1 /* 2 * fs/libfs.c 3 * Library for filesystems writers. 4 */ 5 6 #include <linux/blkdev.h> 7 #include <linux/export.h> 8 #include <linux/pagemap.h> 9 #include <linux/slab.h> 10 #include <linux/mount.h> 11 #include <linux/vfs.h> 12 #include <linux/quotaops.h> 13 #include <linux/mutex.h> 14 #include <linux/namei.h> 15 #include <linux/exportfs.h> 16 #include <linux/writeback.h> 17 #include <linux/buffer_head.h> /* sync_mapping_buffers */ 18 19 #include <linux/uaccess.h> 20 21 #include "internal.h" 22 23 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry, 24 struct kstat *stat) 25 { 26 struct inode *inode = d_inode(dentry); 27 generic_fillattr(inode, stat); 28 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9); 29 return 0; 30 } 31 EXPORT_SYMBOL(simple_getattr); 32 33 int simple_statfs(struct dentry *dentry, struct kstatfs *buf) 34 { 35 buf->f_type = dentry->d_sb->s_magic; 36 buf->f_bsize = PAGE_SIZE; 37 buf->f_namelen = NAME_MAX; 38 return 0; 39 } 40 EXPORT_SYMBOL(simple_statfs); 41 42 /* 43 * Retaining negative dentries for an in-memory filesystem just wastes 44 * memory and lookup time: arrange for them to be deleted immediately. 45 */ 46 int always_delete_dentry(const struct dentry *dentry) 47 { 48 return 1; 49 } 50 EXPORT_SYMBOL(always_delete_dentry); 51 52 const struct dentry_operations simple_dentry_operations = { 53 .d_delete = always_delete_dentry, 54 }; 55 EXPORT_SYMBOL(simple_dentry_operations); 56 57 /* 58 * Lookup the data. This is trivial - if the dentry didn't already 59 * exist, we know it is negative. Set d_op to delete negative dentries. 60 */ 61 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) 62 { 63 if (dentry->d_name.len > NAME_MAX) 64 return ERR_PTR(-ENAMETOOLONG); 65 if (!dentry->d_sb->s_d_op) 66 d_set_d_op(dentry, &simple_dentry_operations); 67 d_add(dentry, NULL); 68 return NULL; 69 } 70 EXPORT_SYMBOL(simple_lookup); 71 72 int dcache_dir_open(struct inode *inode, struct file *file) 73 { 74 file->private_data = d_alloc_cursor(file->f_path.dentry); 75 76 return file->private_data ? 0 : -ENOMEM; 77 } 78 EXPORT_SYMBOL(dcache_dir_open); 79 80 int dcache_dir_close(struct inode *inode, struct file *file) 81 { 82 dput(file->private_data); 83 return 0; 84 } 85 EXPORT_SYMBOL(dcache_dir_close); 86 87 /* parent is locked at least shared */ 88 static struct dentry *next_positive(struct dentry *parent, 89 struct list_head *from, 90 int count) 91 { 92 unsigned *seq = &parent->d_inode->i_dir_seq, n; 93 struct dentry *res; 94 struct list_head *p; 95 bool skipped; 96 int i; 97 98 retry: 99 i = count; 100 skipped = false; 101 n = smp_load_acquire(seq) & ~1; 102 res = NULL; 103 rcu_read_lock(); 104 for (p = from->next; p != &parent->d_subdirs; p = p->next) { 105 struct dentry *d = list_entry(p, struct dentry, d_child); 106 if (!simple_positive(d)) { 107 skipped = true; 108 } else if (!--i) { 109 res = d; 110 break; 111 } 112 } 113 rcu_read_unlock(); 114 if (skipped) { 115 smp_rmb(); 116 if (unlikely(*seq != n)) 117 goto retry; 118 } 119 return res; 120 } 121 122 static void move_cursor(struct dentry *cursor, struct list_head *after) 123 { 124 struct dentry *parent = cursor->d_parent; 125 unsigned n, *seq = &parent->d_inode->i_dir_seq; 126 spin_lock(&parent->d_lock); 127 for (;;) { 128 n = *seq; 129 if (!(n & 1) && cmpxchg(seq, n, n + 1) == n) 130 break; 131 cpu_relax(); 132 } 133 __list_del(cursor->d_child.prev, cursor->d_child.next); 134 if (after) 135 list_add(&cursor->d_child, after); 136 else 137 list_add_tail(&cursor->d_child, &parent->d_subdirs); 138 smp_store_release(seq, n + 2); 139 spin_unlock(&parent->d_lock); 140 } 141 142 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) 143 { 144 struct dentry *dentry = file->f_path.dentry; 145 switch (whence) { 146 case 1: 147 offset += file->f_pos; 148 case 0: 149 if (offset >= 0) 150 break; 151 default: 152 return -EINVAL; 153 } 154 if (offset != file->f_pos) { 155 file->f_pos = offset; 156 if (file->f_pos >= 2) { 157 struct dentry *cursor = file->private_data; 158 struct dentry *to; 159 loff_t n = file->f_pos - 2; 160 161 inode_lock_shared(dentry->d_inode); 162 to = next_positive(dentry, &dentry->d_subdirs, n); 163 move_cursor(cursor, to ? &to->d_child : NULL); 164 inode_unlock_shared(dentry->d_inode); 165 } 166 } 167 return offset; 168 } 169 EXPORT_SYMBOL(dcache_dir_lseek); 170 171 /* Relationship between i_mode and the DT_xxx types */ 172 static inline unsigned char dt_type(struct inode *inode) 173 { 174 return (inode->i_mode >> 12) & 15; 175 } 176 177 /* 178 * Directory is locked and all positive dentries in it are safe, since 179 * for ramfs-type trees they can't go away without unlink() or rmdir(), 180 * both impossible due to the lock on directory. 