Christoph Hellwig | f46b5a6 | 2008-06-11 21:53:53 -0400 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright (C) 2007 Oracle. All rights reserved. |
| 3 | * |
| 4 | * This program is free software; you can redistribute it and/or |
| 5 | * modify it under the terms of the GNU General Public |
| 6 | * License v2 as published by the Free Software Foundation. |
| 7 | * |
| 8 | * This program is distributed in the hope that it will be useful, |
| 9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 11 | * General Public License for more details. |
| 12 | * |
| 13 | * You should have received a copy of the GNU General Public |
| 14 | * License along with this program; if not, write to the |
| 15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, |
| 16 | * Boston, MA 021110-1307, USA. |
| 17 | */ |
| 18 | |
| 19 | #include <linux/kernel.h> |
| 20 | #include <linux/bio.h> |
| 21 | #include <linux/buffer_head.h> |
| 22 | #include <linux/file.h> |
| 23 | #include <linux/fs.h> |
| 24 | #include <linux/pagemap.h> |
| 25 | #include <linux/highmem.h> |
| 26 | #include <linux/time.h> |
| 27 | #include <linux/init.h> |
| 28 | #include <linux/string.h> |
| 29 | #include <linux/smp_lock.h> |
| 30 | #include <linux/backing-dev.h> |
| 31 | #include <linux/mpage.h> |
| 32 | #include <linux/swap.h> |
| 33 | #include <linux/writeback.h> |
| 34 | #include <linux/statfs.h> |
| 35 | #include <linux/compat.h> |
| 36 | #include <linux/bit_spinlock.h> |
| 37 | #include <linux/version.h> |
| 38 | #include <linux/xattr.h> |
| 39 | #include "ctree.h" |
| 40 | #include "disk-io.h" |
| 41 | #include "transaction.h" |
| 42 | #include "btrfs_inode.h" |
| 43 | #include "ioctl.h" |
| 44 | #include "print-tree.h" |
| 45 | #include "volumes.h" |
| 46 | |
| 47 | |
| 48 | |
| 49 | static noinline int create_subvol(struct btrfs_root *root, char *name, |
| 50 | int namelen) |
| 51 | { |
| 52 | struct btrfs_trans_handle *trans; |
| 53 | struct btrfs_key key; |
| 54 | struct btrfs_root_item root_item; |
| 55 | struct btrfs_inode_item *inode_item; |
| 56 | struct extent_buffer *leaf; |
| 57 | struct btrfs_root *new_root = root; |
| 58 | struct inode *dir; |
| 59 | int ret; |
| 60 | int err; |
| 61 | u64 objectid; |
| 62 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; |
| 63 | unsigned long nr = 1; |
| 64 | |
| 65 | mutex_lock(&root->fs_info->fs_mutex); |
| 66 | ret = btrfs_check_free_space(root, 1, 0); |
| 67 | if (ret) |
| 68 | goto fail_commit; |
| 69 | |
| 70 | trans = btrfs_start_transaction(root, 1); |
| 71 | BUG_ON(!trans); |
| 72 | |
| 73 | ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, |
| 74 | 0, &objectid); |
| 75 | if (ret) |
| 76 | goto fail; |
| 77 | |
| 78 | leaf = __btrfs_alloc_free_block(trans, root, root->leafsize, |
| 79 | objectid, trans->transid, 0, 0, |
| 80 | 0, 0); |
| 81 | if (IS_ERR(leaf)) |
| 82 | return PTR_ERR(leaf); |
| 83 | |
| 84 | btrfs_set_header_nritems(leaf, 0); |
| 85 | btrfs_set_header_level(leaf, 0); |
| 86 | btrfs_set_header_bytenr(leaf, leaf->start); |
| 87 | btrfs_set_header_generation(leaf, trans->transid); |
| 88 | btrfs_set_header_owner(leaf, objectid); |
| 89 | |
