16cbd5570SChris Mason /* 26cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 36cbd5570SChris Mason * 46cbd5570SChris Mason * This program is free software; you can redistribute it and/or 56cbd5570SChris Mason * modify it under the terms of the GNU General Public 66cbd5570SChris Mason * License v2 as published by the Free Software Foundation. 76cbd5570SChris Mason * 86cbd5570SChris Mason * This program is distributed in the hope that it will be useful, 96cbd5570SChris Mason * but WITHOUT ANY WARRANTY; without even the implied warranty of 106cbd5570SChris Mason * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 116cbd5570SChris Mason * General Public License for more details. 126cbd5570SChris Mason * 136cbd5570SChris Mason * You should have received a copy of the GNU General Public 146cbd5570SChris Mason * License along with this program; if not, write to the 156cbd5570SChris Mason * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 166cbd5570SChris Mason * Boston, MA 021110-1307, USA. 176cbd5570SChris Mason */ 186cbd5570SChris Mason 19e20d96d6SChris Mason #include <linux/fs.h> 20d98237b3SChris Mason #include <linux/blkdev.h> 2187cbda5cSChris Mason #include <linux/scatterlist.h> 2222b0ebdaSChris Mason #include <linux/swap.h> 230f7d52f4SChris Mason #include <linux/radix-tree.h> 2435b7e476SChris Mason #include <linux/writeback.h> 25d397712bSChris Mason #include <linux/buffer_head.h> 26ce9adaa5SChris Mason #include <linux/workqueue.h> 27a74a4b97SChris Mason #include <linux/kthread.h> 28a74a4b97SChris Mason #include <linux/freezer.h> 295a0e3ad6STejun Heo #include <linux/slab.h> 30784b4e29SChris Mason #include <linux/migrate.h> 317a36ddecSDavid Sterba #include <linux/ratelimit.h> 326463fe58SStefan Behrens #include <linux/uuid.h> 33803b2f54SStefan Behrens #include <linux/semaphore.h> 347e75bf3fSDavid Sterba #include <asm/unaligned.h> 35eb60ceacSChris Mason #include "ctree.h" 36eb60ceacSChris Mason #include "disk-io.h" 370b947affSFilipe David Borba Manana #include "hash.h" 38e089f05cSChris Mason #include "transaction.h" 390f7d52f4SChris Mason #include "btrfs_inode.h" 400b86a832SChris Mason #include "volumes.h" 41db94535dSChris Mason #include "print-tree.h" 42925baeddSChris Mason #include "locking.h" 43e02119d5SChris Mason #include "tree-log.h" 44fa9c0d79SChris Mason #include "free-space-cache.h" 4570f6d82eSOmar Sandoval #include "free-space-tree.h" 46581bb050SLi Zefan #include "inode-map.h" 4721adbd5cSStefan Behrens #include "check-integrity.h" 48606686eeSJosef Bacik #include "rcu-string.h" 498dabb742SStefan Behrens #include "dev-replace.h" 5053b381b3SDavid Woodhouse #include "raid56.h" 515ac1d209SJeff Mahoney #include "sysfs.h" 52fcebe456SJosef Bacik #include "qgroup.h" 53eb60ceacSChris Mason 54de0022b9SJosef Bacik #ifdef CONFIG_X86 55de0022b9SJosef Bacik #include <asm/cpufeature.h> 56de0022b9SJosef Bacik #endif 57de0022b9SJosef Bacik 58319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 59319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 60319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 61319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 62319e4d06SQu Wenruo BTRFS_SUPER_FLAG_METADUMP) 63319e4d06SQu Wenruo 64e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops; 658b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 664df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root); 67fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 68acce952bSliubo int read_only); 69143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 70acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 71acce952bSliubo struct btrfs_root *root); 72143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 73acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 74acce952bSliubo struct extent_io_tree *dirty_pages, 75acce952bSliubo int mark); 76acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 77acce952bSliubo struct extent_io_tree *pinned_extents); 7848a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root); 7948a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root); 80ce9adaa5SChris Mason 81d352ac68SChris Mason /* 8297eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 8397eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 84d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 85d352ac68SChris Mason */ 8697eb6b69SDavid Sterba struct btrfs_end_io_wq { 87ce9adaa5SChris Mason struct bio *bio; 88ce9adaa5SChris Mason bio_end_io_t *end_io; 89ce9adaa5SChris Mason void *private; 90ce9adaa5SChris Mason struct btrfs_fs_info *info; 91ce9adaa5SChris Mason int error; 92bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 93ce9adaa5SChris Mason struct list_head list; 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, 10497eb6b69SDavid Sterba SLAB_RECLAIM_ACCOUNT | 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 if (btrfs_end_io_wq_cache) 11497eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 11597eb6b69SDavid Sterba } 11697eb6b69SDavid Sterba 117d352ac68SChris Mason /* 118d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 119d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 120d352ac68SChris Mason * just before they are sent down the IO stack. 