16cbd5570SChris Mason /* 26cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 36cbd5570SChris Mason * 46cbd5570SChris Mason * This program is free software; you can redistribute it and/or 56cbd5570SChris Mason * modify it under the terms of the GNU General Public 66cbd5570SChris Mason * License v2 as published by the Free Software Foundation. 76cbd5570SChris Mason * 86cbd5570SChris Mason * This program is distributed in the hope that it will be useful, 96cbd5570SChris Mason * but WITHOUT ANY WARRANTY; without even the implied warranty of 106cbd5570SChris Mason * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 116cbd5570SChris Mason * General Public License for more details. 126cbd5570SChris Mason * 136cbd5570SChris Mason * You should have received a copy of the GNU General Public 146cbd5570SChris Mason * License along with this program; if not, write to the 156cbd5570SChris Mason * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 166cbd5570SChris Mason * Boston, MA 021110-1307, USA. 176cbd5570SChris Mason */ 186cbd5570SChris Mason 19e20d96d6SChris Mason #include <linux/fs.h> 20d98237b3SChris Mason #include <linux/blkdev.h> 2187cbda5cSChris Mason #include <linux/scatterlist.h> 2222b0ebdaSChris Mason #include <linux/swap.h> 230f7d52f4SChris Mason #include <linux/radix-tree.h> 2435b7e476SChris Mason #include <linux/writeback.h> 25d397712bSChris Mason #include <linux/buffer_head.h> 26ce9adaa5SChris Mason #include <linux/workqueue.h> 27a74a4b97SChris Mason #include <linux/kthread.h> 28a74a4b97SChris Mason #include <linux/freezer.h> 29163e783eSDavid Woodhouse #include <linux/crc32c.h> 305a0e3ad6STejun Heo #include <linux/slab.h> 31784b4e29SChris Mason #include <linux/migrate.h> 327a36ddecSDavid Sterba #include <linux/ratelimit.h> 337e75bf3fSDavid Sterba #include <asm/unaligned.h> 344b4e25f2SChris Mason #include "compat.h" 35eb60ceacSChris Mason #include "ctree.h" 36eb60ceacSChris Mason #include "disk-io.h" 37e089f05cSChris Mason #include "transaction.h" 380f7d52f4SChris Mason #include "btrfs_inode.h" 390b86a832SChris Mason #include "volumes.h" 40db94535dSChris Mason #include "print-tree.h" 418b712842SChris Mason #include "async-thread.h" 42925baeddSChris Mason #include "locking.h" 43e02119d5SChris Mason #include "tree-log.h" 44fa9c0d79SChris Mason #include "free-space-cache.h" 45581bb050SLi Zefan #include "inode-map.h" 4621adbd5cSStefan Behrens #include "check-integrity.h" 47606686eeSJosef Bacik #include "rcu-string.h" 48eb60ceacSChris Mason 49de0022b9SJosef Bacik #ifdef CONFIG_X86 50de0022b9SJosef Bacik #include <asm/cpufeature.h> 51de0022b9SJosef Bacik #endif 52de0022b9SJosef Bacik 53d1310b2eSChris Mason static struct extent_io_ops btree_extent_io_ops; 548b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 554df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root); 56fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 57acce952bSliubo int read_only); 58143bede5SJeff Mahoney static void btrfs_destroy_ordered_operations(struct btrfs_root *root); 59143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 60acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 61acce952bSliubo struct btrfs_root *root); 62143bede5SJeff Mahoney static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); 63143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 64acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 65acce952bSliubo struct extent_io_tree *dirty_pages, 66acce952bSliubo int mark); 67acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 68acce952bSliubo struct extent_io_tree *pinned_extents); 69ce9adaa5SChris Mason 70d352ac68SChris Mason /* 71d352ac68SChris Mason * end_io_wq structs are used to do processing in task context when an IO is 72d352ac68SChris Mason * complete. This is used during reads to verify checksums, and it is used 73d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 74d352ac68SChris Mason */ 75ce9adaa5SChris Mason struct end_io_wq { 76ce9adaa5SChris Mason struct bio *bio; 77ce9adaa5SChris Mason bio_end_io_t *end_io; 78ce9adaa5SChris Mason void *private; 79ce9adaa5SChris Mason struct btrfs_fs_info *info; 80ce9adaa5SChris Mason int error; 8122c59948SChris Mason int metadata; 82ce9adaa5SChris Mason struct list_head list; 838b712842SChris Mason struct btrfs_work work; 84ce9adaa5SChris Mason }; 850da5468fSChris Mason 86d352ac68SChris Mason /* 87d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 88d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 89d352ac68SChris Mason * just before they are sent down the IO stack. 90d352ac68SChris Mason */ 9144b8bd7eSChris Mason struct async_submit_bio { 9244b8bd7eSChris Mason struct inode *inode; 9344b8bd7eSChris Mason struct bio *bio; 9444b8bd7eSChris Mason struct list_head list; 954a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start; 964a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done; 9744b8bd7eSChris Mason int rw; 9844b8bd7eSChris Mason int mirror_num; 99c8b97818SChris Mason unsigned long bio_flags; 100eaf25d93SChris Mason /* 101eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 102eaf25d93SChris Mason * can't tell us where in the file the bio should go 103eaf25d93SChris Mason */ 104eaf25d93SChris Mason u64 bio_offset; 1058b712842SChris Mason struct btrfs_work work; 10679787eaaSJeff Mahoney int error; 10744b8bd7eSChris Mason }; 10844b8bd7eSChris Mason 10985d4e461SChris Mason /* 11085d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 11185d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 11285d4e461SChris Mason * the level the eb occupies in the tree. 1134008c04aSChris Mason * 11485d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 11585d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 11685d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 11785d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 11885d4e461SChris Mason * have separate keysets. 1194008c04aSChris Mason * 12085d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 12185d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 12285d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1234008c04aSChris Mason * 12485d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 12585d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 12685d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 12785d4e461SChris Mason * 12885d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 12985d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 13085d4e461SChris Mason * needs update as well. 1314008c04aSChris Mason */ 1324008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1334008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1344008c04aSChris Mason # error 1354008c04aSChris Mason # endif 13685d4e461SChris Mason 13785d4e461SChris Mason static struct btrfs_lockdep_keyset { 13885d4e461SChris Mason u64 id; /* root objectid */ 13985d4e461SChris Mason const char *name_stem; /* lock name stem */ 14085d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 14185d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 14285d4e461SChris Mason } btrfs_lockdep_keysets[] = { 14385d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 14485d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 14585d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 14685d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 14785d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 14885d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 14985d4e461SChris Mason { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, 15085d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 15185d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 15285d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 15385d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1544008c04aSChris Mason }; 15585d4e461SChris Mason 15685d4e461SChris Mason void __init btrfs_init_lockdep(void) 15785d4e461SChris Mason { 15885d4e461SChris Mason int i, j; 15985d4e461SChris Mason 16085d4e461SChris Mason /* initialize lockdep class names */ 16185d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 16285d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 16385d4e461SChris Mason 16485d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 16585d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 16685d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 16785d4e461SChris Mason } 16885d4e461SChris Mason } 16985d4e461SChris Mason 17085d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 17185d4e461SChris Mason int level) 17285d4e461SChris Mason { 17385d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 17485d4e461SChris Mason 17585d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 17685d4e461SChris Mason 17785d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 17885d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 17985d4e461SChris Mason if (ks->id == objectid) 18085d4e461SChris Mason break; 18185d4e461SChris Mason 18285d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 18385d4e461SChris Mason &ks->keys[level], ks->names[level]); 18485d4e461SChris Mason } 18585d4e461SChris Mason 1864008c04aSChris Mason #endif 1874008c04aSChris Mason 188d352ac68SChris Mason /* 189d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 190d352ac68SChris Mason * that covers the entire device 191d352ac68SChris Mason */ 192b2950863SChristoph Hellwig static struct extent_map *btree_get_extent(struct inode *inode, 193306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 1945f39d397SChris Mason int create) 1955f39d397SChris Mason { 1965f39d397SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1975f39d397SChris Mason struct extent_map *em; 1985f39d397SChris Mason int ret; 1995f39d397SChris Mason 200890871beSChris Mason read_lock(&em_tree->lock); 201d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 202a061fc8dSChris Mason if (em) { 203a061fc8dSChris Mason em->bdev = 204a061fc8dSChris Mason BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 205890871beSChris Mason read_unlock(&em_tree->lock); 2065f39d397SChris Mason goto out; 207a061fc8dSChris Mason } 208890871beSChris Mason read_unlock(&em_tree->lock); 2097b13b7b1SChris Mason 210172ddd60SDavid Sterba em = alloc_extent_map(); 2115f39d397SChris Mason if (!em) { 2125f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2135f39d397SChris Mason goto out; 2145f39d397SChris Mason } 2155f39d397SChris Mason em->start = 0; 2160afbaf8cSChris Mason em->len = (u64)-1; 217c8b97818SChris Mason em->block_len = (u64)-1; 2185f39d397SChris Mason em->block_start = 0; 219a061fc8dSChris Mason em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 220d1310b2eSChris Mason 221890871beSChris Mason write_lock(&em_tree->lock); 2225f39d397SChris Mason ret = add_extent_mapping(em_tree, em); 2235f39d397SChris Mason if (ret == -EEXIST) { 2245f39d397SChris Mason free_extent_map(em); 2257b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 226b4f359abSTsutomu Itoh if (!em) 2270433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2285f39d397SChris Mason } else if (ret) { 2297b13b7b1SChris Mason free_extent_map(em); 2300433f20dSTsutomu Itoh em = ERR_PTR(ret); 2315f39d397SChris Mason } 232890871beSChris Mason write_unlock(&em_tree->lock); 2337b13b7b1SChris Mason 2345f39d397SChris Mason out: 2355f39d397SChris Mason return em; 2365f39d397SChris Mason } 2375f39d397SChris Mason 23819c00ddcSChris Mason u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) 23919c00ddcSChris Mason { 240163e783eSDavid Woodhouse return crc32c(seed, data, len); 24119c00ddcSChris Mason } 24219c00ddcSChris Mason 24319c00ddcSChris Mason void btrfs_csum_final(u32 crc, char *result) 24419c00ddcSChris Mason { 2457e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 24619c00ddcSChris Mason } 24719c00ddcSChris Mason 248d352ac68SChris Mason /* 249d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 250d352ac68SChris Mason * into the csum field of the block. 251d352ac68SChris Mason */ 25219c00ddcSChris Mason static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, 25319c00ddcSChris Mason int verify) 25419c00ddcSChris Mason { 2556c41761fSDavid Sterba u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); 256607d432dSJosef Bacik char *result = NULL; 25719c00ddcSChris Mason unsigned long len; 25819c00ddcSChris Mason unsigned long cur_len; 25919c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 26019c00ddcSChris Mason char *kaddr; 26119c00ddcSChris Mason unsigned long map_start; 26219c00ddcSChris Mason unsigned long map_len; 26319c00ddcSChris Mason int err; 26419c00ddcSChris Mason u32 crc = ~(u32)0; 265607d432dSJosef Bacik unsigned long inline_result; 26619c00ddcSChris Mason 26719c00ddcSChris Mason len = buf->len - offset; 26819c00ddcSChris Mason while (len > 0) { 26919c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 270a6591715SChris Mason &kaddr, &map_start, &map_len); 271d397712bSChris Mason if (err) 27219c00ddcSChris Mason return 1; 27319c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 27419c00ddcSChris Mason crc = btrfs_csum_data(root, kaddr + offset - map_start, 27519c00ddcSChris Mason crc, cur_len); 27619c00ddcSChris Mason len -= cur_len; 27719c00ddcSChris Mason offset += cur_len; 27819c00ddcSChris Mason } 279607d432dSJosef Bacik if (csum_size > sizeof(inline_result)) { 280607d432dSJosef Bacik result = kzalloc(csum_size * sizeof(char), GFP_NOFS); 281607d432dSJosef Bacik if (!result) 282607d432dSJosef Bacik return 1; 283607d432dSJosef Bacik } else { 284607d432dSJosef Bacik result = (char *)&inline_result; 285607d432dSJosef Bacik } 286607d432dSJosef Bacik 28719c00ddcSChris Mason btrfs_csum_final(crc, result); 28819c00ddcSChris Mason 28919c00ddcSChris Mason if (verify) { 290607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 291e4204dedSChris Mason u32 val; 292e4204dedSChris Mason u32 found = 0; 293607d432dSJosef Bacik memcpy(&found, result, csum_size); 294e4204dedSChris Mason 295607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 2967a36ddecSDavid Sterba printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " 297193f284dSChris Mason "failed on %llu wanted %X found %X " 298193f284dSChris Mason "level %d\n", 29919c00ddcSChris Mason root->fs_info->sb->s_id, 300193f284dSChris Mason (unsigned long long)buf->start, val, found, 301193f284dSChris Mason btrfs_header_level(buf)); 302607d432dSJosef Bacik if (result != (char *)&inline_result) 303607d432dSJosef Bacik kfree(result); 30419c00ddcSChris Mason return 1; 30519c00ddcSChris Mason } 30619c00ddcSChris Mason } else { 307607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 30819c00ddcSChris Mason } 309607d432dSJosef Bacik if (result != (char *)&inline_result) 310607d432dSJosef Bacik kfree(result); 31119c00ddcSChris Mason return 0; 31219c00ddcSChris Mason } 31319c00ddcSChris Mason 314d352ac68SChris Mason /* 315d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 316d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 317d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 318d352ac68SChris Mason * in the wrong place. 319d352ac68SChris Mason */ 3201259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 321b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 322b9fab919SChris Mason int atomic) 3231259ab75SChris Mason { 3242ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3251259ab75SChris Mason int ret; 3261259ab75SChris Mason 3271259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3281259ab75SChris Mason return 0; 3291259ab75SChris Mason 330b9fab919SChris Mason if (atomic) 331b9fab919SChris Mason return -EAGAIN; 332b9fab919SChris Mason 3332ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 334d0082371SJeff Mahoney 0, &cached_state); 3350b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3361259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3371259ab75SChris Mason ret = 0; 3381259ab75SChris Mason goto out; 3391259ab75SChris Mason } 3407a36ddecSDavid Sterba printk_ratelimited("parent transid verify failed on %llu wanted %llu " 341193f284dSChris Mason "found %llu\n", 3421259ab75SChris Mason (unsigned long long)eb->start, 3431259ab75SChris Mason (unsigned long long)parent_transid, 3441259ab75SChris Mason (unsigned long long)btrfs_header_generation(eb)); 3451259ab75SChris Mason ret = 1; 3460b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 34733958dc6SChris Mason out: 3482ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 3492ac55d41SJosef Bacik &cached_state, GFP_NOFS); 3501259ab75SChris Mason return ret; 3511259ab75SChris Mason } 3521259ab75SChris Mason 353d352ac68SChris Mason /* 354d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 355d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 356d352ac68SChris Mason */ 357f188591eSChris Mason static int btree_read_extent_buffer_pages(struct btrfs_root *root, 358f188591eSChris Mason struct extent_buffer *eb, 359ca7a79adSChris Mason u64 start, u64 parent_transid) 360f188591eSChris Mason { 361f188591eSChris Mason struct extent_io_tree *io_tree; 362ea466794SJosef Bacik int failed = 0; 363f188591eSChris Mason int ret; 364f188591eSChris Mason int num_copies = 0; 365f188591eSChris Mason int mirror_num = 0; 366ea466794SJosef Bacik int failed_mirror = 0; 367f188591eSChris Mason 368a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 369f188591eSChris Mason io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 370f188591eSChris Mason while (1) { 371bb82ab88SArne Jansen ret = read_extent_buffer_pages(io_tree, eb, start, 372bb82ab88SArne Jansen WAIT_COMPLETE, 373f188591eSChris Mason btree_get_extent, mirror_num); 374256dd1bbSStefan Behrens if (!ret) { 375256dd1bbSStefan Behrens if (!verify_parent_transid(io_tree, eb, 376b9fab919SChris Mason parent_transid, 0)) 377ea466794SJosef Bacik break; 378256dd1bbSStefan Behrens else 379256dd1bbSStefan Behrens ret = -EIO; 380256dd1bbSStefan Behrens } 381d397712bSChris Mason 382a826d6dcSJosef Bacik /* 383a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 384a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 385a826d6dcSJosef Bacik * any less wrong. 386a826d6dcSJosef Bacik */ 387a826d6dcSJosef Bacik if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 388ea466794SJosef Bacik break; 389ea466794SJosef Bacik 3905d964051SStefan Behrens num_copies = btrfs_num_copies(root->fs_info, 391f188591eSChris Mason eb->start, eb->len); 3924235298eSChris Mason if (num_copies == 1) 393ea466794SJosef Bacik break; 3944235298eSChris Mason 3955cf1ab56SJosef Bacik if (!failed_mirror) { 3965cf1ab56SJosef Bacik failed = 1; 3975cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 3985cf1ab56SJosef Bacik } 3995cf1ab56SJosef Bacik 400f188591eSChris Mason mirror_num++; 401ea466794SJosef Bacik if (mirror_num == failed_mirror) 402ea466794SJosef Bacik mirror_num++; 403ea466794SJosef Bacik 4044235298eSChris Mason if (mirror_num > num_copies) 405ea466794SJosef Bacik break; 406f188591eSChris Mason } 407ea466794SJosef Bacik 408c0901581SStefan Behrens if (failed && !ret && failed_mirror) 409ea466794SJosef Bacik repair_eb_io_failure(root, eb, failed_mirror); 410ea466794SJosef Bacik 411ea466794SJosef Bacik return ret; 412f188591eSChris Mason } 41319c00ddcSChris Mason 414d352ac68SChris Mason /* 415d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 416d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 417d352ac68SChris Mason */ 418d397712bSChris Mason 419b2950863SChristoph Hellwig static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) 42019c00ddcSChris Mason { 421d1310b2eSChris Mason struct extent_io_tree *tree; 42235ebb934SChris Mason u64 start = (u64)page->index << PAGE_CACHE_SHIFT; 42319c00ddcSChris Mason u64 found_start; 42419c00ddcSChris Mason struct extent_buffer *eb; 425f188591eSChris Mason 426d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 42719c00ddcSChris Mason 4284f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4294f2de97aSJosef Bacik if (page != eb->pages[0]) 4304f2de97aSJosef Bacik return 0; 43119c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 43219c00ddcSChris Mason if (found_start != start) { 43355c69072SChris Mason WARN_ON(1); 4344f2de97aSJosef Bacik return 0; 43555c69072SChris Mason } 43655c69072SChris Mason if (!PageUptodate(page)) { 43755c69072SChris Mason WARN_ON(1); 4384f2de97aSJosef Bacik return 0; 43919c00ddcSChris Mason } 44019c00ddcSChris Mason csum_tree_block(root, eb, 0); 44119c00ddcSChris Mason return 0; 44219c00ddcSChris Mason } 44319c00ddcSChris Mason 4442b82032cSYan Zheng static int check_tree_block_fsid(struct btrfs_root *root, 4452b82032cSYan Zheng struct extent_buffer *eb) 4462b82032cSYan Zheng { 4472b82032cSYan Zheng struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 4482b82032cSYan Zheng u8 fsid[BTRFS_UUID_SIZE]; 4492b82032cSYan Zheng int ret = 1; 4502b82032cSYan Zheng 4512b82032cSYan Zheng read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), 4522b82032cSYan Zheng BTRFS_FSID_SIZE); 4532b82032cSYan Zheng while (fs_devices) { 4542b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 4552b82032cSYan Zheng ret = 0; 4562b82032cSYan Zheng break; 4572b82032cSYan Zheng } 4582b82032cSYan Zheng fs_devices = fs_devices->seed; 4592b82032cSYan Zheng } 4602b82032cSYan Zheng return ret; 4612b82032cSYan Zheng } 4622b82032cSYan Zheng 463a826d6dcSJosef Bacik #define CORRUPT(reason, eb, root, slot) \ 464a826d6dcSJosef Bacik printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ 465a826d6dcSJosef Bacik "root=%llu, slot=%d\n", reason, \ 466a826d6dcSJosef Bacik (unsigned long long)btrfs_header_bytenr(eb), \ 467a826d6dcSJosef Bacik (unsigned long long)root->objectid, slot) 468a826d6dcSJosef Bacik 469a826d6dcSJosef Bacik static noinline int check_leaf(struct btrfs_root *root, 470a826d6dcSJosef Bacik struct extent_buffer *leaf) 471a826d6dcSJosef Bacik { 472a826d6dcSJosef Bacik struct btrfs_key key; 473a826d6dcSJosef Bacik struct btrfs_key leaf_key; 474a826d6dcSJosef Bacik u32 nritems = btrfs_header_nritems(leaf); 475a826d6dcSJosef Bacik int slot; 476a826d6dcSJosef Bacik 477a826d6dcSJosef Bacik if (nritems == 0) 478a826d6dcSJosef Bacik return 0; 479a826d6dcSJosef Bacik 480a826d6dcSJosef Bacik /* Check the 0 item */ 481a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 482a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 483a826d6dcSJosef Bacik CORRUPT("invalid item offset size pair", leaf, root, 0); 484a826d6dcSJosef Bacik return -EIO; 485a826d6dcSJosef Bacik } 486a826d6dcSJosef Bacik 487a826d6dcSJosef Bacik /* 488a826d6dcSJosef Bacik * Check to make sure each items keys are in the correct order and their 489a826d6dcSJosef Bacik * offsets make sense. We only have to loop through nritems-1 because 490a826d6dcSJosef Bacik * we check the current slot against the next slot, which verifies the 491a826d6dcSJosef Bacik * next slot's offset+size makes sense and that the current's slot 492a826d6dcSJosef Bacik * offset is correct. 