| 90 | write_extent_buffer(leaf, root->fs_info->fsid, |
| 91 | (unsigned long)btrfs_header_fsid(leaf), |
| 92 | BTRFS_FSID_SIZE); |
| 93 | btrfs_mark_buffer_dirty(leaf); |
| 94 | |
| 95 | inode_item = &root_item.inode; |
| 96 | memset(inode_item, 0, sizeof(*inode_item)); |
| 97 | inode_item->generation = cpu_to_le64(1); |
| 98 | inode_item->size = cpu_to_le64(3); |
| 99 | inode_item->nlink = cpu_to_le32(1); |
| 100 | inode_item->nblocks = cpu_to_le64(1); |
| 101 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); |
| 102 | |
| 103 | btrfs_set_root_bytenr(&root_item, leaf->start); |
| 104 | btrfs_set_root_level(&root_item, 0); |
| 105 | btrfs_set_root_refs(&root_item, 1); |
| 106 | btrfs_set_root_used(&root_item, 0); |
| 107 | |
| 108 | memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); |
| 109 | root_item.drop_level = 0; |
| 110 | |
| 111 | free_extent_buffer(leaf); |
| 112 | leaf = NULL; |
| 113 | |
| 114 | btrfs_set_root_dirid(&root_item, new_dirid); |
| 115 | |
| 116 | key.objectid = objectid; |
| 117 | key.offset = 1; |
| 118 | btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); |
| 119 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, |
| 120 | &root_item); |
| 121 | if (ret) |
| 122 | goto fail; |
| 123 | |
| 124 | /* |
| 125 | * insert the directory item |
| 126 | */ |
| 127 | key.offset = (u64)-1; |
| 128 | dir = root->fs_info->sb->s_root->d_inode; |
| 129 | ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, |
| 130 | name, namelen, dir->i_ino, &key, |
| 131 | BTRFS_FT_DIR); |
| 132 | if (ret) |
| 133 | goto fail; |
| 134 | |
| 135 | ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, |
| 136 | name, namelen, objectid, |
| 137 | root->fs_info->sb->s_root->d_inode->i_ino); |
| 138 | if (ret) |
| 139 | goto fail; |
| 140 | |
| 141 | ret = btrfs_commit_transaction(trans, root); |
| 142 | if (ret) |
| 143 | goto fail_commit; |
| 144 | |
| 145 | new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen); |
| 146 | BUG_ON(!new_root); |
| 147 | |
| 148 | trans = btrfs_start_transaction(new_root, 1); |
| 149 | BUG_ON(!trans); |
| 150 | |
| 151 | ret = btrfs_create_subvol_root(new_root, trans, new_dirid, |
| 152 | BTRFS_I(dir)->block_group); |
| 153 | if (ret) |
| 154 | goto fail; |
| 155 | |
| 156 | /* Invalidate existing dcache entry for new subvolume. */ |
| 157 | btrfs_invalidate_dcache_root(root, name, namelen); |
| 158 | |
| 159 | fail: |
| 160 | nr = trans->blocks_used; |
| 161 | err = btrfs_commit_transaction(trans, new_root); |
| 162 | if (err && !ret) |
| 163 | ret = err; |
| 164 | fail_commit: |
| 165 | mutex_unlock(&root->fs_info->fs_mutex); |
| 166 | btrfs_btree_balance_dirty(root, nr); |
| 167 | btrfs_throttle(root); |
| 168 | return ret; |
| 169 | } |
| 170 | |
| 171 | static int create_snapshot(struct btrfs_root *root, char *name, int namelen) |
| 172 | { |
| 173 | struct btrfs_pending_snapshot *pending_snapshot; |
| 174 | struct btrfs_trans_handle *trans; |
| 175 | int ret; |
| 176 | int err; |
| 177 | unsigned long nr = 0; |
| 178 | |
| 179 | if (!root->ref_cows) |
| 180 | return -EINVAL; |
| 181 | |
| 182 | mutex_lock(&root->fs_info->fs_mutex); |
