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/bug.h>
31 static void proc_evict_inode(struct inode *inode)
33 struct ctl_table_header *head;
34 struct proc_inode *ei = PROC_I(inode);
36 truncate_inode_pages_final(&inode->i_data);
39 /* Stop tracking associated processes */
41 proc_pid_evict_inode(ei);
45 RCU_INIT_POINTER(ei->sysctl, NULL);
46 proc_sys_evict_inode(inode, head);
50 static struct kmem_cache *proc_inode_cachep __ro_after_init;
51 static struct kmem_cache *pde_opener_cache __ro_after_init;
53 static struct inode *proc_alloc_inode(struct super_block *sb)
55 struct proc_inode *ei;
57 ei = alloc_inode_sb(sb, proc_inode_cachep, GFP_KERNEL);
62 ei->op.proc_get_link = NULL;
65 ei->sysctl_entry = NULL;
66 INIT_HLIST_NODE(&ei->sibling_inodes);
68 return &ei->vfs_inode;
71 static void proc_free_inode(struct inode *inode)
73 struct proc_inode *ei = PROC_I(inode);
77 /* Let go of any associated proc directory entry */
80 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
83 static void init_once(void *foo)
85 struct proc_inode *ei = (struct proc_inode *) foo;
87 inode_init_once(&ei->vfs_inode);
90 void __init proc_init_kmemcache(void)
92 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
93 sizeof(struct proc_inode),
94 0, (SLAB_RECLAIM_ACCOUNT|
95 SLAB_MEM_SPREAD|SLAB_ACCOUNT|
99 kmem_cache_create("pde_opener", sizeof(struct pde_opener), 0,
100 SLAB_ACCOUNT|SLAB_PANIC, NULL);
101 proc_dir_entry_cache = kmem_cache_create_usercopy(
102 "proc_dir_entry", SIZEOF_PDE, 0, SLAB_PANIC,
103 offsetof(struct proc_dir_entry, inline_name),
104 SIZEOF_PDE_INLINE_NAME, NULL);
105 BUILD_BUG_ON(sizeof(struct proc_dir_entry) >= SIZEOF_PDE);
108 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock)
110 struct hlist_node *node;
111 struct super_block *old_sb = NULL;
114 while ((node = hlist_first_rcu(inodes))) {
115 struct proc_inode *ei = hlist_entry(node, struct proc_inode, sibling_inodes);
116 struct super_block *sb;
120 hlist_del_init_rcu(&ei->sibling_inodes);
123 inode = &ei->vfs_inode;
125 if ((sb != old_sb) && !atomic_inc_not_zero(&sb->s_active))
127 inode = igrab(inode);
131 deactivate_super(old_sb);
134 if (unlikely(!inode)) {
139 if (S_ISDIR(inode->i_mode)) {
140 struct dentry *dir = d_find_any_alias(inode);
146 struct dentry *dentry;
147 while ((dentry = d_find_alias(inode))) {
148 d_invalidate(dentry);
158 deactivate_super(old_sb);
161 static inline const char *hidepid2str(enum proc_hidepid v)
164 case HIDEPID_OFF: return "off";
165 case HIDEPID_NO_ACCESS: return "noaccess";
166 case HIDEPID_INVISIBLE: return "invisible";
167 case HIDEPID_NOT_PTRACEABLE: return "ptraceable";
169 WARN_ONCE(1, "bad hide_pid value: %d\n", v);
173 static int proc_show_options(struct seq_file *seq, struct dentry *root)
175 struct proc_fs_info *fs_info = proc_sb_info(root->d_sb);
177 if (!gid_eq(fs_info->pid_gid, GLOBAL_ROOT_GID))
178 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, fs_info->pid_gid));
179 if (fs_info->hide_pid != HIDEPID_OFF)
180 seq_printf(seq, ",hidepid=%s", hidepid2str(fs_info->hide_pid));
181 if (fs_info->pidonly != PROC_PIDONLY_OFF)
182 seq_printf(seq, ",subset=pid");
187 const struct super_operations proc_sops = {
188 .alloc_inode = proc_alloc_inode,
189 .free_inode = proc_free_inode,
190 .drop_inode = generic_delete_inode,
191 .evict_inode = proc_evict_inode,
192 .statfs = simple_statfs,
193 .show_options = proc_show_options,
196 enum {BIAS = -1U<<31};
198 static inline int use_pde(struct proc_dir_entry *pde)
200 return likely(atomic_inc_unless_negative(&pde->in_use));
203 static void unuse_pde(struct proc_dir_entry *pde)
205 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
206 complete(pde->pde_unload_completion);
210 * At most 2 contexts can enter this function: the one doing the last
211 * close on the descriptor and whoever is deleting PDE itself.
