1 /* 2 * proc/fs/generic.c --- generic routines for the proc-fs 3 * 4 * This file contains generic proc-fs routines for handling 5 * directories and files. 6 * 7 * Copyright (C) 1991, 1992 Linus Torvalds. 8 * Copyright (C) 1997 Theodore Ts'o 9 */ 10 11 #include <linux/errno.h> 12 #include <linux/time.h> 13 #include <linux/proc_fs.h> 14 #include <linux/stat.h> 15 #include <linux/mm.h> 16 #include <linux/module.h> 17 #include <linux/slab.h> 18 #include <linux/printk.h> 19 #include <linux/mount.h> 20 #include <linux/init.h> 21 #include <linux/idr.h> 22 #include <linux/namei.h> 23 #include <linux/bitops.h> 24 #include <linux/spinlock.h> 25 #include <linux/completion.h> 26 #include <asm/uaccess.h> 27 28 #include "internal.h" 29 30 static DEFINE_SPINLOCK(proc_subdir_lock); 31 32 static int proc_match(unsigned int len, const char *name, struct proc_dir_entry *de) 33 { 34 if (de->namelen != len) 35 return 0; 36 return !memcmp(name, de->name, len); 37 } 38 39 static int proc_notify_change(struct dentry *dentry, struct iattr *iattr) 40 { 41 struct inode *inode = dentry->d_inode; 42 struct proc_dir_entry *de = PDE(inode); 43 int error; 44 45 error = inode_change_ok(inode, iattr); 46 if (error) 47 return error; 48 49 setattr_copy(inode, iattr); 50 mark_inode_dirty(inode); 51 52 proc_set_user(de, inode->i_uid, inode->i_gid); 53 de->mode = inode->i_mode; 54 return 0; 55 } 56 57 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry, 58 struct kstat *stat) 59 { 60 struct inode *inode = dentry->d_inode; 61 struct proc_dir_entry *de = PROC_I(inode)->pde; 62 if (de && de->nlink) 63 set_nlink(inode, de->nlink); 64 65 generic_fillattr(inode, stat); 66 return 0; 67 } 68 69 static const struct inode_operations proc_file_inode_operations = { 70 .setattr = proc_notify_change, 71 }; 72 73 /* 74 * This function parses a name such as "tty/driver/serial", and 75 * returns the struct proc_dir_entry for "/proc/tty/driver", and 76 * returns "serial" in residual. 77 */ 78 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret, 79 const char **residual) 80 { 81 const char *cp = name, *next; 82 struct proc_dir_entry *de; 83 unsigned int len; 84 85 de = *ret; 86 if (!de) 87 de = &proc_root; 88 89 while (1) { 90 next = strchr(cp, '/'); 91 if (!next) 92 break; 93 94 len = next - cp; 95 for (de = de->subdir; de ; de = de->next) { 96 if (proc_match(len, cp, de)) 97 break; 98 } 99 if (!de) { 100 WARN(1, "name '%s'\n", name); 101 return -ENOENT; 102 } 103 cp += len + 1; 104 } 105 *residual = cp; 106 *ret = de; 107 return 0; 108 } 109 110 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret, 111 const char **residual) 112 { 113 int rv; 114 115 spin_lock(&proc_subdir_lock); 116 rv = __xlate_proc_name(name, ret, residual); 117 spin_unlock(&proc_subdir_lock); 118 return rv; 119 } 120 121 static DEFINE_IDA(proc_inum_ida); 122 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */ 123 124 #define PROC_DYNAMIC_FIRST 0xF0000000U 125 126 /* 127 * Return an inode number between PROC_DYNAMIC_FIRST and 128 * 0xffffffff, or zero on failure. 