1
2
3
4
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 const struct proc_ns_operations *ns_ops;
36 void *ns;
37
38 truncate_inode_pages_final(&inode->i_data);
39 clear_inode(inode);
40
41
42 put_pid(PROC_I(inode)->pid);
43
44
45 de = PROC_I(inode)->pde;
46 if (de)
47 pde_put(de);
48
49 head = PROC_I(inode)->sysctl;
50 if (head) {
51 rcu_assign_pointer(PROC_I(inode)->sysctl, NULL);
52 proc_sys_evict_inode(inode, head);
53 }
54
55 ns_ops = PROC_I(inode)->ns.ns_ops;
56 ns = PROC_I(inode)->ns.ns;
57 if (ns_ops && ns)
58 ns_ops->put(ns);
59}
60
61static struct kmem_cache * proc_inode_cachep;
62
63static struct inode *proc_alloc_inode(struct super_block *sb)
64{
65 struct proc_inode *ei;
66 struct inode *inode;
67
68 ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
69 if (!ei)
70 return NULL;
71 ei->pid = NULL;
72 ei->fd = 0;
73 ei->op.proc_get_link = NULL;
74 ei->pde = NULL;
75 ei->sysctl = NULL;
76 ei->sysctl_entry = NULL;
77 ei->ns.ns = NULL;
78 ei->ns.ns_ops = NULL;
79 inode = &ei->vfs_inode;
80 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
81 return inode;
82}
83
84static void proc_i_callback(struct rcu_head *head)
85{
86 struct inode *inode = container_of(head, struct inode, i_rcu);
87 kmem_cache_free(proc_inode_cachep, PROC_I(inode));
88}
89
90static void proc_destroy_inode(struct inode *inode)
91{
92 call_rcu(&inode->i_rcu, proc_i_callback);
93}
94
95static void init_once(void *foo)
96{
97 struct proc_inode *ei = (struct proc_inode *) foo;
98
99 inode_init_once(&ei->vfs_inode);
100}
101
102void __init proc_init_inodecache(void)
103{
104 proc_inode_cachep = kmem_cache_create("proc_inode_cache",
105 sizeof(struct proc_inode),
106 0, (SLAB_RECLAIM_ACCOUNT|
107 SLAB_MEM_SPREAD|SLAB_ACCOUNT|
108 SLAB_PANIC),
109 init_once);
110}
111
112static int proc_show_options(struct seq_file *seq, struct dentry *root)
113{
114 struct super_block *sb = root->d_sb;
115 struct pid_namespace *pid = sb->s_fs_info;
116
117 if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
118 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
119 if (pid->hide_pid != 0)
120 seq_printf(seq, ",hidepid=%u", pid->hide_pid);
121
122 return 0;
123}
124
125static const struct super_operations proc_sops = {
126 .alloc_inode = proc_alloc_inode,
127 .destroy_inode = proc_destroy_inode,
128 .drop_inode = generic_delete_inode,
129 .evict_inode = proc_evict_inode,
130 .statfs = simple_statfs,
131 .remount_fs = proc_remount,
132 .show_options = proc_show_options,
133};
134
135enum {BIAS = -1U<<31};
136
137static inline int use_pde(struct proc_dir_entry *pde)
138{
139 return atomic_inc_unless_negative(&pde->in_use);
140}
141
142static void unuse_pde(struct proc_dir_entry *pde)
143{
144 if (atomic_dec_return(&pde->in_use) == BIAS)
145 complete(pde->pde_unload_completion);
146}
147
148
149static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
150{
151 if (pdeo->closing) {
152
153 DECLARE_COMPLETION_ONSTACK(c);
154 pdeo->c = &c;
155 spin_unlock(&pde->pde_unload_lock);
156 wait_for_completion(&c);
157 spin_lock(&pde->pde_unload_lock);
158 } else {
159 struct file *file;
160 pdeo->closing = 1;
161 spin_unlock(&pde->pde_unload_lock);
162 file = pdeo->file;
163 pde->proc_fops->release(file_inode(file), file);
164 spin_lock(&pde->pde_unload_lock);
165 list_del_init(&pdeo->lh);
166 if (pdeo->c)
167 complete(pdeo->c);
168 kfree(pdeo);
169 }
170}
171
172void proc_entry_rundown(struct proc_dir_entry *de)
173{