213 * First to enter calls ->proc_release hook and signals its completion
214 * to the second one which waits and then does nothing.
216 * PDE is locked on entry, unlocked on exit.
218 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
219 __releases(&pde->pde_unload_lock)
222 * close() (proc_reg_release()) can't delete an entry and proceed:
223 * ->release hook needs to be available at the right moment.
225 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
226 * "struct file" needs to be available at the right moment.
229 /* somebody else is doing that, just wait */
230 DECLARE_COMPLETION_ONSTACK(c);
232 spin_unlock(&pde->pde_unload_lock);
233 wait_for_completion(&c);
236 struct completion *c;
238 pdeo->closing = true;
239 spin_unlock(&pde->pde_unload_lock);
242 pde->proc_ops->proc_release(file_inode(file), file);
244 spin_lock(&pde->pde_unload_lock);
245 /* Strictly after ->proc_release, see above. */
248 spin_unlock(&pde->pde_unload_lock);
251 kmem_cache_free(pde_opener_cache, pdeo);
255 void proc_entry_rundown(struct proc_dir_entry *de)
257 DECLARE_COMPLETION_ONSTACK(c);
258 /* Wait until all existing callers into module are done. */
259 de->pde_unload_completion = &c;
260 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
261 wait_for_completion(&c);
263 /* ->pde_openers list can't grow from now on. */
265 spin_lock(&de->pde_unload_lock);
266 while (!list_empty(&de->pde_openers)) {
267 struct pde_opener *pdeo;
268 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
269 close_pdeo(de, pdeo);
270 spin_lock(&de->pde_unload_lock);
272 spin_unlock(&de->pde_unload_lock);
275 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
277 struct proc_dir_entry *pde = PDE(file_inode(file));
280 if (pde_is_permanent(pde)) {
281 return pde->proc_ops->proc_lseek(file, offset, whence);
282 } else if (use_pde(pde)) {
283 rv = pde->proc_ops->proc_lseek(file, offset, whence);
289 static ssize_t proc_reg_read_iter(struct kiocb *iocb, struct iov_iter *iter)
291 struct proc_dir_entry *pde = PDE(file_inode(iocb->ki_filp));
294 if (pde_is_permanent(pde))
295 return pde->proc_ops->proc_read_iter(iocb, iter);
299 ret = pde->proc_ops->proc_read_iter(iocb, iter);
304 static ssize_t pde_read(struct proc_dir_entry *pde, struct file *file, char __user *buf, size_t count, loff_t *ppos)
306 typeof_member(struct proc_ops, proc_read) read;
308 read = pde->proc_ops->proc_read;
310 return read(file, buf, count, ppos);
314 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
316 struct proc_dir_entry *pde = PDE(file_inode(file));
319 if (pde_is_permanent(pde)) {
320 return pde_read(pde, file, buf, count, ppos);
321 } else if (use_pde(pde)) {
322 rv = pde_read(pde, file, buf, count, ppos);
328 static ssize_t pde_write(struct proc_dir_entry *pde, struct file *file, const char __user *buf, size_t count, loff_t *ppos)
330 typeof_member(struct proc_ops, proc_write) write;
332 write = pde->proc_ops->proc_write;
334 return write(file, buf, count, ppos);
338 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
340 struct proc_dir_entry *pde = PDE(file_inode(file));
343 if (pde_is_permanent(pde)) {
344 return pde_write(pde, file, buf, count, ppos);
345 } else if (use_pde(pde)) {
346 rv = pde_write(pde, file, buf, count, ppos);
352 static __poll_t pde_poll(struct proc_dir_entry *pde, struct file *file, struct poll_table_struct *pts)