493a826d6dcSJosef Bacik */ 494a826d6dcSJosef Bacik for (slot = 0; slot < nritems - 1; slot++) { 495a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &leaf_key, slot); 496a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &key, slot + 1); 497a826d6dcSJosef Bacik 498a826d6dcSJosef Bacik /* Make sure the keys are in the right order */ 499a826d6dcSJosef Bacik if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { 500a826d6dcSJosef Bacik CORRUPT("bad key order", leaf, root, slot); 501a826d6dcSJosef Bacik return -EIO; 502a826d6dcSJosef Bacik } 503a826d6dcSJosef Bacik 504a826d6dcSJosef Bacik /* 505a826d6dcSJosef Bacik * Make sure the offset and ends are right, remember that the 506a826d6dcSJosef Bacik * item data starts at the end of the leaf and grows towards the 507a826d6dcSJosef Bacik * front. 508a826d6dcSJosef Bacik */ 509a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, slot) != 510a826d6dcSJosef Bacik btrfs_item_end_nr(leaf, slot + 1)) { 511a826d6dcSJosef Bacik CORRUPT("slot offset bad", leaf, root, slot); 512a826d6dcSJosef Bacik return -EIO; 513a826d6dcSJosef Bacik } 514a826d6dcSJosef Bacik 515a826d6dcSJosef Bacik /* 516a826d6dcSJosef Bacik * Check to make sure that we don't point outside of the leaf, 517a826d6dcSJosef Bacik * just incase all the items are consistent to eachother, but 518a826d6dcSJosef Bacik * all point outside of the leaf. 519a826d6dcSJosef Bacik */ 520a826d6dcSJosef Bacik if (btrfs_item_end_nr(leaf, slot) > 521a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 522a826d6dcSJosef Bacik CORRUPT("slot end outside of leaf", leaf, root, slot); 523a826d6dcSJosef Bacik return -EIO; 524a826d6dcSJosef Bacik } 525a826d6dcSJosef Bacik } 526a826d6dcSJosef Bacik 527a826d6dcSJosef Bacik return 0; 528a826d6dcSJosef Bacik } 529a826d6dcSJosef Bacik 530727011e0SChris Mason struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, 531727011e0SChris Mason struct page *page, int max_walk) 532727011e0SChris Mason { 533727011e0SChris Mason struct extent_buffer *eb; 534727011e0SChris Mason u64 start = page_offset(page); 535727011e0SChris Mason u64 target = start; 536727011e0SChris Mason u64 min_start; 537727011e0SChris Mason 538727011e0SChris Mason if (start < max_walk) 539727011e0SChris Mason min_start = 0; 540727011e0SChris Mason else 541727011e0SChris Mason min_start = start - max_walk; 542727011e0SChris Mason 543727011e0SChris Mason while (start >= min_start) { 544727011e0SChris Mason eb = find_extent_buffer(tree, start, 0); 545727011e0SChris Mason if (eb) { 546727011e0SChris Mason /* 547727011e0SChris Mason * we found an extent buffer and it contains our page 548727011e0SChris Mason * horray! 549727011e0SChris Mason */ 550727011e0SChris Mason if (eb->start <= target && 551727011e0SChris Mason eb->start + eb->len > target) 552727011e0SChris Mason return eb; 553727011e0SChris Mason 554727011e0SChris Mason /* we found an extent buffer that wasn't for us */ 555727011e0SChris Mason free_extent_buffer(eb); 556727011e0SChris Mason return NULL; 557727011e0SChris Mason } 558727011e0SChris Mason if (start == 0) 559727011e0SChris Mason break; 560727011e0SChris Mason start -= PAGE_CACHE_SIZE; 561727011e0SChris Mason } 562727011e0SChris Mason return NULL; 563727011e0SChris Mason } 564727011e0SChris Mason 565b2950863SChristoph Hellwig static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, 5665cf1ab56SJosef Bacik struct extent_state *state, int mirror) 567ce9adaa5SChris Mason { 568ce9adaa5SChris Mason struct extent_io_tree *tree; 569ce9adaa5SChris Mason u64 found_start; 570ce9adaa5SChris Mason int found_level; 571ce9adaa5SChris Mason struct extent_buffer *eb; 572ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 573f188591eSChris Mason int ret = 0; 574727011e0SChris Mason int reads_done; 575ce9adaa5SChris Mason 576ce9adaa5SChris Mason if (!page->private) 577ce9adaa5SChris Mason goto out; 578d397712bSChris Mason 579727011e0SChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 5804f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 581d397712bSChris Mason 5820b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 5830b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 5840b32f4bbSJosef Bacik */ 5850b32f4bbSJosef Bacik extent_buffer_get(eb); 5860b32f4bbSJosef Bacik 5870b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 588727011e0SChris Mason if (!reads_done) 589727011e0SChris Mason goto err; 590f188591eSChris Mason 5915cf1ab56SJosef Bacik eb->read_mirror = mirror; 592ea466794SJosef Bacik if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { 593ea466794SJosef Bacik ret = -EIO; 594ea466794SJosef Bacik goto err; 595ea466794SJosef Bacik } 596ea466794SJosef Bacik 597ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 598727011e0SChris Mason if (found_start != eb->start) { 5997a36ddecSDavid Sterba printk_ratelimited(KERN_INFO "btrfs bad tree block start " 600193f284dSChris Mason "%llu %llu\n", 601a1b32a59SChris Mason (unsigned long long)found_start, 602a1b32a59SChris Mason (unsigned long long)eb->start); 603f188591eSChris Mason ret = -EIO; 604ce9adaa5SChris Mason goto err; 605ce9adaa5SChris Mason } 6062b82032cSYan Zheng if (check_tree_block_fsid(root, eb)) { 6077a36ddecSDavid Sterba printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", 608d397712bSChris Mason (unsigned long long)eb->start); 6091259ab75SChris Mason ret = -EIO; 6101259ab75SChris Mason goto err; 6111259ab75SChris Mason } 612ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 613ce9adaa5SChris Mason 61485d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 61585d4e461SChris Mason eb, found_level); 6164008c04aSChris Mason 617ce9adaa5SChris Mason ret = csum_tree_block(root, eb, 1); 618a826d6dcSJosef Bacik if (ret) { 619f188591eSChris Mason ret = -EIO; 620a826d6dcSJosef Bacik goto err; 621a826d6dcSJosef Bacik } 622a826d6dcSJosef Bacik 623a826d6dcSJosef Bacik /* 624a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 625a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 626a826d6dcSJosef Bacik * return -EIO. 627a826d6dcSJosef Bacik */ 628a826d6dcSJosef Bacik if (found_level == 0 && check_leaf(root, eb)) { 629a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 630a826d6dcSJosef Bacik ret = -EIO; 631a826d6dcSJosef Bacik } 632ce9adaa5SChris Mason 6330b32f4bbSJosef Bacik if (!ret) 6340b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 635ce9adaa5SChris Mason err: 6364bb31e92SArne Jansen if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { 6374bb31e92SArne Jansen clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); 6384bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, ret); 6394bb31e92SArne Jansen } 6404bb31e92SArne Jansen 6410b32f4bbSJosef Bacik if (ret) 6420b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 6430b32f4bbSJosef Bacik free_extent_buffer(eb); 644ce9adaa5SChris Mason out: 645f188591eSChris Mason return ret; 646ce9adaa5SChris Mason } 647ce9adaa5SChris Mason 648ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6494bb31e92SArne Jansen { 6504bb31e92SArne Jansen struct extent_buffer *eb; 6514bb31e92SArne Jansen struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 6524bb31e92SArne Jansen 6534f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 654ea466794SJosef Bacik set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); 6555cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 656ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 6574bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, -EIO); 6584bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6594bb31e92SArne Jansen } 6604bb31e92SArne Jansen 661ce9adaa5SChris Mason static void end_workqueue_bio(struct bio *bio, int err) 662ce9adaa5SChris Mason { 663ce9adaa5SChris Mason struct end_io_wq *end_io_wq = bio->bi_private; 664ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 665ce9adaa5SChris Mason 666ce9adaa5SChris Mason fs_info = end_io_wq->info; 667ce9adaa5SChris Mason end_io_wq->error = err; 6688b712842SChris Mason end_io_wq->work.func = end_workqueue_fn; 6698b712842SChris Mason end_io_wq->work.flags = 0; 670d20f7043SChris Mason 6717b6d91daSChristoph Hellwig if (bio->bi_rw & REQ_WRITE) { 6720cb59c99SJosef Bacik if (end_io_wq->metadata == 1) 673cad321adSChris Mason btrfs_queue_worker(&fs_info->endio_meta_write_workers, 674cad321adSChris Mason &end_io_wq->work); 6750cb59c99SJosef Bacik else if (end_io_wq->metadata == 2) 6760cb59c99SJosef Bacik btrfs_queue_worker(&fs_info->endio_freespace_worker, 6770cb59c99SJosef Bacik &end_io_wq->work); 678cad321adSChris Mason else 679e6dcd2dcSChris Mason btrfs_queue_worker(&fs_info->endio_write_workers, 680e6dcd2dcSChris Mason &end_io_wq->work); 681d20f7043SChris Mason } else { 682d20f7043SChris Mason if (end_io_wq->metadata) 683d20f7043SChris Mason btrfs_queue_worker(&fs_info->endio_meta_workers, 684d20f7043SChris Mason &end_io_wq->work); 685e6dcd2dcSChris Mason else 686d20f7043SChris Mason btrfs_queue_worker(&fs_info->endio_workers, 687d20f7043SChris Mason &end_io_wq->work); 688d20f7043SChris Mason } 689ce9adaa5SChris Mason } 690ce9adaa5SChris Mason 6910cb59c99SJosef Bacik /* 6920cb59c99SJosef Bacik * For the metadata arg you want 6930cb59c99SJosef Bacik * 6940cb59c99SJosef Bacik * 0 - if data 6950cb59c99SJosef Bacik * 1 - if normal metadta 6960cb59c99SJosef Bacik * 2 - if writing to the free space cache area 6970cb59c99SJosef Bacik */ 69822c59948SChris Mason int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 69922c59948SChris Mason int metadata) 7000b86a832SChris Mason { 701ce9adaa5SChris Mason struct end_io_wq *end_io_wq; 702ce9adaa5SChris Mason end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); 703ce9adaa5SChris Mason if (!end_io_wq) 704ce9adaa5SChris Mason return -ENOMEM; 705ce9adaa5SChris Mason 706ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 707ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 70822c59948SChris Mason end_io_wq->info = info; 709ce9adaa5SChris Mason end_io_wq->error = 0; 710ce9adaa5SChris Mason end_io_wq->bio = bio; 71122c59948SChris Mason end_io_wq->metadata = metadata; 712ce9adaa5SChris Mason 713ce9adaa5SChris Mason bio->bi_private = end_io_wq; 714ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 71522c59948SChris Mason return 0; 71622c59948SChris Mason } 71722c59948SChris Mason 718b64a2851SChris Mason unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 7194854ddd0SChris Mason { 7204854ddd0SChris Mason unsigned long limit = min_t(unsigned long, 7214854ddd0SChris Mason info->workers.max_workers, 7224854ddd0SChris Mason info->fs_devices->open_devices); 7234854ddd0SChris Mason return 256 * limit; 7244854ddd0SChris Mason } 7254854ddd0SChris Mason 7264a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7274a69a410SChris Mason { 7284a69a410SChris Mason struct async_submit_bio *async; 72979787eaaSJeff Mahoney int ret; 7304a69a410SChris Mason 7314a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 73279787eaaSJeff Mahoney ret = async->submit_bio_start(async->inode, async->rw, async->bio, 733eaf25d93SChris Mason async->mirror_num, async->bio_flags, 734eaf25d93SChris Mason async->bio_offset); 73579787eaaSJeff Mahoney if (ret) 73679787eaaSJeff Mahoney async->error = ret; 7374a69a410SChris Mason } 7384a69a410SChris Mason 7394a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7408b712842SChris Mason { 7418b712842SChris Mason struct btrfs_fs_info *fs_info; 7428b712842SChris Mason struct async_submit_bio *async; 7434854ddd0SChris Mason int limit; 7448b712842SChris Mason 7458b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7468b712842SChris Mason fs_info = BTRFS_I(async->inode)->root->fs_info; 7474854ddd0SChris Mason 748b64a2851SChris Mason limit = btrfs_async_submit_limit(fs_info); 7494854ddd0SChris Mason limit = limit * 2 / 3; 7504854ddd0SChris Mason 75166657b31SJosef Bacik if (atomic_dec_return(&fs_info->nr_async_submits) < limit && 752b64a2851SChris Mason waitqueue_active(&fs_info->async_submit_wait)) 7534854ddd0SChris Mason wake_up(&fs_info->async_submit_wait); 7544854ddd0SChris Mason 75579787eaaSJeff Mahoney /* If an error occured we just want to clean up the bio and move on */ 75679787eaaSJeff Mahoney if (async->error) { 75779787eaaSJeff Mahoney bio_endio(async->bio, async->error); 75879787eaaSJeff Mahoney return; 75979787eaaSJeff Mahoney } 76079787eaaSJeff Mahoney 7614a69a410SChris Mason async->submit_bio_done(async->inode, async->rw, async->bio, 762eaf25d93SChris Mason async->mirror_num, async->bio_flags, 763eaf25d93SChris Mason async->bio_offset); 7644a69a410SChris Mason } 7654a69a410SChris Mason 7664a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7674a69a410SChris Mason { 7684a69a410SChris Mason struct async_submit_bio *async; 7694a69a410SChris Mason 7704a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7718b712842SChris Mason kfree(async); 7728b712842SChris Mason } 7738b712842SChris Mason 77444b8bd7eSChris Mason int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, 77544b8bd7eSChris Mason int rw, struct bio *bio, int mirror_num, 776c8b97818SChris Mason unsigned long bio_flags, 777eaf25d93SChris Mason u64 bio_offset, 7784a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start, 7794a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done) 78044b8bd7eSChris Mason { 78144b8bd7eSChris Mason struct async_submit_bio *async; 78244b8bd7eSChris Mason 78344b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 78444b8bd7eSChris Mason if (!async) 78544b8bd7eSChris Mason return -ENOMEM; 78644b8bd7eSChris Mason 78744b8bd7eSChris Mason async->inode = inode; 78844b8bd7eSChris Mason async->rw = rw; 78944b8bd7eSChris Mason async->bio = bio; 79044b8bd7eSChris Mason async->mirror_num = mirror_num; 7914a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7924a69a410SChris Mason async->submit_bio_done = submit_bio_done; 7934a69a410SChris Mason 7944a69a410SChris Mason async->work.func = run_one_async_start; 7954a69a410SChris Mason async->work.ordered_func = run_one_async_done; 7964a69a410SChris Mason async->work.ordered_free = run_one_async_free; 7974a69a410SChris Mason 7988b712842SChris Mason async->work.flags = 0; 799c8b97818SChris Mason async->bio_flags = bio_flags; 800eaf25d93SChris Mason async->bio_offset = bio_offset; 8018c8bee1dSChris Mason 80279787eaaSJeff Mahoney async->error = 0; 80379787eaaSJeff Mahoney 804cb03c743SChris Mason atomic_inc(&fs_info->nr_async_submits); 805d313d7a3SChris Mason 8067b6d91daSChristoph Hellwig if (rw & REQ_SYNC) 807d313d7a3SChris Mason btrfs_set_work_high_prio(&async->work); 808d313d7a3SChris Mason 8098b712842SChris Mason btrfs_queue_worker(&fs_info->workers, &async->work); 8109473f16cSChris Mason 811771ed689SChris Mason while (atomic_read(&fs_info->async_submit_draining) && 812771ed689SChris Mason atomic_read(&fs_info->nr_async_submits)) { 813771ed689SChris Mason wait_event(fs_info->async_submit_wait, 814771ed689SChris Mason (atomic_read(&fs_info->nr_async_submits) == 0)); 815771ed689SChris Mason } 816771ed689SChris Mason 81744b8bd7eSChris Mason return 0; 81844b8bd7eSChris Mason } 81944b8bd7eSChris Mason 820ce3ed71aSChris Mason static int btree_csum_one_bio(struct bio *bio) 821ce3ed71aSChris Mason { 822ce3ed71aSChris Mason struct bio_vec *bvec = bio->bi_io_vec; 823ce3ed71aSChris Mason int bio_index = 0; 824ce3ed71aSChris Mason struct btrfs_root *root; 82579787eaaSJeff Mahoney int ret = 0; 826ce3ed71aSChris Mason 827ce3ed71aSChris Mason WARN_ON(bio->bi_vcnt <= 0); 828ce3ed71aSChris Mason while (bio_index < bio->bi_vcnt) { 829ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 83079787eaaSJeff Mahoney ret = csum_dirty_buffer(root, bvec->bv_page); 83179787eaaSJeff Mahoney if (ret) 83279787eaaSJeff Mahoney break; 833ce3ed71aSChris Mason bio_index++; 834ce3ed71aSChris Mason bvec++; 835ce3ed71aSChris Mason } 83679787eaaSJeff Mahoney return ret; 837ce3ed71aSChris Mason } 838ce3ed71aSChris Mason 8394a69a410SChris Mason static int __btree_submit_bio_start(struct inode *inode, int rw, 8404a69a410SChris Mason struct bio *bio, int mirror_num, 841eaf25d93SChris Mason unsigned long bio_flags, 842eaf25d93SChris Mason u64 bio_offset) 84322c59948SChris Mason { 8448b712842SChris Mason /* 8458b712842SChris Mason * when we're called for a write, we're already in the async 8465443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8478b712842SChris Mason */ 84879787eaaSJeff Mahoney return btree_csum_one_bio(bio); 84922c59948SChris Mason } 85022c59948SChris Mason 8514a69a410SChris Mason static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 852eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 853eaf25d93SChris Mason u64 bio_offset) 8544a69a410SChris Mason { 8558b712842SChris Mason /* 8564a69a410SChris Mason * when we're called for a write, we're already in the async 8574a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 8588b712842SChris Mason */ 8598b712842SChris Mason return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 8600b86a832SChris Mason } 8610b86a832SChris Mason 862de0022b9SJosef Bacik static int check_async_write(struct inode *inode, unsigned long bio_flags) 863de0022b9SJosef Bacik { 864de0022b9SJosef Bacik if (bio_flags & EXTENT_BIO_TREE_LOG) 865de0022b9SJosef Bacik return 0; 866de0022b9SJosef Bacik #ifdef CONFIG_X86 867de0022b9SJosef Bacik if (cpu_has_xmm4_2) 868de0022b9SJosef Bacik return 0; 869de0022b9SJosef Bacik #endif 870de0022b9SJosef Bacik return 1; 871de0022b9SJosef Bacik } 872de0022b9SJosef Bacik 87344b8bd7eSChris Mason static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 874eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 875eaf25d93SChris Mason u64 bio_offset) 87644b8bd7eSChris Mason { 877de0022b9SJosef Bacik int async = check_async_write(inode, bio_flags); 8784a69a410SChris Mason int ret; 879cad321adSChris Mason 8807b6d91daSChristoph Hellwig if (!(rw & REQ_WRITE)) { 881f3f266abSChris Mason 882cad321adSChris Mason /* 883cad321adSChris Mason * called for a read, do the setup so that checksum validation 884cad321adSChris Mason * can happen in the async kernel threads 885cad321adSChris Mason */ 886f3f266abSChris Mason ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 887f3f266abSChris Mason bio, 1); 8881d4284bdSChris Mason if (ret) 8891d4284bdSChris Mason return ret; 8904a69a410SChris Mason return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 8916f3577bdSChris Mason mirror_num, 0); 892de0022b9SJosef Bacik } else if (!async) { 893de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 894de0022b9SJosef Bacik if (ret) 895de0022b9SJosef Bacik return ret; 896de0022b9SJosef Bacik return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 897de0022b9SJosef Bacik mirror_num, 0); 89844b8bd7eSChris Mason } 899d313d7a3SChris Mason 900cad321adSChris Mason /* 901cad321adSChris Mason * kthread helpers are used to submit writes so that checksumming 902cad321adSChris Mason * can happen in parallel across all CPUs 903cad321adSChris Mason */ 90444b8bd7eSChris Mason return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 905c8b97818SChris Mason inode, rw, bio, mirror_num, 0, 906eaf25d93SChris Mason bio_offset, 9074a69a410SChris Mason __btree_submit_bio_start, 9084a69a410SChris Mason __btree_submit_bio_done); 90944b8bd7eSChris Mason } 91044b8bd7eSChris Mason 9113dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 912784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 913a6bc32b8SMel Gorman struct page *newpage, struct page *page, 914a6bc32b8SMel Gorman enum migrate_mode mode) 915784b4e29SChris Mason { 916784b4e29SChris Mason /* 917784b4e29SChris Mason * we can't safely write a btree page from here, 918784b4e29SChris Mason * we haven't done the locking hook 919784b4e29SChris Mason */ 920784b4e29SChris Mason if (PageDirty(page)) 921784b4e29SChris Mason return -EAGAIN; 922784b4e29SChris Mason /* 923784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 924784b4e29SChris Mason * We must have no buffers or drop them. 925784b4e29SChris Mason */ 926784b4e29SChris Mason if (page_has_private(page) && 927784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 928784b4e29SChris Mason return -EAGAIN; 929a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 930784b4e29SChris Mason } 9313dd1462eSJan Beulich #endif 932784b4e29SChris Mason 9330da5468fSChris Mason 9340da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9350da5468fSChris Mason struct writeback_control *wbc) 9360da5468fSChris Mason { 937d1310b2eSChris Mason struct extent_io_tree *tree; 938d1310b2eSChris Mason tree = &BTRFS_I(mapping->host)->io_tree; 939d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 940b9473439SChris Mason struct btrfs_root *root = BTRFS_I(mapping->host)->root; 941793955bcSChris Mason u64 num_dirty; 94224ab9cd8SChris Mason unsigned long thresh = 32 * 1024 * 1024; 943448d640bSChris Mason 944448d640bSChris Mason if (wbc->for_kupdate) 945448d640bSChris Mason return 0; 946448d640bSChris Mason 947b9473439SChris Mason /* this is a bit racy, but that's ok */ 948b9473439SChris Mason num_dirty = root->fs_info->dirty_metadata_bytes; 949d397712bSChris Mason if (num_dirty < thresh) 950793955bcSChris Mason return 0; 951793955bcSChris Mason } 9520b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9530da5468fSChris Mason } 9540da5468fSChris Mason 955b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9565f39d397SChris Mason { 957d1310b2eSChris Mason struct extent_io_tree *tree; 958d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9598ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 9605f39d397SChris Mason } 9615f39d397SChris Mason 96270dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9635f39d397SChris Mason { 96498509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 96598509cfcSChris Mason return 0; 9660c4e538bSDavid Sterba /* 9670c4e538bSDavid Sterba * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing 9680c4e538bSDavid Sterba * slab allocation from alloc_extent_state down the callchain where 9690c4e538bSDavid Sterba * it'd hit a BUG_ON as those flags are not allowed. 