Linux Audio

Check our new training course

Loading...
v4.6
 
  1/*
  2 *  linux/fs/proc/inode.c
  3 *
  4 *  Copyright (C) 1991, 1992  Linus Torvalds
  5 */
  6
 
  7#include <linux/time.h>
  8#include <linux/proc_fs.h>
  9#include <linux/kernel.h>
 10#include <linux/pid_namespace.h>
 11#include <linux/mm.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
 27#include <asm/uaccess.h>
 28
 29#include "internal.h"
 30
 31static void proc_evict_inode(struct inode *inode)
 32{
 33	struct proc_dir_entry *de;
 34	struct ctl_table_header *head;
 
 35
 36	truncate_inode_pages_final(&inode->i_data);
 37	clear_inode(inode);
 38
 39	/* Stop tracking associated processes */
 40	put_pid(PROC_I(inode)->pid);
 
 
 
 41
 42	/* Let go of any associated proc directory entry */
 43	de = PDE(inode);
 44	if (de)
 45		pde_put(de);
 46	head = PROC_I(inode)->sysctl;
 
 
 
 47	if (head) {
 48		RCU_INIT_POINTER(PROC_I(inode)->sysctl, NULL);
 49		sysctl_head_put(head);
 50	}
 51}
 52
 53static struct kmem_cache * proc_inode_cachep;
 
 54
 55static struct inode *proc_alloc_inode(struct super_block *sb)
 56{
 57	struct proc_inode *ei;
 58	struct inode *inode;
 59
 60	ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
 61	if (!ei)
 62		return NULL;
 63	ei->pid = NULL;
 64	ei->fd = 0;
 65	ei->op.proc_get_link = NULL;
 66	ei->pde = NULL;
 67	ei->sysctl = NULL;
 68	ei->sysctl_entry = NULL;
 
 69	ei->ns_ops = NULL;
 70	inode = &ei->vfs_inode;
 71	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 72	return inode;
 73}
 74
 75static void proc_i_callback(struct rcu_head *head)
 76{
 77	struct inode *inode = container_of(head, struct inode, i_rcu);
 78	kmem_cache_free(proc_inode_cachep, PROC_I(inode));
 79}
 80
 81static void proc_destroy_inode(struct inode *inode)
 82{
 83	call_rcu(&inode->i_rcu, proc_i_callback);
 84}
 85
 86static void init_once(void *foo)
 87{
 88	struct proc_inode *ei = (struct proc_inode *) foo;
 89
 90	inode_init_once(&ei->vfs_inode);
 91}
 92
 93void __init proc_init_inodecache(void)
 94{
 95	proc_inode_cachep = kmem_cache_create("proc_inode_cache",
 96					     sizeof(struct proc_inode),
 97					     0, (SLAB_RECLAIM_ACCOUNT|
 98						SLAB_MEM_SPREAD|SLAB_ACCOUNT|
 99						SLAB_PANIC),
100					     init_once);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
101}
102
103static int proc_show_options(struct seq_file *seq, struct dentry *root)
104{
105	struct super_block *sb = root->d_sb;
106	struct pid_namespace *pid = sb->s_fs_info;
107
108	if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
109		seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
110	if (pid->hide_pid != 0)
111		seq_printf(seq, ",hidepid=%u", pid->hide_pid);
 
 
112
113	return 0;
114}
115
116static const struct super_operations proc_sops = {
117	.alloc_inode	= proc_alloc_inode,
118	.destroy_inode	= proc_destroy_inode,
119	.drop_inode	= generic_delete_inode,
120	.evict_inode	= proc_evict_inode,
121	.statfs		= simple_statfs,
122	.remount_fs	= proc_remount,
123	.show_options	= proc_show_options,
124};
125
126enum {BIAS = -1U<<31};
127
128static inline int use_pde(struct proc_dir_entry *pde)
129{
130	return atomic_inc_unless_negative(&pde->in_use);
131}
132
133static void unuse_pde(struct proc_dir_entry *pde)
134{
135	if (atomic_dec_return(&pde->in_use) == BIAS)
136		complete(pde->pde_unload_completion);
137}
138
139/* pde is locked */
140static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
 
141{
 
 
 
 
 
 
 
 
 
 
142	if (pdeo->closing) {
143		/* somebody else is doing that, just wait */
144		DECLARE_COMPLETION_ONSTACK(c);
145		pdeo->c = &c;
146		spin_unlock(&pde->pde_unload_lock);
147		wait_for_completion(&c);
148		spin_lock(&pde->pde_unload_lock);
149	} else {
150		struct file *file;
151		pdeo->closing = 1;
 
 
152		spin_unlock(&pde->pde_unload_lock);
153		file = pdeo->file;
154		pde->proc_fops->release(file_inode(file), file);
155		spin_lock(&pde->pde_unload_lock);
156		list_del_init(&pdeo->lh);
157		if (pdeo->c)
158			complete(pdeo->c);
159		kfree(pdeo);
 
 
 
