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/file.h>
17 #include <linux/limits.h>
18 #include <linux/init.h>
19 #include <linux/module.h>
20 #include <linux/sysctl.h>
21 #include <linux/seq_file.h>
22 #include <linux/slab.h>
23 #include <linux/mount.h>
25 #include <asm/system.h>
26 #include <asm/uaccess.h>
30 static void proc_evict_inode(struct inode *inode)
32 struct proc_dir_entry *de;
33 struct ctl_table_header *head;
34 const struct proc_ns_operations *ns_ops;
37 truncate_inode_pages(&inode->i_data, 0);
40 /* Stop tracking associated processes */
41 put_pid(PROC_I(inode)->pid);
43 /* Let go of any associated proc directory entry */
44 de = PROC_I(inode)->pde;
47 head = PROC_I(inode)->sysctl;
49 rcu_assign_pointer(PROC_I(inode)->sysctl, NULL);
50 sysctl_head_put(head);
52 /* Release any associated namespace */
53 ns_ops = PROC_I(inode)->ns_ops;
54 ns = PROC_I(inode)->ns;
59 static struct kmem_cache * proc_inode_cachep;
61 static struct inode *proc_alloc_inode(struct super_block *sb)
63 struct proc_inode *ei;
66 ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
71 ei->op.proc_get_link = NULL;
74 ei->sysctl_entry = NULL;
77 inode = &ei->vfs_inode;
78 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
82 static void proc_i_callback(struct rcu_head *head)
84 struct inode *inode = container_of(head, struct inode, i_rcu);
85 INIT_LIST_HEAD(&inode->i_dentry);
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 vfsmount *vfs)
115 static const struct super_operations proc_sops = {
116 .alloc_inode = proc_alloc_inode,
117 .destroy_inode = proc_destroy_inode,
118 .drop_inode = generic_delete_inode,
119 .evict_inode = proc_evict_inode,
120 .statfs = simple_statfs,
121 .remount_fs = proc_remount,
122 .show_options = proc_show_options,
125 static void __pde_users_dec(struct proc_dir_entry *pde)
128 if (pde->pde_unload_completion && pde->pde_users == 0)
129 complete(pde->pde_unload_completion);
132 void pde_users_dec(struct proc_dir_entry *pde)
134 spin_lock(&pde->pde_unload_lock);
135 __pde_users_dec(pde);
136 spin_unlock(&pde->pde_unload_lock);
139 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
141 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
143 loff_t (*llseek)(struct file *, loff_t, int);
145 spin_lock(&pde->pde_unload_lock);
147 * remove_proc_entry() is going to delete PDE (as part of module
148 * cleanup sequence). No new callers into module allowed.
150 if (!pde->proc_fops) {
151 spin_unlock(&pde->pde_unload_lock);
155 * Bump refcount so that remove_proc_entry will wail for ->llseek to
160 * Save function pointer under lock, to protect against ->proc_fops
161 * NULL'ifying right after ->pde_unload_lock is dropped.
163 llseek = pde->proc_fops->llseek;
164 spin_unlock(&pde->pde_unload_lock);
167 llseek = default_llseek;
168 rv = llseek(file, offset, whence);
174 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
176 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
178 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
180 spin_lock(&pde->pde_unload_lock);
181 if (!pde->proc_fops) {
182 spin_unlock(&pde->pde_unload_lock);
186 read = pde->proc_fops->read;
187 spin_unlock(&pde->pde_unload_lock);
190 rv = read(file, buf, count, ppos);
196 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
198 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
200 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
202 spin_lock(&pde->pde_unload_lock);
203 if (!pde->proc_fops) {
204 spin_unlock(&pde->pde_unload_lock);
208 write = pde->proc_fops->write;
209 spin_unlock(&pde->pde_unload_lock);
212 rv = write(file, buf, count, ppos);
218 static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
220 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
221 unsigned int rv = DEFAULT_POLLMASK;
222 unsigned int (*poll)(struct file *, struct poll_table_struct *);
224 spin_lock(&pde->pde_unload_lock);
225 if (!pde->proc_fops) {
226 spin_unlock(&pde->pde_unload_lock);
230 poll = pde->proc_fops->poll;
231 spin_unlock(&pde->pde_unload_lock);
234 rv = poll(file, pts);
240 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
242 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
244 long (*ioctl)(struct file *, unsigned int, unsigned long);
246 spin_lock(&pde->pde_unload_lock);
247 if (!pde->proc_fops) {
248 spin_unlock(&pde->pde_unload_lock);
252 ioctl = pde->proc_fops->unlocked_ioctl;
253 spin_unlock(&pde->pde_unload_lock);
256 rv = ioctl(file, cmd, arg);
263 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
265 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
267 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
269 spin_lock(&pde->pde_unload_lock);
270 if (!pde->proc_fops) {
271 spin_unlock(&pde->pde_unload_lock);
275 compat_ioctl = pde->proc_fops->compat_ioctl;
276 spin_unlock(&pde->pde_unload_lock);
279 rv = compat_ioctl(file, cmd, arg);
286 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
288 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
290 int (*mmap)(struct file *, struct vm_area_struct *);
292 spin_lock(&pde->pde_unload_lock);
293 if (!pde->proc_fops) {
294 spin_unlock(&pde->pde_unload_lock);
298 mmap = pde->proc_fops->mmap;
299 spin_unlock(&pde->pde_unload_lock);
302 rv = mmap(file, vma);
308 static int proc_reg_open(struct inode *inode, struct file *file)
310 struct proc_dir_entry *pde = PDE(inode);
312 int (*open)(struct inode *, struct file *);
313 int (*release)(struct inode *, struct file *);
314 struct pde_opener *pdeo;
317 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
318 * sequence. ->release won't be called because ->proc_fops will be
319 * cleared. Depending on complexity of ->release, consequences vary.