9700c4e538bSDavid Sterba */ 9710c4e538bSDavid Sterba gfp_flags &= ~GFP_SLAB_BUG_MASK; 9720c4e538bSDavid Sterba 9733083ee2eSJosef Bacik return try_release_extent_buffer(page, gfp_flags); 974d98237b3SChris Mason } 975d98237b3SChris Mason 9765f39d397SChris Mason static void btree_invalidatepage(struct page *page, unsigned long offset) 977d98237b3SChris Mason { 978d1310b2eSChris Mason struct extent_io_tree *tree; 979d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9805f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9815f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9829ad6b7bcSChris Mason if (PagePrivate(page)) { 983d397712bSChris Mason printk(KERN_WARNING "btrfs warning page private not zero " 984d397712bSChris Mason "on page %llu\n", (unsigned long long)page_offset(page)); 9859ad6b7bcSChris Mason ClearPagePrivate(page); 9869ad6b7bcSChris Mason set_page_private(page, 0); 9879ad6b7bcSChris Mason page_cache_release(page); 9889ad6b7bcSChris Mason } 989d98237b3SChris Mason } 990d98237b3SChris Mason 9910b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9920b32f4bbSJosef Bacik { 9930b32f4bbSJosef Bacik struct extent_buffer *eb; 9940b32f4bbSJosef Bacik 9950b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9960b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9970b32f4bbSJosef Bacik BUG_ON(!eb); 9980b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9990b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10000b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 10010b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10020b32f4bbSJosef Bacik } 10030b32f4bbSJosef Bacik 10047f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1005d98237b3SChris Mason .readpage = btree_readpage, 10060da5468fSChris Mason .writepages = btree_writepages, 10075f39d397SChris Mason .releasepage = btree_releasepage, 10085f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10095a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1010784b4e29SChris Mason .migratepage = btree_migratepage, 10115a92bc88SChris Mason #endif 10120b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1013d98237b3SChris Mason }; 1014123abc88SChris Mason 1015ca7a79adSChris Mason int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1016ca7a79adSChris Mason u64 parent_transid) 1017090d1875SChris Mason { 10185f39d397SChris Mason struct extent_buffer *buf = NULL; 10195f39d397SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 1020de428b63SChris Mason int ret = 0; 1021090d1875SChris Mason 1022db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 10235f39d397SChris Mason if (!buf) 1024090d1875SChris Mason return 0; 1025d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1026bb82ab88SArne Jansen buf, 0, WAIT_NONE, btree_get_extent, 0); 10275f39d397SChris Mason free_extent_buffer(buf); 1028de428b63SChris Mason return ret; 1029090d1875SChris Mason } 1030090d1875SChris Mason 1031ab0fff03SArne Jansen int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1032ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1033ab0fff03SArne Jansen { 1034ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 1035ab0fff03SArne Jansen struct inode *btree_inode = root->fs_info->btree_inode; 1036ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1037ab0fff03SArne Jansen int ret; 1038ab0fff03SArne Jansen 1039ab0fff03SArne Jansen buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1040ab0fff03SArne Jansen if (!buf) 1041ab0fff03SArne Jansen return 0; 1042ab0fff03SArne Jansen 1043ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1044ab0fff03SArne Jansen 1045ab0fff03SArne Jansen ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1046ab0fff03SArne Jansen btree_get_extent, mirror_num); 1047ab0fff03SArne Jansen if (ret) { 1048ab0fff03SArne Jansen free_extent_buffer(buf); 1049ab0fff03SArne Jansen return ret; 1050ab0fff03SArne Jansen } 1051ab0fff03SArne Jansen 1052ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1053ab0fff03SArne Jansen free_extent_buffer(buf); 1054ab0fff03SArne Jansen return -EIO; 10550b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1056ab0fff03SArne Jansen *eb = buf; 1057ab0fff03SArne Jansen } else { 1058ab0fff03SArne Jansen free_extent_buffer(buf); 1059ab0fff03SArne Jansen } 1060ab0fff03SArne Jansen return 0; 1061ab0fff03SArne Jansen } 1062ab0fff03SArne Jansen 10630999df54SChris Mason struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, 10640999df54SChris Mason u64 bytenr, u32 blocksize) 10650999df54SChris Mason { 10660999df54SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 10670999df54SChris Mason struct extent_buffer *eb; 10680999df54SChris Mason eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, 1069f09d1f60SDavid Sterba bytenr, blocksize); 10700999df54SChris Mason return eb; 10710999df54SChris Mason } 10720999df54SChris Mason 10730999df54SChris Mason struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 10740999df54SChris Mason u64 bytenr, u32 blocksize) 10750999df54SChris Mason { 10760999df54SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 10770999df54SChris Mason struct extent_buffer *eb; 10780999df54SChris Mason 10790999df54SChris Mason eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, 1080727011e0SChris Mason bytenr, blocksize); 10810999df54SChris Mason return eb; 10820999df54SChris Mason } 10830999df54SChris Mason 10840999df54SChris Mason 1085e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1086e02119d5SChris Mason { 1087727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 10888aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1089e02119d5SChris Mason } 1090e02119d5SChris Mason 1091e02119d5SChris Mason int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1092e02119d5SChris Mason { 1093727011e0SChris Mason return filemap_fdatawait_range(buf->pages[0]->mapping, 10948aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1095e02119d5SChris Mason } 1096e02119d5SChris Mason 1097db94535dSChris Mason struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1098ca7a79adSChris Mason u32 blocksize, u64 parent_transid) 1099e20d96d6SChris Mason { 11005f39d397SChris Mason struct extent_buffer *buf = NULL; 110119c00ddcSChris Mason int ret; 110219c00ddcSChris Mason 1103db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 11045f39d397SChris Mason if (!buf) 1105d98237b3SChris Mason return NULL; 1106e4204dedSChris Mason 1107ca7a79adSChris Mason ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 11085f39d397SChris Mason return buf; 1109ce9adaa5SChris Mason 1110eb60ceacSChris Mason } 1111eb60ceacSChris Mason 1112d5c13f92SJeff Mahoney void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, 11135f39d397SChris Mason struct extent_buffer *buf) 1114ed2ff2cbSChris Mason { 111555c69072SChris Mason if (btrfs_header_generation(buf) == 1116925baeddSChris Mason root->fs_info->running_transaction->transid) { 1117b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1118b4ce94deSChris Mason 1119b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1120b9473439SChris Mason spin_lock(&root->fs_info->delalloc_lock); 1121b9473439SChris Mason if (root->fs_info->dirty_metadata_bytes >= buf->len) 1122b9473439SChris Mason root->fs_info->dirty_metadata_bytes -= buf->len; 1123d5c13f92SJeff Mahoney else { 1124d5c13f92SJeff Mahoney spin_unlock(&root->fs_info->delalloc_lock); 1125d5c13f92SJeff Mahoney btrfs_panic(root->fs_info, -EOVERFLOW, 1126d5c13f92SJeff Mahoney "Can't clear %lu bytes from " 1127533574c6SJoe Perches " dirty_mdatadata_bytes (%llu)", 1128d5c13f92SJeff Mahoney buf->len, 1129d5c13f92SJeff Mahoney root->fs_info->dirty_metadata_bytes); 1130d5c13f92SJeff Mahoney } 1131b9473439SChris Mason spin_unlock(&root->fs_info->delalloc_lock); 1132b9473439SChris Mason } 1133b4ce94deSChris Mason 1134b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1135b9473439SChris Mason btrfs_set_lock_blocking(buf); 11360b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1137925baeddSChris Mason } 11385f39d397SChris Mason } 11395f39d397SChris Mason 1140143bede5SJeff Mahoney static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, 114187ee04ebSChris Mason u32 stripesize, struct btrfs_root *root, 11429f5fae2fSChris Mason struct btrfs_fs_info *fs_info, 1143e20d96d6SChris Mason u64 objectid) 1144d97e63b6SChris Mason { 1145cfaa7295SChris Mason root->node = NULL; 1146a28ec197SChris Mason root->commit_root = NULL; 1147db94535dSChris Mason root->sectorsize = sectorsize; 1148db94535dSChris Mason root->nodesize = nodesize; 1149db94535dSChris Mason root->leafsize = leafsize; 115087ee04ebSChris Mason root->stripesize = stripesize; 1151123abc88SChris Mason root->ref_cows = 0; 11520b86a832SChris Mason root->track_dirty = 0; 1153c71bf099SYan, Zheng root->in_radix = 0; 1154d68fc57bSYan, Zheng root->orphan_item_inserted = 0; 1155d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11560b86a832SChris Mason 11570f7d52f4SChris Mason root->objectid = objectid; 11580f7d52f4SChris Mason root->last_trans = 0; 115913a8a7c8SYan, Zheng root->highest_objectid = 0; 116058176a96SJosef Bacik root->name = NULL; 11616bef4d31SEric Paris root->inode_tree = RB_ROOT; 116216cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1163f0486c68SYan, Zheng root->block_rsv = NULL; 1164d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 11650b86a832SChris Mason 11660b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11675d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1168d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 11695d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1170f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 1171a2135011SChris Mason mutex_init(&root->objectid_mutex); 1172e02119d5SChris Mason mutex_init(&root->log_mutex); 11737237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11747237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11757237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11767237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11777237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11787237f183SYan Zheng atomic_set(&root->log_writers, 0); 11792ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11808a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 11817237f183SYan Zheng root->log_transid = 0; 1182257c62e1SChris Mason root->last_log_commit = 0; 1183d0c803c4SChris Mason extent_io_tree_init(&root->dirty_log_pages, 1184f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 1185017e5369SChris Mason 11863768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 11873768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 11886702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 118958176a96SJosef Bacik memset(&root->root_kobj, 0, sizeof(root->root_kobj)); 11903f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 119158176a96SJosef Bacik init_completion(&root->kobj_unregister); 11926702ed49SChris Mason root->defrag_running = 0; 11934d775673SChris Mason root->root_key.objectid = objectid; 11940ee5dc67SAl Viro root->anon_dev = 0; 11958ea05e3aSAlexander Block 11968ea05e3aSAlexander Block spin_lock_init(&root->root_times_lock); 11973768f368SChris Mason } 11983768f368SChris Mason 1199200a5c17SJeff Mahoney static int __must_check find_and_setup_root(struct btrfs_root *tree_root, 12009f5fae2fSChris Mason struct btrfs_fs_info *fs_info, 12019f5fae2fSChris Mason u64 objectid, 1202e20d96d6SChris Mason struct btrfs_root *root) 12033768f368SChris Mason { 12043768f368SChris Mason int ret; 1205db94535dSChris Mason u32 blocksize; 120684234f3aSYan Zheng u64 generation; 12073768f368SChris Mason 1208db94535dSChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 120987ee04ebSChris Mason tree_root->sectorsize, tree_root->stripesize, 121087ee04ebSChris Mason root, fs_info, objectid); 12113768f368SChris Mason ret = btrfs_find_last_root(tree_root, objectid, 12123768f368SChris Mason &root->root_item, &root->root_key); 12134df27c4dSYan, Zheng if (ret > 0) 12144df27c4dSYan, Zheng return -ENOENT; 1215200a5c17SJeff Mahoney else if (ret < 0) 1216200a5c17SJeff Mahoney return ret; 12173768f368SChris Mason 121884234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1219db94535dSChris Mason blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); 1220af31f5e5SChris Mason root->commit_root = NULL; 1221db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 122284234f3aSYan Zheng blocksize, generation); 1223b9fab919SChris Mason if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) { 122468433b73SChris Mason free_extent_buffer(root->node); 1225af31f5e5SChris Mason root->node = NULL; 122668433b73SChris Mason return -EIO; 122768433b73SChris Mason } 12284df27c4dSYan, Zheng root->commit_root = btrfs_root_node(root); 1229d97e63b6SChris Mason return 0; 1230d97e63b6SChris Mason } 1231d97e63b6SChris Mason 1232f84a8bd6SAl Viro static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) 12336f07e42eSAl Viro { 12346f07e42eSAl Viro struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); 12356f07e42eSAl Viro if (root) 12366f07e42eSAl Viro root->fs_info = fs_info; 12376f07e42eSAl Viro return root; 12386f07e42eSAl Viro } 12396f07e42eSAl Viro 124020897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 124120897f5cSArne Jansen struct btrfs_fs_info *fs_info, 124220897f5cSArne Jansen u64 objectid) 124320897f5cSArne Jansen { 124420897f5cSArne Jansen struct extent_buffer *leaf; 124520897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 124620897f5cSArne Jansen struct btrfs_root *root; 124720897f5cSArne Jansen struct btrfs_key key; 124820897f5cSArne Jansen int ret = 0; 124920897f5cSArne Jansen u64 bytenr; 125020897f5cSArne Jansen 125120897f5cSArne Jansen root = btrfs_alloc_root(fs_info); 125220897f5cSArne Jansen if (!root) 125320897f5cSArne Jansen return ERR_PTR(-ENOMEM); 125420897f5cSArne Jansen 125520897f5cSArne Jansen __setup_root(tree_root->nodesize, tree_root->leafsize, 125620897f5cSArne Jansen tree_root->sectorsize, tree_root->stripesize, 125720897f5cSArne Jansen root, fs_info, objectid); 125820897f5cSArne Jansen root->root_key.objectid = objectid; 125920897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 126020897f5cSArne Jansen root->root_key.offset = 0; 126120897f5cSArne Jansen 126220897f5cSArne Jansen leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 126320897f5cSArne Jansen 0, objectid, NULL, 0, 0, 0); 126420897f5cSArne Jansen if (IS_ERR(leaf)) { 126520897f5cSArne Jansen ret = PTR_ERR(leaf); 126620897f5cSArne Jansen goto fail; 126720897f5cSArne Jansen } 126820897f5cSArne Jansen 126920897f5cSArne Jansen bytenr = leaf->start; 127020897f5cSArne Jansen memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 127120897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 127220897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 127320897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 127420897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 127520897f5cSArne Jansen root->node = leaf; 127620897f5cSArne Jansen 127720897f5cSArne Jansen write_extent_buffer(leaf, fs_info->fsid, 127820897f5cSArne Jansen (unsigned long)btrfs_header_fsid(leaf), 127920897f5cSArne Jansen BTRFS_FSID_SIZE); 128020897f5cSArne Jansen write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 128120897f5cSArne Jansen (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 128220897f5cSArne Jansen BTRFS_UUID_SIZE); 128320897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 128420897f5cSArne Jansen 128520897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 128620897f5cSArne Jansen root->track_dirty = 1; 128720897f5cSArne Jansen 128820897f5cSArne Jansen 128920897f5cSArne Jansen root->root_item.flags = 0; 129020897f5cSArne Jansen root->root_item.byte_limit = 0; 129120897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 129220897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 129320897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 129420897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 129520897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 129620897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 129720897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 129820897f5cSArne Jansen root->root_item.drop_level = 0; 129920897f5cSArne Jansen 130020897f5cSArne Jansen key.objectid = objectid; 130120897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 130220897f5cSArne Jansen key.offset = 0; 130320897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 130420897f5cSArne Jansen if (ret) 130520897f5cSArne Jansen goto fail; 130620897f5cSArne Jansen 130720897f5cSArne Jansen btrfs_tree_unlock(leaf); 130820897f5cSArne Jansen 130920897f5cSArne Jansen fail: 131020897f5cSArne Jansen if (ret) 131120897f5cSArne Jansen return ERR_PTR(ret); 131220897f5cSArne Jansen 131320897f5cSArne Jansen return root; 131420897f5cSArne Jansen } 131520897f5cSArne Jansen 13167237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1317e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13180f7d52f4SChris Mason { 13190f7d52f4SChris Mason struct btrfs_root *root; 13200f7d52f4SChris Mason struct btrfs_root *tree_root = fs_info->tree_root; 13217237f183SYan Zheng struct extent_buffer *leaf; 1322e02119d5SChris Mason 13236f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1324e02119d5SChris Mason if (!root) 13257237f183SYan Zheng return ERR_PTR(-ENOMEM); 1326e02119d5SChris Mason 1327e02119d5SChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 1328e02119d5SChris Mason tree_root->sectorsize, tree_root->stripesize, 1329e02119d5SChris Mason root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1330e02119d5SChris Mason 1331e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1332e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1333e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 13347237f183SYan Zheng /* 13357237f183SYan Zheng * log trees do not get reference counted because they go away 13367237f183SYan Zheng * before a real commit is actually done. They do store pointers 13377237f183SYan Zheng * to file data extents, and those reference counts still get 13387237f183SYan Zheng * updated (along with back refs to the log tree). 13397237f183SYan Zheng */ 1340e02119d5SChris Mason root->ref_cows = 0; 1341e02119d5SChris Mason 13425d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 134366d7e7f0SArne Jansen BTRFS_TREE_LOG_OBJECTID, NULL, 13445581a51aSJan Schmidt 0, 0, 0); 13457237f183SYan Zheng if (IS_ERR(leaf)) { 13467237f183SYan Zheng kfree(root); 13477237f183SYan Zheng return ERR_CAST(leaf); 13487237f183SYan Zheng } 1349e02119d5SChris Mason 13505d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 13515d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 13525d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 13535d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 13545d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 13557237f183SYan Zheng root->node = leaf; 1356e02119d5SChris Mason 1357e02119d5SChris Mason write_extent_buffer(root->node, root->fs_info->fsid, 1358e02119d5SChris Mason (unsigned long)btrfs_header_fsid(root->node), 1359e02119d5SChris Mason BTRFS_FSID_SIZE); 1360e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1361e02119d5SChris Mason btrfs_tree_unlock(root->node); 13627237f183SYan Zheng return root; 13637237f183SYan Zheng } 13647237f183SYan Zheng 13657237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13667237f183SYan Zheng struct btrfs_fs_info *fs_info) 13677237f183SYan Zheng { 13687237f183SYan Zheng struct btrfs_root *log_root; 13697237f183SYan Zheng 13707237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13717237f183SYan Zheng if (IS_ERR(log_root)) 13727237f183SYan Zheng return PTR_ERR(log_root); 13737237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13747237f183SYan Zheng fs_info->log_root_tree = log_root; 13757237f183SYan Zheng return 0; 13767237f183SYan Zheng } 13777237f183SYan Zheng 13787237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13797237f183SYan Zheng struct btrfs_root *root) 13807237f183SYan Zheng { 13817237f183SYan Zheng struct btrfs_root *log_root; 13827237f183SYan Zheng struct btrfs_inode_item *inode_item; 13837237f183SYan Zheng 13847237f183SYan Zheng log_root = alloc_log_tree(trans, root->fs_info); 13857237f183SYan Zheng if (IS_ERR(log_root)) 13867237f183SYan Zheng return PTR_ERR(log_root); 13877237f183SYan Zheng 13887237f183SYan Zheng log_root->last_trans = trans->transid; 13897237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13907237f183SYan Zheng 13917237f183SYan Zheng inode_item = &log_root->root_item.inode; 13927237f183SYan Zheng inode_item->generation = cpu_to_le64(1); 13937237f183SYan Zheng inode_item->size = cpu_to_le64(3); 13947237f183SYan Zheng inode_item->nlink = cpu_to_le32(1); 13957237f183SYan Zheng inode_item->nbytes = cpu_to_le64(root->leafsize); 13967237f183SYan Zheng inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 13977237f183SYan Zheng 13985d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13997237f183SYan Zheng 14007237f183SYan Zheng WARN_ON(root->log_root); 14017237f183SYan Zheng root->log_root = log_root; 14027237f183SYan Zheng root->log_transid = 0; 1403257c62e1SChris Mason root->last_log_commit = 0; 1404e02119d5SChris Mason return 0; 1405e02119d5SChris Mason } 1406e02119d5SChris Mason 1407e02119d5SChris Mason struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, 1408e02119d5SChris Mason struct btrfs_key *location) 1409e02119d5SChris Mason { 1410e02119d5SChris Mason struct btrfs_root *root; 1411e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 14120f7d52f4SChris Mason struct btrfs_path *path; 14135f39d397SChris Mason struct extent_buffer *l; 141484234f3aSYan Zheng u64 generation; 1415db94535dSChris Mason u32 blocksize; 14160f7d52f4SChris Mason int ret = 0; 14178ea05e3aSAlexander Block int slot; 14180f7d52f4SChris Mason 14196f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 14200cf6c620SChris Mason if (!root) 14210f7d52f4SChris Mason return ERR_PTR(-ENOMEM); 14220f7d52f4SChris Mason if (location->offset == (u64)-1) { 1423db94535dSChris Mason ret = find_and_setup_root(tree_root, fs_info, 14240f7d52f4SChris Mason location->objectid, root); 14250f7d52f4SChris Mason if (ret) { 14260f7d52f4SChris Mason kfree(root); 14270f7d52f4SChris Mason return ERR_PTR(ret); 14280f7d52f4SChris Mason } 142913a8a7c8SYan, Zheng goto out; 14300f7d52f4SChris Mason } 14310f7d52f4SChris Mason 1432db94535dSChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 143387ee04ebSChris Mason tree_root->sectorsize, tree_root->stripesize, 143487ee04ebSChris Mason root, fs_info, location->objectid); 14350f7d52f4SChris Mason 14360f7d52f4SChris Mason path = btrfs_alloc_path(); 1437db5b493aSTsutomu Itoh if (!path) { 1438db5b493aSTsutomu Itoh kfree(root); 1439db5b493aSTsutomu Itoh return ERR_PTR(-ENOMEM); 1440db5b493aSTsutomu Itoh } 14410f7d52f4SChris Mason ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); 144213a8a7c8SYan, Zheng if (ret == 0) { 14435f39d397SChris Mason l = path->nodes[0]; 14448ea05e3aSAlexander Block slot = path->slots[0]; 14458ea05e3aSAlexander Block btrfs_read_root_item(tree_root, l, slot, &root->root_item); 144644b36eb2SYan Zheng memcpy(&root->root_key, location, sizeof(*location)); 144713a8a7c8SYan, Zheng } 14480f7d52f4SChris Mason btrfs_free_path(path); 14490f7d52f4SChris Mason if (ret) { 14505e540f77STsutomu Itoh kfree(root); 145113a8a7c8SYan, Zheng if (ret > 0) 145213a8a7c8SYan, Zheng ret = -ENOENT; 14530f7d52f4SChris Mason return ERR_PTR(ret); 14540f7d52f4SChris Mason } 145513a8a7c8SYan, Zheng 145684234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1457db94535dSChris Mason blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); 1458db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 145984234f3aSYan Zheng blocksize, generation); 14605d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 146179787eaaSJeff Mahoney BUG_ON(!root->node); /* -ENOMEM */ 146213a8a7c8SYan, Zheng out: 146308fe4db1SLi