181 */ 182 183 int dcache_readdir(struct file *file, struct dir_context *ctx) 184 { 185 struct dentry *dentry = file->f_path.dentry; 186 struct dentry *cursor = file->private_data; 187 struct list_head *p = &cursor->d_child; 188 struct dentry *next; 189 bool moved = false; 190 191 if (!dir_emit_dots(file, ctx)) 192 return 0; 193 194 if (ctx->pos == 2) 195 p = &dentry->d_subdirs; 196 while ((next = next_positive(dentry, p, 1)) != NULL) { 197 if (!dir_emit(ctx, next->d_name.name, next->d_name.len, 198 d_inode(next)->i_ino, dt_type(d_inode(next)))) 199 break; 200 moved = true; 201 p = &next->d_child; 202 ctx->pos++; 203 } 204 if (moved) 205 move_cursor(cursor, p); 206 return 0; 207 } 208 EXPORT_SYMBOL(dcache_readdir); 209 210 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) 211 { 212 return -EISDIR; 213 } 214 EXPORT_SYMBOL(generic_read_dir); 215 216 const struct file_operations simple_dir_operations = { 217 .open = dcache_dir_open, 218 .release = dcache_dir_close, 219 .llseek = dcache_dir_lseek, 220 .read = generic_read_dir, 221 .iterate_shared = dcache_readdir, 222 .fsync = noop_fsync, 223 }; 224 EXPORT_SYMBOL(simple_dir_operations); 225 226 const struct inode_operations simple_dir_inode_operations = { 227 .lookup = simple_lookup, 228 }; 229 EXPORT_SYMBOL(simple_dir_inode_operations); 230 231 static const struct super_operations simple_super_operations = { 232 .statfs = simple_statfs, 233 }; 234 235 /* 236 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that 237 * will never be mountable) 238 */ 239 struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name, 240 const struct super_operations *ops, const struct xattr_handler **xattr, 241 const struct dentry_operations *dops, unsigned long magic) 242 { 243 struct super_block *s; 244 struct dentry *dentry; 245 struct inode *root; 246 struct qstr d_name = QSTR_INIT(name, strlen(name)); 247 248 s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL); 249 if (IS_ERR(s)) 250 return ERR_CAST(s); 251 252 s->s_maxbytes = MAX_LFS_FILESIZE; 253 s->s_blocksize = PAGE_SIZE; 254 s->s_blocksize_bits = PAGE_SHIFT; 255 s->s_magic = magic; 256 s->s_op = ops ? ops : &simple_super_operations; 257 s->s_xattr = xattr; 258 s->s_time_gran = 1; 259 root = new_inode(s); 260 if (!root) 261 goto Enomem; 262 /* 263 * since this is the first inode, make it number 1. New inodes created 264 * after this must take care not to collide with it (by passing 265 * max_reserved of 1 to iunique). 266 */ 267 root->i_ino = 1; 268 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; 269 root->i_atime = root->i_mtime = root->i_ctime = current_time(root); 270 dentry = __d_alloc(s, &d_name); 271 if (!dentry) { 272 iput(root); 273 goto Enomem; 274 } 275 d_instantiate(dentry, root); 276 s->s_root = dentry; 277 s->s_d_op = dops; 278 s->s_flags |= MS_ACTIVE; 279 return dget(s->s_root); 280 281 Enomem: 282 deactivate_locked_super(s); 283 return ERR_PTR(-ENOMEM); 284 } 285 EXPORT_SYMBOL(mount_pseudo_xattr); 286 287 int simple_open(struct inode *inode, struct file *file) 288 { 289 if (inode->i_private) 290 file->private_data = inode->i_private; 291 return 0; 292 } 293 EXPORT_SYMBOL(simple_open); 294 295 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) 296 { 297 struct inode *inode = d_inode(old_dentry); 298 299 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); 300 inc_nlink(inode); 301 ihold(inode); 302 dget(dentry); 303 d_instantiate(dentry, inode); 304 return 0; 305 } 306 EXPORT_SYMBOL(simple_link); 307 308 int simple_empty(struct dentry *dentry) 309 { 310 struct dentry *child; 311 int ret = 0; 312 313 spin_lock(&dentry->d_lock); 314 list_for_each_entry(child, &dentry->d_subdirs, d_child) { 315 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); 316 if (simple_positive(child)) { 317 spin_unlock(&child->d_lock); 