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> 285a0e3ad6STejun Heo #include <linux/slab.h> 29784b4e29SChris Mason #include <linux/migrate.h> 307a36ddecSDavid Sterba #include <linux/ratelimit.h> 316463fe58SStefan Behrens #include <linux/uuid.h> 32803b2f54SStefan Behrens #include <linux/semaphore.h> 337e75bf3fSDavid Sterba #include <asm/unaligned.h> 34eb60ceacSChris Mason #include "ctree.h" 35eb60ceacSChris Mason #include "disk-io.h" 360b947affSFilipe David Borba Manana #include "hash.h" 37e089f05cSChris Mason #include "transaction.h" 380f7d52f4SChris Mason #include "btrfs_inode.h" 390b86a832SChris Mason #include "volumes.h" 40db94535dSChris Mason #include "print-tree.h" 41925baeddSChris Mason #include "locking.h" 42e02119d5SChris Mason #include "tree-log.h" 43fa9c0d79SChris Mason #include "free-space-cache.h" 4470f6d82eSOmar Sandoval #include "free-space-tree.h" 45581bb050SLi Zefan #include "inode-map.h" 4621adbd5cSStefan Behrens #include "check-integrity.h" 47606686eeSJosef Bacik #include "rcu-string.h" 488dabb742SStefan Behrens #include "dev-replace.h" 4953b381b3SDavid Woodhouse #include "raid56.h" 505ac1d209SJeff Mahoney #include "sysfs.h" 51fcebe456SJosef Bacik #include "qgroup.h" 52ebb8765bSAnand Jain #include "compression.h" 53557ea5ddSQu Wenruo #include "tree-checker.h" 54fd708b81SJosef Bacik #include "ref-verify.h" 55eb60ceacSChris Mason 56de0022b9SJosef Bacik #ifdef CONFIG_X86 57de0022b9SJosef Bacik #include <asm/cpufeature.h> 58de0022b9SJosef Bacik #endif 59de0022b9SJosef Bacik 60319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 61319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 62319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 63319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 64319e4d06SQu Wenruo BTRFS_SUPER_FLAG_METADUMP) 65319e4d06SQu Wenruo 66e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops; 678b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 684df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root); 693d3a126aSDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info); 70143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 71acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 722ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 73143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 742ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 75acce952bSliubo struct extent_io_tree *dirty_pages, 76acce952bSliubo int mark); 772ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 78acce952bSliubo struct extent_io_tree *pinned_extents); 792ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 802ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 81ce9adaa5SChris Mason 82d352ac68SChris Mason /* 8397eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 8497eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 85d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 86d352ac68SChris Mason */ 8797eb6b69SDavid Sterba struct btrfs_end_io_wq { 88ce9adaa5SChris Mason struct bio *bio; 89ce9adaa5SChris Mason bio_end_io_t *end_io; 90ce9adaa5SChris Mason void *private; 91ce9adaa5SChris Mason struct btrfs_fs_info *info; 924e4cbee9SChristoph Hellwig blk_status_t status; 93bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 948b712842SChris Mason struct btrfs_work work; 95ce9adaa5SChris Mason }; 960da5468fSChris Mason 9797eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 9897eb6b69SDavid Sterba 9997eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 10097eb6b69SDavid Sterba { 10197eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 10297eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 10397eb6b69SDavid Sterba 0, 104fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 10597eb6b69SDavid Sterba NULL); 10697eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 10797eb6b69SDavid Sterba return -ENOMEM; 10897eb6b69SDavid Sterba return 0; 10997eb6b69SDavid Sterba } 11097eb6b69SDavid Sterba 11197eb6b69SDavid Sterba void btrfs_end_io_wq_exit(void) 11297eb6b69SDavid Sterba { 11397eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 11497eb6b69SDavid Sterba } 11597eb6b69SDavid Sterba 116d352ac68SChris Mason /* 117d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 118d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 119d352ac68SChris Mason * just before they are sent down the IO stack. 120d352ac68SChris Mason */ 12144b8bd7eSChris Mason struct async_submit_bio { 122c6100a4bSJosef Bacik void *private_data; 123c6100a4bSJosef Bacik struct btrfs_fs_info *fs_info; 12444b8bd7eSChris Mason struct bio *bio; 1254a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start; 1264a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done; 12744b8bd7eSChris Mason int mirror_num; 128c8b97818SChris Mason unsigned long bio_flags; 129eaf25d93SChris Mason /* 130eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 131eaf25d93SChris Mason * can't tell us where in the file the bio should go 132eaf25d93SChris Mason */ 133eaf25d93SChris Mason u64 bio_offset; 1348b712842SChris Mason struct btrfs_work work; 1354e4cbee9SChristoph Hellwig blk_status_t status; 13644b8bd7eSChris Mason }; 13744b8bd7eSChris Mason 13885d4e461SChris Mason /* 13985d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 14085d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 14185d4e461SChris Mason * the level the eb occupies in the tree. 1424008c04aSChris Mason * 14385d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 14485d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 14585d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 14685d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 14785d4e461SChris Mason * have separate keysets. 1484008c04aSChris Mason * 14985d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 15085d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 15185d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1524008c04aSChris Mason * 15385d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 15485d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 15585d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 15685d4e461SChris Mason * 15785d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 15885d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 15985d4e461SChris Mason * needs update as well. 1604008c04aSChris Mason */ 1614008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1624008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1634008c04aSChris Mason # error 1644008c04aSChris Mason # endif 16585d4e461SChris Mason 16685d4e461SChris Mason static struct btrfs_lockdep_keyset { 16785d4e461SChris Mason u64 id; /* root objectid */ 16885d4e461SChris Mason const char *name_stem; /* lock name stem */ 16985d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 17085d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 17185d4e461SChris Mason } btrfs_lockdep_keysets[] = { 17285d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 17385d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 17485d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 17585d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 17685d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 17785d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 17860b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 17985d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 18085d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 18185d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 18213fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1836b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 18485d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1854008c04aSChris Mason }; 18685d4e461SChris Mason 18785d4e461SChris Mason void __init btrfs_init_lockdep(void) 18885d4e461SChris Mason { 18985d4e461SChris Mason int i, j; 19085d4e461SChris Mason 19185d4e461SChris Mason /* initialize lockdep class names */ 19285d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 19385d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 19485d4e461SChris Mason 19585d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 19685d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 19785d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 19885d4e461SChris Mason } 19985d4e461SChris Mason } 20085d4e461SChris Mason 20185d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 20285d4e461SChris Mason int level) 20385d4e461SChris Mason { 20485d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 20585d4e461SChris Mason 20685d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 20785d4e461SChris Mason 20885d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 20985d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 21085d4e461SChris Mason if (ks->id == objectid) 21185d4e461SChris Mason break; 21285d4e461SChris Mason 21385d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 21485d4e461SChris Mason &ks->keys[level], ks->names[level]); 21585d4e461SChris Mason } 21685d4e461SChris Mason 2174008c04aSChris Mason #endif 2184008c04aSChris Mason 219d352ac68SChris Mason /* 220d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 221d352ac68SChris Mason * that covers the entire device 222d352ac68SChris Mason */ 2236af49dbdSDavid Sterba struct extent_map *btree_get_extent(struct btrfs_inode *inode, 224306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 2255f39d397SChris Mason int create) 2265f39d397SChris Mason { 227fc4f21b1SNikolay Borisov struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb); 228fc4f21b1SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 2295f39d397SChris Mason struct extent_map *em; 2305f39d397SChris Mason int ret; 2315f39d397SChris Mason 232890871beSChris Mason read_lock(&em_tree->lock); 233d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 234a061fc8dSChris Mason if (em) { 2350b246afaSJeff Mahoney em->bdev = fs_info->fs_devices->latest_bdev; 236890871beSChris Mason read_unlock(&em_tree->lock); 2375f39d397SChris Mason goto out; 238a061fc8dSChris Mason } 239890871beSChris Mason read_unlock(&em_tree->lock); 2407b13b7b1SChris Mason 241172ddd60SDavid Sterba em = alloc_extent_map(); 2425f39d397SChris Mason if (!em) { 2435f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2445f39d397SChris Mason goto out; 2455f39d397SChris Mason } 2465f39d397SChris Mason em->start = 0; 2470afbaf8cSChris Mason em->len = (u64)-1; 248c8b97818SChris Mason em->block_len = (u64)-1; 2495f39d397SChris Mason em->block_start = 0; 2500b246afaSJeff Mahoney em->bdev = fs_info->fs_devices->latest_bdev; 251d1310b2eSChris Mason 252890871beSChris Mason write_lock(&em_tree->lock); 25309a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2545f39d397SChris Mason if (ret == -EEXIST) { 2555f39d397SChris Mason free_extent_map(em); 2567b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 257b4f359abSTsutomu Itoh if (!em) 2580433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2595f39d397SChris Mason } else if (ret) { 2607b13b7b1SChris Mason free_extent_map(em); 2610433f20dSTsutomu Itoh em = ERR_PTR(ret); 2625f39d397SChris Mason } 263890871beSChris Mason write_unlock(&em_tree->lock); 2647b13b7b1SChris Mason 2655f39d397SChris Mason out: 2665f39d397SChris Mason return em; 2675f39d397SChris Mason } 2685f39d397SChris Mason 2699ed57367SDavid Sterba u32 btrfs_csum_data(const char *data, u32 seed, size_t len) 27019c00ddcSChris Mason { 2710b947affSFilipe David Borba Manana return btrfs_crc32c(seed, data, len); 27219c00ddcSChris Mason } 27319c00ddcSChris Mason 2740b5e3dafSDomagoj Tršan void btrfs_csum_final(u32 crc, u8 *result) 27519c00ddcSChris Mason { 2767e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 27719c00ddcSChris Mason } 27819c00ddcSChris Mason 279d352ac68SChris Mason /* 280d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 281d352ac68SChris Mason * into the csum field of the block. 282d352ac68SChris Mason */ 28301d58472SDaniel Dressler static int csum_tree_block(struct btrfs_fs_info *fs_info, 28401d58472SDaniel Dressler struct extent_buffer *buf, 28519c00ddcSChris Mason int verify) 28619c00ddcSChris Mason { 28701d58472SDaniel Dressler u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 28871a63551SDavid Sterba char result[BTRFS_CSUM_SIZE]; 28919c00ddcSChris Mason unsigned long len; 29019c00ddcSChris Mason unsigned long cur_len; 29119c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 29219c00ddcSChris Mason char *kaddr; 29319c00ddcSChris Mason unsigned long map_start; 29419c00ddcSChris Mason unsigned long map_len; 29519c00ddcSChris Mason int err; 29619c00ddcSChris Mason u32 crc = ~(u32)0; 29719c00ddcSChris Mason 29819c00ddcSChris Mason len = buf->len - offset; 29919c00ddcSChris Mason while (len > 0) { 30019c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 301a6591715SChris Mason &kaddr, &map_start, &map_len); 302d397712bSChris Mason if (err) 3038bd98f0eSAlex Lyakas return err; 30419c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 305b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 30619c00ddcSChris Mason crc, cur_len); 30719c00ddcSChris Mason len -= cur_len; 30819c00ddcSChris Mason offset += cur_len; 30919c00ddcSChris Mason } 31071a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 311607d432dSJosef Bacik 31219c00ddcSChris Mason btrfs_csum_final(crc, result); 31319c00ddcSChris Mason 31419c00ddcSChris Mason if (verify) { 315607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 316e4204dedSChris Mason u32 val; 317e4204dedSChris Mason u32 found = 0; 318607d432dSJosef Bacik memcpy(&found, result, csum_size); 319e4204dedSChris Mason 320607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 32194647322SDavid Sterba btrfs_warn_rl(fs_info, 3225d163e0eSJeff Mahoney "%s checksum verify failed on %llu wanted %X found %X level %d", 32301d58472SDaniel Dressler fs_info->sb->s_id, buf->start, 324c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 3258bd98f0eSAlex Lyakas return -EUCLEAN; 32619c00ddcSChris Mason } 32719c00ddcSChris Mason } else { 328607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 32919c00ddcSChris Mason } 33071a63551SDavid Sterba 33119c00ddcSChris Mason return 0; 33219c00ddcSChris Mason } 33319c00ddcSChris Mason 334d352ac68SChris Mason /* 335d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 336d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 337d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 338d352ac68SChris Mason * in the wrong place. 339d352ac68SChris Mason */ 3401259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 341b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 342b9fab919SChris Mason int atomic) 3431259ab75SChris Mason { 3442ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3451259ab75SChris Mason int ret; 3462755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 3471259ab75SChris Mason 3481259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3491259ab75SChris Mason return 0; 3501259ab75SChris Mason 351b9fab919SChris Mason if (atomic) 352b9fab919SChris Mason return -EAGAIN; 353b9fab919SChris Mason 354a26e8c9fSJosef Bacik if (need_lock) { 355a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 356a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 357a26e8c9fSJosef Bacik } 358a26e8c9fSJosef Bacik 3592ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 360ff13db41SDavid Sterba &cached_state); 3610b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3621259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3631259ab75SChris Mason ret = 0; 3641259ab75SChris Mason goto out; 3651259ab75SChris Mason } 36694647322SDavid Sterba btrfs_err_rl(eb->fs_info, 36794647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 36894647322SDavid Sterba eb->start, 36929549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3701259ab75SChris Mason ret = 1; 371a26e8c9fSJosef Bacik 372a26e8c9fSJosef Bacik /* 373a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 374a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 37501327610SNicholas D Steeves * block that has been freed and re-allocated. So don't clear uptodate 376a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 377a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 378a26e8c9fSJosef Bacik * making everybody very sad. 379a26e8c9fSJosef Bacik */ 380a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3810b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 38233958dc6SChris Mason out: 3832ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 3842ac55d41SJosef Bacik &cached_state, GFP_NOFS); 385472b909fSJosef Bacik if (need_lock) 386a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3871259ab75SChris Mason return ret; 3881259ab75SChris Mason } 3891259ab75SChris Mason 390d352ac68SChris Mason /* 3911104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3921104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3931104a885SDavid Sterba */ 394ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 395ab8d0fc4SJeff Mahoney char *raw_disk_sb) 3961104a885SDavid Sterba { 3971104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3981104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 3991104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 4001104a885SDavid Sterba int ret = 0; 4011104a885SDavid Sterba 4021104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 4031104a885SDavid Sterba u32 crc = ~(u32)0; 4041104a885SDavid Sterba const int csum_size = sizeof(crc); 4051104a885SDavid Sterba char result[csum_size]; 4061104a885SDavid Sterba 4071104a885SDavid Sterba /* 4081104a885SDavid Sterba * The super_block structure does not span the whole 4091104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 41001327610SNicholas D Steeves * is filled with zeros and is included in the checksum. 4111104a885SDavid Sterba */ 4121104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 4131104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 4141104a885SDavid Sterba btrfs_csum_final(crc, result); 4151104a885SDavid Sterba 4161104a885SDavid Sterba if (memcmp(raw_disk_sb, result, csum_size)) 4171104a885SDavid Sterba ret = 1; 4181104a885SDavid Sterba } 4191104a885SDavid Sterba 4201104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 421ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "unsupported checksum algorithm %u", 4221104a885SDavid Sterba csum_type); 4231104a885SDavid Sterba ret = 1; 4241104a885SDavid Sterba } 4251104a885SDavid Sterba 4261104a885SDavid Sterba return ret; 4271104a885SDavid Sterba } 4281104a885SDavid Sterba 4291104a885SDavid Sterba /* 430d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 431d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 432d352ac68SChris Mason */ 4332ff7e61eSJeff Mahoney static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info, 434f188591eSChris Mason struct extent_buffer *eb, 4358436ea91SJosef Bacik u64 parent_transid) 436f188591eSChris Mason { 437f188591eSChris Mason struct extent_io_tree *io_tree; 438ea466794SJosef Bacik int failed = 0; 439f188591eSChris Mason int ret; 440f188591eSChris Mason int num_copies = 0; 441f188591eSChris Mason int mirror_num = 0; 442ea466794SJosef Bacik int failed_mirror = 0; 443f188591eSChris Mason 444a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 4450b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 446f188591eSChris Mason while (1) { 4478436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, 4486af49dbdSDavid Sterba mirror_num); 449256dd1bbSStefan Behrens if (!ret) { 450256dd1bbSStefan Behrens if (!verify_parent_transid(io_tree, eb, 451b9fab919SChris Mason parent_transid, 0)) 452ea466794SJosef Bacik break; 453256dd1bbSStefan Behrens else 454256dd1bbSStefan Behrens ret = -EIO; 455256dd1bbSStefan Behrens } 456d397712bSChris Mason 457a826d6dcSJosef Bacik /* 458a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 459a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 460a826d6dcSJosef Bacik * any less wrong. 461a826d6dcSJosef Bacik */ 462a826d6dcSJosef Bacik if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 463ea466794SJosef Bacik break; 464ea466794SJosef Bacik 4650b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 466f188591eSChris Mason eb->start, eb->len); 4674235298eSChris Mason if (num_copies == 1) 468ea466794SJosef Bacik break; 4694235298eSChris Mason 4705cf1ab56SJosef Bacik if (!failed_mirror) { 4715cf1ab56SJosef Bacik failed = 1; 4725cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4735cf1ab56SJosef Bacik } 4745cf1ab56SJosef Bacik 475f188591eSChris Mason mirror_num++; 476ea466794SJosef Bacik if (mirror_num == failed_mirror) 477ea466794SJosef Bacik mirror_num++; 478ea466794SJosef Bacik 4794235298eSChris Mason if (mirror_num > num_copies) 480ea466794SJosef Bacik break; 481f188591eSChris Mason } 482ea466794SJosef Bacik 483c0901581SStefan Behrens if (failed && !ret && failed_mirror) 4842ff7e61eSJeff Mahoney repair_eb_io_failure(fs_info, eb, failed_mirror); 485ea466794SJosef Bacik 486ea466794SJosef Bacik return ret; 487f188591eSChris Mason } 48819c00ddcSChris Mason 489d352ac68SChris Mason /* 490d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 491d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 492d352ac68SChris Mason */ 493d397712bSChris Mason 49401d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 49519c00ddcSChris Mason { 4964eee4fa4SMiao Xie u64 start = page_offset(page); 49719c00ddcSChris Mason u64 found_start; 49819c00ddcSChris Mason struct extent_buffer *eb; 499f188591eSChris Mason 5004f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5014f2de97aSJosef Bacik if (page != eb->pages[0]) 5024f2de97aSJosef Bacik return 0; 5030f805531SAlex Lyakas 50419c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 5050f805531SAlex Lyakas /* 5060f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5070f805531SAlex Lyakas * It is useful to know what went wrong. 