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)
114 struct proc_inode *ei;
115 struct hlist_node *node;
116 struct super_block *old_sb = NULL;
120 struct super_block *sb;
121 node = hlist_first_rcu(inodes);
124 ei = hlist_entry(node, struct proc_inode, sibling_inodes);
126 hlist_del_init_rcu(&ei->sibling_inodes);
129 inode = &ei->vfs_inode;
131 if ((sb != old_sb) && !atomic_inc_not_zero(&sb->s_active))
133 inode = igrab(inode);
137 deactivate_super(old_sb);
140 if (unlikely(!inode)) {
145 if (S_ISDIR(inode->i_mode)) {
146 struct dentry *dir = d_find_any_alias(inode);
152 struct dentry *dentry;
153 while ((dentry = d_find_alias(inode))) {
154 d_invalidate(dentry);
164 deactivate_super(old_sb);
167 static inline const char *hidepid2str(enum proc_hidepid v)
170 case HIDEPID_OFF: return "off";
171 case HIDEPID_NO_ACCESS: return "noaccess";
172 case HIDEPID_INVISIBLE: return "invisible";
173 case HIDEPID_NOT_PTRACEABLE: return "ptraceable";
175 WARN_ONCE(1, "bad hide_pid value: %d\n", v);
179 static int proc_show_options(struct seq_file *seq, struct dentry *root)
181 struct proc_fs_info *fs_info = proc_sb_info(root->d_sb);
183 if (!gid_eq(fs_info->pid_gid, GLOBAL_ROOT_GID))
184 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, fs_info->pid_gid));
185 if (fs_info->hide_pid != HIDEPID_OFF)
186 seq_printf(seq, ",hidepid=%s", hidepid2str(fs_info->hide_pid));
187 if (fs_info->pidonly != PROC_PIDONLY_OFF)
188 seq_printf(seq, ",subset=pid");
193 const struct super_operations proc_sops = {
194 .alloc_inode = proc_alloc_inode,
195 .free_inode = proc_free_inode,
196 .drop_inode = generic_delete_inode,
197 .evict_inode = proc_evict_inode,
198 .statfs = simple_statfs,
199 .show_options = proc_show_options,
202 enum {BIAS = -1U<<31};
204 static inline int use_pde(struct proc_dir_entry *pde)
206 return likely(atomic_inc_unless_negative(&pde->in_use));
209 static void unuse_pde(struct proc_dir_entry *pde)
211 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
212 complete(pde->pde_unload_completion);
216 * At most 2 contexts can enter this function: the one doing the last
217 * close on the descriptor and whoever is deleting PDE itself.
219 * First to enter calls ->proc_release hook and signals its completion
220 * to the second one which waits and then does nothing.
222 * PDE is locked on entry, unlocked on exit.
224 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
225 __releases(&pde->pde_unload_lock)
228 * close() (proc_reg_release()) can't delete an entry and proceed:
229 * ->release hook needs to be available at the right moment.
231 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
232 * "struct file" needs to be available at the right moment.
235 /* somebody else is doing that, just wait */
236 DECLARE_COMPLETION_ONSTACK(c);
238 spin_unlock(&pde->pde_unload_lock);
239 wait_for_completion(&c);
242 struct completion *c;
244 pdeo->closing = true;
245 spin_unlock(&pde->pde_unload_lock);
248 pde->proc_ops->proc_release(file_inode(file), file);
250 spin_lock(&pde->pde_unload_lock);
251 /* Strictly after ->proc_release, see above. */
254 spin_unlock(&pde->pde_unload_lock);
257 kmem_cache_free(pde_opener_cache, pdeo);
261 void proc_entry_rundown(struct proc_dir_entry *de)
263 DECLARE_COMPLETION_ONSTACK(c);
264 /* Wait until all existing callers into module are done. */
265 de->pde_unload_completion = &c;
266 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
267 wait_for_completion(&c);
269 /* ->pde_openers list can't grow from now on. */
271 spin_lock(&de->pde_unload_lock);
272 while (!list_empty(&de->pde_openers)) {
273 struct pde_opener *pdeo;
274 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
275 close_pdeo(de, pdeo);
276 spin_lock(&de->pde_unload_lock);
278 spin_unlock(&de->pde_unload_lock);
281 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
283 struct proc_dir_entry *pde = PDE(file_inode(file));
286 if (pde_is_permanent(pde)) {
287 return pde->proc_ops->proc_lseek(file, offset, whence);
288 } else if (use_pde(pde)) {
289 rv = pde->proc_ops->proc_lseek(file, offset, whence);
295 static ssize_t proc_reg_read_iter(struct kiocb *iocb, struct iov_iter *iter)
297 struct proc_dir_entry *pde = PDE(file_inode(iocb->ki_filp));
300 if (pde_is_permanent(pde))
301 return pde->proc_ops->proc_read_iter(iocb, iter);
305 ret = pde->proc_ops->proc_read_iter(iocb, iter);
310 static ssize_t pde_read(struct proc_dir_entry *pde, struct file *file, char __user *buf, size_t count, loff_t *ppos)