318 goto out; 319 } 320 spin_unlock(&child->d_lock); 321 } 322 ret = 1; 323 out: 324 spin_unlock(&dentry->d_lock); 325 return ret; 326 } 327 EXPORT_SYMBOL(simple_empty); 328 329 int simple_unlink(struct inode *dir, struct dentry *dentry) 330 { 331 struct inode *inode = d_inode(dentry); 332 333 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); 334 drop_nlink(inode); 335 dput(dentry); 336 return 0; 337 } 338 EXPORT_SYMBOL(simple_unlink); 339 340 int simple_rmdir(struct inode *dir, struct dentry *dentry) 341 { 342 if (!simple_empty(dentry)) 343 return -ENOTEMPTY; 344 345 drop_nlink(d_inode(dentry)); 346 simple_unlink(dir, dentry); 347 drop_nlink(dir); 348 return 0; 349 } 350 EXPORT_SYMBOL(simple_rmdir); 351 352 int simple_rename(struct inode *old_dir, struct dentry *old_dentry, 353 struct inode *new_dir, struct dentry *new_dentry, 354 unsigned int flags) 355 { 356 struct inode *inode = d_inode(old_dentry); 357 int they_are_dirs = d_is_dir(old_dentry); 358 359 if (flags & ~RENAME_NOREPLACE) 360 return -EINVAL; 361 362 if (!simple_empty(new_dentry)) 363 return -ENOTEMPTY; 364 365 if (d_really_is_positive(new_dentry)) { 366 simple_unlink(new_dir, new_dentry); 367 if (they_are_dirs) { 368 drop_nlink(d_inode(new_dentry)); 369 drop_nlink(old_dir); 370 } 371 } else if (they_are_dirs) { 372 drop_nlink(old_dir); 373 inc_nlink(new_dir); 374 } 375 376 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime = 377 new_dir->i_mtime = inode->i_ctime = current_time(old_dir); 378 379 return 0; 380 } 381 EXPORT_SYMBOL(simple_rename); 382 383 /** 384 * simple_setattr - setattr for simple filesystem 385 * @dentry: dentry 386 * @iattr: iattr structure 387 * 388 * Returns 0 on success, -error on failure. 389 * 390 * simple_setattr is a simple ->setattr implementation without a proper 391 * implementation of size changes. 392 * 393 * It can either be used for in-memory filesystems or special files 394 * on simple regular filesystems. Anything that needs to change on-disk 395 * or wire state on size changes needs its own setattr method. 396 */ 397 int simple_setattr(struct dentry *dentry, struct iattr *iattr) 398 { 399 struct inode *inode = d_inode(dentry); 400 int error; 401 402 error = setattr_prepare(dentry, iattr); 403 if (error) 404 return error; 405 406 if (iattr->ia_valid & ATTR_SIZE) 407 truncate_setsize(inode, iattr->ia_size); 408 setattr_copy(inode, iattr); 409 mark_inode_dirty(inode); 410 return 0; 411 } 412 EXPORT_SYMBOL(simple_setattr); 413 414 int simple_readpage(struct file *file, struct page *page) 415 { 416 clear_highpage(page); 417 flush_dcache_page(page); 418 SetPageUptodate(page); 419 unlock_page(page); 420 return 0; 421 } 422 EXPORT_SYMBOL(simple_readpage); 423 424 int simple_write_begin(struct file *file, struct address_space *mapping, 425 loff_t pos, unsigned len, unsigned flags, 426 struct page **pagep, void **fsdata) 427 { 428 struct page *page; 429 pgoff_t index; 430 431 index = pos >> PAGE_SHIFT; 432 433 page = grab_cache_page_write_begin(mapping, index, flags); 434 if (!page) 435 return -ENOMEM; 436 437 *pagep = page; 438 439 if (!PageUptodate(page) && (len != PAGE_SIZE)) { 440 unsigned from = pos & (PAGE_SIZE - 1); 441 442 zero_user_segments(page, 0, from, from + len, PAGE_SIZE); 443 } 444 return 0; 445 } 446 EXPORT_SYMBOL(simple_write_begin); 447 448 /** 449 * simple_write_end - .write_end helper for non-block-device FSes 450 * @available: See .write_end of address_space_operations 451 * @file: " 452 * @mapping: " 453 * @pos: " 454 * @len: " 455 * @copied: " 456 * @page: " 457 * @fsdata: " 458 * 459 * simple_write_end does the minimum needed for updating a page after writing is 460 * done. It has the same API signature as the .write_end of 461 * address_space_operations vector. So it can just be set onto .write_end for 462 * FSes that don't need any other processing. i_mutex is assumed to be held. 463 * Block based filesystems should use generic_write_end(). 464 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty 465 * is not called, so a filesystem that actually does store data in .write_inode 466 * should extend on what's done here with a call to mark_inode_dirty() in the 467 * case that i_size has changed. 