| 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/fs.h> |
| 20 | #include <linux/blkdev.h> |
| 21 | #include <linux/crc32c.h> |
| 22 | #include <linux/scatterlist.h> |
| 23 | #include <linux/swap.h> |
| 24 | #include <linux/radix-tree.h> |
| 25 | #include <linux/writeback.h> |
| 26 | #include <linux/buffer_head.h> // for block_sync_page |
| 27 | #include "ctree.h" |
| 28 | #include "disk-io.h" |
| 29 | #include "transaction.h" |
| 30 | #include "btrfs_inode.h" |
| 31 | #include "print-tree.h" |
| 32 | |
| 33 | #if 0 |
| 34 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) |
| 35 | { |
| 36 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
| 37 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", |
| 38 | (unsigned long long)extent_buffer_blocknr(buf), |
| 39 | (unsigned long long)btrfs_header_blocknr(buf)); |
| 40 | return 1; |
| 41 | } |
| 42 | return 0; |
| 43 | } |
| 44 | #endif |
| 45 | |
| 46 | static struct extent_map_ops btree_extent_map_ops; |
| 47 | |
| 48 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
| 49 | u64 bytenr, u32 blocksize) |
| 50 | { |
| 51 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 52 | struct extent_buffer *eb; |
| 53 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, |
| 54 | bytenr, blocksize, GFP_NOFS); |
| 55 | return eb; |
| 56 | } |
| 57 | |
| 58 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, |
| 59 | u64 bytenr, u32 blocksize) |
| 60 | { |
| 61 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 62 | struct extent_buffer *eb; |
| 63 | |
| 64 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, |
| 65 | bytenr, blocksize, NULL, GFP_NOFS); |
| 66 | return eb; |
| 67 | } |
| 68 | |
| 69 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
| 70 | size_t page_offset, u64 start, u64 end, |
| 71 | int create) |
| 72 | { |
| 73 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
| 74 | struct extent_map *em; |
| 75 | int ret; |
| 76 | |
| 77 | again: |
| 78 | em = lookup_extent_mapping(em_tree, start, end); |
| 79 | if (em) { |
| 80 | goto out; |
| 81 | } |
| 82 | em = alloc_extent_map(GFP_NOFS); |
| 83 | if (!em) { |
| 84 | em = ERR_PTR(-ENOMEM); |
| 85 | goto out; |
| 86 | } |
| 87 | em->start = 0; |
| 88 | em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1; |
| 89 | em->block_start = 0; |
| 90 | em->block_end = em->end; |
| 91 | em->bdev = inode->i_sb->s_bdev; |
| 92 | ret = add_extent_mapping(em_tree, em); |
| 93 | if (ret == -EEXIST) { |
| 94 | free_extent_map(em); |
| 95 | em = NULL; |
| 96 | goto again; |
| 97 | } else if (ret) { |
| 98 | em = ERR_PTR(ret); |
| 99 | } |
| 100 | out: |
| 101 | return em; |
| 102 | } |
| 103 | |
| 104 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
| 105 | { |
| 106 | return crc32c(seed, data, len); |
| 107 | } |
| 108 | |
| 109 | void btrfs_csum_final(u32 crc, char *result) |
| 110 | { |
| 111 | *(__le32 *)result = ~cpu_to_le32(crc); |
| 112 | } |
| 113 | |
| 114 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
| 115 | int verify) |
| 116 | { |
| 117 | char result[BTRFS_CRC32_SIZE]; |
| 118 | unsigned long len; |
| 119 | unsigned long cur_len; |
| 120 | unsigned long offset = BTRFS_CSUM_SIZE; |
| 121 | char *map_token = NULL; |
| 122 | char *kaddr; |
| 123 | unsigned long map_start; |
| 124 | unsigned long map_len; |
| 125 | int err; |
| 126 | u32 crc = ~(u32)0; |
| 127 | |
| 128 | len = buf->len - offset; |
| 129 | while(len > 0) { |
| 130 | err = map_private_extent_buffer(buf, offset, 32, |
| 131 | &map_token, &kaddr, |
| 132 | &map_start, &map_len, KM_USER0); |
| 133 | if (err) { |
| 134 | printk("failed to map extent buffer! %lu\n", |
| 135 | offset); |
| 136 | return 1; |
| 137 | } |
| 138 | cur_len = min(len, map_len - (offset - map_start)); |
| 139 | crc = btrfs_csum_data(root, kaddr + offset - map_start, |
| 140 | crc, cur_len); |
| 141 | len -= cur_len; |
| 142 | offset += cur_len; |