| 183 | ret = btrfs_check_free_space(root, 1, 0); |
| 184 | if (ret) |
| 185 | goto fail_unlock; |
| 186 | |
| 187 | pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS); |
| 188 | if (!pending_snapshot) { |
| 189 | ret = -ENOMEM; |
| 190 | goto fail_unlock; |
| 191 | } |
| 192 | pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); |
| 193 | if (!pending_snapshot->name) { |
| 194 | ret = -ENOMEM; |
| 195 | kfree(pending_snapshot); |
| 196 | goto fail_unlock; |
| 197 | } |
| 198 | memcpy(pending_snapshot->name, name, namelen); |
| 199 | pending_snapshot->name[namelen] = '\0'; |
| 200 | trans = btrfs_start_transaction(root, 1); |
| 201 | BUG_ON(!trans); |
| 202 | pending_snapshot->root = root; |
| 203 | list_add(&pending_snapshot->list, |
| 204 | &trans->transaction->pending_snapshots); |
| 205 | ret = btrfs_update_inode(trans, root, root->inode); |
| 206 | err = btrfs_commit_transaction(trans, root); |
| 207 | |
| 208 | fail_unlock: |
| 209 | mutex_unlock(&root->fs_info->fs_mutex); |
| 210 | btrfs_btree_balance_dirty(root, nr); |
| 211 | btrfs_throttle(root); |
| 212 | return ret; |
| 213 | } |
| 214 | |
| 215 | int btrfs_defrag_file(struct file *file) |
| 216 | { |
| 217 | struct inode *inode = fdentry(file)->d_inode; |
| 218 | struct btrfs_root *root = BTRFS_I(inode)->root; |
| 219 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
| 220 | struct page *page; |
| 221 | unsigned long last_index; |
| 222 | unsigned long ra_pages = root->fs_info->bdi.ra_pages; |
| 223 | unsigned long total_read = 0; |
| 224 | u64 page_start; |
| 225 | u64 page_end; |
| 226 | unsigned long i; |
| 227 | int ret; |
| 228 | |
| 229 | mutex_lock(&root->fs_info->fs_mutex); |
| 230 | ret = btrfs_check_free_space(root, inode->i_size, 0); |
| 231 | mutex_unlock(&root->fs_info->fs_mutex); |
| 232 | if (ret) |
| 233 | return -ENOSPC; |
| 234 | |
| 235 | mutex_lock(&inode->i_mutex); |
| 236 | last_index = inode->i_size >> PAGE_CACHE_SHIFT; |
| 237 | for (i = 0; i <= last_index; i++) { |
| 238 | if (total_read % ra_pages == 0) { |
| 239 | btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, |
| 240 | min(last_index, i + ra_pages - 1)); |
| 241 | } |
| 242 | total_read++; |
| 243 | page = grab_cache_page(inode->i_mapping, i); |
| 244 | if (!page) |
| 245 | goto out_unlock; |
| 246 | if (!PageUptodate(page)) { |
| 247 | btrfs_readpage(NULL, page); |
| 248 | lock_page(page); |
| 249 | if (!PageUptodate(page)) { |
| 250 | unlock_page(page); |
| 251 | page_cache_release(page); |
| 252 | goto out_unlock; |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
| 257 | ClearPageDirty(page); |
| 258 | #else |
| 259 | cancel_dirty_page(page, PAGE_CACHE_SIZE); |
| 260 | #endif |
| 261 | wait_on_page_writeback(page); |
| 262 | set_page_extent_mapped(page); |
| 263 | |
| 264 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; |
| 265 | page_end = page_start + PAGE_CACHE_SIZE - 1; |
| 266 | |
| 267 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
| 268 | set_extent_delalloc(io_tree, page_start, |
| 269 | page_end, GFP_NOFS); |
| 270 | |
| 271 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); |
| 272 | set_page_dirty(page); |
