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 proc_dir_entry *de;
34 struct ctl_table_header *head;
35 struct proc_inode *ei = PROC_I(inode);
37 truncate_inode_pages_final(&inode->i_data);
40 /* Stop tracking associated processes */
42 proc_pid_evict_inode(ei);
46 /* Let go of any associated proc directory entry */
55 RCU_INIT_POINTER(ei->sysctl, NULL);
56 proc_sys_evict_inode(inode, head);
60 static struct kmem_cache *proc_inode_cachep __ro_after_init;
61 static struct kmem_cache *pde_opener_cache __ro_after_init;
63 static struct inode *proc_alloc_inode(struct super_block *sb)
65 struct proc_inode *ei;
67 ei = alloc_inode_sb(sb, proc_inode_cachep, GFP_KERNEL);
72 ei->op.proc_get_link = NULL;
75 ei->sysctl_entry = NULL;
76 INIT_HLIST_NODE(&ei->sibling_inodes);
78 return &ei->vfs_inode;
81 static void proc_free_inode(struct inode *inode)
83 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
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 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock)
113 struct hlist_node *node;
114 struct super_block *old_sb = NULL;
117 while ((node = hlist_first_rcu(inodes))) {
118 struct proc_inode *ei = hlist_entry(node, struct proc_inode, sibling_inodes);
119 struct super_block *sb;
123 hlist_del_init_rcu(&ei->sibling_inodes);
126 inode = &ei->vfs_inode;
128 if ((sb != old_sb) && !atomic_inc_not_zero(&sb->s_active))
130 inode = igrab(inode);
134 deactivate_super(old_sb);
137 if (unlikely(!inode)) {
142 if (S_ISDIR(inode->i_mode)) {
143 struct dentry *dir = d_find_any_alias(inode);
149 struct dentry *dentry;
150 while ((dentry = d_find_alias(inode))) {
151 d_invalidate(dentry);
161 deactivate_super(old_sb);
164 static inline const char *hidepid2str(enum proc_hidepid v)
167 case HIDEPID_OFF: return "off";
168 case HIDEPID_NO_ACCESS: return "noaccess";
169 case HIDEPID_INVISIBLE: return "invisible";
170 case HIDEPID_NOT_PTRACEABLE: return "ptraceable";
172 WARN_ONCE(1, "bad hide_pid value: %d\n", v);
176 static int proc_show_options(struct seq_file *seq, struct dentry *root)
178 struct proc_fs_info *fs_info = proc_sb_info(root->d_sb);
180 if (!gid_eq(fs_info->pid_gid, GLOBAL_ROOT_GID))
181 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, fs_info->pid_gid));
182 if (fs_info->hide_pid != HIDEPID_OFF)
183 seq_printf(seq, ",hidepid=%s", hidepid2str(fs_info->hide_pid));
184 if (fs_info->pidonly != PROC_PIDONLY_OFF)
185 seq_printf(seq, ",subset=pid");
190 const struct super_operations proc_sops = {
191 .alloc_inode = proc_alloc_inode,
192 .free_inode = proc_free_inode,
193 .drop_inode = generic_delete_inode,
194 .evict_inode = proc_evict_inode,
195 .statfs = simple_statfs,
196 .show_options = proc_show_options,
199 enum {BIAS = -1U<<31};
201 static inline int use_pde(struct proc_dir_entry *pde)
203 return likely(atomic_inc_unless_negative(&pde->in_use));
206 static void unuse_pde(struct proc_dir_entry *pde)
208 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
209 complete(pde->pde_unload_completion);
213 * At most 2 contexts can enter this function: the one doing the last
214 * close on the descriptor and whoever is deleting PDE itself.
216 * First to enter calls ->proc_release hook and signals its completion
217 * to the second one which waits and then does nothing.
219 * PDE is locked on entry, unlocked on exit.
221 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
222 __releases(&pde->pde_unload_lock)
225 * close() (proc_reg_release()) can't delete an entry and proceed:
226 * ->release hook needs to be available at the right moment.
