2 * linux/fs/proc/inode.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/time.h>
8 #include <linux/proc_fs.h>
9 #include <linux/kernel.h>
10 #include <linux/pid_namespace.h>
12 #include <linux/string.h>
13 #include <linux/stat.h>
14 #include <linux/completion.h>
15 #include <linux/poll.h>
16 #include <linux/printk.h>
17 #include <linux/file.h>
18 #include <linux/limits.h>
19 #include <linux/init.h>
20 #include <linux/module.h>
21 #include <linux/sysctl.h>
22 #include <linux/seq_file.h>
23 #include <linux/slab.h>
24 #include <linux/mount.h>
25 #include <linux/magic.h>
27 #include <asm/uaccess.h>
31 static void proc_evict_inode(struct inode *inode)
33 struct proc_dir_entry *de;
34 struct ctl_table_header *head;
35 const struct proc_ns_operations *ns_ops;
38 truncate_inode_pages(&inode->i_data, 0);
41 /* Stop tracking associated processes */
42 put_pid(PROC_I(inode)->pid);
44 /* Let go of any associated proc directory entry */
45 de = PROC_I(inode)->pde;
48 head = PROC_I(inode)->sysctl;
50 rcu_assign_pointer(PROC_I(inode)->sysctl, NULL);
51 sysctl_head_put(head);
53 /* Release any associated namespace */
54 ns_ops = PROC_I(inode)->ns.ns_ops;
55 ns = PROC_I(inode)->ns.ns;
60 static struct kmem_cache * proc_inode_cachep;
62 static struct inode *proc_alloc_inode(struct super_block *sb)
64 struct proc_inode *ei;
67 ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
72 ei->op.proc_get_link = NULL;
75 ei->sysctl_entry = NULL;
78 inode = &ei->vfs_inode;
79 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
83 static void proc_i_callback(struct rcu_head *head)
85 struct inode *inode = container_of(head, struct inode, i_rcu);
86 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
89 static void proc_destroy_inode(struct inode *inode)
91 call_rcu(&inode->i_rcu, proc_i_callback);
94 static void init_once(void *foo)
96 struct proc_inode *ei = (struct proc_inode *) foo;
98 inode_init_once(&ei->vfs_inode);
101 void __init proc_init_inodecache(void)
103 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
104 sizeof(struct proc_inode),
105 0, (SLAB_RECLAIM_ACCOUNT|
106 SLAB_MEM_SPREAD|SLAB_PANIC),
110 static int proc_show_options(struct seq_file *seq, struct dentry *root)
112 struct super_block *sb = root->d_sb;
113 struct pid_namespace *pid = sb->s_fs_info;
115 if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
116 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
117 if (pid->hide_pid != 0)
118 seq_printf(seq, ",hidepid=%u", pid->hide_pid);
123 static const struct super_operations proc_sops = {
124 .alloc_inode = proc_alloc_inode,
125 .destroy_inode = proc_destroy_inode,
126 .drop_inode = generic_delete_inode,
127 .evict_inode = proc_evict_inode,
128 .statfs = simple_statfs,
129 .remount_fs = proc_remount,
130 .show_options = proc_show_options,
133 enum {BIAS = -1U<<31};
135 static inline int use_pde(struct proc_dir_entry *pde)
137 return atomic_inc_unless_negative(&pde->in_use);
140 static void unuse_pde(struct proc_dir_entry *pde)
142 if (atomic_dec_return(&pde->in_use) == BIAS)
143 complete(pde->pde_unload_completion);
147 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
150 /* somebody else is doing that, just wait */
151 DECLARE_COMPLETION_ONSTACK(c);
153 spin_unlock(&pde->pde_unload_lock);
154 wait_for_completion(&c);
155 spin_lock(&pde->pde_unload_lock);
159 spin_unlock(&pde->pde_unload_lock);
161 pde->proc_fops->release(file_inode(file), file);
162 spin_lock(&pde->pde_unload_lock);
163 list_del_init(&pdeo->lh);
170 void proc_entry_rundown(struct proc_dir_entry *de)
172 DECLARE_COMPLETION_ONSTACK(c);
173 /* Wait until all existing callers into module are done. */
174 de->pde_unload_completion = &c;
175 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
176 wait_for_completion(&c);
178 spin_lock(&de->pde_unload_lock);
179 while (!list_empty(&de->pde_openers)) {
180 struct pde_opener *pdeo;
181 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
182 close_pdeo(de, pdeo);
184 spin_unlock(&de->pde_unload_lock);
187 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
189 struct proc_dir_entry *pde = PDE(file_inode(file));
192 loff_t (*llseek)(struct file *, loff_t, int);
193 llseek = pde->proc_fops->llseek;
195 llseek = default_llseek;
196 rv = llseek(file, offset, whence);
202 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
204 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
205 struct proc_dir_entry *pde = PDE(file_inode(file));
208 read = pde->proc_fops->read;
210 rv = read(file, buf, count, ppos);
216 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
218 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
219 struct proc_dir_entry *pde = PDE(file_inode(file));
222 write = pde->proc_fops->write;
