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 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 de->uid = inode->i_uid; 53 de->gid = inode->i_gid; 54 de->mode = inode->i_mode; 55 return 0; 56 } 57 58 static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry, 59 struct kstat *stat) 60 { 61 struct inode *inode = dentry->d_inode; 62 struct proc_dir_entry *de = PROC_I(inode)->pde; 63 if (de && de->nlink) 64 set_nlink(inode, de->nlink); 65 66 generic_fillattr(inode, stat); 67 return 0; 68 } 69 70 static const struct inode_operations proc_file_inode_operations = { 71 .setattr = proc_notify_change, 72 }; 73 74 /* 75 * This function parses a name such as "tty/driver/serial", and 76 * returns the struct proc_dir_entry for "/proc/tty/driver", and 77 * returns "serial" in residual. 78 */ 79 static int __xlate_proc_name(const char *name, struct proc_dir_entry **ret, 80 const char **residual) 81 { 82 const char *cp = name, *next; 83 struct proc_dir_entry *de; 84 unsigned int len; 85 86 de = *ret; 87 if (!de) 88 de = &proc_root; 89 90 while (1) { 91 next = strchr(cp, '/'); 92 if (!next) 93 break; 94 95 len = next - cp; 96 for (de = de->subdir; de ; de = de->next) { 97 if (proc_match(len, cp, de)) 98 break; 99 } 100 if (!de) { 101 WARN(1, "name '%s'\n", name); 102 return -ENOENT; 103 } 104 cp += len + 1; 105 } 106 *residual = cp; 107 *ret = de; 108 return 0; 109 } 110 111 static int xlate_proc_name(const char *name, struct proc_dir_entry **ret, 112 const char **residual) 113 { 114 int rv; 115 116 spin_lock(&proc_subdir_lock); 117 rv = __xlate_proc_name(name, ret, residual); 118 spin_unlock(&proc_subdir_lock); 119 return rv; 120 } 121 122 static DEFINE_IDA(proc_inum_ida); 123 static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */ 124 125 #define PROC_DYNAMIC_FIRST 0xF0000000U 126 127 /* 128 * Return an inode number between PROC_DYNAMIC_FIRST and 129 * 0xffffffff, or zero on failure. 130 */ 131 int proc_alloc_inum(unsigned int *inum) 132 { 133 unsigned int i; 134 int error; 135 136 retry: 137 if (!ida_pre_get(&proc_inum_ida, GFP_KERNEL)) 138 return -ENOMEM; 139 140 spin_lock_irq(&proc_inum_lock); 141 error = ida_get_new(&proc_inum_ida, &i); 142 spin_unlock_irq(&proc_inum_lock); 143 if (error == -EAGAIN) 144 goto retry; 145 else if (error) 146 return error; 147 148 if (i > UINT_MAX - PROC_DYNAMIC_FIRST) { 149 spin_lock_irq(&proc_inum_lock); 150 ida_remove(&proc_inum_ida, i); 151 spin_unlock_irq(&proc_inum_lock); 152 return -ENOSPC; 153 } 154 *inum = PROC_DYNAMIC_FIRST + i; 155 return 0; 156 } 157 158 void proc_free_inum(unsigned int inum) 159 { 160 unsigned long flags; 161 spin_lock_irqsave(&proc_inum_lock, flags); 162 ida_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST); 163 spin_unlock_irqrestore(&proc_inum_lock, flags); 164 } 165 166 static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd) 167 { 168 nd_set_link(nd, __PDE_DATA(dentry->d_inode)); 169 return NULL; 170 } 171 172 static const struct inode_operations proc_link_inode_operations = { 173 .readlink = generic_readlink, 174 .follow_link = proc_follow_link, 175 }; 176 177 /* 178 * Don't create negative dentries here, return -ENOENT by hand 179 * instead. 