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>
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;
37 struct proc_inode *ei = PROC_I(inode);
39 truncate_inode_pages_final(&inode->i_data);
42 /* Stop tracking associated processes */
44 proc_pid_evict_inode(ei);
48 /* Let go of any associated proc directory entry */
57 RCU_INIT_POINTER(ei->sysctl, NULL);
58 proc_sys_evict_inode(inode, head);
62 static struct kmem_cache *proc_inode_cachep __ro_after_init;
63 static struct kmem_cache *pde_opener_cache __ro_after_init;
65 static struct inode *proc_alloc_inode(struct super_block *sb)
67 struct proc_inode *ei;
69 ei = kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
74 ei->op.proc_get_link = NULL;
77 ei->sysctl_entry = NULL;
78 INIT_HLIST_NODE(&ei->sibling_inodes);
80 return &ei->vfs_inode;
83 static void proc_free_inode(struct inode *inode)
85 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
88 static void init_once(void *foo)
90 struct proc_inode *ei = (struct proc_inode *) foo;
92 inode_init_once(&ei->vfs_inode);
95 void __init proc_init_kmemcache(void)
97 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
98 sizeof(struct proc_inode),
99 0, (SLAB_RECLAIM_ACCOUNT|
100 SLAB_MEM_SPREAD|SLAB_ACCOUNT|
104 kmem_cache_create("pde_opener", sizeof(struct pde_opener), 0,
105 SLAB_ACCOUNT|SLAB_PANIC, NULL);
106 proc_dir_entry_cache = kmem_cache_create_usercopy(
107 "proc_dir_entry", SIZEOF_PDE, 0, SLAB_PANIC,
108 offsetof(struct proc_dir_entry, inline_name),
109 SIZEOF_PDE_INLINE_NAME, NULL);
110 BUILD_BUG_ON(sizeof(struct proc_dir_entry) >= SIZEOF_PDE);
113 void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock)
116 struct proc_inode *ei;
117 struct hlist_node *node;
118 struct super_block *old_sb = NULL;
122 struct super_block *sb;
123 node = hlist_first_rcu(inodes);
126 ei = hlist_entry(node, struct proc_inode, sibling_inodes);
128 hlist_del_init_rcu(&ei->sibling_inodes);
131 inode = &ei->vfs_inode;
133 if ((sb != old_sb) && !atomic_inc_not_zero(&sb->s_active))
135 inode = igrab(inode);
139 deactivate_super(old_sb);
142 if (unlikely(!inode)) {
147 if (S_ISDIR(inode->i_mode)) {
148 struct dentry *dir = d_find_any_alias(inode);
154 struct dentry *dentry;
155 while ((dentry = d_find_alias(inode))) {
156 d_invalidate(dentry);
166 deactivate_super(old_sb);
169 static inline const char *hidepid2str(enum proc_hidepid v)
172 case HIDEPID_OFF: return "off";
173 case HIDEPID_NO_ACCESS: return "noaccess";
174 case HIDEPID_INVISIBLE: return "invisible";
175 case HIDEPID_NOT_PTRACEABLE: return "ptraceable";
177 WARN_ONCE(1, "bad hide_pid value: %d\n", v);
181 static int proc_show_options(struct seq_file *seq, struct dentry *root)
183 struct proc_fs_info *fs_info = proc_sb_info(root->d_sb);
185 if (!gid_eq(fs_info->pid_gid, GLOBAL_ROOT_GID))
186 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, fs_info->pid_gid));
187 if (fs_info->hide_pid != HIDEPID_OFF)
188 seq_printf(seq, ",hidepid=%s", hidepid2str(fs_info->hide_pid));
189 if (fs_info->pidonly != PROC_PIDONLY_OFF)
190 seq_printf(seq, ",subset=pid");
195 const struct super_operations proc_sops = {
196 .alloc_inode = proc_alloc_inode,
197 .free_inode = proc_free_inode,
198 .drop_inode = generic_delete_inode,
199 .evict_inode = proc_evict_inode,
200 .statfs = simple_statfs,
201 .show_options = proc_show_options,
204 enum {BIAS = -1U<<31};
206 static inline int use_pde(struct proc_dir_entry *pde)
208 return likely(atomic_inc_unless_negative(&pde->in_use));
211 static void unuse_pde(struct proc_dir_entry *pde)
213 if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
214 complete(pde->pde_unload_completion);
217 /* 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.
228 * Therefore, first process to enter this function does ->release() and
229 * signals its completion to the other process which does nothing.