312 typeof_member(struct proc_ops, proc_read) read;
314 read = pde->proc_ops->proc_read;
316 return read(file, buf, count, ppos);
320 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
322 struct proc_dir_entry *pde = PDE(file_inode(file));
325 if (pde_is_permanent(pde)) {
326 return pde_read(pde, file, buf, count, ppos);
327 } else if (use_pde(pde)) {
328 rv = pde_read(pde, file, buf, count, ppos);
334 static ssize_t pde_write(struct proc_dir_entry *pde, struct file *file, const char __user *buf, size_t count, loff_t *ppos)
336 typeof_member(struct proc_ops, proc_write) write;
338 write = pde->proc_ops->proc_write;
340 return write(file, buf, count, ppos);
344 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
346 struct proc_dir_entry *pde = PDE(file_inode(file));
349 if (pde_is_permanent(pde)) {
350 return pde_write(pde, file, buf, count, ppos);
351 } else if (use_pde(pde)) {
352 rv = pde_write(pde, file, buf, count, ppos);
358 static __poll_t pde_poll(struct proc_dir_entry *pde, struct file *file, struct poll_table_struct *pts)
360 typeof_member(struct proc_ops, proc_poll) poll;
362 poll = pde->proc_ops->proc_poll;
364 return poll(file, pts);
365 return DEFAULT_POLLMASK;
368 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
370 struct proc_dir_entry *pde = PDE(file_inode(file));
371 __poll_t rv = DEFAULT_POLLMASK;
373 if (pde_is_permanent(pde)) {
374 return pde_poll(pde, file, pts);
375 } else if (use_pde(pde)) {
376 rv = pde_poll(pde, file, pts);
382 static long pde_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
384 typeof_member(struct proc_ops, proc_ioctl) ioctl;
386 ioctl = pde->proc_ops->proc_ioctl;
388 return ioctl(file, cmd, arg);
392 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
394 struct proc_dir_entry *pde = PDE(file_inode(file));
397 if (pde_is_permanent(pde)) {
398 return pde_ioctl(pde, file, cmd, arg);
399 } else if (use_pde(pde)) {
400 rv = pde_ioctl(pde, file, cmd, arg);
407 static long pde_compat_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
409 typeof_member(struct proc_ops, proc_compat_ioctl) compat_ioctl;
411 compat_ioctl = pde->proc_ops->proc_compat_ioctl;
413 return compat_ioctl(file, cmd, arg);
417 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
419 struct proc_dir_entry *pde = PDE(file_inode(file));
421 if (pde_is_permanent(pde)) {
422 return pde_compat_ioctl(pde, file, cmd, arg);
423 } else if (use_pde(pde)) {
424 rv = pde_compat_ioctl(pde, file, cmd, arg);
431 static int pde_mmap(struct proc_dir_entry *pde, struct file *file, struct vm_area_struct *vma)
433 typeof_member(struct proc_ops, proc_mmap) mmap;
435 mmap = pde->proc_ops->proc_mmap;
437 return mmap(file, vma);
441 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
443 struct proc_dir_entry *pde = PDE(file_inode(file));
446 if (pde_is_permanent(pde)) {
447 return pde_mmap(pde, file, vma);
448 } else if (use_pde(pde)) {
449 rv = pde_mmap(pde, file, vma);
456 pde_get_unmapped_area(struct proc_dir_entry *pde, struct file *file, unsigned long orig_addr,
457 unsigned long len, unsigned long pgoff,
460 typeof_member(struct proc_ops, proc_get_unmapped_area) get_area;
462 get_area = pde->proc_ops->proc_get_unmapped_area;
465 get_area = current->mm->get_unmapped_area;
468 return get_area(file, orig_addr, len, pgoff, flags);
473 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
474 unsigned long len, unsigned long pgoff,
477 struct proc_dir_entry *pde = PDE(file_inode(file));
478 unsigned long rv = -EIO;
480 if (pde_is_permanent(pde)) {
481 return pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
482 } else if (use_pde(pde)) {
483 rv = pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
489 static int proc_reg_open(struct inode *inode, struct file *file)
491 struct proc_dir_entry *pde = PDE(inode);
493 typeof_member(struct proc_ops, proc_open) open;
494 typeof_member(struct proc_ops, proc_release) release;
495 struct pde_opener *pdeo;
497 if (!pde->proc_ops->proc_lseek)
498 file->f_mode &= ~FMODE_LSEEK;
500 if (pde_is_permanent(pde)) {
501 open = pde->proc_ops->proc_open;
503 rv = open(inode, file);
509 * 1) PDE's ->release hook will be called no matter what
510 * either normally by close()/->release, or forcefully by
511 * rmmod/remove_proc_entry.
513 * 2) rmmod isn't blocked by opening file in /proc and sitting on
514 * the descriptor (including "rmmod foo </proc/foo" scenario).