129 */ 130 int proc_alloc_inum(unsigned int *inum) 131 { 132 unsigned int i; 133 int error; 134 135 retry: 136 if (!ida_pre_get(&proc_inum_ida, GFP_KERNEL)) 137 return -ENOMEM; 138 139 spin_lock_irq(&proc_inum_lock); 140 error = ida_get_new(&proc_inum_ida, &i); 141 spin_unlock_irq(&proc_inum_lock); 142 if (error == -EAGAIN) 143 goto retry; 144 else if (error) 145 return error; 146 147 if (i > UINT_MAX - PROC_DYNAMIC_FIRST) { 148 spin_lock_irq(&proc_inum_lock); 149 ida_remove(&proc_inum_ida, i); 150 spin_unlock_irq(&proc_inum_lock); 151 return -ENOSPC; 152 } 153 *inum = PROC_DYNAMIC_FIRST + i; 154 return 0; 155 } 156 157 void proc_free_inum(unsigned int inum) 158 { 159 unsigned long flags; 160 spin_lock_irqsave(&proc_inum_lock, flags); 161 ida_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST); 162 spin_unlock_irqrestore(&proc_inum_lock, flags); 163 } 164 165 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd) 166 { 167 nd_set_link(nd, __PDE_DATA(dentry->d_inode)); 168 return NULL; 169 } 170 171 static const struct inode_operations proc_link_inode_operations = { 172 .readlink = generic_readlink, 173 .follow_link = proc_follow_link, 174 }; 175 176 /* 177 * Don't create negative dentries here, return -ENOENT by hand 178 * instead. 179 */ 180 struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir, 181 struct dentry *dentry) 182 { 183 struct inode *inode; 184 185 spin_lock(&proc_subdir_lock); 186 for (de = de->subdir; de ; de = de->next) { 187 if (de->namelen != dentry->d_name.len) 188 continue; 189 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) { 190 pde_get(de); 191 spin_unlock(&proc_subdir_lock); 192 inode = proc_get_inode(dir->i_sb, de); 193 if (!inode) 194 return ERR_PTR(-ENOMEM); 195 d_set_d_op(dentry, &simple_dentry_operations); 196 d_add(dentry, inode); 197 return NULL; 198 } 199 } 200 spin_unlock(&proc_subdir_lock); 201 return ERR_PTR(-ENOENT); 202 } 203 204 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry, 205 unsigned int flags) 206 { 207 return proc_lookup_de(PDE(dir), dir, dentry); 208 } 209 210 /* 211 * This returns non-zero if at EOF, so that the /proc 212 * root directory can use this and check if it should 213 * continue with the <pid> entries.. 214 * 215 * Note that the VFS-layer doesn't care about the return 216 * value of the readdir() call, as long as it's non-negative 217 * for success.. 218 */ 219 int proc_readdir_de(struct proc_dir_entry *de, struct file *file, 220 struct dir_context *ctx) 221 { 222 int i; 223 224 if (!dir_emit_dots(file, ctx)) 225 return 0; 226 227 spin_lock(&proc_subdir_lock); 228 de = de->subdir; 229 i = ctx->pos - 2; 230 for (;;) { 231 if (!de) { 232 spin_unlock(&proc_subdir_lock); 233 return 0; 234 } 235 if (!i) 236 break; 237 de = de->next; 238 i--; 239 } 240 241 do { 242 struct proc_dir_entry *next; 243 pde_get(de); 244 spin_unlock(&proc_subdir_lock); 245 if (!dir_emit(ctx, de->name, de->namelen, 246 de->low_ino, de->mode >> 12)) { 247 pde_put(de); 248 return 0; 249 } 250 spin_lock(&proc_subdir_lock); 251 ctx->pos++; 252 next = de->next; 253 pde_put(de); 254 de = next; 255 } while (de); 256 spin_unlock(&proc_subdir_lock); 257 return 1; 258 } 259 260 int proc_readdir(struct file *file, struct dir_context *ctx) 261 { 262 struct inode *inode = file_inode(file); 263 264 return proc_readdir_de(PDE(inode), file, ctx); 265 } 266 267 /* 268 * These are the generic /proc directory operations. They 269 * use the in-memory "struct proc_dir_entry" tree to parse 270 * the /proc directory. 271 */ 272 static const struct file_operations proc_dir_operations = { 273 .llseek = generic_file_llseek, 274 .read = generic_read_dir, 275 .iterate = proc_readdir, 276 }; 277 278 /* 279 * proc directories can do almost nothing.. 