf, count, ppos);
236 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
238 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
239 struct proc_dir_entry *pde = PDE(file_inode(file));
242 write = pde->proc_fops->write;
244 rv = write(file, buf, count, ppos);
250 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
252 struct proc_dir_entry *pde = PDE(file_inode(file));
253 __poll_t rv = DEFAULT_POLLMASK;
254 __poll_t (*poll)(struct file *, struct poll_table_struct *);
256 poll = pde->proc_fops->poll;
258 rv = poll(file, pts);
264 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
266 struct proc_dir_entry *pde = PDE(file_inode(file));
268 long (*ioctl)(struct file *, unsigned int, unsigned long);
270 ioctl = pde->proc_fops->unlocked_ioctl;
272 rv = ioctl(file, cmd, arg);
279 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
281 struct proc_dir_entry *pde = PDE(file_inode(file));
283 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
285 compat_ioctl = pde->proc_fops->compat_ioctl;
287 rv = compat_ioctl(file, cmd, arg);
294 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
296 struct proc_dir_entry *pde = PDE(file_inode(file));
298 int (*mmap)(struct file *, struct vm_area_struct *);
300 mmap = pde->proc_fops->mmap;
302 rv = mmap(file, vma);
309 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
310 unsigned long len, unsigned long pgoff,
313 struct proc_dir_entry *pde = PDE(file_inode(file));
314 unsigned long rv = -EIO;
317 typeof(proc_reg_get_unmapped_area) *get_area;
319 get_area = pde->proc_fops->get_unmapped_area;
322 get_area = current->mm->get_unmapped_area;
326 rv = get_area(file, orig_addr, len, pgoff, flags);
334 static int proc_reg_open(struct inode *inode, struct file *file)
336 struct proc_dir_entry *pde = PDE(inode);
338 int (*open)(struct inode *, struct file *);
339 int (*release)(struct inode *, struct file *);
340 struct pde_opener *pdeo;
344 * 1) PDE's ->release hook will be called no matter what
345 * either normally by close()/->release, or forcefully by
346 * rmmod/remove_proc_entry.
348 * 2) rmmod isn't blocked by opening file in /proc and sitting on
349 * the descriptor (including "rmmod foo </proc/foo" scenario).
351 * Save every "struct file" with custom ->release hook.
356 release = pde->proc_fops->release;
358 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
365 open = pde->proc_fops->open;
367 rv = open(inode, file);
371 /* To know what to release. */
373 pdeo->closing = false;
375 spin_lock(&pde->pde_unload_lock);
376 list_add(&pdeo->lh, &pde->pde_openers);
377 spin_unlock(&pde->pde_unload_lock);
379 kmem_cache_free(pde_opener_cache, pdeo);
387 static int proc_reg_release(struct inode *inode, struct file *file)
389 struct proc_dir_entry *pde = PDE(inode);
390 struct pde_opener *pdeo;
391 spin_lock(&pde->pde_unload_lock);
392 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
393 if (pdeo->file == file) {
394 close_pdeo(pde, pdeo);
398 spin_unlock(&pde->pde_unload_lock);
402 static const struct file_operations proc_reg_file_ops = {
403 .llseek = proc_reg_llseek,
404 .read = proc_reg_read,
405 .write = proc_reg_write,
406 .poll = proc_reg_poll,
407 .unlocked_ioctl = proc_reg_unlocked_ioctl,
409 .compat_ioctl = proc_reg_compat_ioctl,
411 .mmap = proc_reg_mmap,
412 .get_unmapped_area = proc_reg_get_unmapped_area,
413 .open = proc_reg_open,
414 .release = proc_reg_release,
418 static const struct file_operations proc_reg_file_ops_no_compat = {
419 .llseek = proc_reg_llseek,
420 .read = proc_reg_read,
421 .write = proc_reg_write,
422 .poll = proc_reg_poll,
423 .unlocked_ioctl = proc_reg_unlocked_ioctl,
424 .mmap = proc_reg_mmap,
425 .get_unmapped_area = proc_reg_get_unmapped_area,
426 .open = proc_reg_open,
427 .release = proc_reg_release,
431 static void proc_put_link(void *p)
436 static const char *proc_get_link(struct dentry *dentry,
438 struct delayed_call *done)
440 struct proc_dir_entry *pde = PDE(inode);
442 return ERR_PTR(-EINVAL);
443 set_delayed_call(done, proc_put_link, pde);
447 const struct inode_operations proc_link_inode_operations = {
448 .get_link = proc_get_link,
451 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
453 struct inode *inode = new_inode_pseudo(sb);
456 inode->i_ino = de->low_ino;
457 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
458 PROC_I(inode)->pde = de;
460 if (is_empty_pde(de)) {
461 make_empty_dir_inode(inode);
465 inode->i_mode = de->mode;
466 inode->i_uid = de->uid;
467 inode->i_gid = de->gid;
470 inode->i_size = de->size;
472 set_nlink(inode, de->nlink);
473 WARN_ON(!de->proc_iops);
474 inode->i_op = de->proc_iops;
476 if (S_ISREG(inode->i_mode)) {
478 if (!de->proc_fops->compat_ioctl)
480 &proc_reg_file_ops_no_compat;
483 inode->i_fop = &proc_reg_file_ops;
485 inode->i_fop = de->proc_fops;
493 int proc_fill_super(struct super_block *s, void *data, int silent)
495 struct pid_namespace *ns = get_pid_ns(s->s_fs_info);
496 struct inode *root_inode;
499 if (!proc_parse_options(data, ns))
502 /* User space would break if executables or devices appear on proc */
503 s->s_iflags |= SB_I_USERNS_VISIBLE | SB_I_NOEXEC | SB_I_NODEV;
504 s->s_flags |= SB_NODIRATIME | SB_NOSUID | SB_NOEXEC;
505 s->s_blocksize = 1024;
506 s->s_blocksize_bits = 10;
507 s->s_magic = PROC_SUPER_MAGIC;
508 s->s_op = &proc_sops;
512 * procfs isn't actually a stacking filesystem; however, there is
513 * too much magic going on inside it to permit stacking things on
516 s->s_stack_depth = FILESYSTEM_MAX_STACK_DEPTH;
519 root_inode = proc_get_inode(s, &proc_root);
521 pr_err("proc_fill_super: get root inode failed\n");
525 s->s_root = d_make_root(root_inode);
527 pr_err("proc_fill_super: allocate dentry failed\n");
531 ret = proc_setup_self(s);
535 return proc_setup_thread_self(s);