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>
11 #include <linux/string.h>
12 #include <linux/stat.h>
13 #include <linux/completion.h>
14 #include <linux/poll.h>
15 #include <linux/file.h>
16 #include <linux/limits.h>
17 #include <linux/init.h>
18 #include <linux/module.h>
19 #include <linux/smp_lock.h>
20 #include <linux/sysctl.h>
22 #include <asm/system.h>
23 #include <asm/uaccess.h>
27 struct proc_dir_entry *de_get(struct proc_dir_entry *de)
29 atomic_inc(&de->count);
34 * Decrements the use count and checks for deferred deletion.
36 void de_put(struct proc_dir_entry *de)
38 if (!atomic_read(&de->count)) {
39 printk("de_put: entry %s already free!\n", de->name);
43 if (atomic_dec_and_test(&de->count))
48 * Decrement the use count of the proc_dir_entry.
50 static void proc_delete_inode(struct inode *inode)
52 struct proc_dir_entry *de;
54 truncate_inode_pages(&inode->i_data, 0);
56 /* Stop tracking associated processes */
57 put_pid(PROC_I(inode)->pid);
59 /* Let go of any associated proc directory entry */
60 de = PROC_I(inode)->pde;
63 if (PROC_I(inode)->sysctl)
64 sysctl_head_put(PROC_I(inode)->sysctl);
68 struct vfsmount *proc_mnt;
70 static struct kmem_cache * proc_inode_cachep;
72 static struct inode *proc_alloc_inode(struct super_block *sb)
74 struct proc_inode *ei;
77 ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
82 ei->op.proc_get_link = NULL;
85 ei->sysctl_entry = NULL;
86 inode = &ei->vfs_inode;
87 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
91 static void proc_destroy_inode(struct inode *inode)
93 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
96 static void init_once(void *foo)
98 struct proc_inode *ei = (struct proc_inode *) foo;
100 inode_init_once(&ei->vfs_inode);
103 void __init proc_init_inodecache(void)
105 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
106 sizeof(struct proc_inode),
107 0, (SLAB_RECLAIM_ACCOUNT|
108 SLAB_MEM_SPREAD|SLAB_PANIC),
112 static const struct super_operations proc_sops = {
113 .alloc_inode = proc_alloc_inode,
114 .destroy_inode = proc_destroy_inode,
115 .drop_inode = generic_delete_inode,
116 .delete_inode = proc_delete_inode,
117 .statfs = simple_statfs,
120 static void __pde_users_dec(struct proc_dir_entry *pde)
123 if (pde->pde_unload_completion && pde->pde_users == 0)
124 complete(pde->pde_unload_completion);
127 void pde_users_dec(struct proc_dir_entry *pde)
129 spin_lock(&pde->pde_unload_lock);
130 __pde_users_dec(pde);
131 spin_unlock(&pde->pde_unload_lock);
134 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
136 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
138 loff_t (*llseek)(struct file *, loff_t, int);
140 spin_lock(&pde->pde_unload_lock);
142 * remove_proc_entry() is going to delete PDE (as part of module
143 * cleanup sequence). No new callers into module allowed.
145 if (!pde->proc_fops) {
146 spin_unlock(&pde->pde_unload_lock);
150 * Bump refcount so that remove_proc_entry will wail for ->llseek to
155 * Save function pointer under lock, to protect against ->proc_fops
156 * NULL'ifying right after ->pde_unload_lock is dropped.
158 llseek = pde->proc_fops->llseek;
159 spin_unlock(&pde->pde_unload_lock);
162 llseek = default_llseek;
163 rv = llseek(file, offset, whence);
169 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
171 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
173 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
175 spin_lock(&pde->pde_unload_lock);
176 if (!pde->proc_fops) {
177 spin_unlock(&pde->pde_unload_lock);
181 read = pde->proc_fops->read;
182 spin_unlock(&pde->pde_unload_lock);
185 rv = read(file, buf, count, ppos);
191 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
193 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
195 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
197 spin_lock(&pde->pde_unload_lock);
198 if (!pde->proc_fops) {
199 spin_unlock(&pde->pde_unload_lock);
203 write = pde->proc_fops->write;
204 spin_unlock(&pde->pde_unload_lock);
207 rv = write(file, buf, count, ppos);
213 static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
215 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
216 unsigned int rv = DEFAULT_POLLMASK;
217 unsigned int (*poll)(struct file *, struct poll_table_struct *);
219 spin_lock(&pde->pde_unload_lock);
220 if (!pde->proc_fops) {
221 spin_unlock(&pde->pde_unload_lock);
225 poll = pde->proc_fops->poll;
226 spin_unlock(&pde->pde_unload_lock);
229 rv = poll(file, pts);
235 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
237 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
239 long (*unlocked_ioctl)(struct file *, unsigned int, unsigned long);
240 int (*ioctl)(struct inode *, struct file *, unsigned int, unsigned long);
242 spin_lock(&pde->pde_unload_lock);
243 if (!pde->proc_fops) {
244 spin_unlock(&pde->pde_unload_lock);
248 unlocked_ioctl = pde->proc_fops->unlocked_ioctl;
249 ioctl = pde->proc_fops->ioctl;
250 spin_unlock(&pde->pde_unload_lock);
252 if (unlocked_ioctl) {
253 rv = unlocked_ioctl(file, cmd, arg);
254 if (rv == -ENOIOCTLCMD)
258 rv = ioctl(file->f_path.dentry->d_inode, file, cmd, arg);
267 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
269 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
271 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
273 spin_lock(&pde->pde_unload_lock);
274 if (!pde->proc_fops) {
275 spin_unlock(&pde->pde_unload_lock);
279 compat_ioctl = pde->proc_fops->compat_ioctl;
280 spin_unlock(&pde->pde_unload_lock);
283 rv = compat_ioctl(file, cmd, arg);
290 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
292 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
294 int (*mmap)(struct file *, struct vm_area_struct *);
296 spin_lock(&pde->pde_unload_lock);
297 if (!pde->proc_fops) {
298 spin_unlock(&pde->pde_unload_lock);
302 mmap = pde->proc_fops->mmap;
303 spin_unlock(&pde->pde_unload_lock);
306 rv = mmap(file, vma);
312 static int proc_reg_open(struct inode *inode, struct file *file)
314 struct proc_dir_entry *pde = PDE(inode);
316 int (*open)(struct inode *, struct file *);
317 int (*release)(struct inode *, struct file *);
318 struct pde_opener *pdeo;
321 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
322 * sequence. ->release won't be called because ->proc_fops will be
323 * cleared. Depending on complexity of ->release, consequences vary.