280 */ 281 static const struct inode_operations proc_dir_inode_operations = { 282 .lookup = proc_lookup, 283 .getattr = proc_getattr, 284 .setattr = proc_notify_change, 285 }; 286 287 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp) 288 { 289 struct proc_dir_entry *tmp; 290 int ret; 291 292 ret = proc_alloc_inum(&dp->low_ino); 293 if (ret) 294 return ret; 295 296 if (S_ISDIR(dp->mode)) { 297 dp->proc_fops = &proc_dir_operations; 298 dp->proc_iops = &proc_dir_inode_operations; 299 dir->nlink++; 300 } else if (S_ISLNK(dp->mode)) { 301 dp->proc_iops = &proc_link_inode_operations; 302 } else if (S_ISREG(dp->mode)) { 303 BUG_ON(dp->proc_fops == NULL); 304 dp->proc_iops = &proc_file_inode_operations; 305 } else { 306 WARN_ON(1); 307 return -EINVAL; 308 } 309 310 spin_lock(&proc_subdir_lock); 311 312 for (tmp = dir->subdir; tmp; tmp = tmp->next) 313 if (strcmp(tmp->name, dp->name) == 0) { 314 WARN(1, "proc_dir_entry '%s/%s' already registered\n", 315 dir->name, dp->name); 316 break; 317 } 318 319 dp->next = dir->subdir; 320 dp->parent = dir; 321 dir->subdir = dp; 322 spin_unlock(&proc_subdir_lock); 323 324 return 0; 325 } 326 327 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent, 328 const char *name, 329 umode_t mode, 330 nlink_t nlink) 331 { 332 struct proc_dir_entry *ent = NULL; 333 const char *fn; 334 struct qstr qstr; 335 336 if (xlate_proc_name(name, parent, &fn) != 0) 337 goto out; 338 qstr.name = fn; 339 qstr.len = strlen(fn); 340 if (qstr.len == 0 || qstr.len >= 256) { 341 WARN(1, "name len %u\n", qstr.len); 342 return NULL; 343 } 344 if (*parent == &proc_root && name_to_int(&qstr) != ~0U) { 345 WARN(1, "create '/proc/%s' by hand\n", qstr.name); 346 return NULL; 347 } 348 349 ent = kzalloc(sizeof(struct proc_dir_entry) + qstr.len + 1, GFP_KERNEL); 350 if (!ent) 351 goto out; 352 353 memcpy(ent->name, fn, qstr.len + 1); 354 ent->namelen = qstr.len; 355 ent->mode = mode; 356 ent->nlink = nlink; 357 atomic_set(&ent->count, 1); 358 spin_lock_init(&ent->pde_unload_lock); 359 INIT_LIST_HEAD(&ent->pde_openers); 360 out: 361 return ent; 362 } 363 364 struct proc_dir_entry *proc_symlink(const char *name, 365 struct proc_dir_entry *parent, const char *dest) 366 { 367 struct proc_dir_entry *ent; 368 369 ent = __proc_create(&parent, name, 370 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1); 371 372 if (ent) { 373 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL); 374 if (ent->data) { 375 strcpy((char*)ent->data,dest); 376 if (proc_register(parent, ent) < 0) { 377 kfree(ent->data); 378 kfree(ent); 379 ent = NULL; 380 } 381 } else { 382 kfree(ent); 383 ent = NULL; 384 } 385 } 386 return ent; 387 } 388 EXPORT_SYMBOL(proc_symlink); 389 390 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode, 391 struct proc_dir_entry *parent, void *data) 392 { 393 struct proc_dir_entry *ent; 394 395 if (mode == 0) 396 mode = S_IRUGO | S_IXUGO; 397 398 ent = __proc_create(&parent, name, S_IFDIR | mode, 2); 399 if (ent) { 400 ent->data = data; 401 if (proc_register(parent, ent) < 0) { 402 kfree(ent); 403 ent = NULL; 404 } 405 } 406 return ent; 407 } 408 EXPORT_SYMBOL_GPL(proc_mkdir_data); 409 410 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode, 411 struct proc_dir_entry *parent) 412 { 413 return proc_mkdir_data(name, mode, parent, NULL); 414 } 415 EXPORT_SYMBOL(proc_mkdir_mode); 416 417 struct proc_dir_entry *proc_mkdir(const char *name, 418 struct proc_dir_entry *parent) 419 { 420 return proc_mkdir_data(name, 0, parent, NULL); 421 } 422 EXPORT_SYMBOL(proc_mkdir); 423 424 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode, 425 struct proc_dir_entry *parent, 426 const struct file_operations *proc_fops, 427 void *data) 428 { 429 struct proc_dir_entry *pde; 430 if ((mode & S_IFMT) == 0) 431 mode |= S_IFREG; 432 433 if (!S_ISREG(mode)) { 434 WARN_ON(1); /* use proc_mkdir() */ 435 return NULL; 436 } 437 438 if ((mode & S_IALLUGO) == 