160	}
161}
162
163void proc_entry_rundown(struct proc_dir_entry *de)
164{
165	DECLARE_COMPLETION_ONSTACK(c);
166	/* Wait until all existing callers into module are done. */
167	de->pde_unload_completion = &c;
168	if (atomic_add_return(BIAS, &de->in_use) != BIAS)
169		wait_for_completion(&c);
170
 
 
171	spin_lock(&de->pde_unload_lock);
172	while (!list_empty(&de->pde_openers)) {
173		struct pde_opener *pdeo;
174		pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
175		close_pdeo(de, pdeo);
 
176	}
177	spin_unlock(&de->pde_unload_lock);
178}
179
180static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
181{
182	struct proc_dir_entry *pde = PDE(file_inode(file));
183	loff_t rv = -EINVAL;
184	if (use_pde(pde)) {
185		loff_t (*llseek)(struct file *, loff_t, int);
186		llseek = pde->proc_fops->llseek;
187		if (!llseek)
188			llseek = default_llseek;
189		rv = llseek(file, offset, whence);
190		unuse_pde(pde);
191	}
192	return rv;
193}
194
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
195static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
196{
197	ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
198	struct proc_dir_entry *pde = PDE(file_inode(file));
199	ssize_t rv = -EIO;
200	if (use_pde(pde)) {
201		read = pde->proc_fops->read;
202		if (read)
203			rv = read(file, buf, count, ppos);
 
204		unuse_pde(pde);
205	}
206	return rv;
207}
208
 
 
 
 
 
 
 
 
 
 
209static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
210{
211	ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
212	struct proc_dir_entry *pde = PDE(file_inode(file));
213	ssize_t rv = -EIO;
214	if (use_pde(pde)) {
215		write = pde->proc_fops->write;
216		if (write)
217			rv = write(file, buf, count, ppos);
 
218		unuse_pde(pde);
219	}
220	return rv;
221}
222
223static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
 
 
 
 
 
 
 
 
 
 
224{
225	struct proc_dir_entry *pde = PDE(file_inode(file));
226	unsigned int rv = DEFAULT_POLLMASK;
227	unsigned int (*poll)(struct file *, struct poll_table_struct *);
228	if (use_pde(pde)) {
229		poll = pde->proc_fops->poll;
230		if (poll)
231			rv = poll(file, pts);
232		unuse_pde(pde);
233	}
234	return rv;
235}
236
 
 
 
 
 
 
 
 
 
 
237static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
238{
239	struct proc_dir_entry *pde = PDE(file_inode(file));
240	long rv = -ENOTTY;
241	long (*ioctl)(struct file *, unsigned int, unsigned long);
242	if (use_pde(pde)) {
243		ioctl = pde->proc_fops->unlocked_ioctl;
244		if (ioctl)
245			rv = ioctl(file, cmd, arg);
246		unuse_pde(pde);
247	}
248	return rv;
249}
250
251#ifdef CONFIG_COMPAT
 
 
 
 
 
 
 
 
 
 
252static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
253{
254	struct proc_dir_entry *pde = PDE(file_inode(file));
255	long rv = -ENOTTY;
256	long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
257	if (use_pde(pde)) {
258		compat_ioctl = pde->proc_fops->compat_ioctl;
259		if (compat_ioctl)
260			rv = compat_ioctl(file, cmd, arg);
261		unuse_pde(pde);
262	}
263	return rv;
264}
265#endif
266
 
 
 
 
 
 
 
 
 
 
267static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
268{
269	struct proc_dir_entry *pde = PDE(file_inode(file));
270	int rv = -EIO;
271	int (*mmap)(struct file *, struct vm_area_struct *);
272	if (use_pde(pde)) {
273		mmap = pde->proc_fops->mmap;
274		if (mmap)
275			rv = mmap(file, vma);
276		unuse_pde(pde);
277	}
278	return rv;
279}
280
281static unsigned long
282proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
283			   unsigned long len, unsigned long pgoff,
284			   unsigned long flags)
285{
286	struct proc_dir_entry *pde = PDE(file_inode(file));
287	unsigned long rv = -EIO;
288
289	if (use_pde(pde)) {
290		typeof(proc_reg_get_unmapped_area) *get_area;
291
292		get_area = pde->proc_fops->get_unmapped_area;
293#ifdef CONFIG_MMU
294		if (!get_area)
295			get_area = current->mm->get_unmapped_area;
296#endif
 
 
 
 
297
298		if (get_area)
299			rv = get_area(file, orig_addr, len, pgoff, flags);
300		else
301			rv = orig_addr;
 
 
 
 
 
 
 
 
302		unuse_pde(pde);
303	}
304	return rv;
305}
306
307static int proc_reg_open(struct inode *inode, struct file *file)
308{
309	struct proc_dir_entry *pde = PDE(inode);
310	int rv = 0;
311	int (*open)(struct inode *, struct file *);
312	int (*release)(struct inode *, struct file *);
313	struct pde_opener *pdeo;
314
 
 
 
 
 
 
 
315	/*
316	 * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
317	 * sequence. ->release won't be called because ->proc_fops will be
318	 * cleared. Depending on complexity of ->release, consequences vary.
 