354 typeof_member(struct proc_ops, proc_poll) poll;
356 poll = pde->proc_ops->proc_poll;
358 return poll(file, pts);
359 return DEFAULT_POLLMASK;
362 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
364 struct proc_dir_entry *pde = PDE(file_inode(file));
365 __poll_t rv = DEFAULT_POLLMASK;
367 if (pde_is_permanent(pde)) {
368 return pde_poll(pde, file, pts);
369 } else if (use_pde(pde)) {
370 rv = pde_poll(pde, file, pts);
376 static long pde_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
378 typeof_member(struct proc_ops, proc_ioctl) ioctl;
380 ioctl = pde->proc_ops->proc_ioctl;
382 return ioctl(file, cmd, arg);
386 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
388 struct proc_dir_entry *pde = PDE(file_inode(file));
391 if (pde_is_permanent(pde)) {
392 return pde_ioctl(pde, file, cmd, arg);
393 } else if (use_pde(pde)) {
394 rv = pde_ioctl(pde, file, cmd, arg);
401 static long pde_compat_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
403 typeof_member(struct proc_ops, proc_compat_ioctl) compat_ioctl;
405 compat_ioctl = pde->proc_ops->proc_compat_ioctl;
407 return compat_ioctl(file, cmd, arg);
411 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
413 struct proc_dir_entry *pde = PDE(file_inode(file));
415 if (pde_is_permanent(pde)) {
416 return pde_compat_ioctl(pde, file, cmd, arg);
417 } else if (use_pde(pde)) {
418 rv = pde_compat_ioctl(pde, file, cmd, arg);
425 static int pde_mmap(struct proc_dir_entry *pde, struct file *file, struct vm_area_struct *vma)
427 typeof_member(struct proc_ops, proc_mmap) mmap;
429 mmap = pde->proc_ops->proc_mmap;
431 return mmap(file, vma);
435 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
437 struct proc_dir_entry *pde = PDE(file_inode(file));
440 if (pde_is_permanent(pde)) {
441 return pde_mmap(pde, file, vma);
442 } else if (use_pde(pde)) {
443 rv = pde_mmap(pde, file, vma);
450 pde_get_unmapped_area(struct proc_dir_entry *pde, struct file *file, unsigned long orig_addr,
451 unsigned long len, unsigned long pgoff,
454 typeof_member(struct proc_ops, proc_get_unmapped_area) get_area;
456 get_area = pde->proc_ops->proc_get_unmapped_area;
459 get_area = current->mm->get_unmapped_area;
462 return get_area(file, orig_addr, len, pgoff, flags);
467 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
468 unsigned long len, unsigned long pgoff,
471 struct proc_dir_entry *pde = PDE(file_inode(file));
472 unsigned long rv = -EIO;
474 if (pde_is_permanent(pde)) {
475 return pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
476 } else if (use_pde(pde)) {
477 rv = pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
483 static int proc_reg_open(struct inode *inode, struct file *file)
485 struct proc_dir_entry *pde = PDE(inode);
487 typeof_member(struct proc_ops, proc_open) open;
488 typeof_member(struct proc_ops, proc_release) release;
489 struct pde_opener *pdeo;
491 if (!pde->proc_ops->proc_lseek)
492 file->f_mode &= ~FMODE_LSEEK;
494 if (pde_is_permanent(pde)) {
495 open = pde->proc_ops->proc_open;
497 rv = open(inode, file);
503 * 1) PDE's ->release hook will be called no matter what
504 * either normally by close()/->release, or forcefully by
505 * rmmod/remove_proc_entry.
507 * 2) rmmod isn't blocked by opening file in /proc and sitting on
508 * the descriptor (including "rmmod foo </proc/foo" scenario).
510 * Save every "struct file" with custom ->release hook.