| 143 | unmap_extent_buffer(buf, map_token, KM_USER0); |
| 144 | } |
| 145 | btrfs_csum_final(crc, result); |
| 146 | |
| 147 | if (verify) { |
| 148 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
| 149 | printk("btrfs: %s checksum verify failed on %llu\n", |
| 150 | root->fs_info->sb->s_id, |
| 151 | buf->start); |
| 152 | return 1; |
| 153 | } |
| 154 | } else { |
| 155 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); |
| 156 | } |
| 157 | return 0; |
| 158 | } |
| 159 | |
| 160 | |
| 161 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
| 162 | { |
| 163 | struct extent_map_tree *tree; |
| 164 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
| 165 | u64 found_start; |
| 166 | int found_level; |
| 167 | unsigned long len; |
| 168 | struct extent_buffer *eb; |
| 169 | tree = &BTRFS_I(page->mapping->host)->extent_tree; |
| 170 | |
| 171 | if (page->private == EXTENT_PAGE_PRIVATE) |
| 172 | goto out; |
| 173 | if (!page->private) |
| 174 | goto out; |
| 175 | len = page->private >> 2; |
| 176 | if (len == 0) { |
| 177 | WARN_ON(1); |
| 178 | } |
| 179 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
| 180 | read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1); |
| 181 | found_start = btrfs_header_bytenr(eb); |
| 182 | if (found_start != start) { |
| 183 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", |
| 184 | start, found_start, len); |
| 185 | } |
| 186 | found_level = btrfs_header_level(eb); |
| 187 | csum_tree_block(root, eb, 0); |
| 188 | free_extent_buffer(eb); |
| 189 | out: |
| 190 | return 0; |
| 191 | } |
| 192 | |
| 193 | static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) |
| 194 | { |
| 195 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
| 196 | |
| 197 | csum_dirty_buffer(root, page); |
| 198 | return 0; |
| 199 | } |
| 200 | |
| 201 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
| 202 | { |
| 203 | struct extent_map_tree *tree; |
| 204 | tree = &BTRFS_I(page->mapping->host)->extent_tree; |
| 205 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
| 206 | } |
| 207 | |
| 208 | static int btree_writepages(struct address_space *mapping, |
| 209 | struct writeback_control *wbc) |
| 210 | { |
| 211 | struct extent_map_tree *tree; |
| 212 | tree = &BTRFS_I(mapping->host)->extent_tree; |
| 213 | if (wbc->sync_mode == WB_SYNC_NONE) { |
| 214 | u64 num_dirty; |
| 215 | u64 start = 0; |
| 216 | unsigned long thresh = 96 * 1024 * 1024; |
| 217 | |
| 218 | if (wbc->for_kupdate) |
| 219 | return 0; |
| 220 | |
| 221 | if (current_is_pdflush()) { |
| 222 | thresh = 96 * 1024 * 1024; |
| 223 | } else { |
| 224 | thresh = 8 * 1024 * 1024; |
| 225 | } |
| 226 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
| 227 | thresh, EXTENT_DIRTY); |
| 228 | if (num_dirty < thresh) { |
| 229 | return 0; |
| 230 | } |
| 231 | } |
| 232 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
| 233 | } |
| 234 | |
| 235 | int btree_readpage(struct file *file, struct page *page) |
| 236 | { |
| 237 | struct extent_map_tree *tree; |
| 238 | tree = &BTRFS_I(page->mapping->host)->extent_tree; |
| 239 | return extent_read_full_page(tree, page, btree_get_extent); |
| 240 | } |
| 241 | |
| 242 | static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) |
| 243 | { |
| 244 | struct extent_map_tree *tree; |
| 245 | int ret; |
| 246 | |
| 247 | tree = &BTRFS_I(page->mapping->host)->extent_tree; |
| 248 | ret = try_release_extent_mapping(tree, page); |
| 249 | if (ret == 1) { |
| 250 | ClearPagePrivate(page); |
| 251 | set_page_private(page, 0); |
| 252 | page_cache_release(page); |
| 253 | } |
| 254 | return ret; |
| 255 | } |
| 256 | |
| 257 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
| 258 | { |
| 259 | struct extent_map_tree *tree; |
| 260 | tree = &BTRFS_I(page->mapping->host)->extent_tree; |
| 261 | extent_invalidatepage(tree, page, offset); |