Zefan if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { 14640f7d52f4SChris Mason root->ref_cows = 1; 146508fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 146608fe4db1SLi Zefan } 146713a8a7c8SYan, Zheng 14685eda7b5eSChris Mason return root; 14695eda7b5eSChris Mason } 14705eda7b5eSChris Mason 1471edbd8d4eSChris Mason struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, 1472edbd8d4eSChris Mason struct btrfs_key *location) 14735eda7b5eSChris Mason { 14745eda7b5eSChris Mason struct btrfs_root *root; 14755eda7b5eSChris Mason int ret; 14765eda7b5eSChris Mason 1477edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1478edbd8d4eSChris Mason return fs_info->tree_root; 1479edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1480edbd8d4eSChris Mason return fs_info->extent_root; 14818f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 14828f18cf13SChris Mason return fs_info->chunk_root; 14838f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 14848f18cf13SChris Mason return fs_info->dev_root; 14850403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 14860403e47eSYan Zheng return fs_info->csum_root; 1487bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1488bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1489bcef60f2SArne Jansen ERR_PTR(-ENOENT); 14904df27c4dSYan, Zheng again: 14914df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 14925eda7b5eSChris Mason root = radix_tree_lookup(&fs_info->fs_roots_radix, 14935eda7b5eSChris Mason (unsigned long)location->objectid); 14944df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 14955eda7b5eSChris Mason if (root) 14965eda7b5eSChris Mason return root; 14975eda7b5eSChris Mason 1498e02119d5SChris Mason root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); 14995eda7b5eSChris Mason if (IS_ERR(root)) 15005eda7b5eSChris Mason return root; 15013394e160SChris Mason 1502581bb050SLi Zefan root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1503581bb050SLi Zefan root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1504581bb050SLi Zefan GFP_NOFS); 150535a30d7cSDavid Sterba if (!root->free_ino_pinned || !root->free_ino_ctl) { 150635a30d7cSDavid Sterba ret = -ENOMEM; 1507581bb050SLi Zefan goto fail; 150835a30d7cSDavid Sterba } 1509581bb050SLi Zefan 1510581bb050SLi Zefan btrfs_init_free_ino_ctl(root); 1511581bb050SLi Zefan mutex_init(&root->fs_commit_mutex); 1512581bb050SLi Zefan spin_lock_init(&root->cache_lock); 1513581bb050SLi Zefan init_waitqueue_head(&root->cache_wait); 1514581bb050SLi Zefan 15150ee5dc67SAl Viro ret = get_anon_bdev(&root->anon_dev); 1516ac08aedfSChris Mason if (ret) 1517ac08aedfSChris Mason goto fail; 15183394e160SChris Mason 1519d68fc57bSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) { 1520d68fc57bSYan, Zheng ret = -ENOENT; 1521d68fc57bSYan, Zheng goto fail; 1522d68fc57bSYan, Zheng } 1523d68fc57bSYan, Zheng 1524d68fc57bSYan, Zheng ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); 1525d68fc57bSYan, Zheng if (ret < 0) 1526d68fc57bSYan, Zheng goto fail; 1527d68fc57bSYan, Zheng if (ret == 0) 1528d68fc57bSYan, Zheng root->orphan_item_inserted = 1; 1529d68fc57bSYan, Zheng 15304df27c4dSYan, Zheng ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 15314df27c4dSYan, Zheng if (ret) 15324df27c4dSYan, Zheng goto fail; 15334df27c4dSYan, Zheng 15344df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 15352619ba1fSChris Mason ret = radix_tree_insert(&fs_info->fs_roots_radix, 15362619ba1fSChris Mason (unsigned long)root->root_key.objectid, 15370f7d52f4SChris Mason root); 1538d68fc57bSYan, Zheng if (ret == 0) 15394df27c4dSYan, Zheng root->in_radix = 1; 1540d68fc57bSYan, Zheng 15414df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 15424df27c4dSYan, Zheng radix_tree_preload_end(); 15430f7d52f4SChris Mason if (ret) { 15444df27c4dSYan, Zheng if (ret == -EEXIST) { 15454df27c4dSYan, Zheng free_fs_root(root); 15464df27c4dSYan, Zheng goto again; 15470f7d52f4SChris Mason } 15484df27c4dSYan, Zheng goto fail; 15494df27c4dSYan, Zheng } 15504df27c4dSYan, Zheng 1551edbd8d4eSChris Mason ret = btrfs_find_dead_roots(fs_info->tree_root, 15525d4f98a2SYan Zheng root->root_key.objectid); 15534df27c4dSYan, Zheng WARN_ON(ret); 1554edbd8d4eSChris Mason return root; 15554df27c4dSYan, Zheng fail: 15564df27c4dSYan, Zheng free_fs_root(root); 15574df27c4dSYan, Zheng return ERR_PTR(ret); 1558edbd8d4eSChris Mason } 1559edbd8d4eSChris Mason 156004160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 156104160088SChris Mason { 156204160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 156304160088SChris Mason int ret = 0; 156404160088SChris Mason struct btrfs_device *device; 156504160088SChris Mason struct backing_dev_info *bdi; 1566b7967db7SChris Mason 15671f78160cSXiao Guangrong rcu_read_lock(); 15681f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1569dfe25020SChris Mason if (!device->bdev) 1570dfe25020SChris Mason continue; 157104160088SChris Mason bdi = blk_get_backing_dev_info(device->bdev); 157204160088SChris Mason if (bdi && bdi_congested(bdi, bdi_bits)) { 157304160088SChris Mason ret = 1; 157404160088SChris Mason break; 157504160088SChris Mason } 157604160088SChris Mason } 15771f78160cSXiao Guangrong rcu_read_unlock(); 157804160088SChris Mason return ret; 157904160088SChris Mason } 158004160088SChris Mason 158138b66988SChris Mason /* 1582ad081f14SJens Axboe * If this fails, caller must call bdi_destroy() to get rid of the 1583ad081f14SJens Axboe * bdi again. 1584ad081f14SJens Axboe */ 158504160088SChris Mason static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 158604160088SChris Mason { 1587ad081f14SJens Axboe int err; 1588ad081f14SJens Axboe 1589ad081f14SJens Axboe bdi->capabilities = BDI_CAP_MAP_COPY; 1590e6d086d8SJens Axboe err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); 1591ad081f14SJens Axboe if (err) 1592ad081f14SJens Axboe return err; 1593ad081f14SJens Axboe 15944575c9ccSChris Mason bdi->ra_pages = default_backing_dev_info.ra_pages; 159504160088SChris Mason bdi->congested_fn = btrfs_congested_fn; 159604160088SChris Mason bdi->congested_data = info; 159704160088SChris Mason return 0; 159804160088SChris Mason } 159904160088SChris Mason 16008b712842SChris Mason /* 16018b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16028b712842SChris Mason * functions. This is where read checksum verification actually happens 16038b712842SChris Mason */ 16048b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1605ce9adaa5SChris Mason { 1606ce9adaa5SChris Mason struct bio *bio; 16078b712842SChris Mason struct end_io_wq *end_io_wq; 16088b712842SChris Mason struct btrfs_fs_info *fs_info; 1609ce9adaa5SChris Mason int error; 1610ce9adaa5SChris Mason 16118b712842SChris Mason end_io_wq = container_of(work, struct end_io_wq, work); 1612ce9adaa5SChris Mason bio = end_io_wq->bio; 16138b712842SChris Mason fs_info = end_io_wq->info; 16148b712842SChris Mason 1615ce9adaa5SChris Mason error = end_io_wq->error; 1616ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1617ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 1618ce9adaa5SChris Mason kfree(end_io_wq); 1619ce9adaa5SChris Mason bio_endio(bio, error); 1620ce9adaa5SChris Mason } 162144b8bd7eSChris Mason 1622a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1623a74a4b97SChris Mason { 1624a74a4b97SChris Mason struct btrfs_root *root = arg; 1625a74a4b97SChris Mason 1626a74a4b97SChris Mason do { 162776dda93cSYan, Zheng if (!(root->fs_info->sb->s_flags & MS_RDONLY) && 162876dda93cSYan, Zheng mutex_trylock(&root->fs_info->cleaner_mutex)) { 162924bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 1630a74a4b97SChris Mason btrfs_clean_old_snapshots(root); 1631a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 16324cb5300bSChris Mason btrfs_run_defrag_inodes(root->fs_info); 163376dda93cSYan, Zheng } 1634a74a4b97SChris Mason 1635a0acae0eSTejun Heo if (!try_to_freeze()) { 1636a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 16378929ecfaSYan, Zheng if (!kthread_should_stop()) 1638a74a4b97SChris Mason schedule(); 1639a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1640a74a4b97SChris Mason } 1641a74a4b97SChris Mason } while (!kthread_should_stop()); 1642a74a4b97SChris Mason return 0; 1643a74a4b97SChris Mason } 1644a74a4b97SChris Mason 1645a74a4b97SChris Mason static int transaction_kthread(void *arg) 1646a74a4b97SChris Mason { 1647a74a4b97SChris Mason struct btrfs_root *root = arg; 1648a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1649a74a4b97SChris Mason struct btrfs_transaction *cur; 16508929ecfaSYan, Zheng u64 transid; 1651a74a4b97SChris Mason unsigned long now; 1652a74a4b97SChris Mason unsigned long delay; 1653914b2007SJan Kara bool cannot_commit; 1654a74a4b97SChris Mason 1655a74a4b97SChris Mason do { 1656914b2007SJan Kara cannot_commit = false; 1657a74a4b97SChris Mason delay = HZ * 30; 1658a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 1659a74a4b97SChris Mason 1660a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 1661a74a4b97SChris Mason cur = root->fs_info->running_transaction; 1662a74a4b97SChris Mason if (!cur) { 1663a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1664a74a4b97SChris Mason goto sleep; 1665a74a4b97SChris Mason } 166631153d81SYan Zheng 1667a74a4b97SChris Mason now = get_seconds(); 16688929ecfaSYan, Zheng if (!cur->blocked && 16698929ecfaSYan, Zheng (now < cur->start_time || now - cur->start_time < 30)) { 1670a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1671a74a4b97SChris Mason delay = HZ * 5; 1672a74a4b97SChris Mason goto sleep; 1673a74a4b97SChris Mason } 16748929ecfaSYan, Zheng transid = cur->transid; 1675a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 167656bec294SChris Mason 167779787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1678354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1679914b2007SJan Kara if (IS_ERR(trans)) { 1680354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1681914b2007SJan Kara cannot_commit = true; 168279787eaaSJeff Mahoney goto sleep; 1683914b2007SJan Kara } 16848929ecfaSYan, Zheng if (transid == trans->transid) { 168579787eaaSJeff Mahoney btrfs_commit_transaction(trans, root); 16868929ecfaSYan, Zheng } else { 16878929ecfaSYan, Zheng btrfs_end_transaction(trans, root); 16888929ecfaSYan, Zheng } 1689a74a4b97SChris Mason sleep: 1690a74a4b97SChris Mason wake_up_process(root->fs_info->cleaner_kthread); 1691a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1692a74a4b97SChris Mason 1693a0acae0eSTejun Heo if (!try_to_freeze()) { 1694a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 16958929ecfaSYan, Zheng if (!kthread_should_stop() && 1696914b2007SJan Kara (!btrfs_transaction_blocked(root->fs_info) || 1697914b2007SJan Kara cannot_commit)) 1698a74a4b97SChris Mason schedule_timeout(delay); 1699a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1700a74a4b97SChris Mason } 1701a74a4b97SChris Mason } while (!kthread_should_stop()); 1702a74a4b97SChris Mason return 0; 1703a74a4b97SChris Mason } 1704a74a4b97SChris Mason 1705af31f5e5SChris Mason /* 1706af31f5e5SChris Mason * this will find the highest generation in the array of 1707af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1708af31f5e5SChris Mason * or -1 if we can't find anything. 1709af31f5e5SChris Mason * 1710af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1711af31f5e5SChris Mason * generation of the latest root in the array with the generation 1712af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1713af31f5e5SChris Mason */ 1714af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1715af31f5e5SChris Mason { 1716af31f5e5SChris Mason u64 cur; 1717af31f5e5SChris Mason int newest_index = -1; 1718af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1719af31f5e5SChris Mason int i; 1720af31f5e5SChris Mason 1721af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1722af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1723af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1724af31f5e5SChris Mason if (cur == newest_gen) 1725af31f5e5SChris Mason newest_index = i; 1726af31f5e5SChris Mason } 1727af31f5e5SChris Mason 1728af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1729af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1730af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1731af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1732af31f5e5SChris Mason if (cur == newest_gen) 1733af31f5e5SChris Mason newest_index = 0; 1734af31f5e5SChris Mason } 1735af31f5e5SChris Mason return newest_index; 1736af31f5e5SChris Mason } 1737af31f5e5SChris Mason 1738af31f5e5SChris Mason 1739af31f5e5SChris Mason /* 1740af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1741af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1742af31f5e5SChris Mason * field in the fs_info struct 1743af31f5e5SChris Mason */ 1744af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1745af31f5e5SChris Mason u64 newest_gen) 1746af31f5e5SChris Mason { 1747af31f5e5SChris Mason int newest_index = -1; 1748af31f5e5SChris Mason 1749af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1750af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1751af31f5e5SChris Mason if (newest_index == -1) { 1752af31f5e5SChris Mason info->backup_root_index = 0; 1753af31f5e5SChris Mason } else { 1754af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1755af31f5e5SChris Mason } 1756af31f5e5SChris Mason } 1757af31f5e5SChris Mason 1758af31f5e5SChris Mason /* 1759af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1760af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1761af31f5e5SChris Mason * done 1762af31f5e5SChris Mason */ 1763af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1764af31f5e5SChris Mason { 1765af31f5e5SChris Mason int next_backup; 1766af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1767af31f5e5SChris Mason int last_backup; 1768af31f5e5SChris Mason 1769af31f5e5SChris Mason next_backup = info->backup_root_index; 1770af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1771af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1772af31f5e5SChris Mason 1773af31f5e5SChris Mason /* 1774af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1775af31f5e5SChris Mason * this happens only at umount 1776af31f5e5SChris Mason */ 1777af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1778af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1779af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1780af31f5e5SChris Mason next_backup = last_backup; 1781af31f5e5SChris Mason 1782af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1783af31f5e5SChris Mason 1784af31f5e5SChris Mason /* 1785af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1786af31f5e5SChris Mason * regardless of which ones we use today 1787af31f5e5SChris Mason */ 1788af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1789af31f5e5SChris Mason 1790af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1791af31f5e5SChris Mason 1792af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1793af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1794af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1795af31f5e5SChris Mason 1796af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1797af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1798af31f5e5SChris Mason 1799af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1800af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1801af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1802af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1803af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1804af31f5e5SChris Mason 1805af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1806af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1807af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1808af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1809af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1810af31f5e5SChris Mason 18117c7e82a7SChris Mason /* 18127c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 18137c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 18147c7e82a7SChris Mason */ 18157c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 18167c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 18177c7e82a7SChris Mason info->fs_root->node->start); 1818af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1819af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1820af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1821af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 18227c7e82a7SChris Mason } 1823af31f5e5SChris Mason 1824af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1825af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1826af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1827af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1828af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1829af31f5e5SChris Mason 1830af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1831af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1832af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1833af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1834af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1835af31f5e5SChris Mason 1836af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1837af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1838af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1839af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1840af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1841af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1842af31f5e5SChris Mason 1843af31f5e5SChris Mason /* 1844af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1845af31f5e5SChris Mason * for the next commit 1846af31f5e5SChris Mason */ 1847af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1848af31f5e5SChris Mason &info->super_for_commit->super_roots, 1849af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1850af31f5e5SChris Mason } 1851af31f5e5SChris Mason 1852af31f5e5SChris Mason /* 1853af31f5e5SChris Mason * this copies info out of the root backup array and back into 1854af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 1855af31f5e5SChris Mason * the array, so you send it the number of backups you've already 1856af31f5e5SChris Mason * tried and the last backup index you used. 1857af31f5e5SChris Mason * 1858af31f5e5SChris Mason * this returns -1 when it has tried all the backups 1859af31f5e5SChris Mason */ 1860af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 1861af31f5e5SChris Mason struct btrfs_super_block *super, 1862af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 1863af31f5e5SChris Mason { 1864af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1865af31f5e5SChris Mason int newest = *backup_index; 1866af31f5e5SChris Mason 1867af31f5e5SChris Mason if (*num_backups_tried == 0) { 1868af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 1869af31f5e5SChris Mason 1870af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 1871af31f5e5SChris Mason if (newest == -1) 1872af31f5e5SChris Mason return -1; 1873af31f5e5SChris Mason 1874af31f5e5SChris Mason *backup_index = newest; 1875af31f5e5SChris Mason *num_backups_tried = 1; 1876af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 1877af31f5e5SChris Mason /* we've tried all the backups, all done */ 1878af31f5e5SChris Mason return -1; 1879af31f5e5SChris Mason } else { 1880af31f5e5SChris Mason /* jump to the next oldest backup */ 1881af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 1882af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1883af31f5e5SChris Mason *backup_index = newest; 1884af31f5e5SChris Mason *num_backups_tried += 1; 1885af31f5e5SChris Mason } 1886af31f5e5SChris Mason root_backup = super->super_roots + newest; 1887af31f5e5SChris Mason 1888af31f5e5SChris Mason btrfs_set_super_generation(super, 1889af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 1890af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1891af31f5e5SChris Mason btrfs_set_super_root_level(super, 1892af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 1893af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1894af31f5e5SChris Mason 1895af31f5e5SChris Mason /* 1896af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 1897af31f5e5SChris Mason * need a fsck 1898af31f5e5SChris Mason */ 1899af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1900af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1901af31f5e5SChris Mason return 0; 1902af31f5e5SChris Mason } 1903af31f5e5SChris Mason 1904af31f5e5SChris Mason /* helper to cleanup tree roots */ 1905af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 1906af31f5e5SChris Mason { 1907af31f5e5SChris Mason free_extent_buffer(info->tree_root->node); 1908af31f5e5SChris Mason free_extent_buffer(info->tree_root->commit_root); 1909af31f5e5SChris Mason free_extent_buffer(info->dev_root->node); 1910af31f5e5SChris Mason free_extent_buffer(info->dev_root->commit_root); 1911af31f5e5SChris Mason free_extent_buffer(info->extent_root->node); 1912af31f5e5SChris Mason free_extent_buffer(info->extent_root->commit_root); 1913af31f5e5SChris Mason free_extent_buffer(info->csum_root->node); 1914af31f5e5SChris Mason free_extent_buffer(info->csum_root->commit_root); 1915bcef60f2SArne Jansen if (info->quota_root) { 1916bcef60f2SArne Jansen free_extent_buffer(info->quota_root->node); 1917bcef60f2SArne Jansen free_extent_buffer(info->quota_root->commit_root); 1918bcef60f2SArne Jansen } 1919af31f5e5SChris Mason 1920af31f5e5SChris Mason info->tree_root->node = NULL; 1921af31f5e5SChris Mason info->tree_root->commit_root = NULL; 1922af31f5e5SChris Mason info->dev_root->node = NULL; 1923af31f5e5SChris Mason info->dev_root->commit_root = NULL; 1924af31f5e5SChris Mason info->extent_root->node = NULL; 1925af31f5e5SChris Mason info->extent_root->commit_root = NULL; 1926af31f5e5SChris Mason info->csum_root->node = NULL; 1927af31f5e5SChris Mason info->csum_root->commit_root = NULL; 1928bcef60f2SArne Jansen if (info->quota_root) { 1929bcef60f2SArne Jansen info->quota_root->node = NULL; 1930bcef60f2SArne Jansen info->quota_root->commit_root = NULL; 1931bcef60f2SArne Jansen } 1932af31f5e5SChris Mason 1933af31f5e5SChris Mason if (chunk_root) { 1934af31f5e5SChris Mason free_extent_buffer(info->chunk_root->node); 1935af31f5e5SChris Mason free_extent_buffer(info->chunk_root->commit_root); 1936af31f5e5SChris Mason info->chunk_root->node = NULL; 1937af31f5e5SChris Mason info->chunk_root->commit_root = NULL; 1938af31f5e5SChris Mason } 1939af31f5e5SChris Mason } 1940af31f5e5SChris Mason 1941af31f5e5SChris Mason 1942ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 1943dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 1944dfe25020SChris Mason char *options) 1945eb60ceacSChris Mason { 1946db94535dSChris Mason u32 sectorsize; 1947db94535dSChris Mason u32 nodesize; 1948db94535dSChris Mason u32 leafsize; 1949db94535dSChris