174 DECLARE_COMPLETION_ONSTACK(c);
175
176 de->pde_unload_completion = &c;
177 if (atomic_add_return(BIAS, &de->in_use) != BIAS)
178 wait_for_completion(&c);
179
180 spin_lock(&de->pde_unload_lock);
181 while (!list_empty(&de->pde_openers)) {
182 struct pde_opener *pdeo;
183 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
184 close_pdeo(de, pdeo);
185 }
186 spin_unlock(&de->pde_unload_lock);
187}
188
189static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
190{
191 struct proc_dir_entry *pde = PDE(file_inode(file));
192 loff_t rv = -EINVAL;
193 if (use_pde(pde)) {
194 loff_t (*llseek)(struct file *, loff_t, int);
195 llseek = pde->proc_fops->llseek;
196 if (!llseek)
197 llseek = default_llseek;
198 rv = llseek(file, offset, whence);
199 unuse_pde(pde);
200 }
201 return rv;
202}
203
204static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
205{
206 ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
207 struct proc_dir_entry *pde = PDE(file_inode(file));
208 ssize_t rv = -EIO;
209 if (use_pde(pde)) {
210 read = pde->proc_fops->read;
211 if (read)
212 rv = read(file, buf, count, ppos);
213 unuse_pde(pde);
214 }
215 return rv;
216}
217
218static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
219{
220 ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
221 struct proc_dir_entry *pde = PDE(file_inode(file));
222 ssize_t rv = -EIO;
223 if (use_pde(pde)) {
224 write = pde->proc_fops->write;
225 if (write)
226 rv = write(file, buf, count, ppos);
227 unuse_pde(pde);
228 }
229 return rv;
230}
231
232static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
233{
234 struct proc_dir_entry *pde = PDE(file_inode(file));
235 unsigned int rv = DEFAULT_POLLMASK;
236 unsigned int (*poll)(struct file *, struct poll_table_struct *);
237 if (use_pde(pde)) {
238 poll = pde->proc_fops->poll;
239 if (poll)
240 rv = poll(file, pts);
241 unuse_pde(pde);
242 }
243 return rv;
244}
245
246static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
247{
248 struct proc_dir_entry *pde = PDE(file_inode(file));
249 long rv = -ENOTTY;
250 long (*ioctl)(struct file *, unsigned int, unsigned long);
251 if (use_pde(pde)) {
252 ioctl = pde->proc_fops->unlocked_ioctl;
253 if (ioctl)
254 rv = ioctl(file, cmd, arg);
255 unuse_pde(pde);
256 }
257 return rv;
258}
259
260#ifdef CONFIG_COMPAT
261static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
262{
263 struct proc_dir_entry *pde = PDE(file_inode(file));
264 long rv = -ENOTTY;
265 long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
266 if (use_pde(pde)) {
267 compat_ioctl = pde->proc_fops->compat_ioctl;
268 if (compat_ioctl)
269 rv = compat_ioctl(file, cmd, arg);
270 unuse_pde(pde);
271 }
272 return rv;
273}
274#endif
275
276static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
277{
278 struct proc_dir_entry *pde = PDE(file_inode(file));
279 int rv = -EIO;
280 int (*mmap)(struct file *, struct vm_area_struct *);
281 if (use_pde(pde)) {
282 mmap = pde->proc_fops->mmap;
283 if (mmap)
284 rv = mmap(file, vma);
285 unuse_pde(pde);
286 }
287 return rv;
288}
289
290static int proc_reg_open(struct inode *inode, struct file *file)
291{
292 struct proc_dir_entry *pde = PDE(inode);
293 int rv = 0;
294 int (*open)(struct inode *, struct file *);
295 int (*release)(struct inode *, struct file *);
296 struct pde_opener *pdeo;
297
298
299
300
301
302
303
304
305
306
307
308 pdeo = kzalloc(sizeof(struct pde_opener), GFP_KERNEL);
309 if (!pdeo)
