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/sysctl.h>
20 #include <linux/slab.h>
22 #include <asm/system.h>
23 #include <asm/uaccess.h>
27 static void proc_evict_inode(struct inode *inode)
29 struct proc_dir_entry *de;
30 struct ctl_table_header *head;
31 const struct proc_ns_operations *ns_ops;
33 truncate_inode_pages(&inode->i_data, 0);
36 /* Stop tracking associated processes */
37 put_pid(PROC_I(inode)->pid);
39 /* Let go of any associated proc directory entry */
40 de = PROC_I(inode)->pde;
43 head = PROC_I(inode)->sysctl;
45 rcu_assign_pointer(PROC_I(inode)->sysctl, NULL);
46 sysctl_head_put(head);
48 /* Release any associated namespace */
49 ns_ops = PROC_I(inode)->ns_ops;
50 if (ns_ops && ns_ops->put)
51 ns_ops->put(PROC_I(inode)->ns);
54 static struct kmem_cache * proc_inode_cachep;
56 static struct inode *proc_alloc_inode(struct super_block *sb)
58 struct proc_inode *ei;
61 ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
66 ei->op.proc_get_link = NULL;
69 ei->sysctl_entry = NULL;
72 inode = &ei->vfs_inode;
73 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
77 static void proc_i_callback(struct rcu_head *head)
79 struct inode *inode = container_of(head, struct inode, i_rcu);
80 INIT_LIST_HEAD(&inode->i_dentry);
81 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
84 static void proc_destroy_inode(struct inode *inode)
86 call_rcu(&inode->i_rcu, proc_i_callback);
89 static void init_once(void *foo)
91 struct proc_inode *ei = (struct proc_inode *) foo;
93 inode_init_once(&ei->vfs_inode);
96 void __init proc_init_inodecache(void)
98 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
99 sizeof(struct proc_inode),
100 0, (SLAB_RECLAIM_ACCOUNT|
101 SLAB_MEM_SPREAD|SLAB_PANIC),
105 static const struct super_operations proc_sops = {
106 .alloc_inode = proc_alloc_inode,
107 .destroy_inode = proc_destroy_inode,
108 .drop_inode = generic_delete_inode,
109 .evict_inode = proc_evict_inode,
110 .statfs = simple_statfs,
113 static void __pde_users_dec(struct proc_dir_entry *pde)
116 if (pde->pde_unload_completion && pde->pde_users == 0)
117 complete(pde->pde_unload_completion);
120 void pde_users_dec(struct proc_dir_entry *pde)
122 spin_lock(&pde->pde_unload_lock);
123 __pde_users_dec(pde);
124 spin_unlock(&pde->pde_unload_lock);
127 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
129 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
131 loff_t (*llseek)(struct file *, loff_t, int);
133 spin_lock(&pde->pde_unload_lock);
135 * remove_proc_entry() is going to delete PDE (as part of module
136 * cleanup sequence). No new callers into module allowed.
138 if (!pde->proc_fops) {
139 spin_unlock(&pde->pde_unload_lock);
143 * Bump refcount so that remove_proc_entry will wail for ->llseek to
148 * Save function pointer under lock, to protect against ->proc_fops
149 * NULL'ifying right after ->pde_unload_lock is dropped.
151 llseek = pde->proc_fops->llseek;
152 spin_unlock(&pde->pde_unload_lock);
155 llseek = default_llseek;
156 rv = llseek(file, offset, whence);
162 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
164 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
166 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
168 spin_lock(&pde->pde_unload_lock);
169 if (!pde->proc_fops) {
170 spin_unlock(&pde->pde_unload_lock);
174 read = pde->proc_fops->read;
175 spin_unlock(&pde->pde_unload_lock);
178 rv = read(file, buf, count, ppos);
184 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
186 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
188 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
190 spin_lock(&pde->pde_unload_lock);
191 if (!pde->proc_fops) {
192 spin_unlock(&pde->pde_unload_lock);
196 write = pde->proc_fops->write;
197 spin_unlock(&pde->pde_unload_lock);
200 rv = write(file, buf, count, ppos);
206 static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
208 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
209 unsigned int rv = DEFAULT_POLLMASK;
210 unsigned int (*poll)(struct file *, struct poll_table_struct *);
212 spin_lock(&pde->pde_unload_lock);
213 if (!pde->proc_fops) {
214 spin_unlock(&pde->pde_unload_lock);
218 poll = pde->proc_fops->poll;
219 spin_unlock(&pde->pde_unload_lock);
222 rv = poll(file, pts);
228 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
230 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
232 long (*ioctl)(struct file *, unsigned int, unsigned long);
234 spin_lock(&pde->pde_unload_lock);
235 if (!pde->proc_fops) {
236 spin_unlock(&pde->pde_unload_lock);
240 ioctl = pde->proc_fops->unlocked_ioctl;
241 spin_unlock(&pde->pde_unload_lock);
244 rv = ioctl(file, cmd, arg);
251 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
253 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
255 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
257 spin_lock(&pde->pde_unload_lock);
258 if (!pde->proc_fops) {
259 spin_unlock(&pde->pde_unload_lock);
263 compat_ioctl = pde->proc_fops->compat_ioctl;
264 spin_unlock(&pde->pde_unload_lock);
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->f_path.dentry->d_inode);
278 int (*mmap)(struct file *, struct vm_area_struct *);
280 spin_lock(&pde->pde_unload_lock);
281 if (!pde->proc_fops) {
282 spin_unlock(&pde->pde_unload_lock);
286 mmap = pde->proc_fops->mmap;
287 spin_unlock(&pde->pde_unload_lock);
290 rv = mmap(file, vma);
296 static int proc_reg_open(struct inode *inode, struct file *file)
298 struct proc_dir_entry *pde = PDE(inode);
300 int (*open)(struct inode *, struct file *);
301 int (*release)(struct inode *, struct file *);
302 struct pde_opener *pdeo;
305 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
306 * sequence. ->release won't be called because ->proc_fops will be
307 * cleared. Depending on complexity of ->release, consequences vary.