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);
244 pde->proc_ops->proc_release(file_inode(file), file);
245 spin_lock(&pde->pde_unload_lock);
246 /* After ->release. */
249 spin_unlock(&pde->pde_unload_lock);
252 kmem_cache_free(pde_opener_cache, pdeo);
256 void proc_entry_rundown(struct proc_dir_entry *de)
258 DECLARE_COMPLETION_ONSTACK(c);
259 /* Wait until all existing callers into module are done. */
260 de->pde_unload_completion = &c;
261 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
262 wait_for_completion(&c);
264 /* ->pde_openers list can't grow from now on. */
266 spin_lock(&de->pde_unload_lock);
267 while (!list_empty(&de->pde_openers)) {
268 struct pde_opener *pdeo;
269 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
270 close_pdeo(de, pdeo);
271 spin_lock(&de->pde_unload_lock);
273 spin_unlock(&de->pde_unload_lock);
276 static loff_t pde_lseek(struct proc_dir_entry *pde, struct file *file, loff_t offset, int whence)
278 typeof_member(struct proc_ops, proc_lseek) lseek;
280 lseek = pde->proc_ops->proc_lseek;
282 lseek = default_llseek;
283 return lseek(file, offset, whence);
286 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
288 struct proc_dir_entry *pde = PDE(file_inode(file));
291 if (pde_is_permanent(pde)) {
292 return pde_lseek(pde, file, offset, whence);
293 } else if (use_pde(pde)) {
294 rv = pde_lseek(pde, file, offset, whence);
300 static ssize_t pde_read(struct proc_dir_entry *pde, struct file *file, char __user *buf, size_t count, loff_t *ppos)
302 typeof_member(struct proc_ops, proc_read) read;
304 read = pde->proc_ops->proc_read;
306 return read(file, buf, count, ppos);
310 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
312 struct proc_dir_entry *pde = PDE(file_inode(file));
315 if (pde_is_permanent(pde)) {
316 return pde_read(pde, file, buf, count, ppos);
317 } else if (use_pde(pde)) {
318 rv = pde_read(pde, file, buf, count, ppos);
324 static ssize_t pde_write(struct proc_dir_entry *pde, struct file *file, const char __user *buf, size_t count, loff_t *ppos)
326 typeof_member(struct proc_ops, proc_write) write;
328 write = pde->proc_ops->proc_write;
330 return write(file, buf, count, ppos);
334 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
336 struct proc_dir_entry *pde = PDE(file_inode(file));
339 if (pde_is_permanent(pde)) {
340 return pde_write(pde, file, buf, count, ppos);
341 } else if (use_pde(pde)) {
342 rv = pde_write(pde, file, buf, count, ppos);
348 static __poll_t pde_poll(struct proc_dir_entry *pde, struct file *file, struct poll_table_struct *pts)
350 typeof_member(struct proc_ops, proc_poll) poll;
352 poll = pde->proc_ops->proc_poll;
354 return poll(file, pts);
355 return DEFAULT_POLLMASK;
358 static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
360 struct proc_dir_entry *pde = PDE(file_inode(file));
361 __poll_t rv = DEFAULT_POLLMASK;
363 if (pde_is_permanent(pde)) {
364 return pde_poll(pde, file, pts);
365 } else if (use_pde(pde)) {
366 rv = pde_poll(pde, file, pts);
372 static long pde_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
374 typeof_member(struct proc_ops, proc_ioctl) ioctl;
376 ioctl = pde->proc_ops->proc_ioctl;
378 return ioctl(file, cmd, arg);
382 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
384 struct proc_dir_entry *pde = PDE(file_inode(file));
387 if (pde_is_permanent(pde)) {
388 return pde_ioctl(pde, file, cmd, arg);
389 } else if (use_pde(pde)) {
390 rv = pde_ioctl(pde, file, cmd, arg);
397 static long pde_compat_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
399 typeof_member(struct proc_ops, proc_compat_ioctl) compat_ioctl;
401 compat_ioctl = pde->proc_ops->proc_compat_ioctl;
403 return compat_ioctl(file, cmd, arg);
407 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
409 struct proc_dir_entry *pde = PDE(file_inode(file));
411 if (pde_is_permanent(pde)) {
412 return pde_compat_ioctl(pde, file, cmd, arg);
413 } else if (use_pde(pde)) {
414 rv = pde_compat_ioctl(pde, file, cmd, arg);
421 static int pde_mmap(struct proc_dir_entry *pde, struct file *file, struct vm_area_struct *vma)
423 typeof_member(struct proc_ops, proc_mmap) mmap;
425 mmap = pde->proc_ops->proc_mmap;
427 return mmap(file, vma);
431 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