0) 439 mode |= S_IRUGO; 440 pde = __proc_create(&parent, name, mode, 1); 441 if (!pde) 442 goto out; 443 pde->proc_fops = proc_fops; 444 pde->data = data; 445 if (proc_register(parent, pde) < 0) 446 goto out_free; 447 return pde; 448 out_free: 449 kfree(pde); 450 out: 451 return NULL; 452 } 453 EXPORT_SYMBOL(proc_create_data); 454 455 void proc_set_size(struct proc_dir_entry *de, loff_t size) 456 { 457 de->size = size; 458 } 459 EXPORT_SYMBOL(proc_set_size); 460 461 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid) 462 { 463 de->uid = uid; 464 de->gid = gid; 465 } 466 EXPORT_SYMBOL(proc_set_user); 467 468 static void free_proc_entry(struct proc_dir_entry *de) 469 { 470 proc_free_inum(de->low_ino); 471 472 if (S_ISLNK(de->mode)) 473 kfree(de->data); 474 kfree(de); 475 } 476 477 void pde_put(struct proc_dir_entry *pde) 478 { 479 if (atomic_dec_and_test(&pde->count)) 480 free_proc_entry(pde); 481 } 482 483 /* 484 * Remove a /proc entry and free it if it's not currently in use. 485 */ 486 void remove_proc_entry(const char *name, struct proc_dir_entry *parent) 487 { 488 struct proc_dir_entry **p; 489 struct proc_dir_entry *de = NULL; 490 const char *fn = name; 491 unsigned int len; 492 493 spin_lock(&proc_subdir_lock); 494 if (__xlate_proc_name(name, &parent, &fn) != 0) { 495 spin_unlock(&proc_subdir_lock); 496 return; 497 } 498 len = strlen(fn); 499 500 for (p = &parent->subdir; *p; p=&(*p)->next ) { 501 if (proc_match(len, fn, *p)) { 502 de = *p; 503 *p = de->next; 504 de->next = NULL; 505 break; 506 } 507 } 508 spin_unlock(&proc_subdir_lock); 509 if (!de) { 510 WARN(1, "name '%s'\n", name); 511 return; 512 } 513 514 proc_entry_rundown(de); 515 516 if (S_ISDIR(de->mode)) 517 parent->nlink--; 518 de->nlink = 0; 519 WARN(de->subdir, "%s: removing non-empty directory " 520 "'%s/%s', leaking at least '%s'\n", __func__, 521 de->parent->name, de->name, de->subdir->name); 522 pde_put(de); 523 } 524 EXPORT_SYMBOL(remove_proc_entry); 525 526 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent) 527 { 528 struct proc_dir_entry **p; 529 struct proc_dir_entry *root = NULL, *de, *next; 530 const char *fn = name; 531 unsigned int len; 532 533 spin_lock(&proc_subdir_lock); 534 if (__xlate_proc_name(name, &parent, &fn) != 0) { 535 spin_unlock(&proc_subdir_lock); 536 return -ENOENT; 537 } 538 len = strlen(fn); 539 540 for (p = &parent->subdir; *p; p=&(*p)->next ) { 541 if (proc_match(len, fn, *p)) { 542 root = *p; 543 *p = root->next; 544 root->next = NULL; 545 break; 546 } 547 } 548 if (!root) { 549 spin_unlock(&proc_subdir_lock); 550 return -ENOENT; 551 } 552 de = root; 553 while (1) { 554 next = de->subdir; 555 if (next) { 556 de->subdir = next->next; 557 next->next = NULL; 558 de = next; 559 continue; 560 } 561 spin_unlock(&proc_subdir_lock); 562 563 proc_entry_rundown(de); 564 next = de->parent; 565 if (S_ISDIR(de->mode)) 566 next->nlink--; 567 de->nlink = 0; 568 if (de == root) 569 break; 570 pde_put(de); 571 572 spin_lock(&proc_subdir_lock); 573 de = next; 574 } 575 pde_put(root); 576 return 0; 577 } 578 EXPORT_SYMBOL(remove_proc_subtree); 579 580 void *proc_get_parent_data(const struct inode *inode) 581 { 582 struct proc_dir_entry *de = PDE(inode); 583 return de->parent->data; 584 } 585 EXPORT_SYMBOL_GPL(proc_get_parent_data); 586 587 void proc_remove(struct proc_dir_entry *de) 588 { 589 if (de) 590 remove_proc_subtree(de->name, de->parent); 591 } 592 EXPORT_SYMBOL(proc_remove); 593 594 void *PDE_DATA(const struct inode *inode) 595 { 596 return __PDE_DATA(inode); 597 } 598 EXPORT_SYMBOL(PDE_DATA); 599