inode.c 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502
  1. /*
  2. * linux/fs/proc/inode.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. #include <linux/time.h>
  7. #include <linux/proc_fs.h>
  8. #include <linux/kernel.h>
  9. #include <linux/pid_namespace.h>
  10. #include <linux/mm.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/seq_file.h>
  21. #include <linux/slab.h>
  22. #include <linux/mount.h>
  23. #include <asm/uaccess.h>
  24. #include "internal.h"
  25. static void proc_evict_inode(struct inode *inode)
  26. {
  27. struct proc_dir_entry *de;
  28. struct ctl_table_header *head;
  29. const struct proc_ns_operations *ns_ops;
  30. void *ns;
  31. truncate_inode_pages(&inode->i_data, 0);
  32. end_writeback(inode);
  33. /* Stop tracking associated processes */
  34. put_pid(PROC_I(inode)->pid);
  35. /* Let go of any associated proc directory entry */
  36. de = PROC_I(inode)->pde;
  37. if (de)
  38. pde_put(de);
  39. head = PROC_I(inode)->sysctl;
  40. if (head) {
  41. rcu_assign_pointer(PROC_I(inode)->sysctl, NULL);
  42. sysctl_head_put(head);
  43. }
  44. /* Release any associated namespace */
  45. ns_ops = PROC_I(inode)->ns_ops;
  46. ns = PROC_I(inode)->ns;
  47. if (ns_ops && ns)
  48. ns_ops->put(ns);
  49. }
  50. static struct kmem_cache * proc_inode_cachep;
  51. static struct inode *proc_alloc_inode(struct super_block *sb)
  52. {
  53. struct proc_inode *ei;
  54. struct inode *inode;
  55. ei = (struct proc_inode *)kmem_cache_alloc(proc_inode_cachep, GFP_KERNEL);
  56. if (!ei)
  57. return NULL;
  58. ei->pid = NULL;
  59. ei->fd = 0;
  60. ei->op.proc_get_link = NULL;
  61. ei->pde = NULL;
  62. ei->sysctl = NULL;
  63. ei->sysctl_entry = NULL;
  64. ei->ns = NULL;
  65. ei->ns_ops = NULL;
  66. inode = &ei->vfs_inode;
  67. inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
  68. return inode;
  69. }
  70. static void proc_i_callback(struct rcu_head *head)
  71. {
  72. struct inode *inode = container_of(head, struct inode, i_rcu);
  73. kmem_cache_free(proc_inode_cachep, PROC_I(inode));
  74. }
  75. static void proc_destroy_inode(struct inode *inode)
  76. {
  77. call_rcu(&inode->i_rcu, proc_i_callback);
  78. }
  79. static void init_once(void *foo)
  80. {
  81. struct proc_inode *ei = (struct proc_inode *) foo;
  82. inode_init_once(&ei->vfs_inode);
  83. }
  84. void __init proc_init_inodecache(void)
  85. {
  86. proc_inode_cachep = kmem_cache_create("proc_inode_cache",
  87. sizeof(struct proc_inode),
  88. 0, (SLAB_RECLAIM_ACCOUNT|
  89. SLAB_MEM_SPREAD|SLAB_PANIC),
  90. init_once);
  91. }
  92. static int proc_show_options(struct seq_file *seq, struct dentry *root)
  93. {
  94. struct super_block *sb = root->d_sb;
  95. struct pid_namespace *pid = sb->s_fs_info;
  96. if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
  97. seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
  98. if (pid->hide_pid != 0)
  99. seq_printf(seq, ",hidepid=%u", pid->hide_pid);
  100. return 0;
  101. }
  102. static const struct super_operations proc_sops = {
  103. .alloc_inode = proc_alloc_inode,
  104. .destroy_inode = proc_destroy_inode,
  105. .drop_inode = generic_delete_inode,
  106. .evict_inode = proc_evict_inode,
  107. .statfs = simple_statfs,
  108. .remount_fs = proc_remount,
  109. .show_options = proc_show_options,
  110. };
  111. static void __pde_users_dec(struct proc_dir_entry *pde)
  112. {
  113. pde->pde_users--;
  114. if (pde->pde_unload_completion && pde->pde_users == 0)
  115. complete(pde->pde_unload_completion);
  116. }
  117. void pde_users_dec(struct proc_dir_entry *pde)
  118. {
  119. spin_lock(&pde->pde_unload_lock);
  120. __pde_users_dec(pde);
  121. spin_unlock(&pde->pde_unload_lock);
  122. }
  123. static loff_t proc_reg_llseek(struct file *file, loff_t offset, int whence)
  124. {
  125. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  126. loff_t rv = -EINVAL;
  127. loff_t (*llseek)(struct file *, loff_t, int);
  128. spin_lock(&pde->pde_unload_lock);
  129. /*
  130. * remove_proc_entry() is going to delete PDE (as part of module
  131. * cleanup sequence). No new callers into module allowed.