319	 *
320	 * We can't wait for mercy when close will be done for real, it's
321	 * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
322	 * by hand in remove_proc_entry(). For this, save opener's credentials
323	 * for later.
324	 */
325	pdeo = kzalloc(sizeof(struct pde_opener), GFP_KERNEL);
326	if (!pdeo)
327		return -ENOMEM;
328
329	if (!use_pde(pde)) {
330		kfree(pdeo);
331		return -ENOENT;
 
 
 
 
 
 
 
 
332	}
333	open = pde->proc_fops->open;
334	release = pde->proc_fops->release;
335
 
336	if (open)
337		rv = open(inode, file);
338
339	if (rv == 0 && release) {
340		/* To know what to release. */
341		pdeo->file = file;
342		/* Strictly for "too late" ->release in proc_reg_release(). */
343		spin_lock(&pde->pde_unload_lock);
344		list_add(&pdeo->lh, &pde->pde_openers);
345		spin_unlock(&pde->pde_unload_lock);
346	} else
347		kfree(pdeo);
 
 
 
348
 
349	unuse_pde(pde);
350	return rv;
351}
352
353static int proc_reg_release(struct inode *inode, struct file *file)
354{
355	struct proc_dir_entry *pde = PDE(inode);
356	struct pde_opener *pdeo;
 
 
 
 
 
 
 
 
 
 
 
357	spin_lock(&pde->pde_unload_lock);
358	list_for_each_entry(pdeo, &pde->pde_openers, lh) {
359		if (pdeo->file == file) {
360			close_pdeo(pde, pdeo);
361			break;
362		}
363	}
364	spin_unlock(&pde->pde_unload_lock);
365	return 0;
366}
367
368static const struct file_operations proc_reg_file_ops = {
369	.llseek		= proc_reg_llseek,
370	.read		= proc_reg_read,
371	.write		= proc_reg_write,
372	.poll		= proc_reg_poll,
373	.unlocked_ioctl	= proc_reg_unlocked_ioctl,
374#ifdef CONFIG_COMPAT
375	.compat_ioctl	= proc_reg_compat_ioctl,
376#endif
 
 
 
 
 
 
 
 
 
 
377	.mmap		= proc_reg_mmap,
378	.get_unmapped_area = proc_reg_get_unmapped_area,
379	.open		= proc_reg_open,
380	.release	= proc_reg_release,
381};
382
383#ifdef CONFIG_COMPAT
384static const struct file_operations proc_reg_file_ops_no_compat = {
385	.llseek		= proc_reg_llseek,
386	.read		= proc_reg_read,
387	.write		= proc_reg_write,
388	.poll		= proc_reg_poll,
389	.unlocked_ioctl	= proc_reg_unlocked_ioctl,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
390	.mmap		= proc_reg_mmap,
391	.get_unmapped_area = proc_reg_get_unmapped_area,
392	.open		= proc_reg_open,
393	.release	= proc_reg_release,
394};
395#endif
396
397static void proc_put_link(void *p)
398{
399	unuse_pde(p);
400}
401
402static const char *proc_get_link(struct dentry *dentry,
403				 struct inode *inode,
404				 struct delayed_call *done)
405{
406	struct proc_dir_entry *pde = PDE(inode);
407	if (unlikely(!use_pde(pde)))
408		return ERR_PTR(-EINVAL);
409	set_delayed_call(done, proc_put_link, pde);
410	return pde->data;
411}
412
413const struct inode_operations proc_link_inode_operations = {
414	.readlink	= generic_readlink,
415	.get_link	= proc_get_link,
416};
417
418struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
419{
420	struct inode *inode = new_inode_pseudo(sb);
421
422	if (inode) {
423		inode->i_ino = de->low_ino;
424		inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
425		PROC_I(inode)->pde = de;
426
427		if (is_empty_pde(de)) {
428			make_empty_dir_inode(inode);
429			return inode;
430		}
431		if (de->mode) {
432			inode->i_mode = de->mode;
433			inode->i_uid = de->uid;
434			inode->i_gid = de->gid;
435		}
436		if (de->size)
437			inode->i_size = de->size;
438		if (de->nlink)
439			set_nlink(inode, de->nlink);
440		WARN_ON(!de->proc_iops);
 
 
 
 
 
441		inode->i_op = de->proc_iops;
442		if (de->proc_fops) {
443			if (S_ISREG(inode->i_mode)) {
 