180 */ 181 struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir, 182 struct dentry *dentry) 183 { 184 struct inode *inode; 185 186 spin_lock(&proc_subdir_lock); 187 for (de = de->subdir; de ; de = de->next) { 188 if (de->namelen != dentry->d_name.len) 189 continue; 190 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) { 191 pde_get(de); 192 spin_unlock(&proc_subdir_lock); 193 inode = proc_get_inode(dir->i_sb, de); 194 if (!inode) 195 return ERR_PTR(-ENOMEM); 196 d_set_d_op(dentry, &simple_dentry_operations); 197 d_add(dentry, inode); 198 return NULL; 199 } 200 } 201 spin_unlock(&proc_subdir_lock); 202 return ERR_PTR(-ENOENT); 203 } 204 205 struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry, 206 unsigned int flags) 207 { 208 return proc_lookup_de(PDE(dir), dir, dentry); 209 } 210 211 /* 212 * This returns non-zero if at EOF, so that the /proc 213 * root directory can use this and check if it should 214 * continue with the <pid> entries.. 215 * 216 * Note that the VFS-layer doesn't care about the return 217 * value of the readdir() call, as long as it's non-negative 218 * for success.. 219 */ 220 int proc_readdir_de(struct proc_dir_entry *de, struct file *file, 221 struct dir_context *ctx) 222 { 223 int i; 224 225 if (!dir_emit_dots(file, ctx)) 226 return 0; 227 228 spin_lock(&proc_subdir_lock); 229 de = de->subdir; 230 i = ctx->pos - 2; 231 for (;;) { 232 if (!de) { 233 spin_unlock(&proc_subdir_lock); 234 return 0; 235 } 236 if (!i) 237 break; 238 de = de->next; 239 i--; 240 } 241 242 do { 243 struct proc_dir_entry *next; 244 pde_get(de); 245 spin_unlock(&proc_subdir_lock); 246 if (!dir_emit(ctx, de->name, de->namelen, 247 de->low_ino, de->mode >> 12)) { 248 pde_put(de); 249 return 0; 250 } 251 spin_lock(&proc_subdir_lock); 252 ctx->pos++; 253 next = de->next; 254 pde_put(de); 255 de = next; 256 } while (de); 257 spin_unlock(&proc_subdir_lock); 258 return 1; 259 } 260 261 int proc_readdir(struct file *file, struct dir_context *ctx) 262 { 263 struct inode *inode = file_inode(file); 264 265 return proc_readdir_de(PDE(inode), file, ctx); 266 } 267 268 /* 269 * These are the generic /proc directory operations. They 270 * use the in-memory "struct proc_dir_entry" tree to parse 271 * the /proc directory. 272 */ 273 static const struct file_operations proc_dir_operations = { 274 .llseek = generic_file_llseek, 275 .read = generic_read_dir, 276 .iterate = proc_readdir, 277 }; 278 279 /* 280 * proc directories can do almost nothing.. 281 */ 282 static const struct inode_operations proc_dir_inode_operations = { 283 .lookup = proc_lookup, 284 .getattr = proc_getattr, 285 .setattr = proc_notify_change, 286 }; 287 288 static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp) 289 { 290 struct proc_dir_entry *tmp; 291 int ret; 292 293 ret = proc_alloc_inum(&dp->low_ino); 294 if (ret) 295 return ret; 296 297 if (S_ISDIR(dp->mode)) { 298 dp->proc_fops = &proc_dir_operations; 299 dp->proc_iops = &proc_dir_inode_operations; 300 dir->nlink++; 301 } else if (S_ISLNK(dp->mode)) { 302 dp->proc_iops = &proc_link_inode_operations; 303 } else if (S_ISREG(dp->mode)) { 304 BUG_ON(dp->proc_fops == NULL); 305 dp->proc_iops = &proc_file_inode_operations; 306 } else { 307 WARN_ON(1); 308 return -EINVAL; 309 } 310 311 spin_lock(&proc_subdir_lock); 312 313 for (tmp = dir->subdir; tmp; tmp = tmp->next) 314 if (strcmp(tmp->name, dp->name) == 