Mason u32 blocksize; 195087ee04ebSChris Mason u32 stripesize; 195184234f3aSYan Zheng u64 generation; 1952f2b636e8SJosef Bacik u64 features; 19533de4586cSChris Mason struct btrfs_key location; 1954a061fc8dSChris Mason struct buffer_head *bh; 19554d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 1956815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1957f84a8bd6SAl Viro struct btrfs_root *tree_root; 19584d34b278SIlya Dryomov struct btrfs_root *extent_root; 19594d34b278SIlya Dryomov struct btrfs_root *csum_root; 19604d34b278SIlya Dryomov struct btrfs_root *chunk_root; 19614d34b278SIlya Dryomov struct btrfs_root *dev_root; 1962bcef60f2SArne Jansen struct btrfs_root *quota_root; 1963e02119d5SChris Mason struct btrfs_root *log_tree_root; 1964eb60ceacSChris Mason int ret; 1965e58ca020SYan int err = -EINVAL; 1966af31f5e5SChris Mason int num_backups_tried = 0; 1967af31f5e5SChris Mason int backup_index = 0; 19684543df7eSChris Mason 1969f84a8bd6SAl Viro tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); 19706f07e42eSAl Viro extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); 19716f07e42eSAl Viro csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); 19726f07e42eSAl Viro chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); 19736f07e42eSAl Viro dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); 1974bcef60f2SArne Jansen quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info); 19758790d502SChris Mason 1976f84a8bd6SAl Viro if (!tree_root || !extent_root || !csum_root || 1977bcef60f2SArne Jansen !chunk_root || !dev_root || !quota_root) { 197839279cc3SChris Mason err = -ENOMEM; 197939279cc3SChris Mason goto fail; 198039279cc3SChris Mason } 198176dda93cSYan, Zheng 198276dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 198376dda93cSYan, Zheng if (ret) { 198476dda93cSYan, Zheng err = ret; 198576dda93cSYan, Zheng goto fail; 198676dda93cSYan, Zheng } 198776dda93cSYan, Zheng 198876dda93cSYan, Zheng ret = setup_bdi(fs_info, &fs_info->bdi); 198976dda93cSYan, Zheng if (ret) { 199076dda93cSYan, Zheng err = ret; 199176dda93cSYan, Zheng goto fail_srcu; 199276dda93cSYan, Zheng } 199376dda93cSYan, Zheng 199476dda93cSYan, Zheng fs_info->btree_inode = new_inode(sb); 199576dda93cSYan, Zheng if (!fs_info->btree_inode) { 199676dda93cSYan, Zheng err = -ENOMEM; 199776dda93cSYan, Zheng goto fail_bdi; 199876dda93cSYan, Zheng } 199976dda93cSYan, Zheng 2000a6591715SChris Mason mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 20011561dedaSMiao Xie 200276dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 20038fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2004facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 200524bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2006ea8c2819SChris Mason INIT_LIST_HEAD(&fs_info->delalloc_inodes); 20075a3f23d5SChris Mason INIT_LIST_HEAD(&fs_info->ordered_operations); 200811833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 20091832a6d5SChris Mason spin_lock_init(&fs_info->delalloc_lock); 2010a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 201176dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 201224bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 20134cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 20142bf64758SJosef Bacik spin_lock_init(&fs_info->free_chunk_lock); 2015f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2016f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 20177585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 201819c00ddcSChris Mason 201958176a96SJosef Bacik init_completion(&fs_info->kobj_unregister); 20200b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 20216324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2022f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 20230b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 202466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 202566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 202666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 202766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELALLOC); 202866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 202966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 203066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 203166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 203266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2033cb03c743SChris Mason atomic_set(&fs_info->nr_async_submits, 0); 2034771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 20358c8bee1dSChris Mason atomic_set(&fs_info->async_submit_draining, 0); 20360986fe9eSChris Mason atomic_set(&fs_info->nr_async_bios, 0); 20374cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2038f29021b2SJan Schmidt atomic_set(&fs_info->tree_mod_seq, 0); 2039e20d96d6SChris Mason fs_info->sb = sb; 20406f568d35SChris Mason fs_info->max_inline = 8192 * 1024; 20419ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 20424cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2043a4abeea4SJosef Bacik fs_info->trans_no_join = 0; 20442bf64758SJosef Bacik fs_info->free_chunk_space = 0; 2045f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 2046c8b97818SChris Mason 204790519d66SArne Jansen /* readahead state */ 204890519d66SArne Jansen INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); 204990519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 20502c90e5d6SChris Mason 2051b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2052b34b086cSChris Mason num_online_cpus() + 2, 8); 20530afbaf8cSChris Mason 20543eaa2885SChris Mason INIT_LIST_HEAD(&fs_info->ordered_extents); 20553eaa2885SChris Mason spin_lock_init(&fs_info->ordered_extent_lock); 205616cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 205716cdcec7SMiao Xie GFP_NOFS); 205816cdcec7SMiao Xie if (!fs_info->delayed_root) { 205916cdcec7SMiao Xie err = -ENOMEM; 206016cdcec7SMiao Xie goto fail_iput; 206116cdcec7SMiao Xie } 206216cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 20633eaa2885SChris Mason 2064a2de733cSArne Jansen mutex_init(&fs_info->scrub_lock); 2065a2de733cSArne Jansen atomic_set(&fs_info->scrubs_running, 0); 2066a2de733cSArne Jansen atomic_set(&fs_info->scrub_pause_req, 0); 2067a2de733cSArne Jansen atomic_set(&fs_info->scrubs_paused, 0); 2068a2de733cSArne Jansen atomic_set(&fs_info->scrub_cancel_req, 0); 2069a2de733cSArne Jansen init_waitqueue_head(&fs_info->scrub_pause_wait); 2070a2de733cSArne Jansen init_rwsem(&fs_info->scrub_super_lock); 2071a2de733cSArne Jansen fs_info->scrub_workers_refcnt = 0; 207221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 207321adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 207421adbd5cSStefan Behrens #endif 2075a2de733cSArne Jansen 2076c9e9f97bSIlya Dryomov spin_lock_init(&fs_info->balance_lock); 2077c9e9f97bSIlya Dryomov mutex_init(&fs_info->balance_mutex); 2078837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_running, 0); 2079837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_pause_req, 0); 2080a7e99c69SIlya Dryomov atomic_set(&fs_info->balance_cancel_req, 0); 2081c9e9f97bSIlya Dryomov fs_info->balance_ctl = NULL; 2082837d5b6eSIlya Dryomov init_waitqueue_head(&fs_info->balance_wait_q); 2083a061fc8dSChris Mason 20840afbaf8cSChris Mason sb->s_blocksize = 4096; 20850afbaf8cSChris Mason sb->s_blocksize_bits = blksize_bits(4096); 208632a88aa1SJens Axboe sb->s_bdi = &fs_info->bdi; 20870afbaf8cSChris Mason 208876dda93cSYan, Zheng fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2089bfe86848SMiklos Szeredi set_nlink(fs_info->btree_inode, 1); 20900afbaf8cSChris Mason /* 20910afbaf8cSChris Mason * we set the i_size on the btree inode to the max possible int. 20920afbaf8cSChris Mason * the real end of the address space is determined by all of 20930afbaf8cSChris Mason * the devices in the system 20940afbaf8cSChris Mason */ 20950afbaf8cSChris Mason fs_info->btree_inode->i_size = OFFSET_MAX; 2096d98237b3SChris Mason fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 209704160088SChris Mason fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; 209804160088SChris Mason 20995d4f98a2SYan Zheng RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2100d1310b2eSChris Mason extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2101f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 21020b32f4bbSJosef Bacik BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2103a8067e02SDavid Sterba extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 21040da5468fSChris Mason 2105d1310b2eSChris Mason BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2106d1310b2eSChris Mason 21070f7d52f4SChris Mason BTRFS_I(fs_info->btree_inode)->root = tree_root; 21080f7d52f4SChris Mason memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 21090f7d52f4SChris Mason sizeof(struct btrfs_key)); 211072ac3c0dSJosef Bacik set_bit(BTRFS_INODE_DUMMY, 211172ac3c0dSJosef Bacik &BTRFS_I(fs_info->btree_inode)->runtime_flags); 211222b0ebdaSChris Mason insert_inode_hash(fs_info->btree_inode); 211339279cc3SChris Mason 2114e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 21156bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2116925baeddSChris Mason 211711833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[0], 2118f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 211911833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[1], 2120f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 212111833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2122e6dcd2dcSChris Mason fs_info->do_barriers = 1; 2123f9295749SChris Mason 2124d98237b3SChris Mason 21255a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 212630ae8467SChris Mason mutex_init(&fs_info->tree_log_mutex); 2127925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2128925baeddSChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2129a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 2130a74a4b97SChris Mason mutex_init(&fs_info->volume_mutex); 2131276e680dSYan Zheng init_rwsem(&fs_info->extent_commit_sem); 2132c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 213376dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2134fa9c0d79SChris Mason 2135416ac51dSArne Jansen spin_lock_init(&fs_info->qgroup_lock); 2136416ac51dSArne Jansen fs_info->qgroup_tree = RB_ROOT; 2137416ac51dSArne Jansen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2138416ac51dSArne Jansen fs_info->qgroup_seq = 1; 2139416ac51dSArne Jansen fs_info->quota_enabled = 0; 2140416ac51dSArne Jansen fs_info->pending_quota_state = 0; 2141416ac51dSArne Jansen 2142fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2143fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2144fa9c0d79SChris Mason 21457d9eb12cSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 21463768f368SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2147bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 21484854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 21499a8dd150SChris Mason 21500b86a832SChris Mason __setup_root(4096, 4096, 4096, 4096, tree_root, 21512c90e5d6SChris Mason fs_info, BTRFS_ROOT_TREE_OBJECTID); 21527eccb903SChris Mason 21533c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 2154a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 215520b45077SDave Young if (!bh) { 215620b45077SDave Young err = -EINVAL; 215716cdcec7SMiao Xie goto fail_alloc; 215820b45077SDave Young } 215939279cc3SChris Mason 21606c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 21616c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 21626c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2163a061fc8dSChris Mason brelse(bh); 21645f39d397SChris Mason 21656c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 21660b86a832SChris Mason 21676c41761fSDavid Sterba disk_super = fs_info->super_copy; 21680f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 216916cdcec7SMiao Xie goto fail_alloc; 21700f7d52f4SChris Mason 2171acce952bSliubo /* check FS state, whether FS is broken. */ 2172acce952bSliubo fs_info->fs_state |= btrfs_super_flags(disk_super); 2173acce952bSliubo 2174fcd1f065SDavid Sterba ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 2175fcd1f065SDavid Sterba if (ret) { 2176fcd1f065SDavid Sterba printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); 2177fcd1f065SDavid Sterba err = ret; 2178fcd1f065SDavid Sterba goto fail_alloc; 2179fcd1f065SDavid Sterba } 2180acce952bSliubo 218175e7cb7fSLiu Bo /* 2182af31f5e5SChris Mason * run through our array of backup supers and setup 2183af31f5e5SChris Mason * our ring pointer to the oldest one 2184af31f5e5SChris Mason */ 2185af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2186af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2187af31f5e5SChris Mason 2188af31f5e5SChris Mason /* 218975e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 219075e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 219175e7cb7fSLiu Bo */ 219275e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 219375e7cb7fSLiu Bo 21942b82032cSYan Zheng ret = btrfs_parse_options(tree_root, options); 21952b82032cSYan Zheng if (ret) { 21962b82032cSYan Zheng err = ret; 219716cdcec7SMiao Xie goto fail_alloc; 21982b82032cSYan Zheng } 2199dfe25020SChris Mason 2200f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2201f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2202f2b636e8SJosef Bacik if (features) { 2203f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount because of " 2204f2b636e8SJosef Bacik "unsupported optional features (%Lx).\n", 220521380931SJoel Becker (unsigned long long)features); 2206f2b636e8SJosef Bacik err = -EINVAL; 220716cdcec7SMiao Xie goto fail_alloc; 2208f2b636e8SJosef Bacik } 2209f2b636e8SJosef Bacik 2210727011e0SChris Mason if (btrfs_super_leafsize(disk_super) != 2211727011e0SChris Mason btrfs_super_nodesize(disk_super)) { 2212727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2213727011e0SChris Mason "blocksizes don't match. node %d leaf %d\n", 2214727011e0SChris Mason btrfs_super_nodesize(disk_super), 2215727011e0SChris Mason btrfs_super_leafsize(disk_super)); 2216727011e0SChris Mason err = -EINVAL; 2217727011e0SChris Mason goto fail_alloc; 2218727011e0SChris Mason } 2219727011e0SChris Mason if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { 2220727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2221727011e0SChris Mason "blocksize (%d) was too large\n", 2222727011e0SChris Mason btrfs_super_leafsize(disk_super)); 2223727011e0SChris Mason err = -EINVAL; 2224727011e0SChris Mason goto fail_alloc; 2225727011e0SChris Mason } 2226727011e0SChris Mason 22275d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 22285d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 222969e380d1SLi Zefan if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2230a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2231727011e0SChris Mason 2232727011e0SChris Mason /* 2233727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2234727011e0SChris Mason * they are bigger than the page size 2235727011e0SChris Mason */ 2236727011e0SChris Mason if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { 2237727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2238727011e0SChris Mason printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); 2239727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2240727011e0SChris Mason } 2241727011e0SChris Mason 2242bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2243bc3f116fSChris Mason leafsize = btrfs_super_leafsize(disk_super); 2244bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2245bc3f116fSChris Mason stripesize = btrfs_super_stripesize(disk_super); 2246bc3f116fSChris Mason 2247bc3f116fSChris Mason /* 2248bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2249bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2250bc3f116fSChris Mason */ 2251bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2252bc3f116fSChris Mason (sectorsize != leafsize)) { 2253bc3f116fSChris Mason printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " 2254bc3f116fSChris Mason "are not allowed for mixed block groups on %s\n", 2255bc3f116fSChris Mason sb->s_id); 2256bc3f116fSChris Mason goto fail_alloc; 2257bc3f116fSChris Mason } 2258bc3f116fSChris Mason 22595d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 22605d4f98a2SYan Zheng 2261f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2262f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2263f2b636e8SJosef Bacik if (!(sb->s_flags & MS_RDONLY) && features) { 2264f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2265f2b636e8SJosef Bacik "unsupported option features (%Lx).\n", 226621380931SJoel Becker (unsigned long long)features); 2267f2b636e8SJosef Bacik err = -EINVAL; 226816cdcec7SMiao Xie goto fail_alloc; 2269f2b636e8SJosef Bacik } 227061d92c32SChris Mason 227161d92c32SChris Mason btrfs_init_workers(&fs_info->generic_worker, 227261d92c32SChris Mason "genwork", 1, NULL); 227361d92c32SChris Mason 22745443be45SChris Mason btrfs_init_workers(&fs_info->workers, "worker", 227561d92c32SChris Mason fs_info->thread_pool_size, 227661d92c32SChris Mason &fs_info->generic_worker); 2277c8b97818SChris Mason 2278771ed689SChris Mason btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", 227961d92c32SChris Mason fs_info->thread_pool_size, 228061d92c32SChris Mason &fs_info->generic_worker); 2281771ed689SChris Mason 22828ccf6f19SMiao Xie btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc", 22838ccf6f19SMiao Xie fs_info->thread_pool_size, 22848ccf6f19SMiao Xie &fs_info->generic_worker); 22858ccf6f19SMiao Xie 22865443be45SChris Mason btrfs_init_workers(&fs_info->submit_workers, "submit", 2287b720d209SChris Mason min_t(u64, fs_devices->num_devices, 228861d92c32SChris Mason fs_info->thread_pool_size), 228961d92c32SChris Mason &fs_info->generic_worker); 229061b49440SChris Mason 2291bab39bf9SJosef Bacik btrfs_init_workers(&fs_info->caching_workers, "cache", 2292bab39bf9SJosef Bacik 2, &fs_info->generic_worker); 2293bab39bf9SJosef Bacik 229461b49440SChris Mason /* a higher idle thresh on the submit workers makes it much more 229561b49440SChris Mason * likely that bios will be send down in a sane order to the 229661b49440SChris Mason * devices 229761b49440SChris Mason */ 229861b49440SChris Mason fs_info->submit_workers.idle_thresh = 64; 229953863232SChris Mason 2300771ed689SChris Mason fs_info->workers.idle_thresh = 16; 23014a69a410SChris Mason fs_info->workers.ordered = 1; 230261b49440SChris Mason 2303771ed689SChris Mason fs_info->delalloc_workers.idle_thresh = 2; 2304771ed689SChris Mason fs_info->delalloc_workers.ordered = 1; 2305771ed689SChris Mason 230661d92c32SChris Mason btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, 230761d92c32SChris Mason &fs_info->generic_worker); 23085443be45SChris Mason btrfs_init_workers(&fs_info->endio_workers, "endio", 230961d92c32SChris Mason fs_info->thread_pool_size, 231061d92c32SChris Mason &fs_info->generic_worker); 2311d20f7043SChris Mason btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", 231261d92c32SChris Mason fs_info->thread_pool_size, 231361d92c32SChris Mason &fs_info->generic_worker); 2314cad321adSChris Mason btrfs_init_workers(&fs_info->endio_meta_write_workers, 231561d92c32SChris Mason "endio-meta-write", fs_info->thread_pool_size, 231661d92c32SChris Mason &fs_info->generic_worker); 23175443be45SChris Mason btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", 231861d92c32SChris Mason fs_info->thread_pool_size, 231961d92c32SChris Mason &fs_info->generic_worker); 23200cb59c99SJosef Bacik btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", 23210cb59c99SJosef Bacik 1, &fs_info->generic_worker); 232216cdcec7SMiao Xie btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", 232316cdcec7SMiao Xie fs_info->thread_pool_size, 232416cdcec7SMiao Xie &fs_info->generic_worker); 232590519d66SArne Jansen btrfs_init_workers(&fs_info->readahead_workers, "readahead", 232690519d66SArne Jansen fs_info->thread_pool_size, 232790519d66SArne Jansen &fs_info->generic_worker); 232861b49440SChris Mason 232961b49440SChris Mason /* 233061b49440SChris Mason * endios are largely parallel and should have a very 233161b49440SChris Mason * low idle thresh 233261b49440SChris Mason */ 233361b49440SChris Mason fs_info->endio_workers.idle_thresh = 4; 2334b51912c9SChris Mason fs_info->endio_meta_workers.idle_thresh = 4; 2335b51912c9SChris Mason 23369042846bSChris Mason fs_info->endio_write_workers.idle_thresh = 2; 23379042846bSChris Mason fs_info->endio_meta_write_workers.idle_thresh = 2; 233890519d66SArne Jansen fs_info->readahead_workers.idle_thresh = 2; 23399042846bSChris Mason 23400dc3b84aSJosef Bacik /* 23410dc3b84aSJosef Bacik * btrfs_start_workers can really only fail because of ENOMEM so just 23420dc3b84aSJosef Bacik * return -ENOMEM if any of these fail. 23430dc3b84aSJosef Bacik */ 23440dc3b84aSJosef Bacik ret = btrfs_start_workers(&fs_info->workers); 23450dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->generic_worker); 23460dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->submit_workers); 23470dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->delalloc_workers); 23480dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->fixup_workers); 23490dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->endio_workers); 23500dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->endio_meta_workers); 23510dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); 23520dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->endio_write_workers); 23530dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); 23540dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->delayed_workers); 23550dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->caching_workers); 23560dc3b84aSJosef Bacik ret |= btrfs_start_workers(&fs_info->readahead_workers); 23578ccf6f19SMiao Xie ret |= btrfs_start_workers(&fs_info->flush_workers); 23580dc3b84aSJosef Bacik if (ret) { 2359fed425c7SIlya Dryomov err = -ENOMEM; 23600dc3b84aSJosef Bacik goto fail_sb_buffer; 23610dc3b84aSJosef Bacik } 23624543df7eSChris Mason 23634575c9ccSChris Mason fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2364c8b97818SChris Mason fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2365c8b97818SChris Mason 4 * 1024 * 1024 / PAGE_CACHE_SIZE); 23664575c9ccSChris Mason 2367db94535dSChris Mason tree_root->nodesize = nodesize; 2368db94535dSChris Mason tree_root->leafsize = leafsize; 2369db94535dSChris Mason tree_root->sectorsize = sectorsize; 237087ee04ebSChris Mason tree_root->stripesize = stripesize; 2371a061fc8dSChris Mason 2372a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2373a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2374db94535dSChris Mason 237539279cc3SChris Mason if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, 237639279cc3SChris Mason sizeof(disk_super->magic))) { 2377d397712bSChris Mason printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); 237839279cc3SChris Mason goto fail_sb_buffer; 237939279cc3SChris Mason } 238019c00ddcSChris Mason 23818d082fb7SLiu Bo if (sectorsize != PAGE_SIZE) { 23828d082fb7SLiu Bo printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) " 23838d082fb7SLiu Bo "found on %s\n", (unsigned long)sectorsize, sb->s_id); 2384941b2ddfSKeith Mannthey goto fail_sb_buffer; 2385941b2ddfSKeith Mannthey } 2386941b2ddfSKeith Mannthey 2387925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 2388e4404d6eSYan Zheng ret = btrfs_read_sys_array(tree_root); 2389925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 239084eed90fSChris Mason if (ret) { 2391d397712bSChris Mason printk(KERN_WARNING "btrfs: failed to read the system " 2392d397712bSChris Mason "array on %s\n", sb->s_id); 23935d4f98a2SYan Zheng goto fail_sb_buffer; 239484eed90fSChris Mason } 23950b86a832SChris Mason 23960b86a832SChris Mason blocksize = btrfs_level_size(tree_root, 23970b86a832SChris Mason btrfs_super_chunk_root_level(disk_super)); 239884234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 23990b86a832SChris Mason 24000b86a832SChris Mason __setup_root(nodesize, leafsize, sectorsize, stripesize, 24010b86a832SChris Mason chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 24020b86a832SChris Mason 24030b86a832SChris Mason chunk_root->node = read_tree_block(chunk_root, 24040b86a832SChris Mason btrfs_super_chunk_root(disk_super), 240584234f3aSYan Zheng blocksize, generation); 240679787eaaSJeff Mahoney BUG_ON(!chunk_root->node); /* -ENOMEM */ 240783121942SDavid Woodhouse if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { 240883121942SDavid Woodhouse printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", 240983121942SDavid Woodhouse sb->s_id); 2410af31f5e5SChris Mason goto fail_tree_roots; 241183121942SDavid Woodhouse } 24125d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 24135d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 24140b86a832SChris Mason 2415e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2416e17cade2SChris Mason (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), 2417e17cade2SChris Mason BTRFS_UUID_SIZE); 2418e17cade2SChris Mason 24190b86a832SChris Mason ret = btrfs_read_chunk_tree(chunk_root); 24202b82032cSYan Zheng if (ret) { 2421d397712bSChris Mason printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", 2422d397712bSChris Mason sb->s_id); 2423af31f5e5SChris Mason goto fail_tree_roots; 24242b82032cSYan Zheng } 24250b86a832SChris Mason 2426dfe25020SChris Mason btrfs_close_extra_devices(fs_devices); 2427dfe25020SChris Mason 2428a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 2429a6b0d5c8SChris Mason printk(KERN_CRIT "btrfs: failed to read devices on %s\n", 2430a6b0d5c8SChris Mason sb->s_id); 2431a6b0d5c8SChris Mason goto fail_tree_roots; 2432a6b0d5c8SChris Mason } 2433a6b0d5c8SChris Mason 2434af31f5e5SChris Mason retry_root_backup: 2435db94535dSChris Mason blocksize = btrfs_level_size(tree_root, 2436db94535dSChris Mason btrfs_super_root_level(disk_super)); 243784234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 24380b86a832SChris Mason 2439e20d96d6SChris Mason tree_root->node = read_tree_block(tree_root, 2440db94535dSChris Mason btrfs_super_root(disk_super), 244184234f3aSYan Zheng blocksize, generation); 2442af31f5e5SChris Mason if (!tree_root->node || 2443af31f5e5SChris Mason !