 
444#ifdef CONFIG_COMPAT
445				if (!de->proc_fops->compat_ioctl)
446					inode->i_fop =
447						&proc_reg_file_ops_no_compat;
448				else
449#endif
450					inode->i_fop = &proc_reg_file_ops;
451			} else {
452				inode->i_fop = de->proc_fops;
453			}
454		}
455	} else
456	       pde_put(de);
457	return inode;
458}
459
460int proc_fill_super(struct super_block *s)
461{
462	struct inode *root_inode;
463	int ret;
464
465	s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
466	s->s_blocksize = 1024;
467	s->s_blocksize_bits = 10;
468	s->s_magic = PROC_SUPER_MAGIC;
469	s->s_op = &proc_sops;
470	s->s_time_gran = 1;
471	
472	pde_get(&proc_root);
473	root_inode = proc_get_inode(s, &proc_root);
474	if (!root_inode) {
475		pr_err("proc_fill_super: get root inode failed\n");
476		return -ENOMEM;
477	}
478
479	s->s_root = d_make_root(root_inode);
480	if (!s->s_root) {
481		pr_err("proc_fill_super: allocate dentry failed\n");
482		return -ENOMEM;
483	}
484
485	ret = proc_setup_self(s);
486	if (ret) {
487		return ret;
488	}
489	return proc_setup_thread_self(s);
490}
v5.14.15
  1// SPDX-License-Identifier: GPL-2.0
  2/*
  3 *  linux/fs/proc/inode.c
  4 *
  5 *  Copyright (C) 1991, 1992  Linus Torvalds
  6 */
  7
  8#include <linux/cache.h>
  9#include <linux/time.h>
 10#include <linux/proc_fs.h>
 11#include <linux/kernel.h>
 12#include <linux/pid_namespace.h>
 13#include <linux/mm.h>
 14#include <linux/string.h>
 15#include <linux/stat.h>
 16#include <linux/completion.h>
 17#include <linux/poll.h>
 18#include <linux/printk.h>
 19#include <linux/file.h>
 20#include <linux/limits.h>
 21#include <linux/init.h>
 22#include <linux/module.h>
 23#include <linux/sysctl.h>
 24#include <linux/seq_file.h>
 25#include <linux/slab.h>
 26#include <linux/mount.h>
 27#include <linux/bug.h>
 28
 29#include <linux/uaccess.h>
 30
 31#include "internal.h"
 32
 33static void proc_evict_inode(struct inode *inode)
 34{
 35	struct proc_dir_entry *de;
 36	struct ctl_table_header *head;
 37	struct proc_inode *ei = PROC_I(inode);
 38
 39	truncate_inode_pages_final(&inode->i_data);
 40	clear_inode(inode);
 41
 42	/* Stop tracking associated processes */
 43	if (ei->pid) {
 44		proc_pid_evict_inode(ei);
 45		ei->pid = NULL;
 46	}
 47
 48	/* Let go of any associated proc directory entry */
 49	de = ei->pde;
 50	if (de) {
 51		pde_put(de);
 52		ei->pde = NULL;
 53	}
 54
 55	head = ei->sysctl;
 56	if (head) {
 57		RCU_INIT_POINTER(ei->sysctl, NULL);
 58		proc_sys_evict_inode(inode, head);
 59	}
 60}
 61
 62static struct kmem_cache *proc_inode_cachep __ro_after_init;
 63static struct kmem_cache *pde_opener_cache __ro_after_init;
 64
 65static struct inode *proc_alloc_inode(struct super_block *sb)
 66{
 67	struct proc_inode *ei;
 
 68
 69	ei = kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
 70	if (!ei)
 71		return NULL;
 72	ei->pid = NULL;
 73	ei->fd = 0;
 74	ei->op.proc_get_link = NULL;
 75	ei->pde = NULL;
 76	ei->sysctl = NULL;
 77	ei->sysctl_entry = NULL;
 78	INIT_HLIST_NODE(&ei->sibling_inodes);
 79	ei->ns_ops = NULL;
 80	return &ei->vfs_inode;
 
 
 81}
 82
 83static void proc_free_inode(struct inode *inode)
 84{
 
 85	kmem_cache_free(proc_inode_cachep, PROC_I(inode));
 86}
 87
 
 
 
 
 