0) { 315 WARN(1, "proc_dir_entry '%s/%s' already registered\n", 316 dir->name, dp->name); 317 break; 318 } 319 320 dp->next = dir->subdir; 321 dp->parent = dir; 322 dir->subdir = dp; 323 spin_unlock(&proc_subdir_lock); 324 325 return 0; 326 } 327 328 static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent, 329 const char *name, 330 umode_t mode, 331 nlink_t nlink) 332 { 333 struct proc_dir_entry *ent = NULL; 334 const char *fn = name; 335 unsigned int len; 336 337 /* make sure name is valid */ 338 if (!name || !strlen(name)) 339 goto out; 340 341 if (xlate_proc_name(name, parent, &fn) != 0) 342 goto out; 343 344 /* At this point there must not be any '/' characters beyond *fn */ 345 if (strchr(fn, '/')) 346 goto out; 347 348 len = strlen(fn); 349 350 ent = kzalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL); 351 if (!ent) 352 goto out; 353 354 memcpy(ent->name, fn, len + 1); 355 ent->namelen = len; 356 ent->mode = mode; 357 ent->nlink = nlink; 358 atomic_set(&ent->count, 1); 359 spin_lock_init(&ent->pde_unload_lock); 360 INIT_LIST_HEAD(&ent->pde_openers); 361 out: 362 return ent; 363 } 364 365 struct proc_dir_entry *proc_symlink(const char *name, 366 struct proc_dir_entry *parent, const char *dest) 367 { 368 struct proc_dir_entry *ent; 369 370 ent = __proc_create(&parent, name, 371 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1); 372 373 if (ent) { 374 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL); 375 if (ent->data) { 376 strcpy((char*)ent->data,dest); 377 if (proc_register(parent, ent) < 0) { 378 kfree(ent->data); 379 kfree(ent); 380 ent = NULL; 381 } 382 } else { 383 kfree(ent); 384 ent = NULL; 385 } 386 } 387 return ent; 388 } 389 EXPORT_SYMBOL(proc_symlink); 390 391 struct proc_dir_entry *proc_mkdir_data(const char *name, umode_t mode, 392 struct proc_dir_entry *parent, void *data) 393 { 394 struct proc_dir_entry *ent; 395 396 if (mode == 0) 397 mode = S_IRUGO | S_IXUGO; 398 399 ent = __proc_create(&parent, name, S_IFDIR | mode, 2); 400 if (ent) { 401 ent->data = data; 402 if (proc_register(parent, ent) < 0) { 403 kfree(ent); 404 ent = NULL; 405 } 406 } 407 return ent; 408 } 409 EXPORT_SYMBOL_GPL(proc_mkdir_data); 410 411 struct proc_dir_entry *proc_mkdir_mode(const char *name, umode_t mode, 412 struct proc_dir_entry *parent) 413 { 414 return proc_mkdir_data(name, mode, parent, NULL); 415 } 416 EXPORT_SYMBOL(proc_mkdir_mode); 417 418 struct proc_dir_entry *proc_mkdir(const char *name, 419 struct proc_dir_entry *parent) 420 { 421 return proc_mkdir_data(name, 0, parent, NULL); 422 } 423 EXPORT_SYMBOL(proc_mkdir); 424 425 struct proc_dir_entry *proc_create_data(const char *name, umode_t mode, 426 struct proc_dir_entry *parent, 427 const struct file_operations *proc_fops, 428 void *data) 429 { 430 struct proc_dir_entry *pde; 431 if ((mode & S_IFMT) == 0) 432 mode |= S_IFREG; 433 434 if (!S_ISREG(mode)) { 435 WARN_ON(1); /* use proc_mkdir() */ 436 return NULL; 437 } 438 439 if ((mode & S_IALLUGO) == 0) 440 mode |= S_IRUGO; 441 pde = __proc_create(&parent, name, mode, 1); 442 if (!pde) 443 goto out; 444 pde->proc_fops = proc_fops; 445 pde->data = data; 446 if (proc_register(parent, pde) < 0) 447 goto out_free; 448 return pde; 449 out_free: 450 kfree(pde); 451 out: 452 return NULL; 453 } 454 