  132. */
  133. if (!pde->proc_fops) {
  134. spin_unlock(&pde->pde_unload_lock);
  135. return rv;
  136. }
  137. /*
  138. * Bump refcount so that remove_proc_entry will wail for ->llseek to
  139. * complete.
  140. */
  141. pde->pde_users++;
  142. /*
  143. * Save function pointer under lock, to protect against ->proc_fops
  144. * NULL'ifying right after ->pde_unload_lock is dropped.
  145. */
  146. llseek = pde->proc_fops->llseek;
  147. spin_unlock(&pde->pde_unload_lock);
  148. if (!llseek)
  149. llseek = default_llseek;
  150. rv = llseek(file, offset, whence);
  151. pde_users_dec(pde);
  152. return rv;
  153. }
  154. static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
  155. {
  156. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  157. ssize_t rv = -EIO;
  158. ssize_t (*read)(struct file *, char __user *, size_t, loff_t *);
  159. spin_lock(&pde->pde_unload_lock);
  160. if (!pde->proc_fops) {
  161. spin_unlock(&pde->pde_unload_lock);
  162. return rv;
  163. }
  164. pde->pde_users++;
  165. read = pde->proc_fops->read;
  166. spin_unlock(&pde->pde_unload_lock);
  167. if (read)
  168. rv = read(file, buf, count, ppos);
  169. pde_users_dec(pde);
  170. return rv;
  171. }
  172. static ssize_t proc_reg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
  173. {
  174. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  175. ssize_t rv = -EIO;
  176. ssize_t (*write)(struct file *, const char __user *, size_t, loff_t *);
  177. spin_lock(&pde->pde_unload_lock);
  178. if (!pde->proc_fops) {
  179. spin_unlock(&pde->pde_unload_lock);
  180. return rv;
  181. }
  182. pde->pde_users++;
  183. write = pde->proc_fops->write;
  184. spin_unlock(&pde->pde_unload_lock);
  185. if (write)
  186. rv = write(file, buf, count, ppos);
  187. pde_users_dec(pde);
  188. return rv;
  189. }
  190. static unsigned int proc_reg_poll(struct file *file, struct poll_table_struct *pts)
  191. {
  192. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  193. unsigned int rv = DEFAULT_POLLMASK;
  194. unsigned int (*poll)(struct file *, struct poll_table_struct *);
  195. spin_lock(&pde->pde_unload_lock);
  196. if (!pde->proc_fops) {
  197. spin_unlock(&pde->pde_unload_lock);
  198. return rv;
  199. }
  200. pde->pde_users++;
  201. poll = pde->proc_fops->poll;
  202. spin_unlock(&pde->pde_unload_lock);
  203. if (poll)
  204. rv = poll(file, pts);
  205. pde_users_dec(pde);
  206. return rv;
  207. }
  208. static long proc_reg_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
  209. {
  210. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  211. long rv = -ENOTTY;
  212. long (*ioctl)(struct file *, unsigned int, unsigned long);
  213. spin_lock(&pde->pde_unload_lock);
  214. if (!pde->proc_fops) {
  215. spin_unlock(&pde->pde_unload_lock);
  216. return rv;
  217. }
  218. pde->pde_users++;
  219. ioctl = pde->proc_fops->unlocked_ioctl;
  220. spin_unlock(&pde->pde_unload_lock);
  221. if (ioctl)
  222. rv = ioctl(file, cmd, arg);
  223. pde_users_dec(pde);
  224. return rv;
  225. }
  226. #ifdef CONFIG_COMPAT
  227. static long proc_reg_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
  228. {
  229. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  230. long rv = -ENOTTY;
  231. long (*compat_ioctl)(struct file *, unsigned int, unsigned long);
  232. spin_lock(&pde->pde_unload_lock);
  233. if (!pde->proc_fops) {
  234. spin_unlock(&pde->pde_unload_lock);
  235. return rv;
  236. }