| 262 | btree_releasepage(page, GFP_NOFS); |
| 263 | } |
| 264 | |
| 265 | #if 0 |
| 266 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
| 267 | { |
| 268 | struct buffer_head *bh; |
| 269 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
| 270 | struct buffer_head *head; |
| 271 | if (!page_has_buffers(page)) { |
| 272 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, |
| 273 | (1 << BH_Dirty)|(1 << BH_Uptodate)); |
| 274 | } |
| 275 | head = page_buffers(page); |
| 276 | bh = head; |
| 277 | do { |
| 278 | if (buffer_dirty(bh)) |
| 279 | csum_tree_block(root, bh, 0); |
| 280 | bh = bh->b_this_page; |
| 281 | } while (bh != head); |
| 282 | return block_write_full_page(page, btree_get_block, wbc); |
| 283 | } |
| 284 | #endif |
| 285 | |
| 286 | static struct address_space_operations btree_aops = { |
| 287 | .readpage = btree_readpage, |
| 288 | .writepage = btree_writepage, |
| 289 | .writepages = btree_writepages, |
| 290 | .releasepage = btree_releasepage, |
| 291 | .invalidatepage = btree_invalidatepage, |
| 292 | .sync_page = block_sync_page, |
| 293 | }; |
| 294 | |
| 295 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) |
| 296 | { |
| 297 | struct extent_buffer *buf = NULL; |
| 298 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 299 | int ret = 0; |
| 300 | |
| 301 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
| 302 | if (!buf) |
| 303 | return 0; |
| 304 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, |
| 305 | buf, 0, 0); |
| 306 | free_extent_buffer(buf); |
| 307 | return ret; |
| 308 | } |
| 309 | |
| 310 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
| 311 | u32 blocksize) |
| 312 | { |
| 313 | struct extent_buffer *buf = NULL; |
| 314 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 315 | struct extent_map_tree *extent_tree; |
| 316 | int ret; |
| 317 | |
| 318 | extent_tree = &BTRFS_I(btree_inode)->extent_tree; |
| 319 | |
| 320 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
| 321 | if (!buf) |
| 322 | return NULL; |
| 323 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, |
| 324 | buf, 0, 1); |
| 325 | if (buf->flags & EXTENT_CSUM) { |
| 326 | return buf; |
| 327 | } |
| 328 | if (test_range_bit(extent_tree, buf->start, buf->start + buf->len - 1, |
| 329 | EXTENT_CSUM, 1)) { |
| 330 | buf->flags |= EXTENT_CSUM; |
| 331 | return buf; |
| 332 | } |
| 333 | ret = csum_tree_block(root, buf, 1); |
| 334 | set_extent_bits(extent_tree, buf->start, |
| 335 | buf->start + buf->len - 1, |
| 336 | EXTENT_CSUM, GFP_NOFS); |
| 337 | buf->flags |= EXTENT_CSUM; |
| 338 | return buf; |
| 339 | } |
| 340 | |
| 341 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
| 342 | struct extent_buffer *buf) |
| 343 | { |
| 344 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 345 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); |
| 346 | return 0; |
| 347 | } |
| 348 | |
| 349 | int wait_on_tree_block_writeback(struct btrfs_root *root, |
| 350 | struct extent_buffer *buf) |
| 351 | { |
| 352 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 353 | wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree, |
| 354 | buf); |
| 355 | return 0; |
| 356 | } |
| 357 | |
| 358 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
| 359 | u32 stripesize, struct btrfs_root *root, |
| 360 | struct btrfs_fs_info *fs_info, |
| 361 | u64 objectid) |
| 362 | { |
| 363 | root->node = NULL; |
| 364 | root->inode = NULL; |
| 365 | root->commit_root = NULL; |
| 366 | root->sectorsize = sectorsize; |
| 367 | root->nodesize = nodesize; |
| 368 | root->leafsize = leafsize; |
| 369 | root->stripesize = stripesize; |
| 370 | root->ref_cows = 0; |
| 371 | root->fs_info = fs_info; |
| 372 | root->objectid = objectid; |
| 373 | root->last_trans = 0; |
| 374 | root->highest_inode = 0; |
| 375 | root->last_inode_alloc = 0; |
| 376 | root->name = NULL; |