468 * 469 * Use *ONLY* with simple_readpage() 470 */ 471 int simple_write_end(struct file *file, struct address_space *mapping, 472 loff_t pos, unsigned len, unsigned copied, 473 struct page *page, void *fsdata) 474 { 475 struct inode *inode = page->mapping->host; 476 loff_t last_pos = pos + copied; 477 478 /* zero the stale part of the page if we did a short copy */ 479 if (!PageUptodate(page)) { 480 if (copied < len) { 481 unsigned from = pos & (PAGE_SIZE - 1); 482 483 zero_user(page, from + copied, len - copied); 484 } 485 SetPageUptodate(page); 486 } 487 /* 488 * No need to use i_size_read() here, the i_size 489 * cannot change under us because we hold the i_mutex. 490 */ 491 if (last_pos > inode->i_size) 492 i_size_write(inode, last_pos); 493 494 set_page_dirty(page); 495 unlock_page(page); 496 put_page(page); 497 498 return copied; 499 } 500 EXPORT_SYMBOL(simple_write_end); 501 502 /* 503 * the inodes created here are not hashed. If you use iunique to generate 504 * unique inode values later for this filesystem, then you must take care 505 * to pass it an appropriate max_reserved value to avoid collisions. 506 */ 507 int simple_fill_super(struct super_block *s, unsigned long magic, 508 struct tree_descr *files) 509 { 510 struct inode *inode; 511 struct dentry *root; 512 struct dentry *dentry; 513 int i; 514 515 s->s_blocksize = PAGE_SIZE; 516 s->s_blocksize_bits = PAGE_SHIFT; 517 s->s_magic = magic; 518 s->s_op = &simple_super_operations; 519 s->s_time_gran = 1; 520 521 inode = new_inode(s); 522 if (!inode) 523 return -ENOMEM; 524 /* 525 * because the root inode is 1, the files array must not contain an 526 * entry at index 1 527 */ 528 inode->i_ino = 1; 529 inode->i_mode = S_IFDIR | 0755; 530 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); 531 inode->i_op = &simple_dir_inode_operations; 532 inode->i_fop = &simple_dir_operations; 533 set_nlink(inode, 2); 534 root = d_make_root(inode); 535 if (!root) 536 return -ENOMEM; 537 for (i = 0; !files->name || files->name[0]; i++, files++) { 538 if (!files->name) 539 continue; 540 541 /* warn if it tries to conflict with the root inode */ 542 if (unlikely(i == 1)) 543 printk(KERN_WARNING "%s: %s passed in a files array" 544 "with an index of 1!\n", __func__, 545 s->s_type->name); 546 547 dentry = d_alloc_name(root, files->name); 548 if (!dentry) 549 goto out; 550 inode = new_inode(s); 551 if (!inode) { 552 dput(dentry); 553 goto out; 554 } 555 inode->i_mode = S_IFREG | files->mode; 556 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); 557 inode->i_fop = files->ops; 558 inode->i_ino = i; 559 d_add(dentry, inode); 560 } 561 s->s_root = root; 562 return 0; 563 out: 564 d_genocide(root); 565 shrink_dcache_parent(root); 566 dput(root); 567 return -ENOMEM; 568 } 569 EXPORT_SYMBOL(simple_fill_super); 570 571 static DEFINE_SPINLOCK(pin_fs_lock); 572 573 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) 574 { 575 struct vfsmount *mnt = NULL; 576 spin_lock(&pin_fs_lock); 577 if (unlikely(!*mount)) { 578 spin_unlock(&pin_fs_lock); 579 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL); 580 if (IS_ERR(mnt)) 581 return PTR_ERR(mnt); 582 spin_lock(&pin_fs_lock); 583 if (!*mount) 584 *mount = mnt; 585 } 586 mntget(*mount); 587 ++*count; 588 spin_unlock(&pin_fs_lock); 589 mntput(mnt); 590 return 0; 591 } 592 EXPORT_SYMBOL(simple_pin_fs); 593 594 void simple_release_fs(struct vfsmount **mount, int *count) 595 { 596 struct vfsmount *mnt; 597 spin_lock(&pin_fs_lock); 598 mnt = *mount; 599 if (!--*count) 600 *mount = NULL; 601 spin_unlock(&pin_fs_lock); 602 mntput(mnt); 603 } 604 EXPORT_SYMBOL(simple_release_fs); 605 606 /** 607 * simple_read_from_buffer - copy data from the buffer to user space 608 * @to: the user space buffer to read to 609 * @count: the maximum number of bytes to read 610 * @ppos: the current position in the buffer 611 * @from: the buffer to read from 612 * @available: the size of the buffer 613 * 614 * The simple_read_from_buffer() function reads up to @count bytes from the 615 * buffer @from at offset @ppos into the user space address starting at @to. 616 * 617 * On success, the number of bytes read is returned and the offset @ppos is 618 * advanced by this number, or negative value is returned on error. 