516 * Save every "struct file" with custom ->release hook.
521 release = pde->proc_ops->proc_release;
523 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
530 open = pde->proc_ops->proc_open;
532 rv = open(inode, file);
536 /* To know what to release. */
538 pdeo->closing = false;
540 spin_lock(&pde->pde_unload_lock);
541 list_add(&pdeo->lh, &pde->pde_openers);
542 spin_unlock(&pde->pde_unload_lock);
544 kmem_cache_free(pde_opener_cache, pdeo);
552 static int proc_reg_release(struct inode *inode, struct file *file)
554 struct proc_dir_entry *pde = PDE(inode);
555 struct pde_opener *pdeo;
557 if (pde_is_permanent(pde)) {
558 typeof_member(struct proc_ops, proc_release) release;
560 release = pde->proc_ops->proc_release;
562 return release(inode, file);
567 spin_lock(&pde->pde_unload_lock);
568 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
569 if (pdeo->file == file) {
570 close_pdeo(pde, pdeo);
574 spin_unlock(&pde->pde_unload_lock);
578 static const struct file_operations proc_reg_file_ops = {
579 .llseek = proc_reg_llseek,
580 .read = proc_reg_read,
581 .write = proc_reg_write,
582 .poll = proc_reg_poll,
583 .unlocked_ioctl = proc_reg_unlocked_ioctl,
584 .mmap = proc_reg_mmap,
585 .get_unmapped_area = proc_reg_get_unmapped_area,
586 .open = proc_reg_open,
587 .release = proc_reg_release,
590 static const struct file_operations proc_iter_file_ops = {
591 .llseek = proc_reg_llseek,
592 .read_iter = proc_reg_read_iter,
593 .write = proc_reg_write,
594 .splice_read = copy_splice_read,
595 .poll = proc_reg_poll,
596 .unlocked_ioctl = proc_reg_unlocked_ioctl,
597 .mmap = proc_reg_mmap,
598 .get_unmapped_area = proc_reg_get_unmapped_area,
599 .open = proc_reg_open,
600 .release = proc_reg_release,
604 static const struct file_operations proc_reg_file_ops_compat = {
605 .llseek = proc_reg_llseek,
606 .read = proc_reg_read,
607 .write = proc_reg_write,
608 .poll = proc_reg_poll,
609 .unlocked_ioctl = proc_reg_unlocked_ioctl,
610 .compat_ioctl = proc_reg_compat_ioctl,
611 .mmap = proc_reg_mmap,
612 .get_unmapped_area = proc_reg_get_unmapped_area,
613 .open = proc_reg_open,
614 .release = proc_reg_release,
617 static const struct file_operations proc_iter_file_ops_compat = {
618 .llseek = proc_reg_llseek,
619 .read_iter = proc_reg_read_iter,
620 .splice_read = copy_splice_read,
621 .write = proc_reg_write,
622 .poll = proc_reg_poll,
623 .unlocked_ioctl = proc_reg_unlocked_ioctl,
624 .compat_ioctl = proc_reg_compat_ioctl,
625 .mmap = proc_reg_mmap,
626 .get_unmapped_area = proc_reg_get_unmapped_area,
627 .open = proc_reg_open,
628 .release = proc_reg_release,
632 static void proc_put_link(void *p)
637 static const char *proc_get_link(struct dentry *dentry,
639 struct delayed_call *done)
641 struct proc_dir_entry *pde = PDE(inode);
643 return ERR_PTR(-EINVAL);
644 set_delayed_call(done, proc_put_link, pde);
648 const struct inode_operations proc_link_inode_operations = {
649 .get_link = proc_get_link,
652 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
654 struct inode *inode = new_inode(sb);
661 inode->i_private = de->data;
662 inode->i_ino = de->low_ino;
663 inode->i_mtime = inode->i_atime = inode_set_ctime_current(inode);
664 PROC_I(inode)->pde = de;
665 if (is_empty_pde(de)) {
666 make_empty_dir_inode(inode);
671 inode->i_mode = de->mode;
672 inode->i_uid = de->uid;
673 inode->i_gid = de->gid;
676 inode->i_size = de->size;
678 set_nlink(inode, de->nlink);
680 if (S_ISREG(inode->i_mode)) {
681 inode->i_op = de->proc_iops;
682 if (de->proc_ops->proc_read_iter)
683 inode->i_fop = &proc_iter_file_ops;
685 inode->i_fop = &proc_reg_file_ops;
687 if (de->proc_ops->proc_compat_ioctl) {
688 if (de->proc_ops->proc_read_iter)
689 inode->i_fop = &proc_iter_file_ops_compat;
691 inode->i_fop = &proc_reg_file_ops_compat;
694 } else if (S_ISDIR(inode->i_mode)) {
695 inode->i_op = de->proc_iops;
696 inode->i_fop = de->proc_dir_ops;
697 } else if (S_ISLNK(inode->i_mode)) {
698 inode->i_op = de->proc_iops;