| 273 | unlock_page(page); |
| 274 | page_cache_release(page); |
| 275 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); |
| 276 | } |
| 277 | |
| 278 | out_unlock: |
| 279 | mutex_unlock(&inode->i_mutex); |
| 280 | return 0; |
| 281 | } |
| 282 | |
| 283 | /* |
| 284 | * Called inside transaction, so use GFP_NOFS |
| 285 | */ |
| 286 | |
| 287 | static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) |
| 288 | { |
| 289 | u64 new_size; |
| 290 | u64 old_size; |
| 291 | u64 devid = 1; |
| 292 | struct btrfs_ioctl_vol_args *vol_args; |
| 293 | struct btrfs_trans_handle *trans; |
| 294 | struct btrfs_device *device = NULL; |
| 295 | char *sizestr; |
| 296 | char *devstr = NULL; |
| 297 | int ret = 0; |
| 298 | int namelen; |
| 299 | int mod = 0; |
| 300 | |
| 301 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); |
| 302 | |
| 303 | if (!vol_args) |
| 304 | return -ENOMEM; |
| 305 | |
| 306 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { |
| 307 | ret = -EFAULT; |
| 308 | goto out; |
| 309 | } |
| 310 | namelen = strlen(vol_args->name); |
| 311 | if (namelen > BTRFS_VOL_NAME_MAX) { |
| 312 | ret = -EINVAL; |
| 313 | goto out; |
| 314 | } |
| 315 | |
| 316 | mutex_lock(&root->fs_info->fs_mutex); |
| 317 | sizestr = vol_args->name; |
| 318 | devstr = strchr(sizestr, ':'); |
| 319 | if (devstr) { |
| 320 | char *end; |
| 321 | sizestr = devstr + 1; |
| 322 | *devstr = '\0'; |
| 323 | devstr = vol_args->name; |
| 324 | devid = simple_strtoull(devstr, &end, 10); |
| 325 | printk(KERN_INFO "resizing devid %llu\n", devid); |
| 326 | } |
| 327 | device = btrfs_find_device(root, devid, NULL); |
| 328 | if (!device) { |
| 329 | printk(KERN_INFO "resizer unable to find device %llu\n", devid); |
| 330 | ret = -EINVAL; |
| 331 | goto out_unlock; |
| 332 | } |
| 333 | if (!strcmp(sizestr, "max")) |
| 334 | new_size = device->bdev->bd_inode->i_size; |
| 335 | else { |
| 336 | if (sizestr[0] == '-') { |
| 337 | mod = -1; |
| 338 | sizestr++; |
| 339 | } else if (sizestr[0] == '+') { |
| 340 | mod = 1; |
| 341 | sizestr++; |
| 342 | } |
| 343 | new_size = btrfs_parse_size(sizestr); |
| 344 | if (new_size == 0) { |
| 345 | ret = -EINVAL; |
| 346 | goto out_unlock; |
| 347 | } |
| 348 | } |
| 349 | |
| 350 | old_size = device->total_bytes; |
| 351 | |
| 352 | if (mod < 0) { |
| 353 | if (new_size > old_size) { |
| 354 | ret = -EINVAL; |
| 355 | goto out_unlock; |
| 356 | } |
| 357 | new_size = old_size - new_size; |
| 358 | } else if (mod > 0) { |
| 359 | new_size = old_size + new_size; |
| 360 | } |
| 361 | |
| 362 | if (new_size < 256 * 1024 * 1024) { |
| 363 | ret = -EINVAL; |
| 364 | goto out_unlock; |
| 365 | } |
| 366 | if (new_size > device->bdev->bd_inode->i_size) { |
| 367 | ret = -EFBIG; |
| 368 | goto out_unlock; |
| 369 | } |
| 370 | |
| 371 | do_div(new_size, root->sectorsize); |
| 372 | new_size *= root->sectorsize; |
| 373 | |
| 374 | printk(KERN_INFO "new size for %s is %llu\n", |
| 375 | device->name, (unsigned long long)new_size); |
| 376 | |
| 377 | if (new_size > old_size) { |
| 378 | trans = btrfs_start_transaction(root, 1); |
| 379 | ret = btrfs_grow_device(trans, device, new_size); |