310 return -ENOMEM;
311
312 if (!use_pde(pde)) {
313 kfree(pdeo);
314 return -ENOENT;
315 }
316 open = pde->proc_fops->open;
317 release = pde->proc_fops->release;
318
319 if (open)
320 rv = open(inode, file);
321
322 if (rv == 0 && release) {
323
324 pdeo->file = file;
325
326 spin_lock(&pde->pde_unload_lock);
327 list_add(&pdeo->lh, &pde->pde_openers);
328 spin_unlock(&pde->pde_unload_lock);
329 } else
330 kfree(pdeo);
331
332 unuse_pde(pde);
333 return rv;
334}
335
336static int proc_reg_release(struct inode *inode, struct file *file)
337{
338 struct proc_dir_entry *pde = PDE(inode);
339 struct pde_opener *pdeo;
340 spin_lock(&pde->pde_unload_lock);
341 list_for_each_entry(pdeo, &pde->pde_openers, lh) {
342 if (pdeo->file == file) {
343 close_pdeo(pde, pdeo);
344 break;
345 }
346 }
347 spin_unlock(&pde->pde_unload_lock);
348 return 0;
349}
350
351static const struct file_operations proc_reg_file_ops = {
352 .llseek = proc_reg_llseek,
353 .read = proc_reg_read,
354 .write = proc_reg_write,
355 .poll = proc_reg_poll,
356 .unlocked_ioctl = proc_reg_unlocked_ioctl,
357#ifdef CONFIG_COMPAT
358 .compat_ioctl = proc_reg_compat_ioctl,
359#endif
360 .mmap = proc_reg_mmap,
361 .open = proc_reg_open,
362 .release = proc_reg_release,
363};
364
365#ifdef CONFIG_COMPAT
366static const struct file_operations proc_reg_file_ops_no_compat = {
367 .llseek = proc_reg_llseek,
368 .read = proc_reg_read,
369 .write = proc_reg_write,
370 .poll = proc_reg_poll,
371 .unlocked_ioctl = proc_reg_unlocked_ioctl,
372 .mmap = proc_reg_mmap,
373 .open = proc_reg_open,
374 .release = proc_reg_release,
375};
376#endif
377
378struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
379{
380 struct inode *inode = new_inode_pseudo(sb);
381
382 if (inode) {
383 inode->i_ino = de->low_ino;
384 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
385 PROC_I(inode)->pde = de;
386
387 if (is_empty_pde(de)) {
388 make_empty_dir_inode(inode);
389 return inode;
390 }
391 if (de->mode) {
392 inode->i_mode = de->mode;
393 inode->i_uid = de->uid;
394 inode->i_gid = de->gid;
395 }
396 if (de->size)
397 inode->i_size = de->size;
398 if (de->nlink)
399 set_nlink(inode, de->nlink);
400 WARN_ON(!de->proc_iops);
401 inode->i_op = de->proc_iops;
402 if (de->proc_fops) {
403 if (S_ISREG(inode->i_mode)) {
404#ifdef CONFIG_COMPAT
405 if (!de->proc_fops->compat_ioctl)
406 inode->i_fop =
407 &proc_reg_file_ops_no_compat;
408 else
409#endif
410 inode->i_fop = &proc_reg_file_ops;
411 } else {
412 inode->i_fop = de->proc_fops;
413 }
414 }
415 } else
416 pde_put(de);
417 return inode;
418}
419
420int proc_fill_super(struct super_block *s, void *data, int silent)
421{
422 struct pid_namespace *ns = get_pid_ns(s->s_fs_info);
423 struct inode *root_inode;
424
425 if (!proc_parse_options(data, ns))
426 return -EINVAL;
427
428
429 s->s_iflags |= SB_I_USERNS_VISIBLE | SB_I_NOEXEC | SB_I_NODEV;
430 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
431 s->s_blocksize = 1024;
432 s->s_blocksize_bits = 10;
433 s->s_magic = PROC_SUPER_MAGIC;
434 s->s_op = &proc_sops;
435 s->s_time_gran = 1;
436
437 pde_get(&proc_root);
438 root_inode = proc_get_inode(s, &proc_root);
439 if (!root_inode) {
440 pr_err("proc_fill_super: get root inode failed\n");
441 return -ENOMEM;
442 }
443
444 s->s_root = d_make_root(root_inode);
445 if (!s->s_root) {
446 pr_err("proc_fill_super: allocate dentry failed\n");
447 return -ENOMEM;
448 }
449
450 return proc_setup_self(s);
451}
452