EXPORT_SYMBOL(proc_create_data); 455 456 void proc_set_size(struct proc_dir_entry *de, loff_t size) 457 { 458 de->size = size; 459 } 460 EXPORT_SYMBOL(proc_set_size); 461 462 void proc_set_user(struct proc_dir_entry *de, kuid_t uid, kgid_t gid) 463 { 464 de->uid = uid; 465 de->gid = gid; 466 } 467 EXPORT_SYMBOL(proc_set_user); 468 469 static void free_proc_entry(struct proc_dir_entry *de) 470 { 471 proc_free_inum(de->low_ino); 472 473 if (S_ISLNK(de->mode)) 474 kfree(de->data); 475 kfree(de); 476 } 477 478 void pde_put(struct proc_dir_entry *pde) 479 { 480 if (atomic_dec_and_test(&pde->count)) 481 free_proc_entry(pde); 482 } 483 484 /* 485 * Remove a /proc entry and free it if it's not currently in use. 486 */ 487 void remove_proc_entry(const char *name, struct proc_dir_entry *parent) 488 { 489 struct proc_dir_entry **p; 490 struct proc_dir_entry *de = NULL; 491 const char *fn = name; 492 unsigned int len; 493 494 spin_lock(&proc_subdir_lock); 495 if (__xlate_proc_name(name, &parent, &fn) != 0) { 496 spin_unlock(&proc_subdir_lock); 497 return; 498 } 499 len = strlen(fn); 500 501 for (p = &parent->subdir; *p; p=&(*p)->next ) { 502 if (proc_match(len, fn, *p)) { 503 de = *p; 504 *p = de->next; 505 de->next = NULL; 506 break; 507 } 508 } 509 spin_unlock(&proc_subdir_lock); 510 if (!de) { 511 WARN(1, "name '%s'\n", name); 512 return; 513 } 514 515 proc_entry_rundown(de); 516 517 if (S_ISDIR(de->mode)) 518 parent->nlink--; 519 de->nlink = 0; 520 WARN(de->subdir, "%s: removing non-empty directory " 521 "'%s/%s', leaking at least '%s'\n", __func__, 522 de->parent->name, de->name, de->subdir->name); 523 pde_put(de); 524 } 525 EXPORT_SYMBOL(remove_proc_entry); 526 527 int remove_proc_subtree(const char *name, struct proc_dir_entry *parent) 528 { 529 struct proc_dir_entry **p; 530 struct proc_dir_entry *root = NULL, *de, *next; 531 const char *fn = name; 532 unsigned int len; 533 534 spin_lock(&proc_subdir_lock); 535 if (__xlate_proc_name(name, &parent, &fn) != 0) { 536 spin_unlock(&proc_subdir_lock); 537 return -ENOENT; 538 } 539 len = strlen(fn); 540 541 for (p = &parent->subdir; *p; p=&(*p)->next ) { 542 if (proc_match(len, fn, *p)) { 543 root = *p; 544 *p = root->next; 545 root->next = NULL; 546 break; 547 } 548 } 549 if (!root) { 550 spin_unlock(&proc_subdir_lock); 551 return -ENOENT; 552 } 553 de = root; 554 while (1) { 555 next = de->subdir; 556 if (next) { 557 de->subdir = next->next; 558 next->next = NULL; 559 de = next; 560 continue; 561 } 562 spin_unlock(&proc_subdir_lock); 563 564 proc_entry_rundown(de); 565 next = de->parent; 566 if (S_ISDIR(de->mode)) 567 next->nlink--; 568 de->nlink = 0; 569 if (de == root) 570 break; 571 pde_put(de); 572 573 spin_lock(&proc_subdir_lock); 574 de = next; 575 } 576 pde_put(root); 577 return 0; 578 } 579 EXPORT_SYMBOL(remove_proc_subtree); 580 581 void *proc_get_parent_data(const struct inode *inode) 582 { 583 struct proc_dir_entry *de = PDE(inode); 584 return de->parent->data; 585 } 586 EXPORT_SYMBOL_GPL(proc_get_parent_data); 587 588 void proc_remove(struct proc_dir_entry *de) 589 { 590 if (de) 591 remove_proc_subtree(de->name, de->parent); 592 } 593 EXPORT_SYMBOL(proc_remove); 594 595 void *PDE_DATA(const struct inode *inode) 596 { 597 return __PDE_DATA(inode); 598 } 599 EXPORT_SYMBOL(PDE_DATA); 600