 88static void init_once(void *foo)
 89{
 90	struct proc_inode *ei = (struct proc_inode *) foo;
 91
 92	inode_init_once(&ei->vfs_inode);
 93}
 94
 95void __init proc_init_kmemcache(void)
 96{
 97	proc_inode_cachep = kmem_cache_create("proc_inode_cache",
 98					     sizeof(struct proc_inode),
 99					     0, (SLAB_RECLAIM_ACCOUNT|
100						SLAB_MEM_SPREAD|SLAB_ACCOUNT|
101						SLAB_PANIC),
102					     init_once);
103	pde_opener_cache =
104		kmem_cache_create("pde_opener", sizeof(struct pde_opener), 0,
105				  SLAB_ACCOUNT|SLAB_PANIC, NULL);
106	proc_dir_entry_cache = kmem_cache_create_usercopy(
107		"proc_dir_entry", SIZEOF_PDE, 0, SLAB_PANIC,
108		offsetof(struct proc_dir_entry, inline_name),
109		SIZEOF_PDE_INLINE_NAME, NULL);
110	BUILD_BUG_ON(sizeof(struct proc_dir_entry) >= SIZEOF_PDE);
111}
112
113void proc_invalidate_siblings_dcache(struct hlist_head *inodes, spinlock_t *lock)
114{
115	struct inode *inode;
116	struct proc_inode *ei;
117	struct hlist_node *node;
118	struct super_block *old_sb = NULL;
119
120	rcu_read_lock();
121	for (;;) {
122		struct super_block *sb;
123		node = hlist_first_rcu(inodes);
124		if (!node)
125			break;
126		ei = hlist_entry(node, struct proc_inode, sibling_inodes);
127		spin_lock(lock);
128		hlist_del_init_rcu(&ei->sibling_inodes);
129		spin_unlock(lock);
130
131		inode = &ei->vfs_inode;
132		sb = inode->i_sb;
133		if ((sb != old_sb) && !atomic_inc_not_zero(&sb->s_active))
134			continue;
135		inode = igrab(inode);
136		rcu_read_unlock();
137		if (sb != old_sb) {
138			if (old_sb)
139				deactivate_super(old_sb);
140			old_sb = sb;
141		}
142		if (unlikely(!inode)) {
143			rcu_read_lock();
144			continue;
145		}
146
147		if (S_ISDIR(inode->i_mode)) {
148			struct dentry *dir = d_find_any_alias(inode);
149			if (dir) {
150				d_invalidate(dir);
151				dput(dir);
152			}
153		} else {
154			struct dentry *dentry;
155			while ((dentry = d_find_alias(inode))) {
156				d_invalidate(dentry);
157				dput(dentry);
158			}
159		}
160		iput(inode);
161
162		rcu_read_lock();
163	}
164	rcu_read_unlock();
165	if (old_sb)
166		deactivate_super(old_sb);
167}
168
169static inline const char *hidepid2str(enum proc_hidepid v)
170{
171	switch (v) {
172		case HIDEPID_OFF: return "off";
173		case HIDEPID_NO_ACCESS: return "noaccess";
174		case HIDEPID_INVISIBLE: return "invisible";
175		case HIDEPID_NOT_PTRACEABLE: return "ptraceable";
176	}
177	WARN_ONCE(1, "bad hide_pid value: %d\n", v);
178	return "unknown";
179}
180
181static int proc_show_options(struct seq_file *seq, struct dentry *root)
182{
183	struct proc_fs_info *fs_info = proc_sb_info(root->d_sb);
 
184
185	if (!gid_eq(fs_info->pid_gid, GLOBAL_ROOT_GID))
186		seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, fs_info->pid_gid));
187	if (fs_info->hide_pid != HIDEPID_OFF)
188		seq_printf(seq, ",hidepid=%s", hidepid2str(fs_info->hide_pid));
189	if (fs_info->pidonly != PROC_PIDONLY_OFF)
190		seq_printf(seq, ",subset=pid");
191
192	return 0;
193}
194
195const struct super_operations proc_sops = {
196	.alloc_inode	= proc_alloc_inode,
197	.free_inode	= proc_free_inode,
198	.drop_inode	= generic_delete_inode,
199	.evict_inode	= proc_evict_inode,
200	.statfs		= simple_statfs,
 
201	.show_options	= proc_show_options,
202};
203
204enum {BIAS = -1U<<31};
205
206static inline int use_pde(struct proc_dir_entry *pde)
207{
208	return likely(atomic_inc_unless_negative(&pde->in_use));
209}
210
211static void unuse_pde(struct proc_dir_entry *pde)
212{
213	if (unlikely(atomic_dec_return(&pde->in_use) == BIAS))
214		complete(pde->pde_unload_completion);
215}
216
217/* pde is locked on entry, unlocked on exit */
218static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
219	__releases(&pde->pde_unload_lock)
220{
221	/*
222	 * close() (proc_reg_release()) can't delete an entry and proceed:
223	 * ->release hook needs to be available at the right moment.
224	 *
225	 * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
226	 * "struct file" needs to be available at the right moment.
227	 *
228	 * Therefore, first process to enter this function does ->release() and
229	 * signals its completion to the other process which does nothing.
230	 */
231	if (pdeo->closing) {
232		/* somebody else is doing that, just wait */
233		DECLARE_COMPLETION_ONSTACK(c);
234		pdeo->c = &c;
235		spin_unlock(&pde->pde_unload_lock);
236		wait_for_completion(&c);
 