309 * We can't wait for mercy when close will be done for real, it's
310 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
311 * by hand in remove_proc_entry(). For this, save opener's credentials
314 pdeo = kmalloc(sizeof(struct pde_opener), GFP_KERNEL);
318 spin_lock(&pde->pde_unload_lock);
319 if (!pde->proc_fops) {
320 spin_unlock(&pde->pde_unload_lock);
325 open = pde->proc_fops->open;
326 release = pde->proc_fops->release;
327 spin_unlock(&pde->pde_unload_lock);
330 rv = open(inode, file);
332 spin_lock(&pde->pde_unload_lock);
333 if (rv == 0 && release) {
334 /* To know what to release. */
337 /* Strictly for "too late" ->release in proc_reg_release(). */
338 pdeo->release = release;
339 list_add(&pdeo->lh, &pde->pde_openers);
342 __pde_users_dec(pde);
343 spin_unlock(&pde->pde_unload_lock);
347 static struct pde_opener *find_pde_opener(struct proc_dir_entry *pde,
348 struct inode *inode, struct file *file)
350 struct pde_opener *pdeo;
352 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
353 if (pdeo->inode == inode && pdeo->file == file)
359 static int proc_reg_release(struct inode *inode, struct file *file)
361 struct proc_dir_entry *pde = PDE(inode);
363 int (*release)(struct inode *, struct file *);
364 struct pde_opener *pdeo;
366 spin_lock(&pde->pde_unload_lock);
367 pdeo = find_pde_opener(pde, inode, file);
368 if (!pde->proc_fops) {
370 * Can't simply exit, __fput() will think that everything is OK,
371 * and move on to freeing struct file. remove_proc_entry() will
372 * find slacker in opener's list and will try to do non-trivial
373 * things with struct file. Therefore, remove opener from list.
375 * But if opener is removed from list, who will ->release it?
379 spin_unlock(&pde->pde_unload_lock);
380 rv = pdeo->release(inode, file);
383 spin_unlock(&pde->pde_unload_lock);
387 release = pde->proc_fops->release;
392 spin_unlock(&pde->pde_unload_lock);
395 rv = release(inode, file);
401 static const struct file_operations proc_reg_file_ops = {
402 .llseek = proc_reg_llseek,
403 .read = proc_reg_read,
404 .write = proc_reg_write,
405 .poll = proc_reg_poll,
406 .unlocked_ioctl = proc_reg_unlocked_ioctl,
408 .compat_ioctl = proc_reg_compat_ioctl,
410 .mmap = proc_reg_mmap,
411 .open = proc_reg_open,
412 .release = proc_reg_release,
416 static const struct file_operations proc_reg_file_ops_no_compat = {
417 .llseek = proc_reg_llseek,
418 .read = proc_reg_read,
419 .write = proc_reg_write,
420 .poll = proc_reg_poll,
421 .unlocked_ioctl = proc_reg_unlocked_ioctl,
422 .mmap = proc_reg_mmap,
423 .open = proc_reg_open,
424 .release = proc_reg_release,
428 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
430 struct inode * inode;
432 inode = iget_locked(sb, de->low_ino);
435 if (inode->i_state & I_NEW) {
436 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
437 PROC_I(inode)->fd = 0;
438 PROC_I(inode)->pde = de;
441 inode->i_mode = de->mode;
442 inode->i_uid = de->uid;
443 inode->i_gid = de->gid;
446 inode->i_size = de->size;
448 set_nlink(inode, de->nlink);
450 inode->i_op = de->proc_iops;
452 if (S_ISREG(inode->i_mode)) {
454 if (!de->proc_fops->compat_ioctl)
456 &proc_reg_file_ops_no_compat;
459 inode->i_fop = &proc_reg_file_ops;
461 inode->i_fop = de->proc_fops;
464 unlock_new_inode(inode);
470 int proc_fill_super(struct super_block *s)
472 struct inode * root_inode;
474 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
475 s->s_blocksize = 1024;
476 s->s_blocksize_bits = 10;
477 s->s_magic = PROC_SUPER_MAGIC;
478 s->s_op = &proc_sops;
482 root_inode = proc_get_inode(s, &proc_root);
485 root_inode->i_uid = 0;
486 root_inode->i_gid = 0;
487 s->s_root = d_alloc_root(root_inode);
493 printk("proc_read_super: get root inode failed\n");