| 377 | memset(&root->root_key, 0, sizeof(root->root_key)); |
| 378 | memset(&root->root_item, 0, sizeof(root->root_item)); |
| 379 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
| 380 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
| 381 | init_completion(&root->kobj_unregister); |
| 382 | init_rwsem(&root->snap_sem); |
| 383 | root->defrag_running = 0; |
| 384 | root->defrag_level = 0; |
| 385 | root->root_key.objectid = objectid; |
| 386 | return 0; |
| 387 | } |
| 388 | |
| 389 | static int find_and_setup_root(struct btrfs_root *tree_root, |
| 390 | struct btrfs_fs_info *fs_info, |
| 391 | u64 objectid, |
| 392 | struct btrfs_root *root) |
| 393 | { |
| 394 | int ret; |
| 395 | u32 blocksize; |
| 396 | |
| 397 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
| 398 | tree_root->sectorsize, tree_root->stripesize, |
| 399 | root, fs_info, objectid); |
| 400 | ret = btrfs_find_last_root(tree_root, objectid, |
| 401 | &root->root_item, &root->root_key); |
| 402 | BUG_ON(ret); |
| 403 | |
| 404 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
| 405 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
| 406 | blocksize); |
| 407 | BUG_ON(!root->node); |
| 408 | return 0; |
| 409 | } |
| 410 | |
| 411 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, |
| 412 | struct btrfs_key *location) |
| 413 | { |
| 414 | struct btrfs_root *root; |
| 415 | struct btrfs_root *tree_root = fs_info->tree_root; |
| 416 | struct btrfs_path *path; |
| 417 | struct extent_buffer *l; |
| 418 | u64 highest_inode; |
| 419 | u32 blocksize; |
| 420 | int ret = 0; |
| 421 | |
| 422 | root = kzalloc(sizeof(*root), GFP_NOFS); |
| 423 | if (!root) |
| 424 | return ERR_PTR(-ENOMEM); |
| 425 | if (location->offset == (u64)-1) { |
| 426 | ret = find_and_setup_root(tree_root, fs_info, |
| 427 | location->objectid, root); |
| 428 | if (ret) { |
| 429 | kfree(root); |
| 430 | return ERR_PTR(ret); |
| 431 | } |
| 432 | goto insert; |
| 433 | } |
| 434 | |
| 435 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
| 436 | tree_root->sectorsize, tree_root->stripesize, |
| 437 | root, fs_info, location->objectid); |
| 438 | |
| 439 | path = btrfs_alloc_path(); |
| 440 | BUG_ON(!path); |
| 441 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
| 442 | if (ret != 0) { |
| 443 | if (ret > 0) |
| 444 | ret = -ENOENT; |
| 445 | goto out; |
| 446 | } |
| 447 | l = path->nodes[0]; |
| 448 | read_extent_buffer(l, &root->root_item, |
| 449 | btrfs_item_ptr_offset(l, path->slots[0]), |
| 450 | sizeof(root->root_item)); |
| 451 | memcpy(&root->root_key, location, sizeof(*location)); |
| 452 | ret = 0; |
| 453 | out: |
| 454 | btrfs_release_path(root, path); |
| 455 | btrfs_free_path(path); |
| 456 | if (ret) { |
| 457 | kfree(root); |
| 458 | return ERR_PTR(ret); |
| 459 | } |
| 460 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
| 461 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
| 462 | blocksize); |
| 463 | BUG_ON(!root->node); |
| 464 | insert: |
| 465 | root->ref_cows = 1; |
| 466 | ret = btrfs_find_highest_inode(root, &highest_inode); |
| 467 | if (ret == 0) { |
| 468 | root->highest_inode = highest_inode; |
| 469 | root->last_inode_alloc = highest_inode; |
| 470 | } |
| 471 | return root; |
| 472 | } |
| 473 | |
| 474 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
| 475 | struct btrfs_key *location) |
| 476 | { |
| 477 | struct btrfs_root *root; |
| 478 | int ret; |
| 479 | |
| 480 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
| 481 | return fs_info->tree_root; |
| 482 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) |
| 483 | return fs_info->extent_root; |
| 484 | |
| 485 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
| 486 | (unsigned long)location->objectid); |
| 487 | if (root) |
| 488 | return root; |
| 489 | |
| 490 | root = btrfs_read_fs_root_no_radix(fs_info, location); |
| 491 | if (IS_ERR(root)) |