5080f805531SAlex Lyakas */ 5090f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5100f805531SAlex Lyakas return -EUCLEAN; 5110f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5120f805531SAlex Lyakas return -EUCLEAN; 5130f805531SAlex Lyakas 5140f805531SAlex Lyakas ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, 5150f805531SAlex Lyakas btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); 5160f805531SAlex Lyakas 5178bd98f0eSAlex Lyakas return csum_tree_block(fs_info, eb, 0); 51819c00ddcSChris Mason } 51919c00ddcSChris Mason 52001d58472SDaniel Dressler static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, 5212b82032cSYan Zheng struct extent_buffer *eb) 5222b82032cSYan Zheng { 52301d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 52444880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 5252b82032cSYan Zheng int ret = 1; 5262b82032cSYan Zheng 5270a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5282b82032cSYan Zheng while (fs_devices) { 5292b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5302b82032cSYan Zheng ret = 0; 5312b82032cSYan Zheng break; 5322b82032cSYan Zheng } 5332b82032cSYan Zheng fs_devices = fs_devices->seed; 5342b82032cSYan Zheng } 5352b82032cSYan Zheng return ret; 5362b82032cSYan Zheng } 5372b82032cSYan Zheng 538facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 539facc8a22SMiao Xie u64 phy_offset, struct page *page, 540facc8a22SMiao Xie u64 start, u64 end, int mirror) 541ce9adaa5SChris Mason { 542ce9adaa5SChris Mason u64 found_start; 543ce9adaa5SChris Mason int found_level; 544ce9adaa5SChris Mason struct extent_buffer *eb; 545ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 54602873e43SZhao Lei struct btrfs_fs_info *fs_info = root->fs_info; 547f188591eSChris Mason int ret = 0; 548727011e0SChris Mason int reads_done; 549ce9adaa5SChris Mason 550ce9adaa5SChris Mason if (!page->private) 551ce9adaa5SChris Mason goto out; 552d397712bSChris Mason 5534f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 554d397712bSChris Mason 5550b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 5560b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 5570b32f4bbSJosef Bacik */ 5580b32f4bbSJosef Bacik extent_buffer_get(eb); 5590b32f4bbSJosef Bacik 5600b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 561727011e0SChris Mason if (!reads_done) 562727011e0SChris Mason goto err; 563f188591eSChris Mason 5645cf1ab56SJosef Bacik eb->read_mirror = mirror; 565656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 566ea466794SJosef Bacik ret = -EIO; 567ea466794SJosef Bacik goto err; 568ea466794SJosef Bacik } 569ea466794SJosef Bacik 570ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 571727011e0SChris Mason if (found_start != eb->start) { 57202873e43SZhao Lei btrfs_err_rl(fs_info, "bad tree block start %llu %llu", 57394647322SDavid Sterba found_start, eb->start); 574f188591eSChris Mason ret = -EIO; 575ce9adaa5SChris Mason goto err; 576ce9adaa5SChris Mason } 57702873e43SZhao Lei if (check_tree_block_fsid(fs_info, eb)) { 57802873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 57994647322SDavid Sterba eb->start); 5801259ab75SChris Mason ret = -EIO; 5811259ab75SChris Mason goto err; 5821259ab75SChris Mason } 583ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 5841c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 58502873e43SZhao Lei btrfs_err(fs_info, "bad tree block level %d", 5861c24c3ceSJosef Bacik (int)btrfs_header_level(eb)); 5871c24c3ceSJosef Bacik ret = -EIO; 5881c24c3ceSJosef Bacik goto err; 5891c24c3ceSJosef Bacik } 590ce9adaa5SChris Mason 59185d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 59285d4e461SChris Mason eb, found_level); 5934008c04aSChris Mason 59402873e43SZhao Lei ret = csum_tree_block(fs_info, eb, 1); 5958bd98f0eSAlex Lyakas if (ret) 596a826d6dcSJosef Bacik goto err; 597a826d6dcSJosef Bacik 598a826d6dcSJosef Bacik /* 599a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 600a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 601a826d6dcSJosef Bacik * return -EIO. 602a826d6dcSJosef Bacik */ 60369fc6cbbSQu Wenruo if (found_level == 0 && btrfs_check_leaf_full(root, eb)) { 604a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 605a826d6dcSJosef Bacik ret = -EIO; 606a826d6dcSJosef Bacik } 607ce9adaa5SChris Mason 608557ea5ddSQu Wenruo if (found_level > 0 && btrfs_check_node(root, eb)) 609053ab70fSLiu Bo ret = -EIO; 610053ab70fSLiu Bo 6110b32f4bbSJosef Bacik if (!ret) 6120b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 613ce9adaa5SChris Mason err: 61479fb65a1SJosef Bacik if (reads_done && 61579fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 616d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6174bb31e92SArne Jansen 61853b381b3SDavid Woodhouse if (ret) { 61953b381b3SDavid Woodhouse /* 62053b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 62153b381b3SDavid Woodhouse * again, we have to make sure it has something 62253b381b3SDavid Woodhouse * to decrement 62353b381b3SDavid Woodhouse */ 62453b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6250b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 62653b381b3SDavid Woodhouse } 6270b32f4bbSJosef Bacik free_extent_buffer(eb); 628ce9adaa5SChris Mason out: 629f188591eSChris Mason return ret; 630ce9adaa5SChris Mason } 631ce9adaa5SChris Mason 632ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6334bb31e92SArne Jansen { 6344bb31e92SArne Jansen struct extent_buffer *eb; 6354bb31e92SArne Jansen 6364f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 637656f30dbSFilipe Manana set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); 6385cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 63953b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 640ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 641d48d71aaSDavid Sterba btree_readahead_hook(eb, -EIO); 6424bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6434bb31e92SArne Jansen } 6444bb31e92SArne Jansen 6454246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 646ce9adaa5SChris Mason { 64797eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 648ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6499e0af237SLiu Bo struct btrfs_workqueue *wq; 6509e0af237SLiu Bo btrfs_work_func_t func; 651ce9adaa5SChris Mason 652ce9adaa5SChris Mason fs_info = end_io_wq->info; 6534e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 654d20f7043SChris Mason 65537226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 6569e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 6579e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 6589e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 6599e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 6609e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 6619e0af237SLiu Bo func = btrfs_freespace_write_helper; 6629e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 6639e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 6649e0af237SLiu Bo func = btrfs_endio_raid56_helper; 665d20f7043SChris Mason } else { 6669e0af237SLiu Bo wq = fs_info->endio_write_workers; 6679e0af237SLiu Bo func = btrfs_endio_write_helper; 668d20f7043SChris Mason } 6699e0af237SLiu Bo } else { 6708b110e39SMiao Xie if (unlikely(end_io_wq->metadata == 6718b110e39SMiao Xie BTRFS_WQ_ENDIO_DIO_REPAIR)) { 6728b110e39SMiao Xie wq = fs_info->endio_repair_workers; 6738b110e39SMiao Xie func = btrfs_endio_repair_helper; 6748b110e39SMiao Xie } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 6759e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 6769e0af237SLiu Bo func = btrfs_endio_raid56_helper; 6779e0af237SLiu Bo } else if (end_io_wq->metadata) { 6789e0af237SLiu Bo wq = fs_info->endio_meta_workers; 6799e0af237SLiu Bo func = btrfs_endio_meta_helper; 6809e0af237SLiu Bo } else { 6819e0af237SLiu Bo wq = fs_info->endio_workers; 6829e0af237SLiu Bo func = btrfs_endio_helper; 6839e0af237SLiu Bo } 6849e0af237SLiu Bo } 6859e0af237SLiu Bo 6869e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 6879e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 688ce9adaa5SChris Mason } 689ce9adaa5SChris Mason 6904e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 691bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 6920b86a832SChris Mason { 69397eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 6948b110e39SMiao Xie 69597eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 696ce9adaa5SChris Mason if (!end_io_wq) 6974e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 698ce9adaa5SChris Mason 699ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 700ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 70122c59948SChris Mason end_io_wq->info = info; 7024e4cbee9SChristoph Hellwig end_io_wq->status = 0; 703ce9adaa5SChris Mason end_io_wq->bio = bio; 70422c59948SChris Mason end_io_wq->metadata = metadata; 705ce9adaa5SChris Mason 706ce9adaa5SChris Mason bio->bi_private = end_io_wq; 707ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 70822c59948SChris Mason return 0; 70922c59948SChris Mason } 71022c59948SChris Mason 711b64a2851SChris Mason unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 7124854ddd0SChris Mason { 7134854ddd0SChris Mason unsigned long limit = min_t(unsigned long, 7145cdc7ad3SQu Wenruo info->thread_pool_size, 7154854ddd0SChris Mason info->fs_devices->open_devices); 7164854ddd0SChris Mason return 256 * limit; 7174854ddd0SChris Mason } 7184854ddd0SChris Mason 7194a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7204a69a410SChris Mason { 7214a69a410SChris Mason struct async_submit_bio *async; 7224e4cbee9SChristoph Hellwig blk_status_t ret; 7234a69a410SChris Mason 7244a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 725c6100a4bSJosef Bacik ret = async->submit_bio_start(async->private_data, async->bio, 726eaf25d93SChris Mason async->mirror_num, async->bio_flags, 727eaf25d93SChris Mason async->bio_offset); 72879787eaaSJeff Mahoney if (ret) 7294e4cbee9SChristoph Hellwig async->status = ret; 7304a69a410SChris Mason } 7314a69a410SChris Mason 7324a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7338b712842SChris Mason { 7348b712842SChris Mason struct async_submit_bio *async; 7358b712842SChris Mason 7368b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7374854ddd0SChris Mason 738bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7394e4cbee9SChristoph Hellwig if (async->status) { 7404e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7414246a0b6SChristoph Hellwig bio_endio(async->bio); 74279787eaaSJeff Mahoney return; 74379787eaaSJeff Mahoney } 74479787eaaSJeff Mahoney 745c6100a4bSJosef Bacik async->submit_bio_done(async->private_data, async->bio, async->mirror_num, 74681a75f67SMike Christie async->bio_flags, async->bio_offset); 7474a69a410SChris Mason } 7484a69a410SChris Mason 7494a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7504a69a410SChris Mason { 7514a69a410SChris Mason struct async_submit_bio *async; 7524a69a410SChris Mason 7534a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7548b712842SChris Mason kfree(async); 7558b712842SChris Mason } 7568b712842SChris Mason 7578c27cb35SLinus Torvalds blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 758c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 759c6100a4bSJosef Bacik u64 bio_offset, void *private_data, 7604a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start, 7614a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done) 76244b8bd7eSChris Mason { 76344b8bd7eSChris Mason struct async_submit_bio *async; 76444b8bd7eSChris Mason 76544b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 76644b8bd7eSChris Mason if (!async) 7674e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 76844b8bd7eSChris Mason 769c6100a4bSJosef Bacik async->private_data = private_data; 770c6100a4bSJosef Bacik async->fs_info = fs_info; 77144b8bd7eSChris Mason async->bio = bio; 77244b8bd7eSChris Mason async->mirror_num = mirror_num; 7734a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7744a69a410SChris Mason async->submit_bio_done = submit_bio_done; 7754a69a410SChris Mason 7769e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 7775cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 7784a69a410SChris Mason 779c8b97818SChris Mason async->bio_flags = bio_flags; 780eaf25d93SChris Mason async->bio_offset = bio_offset; 7818c8bee1dSChris Mason 7824e4cbee9SChristoph Hellwig async->status = 0; 78379787eaaSJeff Mahoney 78467f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 7855cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 786d313d7a3SChris Mason 7875cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 78844b8bd7eSChris Mason return 0; 78944b8bd7eSChris Mason } 79044b8bd7eSChris Mason 7914e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 792ce3ed71aSChris Mason { 7932c30c71bSKent Overstreet struct bio_vec *bvec; 794ce3ed71aSChris Mason struct btrfs_root *root; 7952c30c71bSKent Overstreet int i, ret = 0; 796ce3ed71aSChris Mason 797c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 7982c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 799ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 80001d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 80179787eaaSJeff Mahoney if (ret) 80279787eaaSJeff Mahoney break; 803ce3ed71aSChris Mason } 8042c30c71bSKent Overstreet 8054e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 806ce3ed71aSChris Mason } 807ce3ed71aSChris Mason 8088c27cb35SLinus Torvalds static blk_status_t __btree_submit_bio_start(void *private_data, struct bio *bio, 80981a75f67SMike Christie int mirror_num, unsigned long bio_flags, 810eaf25d93SChris Mason u64 bio_offset) 81122c59948SChris Mason { 8128b712842SChris Mason /* 8138b712842SChris Mason * when we're called for a write, we're already in the async 8145443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8158b712842SChris Mason */ 81679787eaaSJeff Mahoney return btree_csum_one_bio(bio); 81722c59948SChris Mason } 81822c59948SChris Mason 8198c27cb35SLinus Torvalds static blk_status_t __btree_submit_bio_done(void *private_data, struct bio *bio, 820eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 821eaf25d93SChris Mason u64 bio_offset) 8224a69a410SChris Mason { 823c6100a4bSJosef Bacik struct inode *inode = private_data; 8244e4cbee9SChristoph Hellwig blk_status_t ret; 82561891923SStefan Behrens 8268b712842SChris Mason /* 8274a69a410SChris Mason * when we're called for a write, we're already in the async 8284a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 8298b712842SChris Mason */ 8302ff7e61eSJeff Mahoney ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1); 8314246a0b6SChristoph Hellwig if (ret) { 8324e4cbee9SChristoph Hellwig bio->bi_status = ret; 8334246a0b6SChristoph Hellwig bio_endio(bio); 8344246a0b6SChristoph Hellwig } 83561891923SStefan Behrens return ret; 8360b86a832SChris Mason } 8370b86a832SChris Mason 83818fdc679SLiu Bo static int check_async_write(struct btrfs_inode *bi) 839de0022b9SJosef Bacik { 8406300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8416300463bSLiu Bo return 0; 842de0022b9SJosef Bacik #ifdef CONFIG_X86 843bc696ca0SBorislav Petkov if (static_cpu_has(X86_FEATURE_XMM4_2)) 844de0022b9SJosef Bacik return 0; 845de0022b9SJosef Bacik #endif 846de0022b9SJosef Bacik return 1; 847de0022b9SJosef Bacik } 848de0022b9SJosef Bacik 8498c27cb35SLinus Torvalds static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, 850eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 851eaf25d93SChris Mason u64 bio_offset) 85244b8bd7eSChris Mason { 853c6100a4bSJosef Bacik struct inode *inode = private_data; 8540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 85518fdc679SLiu Bo int async = check_async_write(BTRFS_I(inode)); 8564e4cbee9SChristoph Hellwig blk_status_t ret; 857cad321adSChris Mason 85837226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 859cad321adSChris Mason /* 860cad321adSChris Mason * called for a read, do the setup so that checksum validation 861cad321adSChris Mason * can happen in the async kernel threads 862cad321adSChris Mason */ 8630b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8640b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8651d4284bdSChris Mason if (ret) 86661891923SStefan Behrens goto out_w_error; 8672ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 868de0022b9SJosef Bacik } else if (!async) { 869de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 870de0022b9SJosef Bacik if (ret) 87161891923SStefan Behrens goto out_w_error; 8722ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 87361891923SStefan Behrens } else { 874cad321adSChris Mason /* 87561891923SStefan Behrens * kthread helpers are used to submit writes so that 87661891923SStefan Behrens * checksumming can happen in parallel across all CPUs 877cad321adSChris Mason */ 878c6100a4bSJosef Bacik ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, 879c6100a4bSJosef Bacik bio_offset, private_data, 8804a69a410SChris Mason __btree_submit_bio_start, 8814a69a410SChris Mason __btree_submit_bio_done); 88244b8bd7eSChris Mason } 88344b8bd7eSChris Mason 8844246a0b6SChristoph Hellwig if (ret) 8854246a0b6SChristoph Hellwig goto out_w_error; 8864246a0b6SChristoph Hellwig return 0; 8874246a0b6SChristoph Hellwig 88861891923SStefan Behrens out_w_error: 8894e4cbee9SChristoph Hellwig bio->bi_status = ret; 8904246a0b6SChristoph Hellwig bio_endio(bio); 89161891923SStefan Behrens return ret; 89261891923SStefan Behrens } 89361891923SStefan Behrens 8943dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 895784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 896a6bc32b8SMel Gorman struct page *newpage, struct page *page, 897a6bc32b8SMel Gorman enum migrate_mode mode) 898784b4e29SChris Mason { 899784b4e29SChris Mason /* 900784b4e29SChris Mason * we can't safely write a btree page from here, 901784b4e29SChris Mason * we haven't done the locking hook 902784b4e29SChris Mason */ 903784b4e29SChris Mason if (PageDirty(page)) 904784b4e29SChris Mason return -EAGAIN; 905784b4e29SChris Mason /* 906784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 907784b4e29SChris Mason * We must have no buffers or drop them. 908784b4e29SChris Mason */ 909784b4e29SChris Mason if (page_has_private(page) && 910784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 911784b4e29SChris Mason return -EAGAIN; 912a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 913784b4e29SChris Mason } 9143dd1462eSJan Beulich #endif 915784b4e29SChris Mason 9160da5468fSChris Mason 9170da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9180da5468fSChris Mason struct writeback_control *wbc) 9190da5468fSChris Mason { 920e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 921e2d84521SMiao Xie int ret; 922e2d84521SMiao Xie 923d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 924448d640bSChris Mason 925448d640bSChris Mason if (wbc->for_kupdate) 926448d640bSChris Mason return 0; 927448d640bSChris Mason 928e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 929b9473439SChris Mason /* this is a bit racy, but that's ok */ 930e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 931e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 932e2d84521SMiao Xie if (ret < 0) 933793955bcSChris Mason return 0; 934793955bcSChris Mason } 9350b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9360da5468fSChris Mason } 9370da5468fSChris Mason 938b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9395f39d397SChris Mason { 940d1310b2eSChris Mason struct extent_io_tree *tree; 941d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9428ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 9435f39d397SChris Mason } 9445f39d397SChris Mason 94570dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9465f39d397SChris Mason { 94798509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 94898509cfcSChris Mason return 0; 9490c4e538bSDavid Sterba 950f7a52a40SDavid Sterba return try_release_extent_buffer(page); 951d98237b3SChris Mason } 952d98237b3SChris Mason 953d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 954d47992f8SLukas Czerner unsigned int length) 955d98237b3SChris Mason { 956d1310b2eSChris Mason struct extent_io_tree *tree; 957d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9585f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9595f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9609ad6b7bcSChris Mason if (PagePrivate(page)) { 961efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 962efe120a0SFrank Holton "page private not zero on page %llu", 963efe120a0SFrank Holton (unsigned long long)page_offset(page)); 9649ad6b7bcSChris Mason ClearPagePrivate(page); 9659ad6b7bcSChris Mason set_page_private(page, 0); 96609cbfeafSKirill A. Shutemov put_page(page); 9679ad6b7bcSChris Mason } 968d98237b3SChris Mason } 969d98237b3SChris Mason 9700b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9710b32f4bbSJosef Bacik { 972bb146eb2SJosef Bacik #ifdef DEBUG 9730b32f4bbSJosef Bacik struct extent_buffer *eb; 9740b32f4bbSJosef Bacik 9750b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9760b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9770b32f4bbSJosef Bacik BUG_ON(!eb); 9780b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9790b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 9800b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 981bb146eb2SJosef Bacik #endif 9820b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 9830b32f4bbSJosef Bacik } 9840b32f4bbSJosef Bacik 9857f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 986d98237b3SChris Mason .readpage = btree_readpage, 9870da5468fSChris Mason .writepages = btree_writepages, 9885f39d397SChris Mason .releasepage = btree_releasepage, 9895f39d397SChris Mason .invalidatepage = btree_invalidatepage, 9905a92bc88SChris Mason #ifdef CONFIG_MIGRATION 991784b4e29SChris Mason .migratepage = btree_migratepage, 9925a92bc88SChris Mason #endif 9930b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 994d98237b3SChris Mason }; 995123abc88SChris Mason 9962ff7e61eSJeff Mahoney void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) 997090d1875SChris Mason { 9985f39d397SChris Mason struct extent_buffer *buf = NULL; 9992ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1000090d1875SChris Mason 10012ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1002c871b0f2SLiu Bo if (IS_ERR(buf)) 10036197d86eSDavid Sterba return; 1004d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 10056af49dbdSDavid Sterba buf, WAIT_NONE, 0); 10065f39d397SChris Mason free_extent_buffer(buf); 1007090d1875SChris Mason } 1008090d1875SChris Mason 10092ff7e61eSJeff Mahoney int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, 1010ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1011ab0fff03SArne Jansen { 1012ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 10132ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1014ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1015ab0fff03SArne Jansen int ret; 1016ab0fff03SArne Jansen 10172ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1018c871b0f2SLiu Bo if (IS_ERR(buf)) 1019ab0fff03SArne Jansen return 0; 1020ab0fff03SArne Jansen 1021ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1022ab0fff03SArne Jansen 10238436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, 10246af49dbdSDavid Sterba mirror_num); 1025ab0fff03SArne Jansen if (ret) { 1026ab0fff03SArne Jansen free_extent_buffer(buf); 1027ab0fff03SArne Jansen return ret; 1028ab0fff03SArne Jansen } 1029ab0fff03SArne Jansen 1030ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1031ab0fff03SArne Jansen free_extent_buffer(buf); 1032ab0fff03SArne Jansen return -EIO; 10330b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1034ab0fff03SArne Jansen *eb = buf; 1035ab0fff03SArne Jansen } else { 1036ab0fff03SArne Jansen free_extent_buffer(buf); 1037ab0fff03SArne Jansen } 1038ab0fff03SArne Jansen return 0; 1039ab0fff03SArne Jansen } 1040ab0fff03SArne Jansen 10412ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10422ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10430308af44SDavid Sterba u64 bytenr) 10440999df54SChris Mason { 10450b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10460b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10470b246afaSJeff Mahoney return alloc_extent_buffer(fs_info, bytenr); 10480999df54SChris Mason } 10490999df54SChris Mason 10500999df54SChris Mason 1051e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1052e02119d5SChris Mason { 1053727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 10548aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1055e02119d5SChris Mason } 1056e02119d5SChris Mason 10573189ff77SJeff Layton void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1058e02119d5SChris Mason { 10593189ff77SJeff Layton filemap_fdatawait_range(buf->pages[0]->mapping, 10608aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1061e02119d5SChris Mason } 1062e02119d5SChris Mason 10632ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 1064ce86cd59SDavid Sterba u64 parent_transid) 1065e20d96d6SChris Mason { 10665f39d397SChris Mason struct extent_buffer *buf = NULL; 106719c00ddcSChris Mason int ret; 106819c00ddcSChris Mason 10692ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1070c871b0f2SLiu Bo if (IS_ERR(buf)) 1071c871b0f2SLiu Bo return buf; 1072e4204dedSChris Mason 10732ff7e61eSJeff Mahoney ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid); 10740f0fe8f7SFilipe David Borba Manana if (ret) { 10750f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 107664c043deSLiu Bo return ERR_PTR(ret); 10770f0fe8f7SFilipe David Borba Manana } 10785f39d397SChris Mason return buf; 1079ce9adaa5SChris Mason 1080eb60ceacSChris Mason } 1081eb60ceacSChris Mason 10827c302b49SDavid Sterba void clean_tree_block(struct btrfs_fs_info *fs_info, 10835f39d397SChris Mason struct extent_buffer *buf) 1084ed2ff2cbSChris Mason { 108555c69072SChris Mason if (btrfs_header_generation(buf) == 1086e2d84521SMiao Xie fs_info->running_transaction->transid) { 1087b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1088b4ce94deSChris Mason 1089b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1090104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1091e2d84521SMiao Xie -buf->len, 1092e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1093b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1094b9473439SChris Mason btrfs_set_lock_blocking(buf); 10950b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1096925baeddSChris Mason } 10975f39d397SChris Mason } 1098ed7b63ebSJosef Bacik } 10995f39d397SChris Mason 11008257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11018257b2dcSMiao Xie { 11028257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11038257b2dcSMiao Xie int ret; 11048257b2dcSMiao Xie 11058257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11068257b2dcSMiao Xie if (!writers) 11078257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11088257b2dcSMiao Xie 1109908c7f19STejun Heo ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); 11108257b2dcSMiao Xie if (ret < 0) { 11118257b2dcSMiao Xie kfree(writers); 11128257b2dcSMiao Xie return ERR_PTR(ret); 11138257b2dcSMiao Xie } 11148257b2dcSMiao Xie 11158257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11168257b2dcSMiao Xie return writers; 11178257b2dcSMiao Xie } 11188257b2dcSMiao Xie 11198257b2dcSMiao Xie static void 11208257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11218257b2dcSMiao Xie { 11228257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 11238257b2dcSMiao Xie kfree(writers); 11248257b2dcSMiao Xie } 11258257b2dcSMiao Xie 1126da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1127e20d96d6SChris Mason u64 objectid) 1128d97e63b6SChris Mason { 11297c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 1130cfaa7295SChris Mason root->node = NULL; 1131a28ec197SChris Mason root->commit_root = NULL; 113227cdeb70SMiao Xie root->state = 0; 1133d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11340b86a832SChris Mason 11350f7d52f4SChris Mason root->objectid = objectid; 11360f7d52f4SChris Mason root->last_trans = 0; 113713a8a7c8SYan, Zheng root->highest_objectid = 0; 1138eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1139199c2a9cSMiao Xie root->nr_ordered_extents = 0; 114058176a96SJosef Bacik root->name = NULL; 11416bef4d31SEric Paris root->inode_tree = RB_ROOT; 114216cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1143f0486c68SYan, Zheng root->block_rsv = NULL; 1144d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 11450b86a832SChris Mason 11460b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11475d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1148eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1149eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1150199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1151199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 11522ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11532ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1154d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 11555d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1156eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1157199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1158f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11592ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11602ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 