433 struct proc_dir_entry *pde = PDE(file_inode(file));
436 if (pde_is_permanent(pde)) {
437 return pde_mmap(pde, file, vma);
438 } else if (use_pde(pde)) {
439 rv = pde_mmap(pde, file, vma);
446 pde_get_unmapped_area(struct proc_dir_entry *pde, struct file *file, unsigned long orig_addr,
447 unsigned long len, unsigned long pgoff,
450 typeof_member(struct proc_ops, proc_get_unmapped_area) get_area;
452 get_area = pde->proc_ops->proc_get_unmapped_area;
455 get_area = current->mm->get_unmapped_area;
458 return get_area(file, orig_addr, len, pgoff, flags);
463 proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
464 unsigned long len, unsigned long pgoff,
467 struct proc_dir_entry *pde = PDE(file_inode(file));
468 unsigned long rv = -EIO;
470 if (pde_is_permanent(pde)) {
471 return pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
472 } else if (use_pde(pde)) {
473 rv = pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
479 static int proc_reg_open(struct inode *inode, struct file *file)
481 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
482 struct proc_dir_entry *pde = PDE(inode);
484 typeof_member(struct proc_ops, proc_open) open;
485 typeof_member(struct proc_ops, proc_release) release;
486 struct pde_opener *pdeo;
488 if (pde_is_permanent(pde)) {
489 open = pde->proc_ops->proc_open;
491 rv = open(inode, file);
495 if (fs_info->pidonly == PROC_PIDONLY_ON)
500 * 1) PDE's ->release hook will be called no matter what
501 * either normally by close()/->release, or forcefully by
502 * rmmod/remove_proc_entry.
504 * 2) rmmod isn't blocked by opening file in /proc and sitting on
505 * the descriptor (including "rmmod foo </proc/foo" scenario).
507 * Save every "struct file" with custom ->release hook.
512 release = pde->proc_ops->proc_release;
514 pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
521 open = pde->proc_ops->proc_open;
523 rv = open(inode, file);
527 /* To know what to release. */
529 pdeo->closing = false;
531 spin_lock(&pde->pde_unload_lock);
532 list_add(&pdeo->lh, &pde->pde_openers);
533 spin_unlock(&pde->pde_unload_lock);
535 kmem_cache_free(pde_opener_cache, pdeo);
543 static int proc_reg_release(struct inode *inode, struct file *file)
545 struct proc_dir_entry *pde = PDE(inode);
546 struct pde_opener *pdeo;
548 if (pde_is_permanent(pde)) {
549 typeof_member(struct proc_ops, proc_release) release;
551 release = pde->proc_ops->proc_release;
553 return release(inode, file);
558 spin_lock(&pde->pde_unload_lock);
559 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
560 if (pdeo->file == file) {
561 close_pdeo(pde, pdeo);
565 spin_unlock(&pde->pde_unload_lock);
569 static const struct file_operations proc_reg_file_ops = {
570 .llseek = proc_reg_llseek,
571 .read = proc_reg_read,
572 .write = proc_reg_write,
573 .poll = proc_reg_poll,
574 .unlocked_ioctl = proc_reg_unlocked_ioctl,
576 .compat_ioctl = proc_reg_compat_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,
585 static const struct file_operations proc_reg_file_ops_no_compat = {
586 .llseek = proc_reg_llseek,
587 .read = proc_reg_read,
588 .write = proc_reg_write,
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 void proc_put_link(void *p)
603 static const char *proc_get_link(struct dentry *dentry,
605 struct delayed_call *done)
607 struct proc_dir_entry *pde = PDE(inode);
609 return ERR_PTR(-EINVAL);
610 set_delayed_call(done, proc_put_link, pde);
614 const struct inode_operations proc_link_inode_operations = {
615 .get_link = proc_get_link,
618 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
620 struct inode *inode = new_inode(sb);
623 inode->i_ino = de->low_ino;
624 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
625 PROC_I(inode)->pde = de;
627 if (is_empty_pde(de)) {
628 make_empty_dir_inode(inode);
632 inode->i_mode = de->mode;
633 inode->i_uid = de->uid;
634 inode->i_gid = de->gid;
637 inode->i_size = de->size;
639 set_nlink(inode, de->nlink);
641 if (S_ISREG(inode->i_mode)) {
642 inode->i_op = de->proc_iops;
643 inode->i_fop = &proc_reg_file_ops;
645 if (!de->proc_ops->proc_compat_ioctl) {
646 inode->i_fop = &proc_reg_file_ops_no_compat;
649 } else if (S_ISDIR(inode->i_mode)) {
650 inode->i_op = de->proc_iops;
651 inode->i_fop = de->proc_dir_ops;
652 } else if (S_ISLNK(inode->i_mode)) {
653 inode->i_op = de->proc_iops;