228 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
229 * "struct file" needs to be available at the right moment.
232 /* somebody else is doing that, just wait */
233 DECLARE_COMPLETION_ONSTACK(c);
235 spin_unlock(&pde->pde_unload_lock);
236 wait_for_completion(&c);
239 struct completion *c;
241 pdeo->closing = true;
242 spin_unlock(&pde->pde_unload_lock);
245 pde->proc_ops->proc_release(file_inode(file), file);
247 spin_lock(&pde->pde_unload_lock);
248 /* Strictly after ->proc_release, see above. */
251 spin_unlock(&pde->pde_unload_lock);
254 kmem_cache_free(pde_opener_cache, pdeo);
258 void proc_entry_rundown(struct proc_dir_entry *de)
260 DECLARE_COMPLETION_ONSTACK(c);
261 /* Wait until all existing callers into module are done. */
262 de->pde_unload_completion = &c;
263 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
264 wait_for_completion(&c);
266 /* ->pde_openers list can't grow from now on. */
268 spin_lock(&de->pde_unload_lock);
269 while (!list_empty(&de->pde_openers)) {
270 struct pde_opener *pdeo;
271 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
272 close_pdeo(de, pdeo);
273 spin_lock(&de->pde_unload_lock);
275 spin_unlock(&de->pde_unload_lock);
278 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
280 struct proc_dir_entry *pde = PDE(file_inode(file));
283 if (pde_is_permanent(pde)) {
284 return pde->proc_ops->proc_lseek(file, offset, whence);
285 } else if (use_pde(pde)) {
286 rv = pde->proc_ops->proc_lseek(file, offset, whence);
292 static ssize_t proc_reg_read_iter(struct kiocb *iocb, struct iov_iter *iter)
294 struct proc_dir_entry *pde = PDE(file_inode(iocb->ki_filp));
297 if (pde_is_permanent(pde))
298 return pde->proc_ops->proc_read_iter(iocb, iter);
302 ret = pde->proc_ops->proc_read_iter(iocb, iter);
307 static ssize_t pde_read(struct proc_dir_entry *pde, struct file *file, char __user *buf, size_t count, loff_t *ppos)
309 typeof_member(struct proc_ops, proc_read) read;
311 read = pde->proc_ops->proc_read;
313 return read(file, buf, count, ppos);
317 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
319 struct proc_dir_entry *pde = PDE(file_inode(file));
322 if (pde_is_permanent(pde)) {
323 return pde_read(pde, file, buf, count, ppos);
324 } else if (use_pde(pde)) {
325 rv = pde_read(pde, file, buf, count, ppos);
331 static ssize_t pde_write(struct proc_dir_entry *pde, struct file *file, const char __user *buf, size_t count, loff_t *ppos)
333 typeof_member(struct proc_ops, proc_write) write;
335 write = pde->proc_ops->proc_write;
337 return write(file, buf, count, ppos);
341 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
343 struct proc_dir_entry *pde = PDE(file_inode(file));
346 if (pde_is_permanent(pde)) {
347 return pde_write(pde, file, buf, count, ppos);
348 } else if (use_pde(pde)) {
349 rv = pde_write(pde, file, buf, count, ppos);
355 static __poll_t pde_poll(struct proc_dir_entry *pde, struct file *file, struct poll_table_struct *pts)
357 typeof_member(struct proc_ops, proc_poll) poll;
359 poll = pde->proc_ops->proc_poll;
361 return poll(file, pts);
362 return DEFAULT_POLLMASK;
365 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