1161a2135011SChris Mason mutex_init(&root->objectid_mutex); 1162e02119d5SChris Mason mutex_init(&root->log_mutex); 116331f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1164573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 11657237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11667237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11677237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11688b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11698b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11707237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11717237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11727237f183SYan Zheng atomic_set(&root->log_writers, 0); 11732ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11748a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 11750700cea7SElena Reshetova refcount_set(&root->refs, 1); 1176ea14b57fSDavid Sterba atomic_set(&root->will_be_snapshotted, 0); 1177ce0dcee6SGoldwyn Rodrigues atomic64_set(&root->qgroup_meta_rsv, 0); 11787237f183SYan Zheng root->log_transid = 0; 1179d1433debSMiao Xie root->log_transid_committed = -1; 1180257c62e1SChris Mason root->last_log_commit = 0; 11817c0260eeSJeff Mahoney if (!dummy) 1182c6100a4bSJosef Bacik extent_io_tree_init(&root->dirty_log_pages, NULL); 1183017e5369SChris Mason 11843768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 11853768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 11866702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 11877c0260eeSJeff Mahoney if (!dummy) 11883f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 118906ea65a3SJosef Bacik else 119006ea65a3SJosef Bacik root->defrag_trans_start = 0; 11914d775673SChris Mason root->root_key.objectid = objectid; 11920ee5dc67SAl Viro root->anon_dev = 0; 11938ea05e3aSAlexander Block 11945f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 11953768f368SChris Mason } 11963768f368SChris Mason 119774e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 119874e4d827SDavid Sterba gfp_t flags) 11996f07e42eSAl Viro { 120074e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12016f07e42eSAl Viro if (root) 12026f07e42eSAl Viro root->fs_info = fs_info; 12036f07e42eSAl Viro return root; 12046f07e42eSAl Viro } 12056f07e42eSAl Viro 120606ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 120706ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1208da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 120906ea65a3SJosef Bacik { 121006ea65a3SJosef Bacik struct btrfs_root *root; 121106ea65a3SJosef Bacik 12127c0260eeSJeff Mahoney if (!fs_info) 12137c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12147c0260eeSJeff Mahoney 12157c0260eeSJeff Mahoney root = btrfs_alloc_root(fs_info, GFP_KERNEL); 121606ea65a3SJosef Bacik if (!root) 121706ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1218da17066cSJeff Mahoney 1219b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1220da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 1221faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 122206ea65a3SJosef Bacik 122306ea65a3SJosef Bacik return root; 122406ea65a3SJosef Bacik } 122506ea65a3SJosef Bacik #endif 122606ea65a3SJosef Bacik 122720897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 122820897f5cSArne Jansen struct btrfs_fs_info *fs_info, 122920897f5cSArne Jansen u64 objectid) 123020897f5cSArne Jansen { 123120897f5cSArne Jansen struct extent_buffer *leaf; 123220897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 123320897f5cSArne Jansen struct btrfs_root *root; 123420897f5cSArne Jansen struct btrfs_key key; 123520897f5cSArne Jansen int ret = 0; 123633d85fdaSQu Wenruo uuid_le uuid = NULL_UUID_LE; 123720897f5cSArne Jansen 123874e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_KERNEL); 123920897f5cSArne Jansen if (!root) 124020897f5cSArne Jansen return ERR_PTR(-ENOMEM); 124120897f5cSArne Jansen 1242da17066cSJeff Mahoney __setup_root(root, fs_info, objectid); 124320897f5cSArne Jansen root->root_key.objectid = objectid; 124420897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 124520897f5cSArne Jansen root->root_key.offset = 0; 124620897f5cSArne Jansen 12474d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 124820897f5cSArne Jansen if (IS_ERR(leaf)) { 124920897f5cSArne Jansen ret = PTR_ERR(leaf); 12501dd05682STsutomu Itoh leaf = NULL; 125120897f5cSArne Jansen goto fail; 125220897f5cSArne Jansen } 125320897f5cSArne Jansen 1254b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 125520897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 125620897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 125720897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 125820897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 125920897f5cSArne Jansen root->node = leaf; 126020897f5cSArne Jansen 1261d24ee97bSDavid Sterba write_extent_buffer_fsid(leaf, fs_info->fsid); 1262d24ee97bSDavid Sterba write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 126320897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 126420897f5cSArne Jansen 126520897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 126627cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 126720897f5cSArne Jansen 126820897f5cSArne Jansen root->root_item.flags = 0; 126920897f5cSArne Jansen root->root_item.byte_limit = 0; 127020897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 127120897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 127220897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 127320897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 127420897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 127520897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 127620897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 127733d85fdaSQu Wenruo if (is_fstree(objectid)) 12786463fe58SStefan Behrens uuid_le_gen(&uuid); 12796463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 128020897f5cSArne Jansen root->root_item.drop_level = 0; 128120897f5cSArne Jansen 128220897f5cSArne Jansen key.objectid = objectid; 128320897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 128420897f5cSArne Jansen key.offset = 0; 128520897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 128620897f5cSArne Jansen if (ret) 128720897f5cSArne Jansen goto fail; 128820897f5cSArne Jansen 128920897f5cSArne Jansen btrfs_tree_unlock(leaf); 129020897f5cSArne Jansen 129120897f5cSArne Jansen return root; 12921dd05682STsutomu Itoh 12931dd05682STsutomu Itoh fail: 12941dd05682STsutomu Itoh if (leaf) { 12951dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 129659885b39STsutomu Itoh free_extent_buffer(root->commit_root); 12971dd05682STsutomu Itoh free_extent_buffer(leaf); 12981dd05682STsutomu Itoh } 12991dd05682STsutomu Itoh kfree(root); 13001dd05682STsutomu Itoh 13011dd05682STsutomu Itoh return ERR_PTR(ret); 130220897f5cSArne Jansen } 130320897f5cSArne Jansen 13047237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1305e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13060f7d52f4SChris Mason { 13070f7d52f4SChris Mason struct btrfs_root *root; 13087237f183SYan Zheng struct extent_buffer *leaf; 1309e02119d5SChris Mason 131074e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1311e02119d5SChris Mason if (!root) 13127237f183SYan Zheng return ERR_PTR(-ENOMEM); 1313e02119d5SChris Mason 1314da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1315e02119d5SChris Mason 1316e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1317e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1318e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 131927cdeb70SMiao Xie 13207237f183SYan Zheng /* 132127cdeb70SMiao Xie * DON'T set REF_COWS for log trees 132227cdeb70SMiao Xie * 13237237f183SYan Zheng * log trees do not get reference counted because they go away 13247237f183SYan Zheng * before a real commit is actually done. They do store pointers 13257237f183SYan Zheng * to file data extents, and those reference counts still get 13267237f183SYan Zheng * updated (along with back refs to the log tree). 13277237f183SYan Zheng */ 1328e02119d5SChris Mason 13294d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13304d75f8a9SDavid Sterba NULL, 0, 0, 0); 13317237f183SYan Zheng if (IS_ERR(leaf)) { 13327237f183SYan Zheng kfree(root); 13337237f183SYan Zheng return ERR_CAST(leaf); 13347237f183SYan Zheng } 1335e02119d5SChris Mason 1336b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 13375d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 13385d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 13395d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 13405d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 13417237f183SYan Zheng root->node = leaf; 1342e02119d5SChris Mason 13430b246afaSJeff Mahoney write_extent_buffer_fsid(root->node, fs_info->fsid); 1344e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1345e02119d5SChris Mason btrfs_tree_unlock(root->node); 13467237f183SYan Zheng return root; 13477237f183SYan Zheng } 13487237f183SYan Zheng 13497237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13507237f183SYan Zheng struct btrfs_fs_info *fs_info) 13517237f183SYan Zheng { 13527237f183SYan Zheng struct btrfs_root *log_root; 13537237f183SYan Zheng 13547237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13557237f183SYan Zheng if (IS_ERR(log_root)) 13567237f183SYan Zheng return PTR_ERR(log_root); 13577237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13587237f183SYan Zheng fs_info->log_root_tree = log_root; 13597237f183SYan Zheng return 0; 13607237f183SYan Zheng } 13617237f183SYan Zheng 13627237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13637237f183SYan Zheng struct btrfs_root *root) 13647237f183SYan Zheng { 13650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13667237f183SYan Zheng struct btrfs_root *log_root; 13677237f183SYan Zheng struct btrfs_inode_item *inode_item; 13687237f183SYan Zheng 13690b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13707237f183SYan Zheng if (IS_ERR(log_root)) 13717237f183SYan Zheng return PTR_ERR(log_root); 13727237f183SYan Zheng 13737237f183SYan Zheng log_root->last_trans = trans->transid; 13747237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13757237f183SYan Zheng 13767237f183SYan Zheng inode_item = &log_root->root_item.inode; 13773cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13783cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13793cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1380da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13810b246afaSJeff Mahoney fs_info->nodesize); 13823cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13837237f183SYan Zheng 13845d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13857237f183SYan Zheng 13867237f183SYan Zheng WARN_ON(root->log_root); 13877237f183SYan Zheng root->log_root = log_root; 13887237f183SYan Zheng root->log_transid = 0; 1389d1433debSMiao Xie root->log_transid_committed = -1; 1390257c62e1SChris Mason root->last_log_commit = 0; 1391e02119d5SChris Mason return 0; 1392e02119d5SChris Mason } 1393e02119d5SChris Mason 139435a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1395cb517eabSMiao Xie struct btrfs_key *key) 1396e02119d5SChris Mason { 1397e02119d5SChris Mason struct btrfs_root *root; 1398e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 13990f7d52f4SChris Mason struct btrfs_path *path; 140084234f3aSYan Zheng u64 generation; 1401cb517eabSMiao Xie int ret; 1402cb517eabSMiao Xie 1403cb517eabSMiao Xie path = btrfs_alloc_path(); 1404cb517eabSMiao Xie if (!path) 1405cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 14060f7d52f4SChris Mason 140774e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1408cb517eabSMiao Xie if (!root) { 1409cb517eabSMiao Xie ret = -ENOMEM; 1410cb517eabSMiao Xie goto alloc_fail; 14110f7d52f4SChris Mason } 14120f7d52f4SChris Mason 1413da17066cSJeff Mahoney __setup_root(root, fs_info, key->objectid); 14140f7d52f4SChris Mason 1415cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1416cb517eabSMiao Xie &root->root_item, &root->root_key); 14170f7d52f4SChris Mason if (ret) { 141813a8a7c8SYan, Zheng if (ret > 0) 141913a8a7c8SYan, Zheng ret = -ENOENT; 1420cb517eabSMiao Xie goto find_fail; 14210f7d52f4SChris Mason } 142213a8a7c8SYan, Zheng 142384234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 14242ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 14252ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 1426ce86cd59SDavid Sterba generation); 142764c043deSLiu Bo if (IS_ERR(root->node)) { 142864c043deSLiu Bo ret = PTR_ERR(root->node); 1429cb517eabSMiao Xie goto find_fail; 1430cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1431cb517eabSMiao Xie ret = -EIO; 143264c043deSLiu Bo free_extent_buffer(root->node); 143364c043deSLiu Bo goto find_fail; 1434416bc658SJosef Bacik } 14355d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 143613a8a7c8SYan, Zheng out: 1437cb517eabSMiao Xie btrfs_free_path(path); 1438cb517eabSMiao Xie return root; 1439cb517eabSMiao Xie 1440cb517eabSMiao Xie find_fail: 1441cb517eabSMiao Xie kfree(root); 1442cb517eabSMiao Xie alloc_fail: 1443cb517eabSMiao Xie root = ERR_PTR(ret); 1444cb517eabSMiao Xie goto out; 1445cb517eabSMiao Xie } 1446cb517eabSMiao Xie 1447cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1448cb517eabSMiao Xie struct btrfs_key *location) 1449cb517eabSMiao Xie { 1450cb517eabSMiao Xie struct btrfs_root *root; 1451cb517eabSMiao Xie 1452cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1453cb517eabSMiao Xie if (IS_ERR(root)) 1454cb517eabSMiao Xie return root; 1455cb517eabSMiao Xie 1456cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 145727cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 145808fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 145908fe4db1SLi Zefan } 146013a8a7c8SYan, Zheng 14615eda7b5eSChris Mason return root; 14625eda7b5eSChris Mason } 14635eda7b5eSChris Mason 1464cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1465cb517eabSMiao Xie { 1466cb517eabSMiao Xie int ret; 14678257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1468cb517eabSMiao Xie 1469cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1470cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1471cb517eabSMiao Xie GFP_NOFS); 1472cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1473cb517eabSMiao Xie ret = -ENOMEM; 1474cb517eabSMiao Xie goto fail; 1475cb517eabSMiao Xie } 1476cb517eabSMiao Xie 14778257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 14788257b2dcSMiao Xie if (IS_ERR(writers)) { 14798257b2dcSMiao Xie ret = PTR_ERR(writers); 14808257b2dcSMiao Xie goto fail; 14818257b2dcSMiao Xie } 14828257b2dcSMiao Xie root->subv_writers = writers; 14838257b2dcSMiao Xie 1484cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 148557cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 148657cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1487cb517eabSMiao Xie 1488cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1489cb517eabSMiao Xie if (ret) 1490876d2cf1SLiu Bo goto fail; 1491f32e48e9SChandan Rajendra 1492f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1493f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1494f32e48e9SChandan Rajendra &root->highest_objectid); 1495f32e48e9SChandan Rajendra if (ret) { 1496f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1497876d2cf1SLiu Bo goto fail; 1498f32e48e9SChandan Rajendra } 1499f32e48e9SChandan Rajendra 1500f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1501f32e48e9SChandan Rajendra 1502f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1503f32e48e9SChandan Rajendra 1504cb517eabSMiao Xie return 0; 1505cb517eabSMiao Xie fail: 1506876d2cf1SLiu Bo /* the caller is responsible to call free_fs_root */ 1507cb517eabSMiao Xie return ret; 1508cb517eabSMiao Xie } 1509cb517eabSMiao Xie 151035bbb97fSJeff Mahoney struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1511cb517eabSMiao Xie u64 root_id) 1512cb517eabSMiao Xie { 1513cb517eabSMiao Xie struct btrfs_root *root; 1514cb517eabSMiao Xie 1515cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1516cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1517cb517eabSMiao Xie (unsigned long)root_id); 1518cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1519cb517eabSMiao Xie return root; 1520cb517eabSMiao Xie } 1521cb517eabSMiao Xie 1522cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1523cb517eabSMiao Xie struct btrfs_root *root) 1524cb517eabSMiao Xie { 1525cb517eabSMiao Xie int ret; 1526cb517eabSMiao Xie 1527e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1528cb517eabSMiao Xie if (ret) 1529cb517eabSMiao Xie return ret; 1530cb517eabSMiao Xie 1531cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1532cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1533cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1534cb517eabSMiao Xie root); 1535cb517eabSMiao Xie if (ret == 0) 153627cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1537cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1538cb517eabSMiao Xie radix_tree_preload_end(); 1539cb517eabSMiao Xie 1540cb517eabSMiao Xie return ret; 1541cb517eabSMiao Xie } 1542cb517eabSMiao Xie 1543c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1544c00869f1SMiao Xie struct btrfs_key *location, 1545c00869f1SMiao Xie bool check_ref) 15465eda7b5eSChris Mason { 15475eda7b5eSChris Mason struct btrfs_root *root; 1548381cf658SDavid Sterba struct btrfs_path *path; 15491d4c08e0SDavid Sterba struct btrfs_key key; 15505eda7b5eSChris Mason int ret; 15515eda7b5eSChris Mason 1552edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1553edbd8d4eSChris Mason return fs_info->tree_root; 1554edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1555edbd8d4eSChris Mason return fs_info->extent_root; 15568f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 15578f18cf13SChris Mason return fs_info->chunk_root; 15588f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 15598f18cf13SChris Mason return fs_info->dev_root; 15600403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 15610403e47eSYan Zheng return fs_info->csum_root; 1562bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1563bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1564bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1565f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1566f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1567f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 156870f6d82eSOmar Sandoval if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 156970f6d82eSOmar Sandoval return fs_info->free_space_root ? fs_info->free_space_root : 157070f6d82eSOmar Sandoval ERR_PTR(-ENOENT); 15714df27c4dSYan, Zheng again: 1572cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 157348475471SStefan Behrens if (root) { 1574c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 157548475471SStefan Behrens return ERR_PTR(-ENOENT); 15765eda7b5eSChris Mason return root; 157748475471SStefan Behrens } 15785eda7b5eSChris Mason 1579cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 15805eda7b5eSChris Mason if (IS_ERR(root)) 15815eda7b5eSChris Mason return root; 15823394e160SChris Mason 1583c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1584d68fc57bSYan, Zheng ret = -ENOENT; 1585d68fc57bSYan, Zheng goto fail; 1586d68fc57bSYan, Zheng } 1587d68fc57bSYan, Zheng 1588cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1589cb517eabSMiao Xie if (ret) 1590cb517eabSMiao Xie goto fail; 1591cb517eabSMiao Xie 1592381cf658SDavid Sterba path = btrfs_alloc_path(); 1593381cf658SDavid Sterba if (!path) { 1594381cf658SDavid Sterba ret = -ENOMEM; 1595381cf658SDavid Sterba goto fail; 1596381cf658SDavid Sterba } 15971d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 15981d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 15991d4c08e0SDavid Sterba key.offset = location->objectid; 16001d4c08e0SDavid Sterba 16011d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1602381cf658SDavid Sterba btrfs_free_path(path); 1603d68fc57bSYan, Zheng if (ret < 0) 1604d68fc57bSYan, Zheng goto fail; 1605d68fc57bSYan, Zheng if (ret == 0) 160627cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1607d68fc57bSYan, Zheng 1608cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16090f7d52f4SChris Mason if (ret) { 16104df27c4dSYan, Zheng if (ret == -EEXIST) { 16114df27c4dSYan, Zheng free_fs_root(root); 16124df27c4dSYan, Zheng goto again; 16130f7d52f4SChris Mason } 16144df27c4dSYan, Zheng goto fail; 16154df27c4dSYan, Zheng } 1616edbd8d4eSChris Mason return root; 16174df27c4dSYan, Zheng fail: 16184df27c4dSYan, Zheng free_fs_root(root); 16194df27c4dSYan, Zheng return ERR_PTR(ret); 1620edbd8d4eSChris Mason } 1621edbd8d4eSChris Mason 162204160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 162304160088SChris Mason { 162404160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 162504160088SChris Mason int ret = 0; 162604160088SChris Mason struct btrfs_device *device; 162704160088SChris Mason struct backing_dev_info *bdi; 1628b7967db7SChris Mason 16291f78160cSXiao Guangrong rcu_read_lock(); 16301f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1631dfe25020SChris Mason if (!device->bdev) 1632dfe25020SChris Mason continue; 1633efa7c9f9SJan Kara bdi = device->bdev->bd_bdi; 1634ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 163504160088SChris Mason ret = 1; 163604160088SChris Mason break; 163704160088SChris Mason } 163804160088SChris Mason } 16391f78160cSXiao Guangrong rcu_read_unlock(); 164004160088SChris Mason return ret; 164104160088SChris Mason } 164204160088SChris Mason 16438b712842SChris Mason /* 16448b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16458b712842SChris Mason * functions. This is where read checksum verification actually happens 16468b712842SChris Mason */ 16478b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1648ce9adaa5SChris Mason { 1649ce9adaa5SChris Mason struct bio *bio; 165097eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1651ce9adaa5SChris Mason 165297eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1653ce9adaa5SChris Mason bio = end_io_wq->bio; 16548b712842SChris Mason 16554e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1656ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1657ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 165897eb6b69SDavid Sterba kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 16594246a0b6SChristoph Hellwig bio_endio(bio); 1660ce9adaa5SChris Mason } 166144b8bd7eSChris Mason 1662a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1663a74a4b97SChris Mason { 1664a74a4b97SChris Mason struct btrfs_root *root = arg; 16650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1666d0278245SMiao Xie int again; 1667da288d28SFilipe Manana struct btrfs_trans_handle *trans; 1668a74a4b97SChris Mason 1669a74a4b97SChris Mason do { 1670d0278245SMiao Xie again = 0; 16719d1a2a3aSDavid Sterba 1672d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 16732ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1674d0278245SMiao Xie goto sleep; 1675d0278245SMiao Xie 167690c711abSZygo Blaxell /* 167790c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 167890c711abSZygo Blaxell * before we have finished mounting the filesystem. 167990c711abSZygo Blaxell */ 16800b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 168190c711abSZygo Blaxell goto sleep; 168290c711abSZygo Blaxell 16830b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1684d0278245SMiao Xie goto sleep; 1685d0278245SMiao Xie 1686dc7f370cSMiao Xie /* 1687dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1688dc7f370cSMiao Xie * during the above check and trylock. 1689dc7f370cSMiao Xie */ 16902ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 16910b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1692dc7f370cSMiao Xie goto sleep; 1693dc7f370cSMiao Xie } 1694dc7f370cSMiao Xie 16950b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_delayed_iput_mutex); 16962ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 16970b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); 1698c2d6cb16SFilipe Manana 16999d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 17000b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1701a74a4b97SChris Mason 1702d0278245SMiao Xie /* 170305323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 170405323cd1SMiao Xie * needn't do anything special here. 1705d0278245SMiao Xie */ 17060b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 170767c5e7d4SFilipe Manana 170867c5e7d4SFilipe Manana /* 170967c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 171067c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 171167c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 171267c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 171367c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 171467c5e7d4SFilipe Manana * unused block groups. 