  237. pde->pde_users++;
  238. compat_ioctl = pde->proc_fops->compat_ioctl;
  239. spin_unlock(&pde->pde_unload_lock);
  240. if (compat_ioctl)
  241. rv = compat_ioctl(file, cmd, arg);
  242. pde_users_dec(pde);
  243. return rv;
  244. }
  245. #endif
  246. static int proc_reg_mmap(struct file *file, struct vm_area_struct *vma)
  247. {
  248. struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
  249. int rv = -EIO;
  250. int (*mmap)(struct file *, struct vm_area_struct *);
  251. spin_lock(&pde->pde_unload_lock);
  252. if (!pde->proc_fops) {
  253. spin_unlock(&pde->pde_unload_lock);
  254. return rv;
  255. }
  256. pde->pde_users++;
  257. mmap = pde->proc_fops->mmap;
  258. spin_unlock(&pde->pde_unload_lock);
  259. if (mmap)
  260. rv = mmap(file, vma);
  261. pde_users_dec(pde);
  262. return rv;
  263. }
  264. static int proc_reg_open(struct inode *inode, struct file *file)
  265. {
  266. struct proc_dir_entry *pde = PDE(inode);
  267. int rv = 0;
  268. int (*open)(struct inode *, struct file *);
  269. int (*release)(struct inode *, struct file *);
  270. struct pde_opener *pdeo;
  271. /*
  272. * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
  273. * sequence. ->release won't be called because ->proc_fops will be
  274. * cleared. Depending on complexity of ->release, consequences vary.
  275. *
  276. * We can't wait for mercy when close will be done for real, it's
  277. * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
  278. * by hand in remove_proc_entry(). For this, save opener's credentials
  279. * for later.
  280. */
  281. pdeo = kmalloc(sizeof(struct pde_opener), GFP_KERNEL);
  282. if (!pdeo)
  283. return -ENOMEM;
  284. spin_lock(&pde->pde_unload_lock);
  285. if (!pde->proc_fops) {
  286. spin_unlock(&pde->pde_unload_lock);
  287. kfree(pdeo);
  288. return -ENOENT;
  289. }
  290. pde->pde_users++;
  291. open = pde->proc_fops->open;
  292. release = pde->proc_fops->release;
  293. spin_unlock(&pde->pde_unload_lock);
  294. if (open)
  295. rv = open(inode, file);
  296. spin_lock(&pde->pde_unload_lock);
  297. if (rv == 0 && release) {
  298. /* To know what to release. */
  299. pdeo->inode = inode;
  300. pdeo->file = file;
  301. /* Strictly for "too late" ->release in proc_reg_release(). */
  302. pdeo->release = release;
  303. list_add(&pdeo->lh, &pde->pde_openers);
  304. } else
  305. kfree(pdeo);
  306. __pde_users_dec(pde);
  307. spin_unlock(&pde->pde_unload_lock);
  308. return rv;
  309. }
  310. static struct pde_opener *find_pde_opener(struct proc_dir_entry *pde,
  311. struct inode *inode, struct file *file)
  312. {
  313. struct pde_opener *pdeo;
  314. list_for_each_entry(pdeo, &pde->pde_openers, lh) {
  315. if (pdeo->inode == inode && pdeo->file == file)
  316. return pdeo;
  317. }
  318. return NULL;
  319. }
  320. static int proc_reg_release(struct inode *inode, struct file *file)
  321. {
  322. struct proc_dir_entry *pde = PDE(inode);
  323. int rv = 0;
  324. int (*release)(struct inode *, struct file *);
  325. struct pde_opener *pdeo;
  326. spin_lock(&pde->pde_unload_lock);
  327. pdeo = find_pde_opener(pde, inode, file);
  328. if (!pde->proc_fops) {
  329. /*
  330. * Can't simply exit, __fput() will think that everything is OK,
  331. * and move on to freeing struct file. remove_proc_entry() will
  332. * find slacker in opener's list and will try to do non-trivial
  333. * things with struct file. Therefore, remove opener from list.