321 * We can't wait for mercy when close will be done for real, it's
322 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
323 * by hand in remove_proc_entry(). For this, save opener's credentials
326 pdeo = kmalloc(sizeof(struct pde_opener), GFP_KERNEL);
330 spin_lock(&pde->pde_unload_lock);
331 if (!pde->proc_fops) {
332 spin_unlock(&pde->pde_unload_lock);
337 open = pde->proc_fops->open;
338 release = pde->proc_fops->release;
339 spin_unlock(&pde->pde_unload_lock);
342 rv = open(inode, file);
344 spin_lock(&pde->pde_unload_lock);
345 if (rv == 0 && release) {
346 /* To know what to release. */
349 /* Strictly for "too late" ->release in proc_reg_release(). */
350 pdeo->release = release;
351 list_add(&pdeo->lh, &pde->pde_openers);
354 __pde_users_dec(pde);
355 spin_unlock(&pde->pde_unload_lock);
359 static struct pde_opener *find_pde_opener(struct proc_dir_entry *pde,
360 struct inode *inode, struct file *file)
362 struct pde_opener *pdeo;
364 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
365 if (pdeo->inode == inode && pdeo->file == file)
371 static int proc_reg_release(struct inode *inode, struct file *file)
373 struct proc_dir_entry *pde = PDE(inode);
375 int (*release)(struct inode *, struct file *);
376 struct pde_opener *pdeo;
378 spin_lock(&pde->pde_unload_lock);
379 pdeo = find_pde_opener(pde, inode, file);
380 if (!pde->proc_fops) {
382 * Can't simply exit, __fput() will think that everything is OK,
383 * and move on to freeing struct file. remove_proc_entry() will
384 * find slacker in opener's list and will try to do non-trivial
385 * things with struct file. Therefore, remove opener from list.
387 * But if opener is removed from list, who will ->release it?
391 spin_unlock(&pde->pde_unload_lock);
392 rv = pdeo->release(inode, file);
395 spin_unlock(&pde->pde_unload_lock);
399 release = pde->proc_fops->release;
404 spin_unlock(&pde->pde_unload_lock);
407 rv = release(inode, file);
413 static const struct file_operations proc_reg_file_ops = {
414 .llseek = proc_reg_llseek,
415 .read = proc_reg_read,
416 .write = proc_reg_write,
417 .poll = proc_reg_poll,
418 .unlocked_ioctl = proc_reg_unlocked_ioctl,
420 .compat_ioctl = proc_reg_compat_ioctl,
422 .mmap = proc_reg_mmap,
423 .open = proc_reg_open,
424 .release = proc_reg_release,
428 static const struct file_operations proc_reg_file_ops_no_compat = {
429 .llseek = proc_reg_llseek,
430 .read = proc_reg_read,
431 .write = proc_reg_write,
432 .poll = proc_reg_poll,
433 .unlocked_ioctl = proc_reg_unlocked_ioctl,
434 .mmap = proc_reg_mmap,
435 .open = proc_reg_open,
436 .release = proc_reg_release,
440 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
442 struct inode * inode;
444 inode = iget_locked(sb, de->low_ino);
447 if (inode->i_state & I_NEW) {
448 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
449 PROC_I(inode)->fd = 0;
450 PROC_I(inode)->pde = de;
453 inode->i_mode = de->mode;
454 inode->i_uid = de->uid;
455 inode->i_gid = de->gid;
458 inode->i_size = de->size;
460 set_nlink(inode, de->nlink);
462 inode->i_op = de->proc_iops;
464 if (S_ISREG(inode->i_mode)) {
466 if (!de->proc_fops->compat_ioctl)
468 &proc_reg_file_ops_no_compat;
471 inode->i_fop = &proc_reg_file_ops;
473 inode->i_fop = de->proc_fops;
476 unlock_new_inode(inode);
482 int proc_fill_super(struct super_block *s)
484 struct inode * root_inode;
486 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
487 s->s_blocksize = 1024;
488 s->s_blocksize_bits = 10;
489 s->s_magic = PROC_SUPER_MAGIC;
490 s->s_op = &proc_sops;
494 root_inode = proc_get_inode(s, &proc_root);
497 root_inode->i_uid = 0;
498 root_inode->i_gid = 0;
499 s->s_root = d_alloc_root(root_inode);
505 printk("proc_read_super: get root inode failed\n");