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>
26 #include <linux/namei.h>
28 #include <asm/uaccess.h>
32 static void proc_evict_inode(struct inode *inode)
34 struct proc_dir_entry *de;
35 struct ctl_table_header *head;
36 const struct proc_ns_operations *ns_ops;
39 truncate_inode_pages(&inode->i_data, 0);
42 /* Stop tracking associated processes */
43 put_pid(PROC_I(inode)->pid);
45 /* Let go of any associated proc directory entry */
46 de = PROC_I(inode)->pde;
49 head = PROC_I(inode)->sysctl;
51 rcu_assign_pointer(PROC_I(inode)->sysctl, NULL);
52 sysctl_head_put(head);
54 /* Release any associated namespace */
55 ns_ops = PROC_I(inode)->ns.ns_ops;
56 ns = PROC_I(inode)->ns.ns;
61 static struct kmem_cache * proc_inode_cachep;
63 static struct inode *proc_alloc_inode(struct super_block *sb)
65 struct proc_inode *ei;
68 ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
73 ei->op.proc_get_link = NULL;
76 ei->sysctl_entry = NULL;
79 inode = &ei->vfs_inode;
80 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
84 static void proc_i_callback(struct rcu_head *head)
86 struct inode *inode = container_of(head, struct inode, i_rcu);
87 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
90 static void proc_destroy_inode(struct inode *inode)
92 call_rcu(&inode->i_rcu, proc_i_callback);
95 static void init_once(void *foo)
97 struct proc_inode *ei = (struct proc_inode *) foo;
99 inode_init_once(&ei->vfs_inode);
102 void __init proc_init_inodecache(void)
104 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
105 sizeof(struct proc_inode),
106 0, (SLAB_RECLAIM_ACCOUNT|
107 SLAB_MEM_SPREAD|SLAB_PANIC),
111 static int proc_show_options(struct seq_file *seq, struct dentry *root)
113 struct super_block *sb = root->d_sb;
114 struct pid_namespace *pid = sb->s_fs_info;
116 if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
117 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
118 if (pid->hide_pid != 0)
119 seq_printf(seq, ",hidepid=%u", pid->hide_pid);
124 static const struct super_operations proc_sops = {
125 .alloc_inode = proc_alloc_inode,
126 .destroy_inode = proc_destroy_inode,
127 .drop_inode = generic_delete_inode,
128 .evict_inode = proc_evict_inode,
129 .statfs = simple_statfs,
130 .remount_fs = proc_remount,
131 .show_options = proc_show_options,
134 enum {BIAS = -1U<<31};
136 static inline int use_pde(struct proc_dir_entry *pde)
138 return atomic_inc_unless_negative(&pde->in_use);
141 static void unuse_pde(struct proc_dir_entry *pde)
143 if (atomic_dec_return(&pde->in_use) == BIAS)
144 complete(pde->pde_unload_completion);
148 static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
151 /* somebody else is doing that, just wait */
152 DECLARE_COMPLETION_ONSTACK(c);
154 spin_unlock(&pde->pde_unload_lock);
155 wait_for_completion(&c);
156 spin_lock(&pde->pde_unload_lock);
160 spin_unlock(&pde->pde_unload_lock);
162 pde->proc_fops->release(file_inode(file), file);
163 spin_lock(&pde->pde_unload_lock);
164 list_del_init(&pdeo->lh);
171 void proc_entry_rundown(struct proc_dir_entry *de)
173 DECLARE_COMPLETION_ONSTACK(c);
174 /* Wait until all existing callers into module are done. */
175 de->pde_unload_completion = &c;
176 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
177 wait_for_completion(&c);
179 spin_lock(&de->pde_unload_lock);
180 while (!list_empty(&de->pde_openers)) {
181 struct pde_opener *pdeo;
182 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
183 close_pdeo(de, pdeo);
185 spin_unlock(&de->pde_unload_lock);
188 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
190 struct proc_dir_entry *pde = PDE(file_inode(file));
193 loff_t (*llseek)(struct file *, loff_t, int);
194 llseek = pde->proc_fops->llseek;
196 llseek = default_llseek;
197 rv = llseek(file, offset, whence);
203 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
205 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
206 struct proc_dir_entry *pde = PDE(file_inode(file));
209 read = pde->proc_fops->read;
211 rv = read(file, buf, count, ppos);
217 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
219 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
220 struct proc_dir_entry *pde = PDE(file_inode(file));
223 write = pde->proc_fops->write;
225 rv = write(file, buf, count, ppos);
231 static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
233 struct proc_dir_entry *pde = PDE(file_inode(file));
234 unsigned int rv = DEFAULT_POLLMASK;
235 unsigned int (*poll)(struct file *, struct poll_table_struct *);
237 poll = pde->proc_fops->poll;
239 rv = poll(file, pts);
245 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
247 struct proc_dir_entry *pde = PDE(file_inode(file));
249 long (*ioctl)(struct file *, unsigned int, unsigned long);
251 ioctl = pde->proc_fops->unlocked_ioctl;
253 rv = ioctl(file, cmd, arg);
260 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
262 struct proc_dir_entry *pde = PDE(file_inode(file));
264 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
266 compat_ioctl = pde->proc_fops->compat_ioctl;
268 rv = compat_ioctl(file, cmd, arg);
275 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
277 struct proc_dir_entry *pde = PDE(file_inode(file));
279 int (*mmap)(struct file *, struct vm_area_struct *);
281 mmap = pde->proc_fops->mmap;
283 rv = mmap(file, vma);
289 static int proc_reg_open(struct inode *inode, struct file *file)
291 struct proc_dir_entry *pde = PDE(inode);
293 int (*open)(struct inode *, struct file *);
294 int (*release)(struct inode *, struct file *);
295 struct pde_opener *pdeo;
298 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
299 * sequence. ->release won't be called because ->proc_fops will be
300 * cleared. Depending on complexity of ->release, consequences vary.