325 * We can't wait for mercy when close will be done for real, it's
326 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
327 * by hand in remove_proc_entry(). For this, save opener's credentials
330 pdeo = kmalloc(sizeof(struct pde_opener), GFP_KERNEL);
334 spin_lock(&pde->pde_unload_lock);
335 if (!pde->proc_fops) {
336 spin_unlock(&pde->pde_unload_lock);
341 open = pde->proc_fops->open;
342 release = pde->proc_fops->release;
343 spin_unlock(&pde->pde_unload_lock);
346 rv = open(inode, file);
348 spin_lock(&pde->pde_unload_lock);
349 if (rv == 0 && release) {
350 /* To know what to release. */
353 /* Strictly for "too late" ->release in proc_reg_release(). */
354 pdeo->release = release;
355 list_add(&pdeo->lh, &pde->pde_openers);
358 __pde_users_dec(pde);
359 spin_unlock(&pde->pde_unload_lock);
363 static struct pde_opener *find_pde_opener(struct proc_dir_entry *pde,
364 struct inode *inode, struct file *file)
366 struct pde_opener *pdeo;
368 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
369 if (pdeo->inode == inode && pdeo->file == file)
375 static int proc_reg_release(struct inode *inode, struct file *file)
377 struct proc_dir_entry *pde = PDE(inode);
379 int (*release)(struct inode *, struct file *);
380 struct pde_opener *pdeo;
382 spin_lock(&pde->pde_unload_lock);
383 pdeo = find_pde_opener(pde, inode, file);
384 if (!pde->proc_fops) {
386 * Can't simply exit, __fput() will think that everything is OK,
387 * and move on to freeing struct file. remove_proc_entry() will
388 * find slacker in opener's list and will try to do non-trivial
389 * things with struct file. Therefore, remove opener from list.
391 * But if opener is removed from list, who will ->release it?
395 spin_unlock(&pde->pde_unload_lock);
396 rv = pdeo->release(inode, file);
399 spin_unlock(&pde->pde_unload_lock);
403 release = pde->proc_fops->release;
408 spin_unlock(&pde->pde_unload_lock);
411 rv = release(inode, file);
417 static const struct file_operations proc_reg_file_ops = {
418 .llseek = proc_reg_llseek,
419 .read = proc_reg_read,
420 .write = proc_reg_write,
421 .poll = proc_reg_poll,
422 .unlocked_ioctl = proc_reg_unlocked_ioctl,
424 .compat_ioctl = proc_reg_compat_ioctl,
426 .mmap = proc_reg_mmap,
427 .open = proc_reg_open,
428 .release = proc_reg_release,
432 static const struct file_operations proc_reg_file_ops_no_compat = {
433 .llseek = proc_reg_llseek,
434 .read = proc_reg_read,
435 .write = proc_reg_write,
436 .poll = proc_reg_poll,
437 .unlocked_ioctl = proc_reg_unlocked_ioctl,
438 .mmap = proc_reg_mmap,
439 .open = proc_reg_open,
440 .release = proc_reg_release,
444 struct inode *proc_get_inode(struct super_block *sb, unsigned int ino,
445 struct proc_dir_entry *de)
447 struct inode * inode;
449 inode = iget_locked(sb, ino);
452 if (inode->i_state & I_NEW) {
453 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
454 PROC_I(inode)->fd = 0;
455 PROC_I(inode)->pde = de;
458 inode->i_mode = de->mode;
459 inode->i_uid = de->uid;
460 inode->i_gid = de->gid;
463 inode->i_size = de->size;
465 inode->i_nlink = de->nlink;
467 inode->i_op = de->proc_iops;
469 if (S_ISREG(inode->i_mode)) {
471 if (!de->proc_fops->compat_ioctl)
473 &proc_reg_file_ops_no_compat;
476 inode->i_fop = &proc_reg_file_ops;
478 inode->i_fop = de->proc_fops;
481 unlock_new_inode(inode);
487 int proc_fill_super(struct super_block *s)
489 struct inode * root_inode;
491 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
492 s->s_blocksize = 1024;
493 s->s_blocksize_bits = 10;
494 s->s_magic = PROC_SUPER_MAGIC;
495 s->s_op = &proc_sops;
499 root_inode = proc_get_inode(s, PROC_ROOT_INO, &proc_root);
502 root_inode->i_uid = 0;
503 root_inode->i_gid = 0;
504 s->s_root = d_alloc_root(root_inode);
510 printk("proc_read_super: get root inode failed\n");