237	} else {
238		struct file *file;
239		struct completion *c;
240
241		pdeo->closing = true;
242		spin_unlock(&pde->pde_unload_lock);
243		file = pdeo->file;
244		pde->proc_ops->proc_release(file_inode(file), file);
245		spin_lock(&pde->pde_unload_lock);
246		/* After ->release. */
247		list_del(&pdeo->lh);
248		c = pdeo->c;
249		spin_unlock(&pde->pde_unload_lock);
250		if (unlikely(c))
251			complete(c);
252		kmem_cache_free(pde_opener_cache, pdeo);
253	}
254}
255
256void proc_entry_rundown(struct proc_dir_entry *de)
257{
258	DECLARE_COMPLETION_ONSTACK(c);
259	/* Wait until all existing callers into module are done. */
260	de->pde_unload_completion = &c;
261	if (atomic_add_return(BIAS, &de->in_use) != BIAS)
262		wait_for_completion(&c);
263
264	/* ->pde_openers list can't grow from now on. */
265
266	spin_lock(&de->pde_unload_lock);
267	while (!list_empty(&de->pde_openers)) {
268		struct pde_opener *pdeo;
269		pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
270		close_pdeo(de, pdeo);
271		spin_lock(&de->pde_unload_lock);
272	}
273	spin_unlock(&de->pde_unload_lock);
274}
275
276static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
277{
278	struct proc_dir_entry *pde = PDE(file_inode(file));
279	loff_t rv = -EINVAL;
280
281	if (pde_is_permanent(pde)) {
282		return pde->proc_ops->proc_lseek(file, offset, whence);
283	} else if (use_pde(pde)) {
284		rv = pde->proc_ops->proc_lseek(file, offset, whence);
 
285		unuse_pde(pde);
286	}
287	return rv;
288}
289
290static ssize_t proc_reg_read_iter(struct kiocb *iocb, struct iov_iter *iter)
291{
292	struct proc_dir_entry *pde = PDE(file_inode(iocb->ki_filp));
293	ssize_t ret;
294
295	if (pde_is_permanent(pde))
296		return pde->proc_ops->proc_read_iter(iocb, iter);
297
298	if (!use_pde(pde))
299		return -EIO;
300	ret = pde->proc_ops->proc_read_iter(iocb, iter);
301	unuse_pde(pde);
302	return ret;
303}
304
305static ssize_t pde_read(struct proc_dir_entry *pde, struct file *file, char __user *buf, size_t count, loff_t *ppos)
306{
307	typeof_member(struct proc_ops, proc_read) read;
308
309	read = pde->proc_ops->proc_read;
310	if (read)
311		return read(file, buf, count, ppos);
312	return -EIO;
313}
314
315static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
316{
 
317	struct proc_dir_entry *pde = PDE(file_inode(file));
318	ssize_t rv = -EIO;
319
320	if (pde_is_permanent(pde)) {
321		return pde_read(pde, file, buf, count, ppos);
322	} else if (use_pde(pde)) {
323		rv = pde_read(pde, file, buf, count, ppos);
324		unuse_pde(pde);
325	}
326	return rv;
327}
328
329static ssize_t pde_write(struct proc_dir_entry *pde, struct file *file, const char __user *buf, size_t count, loff_t *ppos)
330{
331	typeof_member(struct proc_ops, proc_write) write;
332
333	write = pde->proc_ops->proc_write;
334	if (write)
335		return write(file, buf, count, ppos);
336	return -EIO;
337}
338
339static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
340{
 
341	struct proc_dir_entry *pde = PDE(file_inode(file));
342	ssize_t rv = -EIO;
343
344	if (pde_is_permanent(pde)) {
345		return pde_write(pde, file, buf, count, ppos);
346	} else if (use_pde(pde)) {
347		rv = pde_write(pde, file, buf, count, ppos);
348		unuse_pde(pde);
349	}
350	return rv;
351}
352
353static __poll_t pde_poll(struct proc_dir_entry *pde, struct file *file, struct poll_table_struct *pts)
354{
355	typeof_member(struct proc_ops, proc_poll) poll;
356
357	poll = pde->proc_ops->proc_poll;
358	if (poll)
359		return poll(file, pts);
360	return DEFAULT_POLLMASK;
361}
362
363static __poll_t proc_reg_poll(struct file *file, struct poll_table_struct *pts)
364{
365	struct proc_dir_entry *pde = PDE(file_inode(file));
366	__poll_t rv = DEFAULT_POLLMASK;
367
368	if (pde_is_permanent(pde)) {
369		return pde_poll(pde, file, pts);
370	} else if (use_pde(pde)) {
371		rv = pde_poll(pde, file, pts);
372		unuse_pde(pde);
373	}
374	return rv;
375}
376
377static long pde_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
378{
379	typeof_member(struct proc_ops, proc_ioctl) ioctl;
380
381	ioctl = pde->proc_ops->proc_ioctl;
382	if (ioctl)
383		return ioctl(file, cmd, arg);
384	return -ENOTTY;
385}
386
387static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
388{
389	struct proc_dir_entry *pde = PDE(file_inode(file));
390	long rv = -ENOTTY;
391
392	if (pde_is_permanent(pde)) {
393		return pde_ioctl(pde, file, cmd, arg);
394	} else if (use_pde(pde)) {
395		rv = pde_ioctl(pde, file, cmd, arg);
396		unuse_pde(pde);
397	}
398	return rv;
399}
400
401#ifdef CONFIG_COMPAT
402static long pde_compat_ioctl(struct proc_dir_entry *pde, struct file *file, unsigned int cmd, unsigned long arg)
403{
404	typeof_member(struct proc_ops, proc_compat_ioctl) compat_ioctl;
405
406	compat_ioctl = pde->proc_ops->proc_compat_ioctl;
407	if (compat_ioctl)
408		return compat_ioctl(file, cmd, arg);
409	return -ENOTTY;
410}
411
412static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
413{
414	struct proc_dir_entry *pde = PDE(file_inode(file));
415	long rv = -ENOTTY;
416	if (pde_is_permanent(pde)) {
417		return pde_compat_ioctl(pde, file, cmd, arg);
418	} else if (use_pde(pde)) {
419		rv = pde_compat_ioctl(pde, file, cmd, arg);
 