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { 244483121942SDavid Woodhouse printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", 244583121942SDavid Woodhouse sb->s_id); 2446af31f5e5SChris Mason 2447af31f5e5SChris Mason goto recovery_tree_root; 244883121942SDavid Woodhouse } 2449af31f5e5SChris Mason 24505d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 24515d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 2452db94535dSChris Mason 2453db94535dSChris Mason ret = find_and_setup_root(tree_root, fs_info, 2454e20d96d6SChris Mason BTRFS_EXTENT_TREE_OBJECTID, extent_root); 24550b86a832SChris Mason if (ret) 2456af31f5e5SChris Mason goto recovery_tree_root; 24570b86a832SChris Mason extent_root->track_dirty = 1; 24580b86a832SChris Mason 24590b86a832SChris Mason ret = find_and_setup_root(tree_root, fs_info, 24600b86a832SChris Mason BTRFS_DEV_TREE_OBJECTID, dev_root); 24610b86a832SChris Mason if (ret) 2462af31f5e5SChris Mason goto recovery_tree_root; 24635d4f98a2SYan Zheng dev_root->track_dirty = 1; 24643768f368SChris Mason 2465d20f7043SChris Mason ret = find_and_setup_root(tree_root, fs_info, 2466d20f7043SChris Mason BTRFS_CSUM_TREE_OBJECTID, csum_root); 2467d20f7043SChris Mason if (ret) 2468af31f5e5SChris Mason goto recovery_tree_root; 2469d20f7043SChris Mason csum_root->track_dirty = 1; 2470d20f7043SChris Mason 2471bcef60f2SArne Jansen ret = find_and_setup_root(tree_root, fs_info, 2472bcef60f2SArne Jansen BTRFS_QUOTA_TREE_OBJECTID, quota_root); 2473bcef60f2SArne Jansen if (ret) { 2474bcef60f2SArne Jansen kfree(quota_root); 2475bcef60f2SArne Jansen quota_root = fs_info->quota_root = NULL; 2476bcef60f2SArne Jansen } else { 2477bcef60f2SArne Jansen quota_root->track_dirty = 1; 2478bcef60f2SArne Jansen fs_info->quota_enabled = 1; 2479bcef60f2SArne Jansen fs_info->pending_quota_state = 1; 2480bcef60f2SArne Jansen } 2481bcef60f2SArne Jansen 24828929ecfaSYan, Zheng fs_info->generation = generation; 24838929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 24848929ecfaSYan, Zheng 248568310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 248668310a5eSIlya Dryomov if (ret) { 248768310a5eSIlya Dryomov printk(KERN_WARNING "btrfs: failed to recover balance\n"); 248868310a5eSIlya Dryomov goto fail_block_groups; 248968310a5eSIlya Dryomov } 249068310a5eSIlya Dryomov 2491733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2492733f4fbbSStefan Behrens if (ret) { 2493733f4fbbSStefan Behrens printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n", 2494733f4fbbSStefan Behrens ret); 2495733f4fbbSStefan Behrens goto fail_block_groups; 2496733f4fbbSStefan Behrens } 2497733f4fbbSStefan Behrens 2498c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2499c59021f8Sliubo if (ret) { 2500c59021f8Sliubo printk(KERN_ERR "Failed to initial space info: %d\n", ret); 2501c59021f8Sliubo goto fail_block_groups; 2502c59021f8Sliubo } 2503c59021f8Sliubo 25041b1d1f66SJosef Bacik ret = btrfs_read_block_groups(extent_root); 25051b1d1f66SJosef Bacik if (ret) { 25061b1d1f66SJosef Bacik printk(KERN_ERR "Failed to read block groups: %d\n", ret); 25071b1d1f66SJosef Bacik goto fail_block_groups; 25081b1d1f66SJosef Bacik } 25095af3e8ccSStefan Behrens fs_info->num_tolerated_disk_barrier_failures = 25105af3e8ccSStefan Behrens btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2511292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 2512292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 2513292fd7fcSStefan Behrens !(sb->s_flags & MS_RDONLY)) { 2514292fd7fcSStefan Behrens printk(KERN_WARNING 2515292fd7fcSStefan Behrens "Btrfs: too many missing devices, writeable mount is not allowed\n"); 2516292fd7fcSStefan Behrens goto fail_block_groups; 2517292fd7fcSStefan Behrens } 25189078a3e1SChris Mason 2519a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2520a74a4b97SChris Mason "btrfs-cleaner"); 252157506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 25221b1d1f66SJosef Bacik goto fail_block_groups; 2523a74a4b97SChris Mason 2524a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2525a74a4b97SChris Mason tree_root, 2526a74a4b97SChris Mason "btrfs-transaction"); 252757506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 25283f157a2fSChris Mason goto fail_cleaner; 2529a74a4b97SChris Mason 2530c289811cSChris Mason if (!btrfs_test_opt(tree_root, SSD) && 2531c289811cSChris Mason !btrfs_test_opt(tree_root, NOSSD) && 2532c289811cSChris Mason !fs_info->fs_devices->rotating) { 2533c289811cSChris Mason printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " 2534c289811cSChris Mason "mode\n"); 2535c289811cSChris Mason btrfs_set_opt(fs_info->mount_opt, SSD); 2536c289811cSChris Mason } 2537c289811cSChris Mason 253821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 253921adbd5cSStefan Behrens if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 254021adbd5cSStefan Behrens ret = btrfsic_mount(tree_root, fs_devices, 254121adbd5cSStefan Behrens btrfs_test_opt(tree_root, 254221adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 254321adbd5cSStefan Behrens 1 : 0, 254421adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 254521adbd5cSStefan Behrens if (ret) 254621adbd5cSStefan Behrens printk(KERN_WARNING "btrfs: failed to initialize" 254721adbd5cSStefan Behrens " integrity check module %s\n", sb->s_id); 254821adbd5cSStefan Behrens } 254921adbd5cSStefan Behrens #endif 2550bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 2551bcef60f2SArne Jansen if (ret) 2552bcef60f2SArne Jansen goto fail_trans_kthread; 255321adbd5cSStefan Behrens 2554acce952bSliubo /* do not make disk changes in broken FS */ 255568ce9682SStefan Behrens if (btrfs_super_log_root(disk_super) != 0) { 2556e02119d5SChris Mason u64 bytenr = btrfs_super_log_root(disk_super); 2557d18a2c44SChris Mason 25587c2ca468SChris Mason if (fs_devices->rw_devices == 0) { 2559d397712bSChris Mason printk(KERN_WARNING "Btrfs log replay required " 2560d397712bSChris Mason "on RO media\n"); 25617c2ca468SChris Mason err = -EIO; 2562bcef60f2SArne Jansen goto fail_qgroup; 25637c2ca468SChris Mason } 2564e02119d5SChris Mason blocksize = 2565e02119d5SChris Mason btrfs_level_size(tree_root, 2566e02119d5SChris Mason btrfs_super_log_root_level(disk_super)); 2567e02119d5SChris Mason 25686f07e42eSAl Viro log_tree_root = btrfs_alloc_root(fs_info); 2569676e4c86SDan Carpenter if (!log_tree_root) { 2570676e4c86SDan Carpenter err = -ENOMEM; 2571bcef60f2SArne Jansen goto fail_qgroup; 2572676e4c86SDan Carpenter } 2573e02119d5SChris Mason 2574e02119d5SChris Mason __setup_root(nodesize, leafsize, sectorsize, stripesize, 2575e02119d5SChris Mason log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 2576e02119d5SChris Mason 2577e02119d5SChris Mason log_tree_root->node = read_tree_block(tree_root, bytenr, 257884234f3aSYan Zheng blocksize, 257984234f3aSYan Zheng generation + 1); 258079787eaaSJeff Mahoney /* returns with log_tree_root freed on success */ 2581e02119d5SChris Mason ret = btrfs_recover_log_trees(log_tree_root); 258279787eaaSJeff Mahoney if (ret) { 258379787eaaSJeff Mahoney btrfs_error(tree_root->fs_info, ret, 258479787eaaSJeff Mahoney "Failed to recover log tree"); 258579787eaaSJeff Mahoney free_extent_buffer(log_tree_root->node); 258679787eaaSJeff Mahoney kfree(log_tree_root); 258779787eaaSJeff Mahoney goto fail_trans_kthread; 258879787eaaSJeff Mahoney } 2589e556ce2cSYan Zheng 2590e556ce2cSYan Zheng if (sb->s_flags & MS_RDONLY) { 2591e556ce2cSYan Zheng ret = btrfs_commit_super(tree_root); 259279787eaaSJeff Mahoney if (ret) 259379787eaaSJeff Mahoney goto fail_trans_kthread; 2594e556ce2cSYan Zheng } 2595e02119d5SChris Mason } 25961a40e23bSZheng Yan 259776dda93cSYan, Zheng ret = btrfs_find_orphan_roots(tree_root); 259879787eaaSJeff Mahoney if (ret) 259979787eaaSJeff Mahoney goto fail_trans_kthread; 260076dda93cSYan, Zheng 26017c2ca468SChris Mason if (!(sb->s_flags & MS_RDONLY)) { 2602d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 260344c44af2SIlya Dryomov if (ret) 260444c44af2SIlya Dryomov goto fail_trans_kthread; 2605d68fc57bSYan, Zheng 26065d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 2607d7ce5843SMiao Xie if (ret < 0) { 2608d7ce5843SMiao Xie printk(KERN_WARNING 2609d7ce5843SMiao Xie "btrfs: failed to recover relocation\n"); 2610d7ce5843SMiao Xie err = -EINVAL; 2611bcef60f2SArne Jansen goto fail_qgroup; 2612d7ce5843SMiao Xie } 26137c2ca468SChris Mason } 26141a40e23bSZheng Yan 26153de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 26163de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 26173de4586cSChris Mason location.offset = (u64)-1; 26183de4586cSChris Mason 26193de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 26203de4586cSChris Mason if (!fs_info->fs_root) 2621bcef60f2SArne Jansen goto fail_qgroup; 26223140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 26233140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 2624bcef60f2SArne Jansen goto fail_qgroup; 26253140c9a3SDan Carpenter } 2626c289811cSChris Mason 26272b6ba629SIlya Dryomov if (sb->s_flags & MS_RDONLY) 26282b6ba629SIlya Dryomov return 0; 26292b6ba629SIlya Dryomov 2630e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 26312b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 26322b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 26332b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 26342b6ba629SIlya Dryomov close_ctree(tree_root); 26352b6ba629SIlya Dryomov return ret; 26362b6ba629SIlya Dryomov } 2637e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 263859641015SIlya Dryomov 26392b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 26402b6ba629SIlya Dryomov if (ret) { 26412b6ba629SIlya Dryomov printk(KERN_WARNING "btrfs: failed to resume balance\n"); 264266b4ffd1SJosef Bacik close_ctree(tree_root); 26432b6ba629SIlya Dryomov return ret; 2644e3acc2a6SJosef Bacik } 2645e3acc2a6SJosef Bacik 2646ad2b2c80SAl Viro return 0; 264739279cc3SChris Mason 2648bcef60f2SArne Jansen fail_qgroup: 2649bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 26507c2ca468SChris Mason fail_trans_kthread: 26517c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 26523f157a2fSChris Mason fail_cleaner: 2653a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 26547c2ca468SChris Mason 26557c2ca468SChris Mason /* 26567c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 26577c2ca468SChris Mason * kthreads 26587c2ca468SChris Mason */ 26597c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 26607c2ca468SChris Mason invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 26617c2ca468SChris Mason 26621b1d1f66SJosef Bacik fail_block_groups: 26631b1d1f66SJosef Bacik btrfs_free_block_groups(fs_info); 2664af31f5e5SChris Mason 2665af31f5e5SChris Mason fail_tree_roots: 2666af31f5e5SChris Mason free_root_pointers(fs_info, 1); 2667af31f5e5SChris Mason 266839279cc3SChris Mason fail_sb_buffer: 266961d92c32SChris Mason btrfs_stop_workers(&fs_info->generic_worker); 2670306c8b68SChris Mason btrfs_stop_workers(&fs_info->readahead_workers); 2671247e743cSChris Mason btrfs_stop_workers(&fs_info->fixup_workers); 2672771ed689SChris Mason btrfs_stop_workers(&fs_info->delalloc_workers); 26738b712842SChris Mason btrfs_stop_workers(&fs_info->workers); 26748b712842SChris Mason btrfs_stop_workers(&fs_info->endio_workers); 2675d20f7043SChris Mason btrfs_stop_workers(&fs_info->endio_meta_workers); 2676cad321adSChris Mason btrfs_stop_workers(&fs_info->endio_meta_write_workers); 2677e6dcd2dcSChris Mason btrfs_stop_workers(&fs_info->endio_write_workers); 26780cb59c99SJosef Bacik btrfs_stop_workers(&fs_info->endio_freespace_worker); 26791cc127b5SChris Mason btrfs_stop_workers(&fs_info->submit_workers); 268016cdcec7SMiao Xie btrfs_stop_workers(&fs_info->delayed_workers); 2681bab39bf9SJosef Bacik btrfs_stop_workers(&fs_info->caching_workers); 26828ccf6f19SMiao Xie btrfs_stop_workers(&fs_info->flush_workers); 268316cdcec7SMiao Xie fail_alloc: 26844543df7eSChris Mason fail_iput: 2685586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 2686586e46e2SIlya Dryomov 26877c2ca468SChris Mason invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 26884543df7eSChris Mason iput(fs_info->btree_inode); 2689ad081f14SJens Axboe fail_bdi: 26907e662854SQinghuang Feng bdi_destroy(&fs_info->bdi); 269176dda93cSYan, Zheng fail_srcu: 269276dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 26937e662854SQinghuang Feng fail: 2694586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 2695ad2b2c80SAl Viro return err; 2696af31f5e5SChris Mason 2697af31f5e5SChris Mason recovery_tree_root: 2698af31f5e5SChris Mason if (!btrfs_test_opt(tree_root, RECOVERY)) 2699af31f5e5SChris Mason goto fail_tree_roots; 2700af31f5e5SChris Mason 2701af31f5e5SChris Mason free_root_pointers(fs_info, 0); 2702af31f5e5SChris Mason 2703af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 2704af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 2705af31f5e5SChris Mason 2706af31f5e5SChris Mason /* we can't trust the free space cache either */ 2707af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2708af31f5e5SChris Mason 2709af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 2710af31f5e5SChris Mason &num_backups_tried, &backup_index); 2711af31f5e5SChris Mason if (ret == -1) 2712af31f5e5SChris Mason goto fail_block_groups; 2713af31f5e5SChris Mason goto retry_root_backup; 2714eb60ceacSChris Mason } 2715eb60ceacSChris Mason 2716f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 2717f2984462SChris Mason { 2718f2984462SChris Mason if (uptodate) { 2719f2984462SChris Mason set_buffer_uptodate(bh); 2720f2984462SChris Mason } else { 2721442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 2722442a4f63SStefan Behrens bh->b_private; 2723442a4f63SStefan Behrens 2724606686eeSJosef Bacik printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to " 2725606686eeSJosef Bacik "I/O error on %s\n", 2726606686eeSJosef Bacik rcu_str_deref(device->name)); 27271259ab75SChris Mason /* note, we dont' set_buffer_write_io_error because we have 27281259ab75SChris Mason * our own ways of dealing with the IO errors 27291259ab75SChris Mason */ 2730f2984462SChris Mason clear_buffer_uptodate(bh); 2731442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 2732f2984462SChris Mason } 2733f2984462SChris Mason unlock_buffer(bh); 2734f2984462SChris Mason put_bh(bh); 2735f2984462SChris Mason } 2736f2984462SChris Mason 2737a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 2738a512bbf8SYan Zheng { 2739a512bbf8SYan Zheng struct buffer_head *bh; 2740a512bbf8SYan Zheng struct buffer_head *latest = NULL; 2741a512bbf8SYan Zheng struct btrfs_super_block *super; 2742a512bbf8SYan Zheng int i; 2743a512bbf8SYan Zheng u64 transid = 0; 2744a512bbf8SYan Zheng u64 bytenr; 2745a512bbf8SYan Zheng 2746a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 2747a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 2748a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 2749a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 2750a512bbf8SYan Zheng */ 2751a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 2752a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 2753a512bbf8SYan Zheng if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) 2754a512bbf8SYan Zheng break; 2755a512bbf8SYan Zheng bh = __bread(bdev, bytenr / 4096, 4096); 2756a512bbf8SYan Zheng if (!bh) 2757a512bbf8SYan Zheng continue; 2758a512bbf8SYan Zheng 2759a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 2760a512bbf8SYan Zheng if (btrfs_super_bytenr(super) != bytenr || 2761a512bbf8SYan Zheng strncmp((char *)(&super->magic), BTRFS_MAGIC, 2762a512bbf8SYan Zheng sizeof(super->magic))) { 2763a512bbf8SYan Zheng brelse(bh); 2764a512bbf8SYan Zheng continue; 2765a512bbf8SYan Zheng } 2766a512bbf8SYan Zheng 2767a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 2768a512bbf8SYan Zheng brelse(latest); 2769a512bbf8SYan Zheng latest = bh; 2770a512bbf8SYan Zheng transid = btrfs_super_generation(super); 2771a512bbf8SYan Zheng } else { 2772a512bbf8SYan Zheng brelse(bh); 2773a512bbf8SYan Zheng } 2774a512bbf8SYan Zheng } 2775a512bbf8SYan Zheng return latest; 2776a512bbf8SYan Zheng } 2777a512bbf8SYan Zheng 27784eedeb75SHisashi Hifumi /* 27794eedeb75SHisashi Hifumi * this should be called twice, once with wait == 0 and 27804eedeb75SHisashi Hifumi * once with wait == 1. When wait == 0 is done, all the buffer heads 27814eedeb75SHisashi Hifumi * we write are pinned. 27824eedeb75SHisashi Hifumi * 27834eedeb75SHisashi Hifumi * They are released when wait == 1 is done. 27844eedeb75SHisashi Hifumi * max_mirrors must be the same for both runs, and it indicates how 27854eedeb75SHisashi Hifumi * many supers on this one device should be written. 27864eedeb75SHisashi Hifumi * 27874eedeb75SHisashi Hifumi * max_mirrors == 0 means to write them all. 27884eedeb75SHisashi Hifumi */ 2789a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 2790a512bbf8SYan Zheng struct btrfs_super_block *sb, 2791a512bbf8SYan Zheng int do_barriers, int wait, int max_mirrors) 2792a512bbf8SYan Zheng { 2793a512bbf8SYan Zheng struct buffer_head *bh; 2794a512bbf8SYan Zheng int i; 2795a512bbf8SYan Zheng int ret; 2796a512bbf8SYan Zheng int errors = 0; 2797a512bbf8SYan Zheng u32 crc; 2798a512bbf8SYan Zheng u64 bytenr; 2799a512bbf8SYan Zheng 2800a512bbf8SYan Zheng if (max_mirrors == 0) 2801a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 2802a512bbf8SYan Zheng 2803a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 2804a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 2805a512bbf8SYan Zheng if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) 2806a512bbf8SYan Zheng break; 2807a512bbf8SYan Zheng 2808a512bbf8SYan Zheng if (wait) { 2809a512bbf8SYan Zheng bh = __find_get_block(device->bdev, bytenr / 4096, 2810a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 2811a512bbf8SYan Zheng BUG_ON(!bh); 2812a512bbf8SYan Zheng wait_on_buffer(bh); 28134eedeb75SHisashi Hifumi if (!buffer_uptodate(bh)) 28144eedeb75SHisashi Hifumi errors++; 28154eedeb75SHisashi Hifumi 28164eedeb75SHisashi Hifumi /* drop our reference */ 28174eedeb75SHisashi Hifumi brelse(bh); 28184eedeb75SHisashi Hifumi 28194eedeb75SHisashi Hifumi /* drop the reference from the wait == 0 run */ 2820a512bbf8SYan Zheng brelse(bh); 2821a512bbf8SYan Zheng continue; 2822a512bbf8SYan Zheng } else { 2823a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 2824a512bbf8SYan Zheng 2825a512bbf8SYan Zheng crc = ~(u32)0; 2826a512bbf8SYan Zheng crc = btrfs_csum_data(NULL, (char *)sb + 2827a512bbf8SYan Zheng BTRFS_CSUM_SIZE, crc, 2828a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE - 2829a512bbf8SYan Zheng BTRFS_CSUM_SIZE); 2830a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 2831a512bbf8SYan Zheng 28324eedeb75SHisashi Hifumi /* 28334eedeb75SHisashi Hifumi * one reference for us, and we leave it for the 28344eedeb75SHisashi Hifumi * caller 28354eedeb75SHisashi Hifumi */ 2836a512bbf8SYan Zheng bh = __getblk(device->bdev, bytenr / 4096, 2837a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 2838a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 2839a512bbf8SYan Zheng 28404eedeb75SHisashi Hifumi /* one reference for submit_bh */ 2841a512bbf8SYan Zheng get_bh(bh); 28424eedeb75SHisashi Hifumi 28434eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 2844a512bbf8SYan Zheng lock_buffer(bh); 2845a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 2846442a4f63SStefan Behrens bh->b_private = device; 2847a512bbf8SYan Zheng } 2848a512bbf8SYan Zheng 2849387125fcSChris Mason /* 2850387125fcSChris Mason * we fua the first super. The others we allow 2851387125fcSChris Mason * to go down lazy. 2852387125fcSChris Mason */ 285321adbd5cSStefan Behrens ret = btrfsic_submit_bh(WRITE_FUA, bh); 28544eedeb75SHisashi Hifumi if (ret) 2855a512bbf8SYan Zheng errors++; 2856a512bbf8SYan Zheng } 2857a512bbf8SYan Zheng return errors < i ? 