121d352ac68SChris Mason */ 12244b8bd7eSChris Mason struct async_submit_bio { 12344b8bd7eSChris Mason struct inode *inode; 12444b8bd7eSChris Mason struct bio *bio; 12544b8bd7eSChris Mason struct list_head list; 1264a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start; 1274a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done; 12844b8bd7eSChris Mason int rw; 12944b8bd7eSChris Mason int mirror_num; 130c8b97818SChris Mason unsigned long bio_flags; 131eaf25d93SChris Mason /* 132eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 133eaf25d93SChris Mason * can't tell us where in the file the bio should go 134eaf25d93SChris Mason */ 135eaf25d93SChris Mason u64 bio_offset; 1368b712842SChris Mason struct btrfs_work work; 13779787eaaSJeff Mahoney int error; 13844b8bd7eSChris Mason }; 13944b8bd7eSChris Mason 14085d4e461SChris Mason /* 14185d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 14285d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 14385d4e461SChris Mason * the level the eb occupies in the tree. 1444008c04aSChris Mason * 14585d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 14685d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 14785d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 14885d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 14985d4e461SChris Mason * have separate keysets. 1504008c04aSChris Mason * 15185d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 15285d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 15385d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1544008c04aSChris Mason * 15585d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 15685d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 15785d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 15885d4e461SChris Mason * 15985d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 16085d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 16185d4e461SChris Mason * needs update as well. 1624008c04aSChris Mason */ 1634008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1644008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1654008c04aSChris Mason # error 1664008c04aSChris Mason # endif 16785d4e461SChris Mason 16885d4e461SChris Mason static struct btrfs_lockdep_keyset { 16985d4e461SChris Mason u64 id; /* root objectid */ 17085d4e461SChris Mason const char *name_stem; /* lock name stem */ 17185d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 17285d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 17385d4e461SChris Mason } btrfs_lockdep_keysets[] = { 17485d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 17585d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 17685d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 17785d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 17885d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 17985d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 18060b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 18185d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 18285d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 18385d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 18413fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1856b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 18685d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1874008c04aSChris Mason }; 18885d4e461SChris Mason 18985d4e461SChris Mason void __init btrfs_init_lockdep(void) 19085d4e461SChris Mason { 19185d4e461SChris Mason int i, j; 19285d4e461SChris Mason 19385d4e461SChris Mason /* initialize lockdep class names */ 19485d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 19585d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 19685d4e461SChris Mason 19785d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 19885d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 19985d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 20085d4e461SChris Mason } 20185d4e461SChris Mason } 20285d4e461SChris Mason 20385d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 20485d4e461SChris Mason int level) 20585d4e461SChris Mason { 20685d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 20785d4e461SChris Mason 20885d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 20985d4e461SChris Mason 21085d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 21185d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 21285d4e461SChris Mason if (ks->id == objectid) 