171567c5e7d4SFilipe Manana */ 17160b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1717d0278245SMiao Xie sleep: 1718838fe188SJiri Kosina if (!again) { 1719a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 17208929ecfaSYan, Zheng if (!kthread_should_stop()) 1721a74a4b97SChris Mason schedule(); 1722a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1723a74a4b97SChris Mason } 1724a74a4b97SChris Mason } while (!kthread_should_stop()); 1725da288d28SFilipe Manana 1726da288d28SFilipe Manana /* 1727da288d28SFilipe Manana * Transaction kthread is stopped before us and wakes us up. 1728da288d28SFilipe Manana * However we might have started a new transaction and COWed some 1729da288d28SFilipe Manana * tree blocks when deleting unused block groups for example. So 1730da288d28SFilipe Manana * make sure we commit the transaction we started to have a clean 1731da288d28SFilipe Manana * shutdown when evicting the btree inode - if it has dirty pages 1732da288d28SFilipe Manana * when we do the final iput() on it, eviction will trigger a 1733da288d28SFilipe Manana * writeback for it which will fail with null pointer dereferences 1734da288d28SFilipe Manana * since work queues and other resources were already released and 1735da288d28SFilipe Manana * destroyed by the time the iput/eviction/writeback is made. 1736da288d28SFilipe Manana */ 1737da288d28SFilipe Manana trans = btrfs_attach_transaction(root); 1738da288d28SFilipe Manana if (IS_ERR(trans)) { 1739da288d28SFilipe Manana if (PTR_ERR(trans) != -ENOENT) 17400b246afaSJeff Mahoney btrfs_err(fs_info, 1741da288d28SFilipe Manana "cleaner transaction attach returned %ld", 1742da288d28SFilipe Manana PTR_ERR(trans)); 1743da288d28SFilipe Manana } else { 1744da288d28SFilipe Manana int ret; 1745da288d28SFilipe Manana 17463a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 1747da288d28SFilipe Manana if (ret) 17480b246afaSJeff Mahoney btrfs_err(fs_info, 1749da288d28SFilipe Manana "cleaner open transaction commit returned %d", 1750da288d28SFilipe Manana ret); 1751da288d28SFilipe Manana } 1752da288d28SFilipe Manana 1753a74a4b97SChris Mason return 0; 1754a74a4b97SChris Mason } 1755a74a4b97SChris Mason 1756a74a4b97SChris Mason static int transaction_kthread(void *arg) 1757a74a4b97SChris Mason { 1758a74a4b97SChris Mason struct btrfs_root *root = arg; 17590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1760a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1761a74a4b97SChris Mason struct btrfs_transaction *cur; 17628929ecfaSYan, Zheng u64 transid; 1763a74a4b97SChris Mason unsigned long now; 1764a74a4b97SChris Mason unsigned long delay; 1765914b2007SJan Kara bool cannot_commit; 1766a74a4b97SChris Mason 1767a74a4b97SChris Mason do { 1768914b2007SJan Kara cannot_commit = false; 17690b246afaSJeff Mahoney delay = HZ * fs_info->commit_interval; 17700b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1771a74a4b97SChris Mason 17720b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 17730b246afaSJeff Mahoney cur = fs_info->running_transaction; 1774a74a4b97SChris Mason if (!cur) { 17750b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1776a74a4b97SChris Mason goto sleep; 1777a74a4b97SChris Mason } 177831153d81SYan Zheng 1779a74a4b97SChris Mason now = get_seconds(); 17804a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 17818b87dc17SDavid Sterba (now < cur->start_time || 17820b246afaSJeff Mahoney now - cur->start_time < fs_info->commit_interval)) { 17830b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1784a74a4b97SChris Mason delay = HZ * 5; 1785a74a4b97SChris Mason goto sleep; 1786a74a4b97SChris Mason } 17878929ecfaSYan, Zheng transid = cur->transid; 17880b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 178956bec294SChris Mason 179079787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1791354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1792914b2007SJan Kara if (IS_ERR(trans)) { 1793354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1794914b2007SJan Kara cannot_commit = true; 179579787eaaSJeff Mahoney goto sleep; 1796914b2007SJan Kara } 17978929ecfaSYan, Zheng if (transid == trans->transid) { 17983a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 17998929ecfaSYan, Zheng } else { 18003a45bb20SJeff Mahoney btrfs_end_transaction(trans); 18018929ecfaSYan, Zheng } 1802a74a4b97SChris Mason sleep: 18030b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 18040b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1805a74a4b97SChris Mason 18064e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18070b246afaSJeff Mahoney &fs_info->fs_state))) 18082ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 1809a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 18108929ecfaSYan, Zheng if (!kthread_should_stop() && 18110b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1812914b2007SJan Kara cannot_commit)) 1813a74a4b97SChris Mason schedule_timeout(delay); 1814a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1815a74a4b97SChris Mason } while (!kthread_should_stop()); 1816a74a4b97SChris Mason return 0; 1817a74a4b97SChris Mason } 1818a74a4b97SChris Mason 1819af31f5e5SChris Mason /* 1820af31f5e5SChris Mason * this will find the highest generation in the array of 1821af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1822af31f5e5SChris Mason * or -1 if we can't find anything. 1823af31f5e5SChris Mason * 1824af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1825af31f5e5SChris Mason * generation of the latest root in the array with the generation 1826af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1827af31f5e5SChris Mason */ 1828af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1829af31f5e5SChris Mason { 1830af31f5e5SChris Mason u64 cur; 1831af31f5e5SChris Mason int newest_index = -1; 1832af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1833af31f5e5SChris Mason int i; 1834af31f5e5SChris Mason 1835af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1836af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1837af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1838af31f5e5SChris Mason if (cur == newest_gen) 1839af31f5e5SChris Mason newest_index = i; 1840af31f5e5SChris Mason } 1841af31f5e5SChris Mason 1842af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1843af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1844af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1845af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1846af31f5e5SChris Mason if (cur == newest_gen) 1847af31f5e5SChris Mason newest_index = 0; 1848af31f5e5SChris Mason } 1849af31f5e5SChris Mason return newest_index; 1850af31f5e5SChris Mason } 1851af31f5e5SChris Mason 1852af31f5e5SChris Mason 1853af31f5e5SChris Mason /* 1854af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1855af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1856af31f5e5SChris Mason * field in the fs_info struct 1857af31f5e5SChris Mason */ 1858af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1859af31f5e5SChris Mason u64 newest_gen) 1860af31f5e5SChris Mason { 1861af31f5e5SChris Mason int newest_index = -1; 1862af31f5e5SChris Mason 1863af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1864af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1865af31f5e5SChris Mason if (newest_index == -1) { 1866af31f5e5SChris Mason info->backup_root_index = 0; 1867af31f5e5SChris Mason } else { 1868af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1869af31f5e5SChris Mason } 1870af31f5e5SChris Mason } 1871af31f5e5SChris Mason 1872af31f5e5SChris Mason /* 1873af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1874af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1875af31f5e5SChris Mason * done 1876af31f5e5SChris Mason */ 1877af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1878af31f5e5SChris Mason { 1879af31f5e5SChris Mason int next_backup; 1880af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1881af31f5e5SChris Mason int last_backup; 1882af31f5e5SChris Mason 1883af31f5e5SChris Mason next_backup = info->backup_root_index; 1884af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1885af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1886af31f5e5SChris Mason 1887af31f5e5SChris Mason /* 1888af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1889af31f5e5SChris Mason * this happens only at umount 1890af31f5e5SChris Mason */ 1891af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1892af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1893af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1894af31f5e5SChris Mason next_backup = last_backup; 1895af31f5e5SChris Mason 1896af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1897af31f5e5SChris Mason 1898af31f5e5SChris Mason /* 1899af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1900af31f5e5SChris Mason * regardless of which ones we use today 1901af31f5e5SChris Mason */ 1902af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1903af31f5e5SChris Mason 1904af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1905af31f5e5SChris Mason 1906af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1907af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1908af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1909af31f5e5SChris Mason 1910af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1911af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1912af31f5e5SChris Mason 1913af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1914af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1915af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1916af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1917af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1918af31f5e5SChris Mason 1919af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1920af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1921af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1922af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1923af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1924af31f5e5SChris Mason 19257c7e82a7SChris Mason /* 19267c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19277c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19287c7e82a7SChris Mason */ 19297c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19307c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19317c7e82a7SChris Mason info->fs_root->node->start); 1932af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1933af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1934af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1935af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19367c7e82a7SChris Mason } 1937af31f5e5SChris Mason 1938af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1939af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1940af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1941af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1942af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1943af31f5e5SChris Mason 1944af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1945af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1946af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1947af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1948af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1949af31f5e5SChris Mason 1950af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1951af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1952af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1953af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1954af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1955af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1956af31f5e5SChris Mason 1957af31f5e5SChris Mason /* 1958af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1959af31f5e5SChris Mason * for the next commit 1960af31f5e5SChris Mason */ 1961af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1962af31f5e5SChris Mason &info->super_for_commit->super_roots, 1963af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1964af31f5e5SChris Mason } 1965af31f5e5SChris Mason 1966af31f5e5SChris Mason /* 1967af31f5e5SChris Mason * this copies info out of the root backup array and back into 1968af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 1969af31f5e5SChris Mason * the array, so you send it the number of backups you've already 1970af31f5e5SChris Mason * tried and the last backup index you used. 1971af31f5e5SChris Mason * 1972af31f5e5SChris Mason * this returns -1 when it has tried all the backups 1973af31f5e5SChris Mason */ 1974af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 1975af31f5e5SChris Mason struct btrfs_super_block *super, 1976af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 1977af31f5e5SChris Mason { 1978af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1979af31f5e5SChris Mason int newest = *backup_index; 1980af31f5e5SChris Mason 1981af31f5e5SChris Mason if (*num_backups_tried == 0) { 1982af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 1983af31f5e5SChris Mason 1984af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 1985af31f5e5SChris Mason if (newest == -1) 1986af31f5e5SChris Mason return -1; 1987af31f5e5SChris Mason 1988af31f5e5SChris Mason *backup_index = newest; 1989af31f5e5SChris Mason *num_backups_tried = 1; 1990af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 1991af31f5e5SChris Mason /* we've tried all the backups, all done */ 1992af31f5e5SChris Mason return -1; 1993af31f5e5SChris Mason } else { 1994af31f5e5SChris Mason /* jump to the next oldest backup */ 1995af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 1996af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1997af31f5e5SChris Mason *backup_index = newest; 1998af31f5e5SChris Mason *num_backups_tried += 1; 1999af31f5e5SChris Mason } 2000af31f5e5SChris Mason root_backup = super->super_roots + newest; 2001af31f5e5SChris Mason 2002af31f5e5SChris Mason btrfs_set_super_generation(super, 2003af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2004af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2005af31f5e5SChris Mason btrfs_set_super_root_level(super, 2006af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2007af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2008af31f5e5SChris Mason 2009af31f5e5SChris Mason /* 2010af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2011af31f5e5SChris Mason * need a fsck 2012af31f5e5SChris Mason */ 2013af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2014af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2015af31f5e5SChris Mason return 0; 2016af31f5e5SChris Mason } 2017af31f5e5SChris Mason 20187abadb64SLiu Bo /* helper to cleanup workers */ 20197abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20207abadb64SLiu Bo { 2021dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2022afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20235cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2024fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2025fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 20268b110e39SMiao Xie btrfs_destroy_workqueue(fs_info->endio_repair_workers); 2027d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2028fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2029fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2030a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 20315b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2032e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2033736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2034a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2035fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2036a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 2037a9b9477dSFilipe Manana /* 2038a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2039a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2040a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2041a9b9477dSFilipe Manana */ 2042a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2043a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 20447abadb64SLiu Bo } 20457abadb64SLiu Bo 20462e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20472e9f5954SRashika { 20482e9f5954SRashika if (root) { 20492e9f5954SRashika free_extent_buffer(root->node); 20502e9f5954SRashika free_extent_buffer(root->commit_root); 20512e9f5954SRashika root->node = NULL; 20522e9f5954SRashika root->commit_root = NULL; 20532e9f5954SRashika } 20542e9f5954SRashika } 20552e9f5954SRashika 2056af31f5e5SChris Mason /* helper to cleanup tree roots */ 2057af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2058af31f5e5SChris Mason { 20592e9f5954SRashika free_root_extent_buffers(info->tree_root); 2060655b09feSJosef Bacik 20612e9f5954SRashika free_root_extent_buffers(info->dev_root); 20622e9f5954SRashika free_root_extent_buffers(info->extent_root); 20632e9f5954SRashika free_root_extent_buffers(info->csum_root); 20642e9f5954SRashika free_root_extent_buffers(info->quota_root); 20652e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20662e9f5954SRashika if (chunk_root) 20672e9f5954SRashika free_root_extent_buffers(info->chunk_root); 206870f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2069af31f5e5SChris Mason } 2070af31f5e5SChris Mason 2071faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2072171f6537SJosef Bacik { 2073171f6537SJosef Bacik int ret; 2074171f6537SJosef Bacik struct btrfs_root *gang[8]; 2075171f6537SJosef Bacik int i; 2076171f6537SJosef Bacik 2077171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2078171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2079171f6537SJosef Bacik struct btrfs_root, root_list); 2080171f6537SJosef Bacik list_del(&gang[0]->root_list); 2081171f6537SJosef Bacik 208227cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2083cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2084171f6537SJosef Bacik } else { 2085171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2086171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2087b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2088171f6537SJosef Bacik } 2089171f6537SJosef Bacik } 2090171f6537SJosef Bacik 2091171f6537SJosef Bacik while (1) { 2092171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2093171f6537SJosef Bacik (void **)gang, 0, 2094171f6537SJosef Bacik ARRAY_SIZE(gang)); 2095171f6537SJosef Bacik if (!ret) 2096171f6537SJosef Bacik break; 2097171f6537SJosef Bacik for (i = 0; i < ret; i++) 2098cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2099171f6537SJosef Bacik } 21001a4319ccSLiu Bo 21011a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 21021a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 21032ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 21041a4319ccSLiu Bo } 2105171f6537SJosef Bacik } 2106af31f5e5SChris Mason 2107638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2108638aa7edSEric Sandeen { 2109638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2110638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2111638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2112638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2113638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2114638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2115638aa7edSEric Sandeen fs_info->scrub_workers_refcnt = 0; 2116638aa7edSEric Sandeen } 2117638aa7edSEric Sandeen 2118779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2119779a65a4SEric Sandeen { 2120779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2121779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2122779a65a4SEric Sandeen atomic_set(&fs_info->balance_running, 0); 2123779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2124779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2125779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2126779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2127779a65a4SEric Sandeen } 2128779a65a4SEric Sandeen 21296bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2130f37938e0SEric Sandeen { 21312ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 21322ff7e61eSJeff Mahoney 21332ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 21342ff7e61eSJeff Mahoney set_nlink(inode, 1); 2135f37938e0SEric Sandeen /* 2136f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2137f37938e0SEric Sandeen * the real end of the address space is determined by all of 2138f37938e0SEric Sandeen * the devices in the system 2139f37938e0SEric Sandeen */ 21402ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 21412ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2142f37938e0SEric Sandeen 21432ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 2144c6100a4bSJosef Bacik extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); 21452ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.track_uptodate = 0; 21462ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2147f37938e0SEric Sandeen 21482ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; 2149f37938e0SEric Sandeen 21502ff7e61eSJeff Mahoney BTRFS_I(inode)->root = fs_info->tree_root; 21512ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21522ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21532ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2154f37938e0SEric Sandeen } 2155f37938e0SEric Sandeen 2156ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2157ad618368SEric Sandeen { 2158ad618368SEric Sandeen fs_info->dev_replace.lock_owner = 0; 2159ad618368SEric Sandeen atomic_set(&fs_info->dev_replace.nesting_level, 0); 2160ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 216173beece9SLiu Bo rwlock_init(&fs_info->dev_replace.lock); 216273beece9SLiu Bo atomic_set(&fs_info->dev_replace.read_locks, 0); 216373beece9SLiu Bo atomic_set(&fs_info->dev_replace.blocking_readers, 0); 2164ad618368SEric Sandeen init_waitqueue_head(&fs_info->replace_wait); 216573beece9SLiu Bo init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); 2166ad618368SEric Sandeen } 2167ad618368SEric Sandeen 2168f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2169f9e92e40SEric Sandeen { 2170f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2171f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2172f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2173f9e92e40SEric Sandeen fs_info->qgroup_op_tree = RB_ROOT; 2174f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2175f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2176f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2177d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2178f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2179f9e92e40SEric Sandeen } 2180f9e92e40SEric Sandeen 21812a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 21822a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 21832a458198SEric Sandeen { 21842a458198SEric Sandeen int max_active = fs_info->thread_pool_size; 21856f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21862a458198SEric Sandeen 21872a458198SEric Sandeen fs_info->workers = 2188cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2189cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 21902a458198SEric Sandeen 21912a458198SEric Sandeen fs_info->delalloc_workers = 2192cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2193cb001095SJeff Mahoney flags, max_active, 2); 21942a458198SEric Sandeen 21952a458198SEric Sandeen fs_info->flush_workers = 2196cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2197cb001095SJeff Mahoney flags, max_active, 0); 21982a458198SEric Sandeen 21992a458198SEric Sandeen fs_info->caching_workers = 2200cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 22012a458198SEric Sandeen 22022a458198SEric Sandeen /* 22032a458198SEric Sandeen * a higher idle thresh on the submit workers makes it much more 22042a458198SEric Sandeen * likely that bios will be send down in a sane order to the 22052a458198SEric Sandeen * devices 22062a458198SEric Sandeen */ 22072a458198SEric Sandeen fs_info->submit_workers = 2208cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "submit", flags, 22092a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22102a458198SEric Sandeen max_active), 64); 22112a458198SEric Sandeen 22122a458198SEric Sandeen fs_info->fixup_workers = 2213cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 22142a458198SEric Sandeen 22152a458198SEric Sandeen /* 22162a458198SEric Sandeen * endios are largely parallel and should have a very 22172a458198SEric Sandeen * low idle thresh 22182a458198SEric Sandeen */ 22192a458198SEric Sandeen fs_info->endio_workers = 2220cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 22212a458198SEric Sandeen fs_info->endio_meta_workers = 2222cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2223cb001095SJeff Mahoney max_active, 4); 22242a458198SEric Sandeen fs_info->endio_meta_write_workers = 2225cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2226cb001095SJeff Mahoney max_active, 2); 22272a458198SEric Sandeen fs_info->endio_raid56_workers = 2228cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2229cb001095SJeff Mahoney max_active, 4); 22302a458198SEric Sandeen fs_info->endio_repair_workers = 2231cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); 22322a458198SEric Sandeen fs_info->rmw_workers = 2233cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 22342a458198SEric Sandeen fs_info->endio_write_workers = 2235cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2236cb001095SJeff Mahoney max_active, 2); 22372a458198SEric Sandeen fs_info->endio_freespace_worker = 2238cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2239cb001095SJeff Mahoney max_active, 0); 22402a458198SEric Sandeen fs_info->delayed_workers = 2241cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2242cb001095SJeff Mahoney max_active, 0); 22432a458198SEric Sandeen fs_info->readahead_workers = 2244cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2245cb001095SJeff Mahoney max_active, 2); 22462a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2247cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 22482a458198SEric Sandeen fs_info->extent_workers = 2249cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "extent-refs", flags, 