| 380 | btrfs_commit_transaction(trans, root); |
| 381 | } else { |
| 382 | ret = btrfs_shrink_device(device, new_size); |
| 383 | } |
| 384 | |
| 385 | out_unlock: |
| 386 | mutex_unlock(&root->fs_info->fs_mutex); |
| 387 | out: |
| 388 | kfree(vol_args); |
| 389 | return ret; |
| 390 | } |
| 391 | |
| 392 | static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, |
| 393 | void __user *arg) |
| 394 | { |
| 395 | struct btrfs_ioctl_vol_args *vol_args; |
| 396 | struct btrfs_dir_item *di; |
| 397 | struct btrfs_path *path; |
| 398 | u64 root_dirid; |
| 399 | int namelen; |
| 400 | int ret; |
| 401 | |
| 402 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); |
| 403 | |
| 404 | if (!vol_args) |
| 405 | return -ENOMEM; |
| 406 | |
| 407 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { |
| 408 | ret = -EFAULT; |
| 409 | goto out; |
| 410 | } |
| 411 | |
| 412 | namelen = strlen(vol_args->name); |
| 413 | if (namelen > BTRFS_VOL_NAME_MAX) { |
| 414 | ret = -EINVAL; |
| 415 | goto out; |
| 416 | } |
| 417 | if (strchr(vol_args->name, '/')) { |
| 418 | ret = -EINVAL; |
| 419 | goto out; |
| 420 | } |
| 421 | |
| 422 | path = btrfs_alloc_path(); |
| 423 | if (!path) { |
| 424 | ret = -ENOMEM; |
| 425 | goto out; |
| 426 | } |
| 427 | |
| 428 | root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, |
| 429 | mutex_lock(&root->fs_info->fs_mutex); |
| 430 | di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, |
| 431 | path, root_dirid, |
| 432 | vol_args->name, namelen, 0); |
| 433 | mutex_unlock(&root->fs_info->fs_mutex); |
| 434 | btrfs_free_path(path); |
| 435 | |
| 436 | if (di && !IS_ERR(di)) { |
| 437 | ret = -EEXIST; |
| 438 | goto out; |
| 439 | } |
| 440 | |
| 441 | if (IS_ERR(di)) { |
| 442 | ret = PTR_ERR(di); |
| 443 | goto out; |
| 444 | } |
| 445 | |
| 446 | if (root == root->fs_info->tree_root) |
| 447 | ret = create_subvol(root, vol_args->name, namelen); |
| 448 | else |
| 449 | ret = create_snapshot(root, vol_args->name, namelen); |
| 450 | out: |
| 451 | kfree(vol_args); |
| 452 | return ret; |
| 453 | } |
| 454 | |
| 455 | static int btrfs_ioctl_defrag(struct file *file) |
| 456 | { |
| 457 | struct inode *inode = fdentry(file)->d_inode; |
| 458 | struct btrfs_root *root = BTRFS_I(inode)->root; |
| 459 | |
| 460 | switch (inode->i_mode & S_IFMT) { |
| 461 | case S_IFDIR: |
| 462 | mutex_lock(&root->fs_info->fs_mutex); |
| 463 | btrfs_defrag_root(root, 0); |
| 464 | btrfs_defrag_root(root->fs_info->extent_root, 0); |
| 465 | mutex_unlock(&root->fs_info->fs_mutex); |
| 466 | break; |
| 467 | case S_IFREG: |
| 468 | btrfs_defrag_file(file); |
| 469 | break; |
| 470 | } |
| 471 | |
| 472 | return 0; |
| 473 | } |
| 474 | |
| 475 | long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) |
| 476 | { |
| 477 | struct btrfs_ioctl_vol_args *vol_args; |
| 478 | int ret; |
| 479 | |
| 480 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); |
| 481 | |
| 482 | if (!vol_args) |
| 483 | return -ENOMEM; |
| 484 | |
| 485 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { |
| 486 | ret = -EFAULT; |
| 487 | goto out; |
| 488 | } |
| 489 | ret = btrfs_init_new_device(root, vol_args->name); |