  334. *
  335. * But if opener is removed from list, who will ->release it?
  336. */
  337. if (pdeo) {
  338. list_del(&pdeo->lh);
  339. spin_unlock(&pde->pde_unload_lock);
  340. rv = pdeo->release(inode, file);
  341. kfree(pdeo);
  342. } else
  343. spin_unlock(&pde->pde_unload_lock);
  344. return rv;
  345. }
  346. pde->pde_users++;
  347. release = pde->proc_fops->release;
  348. if (pdeo) {
  349. list_del(&pdeo->lh);
  350. kfree(pdeo);
  351. }
  352. spin_unlock(&pde->pde_unload_lock);
  353. if (release)
  354. rv = release(inode, file);
  355. pde_users_dec(pde);
  356. return rv;
  357. }
  358. static const struct file_operations proc_reg_file_ops = {
  359. .llseek = proc_reg_llseek,
  360. .read = proc_reg_read,
  361. .write = proc_reg_write,
  362. .poll = proc_reg_poll,
  363. .unlocked_ioctl = proc_reg_unlocked_ioctl,
  364. #ifdef CONFIG_COMPAT
  365. .compat_ioctl = proc_reg_compat_ioctl,
  366. #endif
  367. .mmap = proc_reg_mmap,
  368. .open = proc_reg_open,
  369. .release = proc_reg_release,
  370. };
  371. #ifdef CONFIG_COMPAT
  372. static const struct file_operations proc_reg_file_ops_no_compat = {
  373. .llseek = proc_reg_llseek,
  374. .read = proc_reg_read,
  375. .write = proc_reg_write,
  376. .poll = proc_reg_poll,
  377. .unlocked_ioctl = proc_reg_unlocked_ioctl,
  378. .mmap = proc_reg_mmap,
  379. .open = proc_reg_open,
  380. .release = proc_reg_release,
  381. };
  382. #endif
  383. struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
  384. {
  385. struct inode *inode = new_inode_pseudo(sb);
  386. if (inode) {
  387. inode->i_ino = de->low_ino;
  388. inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
  389. PROC_I(inode)->fd = 0;
  390. PROC_I(inode)->pde = de;
  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. if (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. }
  416. return inode;
  417. }
  418. int proc_fill_super(struct super_block *s)
  419. {
  420. s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
  421. s->s_blocksize = 1024;
  422. s->s_blocksize_bits = 10;
  423. s->s_magic = PROC_SUPER_MAGIC;
  424. s->s_op = &proc_sops;
  425. s->s_time_gran = 1;
  426. pde_get(&proc_root);
  427. s->s_root = d_make_root(proc_get_inode(s, &proc_root));
  428. if (s->s_root)
  429. return 0;
  430. printk("proc_read_super: get root inode failed\n");
  431. pde_put(&proc_root);
  432. return -ENOMEM;
  433. }