22502a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22512a458198SEric Sandeen max_active), 8); 22522a458198SEric Sandeen 22532a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 22542a458198SEric Sandeen fs_info->submit_workers && fs_info->flush_workers && 22552a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 22562a458198SEric Sandeen fs_info->endio_meta_write_workers && 22572a458198SEric Sandeen fs_info->endio_repair_workers && 22582a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 22592a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 22602a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 22612a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 22622a458198SEric Sandeen fs_info->extent_workers && 22632a458198SEric Sandeen fs_info->qgroup_rescan_workers)) { 22642a458198SEric Sandeen return -ENOMEM; 22652a458198SEric Sandeen } 22662a458198SEric Sandeen 22672a458198SEric Sandeen return 0; 22682a458198SEric Sandeen } 22692a458198SEric Sandeen 227063443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 227163443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 227263443bf5SEric Sandeen { 227363443bf5SEric Sandeen int ret; 227463443bf5SEric Sandeen struct btrfs_root *log_tree_root; 227563443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 227663443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 227763443bf5SEric Sandeen 227863443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2279f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 228063443bf5SEric Sandeen return -EIO; 228163443bf5SEric Sandeen } 228263443bf5SEric Sandeen 228374e4d827SDavid Sterba log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 228463443bf5SEric Sandeen if (!log_tree_root) 228563443bf5SEric Sandeen return -ENOMEM; 228663443bf5SEric Sandeen 2287da17066cSJeff Mahoney __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 228863443bf5SEric Sandeen 22892ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 229063443bf5SEric Sandeen fs_info->generation + 1); 229164c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2292f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22930eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 229464c043deSLiu Bo kfree(log_tree_root); 22950eeff236SLiu Bo return ret; 229664c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2297f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 229863443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 229963443bf5SEric Sandeen kfree(log_tree_root); 230063443bf5SEric Sandeen return -EIO; 230163443bf5SEric Sandeen } 230263443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 230363443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 230463443bf5SEric Sandeen if (ret) { 23050b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 230663443bf5SEric Sandeen "Failed to recover log tree"); 230763443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 230863443bf5SEric Sandeen kfree(log_tree_root); 230963443bf5SEric Sandeen return ret; 231063443bf5SEric Sandeen } 231163443bf5SEric Sandeen 2312bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 23136bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 231463443bf5SEric Sandeen if (ret) 231563443bf5SEric Sandeen return ret; 231663443bf5SEric Sandeen } 231763443bf5SEric Sandeen 231863443bf5SEric Sandeen return 0; 231963443bf5SEric Sandeen } 232063443bf5SEric Sandeen 23216bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 23224bbcaa64SEric Sandeen { 23236bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2324a4f3d2c4SDavid Sterba struct btrfs_root *root; 23254bbcaa64SEric Sandeen struct btrfs_key location; 23264bbcaa64SEric Sandeen int ret; 23274bbcaa64SEric Sandeen 23286bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 23296bccf3abSJeff Mahoney 23304bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 23314bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 23324bbcaa64SEric Sandeen location.offset = 0; 23334bbcaa64SEric Sandeen 2334a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2335a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2336a4f3d2c4SDavid Sterba return PTR_ERR(root); 2337a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2338a4f3d2c4SDavid Sterba fs_info->extent_root = root; 23394bbcaa64SEric Sandeen 23404bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2341a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2342a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2343a4f3d2c4SDavid Sterba return PTR_ERR(root); 2344a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2345a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23464bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 23474bbcaa64SEric Sandeen 23484bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2349a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2350a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2351a4f3d2c4SDavid Sterba return PTR_ERR(root); 2352a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2353a4f3d2c4SDavid Sterba fs_info->csum_root = root; 23544bbcaa64SEric Sandeen 23554bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2356a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2357a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2358a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2359afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2360a4f3d2c4SDavid Sterba fs_info->quota_root = root; 23614bbcaa64SEric Sandeen } 23624bbcaa64SEric Sandeen 23634bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2364a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2365a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2366a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 23674bbcaa64SEric Sandeen if (ret != -ENOENT) 23684bbcaa64SEric Sandeen return ret; 23694bbcaa64SEric Sandeen } else { 2370a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2371a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 23724bbcaa64SEric Sandeen } 23734bbcaa64SEric Sandeen 237470f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 237570f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 237670f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 237770f6d82eSOmar Sandoval if (IS_ERR(root)) 237870f6d82eSOmar Sandoval return PTR_ERR(root); 237970f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 238070f6d82eSOmar Sandoval fs_info->free_space_root = root; 238170f6d82eSOmar Sandoval } 238270f6d82eSOmar Sandoval 23834bbcaa64SEric Sandeen return 0; 23844bbcaa64SEric Sandeen } 23854bbcaa64SEric Sandeen 2386ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2387dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2388dfe25020SChris Mason char *options) 2389eb60ceacSChris Mason { 2390db94535dSChris Mason u32 sectorsize; 2391db94535dSChris Mason u32 nodesize; 239287ee04ebSChris Mason u32 stripesize; 239384234f3aSYan Zheng u64 generation; 2394f2b636e8SJosef Bacik u64 features; 23953de4586cSChris Mason struct btrfs_key location; 2396a061fc8dSChris Mason struct buffer_head *bh; 23974d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2398815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2399f84a8bd6SAl Viro struct btrfs_root *tree_root; 24004d34b278SIlya Dryomov struct btrfs_root *chunk_root; 2401eb60ceacSChris Mason int ret; 2402e58ca020SYan int err = -EINVAL; 2403af31f5e5SChris Mason int num_backups_tried = 0; 2404af31f5e5SChris Mason int backup_index = 0; 24055cdc7ad3SQu Wenruo int max_active; 24066675df31SOmar Sandoval int clear_free_space_tree = 0; 24074543df7eSChris Mason 240874e4d827SDavid Sterba tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 240974e4d827SDavid Sterba chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 2410cb517eabSMiao Xie if (!tree_root || !chunk_root) { 241139279cc3SChris Mason err = -ENOMEM; 241239279cc3SChris Mason goto fail; 241339279cc3SChris Mason } 241476dda93cSYan, Zheng 241576dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 241676dda93cSYan, Zheng if (ret) { 241776dda93cSYan, Zheng err = ret; 241876dda93cSYan, Zheng goto fail; 241976dda93cSYan, Zheng } 242076dda93cSYan, Zheng 2421908c7f19STejun Heo ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2422e2d84521SMiao Xie if (ret) { 2423e2d84521SMiao Xie err = ret; 24249e11ceeeSJan Kara goto fail_srcu; 2425e2d84521SMiao Xie } 242609cbfeafSKirill A. Shutemov fs_info->dirty_metadata_batch = PAGE_SIZE * 2427e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2428e2d84521SMiao Xie 2429908c7f19STejun Heo ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2430963d678bSMiao Xie if (ret) { 2431963d678bSMiao Xie err = ret; 2432963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2433963d678bSMiao Xie } 2434963d678bSMiao Xie 2435908c7f19STejun Heo ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); 2436c404e0dcSMiao Xie if (ret) { 2437c404e0dcSMiao Xie err = ret; 2438c404e0dcSMiao Xie goto fail_delalloc_bytes; 2439c404e0dcSMiao Xie } 2440c404e0dcSMiao Xie 244176dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2442f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 24438fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2444facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 244524bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2446eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 244711833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2448eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2449a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 245076dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 245124bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 24524cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2453f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2454ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2455fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2456f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 245747ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2458f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2459d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 246067c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 24617585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2462573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2463c2d6cb16SFilipe Manana mutex_init(&fs_info->cleaner_delayed_iput_mutex); 2464de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 246519c00ddcSChris Mason 24660b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 24676324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2468f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 246947ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 24700b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 247166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 247266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 247366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 247466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 247566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 247666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 247766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2478771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 24794cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2480fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 24812fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2482fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2483e20d96d6SChris Mason fs_info->sb = sb; 248495ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 24859ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 24864cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2487a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2488f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 24898b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2490f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 249190519d66SArne Jansen /* readahead state */ 2492d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 249390519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2494fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2495c8b97818SChris Mason 2496b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2497b34b086cSChris Mason num_online_cpus() + 2, 8); 24980afbaf8cSChris Mason 2499199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2500199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 250169fe2d75SJosef Bacik 250269fe2d75SJosef Bacik fs_info->btree_inode = new_inode(sb); 250369fe2d75SJosef Bacik if (!fs_info->btree_inode) { 250469fe2d75SJosef Bacik err = -ENOMEM; 250569fe2d75SJosef Bacik goto fail_bio_counter; 250669fe2d75SJosef Bacik } 250769fe2d75SJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 250869fe2d75SJosef Bacik 250916cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 251074e4d827SDavid Sterba GFP_KERNEL); 251116cdcec7SMiao Xie if (!fs_info->delayed_root) { 251216cdcec7SMiao Xie err = -ENOMEM; 251316cdcec7SMiao Xie goto fail_iput; 251416cdcec7SMiao Xie } 251516cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 25163eaa2885SChris Mason 2517638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 251821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 251921adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 252021adbd5cSStefan Behrens #endif 2521779a65a4SEric Sandeen btrfs_init_balance(fs_info); 252221c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2523a2de733cSArne Jansen 25249f6d2510SDavid Sterba sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 25259f6d2510SDavid Sterba sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 2526a061fc8dSChris Mason 25276bccf3abSJeff Mahoney btrfs_init_btree_inode(fs_info); 252839279cc3SChris Mason 2529e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 25306bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2531a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2532925baeddSChris Mason 2533c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[0], NULL); 2534c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[1], NULL); 253511833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2536afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 2537509659cdSChris Mason 25385a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 253979154b1bSChris Mason mutex_init(&fs_info->tree_log_mutex); 2540925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2541a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2542a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 25437d9eb12cSChris Mason mutex_init(&fs_info->volume_mutex); 25441bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 25459e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2546c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 254776dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2548803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2549fa9c0d79SChris Mason 2550ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2551f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2552416ac51dSArne Jansen 2553fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2554fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2555fa9c0d79SChris Mason 2556e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2557f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2558bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 25594854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 25603768f368SChris Mason 256104216820SFilipe Manana INIT_LIST_HEAD(&fs_info->pinned_chunks); 256204216820SFilipe Manana 2563da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2564da17066cSJeff Mahoney fs_info->nodesize = 4096; 2565da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2566da17066cSJeff Mahoney fs_info->stripesize = 4096; 2567da17066cSJeff Mahoney 256853b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 256953b381b3SDavid Woodhouse if (ret) { 257083c8266aSDavid Sterba err = ret; 257153b381b3SDavid Woodhouse goto fail_alloc; 257253b381b3SDavid Woodhouse } 257353b381b3SDavid Woodhouse 2574da17066cSJeff Mahoney __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 25757eccb903SChris Mason 25763c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 25771104a885SDavid Sterba 25781104a885SDavid Sterba /* 25791104a885SDavid Sterba * Read super block and check the signature bytes only 25801104a885SDavid Sterba */ 2581a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 258292fc03fbSAnand Jain if (IS_ERR(bh)) { 258392fc03fbSAnand Jain err = PTR_ERR(bh); 258416cdcec7SMiao Xie goto fail_alloc; 258520b45077SDave Young } 258639279cc3SChris Mason 25871104a885SDavid Sterba /* 25881104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 25891104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 25901104a885SDavid Sterba */ 2591ab8d0fc4SJeff Mahoney if (btrfs_check_super_csum(fs_info, bh->b_data)) { 259205135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 25931104a885SDavid Sterba err = -EINVAL; 2594b2acdddfSAnand Jain brelse(bh); 25951104a885SDavid Sterba goto fail_alloc; 25961104a885SDavid Sterba } 25971104a885SDavid Sterba 25981104a885SDavid Sterba /* 25991104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 26001104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 26011104a885SDavid Sterba * the whole block of INFO_SIZE 26021104a885SDavid Sterba */ 26036c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 26046c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 26056c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2606a061fc8dSChris Mason brelse(bh); 26075f39d397SChris Mason 26086c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 26090b86a832SChris Mason 26103d3a126aSDavid Sterba ret = btrfs_check_super_valid(fs_info); 26111104a885SDavid Sterba if (ret) { 261205135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 26131104a885SDavid Sterba err = -EINVAL; 26141104a885SDavid Sterba goto fail_alloc; 26151104a885SDavid Sterba } 26161104a885SDavid Sterba 26176c41761fSDavid Sterba disk_super = fs_info->super_copy; 26180f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 261916cdcec7SMiao Xie goto fail_alloc; 26200f7d52f4SChris Mason 2621acce952bSliubo /* check FS state, whether FS is broken. */ 262287533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 262387533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2624acce952bSliubo 262575e7cb7fSLiu Bo /* 2626af31f5e5SChris Mason * run through our array of backup supers and setup 2627af31f5e5SChris Mason * our ring pointer to the oldest one 2628af31f5e5SChris Mason */ 2629af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2630af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2631af31f5e5SChris Mason 2632af31f5e5SChris Mason /* 263375e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 263475e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 263575e7cb7fSLiu Bo */ 263675e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 263775e7cb7fSLiu Bo 26382ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 26392b82032cSYan Zheng if (ret) { 26402b82032cSYan Zheng err = ret; 264116cdcec7SMiao Xie goto fail_alloc; 26422b82032cSYan Zheng } 2643dfe25020SChris Mason 2644f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2645f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2646f2b636e8SJosef Bacik if (features) { 264705135f59SDavid Sterba btrfs_err(fs_info, 264805135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 2649c1c9ff7cSGeert Uytterhoeven features); 2650f2b636e8SJosef Bacik err = -EINVAL; 265116cdcec7SMiao Xie goto fail_alloc; 2652f2b636e8SJosef Bacik } 2653f2b636e8SJosef Bacik 26545d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 26555d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 26560b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 2657a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 26585c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 26595c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 2660727011e0SChris Mason 26613173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 266205135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 26633173a18fSJosef Bacik 2664727011e0SChris Mason /* 2665727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2666727011e0SChris Mason * they are bigger than the page size 2667727011e0SChris Mason */ 266809cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 2669727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 267005135f59SDavid Sterba btrfs_info(fs_info, 267105135f59SDavid Sterba "flagging fs with big metadata feature"); 2672727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2673727011e0SChris Mason } 2674727011e0SChris Mason 2675bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2676bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2677b7f67055SChandan Rajendra stripesize = sectorsize; 2678707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2679963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2680bc3f116fSChris Mason 2681da17066cSJeff Mahoney /* Cache block sizes */ 2682da17066cSJeff Mahoney fs_info->nodesize = nodesize; 2683da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 2684da17066cSJeff Mahoney fs_info->stripesize = stripesize; 2685da17066cSJeff Mahoney 2686bc3f116fSChris Mason /* 2687bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2688bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2689bc3f116fSChris Mason */ 2690bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2691707e8a07SDavid Sterba (sectorsize != nodesize)) { 269205135f59SDavid Sterba btrfs_err(fs_info, 269305135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 269405135f59SDavid Sterba nodesize, sectorsize); 2695bc3f116fSChris Mason goto fail_alloc; 2696bc3f116fSChris Mason } 2697bc3f116fSChris Mason 2698ceda0864SMiao Xie /* 2699ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2700ceda0864SMiao Xie * update the flag. 2701ceda0864SMiao Xie */ 27025d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 27035d4f98a2SYan Zheng 2704f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2705f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2706bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 270705135f59SDavid Sterba btrfs_err(fs_info, 270805135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 2709c1c9ff7cSGeert Uytterhoeven features); 2710f2b636e8SJosef Bacik err = -EINVAL; 271116cdcec7SMiao Xie goto fail_alloc; 2712f2b636e8SJosef Bacik } 271361d92c32SChris Mason 27145cdc7ad3SQu Wenruo max_active = fs_info->thread_pool_size; 271561d92c32SChris Mason 27162a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 27172a458198SEric Sandeen if (ret) { 27182a458198SEric Sandeen err = ret; 27190dc3b84aSJosef Bacik goto fail_sb_buffer; 27200dc3b84aSJosef Bacik } 27214543df7eSChris Mason 27229e11ceeeSJan Kara sb->s_bdi->congested_fn = btrfs_congested_fn; 27239e11ceeeSJan Kara sb->s_bdi->congested_data = fs_info; 27249e11ceeeSJan Kara sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 2725d4417e22SNikolay Borisov sb->s_bdi->ra_pages = VM_MAX_READAHEAD * SZ_1K / PAGE_SIZE; 27269e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 27279e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 27284575c9ccSChris Mason 2729a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2730a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2731ee87cf5eSAnand Jain memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE); 2732db94535dSChris Mason 2733925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 27346bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 2735925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 273684eed90fSChris Mason if (ret) { 273705135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 27385d4f98a2SYan Zheng goto fail_sb_buffer; 273984eed90fSChris Mason } 27400b86a832SChris Mason 274184234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 27420b86a832SChris Mason 2743da17066cSJeff Mahoney __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 27440b86a832SChris Mason 27452ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 27460b86a832SChris Mason btrfs_super_chunk_root(disk_super), 2747ce86cd59SDavid Sterba generation); 274864c043deSLiu Bo if (IS_ERR(chunk_root->node) || 274964c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 275005135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 2751e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 2752e5fffbacSchandan free_extent_buffer(chunk_root->node); 275395ab1f64SZhao Lei chunk_root->node = NULL; 2754af31f5e5SChris Mason goto fail_tree_roots; 275583121942SDavid Woodhouse } 27565d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 27575d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 27580b86a832SChris Mason 2759e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2760b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2761e17cade2SChris Mason 27625b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 27632b82032cSYan Zheng if (ret) { 276405135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 2765af31f5e5SChris Mason goto fail_tree_roots; 27662b82032cSYan Zheng } 27670b86a832SChris Mason 27688dabb742SStefan Behrens /* 27698dabb742SStefan Behrens * keep the device that is marked to be the target device for the 27708dabb742SStefan Behrens * dev_replace procedure 27718dabb742SStefan Behrens */ 27729eaed21eSEric Sandeen btrfs_close_extra_devices(fs_devices, 0); 2773dfe25020SChris Mason 2774a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 277505135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 2776a6b0d5c8SChris Mason goto