| 492 | return root; |
| 493 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
| 494 | (unsigned long)root->root_key.objectid, |
| 495 | root); |
| 496 | if (ret) { |
| 497 | free_extent_buffer(root->node); |
| 498 | kfree(root); |
| 499 | return ERR_PTR(ret); |
| 500 | } |
| 501 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
| 502 | root->root_key.objectid, root); |
| 503 | BUG_ON(ret); |
| 504 | |
| 505 | return root; |
| 506 | } |
| 507 | |
| 508 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, |
| 509 | struct btrfs_key *location, |
| 510 | const char *name, int namelen) |
| 511 | { |
| 512 | struct btrfs_root *root; |
| 513 | int ret; |
| 514 | |
| 515 | root = btrfs_read_fs_root_no_name(fs_info, location); |
| 516 | if (!root) |
| 517 | return NULL; |
| 518 | |
| 519 | ret = btrfs_set_root_name(root, name, namelen); |
| 520 | if (ret) { |
| 521 | free_extent_buffer(root->node); |
| 522 | kfree(root); |
| 523 | return ERR_PTR(ret); |
| 524 | } |
| 525 | |
| 526 | ret = btrfs_sysfs_add_root(root); |
| 527 | if (ret) { |
| 528 | free_extent_buffer(root->node); |
| 529 | kfree(root->name); |
| 530 | kfree(root); |
| 531 | return ERR_PTR(ret); |
| 532 | } |
| 533 | return root; |
| 534 | } |
| 535 | #if 0 |
| 536 | static int add_hasher(struct btrfs_fs_info *info, char *type) { |
| 537 | struct btrfs_hasher *hasher; |
| 538 | |
| 539 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); |
| 540 | if (!hasher) |
| 541 | return -ENOMEM; |
| 542 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); |
| 543 | if (!hasher->hash_tfm) { |
| 544 | kfree(hasher); |
| 545 | return -EINVAL; |
| 546 | } |
| 547 | spin_lock(&info->hash_lock); |
| 548 | list_add(&hasher->list, &info->hashers); |
| 549 | spin_unlock(&info->hash_lock); |
| 550 | return 0; |
| 551 | } |
| 552 | #endif |
| 553 | struct btrfs_root *open_ctree(struct super_block *sb) |
| 554 | { |
| 555 | u32 sectorsize; |
| 556 | u32 nodesize; |
| 557 | u32 leafsize; |
| 558 | u32 blocksize; |
| 559 | u32 stripesize; |
| 560 | struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), |
| 561 | GFP_NOFS); |
| 562 | struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), |
| 563 | GFP_NOFS); |
| 564 | struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info), |
| 565 | GFP_NOFS); |
| 566 | int ret; |
| 567 | int err = -EIO; |
| 568 | struct btrfs_super_block *disk_super; |
| 569 | |
| 570 | if (!extent_root || !tree_root || !fs_info) { |
| 571 | err = -ENOMEM; |
| 572 | goto fail; |
| 573 | } |
| 574 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
| 575 | INIT_LIST_HEAD(&fs_info->trans_list); |
| 576 | INIT_LIST_HEAD(&fs_info->dead_roots); |
| 577 | INIT_LIST_HEAD(&fs_info->hashers); |
| 578 | spin_lock_init(&fs_info->hash_lock); |
| 579 | spin_lock_init(&fs_info->delalloc_lock); |
| 580 | |
| 581 | memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj)); |
| 582 | init_completion(&fs_info->kobj_unregister); |
| 583 | sb_set_blocksize(sb, 4096); |
| 584 | fs_info->running_transaction = NULL; |
| 585 | fs_info->last_trans_committed = 0; |
| 586 | fs_info->tree_root = tree_root; |
| 587 | fs_info->extent_root = extent_root; |
| 588 | fs_info->sb = sb; |
| 589 | fs_info->mount_opt = 0; |
| 590 | fs_info->max_extent = (u64)-1; |
| 591 | fs_info->delalloc_bytes = 0; |
| 592 | fs_info->btree_inode = new_inode(sb); |
| 593 | fs_info->btree_inode->i_ino = 1; |
| 594 | fs_info->btree_inode->i_nlink = 1; |
| 595 | fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size; |
| 596 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
| 597 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
| 598 | fs_info->btree_inode->i_mapping, |
| 599 | GFP_NOFS); |
| 600 | BTRFS_I(fs_info->btree_inode)->extent_tree.ops = &btree_extent_map_ops; |
| 601 | |
| 602 | extent_map_tree_init(&fs_info->free_space_cache, |
| 603 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