224 rv = write(file, buf, count, ppos);
230 static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
232 struct proc_dir_entry *pde = PDE(file_inode(file));
233 unsigned int rv = DEFAULT_POLLMASK;
234 unsigned int (*poll)(struct file *, struct poll_table_struct *);
236 poll = pde->proc_fops->poll;
238 rv = poll(file, pts);
244 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
246 struct proc_dir_entry *pde = PDE(file_inode(file));
248 long (*ioctl)(struct file *, unsigned int, unsigned long);
250 ioctl = pde->proc_fops->unlocked_ioctl;
252 rv = ioctl(file, cmd, arg);
259 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
261 struct proc_dir_entry *pde = PDE(file_inode(file));
263 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
265 compat_ioctl = pde->proc_fops->compat_ioctl;
267 rv = compat_ioctl(file, cmd, arg);
274 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
276 struct proc_dir_entry *pde = PDE(file_inode(file));
278 int (*mmap)(struct file *, struct vm_area_struct *);
280 mmap = pde->proc_fops->mmap;
282 rv = mmap(file, vma);
288 static unsigned long proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr, unsigned long len, unsigned long pgoff, unsigned long flags)
290 struct proc_dir_entry *pde = PDE(file_inode(file));
292 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
294 get_unmapped_area = pde->proc_fops->get_unmapped_area;
295 if (get_unmapped_area)
296 rv = get_unmapped_area(file, orig_addr, len, pgoff, flags);
302 static int proc_reg_open(struct inode *inode, struct file *file)
304 struct proc_dir_entry *pde = PDE(inode);
306 int (*open)(struct inode *, struct file *);
307 int (*release)(struct inode *, struct file *);
308 struct pde_opener *pdeo;
311 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
312 * sequence. ->release won't be called because ->proc_fops will be
313 * cleared. Depending on complexity of ->release, consequences vary.
315 * We can't wait for mercy when close will be done for real, it's
316 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
317 * by hand in remove_proc_entry(). For this, save opener's credentials
320 pdeo = kzalloc(sizeof(struct pde_opener), GFP_KERNEL);
328 open = pde->proc_fops->open;
329 release = pde->proc_fops->release;
332 rv = open(inode, file);
334 if (rv == 0 && release) {
335 /* To know what to release. */
337 /* Strictly for "too late" ->release in proc_reg_release(). */
338 spin_lock(&pde->pde_unload_lock);
339 list_add(&pdeo->lh, &pde->pde_openers);
340 spin_unlock(&pde->pde_unload_lock);
348 static int proc_reg_release(struct inode *inode, struct file *file)
350 struct proc_dir_entry *pde = PDE(inode);
351 struct pde_opener *pdeo;
352 spin_lock(&pde->pde_unload_lock);
353 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
354 if (pdeo->file == file) {
355 close_pdeo(pde, pdeo);
359 spin_unlock(&pde->pde_unload_lock);
363 static const struct file_operations proc_reg_file_ops = {
364 .llseek = proc_reg_llseek,
365 .read = proc_reg_read,
366 .write = proc_reg_write,
367 .poll = proc_reg_poll,
368 .unlocked_ioctl = proc_reg_unlocked_ioctl,
370 .compat_ioctl = proc_reg_compat_ioctl,
372 .mmap = proc_reg_mmap,
373 .get_unmapped_area = proc_reg_get_unmapped_area,
374 .open = proc_reg_open,
375 .release = proc_reg_release,
379 static const struct file_operations proc_reg_file_ops_no_compat = {
380 .llseek = proc_reg_llseek,
381 .read = proc_reg_read,
382 .write = proc_reg_write,
383 .poll = proc_reg_poll,
384 .unlocked_ioctl = proc_reg_unlocked_ioctl,
385 .mmap = proc_reg_mmap,
386 .get_unmapped_area = proc_reg_get_unmapped_area,
387 .open = proc_reg_open,
388 .release = proc_reg_release,
392 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
394 struct inode *inode = new_inode_pseudo(sb);
397 inode->i_ino = de->low_ino;
398 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
399 PROC_I(inode)->pde = de;
402 inode->i_mode = de->mode;
403 inode->i_uid = de->uid;
404 inode->i_gid = de->gid;
407 inode->i_size = de->size;
409 set_nlink(inode, de->nlink);
410 WARN_ON(!de->proc_iops);
411 inode->i_op = de->proc_iops;
413 if (S_ISREG(inode->i_mode)) {
415 if (!de->proc_fops->compat_ioctl)
417 &proc_reg_file_ops_no_compat;
420 inode->i_fop = &proc_reg_file_ops;
422 inode->i_fop = de->proc_fops;
430 int proc_fill_super(struct super_block *s)
432 struct inode *root_inode;
434 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
435 s->s_blocksize = 1024;
436 s->s_blocksize_bits = 10;
437 s->s_magic = PROC_SUPER_MAGIC;
438 s->s_op = &proc_sops;
442 root_inode = proc_get_inode(s, &proc_root);
444 pr_err("proc_fill_super: get root inode failed\n");
448 s->s_root = d_make_root(root_inode);
450 pr_err("proc_fill_super: allocate dentry failed\n");
454 return proc_setup_self(s);