| 490 | |
| 491 | out: |
| 492 | kfree(vol_args); |
| 493 | return ret; |
| 494 | } |
| 495 | |
| 496 | long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) |
| 497 | { |
| 498 | struct btrfs_ioctl_vol_args *vol_args; |
| 499 | int ret; |
| 500 | |
| 501 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); |
| 502 | |
| 503 | if (!vol_args) |
| 504 | return -ENOMEM; |
| 505 | |
| 506 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { |
| 507 | ret = -EFAULT; |
| 508 | goto out; |
| 509 | } |
| 510 | ret = btrfs_rm_device(root, vol_args->name); |
| 511 | |
| 512 | out: |
| 513 | kfree(vol_args); |
| 514 | return ret; |
| 515 | } |
| 516 | |
| 517 | int dup_item_to_inode(struct btrfs_trans_handle *trans, |
| 518 | struct btrfs_root *root, |
| 519 | struct btrfs_path *path, |
| 520 | struct extent_buffer *leaf, |
| 521 | int slot, |
| 522 | struct btrfs_key *key, |
| 523 | u64 destino) |
| 524 | { |
| 525 | char *dup; |
| 526 | int len = btrfs_item_size_nr(leaf, slot); |
| 527 | struct btrfs_key ckey = *key; |
| 528 | int ret = 0; |
| 529 | |
| 530 | dup = kmalloc(len, GFP_NOFS); |
| 531 | if (!dup) |
| 532 | return -ENOMEM; |
| 533 | |
| 534 | read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len); |
| 535 | btrfs_release_path(root, path); |
| 536 | |
| 537 | ckey.objectid = destino; |
| 538 | ret = btrfs_insert_item(trans, root, &ckey, dup, len); |
| 539 | kfree(dup); |
| 540 | return ret; |
| 541 | } |
| 542 | |
| 543 | long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) |
| 544 | { |
| 545 | struct inode *inode = fdentry(file)->d_inode; |
| 546 | struct btrfs_root *root = BTRFS_I(inode)->root; |
| 547 | struct file *src_file; |
| 548 | struct inode *src; |
| 549 | struct btrfs_trans_handle *trans; |
| 550 | int ret; |
| 551 | u64 pos; |
| 552 | struct btrfs_path *path; |
| 553 | struct btrfs_key key; |
| 554 | struct extent_buffer *leaf; |
| 555 | u32 nritems; |
| 556 | int slot; |
| 557 | |
| 558 | src_file = fget(src_fd); |
| 559 | if (!src_file) |
| 560 | return -EBADF; |
| 561 | src = src_file->f_dentry->d_inode; |
| 562 | |
| 563 | ret = -EXDEV; |
| 564 | if (src->i_sb != inode->i_sb) |
| 565 | goto out_fput; |
| 566 | |
| 567 | if (inode < src) { |
| 568 | mutex_lock(&inode->i_mutex); |
| 569 | mutex_lock(&src->i_mutex); |
| 570 | } else { |
| 571 | mutex_lock(&src->i_mutex); |
| 572 | mutex_lock(&inode->i_mutex); |
| 573 | } |
| 574 | |
| 575 | ret = -ENOTEMPTY; |
| 576 | if (inode->i_size) |
| 577 | goto out_unlock; |
| 578 | |
| 579 | /* do any pending delalloc/csum calc on src, one way or |
| 580 | another, and lock file content */ |
| 581 | while (1) { |
| 582 | filemap_write_and_wait(src->i_mapping); |
| 583 | lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); |
| 584 | if (BTRFS_I(src)->delalloc_bytes == 0) |
| 585 | break; |
| 586 | unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); |
| 587 | } |
| 588 | |
| 589 | mutex_lock(&root->fs_info->fs_mutex); |
| 590 | trans = btrfs_start_transaction(root, 0); |
| 591 | path = btrfs_alloc_path(); |
| 592 | if (!path) { |
| 593 | ret = -ENOMEM; |
| 594 | goto out; |
| 595 | } |
