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;
36 struct proc_inode *ei = PROC_I(inode);
38 truncate_inode_pages_final(&inode->i_data);
41 /* Stop tracking associated processes */
43 proc_pid_evict_inode(ei);
47 /* Let go of any associated proc directory entry */
56 RCU_INIT_POINTER(ei->sysctl, NULL);
57 proc_sys_evict_inode(inode, head);
61 static struct kmem_cache *proc_inode_cachep __ro_after_init;
62 static struct kmem_cache *pde_opener_cache __ro_after_init;
64 static struct inode *proc_alloc_inode(struct super_block *sb)
66 struct proc_inode *ei;
68 ei = kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
73 ei->op.proc_get_link = NULL;
76 ei->sysctl_entry = NULL;
77 INIT_HLIST_NODE(&ei->sibling_inodes);
79 return &ei->vfs_inode;
82 static void proc_free_inode(struct inode *inode)
84 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
87 static void init_once(void *foo)
89 struct proc_inode *ei = (struct proc_inode *) foo;
91 inode_init_once(&ei->vfs_inode);
94 void __init proc_init_kmemcache(void)
96 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
97 sizeof(struct proc_inode),
98 0, (SLAB_RECLAIM_ACCOUNT|
99 SLAB_MEM_SPREAD|SLAB_ACCOUNT|
103 kmem_cache_create("pde_opener", sizeof(struct pde_opener), 0,
104 SLAB_ACCOUNT|SLAB_PANIC, NULL);
105 proc_dir_entry_cache = kmem_cache_create_usercopy(
106 "proc_dir_entry", SIZEOF_PDE, 0, SLAB_PANIC,
107 offsetof(struct proc_dir_entry, inline_name),
108 SIZEOF_PDE_INLINE_NAME, NULL);
109 BUILD_BUG_ON(sizeof(struct proc_dir_entry) >= SIZEOF_PDE);
112 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock)
115 struct proc_inode *ei;
116 struct hlist_node *node;
117 struct super_block *old_sb = NULL;
121 struct super_block *sb;
122 node = hlist_first_rcu(inodes);
125 ei = hlist_entry(node, struct proc_inode, sibling_inodes);
127 hlist_del_init_rcu(&ei->sibling_inodes);
130 inode = &ei->vfs_inode;
132 if ((sb != old_sb) && !atomic_inc_not_zero(&sb->s_active))
134 inode = igrab(inode);
138 deactivate_super(old_sb);
141 if (unlikely(!inode)) {
146 if (S_ISDIR(inode->i_mode)) {
147 struct dentry *dir = d_find_any_alias(inode);
153 struct dentry *dentry;
154 while ((dentry = d_find_alias(inode))) {
155 d_invalidate(dentry);
165 deactivate_super(old_sb);
168 static int proc_show_options(struct seq_file *seq, struct dentry *root)
170 struct proc_fs_info *fs_info = proc_sb_info(root->d_sb);
172 if (!gid_eq(fs_info->pid_gid, GLOBAL_ROOT_GID))
173 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, fs_info->pid_gid));
174 if (fs_info->hide_pid != HIDEPID_OFF)
175 seq_printf(seq, ",hidepid=%u", fs_info->hide_pid);
180 const struct super_operations proc_sops = {
181 .alloc_inode = proc_alloc_inode,
182 .free_inode = proc_free_inode,
183 .drop_inode = generic_delete_inode,
184 .evict_inode = proc_evict_inode,
185 .statfs = simple_statfs,
186 .show_options = proc_show_options,
189 enum {BIAS = -1U<<31};
191 static inline int use_pde(struct proc_dir_entry *pde)
193 return likely(atomic_inc_unless_negative(&pde->in_use));
196 static void unuse_pde(struct proc_dir_entry *pde)
198 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
199 complete(pde->pde_unload_completion);
202 /* pde is locked on entry, unlocked on exit */
203 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
204 __releases(&pde->pde_unload_lock)
207 * close() (proc_reg_release()) can't delete an entry and proceed:
208 * ->release hook needs to be available at the right moment.
210 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
211 * "struct file" needs to be available at the right moment.
213 * Therefore, first process to enter this function does ->release() and
214 * signals its completion to the other process which does nothing.