fail_tree_roots; 2777a6b0d5c8SChris Mason } 2778a6b0d5c8SChris Mason 2779af31f5e5SChris Mason retry_root_backup: 278084234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 27810b86a832SChris Mason 27822ff7e61eSJeff Mahoney tree_root->node = read_tree_block(fs_info, 2783db94535dSChris Mason btrfs_super_root(disk_super), 2784ce86cd59SDavid Sterba generation); 278564c043deSLiu Bo if (IS_ERR(tree_root->node) || 278664c043deSLiu Bo !extent_buffer_uptodate(tree_root->node)) { 278705135f59SDavid Sterba btrfs_warn(fs_info, "failed to read tree root"); 2788e5fffbacSchandan if (!IS_ERR(tree_root->node)) 2789e5fffbacSchandan free_extent_buffer(tree_root->node); 279095ab1f64SZhao Lei tree_root->node = NULL; 2791af31f5e5SChris Mason goto recovery_tree_root; 279283121942SDavid Woodhouse } 2793af31f5e5SChris Mason 27945d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 27955d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 279669e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2797db94535dSChris Mason 2798f32e48e9SChandan Rajendra mutex_lock(&tree_root->objectid_mutex); 2799f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(tree_root, 2800f32e48e9SChandan Rajendra &tree_root->highest_objectid); 2801f32e48e9SChandan Rajendra if (ret) { 2802f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2803f32e48e9SChandan Rajendra goto recovery_tree_root; 2804f32e48e9SChandan Rajendra } 2805f32e48e9SChandan Rajendra 2806f32e48e9SChandan Rajendra ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2807f32e48e9SChandan Rajendra 2808f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2809f32e48e9SChandan Rajendra 28106bccf3abSJeff Mahoney ret = btrfs_read_roots(fs_info); 28114bbcaa64SEric Sandeen if (ret) 2812af31f5e5SChris Mason goto recovery_tree_root; 2813f7a81ea4SStefan Behrens 28148929ecfaSYan, Zheng fs_info->generation = generation; 28158929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 28168929ecfaSYan, Zheng 281768310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 281868310a5eSIlya Dryomov if (ret) { 281905135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 282068310a5eSIlya Dryomov goto fail_block_groups; 282168310a5eSIlya Dryomov } 282268310a5eSIlya Dryomov 2823733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2824733f4fbbSStefan Behrens if (ret) { 282505135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 2826733f4fbbSStefan Behrens goto fail_block_groups; 2827733f4fbbSStefan Behrens } 2828733f4fbbSStefan Behrens 28298dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 28308dabb742SStefan Behrens if (ret) { 283105135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 28328dabb742SStefan Behrens goto fail_block_groups; 28338dabb742SStefan Behrens } 28348dabb742SStefan Behrens 28359eaed21eSEric Sandeen btrfs_close_extra_devices(fs_devices, 1); 28368dabb742SStefan Behrens 2837b7c35e81SAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices, NULL); 2838b7c35e81SAnand Jain if (ret) { 283905135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 284005135f59SDavid Sterba ret); 2841b7c35e81SAnand Jain goto fail_block_groups; 2842b7c35e81SAnand Jain } 2843b7c35e81SAnand Jain 2844b7c35e81SAnand Jain ret = btrfs_sysfs_add_device(fs_devices); 2845b7c35e81SAnand Jain if (ret) { 284605135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs device interface: %d", 284705135f59SDavid Sterba ret); 2848b7c35e81SAnand Jain goto fail_fsdev_sysfs; 2849b7c35e81SAnand Jain } 2850b7c35e81SAnand Jain 285196f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 28525ac1d209SJeff Mahoney if (ret) { 285305135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 2854b7c35e81SAnand Jain goto fail_fsdev_sysfs; 28555ac1d209SJeff Mahoney } 28565ac1d209SJeff Mahoney 2857c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2858c59021f8Sliubo if (ret) { 285905135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 28602365dd3cSAnand Jain goto fail_sysfs; 2861c59021f8Sliubo } 2862c59021f8Sliubo 28635b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 28641b1d1f66SJosef Bacik if (ret) { 286505135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 28662365dd3cSAnand Jain goto fail_sysfs; 28671b1d1f66SJosef Bacik } 28684330e183SQu Wenruo 28690f0d1272SLinus Torvalds if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info)) { 287005135f59SDavid Sterba btrfs_warn(fs_info, 28714330e183SQu Wenruo "writeable mount is not allowed due to too many missing devices"); 28722365dd3cSAnand Jain goto fail_sysfs; 2873292fd7fcSStefan Behrens } 28749078a3e1SChris Mason 2875a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2876a74a4b97SChris Mason "btrfs-cleaner"); 287757506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 28782365dd3cSAnand Jain goto fail_sysfs; 2879a74a4b97SChris Mason 2880a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2881a74a4b97SChris Mason tree_root, 2882a74a4b97SChris Mason "btrfs-transaction"); 288357506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 28843f157a2fSChris Mason goto fail_cleaner; 2885a74a4b97SChris Mason 2886583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 2887c289811cSChris Mason !fs_info->fs_devices->rotating) { 2888583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 2889c289811cSChris Mason } 2890c289811cSChris Mason 2891572d9ab7SDavid Sterba /* 289201327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 2893572d9ab7SDavid Sterba * not have to wait for transaction commit 2894572d9ab7SDavid Sterba */ 2895572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 28963818aea2SQu Wenruo 289721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 28980b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 28992ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 29000b246afaSJeff Mahoney btrfs_test_opt(fs_info, 290121adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 290221adbd5cSStefan Behrens 1 : 0, 290321adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 290421adbd5cSStefan Behrens if (ret) 290505135f59SDavid Sterba btrfs_warn(fs_info, 290605135f59SDavid Sterba "failed to initialize integrity check module: %d", 290705135f59SDavid Sterba ret); 290821adbd5cSStefan Behrens } 290921adbd5cSStefan Behrens #endif 2910bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 2911bcef60f2SArne Jansen if (ret) 2912bcef60f2SArne Jansen goto fail_trans_kthread; 291321adbd5cSStefan Behrens 2914fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 2915fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 2916fd708b81SJosef Bacik 291796da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 291896da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 29190b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 292063443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 292179787eaaSJeff Mahoney if (ret) { 292263443bf5SEric Sandeen err = ret; 292328c16cbbSWang Shilong goto fail_qgroup; 2924e556ce2cSYan Zheng } 2925e02119d5SChris Mason } 29261a40e23bSZheng Yan 29276bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 292879787eaaSJeff Mahoney if (ret) 292928c16cbbSWang Shilong goto fail_qgroup; 293076dda93cSYan, Zheng 2931bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 2932d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 293344c44af2SIlya Dryomov if (ret) 293428c16cbbSWang Shilong goto fail_qgroup; 293590c711abSZygo Blaxell 293690c711abSZygo Blaxell mutex_lock(&fs_info->cleaner_mutex); 29375d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 293890c711abSZygo Blaxell mutex_unlock(&fs_info->cleaner_mutex); 2939d7ce5843SMiao Xie if (ret < 0) { 294005135f59SDavid Sterba btrfs_warn(fs_info, "failed to recover relocation: %d", 294105135f59SDavid Sterba ret); 2942d7ce5843SMiao Xie err = -EINVAL; 2943bcef60f2SArne Jansen goto fail_qgroup; 2944d7ce5843SMiao Xie } 29457c2ca468SChris Mason } 29461a40e23bSZheng Yan 29473de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 29483de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 2949cb517eabSMiao Xie location.offset = 0; 29503de4586cSChris Mason 29513de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 29523140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 29533140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 2954bcef60f2SArne Jansen goto fail_qgroup; 29553140c9a3SDan Carpenter } 2956c289811cSChris Mason 2957bc98a42cSDavid Howells if (sb_rdonly(sb)) 29582b6ba629SIlya Dryomov return 0; 29592b6ba629SIlya Dryomov 2960f8d468a1SOmar Sandoval if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 2961f8d468a1SOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29626675df31SOmar Sandoval clear_free_space_tree = 1; 29636675df31SOmar Sandoval } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29646675df31SOmar Sandoval !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 29656675df31SOmar Sandoval btrfs_warn(fs_info, "free space tree is invalid"); 29666675df31SOmar Sandoval clear_free_space_tree = 1; 29676675df31SOmar Sandoval } 29686675df31SOmar Sandoval 29696675df31SOmar Sandoval if (clear_free_space_tree) { 2970f8d468a1SOmar Sandoval btrfs_info(fs_info, "clearing free space tree"); 2971f8d468a1SOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 2972f8d468a1SOmar Sandoval if (ret) { 2973f8d468a1SOmar Sandoval btrfs_warn(fs_info, 2974f8d468a1SOmar Sandoval "failed to clear free space tree: %d", ret); 29756bccf3abSJeff Mahoney close_ctree(fs_info); 2976f8d468a1SOmar Sandoval return ret; 2977f8d468a1SOmar Sandoval } 2978f8d468a1SOmar Sandoval } 2979f8d468a1SOmar Sandoval 29800b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 2981511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 298205135f59SDavid Sterba btrfs_info(fs_info, "creating free space tree"); 2983511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 2984511711afSChris Mason if (ret) { 298505135f59SDavid Sterba btrfs_warn(fs_info, 298605135f59SDavid Sterba "failed to create free space tree: %d", ret); 29876bccf3abSJeff Mahoney close_ctree(fs_info); 2988511711afSChris Mason return ret; 2989511711afSChris Mason } 2990511711afSChris Mason } 2991511711afSChris Mason 2992e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 29932b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29942b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29952b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 29966bccf3abSJeff Mahoney close_ctree(fs_info); 29972b6ba629SIlya Dryomov return ret; 29982b6ba629SIlya Dryomov } 2999e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 300059641015SIlya Dryomov 30012b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 30022b6ba629SIlya Dryomov if (ret) { 300305135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume balance: %d", ret); 30046bccf3abSJeff Mahoney close_ctree(fs_info); 30052b6ba629SIlya Dryomov return ret; 3006e3acc2a6SJosef Bacik } 3007e3acc2a6SJosef Bacik 30088dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 30098dabb742SStefan Behrens if (ret) { 301005135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 30116bccf3abSJeff Mahoney close_ctree(fs_info); 30128dabb742SStefan Behrens return ret; 30138dabb742SStefan Behrens } 30148dabb742SStefan Behrens 3015b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3016b382a324SJan Schmidt 30174bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 301805135f59SDavid Sterba btrfs_info(fs_info, "creating UUID tree"); 3019f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3020f7a81ea4SStefan Behrens if (ret) { 302105135f59SDavid Sterba btrfs_warn(fs_info, 302205135f59SDavid Sterba "failed to create the UUID tree: %d", ret); 30236bccf3abSJeff Mahoney close_ctree(fs_info); 3024f7a81ea4SStefan Behrens return ret; 3025f7a81ea4SStefan Behrens } 30260b246afaSJeff Mahoney } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 30274bbcaa64SEric Sandeen fs_info->generation != 30284bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 302905135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 303070f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 303170f80175SStefan Behrens if (ret) { 303205135f59SDavid Sterba btrfs_warn(fs_info, 303305135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 30346bccf3abSJeff Mahoney close_ctree(fs_info); 303570f80175SStefan Behrens return ret; 303670f80175SStefan Behrens } 303770f80175SStefan Behrens } else { 3038afcdd129SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 3039f7a81ea4SStefan Behrens } 3040afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 304147ab2a6cSJosef Bacik 30428dcddfa0SQu Wenruo /* 30438dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 30448dcddfa0SQu Wenruo * no need to keep the flag 30458dcddfa0SQu Wenruo */ 30468dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 30478dcddfa0SQu Wenruo 3048ad2b2c80SAl Viro return 0; 304939279cc3SChris Mason 3050bcef60f2SArne Jansen fail_qgroup: 3051bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 30527c2ca468SChris Mason fail_trans_kthread: 30537c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 30542ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3055faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 30563f157a2fSChris Mason fail_cleaner: 3057a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 30587c2ca468SChris Mason 30597c2ca468SChris Mason /* 30607c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 30617c2ca468SChris Mason * kthreads 30627c2ca468SChris Mason */ 30637c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 30647c2ca468SChris Mason 30652365dd3cSAnand Jain fail_sysfs: 30666618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 30672365dd3cSAnand Jain 3068b7c35e81SAnand Jain fail_fsdev_sysfs: 3069b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3070b7c35e81SAnand Jain 30711b1d1f66SJosef Bacik fail_block_groups: 307254067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3073af31f5e5SChris Mason 3074af31f5e5SChris Mason fail_tree_roots: 3075af31f5e5SChris Mason free_root_pointers(fs_info, 1); 30762b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3077af31f5e5SChris Mason 307839279cc3SChris Mason fail_sb_buffer: 30797abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 30805cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 308116cdcec7SMiao Xie fail_alloc: 30824543df7eSChris Mason fail_iput: 3083586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3084586e46e2SIlya Dryomov 30854543df7eSChris Mason iput(fs_info->btree_inode); 3086c404e0dcSMiao Xie fail_bio_counter: 3087c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 3088963d678bSMiao Xie fail_delalloc_bytes: 3089963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3090e2d84521SMiao Xie fail_dirty_metadata_bytes: 3091e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 309276dda93cSYan, Zheng fail_srcu: 309376dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 30947e662854SQinghuang Feng fail: 309553b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3096586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3097ad2b2c80SAl Viro return err; 3098af31f5e5SChris Mason 3099af31f5e5SChris Mason recovery_tree_root: 31000b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 3101af31f5e5SChris Mason goto fail_tree_roots; 3102af31f5e5SChris Mason 3103af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3104af31f5e5SChris Mason 3105af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3106af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3107af31f5e5SChris Mason 3108af31f5e5SChris Mason /* we can't trust the free space cache either */ 3109af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3110af31f5e5SChris Mason 3111af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3112af31f5e5SChris Mason &num_backups_tried, &backup_index); 3113af31f5e5SChris Mason if (ret == -1) 3114af31f5e5SChris Mason goto fail_block_groups; 3115af31f5e5SChris Mason goto retry_root_backup; 3116eb60ceacSChris Mason } 3117eb60ceacSChris Mason 3118f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3119f2984462SChris Mason { 3120f2984462SChris Mason if (uptodate) { 3121f2984462SChris Mason set_buffer_uptodate(bh); 3122f2984462SChris Mason } else { 3123442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3124442a4f63SStefan Behrens bh->b_private; 3125442a4f63SStefan Behrens 3126fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3127b14af3b4SDavid Sterba "lost page write due to IO error on %s", 3128606686eeSJosef Bacik rcu_str_deref(device->name)); 312901327610SNicholas D Steeves /* note, we don't set_buffer_write_io_error because we have 31301259ab75SChris Mason * our own ways of dealing with the IO errors 31311259ab75SChris Mason */ 3132f2984462SChris Mason clear_buffer_uptodate(bh); 3133442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3134f2984462SChris Mason } 3135f2984462SChris Mason unlock_buffer(bh); 3136f2984462SChris Mason put_bh(bh); 3137f2984462SChris Mason } 3138f2984462SChris Mason 313929c36d72SAnand Jain int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, 314029c36d72SAnand Jain struct buffer_head **bh_ret) 314129c36d72SAnand Jain { 314229c36d72SAnand Jain struct buffer_head *bh; 314329c36d72SAnand Jain struct btrfs_super_block *super; 314429c36d72SAnand Jain u64 bytenr; 314529c36d72SAnand Jain 314629c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 314729c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 314829c36d72SAnand Jain return -EINVAL; 314929c36d72SAnand Jain 31509f6d2510SDavid Sterba bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); 315129c36d72SAnand Jain /* 315229c36d72SAnand Jain * If we fail to read from the underlying devices, as of now 315329c36d72SAnand Jain * the best option we have is to mark it EIO. 315429c36d72SAnand Jain */ 315529c36d72SAnand Jain if (!bh) 315629c36d72SAnand Jain return -EIO; 315729c36d72SAnand Jain 315829c36d72SAnand Jain super = (struct btrfs_super_block *)bh->b_data; 315929c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 316029c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 316129c36d72SAnand Jain brelse(bh); 316229c36d72SAnand Jain return -EINVAL; 316329c36d72SAnand Jain } 316429c36d72SAnand Jain 316529c36d72SAnand Jain *bh_ret = bh; 316629c36d72SAnand Jain return 0; 316729c36d72SAnand Jain } 316829c36d72SAnand Jain 316929c36d72SAnand Jain 3170a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3171a512bbf8SYan Zheng { 3172a512bbf8SYan Zheng struct buffer_head *bh; 3173a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3174a512bbf8SYan Zheng struct btrfs_super_block *super; 3175a512bbf8SYan Zheng int i; 3176a512bbf8SYan Zheng u64 transid = 0; 317792fc03fbSAnand Jain int ret = -EINVAL; 3178a512bbf8SYan Zheng 3179a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3180a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3181a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3182a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3183a512bbf8SYan Zheng */ 3184a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 318529c36d72SAnand Jain ret = btrfs_read_dev_one_super(bdev, i, &bh); 318629c36d72SAnand Jain if (ret) 3187a512bbf8SYan Zheng continue; 3188a512bbf8SYan Zheng 3189a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3190a512bbf8SYan Zheng 3191a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3192a512bbf8SYan Zheng brelse(latest); 3193a512bbf8SYan Zheng latest = bh; 3194a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3195a512bbf8SYan Zheng } else { 3196a512bbf8SYan Zheng brelse(bh); 3197a512bbf8SYan Zheng } 3198a512bbf8SYan Zheng } 319992fc03fbSAnand Jain 320092fc03fbSAnand Jain if (!latest) 320192fc03fbSAnand Jain return ERR_PTR(ret); 320292fc03fbSAnand Jain 3203a512bbf8SYan Zheng return latest; 3204a512bbf8SYan Zheng } 3205a512bbf8SYan Zheng 32064eedeb75SHisashi Hifumi /* 3207abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3208abbb3b8eSDavid Sterba * buffer heads we write are pinned. 32094eedeb75SHisashi Hifumi * 3210abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3211abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3212abbb3b8eSDavid Sterba * same for write and wait phases. 32134eedeb75SHisashi Hifumi * 3214abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or submission fails. 32154eedeb75SHisashi Hifumi */ 3216a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3217abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3218a512bbf8SYan Zheng { 3219a512bbf8SYan Zheng struct buffer_head *bh; 3220a512bbf8SYan Zheng int i; 3221a512bbf8SYan Zheng int ret; 3222a512bbf8SYan Zheng int errors = 0; 3223a512bbf8SYan Zheng u32 crc; 3224a512bbf8SYan Zheng u64 bytenr; 32251b9e619cSOmar Sandoval int op_flags; 3226a512bbf8SYan Zheng 3227a512bbf8SYan Zheng if (max_mirrors == 0) 3228a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3229a512bbf8SYan Zheng 3230a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3231a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3232935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3233935e5cc9SMiao Xie device->commit_total_bytes) 3234a512bbf8SYan Zheng break; 3235a512bbf8SYan Zheng 3236a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3237a512bbf8SYan Zheng 3238a512bbf8SYan Zheng crc = ~(u32)0; 3239abbb3b8eSDavid Sterba crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, 3240abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3241a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3242a512bbf8SYan Zheng 3243abbb3b8eSDavid Sterba /* One reference for us, and we leave it for the caller */ 32449f6d2510SDavid Sterba bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, 3245a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3246634554dcSJosef Bacik if (!bh) { 3247fb456252SJeff Mahoney btrfs_err(device->fs_info, 3248f14d104dSDavid Sterba "couldn't get super buffer head for bytenr %llu", 3249f14d104dSDavid Sterba bytenr); 3250634554dcSJosef Bacik errors++; 3251634554dcSJosef Bacik continue; 3252634554dcSJosef Bacik } 3253634554dcSJosef Bacik 3254a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3255a512bbf8SYan Zheng 32564eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3257a512bbf8SYan Zheng get_bh(bh); 32584eedeb75SHisashi Hifumi 32594eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3260a512bbf8SYan Zheng lock_buffer(bh); 3261a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3262442a4f63SStefan Behrens bh->b_private = device; 3263a512bbf8SYan Zheng 3264387125fcSChris Mason /* 3265387125fcSChris Mason * we fua the first super. The others we allow 3266387125fcSChris Mason * to go down lazy. 3267387125fcSChris Mason */ 32681b9e619cSOmar Sandoval op_flags = REQ_SYNC | REQ_META | REQ_PRIO; 32691b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 32701b9e619cSOmar Sandoval op_flags |= REQ_FUA; 32711b9e619cSOmar Sandoval ret = btrfsic_submit_bh(REQ_OP_WRITE, op_flags, bh); 32724eedeb75SHisashi Hifumi if (ret) 3273a512bbf8SYan Zheng errors++; 3274a512bbf8SYan Zheng } 3275a512bbf8SYan Zheng return errors < i ? 0 : -1; 3276a512bbf8SYan Zheng } 3277a512bbf8SYan Zheng 3278387125fcSChris Mason /* 3279abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3280abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3281abbb3b8eSDavid Sterba * 3282abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or not marked up to 3283abbb3b8eSDavid Sterba * date. 3284abbb3b8eSDavid Sterba */ 3285abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3286abbb3b8eSDavid Sterba { 3287abbb3b8eSDavid Sterba struct buffer_head *bh; 3288abbb3b8eSDavid Sterba int i; 3289abbb3b8eSDavid Sterba int errors = 0; 3290abbb3b8eSDavid Sterba u64 bytenr; 3291abbb3b8eSDavid Sterba 3292abbb3b8eSDavid Sterba if (max_mirrors == 0) 3293abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3294abbb3b8eSDavid Sterba 3295abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3296abbb3b8eSDavid Sterba bytenr = btrfs_sb_offset(i); 3297abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3298abbb3b8eSDavid Sterba device->commit_total_bytes) 3299abbb3b8eSDavid Sterba break; 3300abbb3b8eSDavid Sterba 33019f6d2510SDavid Sterba bh = __find_get_block(device->bdev, 33029f6d2510SDavid Sterba bytenr / BTRFS_BDEV_BLOCKSIZE, 3303abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE); 3304abbb3b8eSDavid Sterba if (!bh) { 3305abbb3b8eSDavid Sterba errors++; 3306abbb3b8eSDavid Sterba continue; 3307abbb3b8eSDavid Sterba } 3308abbb3b8eSDavid Sterba wait_on_buffer(bh); 3309abbb3b8eSDavid Sterba if (!buffer_uptodate(bh)) 3310abbb3b8eSDavid Sterba errors++; 3311abbb3b8eSDavid Sterba 3312abbb3b8eSDavid Sterba /* drop our reference */ 3313abbb3b8eSDavid Sterba brelse(bh); 3314abbb3b8eSDavid Sterba 3315abbb3b8eSDavid Sterba /* drop the reference from the writing run */ 3316abbb3b8eSDavid Sterba brelse(bh); 3317abbb3b8eSDavid Sterba } 3318abbb3b8eSDavid Sterba 3319abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3320abbb3b8eSDavid Sterba } 3321abbb3b8eSDavid Sterba 3322abbb3b8eSDavid Sterba /* 3323387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3324387125fcSChris Mason * for the barrier when it is done 3325387125fcSChris Mason */ 33264246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3327387125fcSChris Mason { 3328387125fcSChris Mason complete(bio->bi_private); 3329387125fcSChris Mason } 3330387125fcSChris Mason 3331387125fcSChris Mason /* 33324fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 33334fc6441aSAnand Jain * done in the waiting counterpart. 