| 604 | extent_map_tree_init(&fs_info->block_group_cache, |
| 605 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
| 606 | extent_map_tree_init(&fs_info->pinned_extents, |
| 607 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
| 608 | extent_map_tree_init(&fs_info->pending_del, |
| 609 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
| 610 | extent_map_tree_init(&fs_info->extent_ins, |
| 611 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
| 612 | fs_info->do_barriers = 1; |
| 613 | fs_info->closing = 0; |
| 614 | fs_info->total_pinned = 0; |
| 615 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
| 616 | INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); |
| 617 | #else |
| 618 | INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); |
| 619 | #endif |
| 620 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
| 621 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, |
| 622 | sizeof(struct btrfs_key)); |
| 623 | insert_inode_hash(fs_info->btree_inode); |
| 624 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
| 625 | |
| 626 | mutex_init(&fs_info->trans_mutex); |
| 627 | mutex_init(&fs_info->fs_mutex); |
| 628 | |
| 629 | #if 0 |
| 630 | ret = add_hasher(fs_info, "crc32c"); |
| 631 | if (ret) { |
| 632 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); |
| 633 | err = -ENOMEM; |
| 634 | goto fail_iput; |
| 635 | } |
| 636 | #endif |
| 637 | __setup_root(512, 512, 512, 512, tree_root, |
| 638 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
| 639 | |
| 640 | fs_info->sb_buffer = read_tree_block(tree_root, |
| 641 | BTRFS_SUPER_INFO_OFFSET, |
| 642 | 512); |
| 643 | |
| 644 | if (!fs_info->sb_buffer) |
| 645 | goto fail_iput; |
| 646 | |
| 647 | read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0, |
| 648 | sizeof(fs_info->super_copy)); |
| 649 | |
| 650 | read_extent_buffer(fs_info->sb_buffer, fs_info->fsid, |
| 651 | (unsigned long)btrfs_super_fsid(fs_info->sb_buffer), |
| 652 | BTRFS_FSID_SIZE); |
| 653 | disk_super = &fs_info->super_copy; |
| 654 | if (!btrfs_super_root(disk_super)) |
| 655 | goto fail_sb_buffer; |
| 656 | |
| 657 | nodesize = btrfs_super_nodesize(disk_super); |
| 658 | leafsize = btrfs_super_leafsize(disk_super); |
| 659 | sectorsize = btrfs_super_sectorsize(disk_super); |
| 660 | stripesize = btrfs_super_stripesize(disk_super); |
| 661 | tree_root->nodesize = nodesize; |
| 662 | tree_root->leafsize = leafsize; |
| 663 | tree_root->sectorsize = sectorsize; |
| 664 | tree_root->stripesize = stripesize; |
| 665 | sb_set_blocksize(sb, sectorsize); |
| 666 | |
| 667 | i_size_write(fs_info->btree_inode, |
| 668 | btrfs_super_total_bytes(disk_super)); |
| 669 | |
| 670 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
| 671 | sizeof(disk_super->magic))) { |
| 672 | printk("btrfs: valid FS not found on %s\n", sb->s_id); |
| 673 | goto fail_sb_buffer; |
| 674 | } |
| 675 | |
| 676 | blocksize = btrfs_level_size(tree_root, |
| 677 | btrfs_super_root_level(disk_super)); |
| 678 | |
| 679 | tree_root->node = read_tree_block(tree_root, |
| 680 | btrfs_super_root(disk_super), |
| 681 | blocksize); |
| 682 | if (!tree_root->node) |
| 683 | goto fail_sb_buffer; |
| 684 | |
| 685 | mutex_lock(&fs_info->fs_mutex); |
| 686 | |
| 687 | ret = find_and_setup_root(tree_root, fs_info, |
| 688 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
| 689 | if (ret) { |
| 690 | mutex_unlock(&fs_info->fs_mutex); |
| 691 | goto fail_tree_root; |
| 692 | } |
| 693 | |
| 694 | btrfs_read_block_groups(extent_root); |
| 695 | |
| 696 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
| 697 | mutex_unlock(&fs_info->fs_mutex); |
| 698 | return tree_root; |
| 699 | |
| 700 | fail_tree_root: |
| 701 | free_extent_buffer(tree_root->node); |
| 702 | fail_sb_buffer: |
| 703 | free_extent_buffer(fs_info->sb_buffer); |
| 704 | fail_iput: |