515 release = pde->proc_ops->proc_release;
517 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
524 open = pde->proc_ops->proc_open;
526 rv = open(inode, file);
530 /* To know what to release. */
532 pdeo->closing = false;
534 spin_lock(&pde->pde_unload_lock);
535 list_add(&pdeo->lh, &pde->pde_openers);
536 spin_unlock(&pde->pde_unload_lock);
538 kmem_cache_free(pde_opener_cache, pdeo);
546 static int proc_reg_release(struct inode *inode, struct file *file)
548 struct proc_dir_entry *pde = PDE(inode);
549 struct pde_opener *pdeo;
551 if (pde_is_permanent(pde)) {
552 typeof_member(struct proc_ops, proc_release) release;
554 release = pde->proc_ops->proc_release;
556 return release(inode, file);
561 spin_lock(&pde->pde_unload_lock);
562 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
563 if (pdeo->file == file) {
564 close_pdeo(pde, pdeo);
568 spin_unlock(&pde->pde_unload_lock);
572 static const struct file_operations proc_reg_file_ops = {
573 .llseek = proc_reg_llseek,
574 .read = proc_reg_read,
575 .write = proc_reg_write,
576 .poll = proc_reg_poll,
577 .unlocked_ioctl = proc_reg_unlocked_ioctl,
578 .mmap = proc_reg_mmap,
579 .get_unmapped_area = proc_reg_get_unmapped_area,
580 .open = proc_reg_open,
581 .release = proc_reg_release,
584 static const struct file_operations proc_iter_file_ops = {
585 .llseek = proc_reg_llseek,
586 .read_iter = proc_reg_read_iter,
587 .write = proc_reg_write,
588 .splice_read = copy_splice_read,
589 .poll = proc_reg_poll,
590 .unlocked_ioctl = proc_reg_unlocked_ioctl,
591 .mmap = proc_reg_mmap,
592 .get_unmapped_area = proc_reg_get_unmapped_area,
593 .open = proc_reg_open,
594 .release = proc_reg_release,
598 static const struct file_operations proc_reg_file_ops_compat = {
599 .llseek = proc_reg_llseek,
600 .read = proc_reg_read,
601 .write = proc_reg_write,
602 .poll = proc_reg_poll,
603 .unlocked_ioctl = proc_reg_unlocked_ioctl,
604 .compat_ioctl = proc_reg_compat_ioctl,
605 .mmap = proc_reg_mmap,
606 .get_unmapped_area = proc_reg_get_unmapped_area,
607 .open = proc_reg_open,
608 .release = proc_reg_release,
611 static const struct file_operations proc_iter_file_ops_compat = {
612 .llseek = proc_reg_llseek,
613 .read_iter = proc_reg_read_iter,
614 .splice_read = copy_splice_read,
615 .write = proc_reg_write,
616 .poll = proc_reg_poll,
617 .unlocked_ioctl = proc_reg_unlocked_ioctl,
618 .compat_ioctl = proc_reg_compat_ioctl,
619 .mmap = proc_reg_mmap,
620 .get_unmapped_area = proc_reg_get_unmapped_area,
621 .open = proc_reg_open,
622 .release = proc_reg_release,
626 static void proc_put_link(void *p)
631 static const char *proc_get_link(struct dentry *dentry,
633 struct delayed_call *done)
635 struct proc_dir_entry *pde = PDE(inode);
637 return ERR_PTR(-EINVAL);
638 set_delayed_call(done, proc_put_link, pde);
642 const struct inode_operations proc_link_inode_operations = {
643 .get_link = proc_get_link,
646 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
648 struct inode *inode = new_inode(sb);
655 inode->i_private = de->data;
656 inode->i_ino = de->low_ino;
657 simple_inode_init_ts(inode);
658 PROC_I(inode)->pde = de;
659 if (is_empty_pde(de)) {
660 make_empty_dir_inode(inode);
665 inode->i_mode = de->mode;
666 inode->i_uid = de->uid;
667 inode->i_gid = de->gid;
670 inode->i_size = de->size;
672 set_nlink(inode, de->nlink);
674 if (S_ISREG(inode->i_mode)) {
675 inode->i_op = de->proc_iops;
676 if (de->proc_ops->proc_read_iter)
677 inode->i_fop = &proc_iter_file_ops;
679 inode->i_fop = &proc_reg_file_ops;
681 if (de->proc_ops->proc_compat_ioctl) {
682 if (de->proc_ops->proc_read_iter)
683 inode->i_fop = &proc_iter_file_ops_compat;
685 inode->i_fop = &proc_reg_file_ops_compat;
688 } else if (S_ISDIR(inode->i_mode)) {
689 inode->i_op = de->proc_iops;
690 inode->i_fop = de->proc_dir_ops;
691 } else if (S_ISLNK(inode->i_mode)) {
692 inode->i_op = de->proc_iops;