3334387125fcSChris Mason */ 33354fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3336387125fcSChris Mason { 3337c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3338e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3339387125fcSChris Mason 3340c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 33414fc6441aSAnand Jain return; 3342387125fcSChris Mason 3343e0ae9994SDavid Sterba bio_reset(bio); 3344387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 334574d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 33468d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3347387125fcSChris Mason init_completion(&device->flush_wait); 3348387125fcSChris Mason bio->bi_private = &device->flush_wait; 3349387125fcSChris Mason 335043a01111SLu Fengqi btrfsic_submit_bio(bio); 3351e0ae9994SDavid Sterba device->flush_bio_sent = 1; 33524fc6441aSAnand Jain } 3353387125fcSChris Mason 33544fc6441aSAnand Jain /* 33554fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 33564fc6441aSAnand Jain */ 33578c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 33584fc6441aSAnand Jain { 33594fc6441aSAnand Jain struct bio *bio = device->flush_bio; 33604fc6441aSAnand Jain 3361e0ae9994SDavid Sterba if (!device->flush_bio_sent) 336258efbc9fSOmar Sandoval return BLK_STS_OK; 33634fc6441aSAnand Jain 3364e0ae9994SDavid Sterba device->flush_bio_sent = 0; 33652980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 33664fc6441aSAnand Jain 33678c27cb35SLinus Torvalds return bio->bi_status; 3368387125fcSChris Mason } 3369387125fcSChris Mason 3370d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3371401b41e5SAnand Jain { 3372d10b82feSQu Wenruo if (!btrfs_check_rw_degradable(fs_info)) 3373401b41e5SAnand Jain return -EIO; 3374387125fcSChris Mason return 0; 3375387125fcSChris Mason } 3376387125fcSChris Mason 3377387125fcSChris Mason /* 3378387125fcSChris Mason * send an empty flush down to each device in parallel, 3379387125fcSChris Mason * then wait for them 3380387125fcSChris Mason */ 3381387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3382387125fcSChris Mason { 3383387125fcSChris Mason struct list_head *head; 3384387125fcSChris Mason struct btrfs_device *dev; 33855af3e8ccSStefan Behrens int errors_wait = 0; 33864e4cbee9SChristoph Hellwig blk_status_t ret; 3387387125fcSChris Mason 33881538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3389387125fcSChris Mason /* send down all the barriers */ 3390387125fcSChris Mason head = &info->fs_devices->devices; 33911538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3392e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3393f88ba6a2SHidetoshi Seto continue; 3394cea7c8bfSAnand Jain if (!dev->bdev) 3395387125fcSChris Mason continue; 3396e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3397ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3398387125fcSChris Mason continue; 3399387125fcSChris Mason 34004fc6441aSAnand Jain write_dev_flush(dev); 340158efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3402387125fcSChris Mason } 3403387125fcSChris Mason 3404387125fcSChris Mason /* wait for all the barriers */ 34051538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3406e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3407f88ba6a2SHidetoshi Seto continue; 3408387125fcSChris Mason if (!dev->bdev) { 34095af3e8ccSStefan Behrens errors_wait++; 3410387125fcSChris Mason continue; 3411387125fcSChris Mason } 3412e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3413ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3414387125fcSChris Mason continue; 3415387125fcSChris Mason 34164fc6441aSAnand Jain ret = wait_dev_flush(dev); 3417401b41e5SAnand Jain if (ret) { 3418401b41e5SAnand Jain dev->last_flush_error = ret; 341966b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 342066b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 34215af3e8ccSStefan Behrens errors_wait++; 3422387125fcSChris Mason } 3423401b41e5SAnand Jain } 3424401b41e5SAnand Jain 3425cea7c8bfSAnand Jain if (errors_wait) { 3426401b41e5SAnand Jain /* 3427401b41e5SAnand Jain * At some point we need the status of all disks 3428401b41e5SAnand Jain * to arrive at the volume status. So error checking 3429401b41e5SAnand Jain * is being pushed to a separate loop. 3430401b41e5SAnand Jain */ 3431d10b82feSQu Wenruo return check_barrier_error(info); 3432401b41e5SAnand Jain } 3433387125fcSChris Mason return 0; 3434387125fcSChris Mason } 3435387125fcSChris Mason 3436943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3437943c6e99SZhao Lei { 34388789f4feSZhao Lei int raid_type; 34398789f4feSZhao Lei int min_tolerated = INT_MAX; 3440943c6e99SZhao Lei 34418789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 34428789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 34438789f4feSZhao Lei min_tolerated = min(min_tolerated, 34448789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 34458789f4feSZhao Lei tolerated_failures); 3446943c6e99SZhao Lei 34478789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 34488789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 34498789f4feSZhao Lei continue; 34508789f4feSZhao Lei if (!(flags & btrfs_raid_group[raid_type])) 34518789f4feSZhao Lei continue; 34528789f4feSZhao Lei min_tolerated = min(min_tolerated, 34538789f4feSZhao Lei btrfs_raid_array[raid_type]. 34548789f4feSZhao Lei tolerated_failures); 34558789f4feSZhao Lei } 3456943c6e99SZhao Lei 34578789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3458ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 34598789f4feSZhao Lei min_tolerated = 0; 34608789f4feSZhao Lei } 34618789f4feSZhao Lei 34628789f4feSZhao Lei return min_tolerated; 3463943c6e99SZhao Lei } 3464943c6e99SZhao Lei 3465eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3466f2984462SChris Mason { 3467e5e9a520SChris Mason struct list_head *head; 3468f2984462SChris Mason struct btrfs_device *dev; 3469a061fc8dSChris Mason struct btrfs_super_block *sb; 3470f2984462SChris Mason struct btrfs_dev_item *dev_item; 3471f2984462SChris Mason int ret; 3472f2984462SChris Mason int do_barriers; 3473a236aed1SChris Mason int max_errors; 3474a236aed1SChris Mason int total_errors = 0; 3475a061fc8dSChris Mason u64 flags; 3476f2984462SChris Mason 34770b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3478fed3b381SLiu Bo 3479fed3b381SLiu Bo /* 3480fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3481fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3482fed3b381SLiu Bo * been consistent on disk. 3483fed3b381SLiu Bo */ 3484fed3b381SLiu Bo if (max_mirrors == 0) 34850b246afaSJeff Mahoney backup_super_roots(fs_info); 3486f2984462SChris Mason 34870b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3488a061fc8dSChris Mason dev_item = &sb->dev_item; 3489e5e9a520SChris Mason 34900b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 34910b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 34920b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3493387125fcSChris Mason 34945af3e8ccSStefan Behrens if (do_barriers) { 34950b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 34965af3e8ccSStefan Behrens if (ret) { 34975af3e8ccSStefan Behrens mutex_unlock( 34980b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 34990b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 35005af3e8ccSStefan Behrens "errors while submitting device barriers."); 35015af3e8ccSStefan Behrens return ret; 35025af3e8ccSStefan Behrens } 35035af3e8ccSStefan Behrens } 3504387125fcSChris Mason 35051538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3506dfe25020SChris Mason if (!dev->bdev) { 3507dfe25020SChris Mason total_errors++; 3508dfe25020SChris Mason continue; 3509dfe25020SChris Mason } 3510e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3511ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3512dfe25020SChris Mason continue; 3513dfe25020SChris Mason 35142b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3515a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3516a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 35177df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3518935e5cc9SMiao Xie dev->commit_total_bytes); 3519ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3520ce7213c7SMiao Xie dev->commit_bytes_used); 3521a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3522a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3523a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3524a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 352544880fdcSAnand Jain memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); 3526a512bbf8SYan Zheng 3527a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3528a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3529f2984462SChris Mason 3530abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3531a236aed1SChris Mason if (ret) 3532a236aed1SChris Mason total_errors++; 3533f2984462SChris Mason } 3534a236aed1SChris Mason if (total_errors > max_errors) { 35350b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3536d397712bSChris Mason total_errors); 35370b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 353879787eaaSJeff Mahoney 35399d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 35400b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 35410b246afaSJeff Mahoney "%d errors while writing supers", 35420b246afaSJeff Mahoney total_errors); 35439d565ba4SStefan Behrens return -EIO; 3544a236aed1SChris Mason } 3545f2984462SChris Mason 3546a512bbf8SYan Zheng total_errors = 0; 35471538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3548dfe25020SChris Mason if (!dev->bdev) 3549dfe25020SChris Mason continue; 3550e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3551ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3552dfe25020SChris Mason continue; 3553dfe25020SChris Mason 3554abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3555a512bbf8SYan Zheng if (ret) 35561259ab75SChris Mason total_errors++; 3557f2984462SChris Mason } 35580b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3559a236aed1SChris Mason if (total_errors > max_errors) { 35600b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 35610b246afaSJeff Mahoney "%d errors while writing supers", 35620b246afaSJeff Mahoney total_errors); 356379787eaaSJeff Mahoney return -EIO; 3564a236aed1SChris Mason } 3565f2984462SChris Mason return 0; 3566f2984462SChris Mason } 3567f2984462SChris Mason 3568cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3569cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3570cb517eabSMiao Xie struct btrfs_root *root) 35712619ba1fSChris Mason { 35724df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 35732619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 35742619ba1fSChris Mason (unsigned long)root->root_key.objectid); 35754df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 357676dda93cSYan, Zheng 357776dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 357876dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 357976dda93cSYan, Zheng 35801c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 35813321719eSLiu Bo btrfs_free_log(NULL, root); 35821c1ea4f7SLiu Bo if (root->reloc_root) { 35831c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->node); 35841c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->commit_root); 35851c1ea4f7SLiu Bo btrfs_put_fs_root(root->reloc_root); 35861c1ea4f7SLiu Bo root->reloc_root = NULL; 35871c1ea4f7SLiu Bo } 35881c1ea4f7SLiu Bo } 35893321719eSLiu Bo 3590faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3591581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3592faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3593581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 35944df27c4dSYan, Zheng free_fs_root(root); 35954df27c4dSYan, Zheng } 35964df27c4dSYan, Zheng 35974df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 35984df27c4dSYan, Zheng { 359957cdc8dbSDavid Sterba iput(root->ino_cache_inode); 36004df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 36012ff7e61eSJeff Mahoney btrfs_free_block_rsv(root->fs_info, root->orphan_block_rsv); 36021cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 36030ee5dc67SAl Viro if (root->anon_dev) 36040ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 36058257b2dcSMiao Xie if (root->subv_writers) 36068257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 36075f39d397SChris Mason free_extent_buffer(root->node); 36085f39d397SChris Mason free_extent_buffer(root->commit_root); 3609581bb050SLi Zefan kfree(root->free_ino_ctl); 3610581bb050SLi Zefan kfree(root->free_ino_pinned); 361158176a96SJosef Bacik kfree(root->name); 3612b0feb9d9SMiao Xie btrfs_put_fs_root(root); 36132619ba1fSChris Mason } 36142619ba1fSChris Mason 3615cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3616cb517eabSMiao Xie { 3617cb517eabSMiao Xie free_fs_root(root); 3618cfaa7295SChris Mason } 3619cfaa7295SChris Mason 3620c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3621c146afadSYan Zheng { 3622c146afadSYan Zheng u64 root_objectid = 0; 3623c146afadSYan Zheng struct btrfs_root *gang[8]; 362465d33fd7SQu Wenruo int i = 0; 362565d33fd7SQu Wenruo int err = 0; 362665d33fd7SQu Wenruo unsigned int ret = 0; 362765d33fd7SQu Wenruo int index; 3628c146afadSYan Zheng 3629c146afadSYan Zheng while (1) { 363065d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3631c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3632c146afadSYan Zheng (void **)gang, root_objectid, 3633c146afadSYan Zheng ARRAY_SIZE(gang)); 363465d33fd7SQu Wenruo if (!ret) { 363565d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3636c146afadSYan Zheng break; 363765d33fd7SQu Wenruo } 36385d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 363966b4ffd1SJosef Bacik 364065d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 364165d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 364265d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 364365d33fd7SQu Wenruo gang[i] = NULL; 364465d33fd7SQu Wenruo continue; 364565d33fd7SQu Wenruo } 364665d33fd7SQu Wenruo /* grab all the search result for later use */ 364765d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 364865d33fd7SQu Wenruo } 364965d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 365065d33fd7SQu Wenruo 365165d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 365265d33fd7SQu Wenruo if (!gang[i]) 365365d33fd7SQu Wenruo continue; 3654c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 365566b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 365666b4ffd1SJosef Bacik if (err) 365765d33fd7SQu Wenruo break; 365865d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3659c146afadSYan Zheng } 3660c146afadSYan Zheng root_objectid++; 3661c146afadSYan Zheng } 366265d33fd7SQu Wenruo 366365d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 366465d33fd7SQu Wenruo for (; i < ret; i++) { 366565d33fd7SQu Wenruo if (gang[i]) 366665d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 366765d33fd7SQu Wenruo } 366865d33fd7SQu Wenruo return err; 3669c146afadSYan Zheng } 3670c146afadSYan Zheng 36716bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 3672c146afadSYan Zheng { 36736bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 3674c146afadSYan Zheng struct btrfs_trans_handle *trans; 3675c146afadSYan Zheng 36760b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 36772ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 36780b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 36790b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 3680c71bf099SYan, Zheng 3681c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 36820b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 36830b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 3684c71bf099SYan, Zheng 36857a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 36863612b495STsutomu Itoh if (IS_ERR(trans)) 36873612b495STsutomu Itoh return PTR_ERR(trans); 36883a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 3689c146afadSYan Zheng } 3690c146afadSYan Zheng 36916bccf3abSJeff Mahoney void close_ctree(struct btrfs_fs_info *fs_info) 3692eb60ceacSChris Mason { 36936bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 3694c146afadSYan Zheng int ret; 3695e089f05cSChris Mason 3696afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 3697a2135011SChris Mason 36987343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 3699d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 37007343dd61SJustin Maggard 3701803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3702803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3703803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3704803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3705803b2f54SStefan Behrens 3706837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3707aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3708837d5b6eSIlya Dryomov 37098dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 37108dabb742SStefan Behrens 3711aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 37124cb5300bSChris Mason 37134cb5300bSChris Mason /* wait for any defraggers to finish */ 37144cb5300bSChris Mason wait_event(fs_info->transaction_wait, 37154cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 37164cb5300bSChris Mason 37174cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 371826176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 37194cb5300bSChris Mason 372021c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 372121c7e756SMiao Xie 3722bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 3723e44163e1SJeff Mahoney /* 3724e44163e1SJeff Mahoney * If the cleaner thread is stopped and there are 3725e44163e1SJeff Mahoney * block groups queued for removal, the deletion will be 3726e44163e1SJeff Mahoney * skipped when we quit the cleaner thread. 3727e44163e1SJeff Mahoney */ 37280b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 3729e44163e1SJeff Mahoney 37306bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 3731d397712bSChris Mason if (ret) 373204892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 3733c146afadSYan Zheng } 3734ed2ff2cbSChris Mason 373587533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 37362ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 3737acce952bSliubo 3738e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3739e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 37408929ecfaSYan, Zheng 3741afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 3742f25784b3SYan Zheng 374304892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 3744bcef60f2SArne Jansen 3745963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 374604892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 3747963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3748b0c68f8bSChris Mason } 374931153d81SYan Zheng 37506618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 3751b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 37525ac1d209SJeff Mahoney 3753faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3754d10c5f31SChris Mason 37551a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 37561a4319ccSLiu Bo 3757de348ee0SWang Shilong /* 3758de348ee0SWang Shilong * we must make sure there is not any read request to 3759de348ee0SWang Shilong * submit after we stopping all workers. 3760de348ee0SWang Shilong */ 3761de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 376296192499SJosef Bacik btrfs_stop_all_workers(fs_info); 376396192499SJosef Bacik 37645cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 37655cdd7db6SFilipe Manana 3766afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 376713e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 37689ad6b7bcSChris Mason 376913e6c37bSJosef Bacik iput(fs_info->btree_inode); 3770d6bfde87SChris Mason 377121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 37720b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 37732ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 377421adbd5cSStefan Behrens #endif 377521adbd5cSStefan Behrens 3776dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 37770b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3778b248a415SChris Mason 3779e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3780963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3781c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 378276dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 37830b86a832SChris Mason 378453b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3785fd708b81SJosef Bacik btrfs_free_ref_cache(fs_info); 378653b381b3SDavid Woodhouse 3787cdfb080eSChris Mason __btrfs_free_block_rsv(root->orphan_block_rsv); 37881cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 378904216820SFilipe Manana 379004216820SFilipe Manana while (!list_empty(&fs_info->pinned_chunks)) { 379104216820SFilipe Manana struct extent_map *em; 379204216820SFilipe Manana 379304216820SFilipe Manana em = list_first_entry(&fs_info->pinned_chunks, 379404216820SFilipe Manana struct extent_map, list); 379504216820SFilipe Manana list_del_init(&em->list); 379604216820SFilipe Manana free_extent_map(em); 379704216820SFilipe Manana } 3798eb60ceacSChris Mason } 3799eb60ceacSChris Mason 3800b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3801b9fab919SChris Mason int atomic) 3802ccd467d6SChris Mason { 38031259ab75SChris Mason int ret; 3804727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 38051259ab75SChris Mason 38060b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 38071259ab75SChris Mason if (!ret) 38081259ab75SChris Mason return ret; 38091259ab75SChris Mason 38101259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3811b9fab919SChris Mason parent_transid, atomic); 3812b9fab919SChris Mason if (ret == -EAGAIN) 3813b9fab919SChris Mason return ret; 38141259ab75SChris Mason return !ret; 38155f39d397SChris Mason } 38166702ed49SChris Mason 38175f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 38185f39d397SChris Mason { 38190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info; 382006ea65a3SJosef Bacik struct btrfs_root *root; 38215f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3822b9473439SChris Mason int was_dirty; 3823b4ce94deSChris Mason 382406ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 382506ea65a3SJosef Bacik /* 382606ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 382706ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 382806ea65a3SJosef Bacik * outside of the sanity tests. 382906ea65a3SJosef Bacik */ 383006ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 383106ea65a3SJosef Bacik return; 383206ea65a3SJosef Bacik #endif 383306ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 38340b246afaSJeff Mahoney fs_info = root->fs_info; 3835b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 38360b246afaSJeff Mahoney if (transid != fs_info->generation) 38375d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 38380b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 38390b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3840e2d84521SMiao Xie if (!was_dirty) 3841104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 3842e2d84521SMiao Xie buf->len, 38430b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 38441f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 384569fc6cbbSQu Wenruo /* 384669fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 384769fc6cbbSQu Wenruo * but item data not updated. 