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)
112 struct super_block *sb = vfs->mnt_sb;
113 struct pid_namespace *pid = sb->s_fs_info;
116 seq_printf(seq, ",gid=%lu", (unsigned long)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 static void __pde_users_dec(struct proc_dir_entry *pde)
136 if (pde->pde_unload_completion && pde->pde_users == 0)
137 complete(pde->pde_unload_completion);
140 void pde_users_dec(struct proc_dir_entry *pde)
142 spin_lock(&pde->pde_unload_lock);
143 __pde_users_dec(pde);
144 spin_unlock(&pde->pde_unload_lock);
147 static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
149 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
151 loff_t (*llseek)(struct file *, loff_t, int);
153 spin_lock(&pde->pde_unload_lock);
155 * remove_proc_entry() is going to delete PDE (as part of module
156 * cleanup sequence). No new callers into module allowed.
158 if (!pde->proc_fops) {
159 spin_unlock(&pde->pde_unload_lock);
163 * Bump refcount so that remove_proc_entry will wail for ->llseek to
168 * Save function pointer under lock, to protect against ->proc_fops
169 * NULL'ifying right after ->pde_unload_lock is dropped.
171 llseek = pde->proc_fops->llseek;
172 spin_unlock(&pde->pde_unload_lock);
175 llseek = default_llseek;
176 rv = llseek(file, offset, whence);
182 static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
184 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
186 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
188 spin_lock(&pde->pde_unload_lock);
189 if (!pde->proc_fops) {
190 spin_unlock(&pde->pde_unload_lock);
194 read = pde->proc_fops->read;
195 spin_unlock(&pde->pde_unload_lock);
198 rv = read(file, buf, count, ppos);
204 static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
206 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
208 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
210 spin_lock(&pde->pde_unload_lock);
211 if (!pde->proc_fops) {
212 spin_unlock(&pde->pde_unload_lock);
216 write = pde->proc_fops->write;
217 spin_unlock(&pde->pde_unload_lock);
220 rv = write(file, buf, count, ppos);
226 static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
228 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
229 unsigned int rv = DEFAULT_POLLMASK;
230 unsigned int (*poll)(struct file *, struct poll_table_struct *);
232 spin_lock(&pde->pde_unload_lock);
233 if (!pde->proc_fops) {
234 spin_unlock(&pde->pde_unload_lock);
238 poll = pde->proc_fops->poll;
239 spin_unlock(&pde->pde_unload_lock);
242 rv = poll(file, pts);
248 static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
250 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
252 long (*ioctl)(struct file *, unsigned int, unsigned long);
254 spin_lock(&pde->pde_unload_lock);
255 if (!pde->proc_fops) {
256 spin_unlock(&pde->pde_unload_lock);
260 ioctl = pde->proc_fops->unlocked_ioctl;
261 spin_unlock(&pde->pde_unload_lock);
264 rv = ioctl(file, cmd, arg);
271 static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
273 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
275 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
277 spin_lock(&pde->pde_unload_lock);
278 if (!pde->proc_fops) {
279 spin_unlock(&pde->pde_unload_lock);
283 compat_ioctl = pde->proc_fops->compat_ioctl;
284 spin_unlock(&pde->pde_unload_lock);
287 rv = compat_ioctl(file, cmd, arg);
294 static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
296 struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
298 int (*mmap)(struct file *, struct vm_area_struct *);
300 spin_lock(&pde->pde_unload_lock);
301 if (!pde->proc_fops) {
302 spin_unlock(&pde->pde_unload_lock);
306 mmap = pde->proc_fops->mmap;
307 spin_unlock(&pde->pde_unload_lock);
310 rv = mmap(file, vma);
316 static int proc_reg_open(struct inode *inode, struct file *file)
318 struct proc_dir_entry *pde = PDE(inode);
320 int (*open)(struct inode *, struct file *);
321 int (*release)(struct inode *, struct file *);
322 struct pde_opener *pdeo;
325 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
326 * sequence. ->release won't be called because ->proc_fops will be
327 * cleared. Depending on complexity of ->release, consequences vary.