0 : -1; 2858a512bbf8SYan Zheng } 2859a512bbf8SYan Zheng 2860387125fcSChris Mason /* 2861387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 2862387125fcSChris Mason * for the barrier when it is done 2863387125fcSChris Mason */ 2864387125fcSChris Mason static void btrfs_end_empty_barrier(struct bio *bio, int err) 2865387125fcSChris Mason { 2866387125fcSChris Mason if (err) { 2867387125fcSChris Mason if (err == -EOPNOTSUPP) 2868387125fcSChris Mason set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); 2869387125fcSChris Mason clear_bit(BIO_UPTODATE, &bio->bi_flags); 2870387125fcSChris Mason } 2871387125fcSChris Mason if (bio->bi_private) 2872387125fcSChris Mason complete(bio->bi_private); 2873387125fcSChris Mason bio_put(bio); 2874387125fcSChris Mason } 2875387125fcSChris Mason 2876387125fcSChris Mason /* 2877387125fcSChris Mason * trigger flushes for one the devices. If you pass wait == 0, the flushes are 2878387125fcSChris Mason * sent down. With wait == 1, it waits for the previous flush. 2879387125fcSChris Mason * 2880387125fcSChris Mason * any device where the flush fails with eopnotsupp are flagged as not-barrier 2881387125fcSChris Mason * capable 2882387125fcSChris Mason */ 2883387125fcSChris Mason static int write_dev_flush(struct btrfs_device *device, int wait) 2884387125fcSChris Mason { 2885387125fcSChris Mason struct bio *bio; 2886387125fcSChris Mason int ret = 0; 2887387125fcSChris Mason 2888387125fcSChris Mason if (device->nobarriers) 2889387125fcSChris Mason return 0; 2890387125fcSChris Mason 2891387125fcSChris Mason if (wait) { 2892387125fcSChris Mason bio = device->flush_bio; 2893387125fcSChris Mason if (!bio) 2894387125fcSChris Mason return 0; 2895387125fcSChris Mason 2896387125fcSChris Mason wait_for_completion(&device->flush_wait); 2897387125fcSChris Mason 2898387125fcSChris Mason if (bio_flagged(bio, BIO_EOPNOTSUPP)) { 2899606686eeSJosef Bacik printk_in_rcu("btrfs: disabling barriers on dev %s\n", 2900606686eeSJosef Bacik rcu_str_deref(device->name)); 2901387125fcSChris Mason device->nobarriers = 1; 29025af3e8ccSStefan Behrens } else if (!bio_flagged(bio, BIO_UPTODATE)) { 2903387125fcSChris Mason ret = -EIO; 2904442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, 2905442a4f63SStefan Behrens BTRFS_DEV_STAT_FLUSH_ERRS); 2906387125fcSChris Mason } 2907387125fcSChris Mason 2908387125fcSChris Mason /* drop the reference from the wait == 0 run */ 2909387125fcSChris Mason bio_put(bio); 2910387125fcSChris Mason device->flush_bio = NULL; 2911387125fcSChris Mason 2912387125fcSChris Mason return ret; 2913387125fcSChris Mason } 2914387125fcSChris Mason 2915387125fcSChris Mason /* 2916387125fcSChris Mason * one reference for us, and we leave it for the 2917387125fcSChris Mason * caller 2918387125fcSChris Mason */ 29199c017abcSJesper Juhl device->flush_bio = NULL; 2920387125fcSChris Mason bio = bio_alloc(GFP_NOFS, 0); 2921387125fcSChris Mason if (!bio) 2922387125fcSChris Mason return -ENOMEM; 2923387125fcSChris Mason 2924387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 2925387125fcSChris Mason bio->bi_bdev = device->bdev; 2926387125fcSChris Mason init_completion(&device->flush_wait); 2927387125fcSChris Mason bio->bi_private = &device->flush_wait; 2928387125fcSChris Mason device->flush_bio = bio; 2929387125fcSChris Mason 2930387125fcSChris Mason bio_get(bio); 293121adbd5cSStefan Behrens btrfsic_submit_bio(WRITE_FLUSH, bio); 2932387125fcSChris Mason 2933387125fcSChris Mason return 0; 2934387125fcSChris Mason } 2935387125fcSChris Mason 2936387125fcSChris Mason /* 2937387125fcSChris Mason * send an empty flush down to each device in parallel, 2938387125fcSChris Mason * then wait for them 2939387125fcSChris Mason */ 2940387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 2941387125fcSChris Mason { 2942387125fcSChris Mason struct list_head *head; 2943387125fcSChris Mason struct btrfs_device *dev; 29445af3e8ccSStefan Behrens int errors_send = 0; 29455af3e8ccSStefan Behrens int errors_wait = 0; 2946387125fcSChris Mason int ret; 2947387125fcSChris Mason 2948387125fcSChris Mason /* send down all the barriers */ 2949387125fcSChris Mason head = &info->fs_devices->devices; 2950387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 2951387125fcSChris Mason if (!dev->bdev) { 29525af3e8ccSStefan Behrens errors_send++; 2953387125fcSChris Mason continue; 2954387125fcSChris Mason } 2955387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 2956387125fcSChris Mason continue; 2957387125fcSChris Mason 2958387125fcSChris Mason ret = write_dev_flush(dev, 0); 2959387125fcSChris Mason if (ret) 29605af3e8ccSStefan Behrens errors_send++; 2961387125fcSChris Mason } 2962387125fcSChris Mason 2963387125fcSChris Mason /* wait for all the barriers */ 2964387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 2965387125fcSChris Mason if (!dev->bdev) { 29665af3e8ccSStefan Behrens errors_wait++; 2967387125fcSChris Mason continue; 2968387125fcSChris Mason } 2969387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 2970387125fcSChris Mason continue; 2971387125fcSChris Mason 2972387125fcSChris Mason ret = write_dev_flush(dev, 1); 2973387125fcSChris Mason if (ret) 29745af3e8ccSStefan Behrens errors_wait++; 2975387125fcSChris Mason } 29765af3e8ccSStefan Behrens if (errors_send > info->num_tolerated_disk_barrier_failures || 29775af3e8ccSStefan Behrens errors_wait > info->num_tolerated_disk_barrier_failures) 2978387125fcSChris Mason return -EIO; 2979387125fcSChris Mason return 0; 2980387125fcSChris Mason } 2981387125fcSChris Mason 29825af3e8ccSStefan Behrens int btrfs_calc_num_tolerated_disk_barrier_failures( 29835af3e8ccSStefan Behrens struct btrfs_fs_info *fs_info) 29845af3e8ccSStefan Behrens { 29855af3e8ccSStefan Behrens struct btrfs_ioctl_space_info space; 29865af3e8ccSStefan Behrens struct btrfs_space_info *sinfo; 29875af3e8ccSStefan Behrens u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 29885af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_SYSTEM, 29895af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_METADATA, 29905af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 29915af3e8ccSStefan Behrens int num_types = 4; 29925af3e8ccSStefan Behrens int i; 29935af3e8ccSStefan Behrens int c; 29945af3e8ccSStefan Behrens int num_tolerated_disk_barrier_failures = 29955af3e8ccSStefan Behrens (int)fs_info->fs_devices->num_devices; 29965af3e8ccSStefan Behrens 29975af3e8ccSStefan Behrens for (i = 0; i < num_types; i++) { 29985af3e8ccSStefan Behrens struct btrfs_space_info *tmp; 29995af3e8ccSStefan Behrens 30005af3e8ccSStefan Behrens sinfo = NULL; 30015af3e8ccSStefan Behrens rcu_read_lock(); 30025af3e8ccSStefan Behrens list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 30035af3e8ccSStefan Behrens if (tmp->flags == types[i]) { 30045af3e8ccSStefan Behrens sinfo = tmp; 30055af3e8ccSStefan Behrens break; 30065af3e8ccSStefan Behrens } 30075af3e8ccSStefan Behrens } 30085af3e8ccSStefan Behrens rcu_read_unlock(); 30095af3e8ccSStefan Behrens 30105af3e8ccSStefan Behrens if (!sinfo) 30115af3e8ccSStefan Behrens continue; 30125af3e8ccSStefan Behrens 30135af3e8ccSStefan Behrens down_read(&sinfo->groups_sem); 30145af3e8ccSStefan Behrens for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 30155af3e8ccSStefan Behrens if (!list_empty(&sinfo->block_groups[c])) { 30165af3e8ccSStefan Behrens u64 flags; 30175af3e8ccSStefan Behrens 30185af3e8ccSStefan Behrens btrfs_get_block_group_info( 30195af3e8ccSStefan Behrens &sinfo->block_groups[c], &space); 30205af3e8ccSStefan Behrens if (space.total_bytes == 0 || 30215af3e8ccSStefan Behrens space.used_bytes == 0) 30225af3e8ccSStefan Behrens continue; 30235af3e8ccSStefan Behrens flags = space.flags; 30245af3e8ccSStefan Behrens /* 30255af3e8ccSStefan Behrens * return 30265af3e8ccSStefan Behrens * 0: if dup, single or RAID0 is configured for 30275af3e8ccSStefan Behrens * any of metadata, system or data, else 30285af3e8ccSStefan Behrens * 1: if RAID5 is configured, or if RAID1 or 30295af3e8ccSStefan Behrens * RAID10 is configured and only two mirrors 30305af3e8ccSStefan Behrens * are used, else 30315af3e8ccSStefan Behrens * 2: if RAID6 is configured, else 30325af3e8ccSStefan Behrens * num_mirrors - 1: if RAID1 or RAID10 is 30335af3e8ccSStefan Behrens * configured and more than 30345af3e8ccSStefan Behrens * 2 mirrors are used. 30355af3e8ccSStefan Behrens */ 30365af3e8ccSStefan Behrens if (num_tolerated_disk_barrier_failures > 0 && 30375af3e8ccSStefan Behrens ((flags & (BTRFS_BLOCK_GROUP_DUP | 30385af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_RAID0)) || 30395af3e8ccSStefan Behrens ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) 30405af3e8ccSStefan Behrens == 0))) 30415af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 0; 30425af3e8ccSStefan Behrens else if (num_tolerated_disk_barrier_failures > 1 30435af3e8ccSStefan Behrens && 30445af3e8ccSStefan Behrens (flags & (BTRFS_BLOCK_GROUP_RAID1 | 30455af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_RAID10))) 30465af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 1; 30475af3e8ccSStefan Behrens } 30485af3e8ccSStefan Behrens } 30495af3e8ccSStefan Behrens up_read(&sinfo->groups_sem); 30505af3e8ccSStefan Behrens } 30515af3e8ccSStefan Behrens 30525af3e8ccSStefan Behrens return num_tolerated_disk_barrier_failures; 30535af3e8ccSStefan Behrens } 30545af3e8ccSStefan Behrens 3055a512bbf8SYan Zheng int write_all_supers(struct btrfs_root *root, int max_mirrors) 3056f2984462SChris Mason { 3057e5e9a520SChris Mason struct list_head *head; 3058f2984462SChris Mason struct btrfs_device *dev; 3059a061fc8dSChris Mason struct btrfs_super_block *sb; 3060f2984462SChris Mason struct btrfs_dev_item *dev_item; 3061f2984462SChris Mason int ret; 3062f2984462SChris Mason int do_barriers; 3063a236aed1SChris Mason int max_errors; 3064a236aed1SChris Mason int total_errors = 0; 3065a061fc8dSChris Mason u64 flags; 3066f2984462SChris Mason 30676c41761fSDavid Sterba max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3068f2984462SChris Mason do_barriers = !btrfs_test_opt(root, NOBARRIER); 3069af31f5e5SChris Mason backup_super_roots(root->fs_info); 3070f2984462SChris Mason 30716c41761fSDavid Sterba sb = root->fs_info->super_for_commit; 3072a061fc8dSChris Mason dev_item = &sb->dev_item; 3073e5e9a520SChris Mason 3074174ba509SChris Mason mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3075e5e9a520SChris Mason head = &root->fs_info->fs_devices->devices; 3076387125fcSChris Mason 30775af3e8ccSStefan Behrens if (do_barriers) { 30785af3e8ccSStefan Behrens ret = barrier_all_devices(root->fs_info); 30795af3e8ccSStefan Behrens if (ret) { 30805af3e8ccSStefan Behrens mutex_unlock( 30815af3e8ccSStefan Behrens &root->fs_info->fs_devices->device_list_mutex); 30825af3e8ccSStefan Behrens btrfs_error(root->fs_info, ret, 30835af3e8ccSStefan Behrens "errors while submitting device barriers."); 30845af3e8ccSStefan Behrens return ret; 30855af3e8ccSStefan Behrens } 30865af3e8ccSStefan Behrens } 3087387125fcSChris Mason 30881f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3089dfe25020SChris Mason if (!dev->bdev) { 3090dfe25020SChris Mason total_errors++; 3091dfe25020SChris Mason continue; 3092dfe25020SChris Mason } 30932b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3094dfe25020SChris Mason continue; 3095dfe25020SChris Mason 30962b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3097a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3098a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 3099a061fc8dSChris Mason btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); 3100a061fc8dSChris Mason btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); 3101a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3102a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3103a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3104a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 31052b82032cSYan Zheng memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3106a512bbf8SYan Zheng 3107a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3108a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3109f2984462SChris Mason 3110a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3111a236aed1SChris Mason if (ret) 3112a236aed1SChris Mason total_errors++; 3113f2984462SChris Mason } 3114a236aed1SChris Mason if (total_errors > max_errors) { 3115d397712bSChris Mason printk(KERN_ERR "btrfs: %d errors while writing supers\n", 3116d397712bSChris Mason total_errors); 311779787eaaSJeff Mahoney 311879787eaaSJeff Mahoney /* This shouldn't happen. FUA is masked off if unsupported */ 3119a236aed1SChris Mason BUG(); 3120a236aed1SChris Mason } 3121f2984462SChris Mason 3122a512bbf8SYan Zheng total_errors = 0; 31231f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3124dfe25020SChris Mason if (!dev->bdev) 3125dfe25020SChris Mason continue; 31262b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3127dfe25020SChris Mason continue; 3128dfe25020SChris Mason 3129a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3130a512bbf8SYan Zheng if (ret) 31311259ab75SChris Mason total_errors++; 3132f2984462SChris Mason } 3133174ba509SChris Mason mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3134a236aed1SChris Mason if (total_errors > max_errors) { 313579787eaaSJeff Mahoney btrfs_error(root->fs_info, -EIO, 313679787eaaSJeff Mahoney "%d errors while writing supers", total_errors); 313779787eaaSJeff Mahoney return -EIO; 3138a236aed1SChris Mason } 3139f2984462SChris Mason return 0; 3140f2984462SChris Mason } 3141f2984462SChris Mason 3142a512bbf8SYan Zheng int write_ctree_super(struct btrfs_trans_handle *trans, 3143a512bbf8SYan Zheng struct btrfs_root *root, int max_mirrors) 3144cfaa7295SChris Mason { 3145e66f709bSChris Mason int ret; 31462c90e5d6SChris Mason 3147a512bbf8SYan Zheng ret = write_all_supers(root, max_mirrors); 31485f39d397SChris Mason return ret; 3149cfaa7295SChris Mason } 3150cfaa7295SChris Mason 3151143bede5SJeff Mahoney void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) 31522619ba1fSChris Mason { 31534df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 31542619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 31552619ba1fSChris Mason (unsigned long)root->root_key.objectid); 31564df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 315776dda93cSYan, Zheng 315876dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 315976dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 316076dda93cSYan, Zheng 3161581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3162581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 31634df27c4dSYan, Zheng free_fs_root(root); 31644df27c4dSYan, Zheng } 31654df27c4dSYan, Zheng 31664df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 31674df27c4dSYan, Zheng { 316882d5902dSLi Zefan iput(root->cache_inode); 31694df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 31700ee5dc67SAl Viro if (root->anon_dev) 31710ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 31725f39d397SChris Mason free_extent_buffer(root->node); 31735f39d397SChris Mason free_extent_buffer(root->commit_root); 3174581bb050SLi Zefan kfree(root->free_ino_ctl); 3175581bb050SLi Zefan kfree(root->free_ino_pinned); 317658176a96SJosef Bacik kfree(root->name); 31772619ba1fSChris Mason kfree(root); 31782619ba1fSChris Mason } 31792619ba1fSChris Mason 3180143bede5SJeff Mahoney static void del_fs_roots(struct btrfs_fs_info *fs_info) 31810f7d52f4SChris Mason { 31820f7d52f4SChris Mason int ret; 31830f7d52f4SChris Mason struct btrfs_root *gang[8]; 31840f7d52f4SChris Mason int i; 31850f7d52f4SChris Mason 318676dda93cSYan, Zheng while (!list_empty(&fs_info->dead_roots)) { 318776dda93cSYan, Zheng gang[0] = list_entry(fs_info->dead_roots.next, 318876dda93cSYan, Zheng struct btrfs_root, root_list); 318976dda93cSYan, Zheng list_del(&gang[0]->root_list); 319076dda93cSYan, Zheng 319176dda93cSYan, Zheng if (gang[0]->in_radix) { 319276dda93cSYan, Zheng btrfs_free_fs_root(fs_info, gang[0]); 319376dda93cSYan, Zheng } else { 319476dda93cSYan, Zheng free_extent_buffer(gang[0]->node); 319576dda93cSYan, Zheng free_extent_buffer(gang[0]->commit_root); 319676dda93cSYan, Zheng kfree(gang[0]); 319776dda93cSYan, Zheng } 319876dda93cSYan, Zheng } 319976dda93cSYan, Zheng 32000f7d52f4SChris Mason while (1) { 32010f7d52f4SChris Mason ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 32020f7d52f4SChris Mason (void **)gang, 0, 32030f7d52f4SChris Mason ARRAY_SIZE(gang)); 32040f7d52f4SChris Mason if (!ret) 32050f7d52f4SChris Mason break; 32062619ba1fSChris Mason for (i = 0; i < ret; i++) 32075eda7b5eSChris Mason btrfs_free_fs_root(fs_info, gang[i]); 32080f7d52f4SChris Mason } 32090f7d52f4SChris Mason } 3210b4100d64SChris Mason 3211c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3212c146afadSYan Zheng { 3213c146afadSYan Zheng u64 root_objectid = 0; 3214c146afadSYan Zheng struct btrfs_root *gang[8]; 3215c146afadSYan Zheng int i; 3216c146afadSYan Zheng int ret; 3217c146afadSYan Zheng 3218c146afadSYan Zheng while (1) { 3219c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3220c146afadSYan Zheng (void **)gang, root_objectid, 3221c146afadSYan Zheng ARRAY_SIZE(gang)); 3222c146afadSYan Zheng if (!ret) 3223c146afadSYan Zheng break; 32245d4f98a2SYan Zheng 32255d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 3226c146afadSYan Zheng for (i = 0; i < ret; i++) { 322766b4ffd1SJosef Bacik int err; 322866b4ffd1SJosef Bacik 3229c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 323066b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 323166b4ffd1SJosef Bacik if (err) 323266b4ffd1SJosef Bacik return err; 3233c146afadSYan Zheng } 3234c146afadSYan Zheng root_objectid++; 3235c146afadSYan Zheng } 3236c146afadSYan Zheng return 0; 3237c146afadSYan Zheng } 3238c146afadSYan Zheng 3239c146afadSYan Zheng int btrfs_commit_super(struct btrfs_root *root) 3240c146afadSYan Zheng { 3241c146afadSYan Zheng struct btrfs_trans_handle *trans; 3242c146afadSYan Zheng int ret; 3243c146afadSYan Zheng 3244c146afadSYan Zheng mutex_lock(&root->fs_info->cleaner_mutex); 324524bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 3246c146afadSYan Zheng btrfs_clean_old_snapshots(root); 3247c146afadSYan Zheng mutex_unlock(&root->fs_info->cleaner_mutex); 3248c71bf099SYan, Zheng 3249c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 3250c71bf099SYan, Zheng down_write(&root->fs_info->cleanup_work_sem); 3251c71bf099SYan, Zheng up_write(&root->fs_info->cleanup_work_sem); 3252c71bf099SYan, Zheng 32537a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 32543612b495STsutomu Itoh if (IS_ERR(trans)) 32553612b495STsutomu Itoh return PTR_ERR(trans); 3256c146afadSYan Zheng ret = btrfs_commit_transaction(trans, root); 325779787eaaSJeff Mahoney if (ret) 325879787eaaSJeff Mahoney return ret; 3259c146afadSYan Zheng /* run commit again to drop the original snapshot */ 32607a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 32613612b495STsutomu Itoh if (IS_ERR(trans)) 32623612b495STsutomu Itoh return PTR_ERR(trans); 326379787eaaSJeff Mahoney ret = btrfs_commit_transaction(trans, root); 326479787eaaSJeff Mahoney if (ret) 326579787eaaSJeff Mahoney return ret; 3266c146afadSYan Zheng ret = btrfs_write_and_wait_transaction(NULL, root); 326779787eaaSJeff Mahoney if (ret) { 326879787eaaSJeff Mahoney btrfs_error(root->fs_info, ret, 326979787eaaSJeff Mahoney "Failed to sync btree inode to disk."); 327079787eaaSJeff Mahoney return ret; 327179787eaaSJeff Mahoney } 3272c146afadSYan Zheng 3273a512bbf8SYan Zheng ret = write_ctree_super(NULL, root, 0); 3274c146afadSYan Zheng return ret; 3275c146afadSYan Zheng } 3276c146afadSYan Zheng 3277e20d96d6SChris Mason int close_ctree(struct btrfs_root *root) 3278eb60ceacSChris Mason { 32790f7d52f4SChris Mason struct btrfs_fs_info *fs_info = root->fs_info; 3280c146afadSYan Zheng int ret; 3281e089f05cSChris Mason 3282facda1e7SChris Mason fs_info->closing = 1; 3283a2135011SChris Mason smp_mb(); 3284a2135011SChris Mason 3285837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3286aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3287837d5b6eSIlya Dryomov 3288aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 32894cb5300bSChris Mason 32904cb5300bSChris Mason /* wait for any defraggers to finish */ 32914cb5300bSChris Mason wait_event(fs_info->transaction_wait, 32924cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 32934cb5300bSChris Mason 32944cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 3295e3029d9fSAl Viro btrfs_run_defrag_inodes(fs_info); 32964cb5300bSChris Mason 3297c146afadSYan Zheng if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3298c146afadSYan Zheng ret = btrfs_commit_super(root); 3299d397712bSChris Mason if (ret) 3300d397712bSChris Mason printk(KERN_ERR "btrfs: commit super ret %d\n", ret); 3301c146afadSYan Zheng } 3302ed2ff2cbSChris Mason 330368ce9682SStefan Behrens if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) 330468ce9682SStefan Behrens btrfs_error_commit_super(root); 3305acce952bSliubo 3306300e4f8aSJosef Bacik btrfs_put_block_group_cache(fs_info); 3307300e4f8aSJosef Bacik 3308e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3309e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 33108929ecfaSYan, Zheng 3311f25784b3SYan Zheng fs_info->closing = 2; 3312f25784b3SYan Zheng smp_mb(); 3313f25784b3SYan Zheng 3314bcef60f2SArne Jansen btrfs_free_qgroup_config(root->fs_info); 3315bcef60f2SArne Jansen 3316b0c68f8bSChris Mason if (fs_info->delalloc_bytes) { 3317d397712bSChris Mason printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", 331821380931SJoel Becker (unsigned long long)fs_info->delalloc_bytes); 3319b0c68f8bSChris Mason } 332031153d81SYan Zheng 33215f39d397SChris Mason free_extent_buffer(fs_info->extent_root->node); 33225d4f98a2SYan Zheng free_extent_buffer(fs_info->extent_root->commit_root); 33235f39d397SChris Mason free_extent_buffer(fs_info->tree_root->node); 33245d4f98a2SYan Zheng free_extent_buffer(fs_info->tree_root->commit_root); 3325e3029d9fSAl Viro free_extent_buffer(fs_info->chunk_root->node); 3326e3029d9fSAl Viro free_extent_buffer(fs_info->chunk_root->commit_root); 3327e3029d9fSAl Viro free_extent_buffer(fs_info->dev_root->node); 3328e3029d9fSAl Viro free_extent_buffer(fs_info->dev_root->commit_root); 3329e3029d9fSAl Viro free_extent_buffer(fs_info->csum_root->node); 3330e3029d9fSAl Viro free_extent_buffer(fs_info->csum_root->commit_root); 3331bcef60f2SArne Jansen if (fs_info->quota_root) { 3332bcef60f2SArne Jansen free_extent_buffer(fs_info->quota_root->node); 3333bcef60f2SArne Jansen