367 struct proc_dir_entry *pde = PDE(file_inode(file));
368 __poll_t rv = DEFAULT_POLLMASK;
370 if (pde_is_permanent(pde)) {
371 return pde_poll(pde, file, pts);
372 } else if (use_pde(pde)) {
373 rv = pde_poll(pde, file, pts);
379 static long pde_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
381 typeof_member(struct proc_ops, proc_ioctl) ioctl;
383 ioctl = pde->proc_ops->proc_ioctl;
385 return ioctl(file, cmd, arg);
389 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
391 struct proc_dir_entry *pde = PDE(file_inode(file));
394 if (pde_is_permanent(pde)) {
395 return pde_ioctl(pde, file, cmd, arg);
396 } else if (use_pde(pde)) {
397 rv = pde_ioctl(pde, file, cmd, arg);
404 static long pde_compat_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
406 typeof_member(struct proc_ops, proc_compat_ioctl) compat_ioctl;
408 compat_ioctl = pde->proc_ops->proc_compat_ioctl;
410 return compat_ioctl(file, cmd, arg);
414 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
416 struct proc_dir_entry *pde = PDE(file_inode(file));
418 if (pde_is_permanent(pde)) {
419 return pde_compat_ioctl(pde, file, cmd, arg);
420 } else if (use_pde(pde)) {
421 rv = pde_compat_ioctl(pde, file, cmd, arg);
428 static int pde_mmap(struct proc_dir_entry *pde, struct file *file, struct vm_area_struct *vma)
430 typeof_member(struct proc_ops, proc_mmap) mmap;
432 mmap = pde->proc_ops->proc_mmap;
434 return mmap(file, vma);
438 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
440 struct proc_dir_entry *pde = PDE(file_inode(file));
443 if (pde_is_permanent(pde)) {
444 return pde_mmap(pde, file, vma);
445 } else if (use_pde(pde)) {
446 rv = pde_mmap(pde, file, vma);
453 pde_get_unmapped_area(struct proc_dir_entry *pde, struct file *file, unsigned long orig_addr,
454 unsigned long len, unsigned long pgoff,
457 typeof_member(struct proc_ops, proc_get_unmapped_area) get_area;
459 get_area = pde->proc_ops->proc_get_unmapped_area;
462 get_area = current->mm->get_unmapped_area;
465 return get_area(file, orig_addr, len, pgoff, flags);
470 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
471 unsigned long len, unsigned long pgoff,
474 struct proc_dir_entry *pde = PDE(file_inode(file));
475 unsigned long rv = -EIO;
477 if (pde_is_permanent(pde)) {
478 return pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
479 } else if (use_pde(pde)) {
480 rv = pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
486 static int proc_reg_open(struct inode *inode, struct file *file)
488 struct proc_dir_entry *pde = PDE(inode);
490 typeof_member(struct proc_ops, proc_open) open;
491 typeof_member(struct proc_ops, proc_release) release;
492 struct pde_opener *pdeo;
494 if (!pde->proc_ops->proc_lseek)
495 file->f_mode &= ~FMODE_LSEEK;
497 if (pde_is_permanent(pde)) {
498 open = pde->proc_ops->proc_open;
500 rv = open(inode, file);
506 * 1) PDE's ->release hook will be called no matter what
507 * either normally by close()/->release, or forcefully by
508 * rmmod/remove_proc_entry.
510 * 2) rmmod isn't blocked by opening file in /proc and sitting on
511 * the descriptor (including "rmmod foo </proc/foo" scenario).
513 * Save every "struct file" with custom ->release hook.