420		unuse_pde(pde);
421	}
422	return rv;
423}
424#endif
425
426static int pde_mmap(struct proc_dir_entry *pde, struct file *file, struct vm_area_struct *vma)
427{
428	typeof_member(struct proc_ops, proc_mmap) mmap;
429
430	mmap = pde->proc_ops->proc_mmap;
431	if (mmap)
432		return mmap(file, vma);
433	return -EIO;
434}
435
436static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
437{
438	struct proc_dir_entry *pde = PDE(file_inode(file));
439	int rv = -EIO;
440
441	if (pde_is_permanent(pde)) {
442		return pde_mmap(pde, file, vma);
443	} else if (use_pde(pde)) {
444		rv = pde_mmap(pde, file, vma);
445		unuse_pde(pde);
446	}
447	return rv;
448}
449
450static unsigned long
451pde_get_unmapped_area(struct proc_dir_entry *pde, struct file *file, unsigned long orig_addr,
452			   unsigned long len, unsigned long pgoff,
453			   unsigned long flags)
454{
455	typeof_member(struct proc_ops, proc_get_unmapped_area) get_area;
 
 
 
 
456
457	get_area = pde->proc_ops->proc_get_unmapped_area;
458#ifdef CONFIG_MMU
459	if (!get_area)
460		get_area = current->mm->get_unmapped_area;
461#endif
462	if (get_area)
463		return get_area(file, orig_addr, len, pgoff, flags);
464	return orig_addr;
465}
466
467static unsigned long
468proc_reg_get_unmapped_area(struct file *file, unsigned long orig_addr,
469			   unsigned long len, unsigned long pgoff,
470			   unsigned long flags)
471{
472	struct proc_dir_entry *pde = PDE(file_inode(file));
473	unsigned long rv = -EIO;
474
475	if (pde_is_permanent(pde)) {
476		return pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
477	} else if (use_pde(pde)) {
478		rv = pde_get_unmapped_area(pde, file, orig_addr, len, pgoff, flags);
479		unuse_pde(pde);
480	}
481	return rv;
482}
483
484static int proc_reg_open(struct inode *inode, struct file *file)
485{
486	struct proc_dir_entry *pde = PDE(inode);
487	int rv = 0;
488	typeof_member(struct proc_ops, proc_open) open;
489	typeof_member(struct proc_ops, proc_release) release;
490	struct pde_opener *pdeo;
491
492	if (pde_is_permanent(pde)) {
493		open = pde->proc_ops->proc_open;
494		if (open)
495			rv = open(inode, file);
496		return rv;
497	}
498
499	/*
500	 * Ensure that
501	 * 1) PDE's ->release hook will be called no matter what
502	 *    either normally by close()/->release, or forcefully by
503	 *    rmmod/remove_proc_entry.
504	 *
505	 * 2) rmmod isn't blocked by opening file in /proc and sitting on
506	 *    the descriptor (including "rmmod foo </proc/foo" scenario).
507	 *
508	 * Save every "struct file" with custom ->release hook.
509	 */
510	if (!use_pde(pde))
 
 
 
 
 
511		return -ENOENT;
512
513	release = pde->proc_ops->proc_release;
514	if (release) {
515		pdeo = kmem_cache_alloc(pde_opener_cache, GFP_KERNEL);
516		if (!pdeo) {
517			rv = -ENOMEM;
518			goto out_unuse;
519		}
520	}
 