free_extent_buffer(fs_info->quota_root->commit_root); 3334bcef60f2SArne Jansen } 3335d20f7043SChris Mason 3336e3029d9fSAl Viro btrfs_free_block_groups(fs_info); 3337c146afadSYan Zheng 33380f7d52f4SChris Mason del_fs_roots(fs_info); 3339d10c5f31SChris Mason 3340c146afadSYan Zheng iput(fs_info->btree_inode); 33419ad6b7bcSChris Mason 334261d92c32SChris Mason btrfs_stop_workers(&fs_info->generic_worker); 3343247e743cSChris Mason btrfs_stop_workers(&fs_info->fixup_workers); 3344771ed689SChris Mason btrfs_stop_workers(&fs_info->delalloc_workers); 33458b712842SChris Mason btrfs_stop_workers(&fs_info->workers); 33468b712842SChris Mason btrfs_stop_workers(&fs_info->endio_workers); 3347d20f7043SChris Mason btrfs_stop_workers(&fs_info->endio_meta_workers); 3348cad321adSChris Mason btrfs_stop_workers(&fs_info->endio_meta_write_workers); 3349e6dcd2dcSChris Mason btrfs_stop_workers(&fs_info->endio_write_workers); 33500cb59c99SJosef Bacik btrfs_stop_workers(&fs_info->endio_freespace_worker); 33511cc127b5SChris Mason btrfs_stop_workers(&fs_info->submit_workers); 335216cdcec7SMiao Xie btrfs_stop_workers(&fs_info->delayed_workers); 3353bab39bf9SJosef Bacik btrfs_stop_workers(&fs_info->caching_workers); 335490519d66SArne Jansen btrfs_stop_workers(&fs_info->readahead_workers); 33558ccf6f19SMiao Xie btrfs_stop_workers(&fs_info->flush_workers); 3356d6bfde87SChris Mason 335721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 335821adbd5cSStefan Behrens if (btrfs_test_opt(root, CHECK_INTEGRITY)) 335921adbd5cSStefan Behrens btrfsic_unmount(root, fs_info->fs_devices); 336021adbd5cSStefan Behrens #endif 336121adbd5cSStefan Behrens 3362dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 33630b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3364b248a415SChris Mason 336504160088SChris Mason bdi_destroy(&fs_info->bdi); 336676dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 33670b86a832SChris Mason 3368eb60ceacSChris Mason return 0; 3369eb60ceacSChris Mason } 3370eb60ceacSChris Mason 3371b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3372b9fab919SChris Mason int atomic) 3373ccd467d6SChris Mason { 33741259ab75SChris Mason int ret; 3375727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 33761259ab75SChris Mason 33770b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 33781259ab75SChris Mason if (!ret) 33791259ab75SChris Mason return ret; 33801259ab75SChris Mason 33811259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3382b9fab919SChris Mason parent_transid, atomic); 3383b9fab919SChris Mason if (ret == -EAGAIN) 3384b9fab919SChris Mason return ret; 33851259ab75SChris Mason return !ret; 33865f39d397SChris Mason } 33876702ed49SChris Mason 33885f39d397SChris Mason int btrfs_set_buffer_uptodate(struct extent_buffer *buf) 33895f39d397SChris Mason { 33900b32f4bbSJosef Bacik return set_extent_buffer_uptodate(buf); 33915f39d397SChris Mason } 33925f39d397SChris Mason 33935f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 33945f39d397SChris Mason { 3395727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 33965f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3397b9473439SChris Mason int was_dirty; 3398b4ce94deSChris Mason 3399b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 340031b1a2bdSJulia Lawall if (transid != root->fs_info->generation) 340131b1a2bdSJulia Lawall WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3402d397712bSChris Mason "found %llu running %llu\n", 3403db94535dSChris Mason (unsigned long long)buf->start, 3404d397712bSChris Mason (unsigned long long)transid, 3405d397712bSChris Mason (unsigned long long)root->fs_info->generation); 34060b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3407b9473439SChris Mason if (!was_dirty) { 3408b9473439SChris Mason spin_lock(&root->fs_info->delalloc_lock); 3409b9473439SChris Mason root->fs_info->dirty_metadata_bytes += buf->len; 3410b9473439SChris Mason spin_unlock(&root->fs_info->delalloc_lock); 3411b9473439SChris Mason } 3412eb60ceacSChris Mason } 3413eb60ceacSChris Mason 3414b53d3f5dSLiu Bo static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 3415b53d3f5dSLiu Bo int flush_delayed) 341635b7e476SChris Mason { 3417188de649SChris Mason /* 3418188de649SChris Mason * looks as though older kernels can get into trouble with 3419188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3420188de649SChris Mason */ 3421d6bfde87SChris Mason u64 num_dirty; 3422771ed689SChris Mason unsigned long thresh = 32 * 1024 * 1024; 3423d6bfde87SChris Mason 34246933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 3425d6bfde87SChris Mason return; 3426d6bfde87SChris Mason 3427b53d3f5dSLiu Bo if (flush_delayed) 342816cdcec7SMiao Xie btrfs_balance_delayed_items(root); 342916cdcec7SMiao Xie 343016cdcec7SMiao Xie num_dirty = root->fs_info->dirty_metadata_bytes; 343116cdcec7SMiao Xie 343216cdcec7SMiao Xie if (num_dirty > thresh) { 343316cdcec7SMiao Xie balance_dirty_pages_ratelimited_nr( 343416cdcec7SMiao Xie root->fs_info->btree_inode->i_mapping, 1); 343516cdcec7SMiao Xie } 343616cdcec7SMiao Xie return; 343716cdcec7SMiao Xie } 343816cdcec7SMiao Xie 3439b53d3f5dSLiu Bo void btrfs_btree_balance_dirty(struct btrfs_root *root) 344016cdcec7SMiao Xie { 3441b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 1); 344235b7e476SChris Mason } 3443b53d3f5dSLiu Bo 3444b53d3f5dSLiu Bo void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 3445b53d3f5dSLiu Bo { 3446b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 0); 3447d6bfde87SChris Mason } 34486b80053dSChris Mason 3449ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 34506b80053dSChris Mason { 3451727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 34520b32f4bbSJosef Bacik return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 34536b80053dSChris Mason } 34540da5468fSChris Mason 3455fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 3456acce952bSliubo int read_only) 3457acce952bSliubo { 3458fcd1f065SDavid Sterba if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) { 3459fcd1f065SDavid Sterba printk(KERN_ERR "btrfs: unsupported checksum algorithm\n"); 3460fcd1f065SDavid Sterba return -EINVAL; 3461fcd1f065SDavid Sterba } 3462acce952bSliubo 3463fcd1f065SDavid Sterba if (read_only) 3464fcd1f065SDavid Sterba return 0; 3465fcd1f065SDavid Sterba 3466fcd1f065SDavid Sterba return 0; 3467fcd1f065SDavid Sterba } 3468fcd1f065SDavid Sterba 346968ce9682SStefan Behrens void btrfs_error_commit_super(struct btrfs_root *root) 3470acce952bSliubo { 3471acce952bSliubo mutex_lock(&root->fs_info->cleaner_mutex); 3472acce952bSliubo btrfs_run_delayed_iputs(root); 3473acce952bSliubo mutex_unlock(&root->fs_info->cleaner_mutex); 3474acce952bSliubo 3475acce952bSliubo down_write(&root->fs_info->cleanup_work_sem); 3476acce952bSliubo up_write(&root->fs_info->cleanup_work_sem); 3477acce952bSliubo 3478acce952bSliubo /* cleanup FS via transaction */ 3479acce952bSliubo btrfs_cleanup_transaction(root); 3480acce952bSliubo } 3481acce952bSliubo 3482143bede5SJeff Mahoney static void btrfs_destroy_ordered_operations(struct btrfs_root *root) 3483acce952bSliubo { 3484acce952bSliubo struct btrfs_inode *btrfs_inode; 3485acce952bSliubo struct list_head splice; 3486acce952bSliubo 3487acce952bSliubo INIT_LIST_HEAD(&splice); 3488acce952bSliubo 3489acce952bSliubo mutex_lock(&root->fs_info->ordered_operations_mutex); 3490acce952bSliubo spin_lock(&root->fs_info->ordered_extent_lock); 3491acce952bSliubo 3492acce952bSliubo list_splice_init(&root->fs_info->ordered_operations, &splice); 3493acce952bSliubo while (!list_empty(&splice)) { 3494acce952bSliubo btrfs_inode = list_entry(splice.next, struct btrfs_inode, 3495acce952bSliubo ordered_operations); 3496acce952bSliubo 3497acce952bSliubo list_del_init(&btrfs_inode->ordered_operations); 3498acce952bSliubo 3499acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3500acce952bSliubo } 3501acce952bSliubo 3502acce952bSliubo spin_unlock(&root->fs_info->ordered_extent_lock); 3503acce952bSliubo mutex_unlock(&root->fs_info->ordered_operations_mutex); 3504acce952bSliubo } 3505acce952bSliubo 3506143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 3507acce952bSliubo { 3508acce952bSliubo struct list_head splice; 3509acce952bSliubo struct btrfs_ordered_extent *ordered; 3510acce952bSliubo struct inode *inode; 3511acce952bSliubo 3512acce952bSliubo INIT_LIST_HEAD(&splice); 3513acce952bSliubo 3514acce952bSliubo spin_lock(&root->fs_info->ordered_extent_lock); 3515acce952bSliubo 3516acce952bSliubo list_splice_init(&root->fs_info->ordered_extents, &splice); 3517acce952bSliubo while (!list_empty(&splice)) { 3518acce952bSliubo ordered = list_entry(splice.next, struct btrfs_ordered_extent, 3519acce952bSliubo root_extent_list); 3520acce952bSliubo 3521acce952bSliubo list_del_init(&ordered->root_extent_list); 3522acce952bSliubo atomic_inc(&ordered->refs); 3523acce952bSliubo 3524acce952bSliubo /* the inode may be getting freed (in sys_unlink path). */ 3525acce952bSliubo inode = igrab(ordered->inode); 3526acce952bSliubo 3527acce952bSliubo spin_unlock(&root->fs_info->ordered_extent_lock); 3528acce952bSliubo if (inode) 3529acce952bSliubo iput(inode); 3530acce952bSliubo 3531acce952bSliubo atomic_set(&ordered->refs, 1); 3532acce952bSliubo btrfs_put_ordered_extent(ordered); 3533acce952bSliubo 3534acce952bSliubo spin_lock(&root->fs_info->ordered_extent_lock); 3535acce952bSliubo } 3536acce952bSliubo 3537acce952bSliubo spin_unlock(&root->fs_info->ordered_extent_lock); 3538acce952bSliubo } 3539acce952bSliubo 354079787eaaSJeff Mahoney int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 3541acce952bSliubo struct btrfs_root *root) 3542acce952bSliubo { 3543acce952bSliubo struct rb_node *node; 3544acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 3545acce952bSliubo struct btrfs_delayed_ref_node *ref; 3546acce952bSliubo int ret = 0; 3547acce952bSliubo 3548acce952bSliubo delayed_refs = &trans->delayed_refs; 3549acce952bSliubo 3550acce952bSliubo spin_lock(&delayed_refs->lock); 3551acce952bSliubo if (delayed_refs->num_entries == 0) { 3552cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 3553acce952bSliubo printk(KERN_INFO "delayed_refs has NO entry\n"); 3554acce952bSliubo return ret; 3555acce952bSliubo } 3556acce952bSliubo 3557b939d1abSJosef Bacik while ((node = rb_first(&delayed_refs->root)) != NULL) { 3558acce952bSliubo ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); 3559acce952bSliubo 3560acce952bSliubo atomic_set(&ref->refs, 1); 3561acce952bSliubo if (btrfs_delayed_ref_is_head(ref)) { 3562acce952bSliubo struct btrfs_delayed_ref_head *head; 3563acce952bSliubo 3564acce952bSliubo head = btrfs_delayed_node_to_head(ref); 3565b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 3566b939d1abSJosef Bacik atomic_inc(&ref->refs); 356779787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 3568b939d1abSJosef Bacik 3569b939d1abSJosef Bacik /* Need to wait for the delayed ref to run */ 3570acce952bSliubo mutex_lock(&head->mutex); 3571b939d1abSJosef Bacik mutex_unlock(&head->mutex); 3572b939d1abSJosef Bacik btrfs_put_delayed_ref(ref); 3573b939d1abSJosef Bacik 3574e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 3575b939d1abSJosef Bacik continue; 3576b939d1abSJosef Bacik } 3577b939d1abSJosef Bacik 3578acce952bSliubo kfree(head->extent_op); 3579acce952bSliubo delayed_refs->num_heads--; 3580acce952bSliubo if (list_empty(&head->cluster)) 3581acce952bSliubo delayed_refs->num_heads_ready--; 3582acce952bSliubo list_del_init(&head->cluster); 3583acce952bSliubo } 3584b939d1abSJosef Bacik ref->in_tree = 0; 3585b939d1abSJosef Bacik rb_erase(&ref->rb_node, &delayed_refs->root); 3586b939d1abSJosef Bacik delayed_refs->num_entries--; 3587b939d1abSJosef Bacik 3588acce952bSliubo spin_unlock(&delayed_refs->lock); 3589acce952bSliubo btrfs_put_delayed_ref(ref); 3590acce952bSliubo 3591acce952bSliubo cond_resched(); 3592acce952bSliubo spin_lock(&delayed_refs->lock); 3593acce952bSliubo } 3594acce952bSliubo 3595acce952bSliubo spin_unlock(&delayed_refs->lock); 3596acce952bSliubo 3597acce952bSliubo return ret; 3598acce952bSliubo } 3599acce952bSliubo 3600143bede5SJeff Mahoney static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) 3601acce952bSliubo { 3602acce952bSliubo struct btrfs_pending_snapshot *snapshot; 3603acce952bSliubo struct list_head splice; 3604acce952bSliubo 3605acce952bSliubo INIT_LIST_HEAD(&splice); 3606acce952bSliubo 3607acce952bSliubo list_splice_init(&t->pending_snapshots, &splice); 3608acce952bSliubo 3609acce952bSliubo while (!list_empty(&splice)) { 3610acce952bSliubo snapshot = list_entry(splice.next, 3611acce952bSliubo struct btrfs_pending_snapshot, 3612acce952bSliubo list); 3613acce952bSliubo 3614acce952bSliubo list_del_init(&snapshot->list); 3615acce952bSliubo 3616acce952bSliubo kfree(snapshot); 3617acce952bSliubo } 3618acce952bSliubo } 3619acce952bSliubo 3620143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 3621acce952bSliubo { 3622acce952bSliubo struct btrfs_inode *btrfs_inode; 3623acce952bSliubo struct list_head splice; 3624acce952bSliubo 3625acce952bSliubo INIT_LIST_HEAD(&splice); 3626acce952bSliubo 3627acce952bSliubo spin_lock(&root->fs_info->delalloc_lock); 36285be76758SDavid Sterba list_splice_init(&root->fs_info->delalloc_inodes, &splice); 3629acce952bSliubo 3630acce952bSliubo while (!list_empty(&splice)) { 3631acce952bSliubo btrfs_inode = list_entry(splice.next, struct btrfs_inode, 3632acce952bSliubo delalloc_inodes); 3633acce952bSliubo 3634acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 3635acce952bSliubo 3636acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3637acce952bSliubo } 3638acce952bSliubo 3639acce952bSliubo spin_unlock(&root->fs_info->delalloc_lock); 3640acce952bSliubo } 3641acce952bSliubo 3642acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 3643acce952bSliubo struct extent_io_tree *dirty_pages, 3644acce952bSliubo int mark) 3645acce952bSliubo { 3646acce952bSliubo int ret; 3647acce952bSliubo struct page *page; 3648acce952bSliubo struct inode *btree_inode = root->fs_info->btree_inode; 3649acce952bSliubo struct extent_buffer *eb; 3650acce952bSliubo u64 start = 0; 3651acce952bSliubo u64 end; 3652acce952bSliubo u64 offset; 3653acce952bSliubo unsigned long index; 3654acce952bSliubo 3655acce952bSliubo while (1) { 3656acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 3657e6138876SJosef Bacik mark, NULL); 3658acce952bSliubo if (ret) 3659acce952bSliubo break; 3660acce952bSliubo 3661acce952bSliubo clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); 3662acce952bSliubo while (start <= end) { 3663acce952bSliubo index = start >> PAGE_CACHE_SHIFT; 3664acce952bSliubo start = (u64)(index + 1) << PAGE_CACHE_SHIFT; 3665acce952bSliubo page = find_get_page(btree_inode->i_mapping, index); 3666acce952bSliubo if (!page) 3667acce952bSliubo continue; 3668acce952bSliubo offset = page_offset(page); 3669acce952bSliubo 3670acce952bSliubo spin_lock(&dirty_pages->buffer_lock); 3671acce952bSliubo eb = radix_tree_lookup( 3672acce952bSliubo &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, 3673acce952bSliubo offset >> PAGE_CACHE_SHIFT); 3674acce952bSliubo spin_unlock(&dirty_pages->buffer_lock); 3675ee670f0aSJosef Bacik if (eb) 3676acce952bSliubo ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, 3677acce952bSliubo &eb->bflags); 3678acce952bSliubo if (PageWriteback(page)) 3679acce952bSliubo end_page_writeback(page); 3680acce952bSliubo 3681acce952bSliubo lock_page(page); 3682acce952bSliubo if (PageDirty(page)) { 3683acce952bSliubo clear_page_dirty_for_io(page); 3684acce952bSliubo spin_lock_irq(&page->mapping->tree_lock); 3685acce952bSliubo radix_tree_tag_clear(&page->mapping->page_tree, 3686acce952bSliubo page_index(page), 3687acce952bSliubo PAGECACHE_TAG_DIRTY); 3688acce952bSliubo spin_unlock_irq(&page->mapping->tree_lock); 3689acce952bSliubo } 3690acce952bSliubo 3691acce952bSliubo unlock_page(page); 3692ee670f0aSJosef Bacik page_cache_release(page); 3693acce952bSliubo } 3694acce952bSliubo } 3695acce952bSliubo 3696acce952bSliubo return ret; 3697acce952bSliubo } 3698acce952bSliubo 3699acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 3700acce952bSliubo struct extent_io_tree *pinned_extents) 3701acce952bSliubo { 3702acce952bSliubo struct extent_io_tree *unpin; 3703acce952bSliubo u64 start; 3704acce952bSliubo u64 end; 3705acce952bSliubo int ret; 3706ed0eaa14SLiu Bo bool loop = true; 3707acce952bSliubo 3708acce952bSliubo unpin = pinned_extents; 3709ed0eaa14SLiu Bo again: 3710acce952bSliubo while (1) { 3711acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 3712e6138876SJosef Bacik EXTENT_DIRTY, NULL); 3713acce952bSliubo if (ret) 3714acce952bSliubo break; 3715acce952bSliubo 3716acce952bSliubo /* opt_discard */ 37175378e607SLi Dongyang if (btrfs_test_opt(root, DISCARD)) 37185378e607SLi Dongyang ret = btrfs_error_discard_extent(root, start, 37195378e607SLi Dongyang end + 1 - start, 37205378e607SLi Dongyang NULL); 3721acce952bSliubo 3722acce952bSliubo clear_extent_dirty(unpin, start, end, GFP_NOFS); 3723acce952bSliubo btrfs_error_unpin_extent_range(root, start, end); 3724acce952bSliubo cond_resched(); 3725acce952bSliubo } 3726acce952bSliubo 3727ed0eaa14SLiu Bo if (loop) { 3728ed0eaa14SLiu Bo if (unpin == &root->fs_info->freed_extents[0]) 3729ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[1]; 3730ed0eaa14SLiu Bo else 3731ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[0]; 3732ed0eaa14SLiu Bo loop = false; 3733ed0eaa14SLiu Bo goto again; 3734ed0eaa14SLiu Bo } 3735ed0eaa14SLiu Bo 3736acce952bSliubo return 0; 3737acce952bSliubo } 3738acce952bSliubo 373949b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 374049b25e05SJeff Mahoney struct btrfs_root *root) 374149b25e05SJeff Mahoney { 374249b25e05SJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, root); 374349b25e05SJeff Mahoney btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, 374449b25e05SJeff Mahoney cur_trans->dirty_pages.dirty_bytes); 374549b25e05SJeff Mahoney 374649b25e05SJeff Mahoney /* FIXME: cleanup wait for commit */ 374749b25e05SJeff Mahoney cur_trans->in_commit = 1; 374849b25e05SJeff Mahoney cur_trans->blocked = 1; 374949b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_blocked_wait); 375049b25e05SJeff Mahoney 375149b25e05SJeff Mahoney cur_trans->blocked = 0; 375249b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_wait); 375349b25e05SJeff Mahoney 375449b25e05SJeff Mahoney cur_trans->commit_done = 1; 375549b25e05SJeff Mahoney wake_up(&cur_trans->commit_wait); 375649b25e05SJeff Mahoney 375767cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 375867cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 375967cde344SMiao Xie 376049b25e05SJeff Mahoney btrfs_destroy_pending_snapshots(cur_trans); 376149b25e05SJeff Mahoney 376249b25e05SJeff Mahoney btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 376349b25e05SJeff Mahoney EXTENT_DIRTY); 37646e841e32SLiu Bo btrfs_destroy_pinned_extent(root, 37656e841e32SLiu Bo root->fs_info->pinned_extents); 376649b25e05SJeff Mahoney 376749b25e05SJeff Mahoney /* 376849b25e05SJeff Mahoney memset(cur_trans, 0, sizeof(*cur_trans)); 376949b25e05SJeff Mahoney kmem_cache_free(btrfs_transaction_cachep, cur_trans); 377049b25e05SJeff Mahoney */ 377149b25e05SJeff Mahoney } 377249b25e05SJeff Mahoney 377349b25e05SJeff Mahoney int btrfs_cleanup_transaction(struct btrfs_root *root) 3774acce952bSliubo { 3775acce952bSliubo struct btrfs_transaction *t; 3776acce952bSliubo LIST_HEAD(list); 3777acce952bSliubo 3778acce952bSliubo mutex_lock(&root->fs_info->transaction_kthread_mutex); 3779acce952bSliubo 3780a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 3781acce952bSliubo list_splice_init(&root->fs_info->trans_list, &list); 3782a4abeea4SJosef Bacik root->fs_info->trans_no_join = 1; 3783a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 3784a4abeea4SJosef Bacik 3785acce952bSliubo while (!list_empty(&list)) { 3786acce952bSliubo t = list_entry(list.next, struct btrfs_transaction, list); 3787acce952bSliubo if (!t) 3788acce952bSliubo break; 3789acce952bSliubo 3790acce952bSliubo btrfs_destroy_ordered_operations(root); 3791acce952bSliubo 3792acce952bSliubo btrfs_destroy_ordered_extents(root); 3793acce952bSliubo 3794acce952bSliubo btrfs_destroy_delayed_refs(t, root); 3795acce952bSliubo 3796acce952bSliubo btrfs_block_rsv_release(root, 3797acce952bSliubo &root->fs_info->trans_block_rsv, 3798acce952bSliubo t->dirty_pages.dirty_bytes); 3799acce952bSliubo 3800acce952bSliubo /* FIXME: cleanup wait for commit */ 3801acce952bSliubo t->in_commit = 1; 3802acce952bSliubo t->blocked = 1; 380366657b31SJosef Bacik smp_mb(); 3804acce952bSliubo if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) 3805acce952bSliubo wake_up(&root->fs_info->transaction_blocked_wait); 3806acce952bSliubo 3807acce952bSliubo t->blocked = 0; 380866657b31SJosef Bacik smp_mb(); 3809acce952bSliubo if (waitqueue_active(&root->fs_info->transaction_wait)) 3810acce952bSliubo wake_up(&root->fs_info->transaction_wait); 3811acce952bSliubo 3812acce952bSliubo t->commit_done = 1; 381366657b31SJosef Bacik smp_mb(); 3814acce952bSliubo if (waitqueue_active(&t->commit_wait)) 3815acce952bSliubo wake_up(&t->commit_wait); 3816acce952bSliubo 381767cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 381867cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 381967cde344SMiao Xie 3820acce952bSliubo btrfs_destroy_pending_snapshots(t); 3821acce952bSliubo 3822acce952bSliubo btrfs_destroy_delalloc_inodes(root); 3823acce952bSliubo 3824a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 3825acce952bSliubo root->fs_info->running_transaction = NULL; 3826a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 3827acce952bSliubo 3828acce952bSliubo btrfs_destroy_marked_extents(root, &t->dirty_pages, 3829acce952bSliubo EXTENT_DIRTY); 3830acce952bSliubo 3831acce952bSliubo btrfs_destroy_pinned_extent(root, 3832acce952bSliubo root->fs_info->pinned_extents); 3833acce952bSliubo 383413c5a93eSJosef Bacik atomic_set(&t->use_count, 0); 3835acce952bSliubo list_del_init(&t->list); 3836acce952bSliubo memset(t, 0, sizeof(*t)); 3837acce952bSliubo kmem_cache_free(btrfs_transaction_cachep, t); 3838acce952bSliubo } 3839acce952bSliubo 3840a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 3841a4abeea4SJosef Bacik root->fs_info->trans_no_join = 0; 3842a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 3843acce952bSliubo mutex_unlock(&root->fs_info->transaction_kthread_mutex); 3844acce952bSliubo 3845acce952bSliubo return 0; 3846acce952bSliubo } 3847acce952bSliubo 3848d1310b2eSChris Mason static struct extent_io_ops btree_extent_io_ops = { 3849ce9adaa5SChris Mason .readpage_end_io_hook = btree_readpage_end_io_hook, 38504bb31e92SArne Jansen .readpage_io_failed_hook = btree_io_failed_hook, 38510b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 3852239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 3853239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 38540da5468fSChris Mason }; 3855