619 **/ 620 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, 621 const void *from, size_t available) 622 { 623 loff_t pos = *ppos; 624 size_t ret; 625 626 if (pos < 0) 627 return -EINVAL; 628 if (pos >= available || !count) 629 return 0; 630 if (count > available - pos) 631 count = available - pos; 632 ret = copy_to_user(to, from + pos, count); 633 if (ret == count) 634 return -EFAULT; 635 count -= ret; 636 *ppos = pos + count; 637 return count; 638 } 639 EXPORT_SYMBOL(simple_read_from_buffer); 640 641 /** 642 * simple_write_to_buffer - copy data from user space to the buffer 643 * @to: the buffer to write to 644 * @available: the size of the buffer 645 * @ppos: the current position in the buffer 646 * @from: the user space buffer to read from 647 * @count: the maximum number of bytes to read 648 * 649 * The simple_write_to_buffer() function reads up to @count bytes from the user 650 * space address starting at @from into the buffer @to at offset @ppos. 651 * 652 * On success, the number of bytes written is returned and the offset @ppos is 653 * advanced by this number, or negative value is returned on error. 654 **/ 655 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 656 const void __user *from, size_t count) 657 { 658 loff_t pos = *ppos; 659 size_t res; 660 661 if (pos < 0) 662 return -EINVAL; 663 if (pos >= available || !count) 664 return 0; 665 if (count > available - pos) 666 count = available - pos; 667 res = copy_from_user(to + pos, from, count); 668 if (res == count) 669 return -EFAULT; 670 count -= res; 671 *ppos = pos + count; 672 return count; 673 } 674 EXPORT_SYMBOL(simple_write_to_buffer); 675 676 /** 677 * memory_read_from_buffer - copy data from the buffer 678 * @to: the kernel space buffer to read to 679 * @count: the maximum number of bytes to read 680 * @ppos: the current position in the buffer 681 * @from: the buffer to read from 682 * @available: the size of the buffer 683 * 684 * The memory_read_from_buffer() function reads up to @count bytes from the 685 * buffer @from at offset @ppos into the kernel space address starting at @to. 686 * 687 * On success, the number of bytes read is returned and the offset @ppos is 688 * advanced by this number, or negative value is returned on error. 689 **/ 690 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, 691 const void *from, size_t available) 692 { 693 loff_t pos = *ppos; 694 695 if (pos < 0) 696 return -EINVAL; 697 if (pos >= available) 698 return 0; 699 if (count > available - pos) 700 count = available - pos; 701 memcpy(to, from + pos, count); 702 *ppos = pos + count; 703 704 return count; 705 } 706 EXPORT_SYMBOL(memory_read_from_buffer); 707 708 /* 709 * Transaction based IO. 710 * The file expects a single write which triggers the transaction, and then 711 * possibly a read which collects the result - which is stored in a 712 * file-local buffer. 713 */ 714 715 void simple_transaction_set(struct file *file, size_t n) 716 { 717 struct simple_transaction_argresp *ar = file->private_data; 718 719 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); 720 721 /* 722 * The barrier ensures that ar->size will really remain zero until 723 * ar->data is ready for reading. 724 */ 725 smp_mb(); 726 ar->size = n; 727 } 728 EXPORT_SYMBOL(simple_transaction_set); 729 730 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) 731 { 732 struct simple_transaction_argresp *ar; 733 static DEFINE_SPINLOCK(simple_transaction_lock); 734 735 if (size > SIMPLE_TRANSACTION_LIMIT - 1) 736 return ERR_PTR(-EFBIG); 737 738 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); 739 if (!ar) 740 return ERR_PTR(-ENOMEM); 741 742 spin_lock(&simple_transaction_lock); 743 744 /* only one write allowed per open */ 745 if (file->private_data) { 746 spin_unlock(&simple_transaction_lock); 747 free_page((unsigned long)ar); 748 return ERR_PTR(-EBUSY); 749 } 750 751 file->private_data = ar; 752 753 spin_unlock(&simple_transaction_lock); 754 755 if (copy_from_user(ar->data, buf, size)) 756 return ERR_PTR(-EFAULT); 757 758 return ar->data; 759 } 760 EXPORT_SYMBOL(simple_transaction_get); 761 762 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) 763 { 764 struct simple_transaction_argresp *ar = file->private_data; 765 766 if (!ar) 767 return 0; 768 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); 