217 /* somebody else is doing that, just wait */
218 DECLARE_COMPLETION_ONSTACK(c);
220 spin_unlock(&pde->pde_unload_lock);
221 wait_for_completion(&c);
224 struct completion *c;
226 pdeo->closing = true;
227 spin_unlock(&pde->pde_unload_lock);
229 pde->proc_ops->proc_release(file_inode(file), file);
230 spin_lock(&pde->pde_unload_lock);
231 /* After ->release. */
234 spin_unlock(&pde->pde_unload_lock);
237 kmem_cache_free(pde_opener_cache, pdeo);
241 void proc_entry_rundown(struct proc_dir_entry *de)
243 DECLARE_COMPLETION_ONSTACK(c);
244 /* Wait until all existing callers into module are done. */
245 de->pde_unload_completion = &c;
246 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
247 wait_for_completion(&c);
249 /* ->pde_openers list can't grow from now on. */
251 spin_lock(&de->pde_unload_lock);
252 while (!list_empty(&de->pde_openers)) {
253 struct pde_opener *pdeo;
254 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
255 close_pdeo(de, pdeo);
256 spin_lock(&de->pde_unload_lock);
258 spin_unlock(&de->pde_unload_lock);
261 static loff_t pde_lseek(struct proc_dir_entry *pde, struct file *file, loff_t offset, int whence)
263 typeof_member(struct proc_ops, proc_lseek) lseek;
265 lseek = pde->proc_ops->proc_lseek;
267 lseek = default_llseek;
268 return lseek(file, offset, whence);
271 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
273 struct proc_dir_entry *pde = PDE(file_inode(file));
276 if (pde_is_permanent(pde)) {
277 return pde_lseek(pde, file, offset, whence);
278 } else if (use_pde(pde)) {
279 rv = pde_lseek(pde, file, offset, whence);
285 static ssize_t pde_read(struct proc_dir_entry *pde, struct file *file, char __user *buf, size_t count, loff_t *ppos)
287 typeof_member(struct proc_ops, proc_read) read;
289 read = pde->proc_ops->proc_read;
291 return read(file, buf, count, ppos);
295 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
297 struct proc_dir_entry *pde = PDE(file_inode(file));
300 if (pde_is_permanent(pde)) {
301 return pde_read(pde, file, buf, count, ppos);
302 } else if (use_pde(pde)) {
303 rv = pde_read(pde, file, buf, count, ppos);
309 static ssize_t pde_write(struct proc_dir_entry *pde, struct file *file, const char __user *buf, size_t count, loff_t *ppos)
311 typeof_member(struct proc_ops, proc_write) write;
313 write = pde->proc_ops->proc_write;
315 return write(file, buf, count, ppos);
319 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
321 struct proc_dir_entry *pde = PDE(file_inode(file));
324 if (pde_is_permanent(pde)) {
325 return pde_write(pde, file, buf, count, ppos);
326 } else if (use_pde(pde)) {
327 rv = pde_write(pde, file, buf, count, ppos);
333 static __poll_t pde_poll(struct proc_dir_entry *pde, struct file *file, struct poll_table_struct *pts)
335 typeof_member(struct proc_ops, proc_poll) poll;
337 poll = pde->proc_ops->proc_poll;
339 return poll(file, pts);
340 return DEFAULT_POLLMASK;
343 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
345 struct proc_dir_entry *pde = PDE(file_inode(file));
346 __poll_t rv = DEFAULT_POLLMASK;
348 if (pde_is_permanent(pde)) {
349 return pde_poll(pde, file, pts);
350 } else if (use_pde(pde)) {
351 rv = pde_poll(pde, file, pts);
357 static long pde_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
359 typeof_member(struct proc_ops, proc_ioctl) ioctl;
361 ioctl = pde->proc_ops->proc_ioctl;
363 return ioctl(file, cmd, arg);
367 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
369 struct proc_dir_entry *pde = PDE(file_inode(file));
372 if (pde_is_permanent(pde)) {
373 return pde_ioctl(pde, file, cmd, arg);
374 } else if (use_pde(pde)) {
375 rv = pde_ioctl(pde, file, cmd, arg);
382 static long pde_compat_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
384 typeof_member(struct proc_ops, proc_compat_ioctl) compat_ioctl;
386 compat_ioctl = pde->proc_ops->proc_compat_ioctl;
388 return compat_ioctl(file, cmd, arg);
392 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
394 struct proc_dir_entry *pde = PDE(file_inode(file));
396 if (pde_is_permanent(pde)) {
397 return pde_compat_ioctl(pde, file, cmd, arg);
398 } else if (use_pde(pde)) {
399 rv = pde_compat_ioctl(pde, file, cmd, arg);
406 static int pde_mmap(struct proc_dir_entry *pde, struct file *file, struct vm_area_struct *vma)
408 typeof_member(struct proc_ops, proc_mmap) mmap;
410 mmap = pde->proc_ops->proc_mmap;
412 return mmap(file, vma);
416 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
418 struct proc_dir_entry *pde = PDE(file_inode(file));
421 if (pde_is_permanent(pde)) {
422 return pde_mmap(pde, file, vma);
423 } else if (use_pde(pde)) {
424 rv = pde_mmap(pde, file, vma);
431 pde_get_unmapped_area(struct proc_dir_entry *pde, struct file *file, unsigned long orig_addr,
432 unsigned long len, unsigned long pgoff,
435 typeof_member(struct proc_ops, proc_get_unmapped_area) get_area;
437 get_area = pde->proc_ops->proc_get_unmapped_area;
440 get_area = current->mm->get_unmapped_area;
443 return get_area(file, orig_addr, len, pgoff, flags);
448 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
449 unsigned long len, unsigned long pgoff,
452 struct proc_dir_entry *pde = PDE(file_inode(file));
453 unsigned long rv = -EIO;
455 if (pde_is_permanent(pde)) {
456 return pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
457 } else if (use_pde(pde)) {
458 rv = pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
464 static int proc_reg_open(struct inode *inode, struct file *file)
466 struct proc_dir_entry *pde = PDE(inode);
468 typeof_member(struct proc_ops, proc_open) open;
469 typeof_member(struct proc_ops, proc_release) release;
470 struct pde_opener *pdeo;
472 if (pde_is_permanent(pde)) {
473 open = pde->proc_ops->proc_open;
475 rv = open(inode, file);
481 * 1) PDE's ->release hook will be called no matter what
482 * either normally by close()/->release, or forcefully by
483 * rmmod/remove_proc_entry.