| 705 | iput(fs_info->btree_inode); |
| 706 | fail: |
| 707 | kfree(extent_root); |
| 708 | kfree(tree_root); |
| 709 | kfree(fs_info); |
| 710 | return ERR_PTR(err); |
| 711 | } |
| 712 | |
| 713 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
| 714 | *root) |
| 715 | { |
| 716 | int ret; |
| 717 | struct extent_buffer *super = root->fs_info->sb_buffer; |
| 718 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 719 | |
| 720 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super); |
| 721 | ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping, |
| 722 | super->start, super->len); |
| 723 | return ret; |
| 724 | } |
| 725 | |
| 726 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
| 727 | { |
| 728 | radix_tree_delete(&fs_info->fs_roots_radix, |
| 729 | (unsigned long)root->root_key.objectid); |
| 730 | btrfs_sysfs_del_root(root); |
| 731 | if (root->inode) |
| 732 | iput(root->inode); |
| 733 | if (root->node) |
| 734 | free_extent_buffer(root->node); |
| 735 | if (root->commit_root) |
| 736 | free_extent_buffer(root->commit_root); |
| 737 | if (root->name) |
| 738 | kfree(root->name); |
| 739 | kfree(root); |
| 740 | return 0; |
| 741 | } |
| 742 | |
| 743 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
| 744 | { |
| 745 | int ret; |
| 746 | struct btrfs_root *gang[8]; |
| 747 | int i; |
| 748 | |
| 749 | while(1) { |
| 750 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
| 751 | (void **)gang, 0, |
| 752 | ARRAY_SIZE(gang)); |
| 753 | if (!ret) |
| 754 | break; |
| 755 | for (i = 0; i < ret; i++) |
| 756 | btrfs_free_fs_root(fs_info, gang[i]); |
| 757 | } |
| 758 | return 0; |
| 759 | } |
| 760 | |
| 761 | int close_ctree(struct btrfs_root *root) |
| 762 | { |
| 763 | int ret; |
| 764 | struct btrfs_trans_handle *trans; |
| 765 | struct btrfs_fs_info *fs_info = root->fs_info; |
| 766 | |
| 767 | fs_info->closing = 1; |
| 768 | btrfs_transaction_flush_work(root); |
| 769 | mutex_lock(&fs_info->fs_mutex); |
| 770 | btrfs_defrag_dirty_roots(root->fs_info); |
| 771 | trans = btrfs_start_transaction(root, 1); |
| 772 | ret = btrfs_commit_transaction(trans, root); |
| 773 | /* run commit again to drop the original snapshot */ |
| 774 | trans = btrfs_start_transaction(root, 1); |
| 775 | btrfs_commit_transaction(trans, root); |
| 776 | ret = btrfs_write_and_wait_transaction(NULL, root); |
| 777 | BUG_ON(ret); |
| 778 | write_ctree_super(NULL, root); |
| 779 | mutex_unlock(&fs_info->fs_mutex); |
| 780 | |
| 781 | if (fs_info->extent_root->node) |
| 782 | free_extent_buffer(fs_info->extent_root->node); |
| 783 | |
| 784 | if (fs_info->tree_root->node) |
| 785 | free_extent_buffer(fs_info->tree_root->node); |
| 786 | |
| 787 | free_extent_buffer(fs_info->sb_buffer); |
| 788 | |
| 789 | btrfs_free_block_groups(root->fs_info); |
| 790 | del_fs_roots(fs_info); |
| 791 | |
| 792 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); |
| 793 | |
| 794 | extent_map_tree_empty_lru(&fs_info->free_space_cache); |
| 795 | extent_map_tree_empty_lru(&fs_info->block_group_cache); |
| 796 | extent_map_tree_empty_lru(&fs_info->pinned_extents); |
| 797 | extent_map_tree_empty_lru(&fs_info->pending_del); |
| 798 | extent_map_tree_empty_lru(&fs_info->extent_ins); |
| 799 | extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
| 800 | |
| 801 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
| 802 | |
| 803 | iput(fs_info->btree_inode); |
| 804 | #if 0 |
| 805 | while(!list_empty(&fs_info->hashers)) { |
| 806 | struct btrfs_hasher *hasher; |
| 807 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, |
| 808 | hashers); |
| 809 | list_del(&hasher->hashers); |
| 810 | crypto_free_hash(&fs_info->hash_tfm); |
| 811 | kfree(hasher); |
| 812 | } |
| 813 | #endif |
| 814 | kfree(fs_info->extent_root); |
| 815 | kfree(fs_info->tree_root); |
| 816 | return 0; |
| 817 | } |
| 818 | |