329 * We can't wait for mercy when close will be done for real, it's
330 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
331 * by hand in remove_proc_entry(). For this, save opener's credentials
334 pdeo = kmalloc(sizeof(struct pde_opener), GFP_KERNEL);
338 spin_lock(&pde->pde_unload_lock);
339 if (!pde->proc_fops) {
340 spin_unlock(&pde->pde_unload_lock);
345 open = pde->proc_fops->open;
346 release = pde->proc_fops->release;
347 spin_unlock(&pde->pde_unload_lock);
350 rv = open(inode, file);
352 spin_lock(&pde->pde_unload_lock);
353 if (rv == 0 && release) {
354 /* To know what to release. */
357 /* Strictly for "too late" ->release in proc_reg_release(). */
358 pdeo->release = release;
359 list_add(&pdeo->lh, &pde->pde_openers);
362 __pde_users_dec(pde);
363 spin_unlock(&pde->pde_unload_lock);
367 static struct pde_opener *find_pde_opener(struct proc_dir_entry *pde,
368 struct inode *inode, struct file *file)
370 struct pde_opener *pdeo;
372 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
373 if (pdeo->inode == inode && pdeo->file == file)
379 static int proc_reg_release(struct inode *inode, struct file *file)
381 struct proc_dir_entry *pde = PDE(inode);
383 int (*release)(struct inode *, struct file *);
384 struct pde_opener *pdeo;
386 spin_lock(&pde->pde_unload_lock);
387 pdeo = find_pde_opener(pde, inode, file);
388 if (!pde->proc_fops) {
390 * Can't simply exit, __fput() will think that everything is OK,
391 * and move on to freeing struct file. remove_proc_entry() will
392 * find slacker in opener's list and will try to do non-trivial
393 * things with struct file. Therefore, remove opener from list.
395 * But if opener is removed from list, who will ->release it?
399 spin_unlock(&pde->pde_unload_lock);
400 rv = pdeo->release(inode, file);
403 spin_unlock(&pde->pde_unload_lock);
407 release = pde->proc_fops->release;
412 spin_unlock(&pde->pde_unload_lock);
415 rv = release(inode, file);
421 static const struct file_operations proc_reg_file_ops = {
422 .llseek = proc_reg_llseek,
423 .read = proc_reg_read,
424 .write = proc_reg_write,
425 .poll = proc_reg_poll,
426 .unlocked_ioctl = proc_reg_unlocked_ioctl,
428 .compat_ioctl = proc_reg_compat_ioctl,
430 .mmap = proc_reg_mmap,
431 .open = proc_reg_open,
432 .release = proc_reg_release,
436 static const struct file_operations proc_reg_file_ops_no_compat = {
437 .llseek = proc_reg_llseek,
438 .read = proc_reg_read,
439 .write = proc_reg_write,
440 .poll = proc_reg_poll,
441 .unlocked_ioctl = proc_reg_unlocked_ioctl,
442 .mmap = proc_reg_mmap,
443 .open = proc_reg_open,
444 .release = proc_reg_release,
448 struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
450 struct inode * inode;
452 inode = iget_locked(sb, de->low_ino);
455 if (inode->i_state & I_NEW) {
456 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
457 PROC_I(inode)->fd = 0;
458 PROC_I(inode)->pde = de;
461 inode->i_mode = de->mode;
462 inode->i_uid = de->uid;
463 inode->i_gid = de->gid;
466 inode->i_size = de->size;
468 set_nlink(inode, de->nlink);
470 inode->i_op = de->proc_iops;
472 if (S_ISREG(inode->i_mode)) {
474 if (!de->proc_fops->compat_ioctl)
476 &proc_reg_file_ops_no_compat;
479 inode->i_fop = &proc_reg_file_ops;
481 inode->i_fop = de->proc_fops;
484 unlock_new_inode(inode);
490 int proc_fill_super(struct super_block *s)
492 struct inode * root_inode;
494 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
495 s->s_blocksize = 1024;
496 s->s_blocksize_bits = 10;
497 s->s_magic = PROC_SUPER_MAGIC;
498 s->s_op = &proc_sops;
502 root_inode = proc_get_inode(s, &proc_root);
505 root_inode->i_uid = 0;
506 root_inode->i_gid = 0;
507 s->s_root = d_alloc_root(root_inode);
513 printk("proc_read_super: get root inode failed\n");