485 * 2) rmmod isn't blocked by opening file in /proc and sitting on
486 * the descriptor (including "rmmod foo </proc/foo" scenario).
488 * Save every "struct file" with custom ->release hook.
493 release = pde->proc_ops->proc_release;
495 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
502 open = pde->proc_ops->proc_open;
504 rv = open(inode, file);
508 /* To know what to release. */
510 pdeo->closing = false;
512 spin_lock(&pde->pde_unload_lock);
513 list_add(&pdeo->lh, &pde->pde_openers);
514 spin_unlock(&pde->pde_unload_lock);
516 kmem_cache_free(pde_opener_cache, pdeo);
524 static int proc_reg_release(struct inode *inode, struct file *file)
526 struct proc_dir_entry *pde = PDE(inode);
527 struct pde_opener *pdeo;
529 if (pde_is_permanent(pde)) {
530 typeof_member(struct proc_ops, proc_release) release;
532 release = pde->proc_ops->proc_release;
534 return release(inode, file);
539 spin_lock(&pde->pde_unload_lock);
540 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
541 if (pdeo->file == file) {
542 close_pdeo(pde, pdeo);
546 spin_unlock(&pde->pde_unload_lock);
550 static const struct file_operations proc_reg_file_ops = {
551 .llseek = proc_reg_llseek,
552 .read = proc_reg_read,
553 .write = proc_reg_write,
554 .poll = proc_reg_poll,
555 .unlocked_ioctl = proc_reg_unlocked_ioctl,
557 .compat_ioctl = proc_reg_compat_ioctl,
559 .mmap = proc_reg_mmap,
560 .get_unmapped_area = proc_reg_get_unmapped_area,
561 .open = proc_reg_open,
562 .release = proc_reg_release,
566 static const struct file_operations proc_reg_file_ops_no_compat = {
567 .llseek = proc_reg_llseek,
568 .read = proc_reg_read,
569 .write = proc_reg_write,
570 .poll = proc_reg_poll,
571 .unlocked_ioctl = proc_reg_unlocked_ioctl,
572 .mmap = proc_reg_mmap,
573 .get_unmapped_area = proc_reg_get_unmapped_area,
574 .open = proc_reg_open,
575 .release = proc_reg_release,
579 static void proc_put_link(void *p)
584 static const char *proc_get_link(struct dentry *dentry,
586 struct delayed_call *done)
588 struct proc_dir_entry *pde = PDE(inode);
590 return ERR_PTR(-EINVAL);
591 set_delayed_call(done, proc_put_link, pde);
595 const struct inode_operations proc_link_inode_operations = {
596 .get_link = proc_get_link,
599 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
601 struct inode *inode = new_inode_pseudo(sb);
604 inode->i_ino = de->low_ino;
605 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
606 PROC_I(inode)->pde = de;
608 if (is_empty_pde(de)) {
609 make_empty_dir_inode(inode);
613 inode->i_mode = de->mode;
614 inode->i_uid = de->uid;
615 inode->i_gid = de->gid;
618 inode->i_size = de->size;
620 set_nlink(inode, de->nlink);
622 if (S_ISREG(inode->i_mode)) {
623 inode->i_op = de->proc_iops;
624 inode->i_fop = &proc_reg_file_ops;
626 if (!de->proc_ops->proc_compat_ioctl) {
627 inode->i_fop = &proc_reg_file_ops_no_compat;
630 } else if (S_ISDIR(inode->i_mode)) {
631 inode->i_op = de->proc_iops;
632 inode->i_fop = de->proc_dir_ops;
633 } else if (S_ISLNK(inode->i_mode)) {
634 inode->i_op = de->proc_iops;