| 819 | int btrfs_buffer_uptodate(struct extent_buffer *buf) |
| 820 | { |
| 821 | struct inode *btree_inode = buf->first_page->mapping->host; |
| 822 | return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf); |
| 823 | } |
| 824 | |
| 825 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) |
| 826 | { |
| 827 | struct inode *btree_inode = buf->first_page->mapping->host; |
| 828 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, |
| 829 | buf); |
| 830 | } |
| 831 | |
| 832 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
| 833 | { |
| 834 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 835 | u64 transid = btrfs_header_generation(buf); |
| 836 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 837 | |
| 838 | if (transid != root->fs_info->generation) { |
| 839 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", |
| 840 | (unsigned long long)buf->start, |
| 841 | transid, root->fs_info->generation); |
| 842 | WARN_ON(1); |
| 843 | } |
| 844 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); |
| 845 | } |
| 846 | |
| 847 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
| 848 | { |
| 849 | balance_dirty_pages_ratelimited_nr( |
| 850 | root->fs_info->btree_inode->i_mapping, 1); |
| 851 | } |
| 852 | |
| 853 | void btrfs_set_buffer_defrag(struct extent_buffer *buf) |
| 854 | { |
| 855 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 856 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 857 | set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start, |
| 858 | buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS); |
| 859 | } |
| 860 | |
| 861 | void btrfs_set_buffer_defrag_done(struct extent_buffer *buf) |
| 862 | { |
| 863 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 864 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 865 | set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start, |
| 866 | buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, |
| 867 | GFP_NOFS); |
| 868 | } |
| 869 | |
| 870 | int btrfs_buffer_defrag(struct extent_buffer *buf) |
| 871 | { |
| 872 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 873 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 874 | return test_range_bit(&BTRFS_I(btree_inode)->extent_tree, |
| 875 | buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0); |
| 876 | } |
| 877 | |
| 878 | int btrfs_buffer_defrag_done(struct extent_buffer *buf) |
| 879 | { |
| 880 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 881 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 882 | return test_range_bit(&BTRFS_I(btree_inode)->extent_tree, |
| 883 | buf->start, buf->start + buf->len - 1, |
| 884 | EXTENT_DEFRAG_DONE, 0); |
| 885 | } |
| 886 | |
| 887 | int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf) |
| 888 | { |
| 889 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 890 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 891 | return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree, |
| 892 | buf->start, buf->start + buf->len - 1, |
| 893 | EXTENT_DEFRAG_DONE, GFP_NOFS); |
| 894 | } |
| 895 | |
| 896 | int btrfs_clear_buffer_defrag(struct extent_buffer *buf) |
| 897 | { |
| 898 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 899 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 900 | return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree, |
| 901 | buf->start, buf->start + buf->len - 1, |
| 902 | EXTENT_DEFRAG, GFP_NOFS); |
| 903 | } |
| 904 | |
| 905 | int btrfs_read_buffer(struct extent_buffer *buf) |
| 906 | { |
| 907 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
| 908 | struct inode *btree_inode = root->fs_info->btree_inode; |
| 909 | return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, |
| 910 | buf, 0, 1); |
| 911 | } |
| 912 | |
| 913 | static struct extent_map_ops btree_extent_map_ops = { |
| 914 | .writepage_io_hook = btree_writepage_io_hook, |
| 915 | }; |