| 596 | key.offset = 0; |
| 597 | key.type = BTRFS_EXTENT_DATA_KEY; |
| 598 | key.objectid = src->i_ino; |
| 599 | pos = 0; |
| 600 | path->reada = 2; |
| 601 | |
| 602 | while (1) { |
| 603 | /* |
| 604 | * note the key will change type as we walk through the |
| 605 | * tree. |
| 606 | */ |
| 607 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); |
| 608 | if (ret < 0) |
| 609 | goto out; |
| 610 | |
| 611 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { |
| 612 | ret = btrfs_next_leaf(root, path); |
| 613 | if (ret < 0) |
| 614 | goto out; |
| 615 | if (ret > 0) |
| 616 | break; |
| 617 | } |
| 618 | leaf = path->nodes[0]; |
| 619 | slot = path->slots[0]; |
| 620 | btrfs_item_key_to_cpu(leaf, &key, slot); |
| 621 | nritems = btrfs_header_nritems(leaf); |
| 622 | |
| 623 | if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || |
| 624 | key.objectid != src->i_ino) |
| 625 | break; |
| 626 | |
| 627 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
| 628 | struct btrfs_file_extent_item *extent; |
| 629 | int found_type; |
| 630 | pos = key.offset; |
| 631 | extent = btrfs_item_ptr(leaf, slot, |
| 632 | struct btrfs_file_extent_item); |
| 633 | found_type = btrfs_file_extent_type(leaf, extent); |
| 634 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
| 635 | u64 len = btrfs_file_extent_num_bytes(leaf, |
| 636 | extent); |
| 637 | u64 ds = btrfs_file_extent_disk_bytenr(leaf, |
| 638 | extent); |
| 639 | u64 dl = btrfs_file_extent_disk_num_bytes(leaf, |
| 640 | extent); |
| 641 | u64 off = btrfs_file_extent_offset(leaf, |
| 642 | extent); |
| 643 | btrfs_insert_file_extent(trans, root, |
| 644 | inode->i_ino, pos, |
| 645 | ds, dl, len, off); |
| 646 | /* ds == 0 means there's a hole */ |
| 647 | if (ds != 0) { |
| 648 | btrfs_inc_extent_ref(trans, root, |
| 649 | ds, dl, |
| 650 | root->root_key.objectid, |
| 651 | trans->transid, |
| 652 | inode->i_ino, pos); |
| 653 | } |
| 654 | pos = key.offset + len; |
| 655 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { |
| 656 | ret = dup_item_to_inode(trans, root, path, |
| 657 | leaf, slot, &key, |
| 658 | inode->i_ino); |
| 659 | if (ret) |
| 660 | goto out; |
| 661 | pos = key.offset + btrfs_item_size_nr(leaf, |
| 662 | slot); |
| 663 | } |
| 664 | } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { |
| 665 | ret = dup_item_to_inode(trans, root, path, leaf, |
| 666 | slot, &key, inode->i_ino); |
| 667 | |
| 668 | if (ret) |
| 669 | goto out; |
| 670 | } |
| 671 | key.offset++; |
| 672 | btrfs_release_path(root, path); |
| 673 | } |
| 674 | |
| 675 | ret = 0; |
| 676 | out: |
| 677 | btrfs_free_path(path); |
| 678 | |
| 679 | inode->i_blocks = src->i_blocks; |
| 680 | i_size_write(inode, src->i_size); |
| 681 | btrfs_update_inode(trans, root, inode); |
| 682 | |
| 683 | unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); |
| 684 | |
| 685 | btrfs_end_transaction(trans, root); |
| 686 | mutex_unlock(&root->fs_info->fs_mutex); |
| 687 | |
| 688 | out_unlock: |
| 689 | mutex_unlock(&src->i_mutex); |
| 690 | mutex_unlock(&inode->i_mutex); |
| 691 | out_fput: |
| 692 | fput(src_file); |
| 693 | return ret; |
| 694 | } |
| 695 | |