302 * We can't wait for mercy when close will be done for real, it's
303 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
304 * by hand in remove_proc_entry(). For this, save opener's credentials
307 pdeo = kzalloc(sizeof(struct pde_opener), GFP_KERNEL);
315 open = pde->proc_fops->open;
316 release = pde->proc_fops->release;
319 rv = open(inode, file);
321 if (rv == 0 && release) {
322 /* To know what to release. */
324 /* Strictly for "too late" ->release in proc_reg_release(). */
325 spin_lock(&pde->pde_unload_lock);
326 list_add(&pdeo->lh, &pde->pde_openers);
327 spin_unlock(&pde->pde_unload_lock);
335 static int proc_reg_release(struct inode *inode, struct file *file)
337 struct proc_dir_entry *pde = PDE(inode);
338 struct pde_opener *pdeo;
339 spin_lock(&pde->pde_unload_lock);
340 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
341 if (pdeo->file == file) {
342 close_pdeo(pde, pdeo);
346 spin_unlock(&pde->pde_unload_lock);
350 static const struct file_operations proc_reg_file_ops = {
351 .llseek = proc_reg_llseek,
352 .read = proc_reg_read,
353 .write = proc_reg_write,
354 .poll = proc_reg_poll,
355 .unlocked_ioctl = proc_reg_unlocked_ioctl,
357 .compat_ioctl = proc_reg_compat_ioctl,
359 .mmap = proc_reg_mmap,
360 .open = proc_reg_open,
361 .release = proc_reg_release,
365 static const struct file_operations proc_reg_file_ops_no_compat = {
366 .llseek = proc_reg_llseek,
367 .read = proc_reg_read,
368 .write = proc_reg_write,
369 .poll = proc_reg_poll,
370 .unlocked_ioctl = proc_reg_unlocked_ioctl,
371 .mmap = proc_reg_mmap,
372 .open = proc_reg_open,
373 .release = proc_reg_release,
377 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
379 struct proc_dir_entry *pde = PDE(dentry->d_inode);
380 if (unlikely(!use_pde(pde)))
381 return ERR_PTR(-EINVAL);
382 nd_set_link(nd, pde->data);
386 static void proc_put_link(struct dentry *dentry, struct nameidata *nd, void *p)
391 const struct inode_operations proc_link_inode_operations = {
392 .readlink = generic_readlink,
393 .follow_link = proc_follow_link,
394 .put_link = proc_put_link,
397 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
399 struct inode *inode = new_inode_pseudo(sb);
402 inode->i_ino = de->low_ino;
403 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
404 PROC_I(inode)->pde = de;
407 inode->i_mode = de->mode;
408 inode->i_uid = de->uid;
409 inode->i_gid = de->gid;
412 inode->i_size = de->size;
414 set_nlink(inode, de->nlink);
415 WARN_ON(!de->proc_iops);
416 inode->i_op = de->proc_iops;
418 if (S_ISREG(inode->i_mode)) {
420 if (!de->proc_fops->compat_ioctl)
422 &proc_reg_file_ops_no_compat;
425 inode->i_fop = &proc_reg_file_ops;
427 inode->i_fop = de->proc_fops;
435 int proc_fill_super(struct super_block *s)
437 struct inode *root_inode;
439 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
440 s->s_blocksize = 1024;
441 s->s_blocksize_bits = 10;
442 s->s_magic = PROC_SUPER_MAGIC;
443 s->s_op = &proc_sops;
447 root_inode = proc_get_inode(s, &proc_root);
449 pr_err("proc_fill_super: get root inode failed\n");
453 s->s_root = d_make_root(root_inode);
455 pr_err("proc_fill_super: allocate dentry failed\n");
459 return proc_setup_self(s);