384869fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 384969fc6cbbSQu Wenruo */ 385069fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 385169fc6cbbSQu Wenruo btrfs_check_leaf_relaxed(root, buf)) { 3852a4f78750SDavid Sterba btrfs_print_leaf(buf); 38531f21ef0aSFilipe Manana ASSERT(0); 38541f21ef0aSFilipe Manana } 38551f21ef0aSFilipe Manana #endif 3856eb60ceacSChris Mason } 3857eb60ceacSChris Mason 38582ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 3859b53d3f5dSLiu Bo int flush_delayed) 386035b7e476SChris Mason { 3861188de649SChris Mason /* 3862188de649SChris Mason * looks as though older kernels can get into trouble with 3863188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3864188de649SChris Mason */ 3865e2d84521SMiao Xie int ret; 3866d6bfde87SChris Mason 38676933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 3868d6bfde87SChris Mason return; 3869d6bfde87SChris Mason 3870b53d3f5dSLiu Bo if (flush_delayed) 38712ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 387216cdcec7SMiao Xie 38730b246afaSJeff Mahoney ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 3874e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 3875e2d84521SMiao Xie if (ret > 0) { 38760b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 387716cdcec7SMiao Xie } 387816cdcec7SMiao Xie } 387916cdcec7SMiao Xie 38802ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 388116cdcec7SMiao Xie { 38822ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 388335b7e476SChris Mason } 3884b53d3f5dSLiu Bo 38852ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 3886b53d3f5dSLiu Bo { 38872ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 3888d6bfde87SChris Mason } 38896b80053dSChris Mason 3890ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 38916b80053dSChris Mason { 3892727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 38932ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 38942ff7e61eSJeff Mahoney 38952ff7e61eSJeff Mahoney return btree_read_extent_buffer_pages(fs_info, buf, parent_transid); 38966b80053dSChris Mason } 38970da5468fSChris Mason 38983d3a126aSDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info) 3899acce952bSliubo { 3900c926093eSDavid Sterba struct btrfs_super_block *sb = fs_info->super_copy; 3901319e4d06SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 3902319e4d06SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 3903c926093eSDavid Sterba int ret = 0; 3904c926093eSDavid Sterba 3905319e4d06SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 3906ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "no valid FS found"); 3907319e4d06SQu Wenruo ret = -EINVAL; 3908319e4d06SQu Wenruo } 3909319e4d06SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) 3910ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "unrecognized super flag: %llu", 3911319e4d06SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 391221e7626bSDavid Sterba if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 3913ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "tree_root level too big: %d >= %d", 391421e7626bSDavid Sterba btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 3915c926093eSDavid Sterba ret = -EINVAL; 3916c926093eSDavid Sterba } 391721e7626bSDavid Sterba if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 3918ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 391921e7626bSDavid Sterba btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 3920c926093eSDavid Sterba ret = -EINVAL; 3921c926093eSDavid Sterba } 392221e7626bSDavid Sterba if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 3923ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "log_root level too big: %d >= %d", 392421e7626bSDavid Sterba btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 3925c926093eSDavid Sterba ret = -EINVAL; 3926c926093eSDavid Sterba } 3927c926093eSDavid Sterba 39281104a885SDavid Sterba /* 3929319e4d06SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 3930319e4d06SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 39311104a885SDavid Sterba */ 3932319e4d06SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 3933319e4d06SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 3934ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 393575d6ad38SDavid Sterba ret = -EINVAL; 393675d6ad38SDavid Sterba } 3937319e4d06SQu Wenruo /* Only PAGE SIZE is supported yet */ 393809cbfeafSKirill A. Shutemov if (sectorsize != PAGE_SIZE) { 3939ab8d0fc4SJeff Mahoney btrfs_err(fs_info, 3940ab8d0fc4SJeff Mahoney "sectorsize %llu not supported yet, only support %lu", 394109cbfeafSKirill A. Shutemov sectorsize, PAGE_SIZE); 3942319e4d06SQu Wenruo ret = -EINVAL; 3943319e4d06SQu Wenruo } 3944319e4d06SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 3945319e4d06SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 3946ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 3947319e4d06SQu Wenruo ret = -EINVAL; 3948319e4d06SQu Wenruo } 3949319e4d06SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 3950ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 3951ab8d0fc4SJeff Mahoney le32_to_cpu(sb->__unused_leafsize), nodesize); 3952319e4d06SQu Wenruo ret = -EINVAL; 3953319e4d06SQu Wenruo } 3954319e4d06SQu Wenruo 3955319e4d06SQu Wenruo /* Root alignment check */ 3956319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 3957ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "tree_root block unaligned: %llu", 3958319e4d06SQu Wenruo btrfs_super_root(sb)); 3959319e4d06SQu Wenruo ret = -EINVAL; 3960319e4d06SQu Wenruo } 3961319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 3962ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 3963319e4d06SQu Wenruo btrfs_super_chunk_root(sb)); 3964319e4d06SQu Wenruo ret = -EINVAL; 3965319e4d06SQu Wenruo } 3966319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 3967ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "log_root block unaligned: %llu", 3968319e4d06SQu Wenruo btrfs_super_log_root(sb)); 396975d6ad38SDavid Sterba ret = -EINVAL; 397075d6ad38SDavid Sterba } 397175d6ad38SDavid Sterba 397244880fdcSAnand Jain if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { 3973ab8d0fc4SJeff Mahoney btrfs_err(fs_info, 3974ab8d0fc4SJeff Mahoney "dev_item UUID does not match fsid: %pU != %pU", 3975c926093eSDavid Sterba fs_info->fsid, sb->dev_item.fsid); 3976c926093eSDavid Sterba ret = -EINVAL; 3977c926093eSDavid Sterba } 3978c926093eSDavid Sterba 3979c926093eSDavid Sterba /* 3980c926093eSDavid Sterba * Hint to catch really bogus numbers, bitflips or so, more exact checks are 3981c926093eSDavid Sterba * done later 3982c926093eSDavid Sterba */ 398399e3ecfcSLiu Bo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 398499e3ecfcSLiu Bo btrfs_err(fs_info, "bytes_used is too small %llu", 398599e3ecfcSLiu Bo btrfs_super_bytes_used(sb)); 398699e3ecfcSLiu Bo ret = -EINVAL; 398799e3ecfcSLiu Bo } 3988b7f67055SChandan Rajendra if (!is_power_of_2(btrfs_super_stripesize(sb))) { 398999e3ecfcSLiu Bo btrfs_err(fs_info, "invalid stripesize %u", 399099e3ecfcSLiu Bo btrfs_super_stripesize(sb)); 399199e3ecfcSLiu Bo ret = -EINVAL; 399299e3ecfcSLiu Bo } 399321e7626bSDavid Sterba if (btrfs_super_num_devices(sb) > (1UL << 31)) 3994ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "suspicious number of devices: %llu", 399521e7626bSDavid Sterba btrfs_super_num_devices(sb)); 399675d6ad38SDavid Sterba if (btrfs_super_num_devices(sb) == 0) { 3997ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "number of devices is 0"); 399875d6ad38SDavid Sterba ret = -EINVAL; 399975d6ad38SDavid Sterba } 4000c926093eSDavid Sterba 400121e7626bSDavid Sterba if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { 4002ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "super offset mismatch %llu != %u", 400321e7626bSDavid Sterba btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 4004c926093eSDavid Sterba ret = -EINVAL; 4005c926093eSDavid Sterba } 4006c926093eSDavid Sterba 4007c926093eSDavid Sterba /* 4008ce7fca5fSDavid Sterba * Obvious sys_chunk_array corruptions, it must hold at least one key 4009ce7fca5fSDavid Sterba * and one chunk 4010ce7fca5fSDavid Sterba */ 4011ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 4012ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "system chunk array too big %u > %u", 4013ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4014ce7fca5fSDavid Sterba BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 4015ce7fca5fSDavid Sterba ret = -EINVAL; 4016ce7fca5fSDavid Sterba } 4017ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 4018ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)) { 4019ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "system chunk array too small %u < %zu", 4020ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4021ce7fca5fSDavid Sterba sizeof(struct btrfs_disk_key) 4022ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)); 4023ce7fca5fSDavid Sterba ret = -EINVAL; 4024ce7fca5fSDavid Sterba } 4025ce7fca5fSDavid Sterba 4026ce7fca5fSDavid Sterba /* 4027c926093eSDavid Sterba * The generation is a global counter, we'll trust it more than the others 4028c926093eSDavid Sterba * but it's still possible that it's the one that's wrong. 4029c926093eSDavid Sterba */ 403021e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 4031ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, 4032ab8d0fc4SJeff Mahoney "suspicious: generation < chunk_root_generation: %llu < %llu", 4033ab8d0fc4SJeff Mahoney btrfs_super_generation(sb), 4034ab8d0fc4SJeff Mahoney btrfs_super_chunk_root_generation(sb)); 403521e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 403621e7626bSDavid Sterba && btrfs_super_cache_generation(sb) != (u64)-1) 4037ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, 4038ab8d0fc4SJeff Mahoney "suspicious: generation < cache_generation: %llu < %llu", 4039ab8d0fc4SJeff Mahoney btrfs_super_generation(sb), 4040ab8d0fc4SJeff Mahoney btrfs_super_cache_generation(sb)); 4041c926093eSDavid Sterba 4042c926093eSDavid Sterba return ret; 4043fcd1f065SDavid Sterba } 4044fcd1f065SDavid Sterba 40452ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4046acce952bSliubo { 40470b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 40482ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 40490b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4050acce952bSliubo 40510b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 40520b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4053acce952bSliubo 4054acce952bSliubo /* cleanup FS via transaction */ 40552ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 4056acce952bSliubo } 4057acce952bSliubo 4058143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4059acce952bSliubo { 4060acce952bSliubo struct btrfs_ordered_extent *ordered; 4061acce952bSliubo 4062199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4063779880efSJosef Bacik /* 4064779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4065779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4066779880efSJosef Bacik */ 4067199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4068779880efSJosef Bacik root_extent_list) 4069779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4070199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4071199c2a9cSMiao Xie } 4072199c2a9cSMiao Xie 4073199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4074199c2a9cSMiao Xie { 4075199c2a9cSMiao Xie struct btrfs_root *root; 4076199c2a9cSMiao Xie struct list_head splice; 4077199c2a9cSMiao Xie 4078199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4079199c2a9cSMiao Xie 4080199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4081199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4082199c2a9cSMiao Xie while (!list_empty(&splice)) { 4083199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4084199c2a9cSMiao Xie ordered_root); 40851de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 40861de2cfdeSJosef Bacik &fs_info->ordered_roots); 4087199c2a9cSMiao Xie 40882a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4089199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4090199c2a9cSMiao Xie 40912a85d9caSLiu Bo cond_resched(); 40922a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4093199c2a9cSMiao Xie } 4094199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 4095acce952bSliubo } 4096acce952bSliubo 409735a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 40982ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4099acce952bSliubo { 4100acce952bSliubo struct rb_node *node; 4101acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4102acce952bSliubo struct btrfs_delayed_ref_node *ref; 4103acce952bSliubo int ret = 0; 4104acce952bSliubo 4105acce952bSliubo delayed_refs = &trans->delayed_refs; 4106acce952bSliubo 4107acce952bSliubo spin_lock(&delayed_refs->lock); 4108d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4109cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 41100b246afaSJeff Mahoney btrfs_info(fs_info, "delayed_refs has NO entry"); 4111acce952bSliubo return ret; 4112acce952bSliubo } 4113acce952bSliubo 4114d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 4115d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 41160e0adbcfSJosef Bacik struct rb_node *n; 4117e78417d1SJosef Bacik bool pin_bytes = false; 4118acce952bSliubo 4119d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4120d7df2c79SJosef Bacik href_node); 4121b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 4122d278850eSJosef Bacik refcount_inc(&head->refs); 412379787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 4124b939d1abSJosef Bacik 4125acce952bSliubo mutex_lock(&head->mutex); 4126b939d1abSJosef Bacik mutex_unlock(&head->mutex); 4127d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4128e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 4129b939d1abSJosef Bacik continue; 4130b939d1abSJosef Bacik } 4131d7df2c79SJosef Bacik spin_lock(&head->lock); 41320e0adbcfSJosef Bacik while ((n = rb_first(&head->ref_tree)) != NULL) { 41330e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 41340e0adbcfSJosef Bacik ref_node); 4135d7df2c79SJosef Bacik ref->in_tree = 0; 41360e0adbcfSJosef Bacik rb_erase(&ref->ref_node, &head->ref_tree); 41370e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 41381d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 41391d57ee94SWang Xiaoguang list_del(&ref->add_list); 4140d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4141d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4142d7df2c79SJosef Bacik } 414354067ae9SJosef Bacik if (head->must_insert_reserved) 4144e78417d1SJosef Bacik pin_bytes = true; 414578a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4146acce952bSliubo delayed_refs->num_heads--; 4147d7df2c79SJosef Bacik if (head->processing == 0) 4148acce952bSliubo delayed_refs->num_heads_ready--; 4149d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4150c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 4151d278850eSJosef Bacik RB_CLEAR_NODE(&head->href_node); 4152d7df2c79SJosef Bacik spin_unlock(&head->lock); 4153acce952bSliubo spin_unlock(&delayed_refs->lock); 4154e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4155acce952bSliubo 4156d7df2c79SJosef Bacik if (pin_bytes) 4157d278850eSJosef Bacik btrfs_pin_extent(fs_info, head->bytenr, 4158d278850eSJosef Bacik head->num_bytes, 1); 4159d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4160acce952bSliubo cond_resched(); 4161acce952bSliubo spin_lock(&delayed_refs->lock); 4162acce952bSliubo } 4163acce952bSliubo 4164acce952bSliubo spin_unlock(&delayed_refs->lock); 4165acce952bSliubo 4166acce952bSliubo return ret; 4167acce952bSliubo } 4168acce952bSliubo 4169143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4170acce952bSliubo { 4171acce952bSliubo struct btrfs_inode *btrfs_inode; 4172acce952bSliubo struct list_head splice; 4173acce952bSliubo 4174acce952bSliubo INIT_LIST_HEAD(&splice); 4175acce952bSliubo 4176eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4177eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4178acce952bSliubo 4179acce952bSliubo while (!list_empty(&splice)) { 4180eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4181acce952bSliubo delalloc_inodes); 4182acce952bSliubo 4183acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 4184df0af1a5SMiao Xie clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 4185df0af1a5SMiao Xie &btrfs_inode->runtime_flags); 4186eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4187acce952bSliubo 4188acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 4189b216cbfbSMiao Xie 4190eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4191acce952bSliubo } 4192acce952bSliubo 4193eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4194eb73c1b7SMiao Xie } 4195eb73c1b7SMiao Xie 4196eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4197eb73c1b7SMiao Xie { 4198eb73c1b7SMiao Xie struct btrfs_root *root; 4199eb73c1b7SMiao Xie struct list_head splice; 4200eb73c1b7SMiao Xie 4201eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4202eb73c1b7SMiao Xie 4203eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4204eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4205eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4206eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4207eb73c1b7SMiao Xie delalloc_root); 4208eb73c1b7SMiao Xie list_del_init(&root->delalloc_root); 4209eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 4210eb73c1b7SMiao Xie BUG_ON(!root); 4211eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4212eb73c1b7SMiao Xie 4213eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 4214eb73c1b7SMiao Xie btrfs_put_fs_root(root); 4215eb73c1b7SMiao Xie 4216eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4217eb73c1b7SMiao Xie } 4218eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4219acce952bSliubo } 4220acce952bSliubo 42212ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4222acce952bSliubo struct extent_io_tree *dirty_pages, 4223acce952bSliubo int mark) 4224acce952bSliubo { 4225acce952bSliubo int ret; 4226acce952bSliubo struct extent_buffer *eb; 4227acce952bSliubo u64 start = 0; 4228acce952bSliubo u64 end; 4229acce952bSliubo 4230acce952bSliubo while (1) { 4231acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4232e6138876SJosef Bacik mark, NULL); 4233acce952bSliubo if (ret) 4234acce952bSliubo break; 4235acce952bSliubo 423691166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4237acce952bSliubo while (start <= end) { 42380b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 42390b246afaSJeff Mahoney start += fs_info->nodesize; 4240fd8b2b61SJosef Bacik if (!eb) 4241acce952bSliubo continue; 4242fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4243acce952bSliubo 4244fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4245fd8b2b61SJosef Bacik &eb->bflags)) 4246fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4247fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4248acce952bSliubo } 4249acce952bSliubo } 4250acce952bSliubo 4251acce952bSliubo return ret; 4252acce952bSliubo } 4253acce952bSliubo 42542ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4255acce952bSliubo struct extent_io_tree *pinned_extents) 4256acce952bSliubo { 4257acce952bSliubo struct extent_io_tree *unpin; 4258acce952bSliubo u64 start; 4259acce952bSliubo u64 end; 4260acce952bSliubo int ret; 4261ed0eaa14SLiu Bo bool loop = true; 4262acce952bSliubo 4263acce952bSliubo unpin = pinned_extents; 4264ed0eaa14SLiu Bo again: 4265acce952bSliubo while (1) { 4266acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4267e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4268acce952bSliubo if (ret) 4269acce952bSliubo break; 4270acce952bSliubo 4271af6f8f60SDavid Sterba clear_extent_dirty(unpin, start, end); 42722ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4273acce952bSliubo cond_resched(); 4274acce952bSliubo } 4275acce952bSliubo 4276ed0eaa14SLiu Bo if (loop) { 42770b246afaSJeff Mahoney if (unpin == &fs_info->freed_extents[0]) 42780b246afaSJeff Mahoney unpin = &fs_info->freed_extents[1]; 4279ed0eaa14SLiu Bo else 42800b246afaSJeff Mahoney unpin = &fs_info->freed_extents[0]; 4281ed0eaa14SLiu Bo loop = false; 4282ed0eaa14SLiu Bo goto again; 4283ed0eaa14SLiu Bo } 4284ed0eaa14SLiu Bo 4285acce952bSliubo return 0; 4286acce952bSliubo } 4287acce952bSliubo 4288c79a1751SLiu Bo static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) 4289c79a1751SLiu Bo { 4290c79a1751SLiu Bo struct inode *inode; 4291c79a1751SLiu Bo 4292c79a1751SLiu Bo inode = cache->io_ctl.inode; 4293c79a1751SLiu Bo if (inode) { 4294c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4295c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4296c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4297c79a1751SLiu Bo iput(inode); 4298c79a1751SLiu Bo } 4299c79a1751SLiu Bo btrfs_put_block_group(cache); 4300c79a1751SLiu Bo } 4301c79a1751SLiu Bo 4302c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 43032ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4304c79a1751SLiu Bo { 4305c79a1751SLiu Bo struct btrfs_block_group_cache *cache; 4306c79a1751SLiu Bo 4307c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4308c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4309c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 4310c79a1751SLiu Bo struct btrfs_block_group_cache, 4311c79a1751SLiu Bo dirty_list); 4312c79a1751SLiu Bo if (!cache) { 43130b246afaSJeff Mahoney btrfs_err(fs_info, "orphan block group dirty_bgs list"); 4314c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4315c79a1751SLiu Bo return; 4316c79a1751SLiu Bo } 4317c79a1751SLiu Bo 4318c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4319c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4320c79a1751SLiu Bo list_del_init(&cache->io_list); 4321c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4322c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4323c79a1751SLiu Bo } 4324c79a1751SLiu Bo 4325c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4326c79a1751SLiu Bo spin_lock(&cache->lock); 4327c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4328c79a1751SLiu Bo spin_unlock(&cache->lock); 4329c79a1751SLiu Bo 4330c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4331c79a1751SLiu Bo btrfs_put_block_group(cache); 4332c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4333c79a1751SLiu Bo } 4334c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4335c79a1751SLiu Bo 4336c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4337c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 4338c79a1751SLiu Bo struct btrfs_block_group_cache, 4339c79a1751SLiu Bo io_list); 4340c79a1751SLiu Bo if (!cache) { 43410b246afaSJeff Mahoney btrfs_err(fs_info, "orphan block group on io_bgs list"); 4342c79a1751SLiu Bo return; 4343c79a1751SLiu Bo } 4344c79a1751SLiu Bo 4345c79a1751SLiu Bo list_del_init(&cache->io_list); 4346c79a1751SLiu Bo spin_lock(&cache->lock); 4347c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4348c79a1751SLiu Bo spin_unlock(&cache->lock); 4349c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4350c79a1751SLiu Bo } 4351c79a1751SLiu Bo } 4352c79a1751SLiu Bo 435349b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 43542ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 435549b25e05SJeff Mahoney { 43562ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4357c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4358c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4359c79a1751SLiu Bo 43602ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 436149b25e05SJeff Mahoney 43624a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 43630b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 436449b25e05SJeff Mahoney 43654a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 43660b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 436749b25e05SJeff Mahoney 4368ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4369ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 437067cde344SMiao Xie 43712ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 437249b25e05SJeff Mahoney EXTENT_DIRTY); 43732ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, 43740b246afaSJeff Mahoney fs_info->pinned_extents); 437549b25e05SJeff Mahoney 43764a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 43774a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 437849b25e05SJeff Mahoney } 437949b25e05SJeff Mahoney 43802ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4381acce952bSliubo { 4382acce952bSliubo struct btrfs_transaction *t; 4383acce952bSliubo 43840b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4385acce952bSliubo 43860b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 43870b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 43880b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4389724e2315SJosef Bacik struct btrfs_transaction, list); 4390724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 43919b64f57dSElena Reshetova refcount_inc(&t->use_count); 43920b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 43932ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4394724e2315SJosef Bacik btrfs_put_transaction(t); 43950b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4396724e2315SJosef Bacik continue; 4397724e2315SJosef Bacik } 43980b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4399724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 44000b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4401724e2315SJosef Bacik /* 4402724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4403724e2315SJosef Bacik * handle open currently for this transaction. 4404724e2315SJosef Bacik */ 4405724e2315SJosef Bacik wait_event(t->writer_wait, 4406724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4407724e2315SJosef Bacik } else { 44080b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4409724e2315SJosef Bacik } 44102ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4411724e2315SJosef Bacik 44120b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 44130b246afaSJeff Mahoney if (t == fs_info->running_transaction) 44140b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4415724e2315SJosef Bacik list_del_init(&t->list); 44160b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4417a4abeea4SJosef Bacik 4418724e2315SJosef Bacik btrfs_put_transaction(t); 44192ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 44200b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4421724e2315SJosef Bacik } 44220b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 44230b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4424ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4425ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 44262ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 44270b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 44280b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4429acce952bSliubo 4430acce952bSliubo return 0; 4431acce952bSliubo } 4432acce952bSliubo 4433c6100a4bSJosef Bacik static struct btrfs_fs_info *btree_fs_info(void *private_data) 4434c6100a4bSJosef Bacik { 4435c6100a4bSJosef Bacik struct inode *inode = private_data; 4436c6100a4bSJosef Bacik return btrfs_sb(inode->i_sb); 4437c6100a4bSJosef Bacik } 4438c6100a4bSJosef Bacik 4439e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 44404d53dddbSDavid Sterba /* mandatory callbacks */ 44410b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 44424d53dddbSDavid Sterba .readpage_end_io_hook = btree_readpage_end_io_hook, 4443239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4444239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 444520a7db8aSDavid Sterba .readpage_io_failed_hook = btree_io_failed_hook, 4446c6100a4bSJosef Bacik .set_range_writeback = btrfs_set_range_writeback, 4447c6100a4bSJosef Bacik .tree_fs_info = btree_fs_info, 44484d53dddbSDavid Sterba 44494d53dddbSDavid Sterba /* optional callbacks */ 44500da5468fSChris Mason }; 4451