| 696 | /* |
| 697 | * there are many ways the trans_start and trans_end ioctls can lead |
| 698 | * to deadlocks. They should only be used by applications that |
| 699 | * basically own the machine, and have a very in depth understanding |
| 700 | * of all the possible deadlocks and enospc problems. |
| 701 | */ |
| 702 | long btrfs_ioctl_trans_start(struct file *file) |
| 703 | { |
| 704 | struct inode *inode = fdentry(file)->d_inode; |
| 705 | struct btrfs_root *root = BTRFS_I(inode)->root; |
| 706 | struct btrfs_trans_handle *trans; |
| 707 | int ret = 0; |
| 708 | |
| 709 | mutex_lock(&root->fs_info->fs_mutex); |
| 710 | if (file->private_data) { |
| 711 | ret = -EINPROGRESS; |
| 712 | goto out; |
| 713 | } |
| 714 | trans = btrfs_start_transaction(root, 0); |
| 715 | if (trans) |
| 716 | file->private_data = trans; |
| 717 | else |
| 718 | ret = -ENOMEM; |
| 719 | /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ |
| 720 | out: |
| 721 | mutex_unlock(&root->fs_info->fs_mutex); |
| 722 | return ret; |
| 723 | } |
| 724 | |
| 725 | /* |
| 726 | * there are many ways the trans_start and trans_end ioctls can lead |
| 727 | * to deadlocks. They should only be used by applications that |
| 728 | * basically own the machine, and have a very in depth understanding |
| 729 | * of all the possible deadlocks and enospc problems. |
| 730 | */ |
| 731 | long btrfs_ioctl_trans_end(struct file *file) |
| 732 | { |
| 733 | struct inode *inode = fdentry(file)->d_inode; |
| 734 | struct btrfs_root *root = BTRFS_I(inode)->root; |
| 735 | struct btrfs_trans_handle *trans; |
| 736 | int ret = 0; |
| 737 | |
| 738 | mutex_lock(&root->fs_info->fs_mutex); |
| 739 | trans = file->private_data; |
| 740 | if (!trans) { |
| 741 | ret = -EINVAL; |
| 742 | goto out; |
| 743 | } |
| 744 | btrfs_end_transaction(trans, root); |
| 745 | file->private_data = 0; |
| 746 | out: |
| 747 | mutex_unlock(&root->fs_info->fs_mutex); |
| 748 | return ret; |
| 749 | } |
| 750 | |
| 751 | long btrfs_ioctl(struct file *file, unsigned int |
| 752 | cmd, unsigned long arg) |
| 753 | { |
| 754 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; |
| 755 | |
| 756 | switch (cmd) { |
| 757 | case BTRFS_IOC_SNAP_CREATE: |
| 758 | return btrfs_ioctl_snap_create(root, (void __user *)arg); |
| 759 | case BTRFS_IOC_DEFRAG: |
| 760 | return btrfs_ioctl_defrag(file); |
| 761 | case BTRFS_IOC_RESIZE: |
| 762 | return btrfs_ioctl_resize(root, (void __user *)arg); |
| 763 | case BTRFS_IOC_ADD_DEV: |
| 764 | return btrfs_ioctl_add_dev(root, (void __user *)arg); |
| 765 | case BTRFS_IOC_RM_DEV: |
| 766 | return btrfs_ioctl_rm_dev(root, (void __user *)arg); |
| 767 | case BTRFS_IOC_BALANCE: |
| 768 | return btrfs_balance(root->fs_info->dev_root); |
| 769 | case BTRFS_IOC_CLONE: |
| 770 | return btrfs_ioctl_clone(file, arg); |
| 771 | case BTRFS_IOC_TRANS_START: |
| 772 | return btrfs_ioctl_trans_start(file); |
| 773 | case BTRFS_IOC_TRANS_END: |
| 774 | return btrfs_ioctl_trans_end(file); |
| 775 | case BTRFS_IOC_SYNC: |
| 776 | btrfs_sync_fs(file->f_dentry->d_sb, 1); |
| 777 | return 0; |
| 778 | } |
| 779 | |
| 780 | return -ENOTTY; |
| 781 | } |