769 } 770 EXPORT_SYMBOL(simple_transaction_read); 771 772 int simple_transaction_release(struct inode *inode, struct file *file) 773 { 774 free_page((unsigned long)file->private_data); 775 return 0; 776 } 777 EXPORT_SYMBOL(simple_transaction_release); 778 779 /* Simple attribute files */ 780 781 struct simple_attr { 782 int (*get)(void *, u64 *); 783 int (*set)(void *, u64); 784 char get_buf[24]; /* enough to store a u64 and "\n\0" */ 785 char set_buf[24]; 786 void *data; 787 const char *fmt; /* format for read operation */ 788 struct mutex mutex; /* protects access to these buffers */ 789 }; 790 791 /* simple_attr_open is called by an actual attribute open file operation 792 * to set the attribute specific access operations. */ 793 int simple_attr_open(struct inode *inode, struct file *file, 794 int (*get)(void *, u64 *), int (*set)(void *, u64), 795 const char *fmt) 796 { 797 struct simple_attr *attr; 798 799 attr = kmalloc(sizeof(*attr), GFP_KERNEL); 800 if (!attr) 801 return -ENOMEM; 802 803 attr->get = get; 804 attr->set = set; 805 attr->data = inode->i_private; 806 attr->fmt = fmt; 807 mutex_init(&attr->mutex); 808 809 file->private_data = attr; 810 811 return nonseekable_open(inode, file); 812 } 813 EXPORT_SYMBOL_GPL(simple_attr_open); 814 815 int simple_attr_release(struct inode *inode, struct file *file) 816 { 817 kfree(file->private_data); 818 return 0; 819 } 820 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ 821 822 /* read from the buffer that is filled with the get function */ 823 ssize_t simple_attr_read(struct file *file, char __user *buf, 824 size_t len, loff_t *ppos) 825 { 826 struct simple_attr *attr; 827 size_t size; 828 ssize_t ret; 829 830 attr = file->private_data; 831 832 if (!attr->get) 833 return -EACCES; 834 835 ret = mutex_lock_interruptible(&attr->mutex); 836 if (ret) 837 return ret; 838 839 if (*ppos) { /* continued read */ 840 size = strlen(attr->get_buf); 841 } else { /* first read */ 842 u64 val; 843 ret = attr->get(attr->data, &val); 844 if (ret) 845 goto out; 846 847 size = scnprintf(attr->get_buf, sizeof(attr->get_buf), 848 attr->fmt, (unsigned long long)val); 849 } 850 851 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); 852 out: 853 mutex_unlock(&attr->mutex); 854 return ret; 855 } 856 EXPORT_SYMBOL_GPL(simple_attr_read); 857 858 /* interpret the buffer as a number to call the set function with */ 859 ssize_t simple_attr_write(struct file *file, const char __user *buf, 860 size_t len, loff_t *ppos) 861 { 862 struct simple_attr *attr; 863 u64 val; 864 size_t size; 865 ssize_t ret; 866 867 attr = file->private_data; 868 if (!attr->set) 869 return -EACCES; 870 871 ret = mutex_lock_interruptible(&attr->mutex); 872 if (ret) 873 return ret; 874 875 ret = -EFAULT; 876 size = min(sizeof(attr->set_buf) - 1, len); 877 if (copy_from_user(attr->set_buf, buf, size)) 878 goto out; 879 880 attr->set_buf[size] = '\0'; 881 val = simple_strtoll(attr->set_buf, NULL, 0); 882 ret = attr->set(attr->data, val); 883 if (ret == 0) 884 ret = len; /* on success, claim we got the whole input */ 885 out: 886 mutex_unlock(&attr->mutex); 887 return ret; 888 } 889 EXPORT_SYMBOL_GPL(simple_attr_write); 890 891 /** 892 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation 893 * @sb: filesystem to do the file handle conversion on 894 * @fid: file handle to convert 895 * @fh_len: length of the file handle in bytes 896 * @fh_type: type of file handle 897 * @get_inode: filesystem callback to retrieve inode 898 * 899 * This function decodes @fid as long as it has one of the well-known 900 * Linux filehandle types and calls @get_inode on it to retrieve the 901 * inode for the object specified in the file handle. 