 
521
522	open = pde->proc_ops->proc_open;
523	if (open)
524		rv = open(inode, file);
525
526	if (release) {
527		if (rv == 0) {
528			/* To know what to release. */
529			pdeo->file = file;
530			pdeo->closing = false;
531			pdeo->c = NULL;
532			spin_lock(&pde->pde_unload_lock);
533			list_add(&pdeo->lh, &pde->pde_openers);
534			spin_unlock(&pde->pde_unload_lock);
535		} else
536			kmem_cache_free(pde_opener_cache, pdeo);
537	}
538
539out_unuse:
540	unuse_pde(pde);
541	return rv;
542}
543
544static int proc_reg_release(struct inode *inode, struct file *file)
545{
546	struct proc_dir_entry *pde = PDE(inode);
547	struct pde_opener *pdeo;
548
549	if (pde_is_permanent(pde)) {
550		typeof_member(struct proc_ops, proc_release) release;
551
552		release = pde->proc_ops->proc_release;
553		if (release) {
554			return release(inode, file);
555		}
556		return 0;
557	}
558
559	spin_lock(&pde->pde_unload_lock);
560	list_for_each_entry(pdeo, &pde->pde_openers, lh) {
561		if (pdeo->file == file) {
562			close_pdeo(pde, pdeo);
563			return 0;
564		}
565	}
566	spin_unlock(&pde->pde_unload_lock);
567	return 0;
568}
569
570static const struct file_operations proc_reg_file_ops = {
571	.llseek		= proc_reg_llseek,
572	.read		= proc_reg_read,
573	.write		= proc_reg_write,
574	.poll		= proc_reg_poll,
575	.unlocked_ioctl	= proc_reg_unlocked_ioctl,
576	.mmap		= proc_reg_mmap,
577	.get_unmapped_area = proc_reg_get_unmapped_area,
578	.open		= proc_reg_open,
579	.release	= proc_reg_release,
580};
581
582static const struct file_operations proc_iter_file_ops = {
583	.llseek		= proc_reg_llseek,
584	.read_iter	= proc_reg_read_iter,
585	.write		= proc_reg_write,
586	.splice_read	= generic_file_splice_read,
587	.poll		= proc_reg_poll,
588	.unlocked_ioctl	= proc_reg_unlocked_ioctl,
589	.mmap		= proc_reg_mmap,
590	.get_unmapped_area = proc_reg_get_unmapped_area,
591	.open		= proc_reg_open,
592	.release	= proc_reg_release,
593};
594
595#ifdef CONFIG_COMPAT
596static const struct file_operations proc_reg_file_ops_compat = {
597	.llseek		= proc_reg_llseek,
598	.read		= proc_reg_read,
599	.write		= proc_reg_write,
600	.poll		= proc_reg_poll,
601	.unlocked_ioctl	= proc_reg_unlocked_ioctl,
602	.compat_ioctl	= proc_reg_compat_ioctl,
603	.mmap		= proc_reg_mmap,
604	.get_unmapped_area = proc_reg_get_unmapped_area,
605	.open		= proc_reg_open,
606	.release	= proc_reg_release,
607};
608
609static const struct file_operations proc_iter_file_ops_compat = {
610	.llseek		= proc_reg_llseek,
611	.read_iter	= proc_reg_read_iter,
612	.splice_read	= generic_file_splice_read,
613	.write		= proc_reg_write,
614	.poll		= proc_reg_poll,
615	.unlocked_ioctl	= proc_reg_unlocked_ioctl,
616	.compat_ioctl	= proc_reg_compat_ioctl,
617	.mmap		= proc_reg_mmap,
618	.get_unmapped_area = proc_reg_get_unmapped_area,
619	.open		= proc_reg_open,
620	.release	= proc_reg_release,
621};
622#endif
623
624static void proc_put_link(void *p)
625{
626	unuse_pde(p);
627}
628
629static const char *proc_get_link(struct dentry *dentry,
630				 struct inode *inode,
631				 struct delayed_call *done)
632{
633	struct proc_dir_entry *pde = PDE(inode);
634	if (!use_pde(pde))
635		return ERR_PTR(-EINVAL);
636	set_delayed_call(done, proc_put_link, pde);
637	return pde->data;
638}
639
640const struct inode_operations proc_link_inode_operations = {
 
641	.get_link	= proc_get_link,
642};
643
644struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
645{
646	struct inode *inode = new_inode(sb);
647
648	if (!inode) {
649		pde_put(de);
650		return NULL;
651	}
652
653	inode->i_ino = de->low_ino;
654	inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
655	PROC_I(inode)->pde = de;
656	if (is_empty_pde(de)) {
657		make_empty_dir_inode(inode);
658		return inode;
659	}
660
661	if (de->mode) {
662		inode->i_mode = de->mode;
663		inode->i_uid = de->uid;
664		inode->i_gid = de->gid;
665	}
666	if (de->size)
667		inode->i_size = de->size;
668	if (de->nlink)
669		set_nlink(inode, de->nlink);
670
671	if (S_ISREG(inode->i_mode)) {
672		inode->i_op = de->proc_iops;
673		if (de->proc_ops->proc_read_iter)
674			inode->i_fop = &proc_iter_file_ops;
675		else
676			inode->i_fop = &proc_reg_file_ops;
677#ifdef CONFIG_COMPAT
678		if (de->proc_ops->proc_compat_ioctl) {
679			if (de->proc_ops->proc_read_iter)
680				inode->i_fop = &proc_iter_file_ops_compat;
681			else
682				inode->i_fop = &proc_reg_file_ops_compat;
 
 
 
 
683		}
684#endif
685	} else if (S_ISDIR(inode->i_mode)) {
686		inode->i_op = de->proc_iops;
687		inode->i_fop = de->proc_dir_ops;
688	} else if (S_ISLNK(inode->i_mode)) {
689		inode->i_op = de->proc_iops;
690		inode->i_fop = NULL;
691	} else {
692		BUG();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
693	}
694	return inode;
695}