518 release = pde->proc_ops->proc_release;
520 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
527 open = pde->proc_ops->proc_open;
529 rv = open(inode, file);
533 /* To know what to release. */
535 pdeo->closing = false;
537 spin_lock(&pde->pde_unload_lock);
538 list_add(&pdeo->lh, &pde->pde_openers);
539 spin_unlock(&pde->pde_unload_lock);
541 kmem_cache_free(pde_opener_cache, pdeo);
549 static int proc_reg_release(struct inode *inode, struct file *file)
551 struct proc_dir_entry *pde = PDE(inode);
552 struct pde_opener *pdeo;
554 if (pde_is_permanent(pde)) {
555 typeof_member(struct proc_ops, proc_release) release;
557 release = pde->proc_ops->proc_release;
559 return release(inode, file);
564 spin_lock(&pde->pde_unload_lock);
565 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
566 if (pdeo->file == file) {
567 close_pdeo(pde, pdeo);
571 spin_unlock(&pde->pde_unload_lock);
575 static const struct file_operations proc_reg_file_ops = {
576 .llseek = proc_reg_llseek,
577 .read = proc_reg_read,
578 .write = proc_reg_write,
579 .poll = proc_reg_poll,
580 .unlocked_ioctl = proc_reg_unlocked_ioctl,
581 .mmap = proc_reg_mmap,
582 .get_unmapped_area = proc_reg_get_unmapped_area,
583 .open = proc_reg_open,
584 .release = proc_reg_release,
587 static const struct file_operations proc_iter_file_ops = {
588 .llseek = proc_reg_llseek,
589 .read_iter = proc_reg_read_iter,
590 .write = proc_reg_write,
591 .splice_read = copy_splice_read,
592 .poll = proc_reg_poll,
593 .unlocked_ioctl = proc_reg_unlocked_ioctl,
594 .mmap = proc_reg_mmap,
595 .get_unmapped_area = proc_reg_get_unmapped_area,
596 .open = proc_reg_open,
597 .release = proc_reg_release,
601 static const struct file_operations proc_reg_file_ops_compat = {
602 .llseek = proc_reg_llseek,
603 .read = proc_reg_read,
604 .write = proc_reg_write,
605 .poll = proc_reg_poll,
606 .unlocked_ioctl = proc_reg_unlocked_ioctl,
607 .compat_ioctl = proc_reg_compat_ioctl,
608 .mmap = proc_reg_mmap,
609 .get_unmapped_area = proc_reg_get_unmapped_area,
610 .open = proc_reg_open,
611 .release = proc_reg_release,
614 static const struct file_operations proc_iter_file_ops_compat = {
615 .llseek = proc_reg_llseek,
616 .read_iter = proc_reg_read_iter,
617 .splice_read = copy_splice_read,
618 .write = proc_reg_write,
619 .poll = proc_reg_poll,
620 .unlocked_ioctl = proc_reg_unlocked_ioctl,
621 .compat_ioctl = proc_reg_compat_ioctl,
622 .mmap = proc_reg_mmap,
623 .get_unmapped_area = proc_reg_get_unmapped_area,
624 .open = proc_reg_open,
625 .release = proc_reg_release,
629 static void proc_put_link(void *p)
634 static const char *proc_get_link(struct dentry *dentry,
636 struct delayed_call *done)
638 struct proc_dir_entry *pde = PDE(inode);
640 return ERR_PTR(-EINVAL);
641 set_delayed_call(done, proc_put_link, pde);
645 const struct inode_operations proc_link_inode_operations = {
646 .get_link = proc_get_link,
649 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
651 struct inode *inode = new_inode(sb);
658 inode->i_private = de->data;
659 inode->i_ino = de->low_ino;
660 simple_inode_init_ts(inode);
661 PROC_I(inode)->pde = de;
662 if (is_empty_pde(de)) {
663 make_empty_dir_inode(inode);
668 inode->i_mode = de->mode;
669 inode->i_uid = de->uid;
670 inode->i_gid = de->gid;
673 inode->i_size = de->size;
675 set_nlink(inode, de->nlink);
677 if (S_ISREG(inode->i_mode)) {
678 inode->i_op = de->proc_iops;
679 if (de->proc_ops->proc_read_iter)
680 inode->i_fop = &proc_iter_file_ops;
682 inode->i_fop = &proc_reg_file_ops;
684 if (de->proc_ops->proc_compat_ioctl) {
685 if (de->proc_ops->proc_read_iter)
686 inode->i_fop = &proc_iter_file_ops_compat;
688 inode->i_fop = &proc_reg_file_ops_compat;
691 } else if (S_ISDIR(inode->i_mode)) {
692 inode->i_op = de->proc_iops;
693 inode->i_fop = de->proc_dir_ops;
694 } else if (S_ISLNK(inode->i_mode)) {
695 inode->i_op = de->proc_iops;