902 */ 903 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, 904 int fh_len, int fh_type, struct inode *(*get_inode) 905 (struct super_block *sb, u64 ino, u32 gen)) 906 { 907 struct inode *inode = NULL; 908 909 if (fh_len < 2) 910 return NULL; 911 912 switch (fh_type) { 913 case FILEID_INO32_GEN: 914 case FILEID_INO32_GEN_PARENT: 915 inode = get_inode(sb, fid->i32.ino, fid->i32.gen); 916 break; 917 } 918 919 return d_obtain_alias(inode); 920 } 921 EXPORT_SYMBOL_GPL(generic_fh_to_dentry); 922 923 /** 924 * generic_fh_to_parent - generic helper for the fh_to_parent export operation 925 * @sb: filesystem to do the file handle conversion on 926 * @fid: file handle to convert 927 * @fh_len: length of the file handle in bytes 928 * @fh_type: type of file handle 929 * @get_inode: filesystem callback to retrieve inode 930 * 931 * This function decodes @fid as long as it has one of the well-known 932 * Linux filehandle types and calls @get_inode on it to retrieve the 933 * inode for the _parent_ object specified in the file handle if it 934 * is specified in the file handle, or NULL otherwise. 935 */ 936 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, 937 int fh_len, int fh_type, struct inode *(*get_inode) 938 (struct super_block *sb, u64 ino, u32 gen)) 939 { 940 struct inode *inode = NULL; 941 942 if (fh_len <= 2) 943 return NULL; 944 945 switch (fh_type) { 946 case FILEID_INO32_GEN_PARENT: 947 inode = get_inode(sb, fid->i32.parent_ino, 948 (fh_len > 3 ? fid->i32.parent_gen : 0)); 949 break; 950 } 951 952 return d_obtain_alias(inode); 953 } 954 EXPORT_SYMBOL_GPL(generic_fh_to_parent); 955 956 /** 957 * __generic_file_fsync - generic fsync implementation for simple filesystems 958 * 959 * @file: file to synchronize 960 * @start: start offset in bytes 961 * @end: end offset in bytes (inclusive) 962 * @datasync: only synchronize essential metadata if true 963 * 964 * This is a generic implementation of the fsync method for simple 965 * filesystems which track all non-inode metadata in the buffers list 966 * hanging off the address_space structure. 967 */ 968 int __generic_file_fsync(struct file *file, loff_t start, loff_t end, 969 int datasync) 970 { 971 struct inode *inode = file->f_mapping->host; 972 int err; 973 int ret; 974 975 err = filemap_write_and_wait_range(inode->i_mapping, start, end); 976 if (err) 977 return err; 978 979 inode_lock(inode); 980 ret = sync_mapping_buffers(inode->i_mapping); 981 if (!(inode->i_state & I_DIRTY_ALL)) 982 goto out; 983 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) 984 goto out; 985 986 err = sync_inode_metadata(inode, 1); 987 if (ret == 0) 988 ret = err; 989 990 out: 991 inode_unlock(inode); 992 return ret; 993 } 994 EXPORT_SYMBOL(__generic_file_fsync); 995 996 /** 997 * generic_file_fsync - generic fsync implementation for simple filesystems 998 * with flush 999 * @file: file to synchronize 1000 * @start: start offset in bytes 1001 * @end: end offset in bytes (inclusive) 1002 * @datasync: only synchronize essential metadata if true 1003 * 1004 */ 1005 1006 int generic_file_fsync(struct file *file, loff_t start, loff_t end, 1007 int datasync) 1008 { 1009 struct inode *inode = file->f_mapping->host; 1010 int err; 1011 1012 err = __generic_file_fsync(file, start, end, datasync); 1013 if (err) 1014 return err; 1015 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL); 1016 } 1017 EXPORT_SYMBOL(generic_file_fsync); 1018 1019 /** 1020 * generic_check_addressable - Check addressability of file system 1021 * @blocksize_bits: log of file system block size 1022 * @num_blocks: number of blocks in file system 1023 * 1024 * Determine whether a file system with @num_blocks blocks (and a 1025 * block size of 2**@blocksize_bits) is addressable by the sector_t 1026 * and page cache of the system. Return 0 if so and -EFBIG otherwise. 1027 */ 1028 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) 1029 { 1030 u64 last_fs_block = num_blocks - 1; 1031 u64 last_fs_page = 1032 last_fs_block >> (PAGE_SHIFT - blocksize_bits); 1033 1034 if (unlikely(num_blocks == 0)) 1035 return 0; 1036 1037 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) 1038 return -EINVAL; 1039 1040 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || 1041 (last_fs_page > (pgoff_t)(~0ULL))) { 1042 return -EFBIG; 1043 } 1044 return 0; 1045 } 1046 EXPORT_SYMBOL(generic_check_addressable); 1047 1048 /* 1049 * No-op implementation of ->fsync for in-memory filesystems. 1050 */ 1051 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) 1052 { 1053 return 0; 1054 } 1055 EXPORT_SYMBOL(noop_fsync); 1056 1057 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ 1058 void kfree_link(void *p) 1059 { 1060 kfree(p); 1061 } 1062 EXPORT_SYMBOL(kfree_link); 1063 1064 /* 1065 * nop .set_page_dirty method so that people can use .page_mkwrite on 1066 * anon inodes. 1067 */ 1068 static int anon_set_page_dirty(struct page *page) 1069 { 1070 return 0; 1071 }; 1072 1073 /* 1074 * A single inode exists for all anon_inode files. Contrary to pipes, 1075 * anon_inode inodes have no associated per-instance data, so we need 1076 * only allocate one of them. 1077 */ 1078 struct inode *alloc_anon_inode(struct super_block *s) 1079 { 1080 static const struct address_space_operations anon_aops = { 1081 .set_page_dirty = anon_set_page_dirty, 1082 }; 1083 struct inode *inode = new_inode_pseudo(s); 1084 1085 if (!inode) 1086 return ERR_PTR(-ENOMEM); 1087 1088 inode->i_ino = get_next_ino(); 1089 inode->i_mapping->a_ops = &anon_aops; 1090 1091 /* 1092 * Mark the inode dirty from the very beginning, 1093 * that way it will never be moved to the dirty 1094 * list because mark_inode_dirty() will think 1095 * that it already _is_ on the dirty list. 1096 */ 1097 inode->i_state = I_DIRTY; 1098 inode->i_mode = S_IRUSR | S_IWUSR; 1099 inode->i_uid = current_fsuid(); 1100 inode->i_gid = current_fsgid(); 1101 inode->i_flags |= S_PRIVATE; 1102 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); 1103 return inode; 1104 } 1105 EXPORT_SYMBOL(alloc_anon_inode); 1106 1107 /** 1108 * simple_nosetlease - generic helper for prohibiting leases 1109 * @filp: file pointer 1110 * @arg: type of lease to obtain 1111 * @flp: new lease supplied for insertion 1112 * @priv: private data for lm_setup operation 1113 * 1114 * Generic helper for filesystems that do not wish to allow leases to be set. 1115 * All arguments are ignored and it just returns -EINVAL. 1116 */ 1117 int 1118 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp, 1119 void **priv) 1120 { 1121 return -EINVAL; 1122 } 1123 EXPORT_SYMBOL(simple_nosetlease); 1124 1125 const char *simple_get_link(struct dentry *dentry, struct inode *inode, 1126 struct delayed_call *done) 1127 { 1128 return inode->i_link; 1129 } 1130 EXPORT_SYMBOL(simple_get_link); 1131 1132 const struct inode_operations simple_symlink_inode_operations = { 1133 .get_link = simple_get_link, 1134 }; 1135 EXPORT_SYMBOL(simple_symlink_inode_operations); 1136 1137 /* 1138 * Operations for a permanently empty directory. 1139 */ 1140 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) 1141 { 1142 return ERR_PTR(-ENOENT); 1143 } 1144 1145 static int empty_dir_getattr(struct vfsmount *mnt, struct dentry *dentry, 1146 struct kstat *stat) 1147 { 1148 struct inode *inode = d_inode(dentry); 1149 generic_fillattr(inode, stat); 1150 return 0; 1151 } 1152 1153 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr) 1154 { 1155 return -EPERM; 1156 } 1157 1158 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) 1159 { 1160 return -EOPNOTSUPP; 1161 } 1162 1163 static const struct inode_operations empty_dir_inode_operations = { 1164 .lookup = empty_dir_lookup, 1165 .permission = generic_permission, 1166 .setattr = empty_dir_setattr, 1167 .getattr = empty_dir_getattr, 1168 .listxattr = empty_dir_listxattr, 1169 }; 1170 1171 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) 1172 { 1173 /* An empty directory has two entries . and .. at offsets 0 and 1 */ 1174 return generic_file_llseek_size(file, offset, whence, 2, 2); 1175 } 1176 1177 static int empty_dir_readdir(struct file *file, struct dir_context *ctx) 1178 { 1179 dir_emit_dots(file, ctx); 1180 return 0; 1181 } 1182 1183 static const struct file_operations empty_dir_operations = { 1184 .llseek = empty_dir_llseek, 1185 .read = generic_read_dir, 1186 .iterate_shared = empty_dir_readdir, 1187 .fsync = noop_fsync, 1188 }; 1189 1190 1191 void make_empty_dir_inode(struct inode *inode) 1192 { 1193 set_nlink(inode, 2); 1194 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; 1195 inode->i_uid = GLOBAL_ROOT_UID; 1196 inode->i_gid = GLOBAL_ROOT_GID; 1197 inode->i_rdev = 0; 1198 inode->i_size = 0; 1199 inode->i_blkbits = PAGE_SHIFT; 1200 inode->i_blocks = 0; 1201 1202 inode->i_op = &empty_dir_inode_operations; 1203 inode->i_opflags &= ~IOP_XATTR; 1204 inode->i_fop = &empty_dir_operations; 1205 } 1206 1207 bool is_empty_dir_inode(struct inode *inode) 1208 { 1209 return (inode->i_fop == &empty_dir_operations) && 1210 (inode->i_op == &empty_dir_inode_operations); 1211 } 1212