21385d4e461SChris Mason break; 21485d4e461SChris Mason 21585d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 21685d4e461SChris Mason &ks->keys[level], ks->names[level]); 21785d4e461SChris Mason } 21885d4e461SChris Mason 2194008c04aSChris Mason #endif 2204008c04aSChris Mason 221d352ac68SChris Mason /* 222d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 223d352ac68SChris Mason * that covers the entire device 224d352ac68SChris Mason */ 225b2950863SChristoph Hellwig static struct extent_map *btree_get_extent(struct inode *inode, 226306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 2275f39d397SChris Mason int create) 2285f39d397SChris Mason { 2295f39d397SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 2305f39d397SChris Mason struct extent_map *em; 2315f39d397SChris Mason int ret; 2325f39d397SChris Mason 233890871beSChris Mason read_lock(&em_tree->lock); 234d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 235a061fc8dSChris Mason if (em) { 236a061fc8dSChris Mason em->bdev = 237a061fc8dSChris Mason BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 238890871beSChris Mason read_unlock(&em_tree->lock); 2395f39d397SChris Mason goto out; 240a061fc8dSChris Mason } 241890871beSChris Mason read_unlock(&em_tree->lock); 2427b13b7b1SChris Mason 243172ddd60SDavid Sterba em = alloc_extent_map(); 2445f39d397SChris Mason if (!em) { 2455f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2465f39d397SChris Mason goto out; 2475f39d397SChris Mason } 2485f39d397SChris Mason em->start = 0; 2490afbaf8cSChris Mason em->len = (u64)-1; 250c8b97818SChris Mason em->block_len = (u64)-1; 2515f39d397SChris Mason em->block_start = 0; 252a061fc8dSChris Mason em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 253d1310b2eSChris Mason 254890871beSChris Mason write_lock(&em_tree->lock); 25509a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2565f39d397SChris Mason if (ret == -EEXIST) { 2575f39d397SChris Mason free_extent_map(em); 2587b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 259b4f359abSTsutomu Itoh if (!em) 2600433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2615f39d397SChris Mason } else if (ret) { 2627b13b7b1SChris Mason free_extent_map(em); 2630433f20dSTsutomu Itoh em = ERR_PTR(ret); 2645f39d397SChris Mason } 265890871beSChris Mason write_unlock(&em_tree->lock); 2667b13b7b1SChris Mason 2675f39d397SChris Mason out: 2685f39d397SChris Mason return em; 2695f39d397SChris Mason } 2705f39d397SChris Mason 271b0496686SLiu Bo u32 btrfs_csum_data(char *data, u32 seed, size_t len) 27219c00ddcSChris Mason { 2730b947affSFilipe David Borba Manana return btrfs_crc32c(seed, data, len); 27419c00ddcSChris Mason } 27519c00ddcSChris Mason 27619c00ddcSChris Mason void btrfs_csum_final(u32 crc, char *result) 27719c00ddcSChris Mason { 2787e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 27919c00ddcSChris Mason } 28019c00ddcSChris Mason 281d352ac68SChris Mason /* 282d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 283d352ac68SChris Mason * into the csum field of the block. 284d352ac68SChris Mason */ 28501d58472SDaniel Dressler static int csum_tree_block(struct btrfs_fs_info *fs_info, 28601d58472SDaniel Dressler struct extent_buffer *buf, 28719c00ddcSChris Mason int verify) 28819c00ddcSChris Mason { 28901d58472SDaniel Dressler u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 290607d432dSJosef Bacik char *result = NULL; 29119c00ddcSChris Mason unsigned long len; 29219c00ddcSChris Mason unsigned long cur_len; 29319c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 29419c00ddcSChris Mason char *kaddr; 29519c00ddcSChris Mason unsigned long map_start; 29619c00ddcSChris Mason unsigned long map_len; 29719c00ddcSChris Mason int err; 29819c00ddcSChris Mason u32 crc = ~(u32)0; 299607d432dSJosef Bacik unsigned long inline_result; 30019c00ddcSChris Mason 30119c00ddcSChris Mason len = buf->len - offset; 30219c00ddcSChris Mason while (len > 0) { 30319c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 304a6591715SChris Mason &kaddr, &map_start, &map_len); 305d397712bSChris Mason if (err) 30619c00ddcSChris Mason return 1; 30719c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 308b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 30919c00ddcSChris Mason crc, cur_len); 31019c00ddcSChris Mason len -= cur_len; 31119c00ddcSChris Mason offset += cur_len; 31219c00ddcSChris Mason } 313607d432dSJosef Bacik if (csum_size > sizeof(inline_result)) { 31431e818feSDavid Sterba result = kzalloc(csum_size, GFP_NOFS); 315607d432dSJosef Bacik if (!result) 316607d432dSJosef Bacik return 1; 317607d432dSJosef Bacik } else { 318607d432dSJosef Bacik result = (char *)&inline_result; 319607d432dSJosef Bacik } 320607d432dSJosef Bacik 32119c00ddcSChris Mason btrfs_csum_final(crc, result); 32219c00ddcSChris Mason 32319c00ddcSChris Mason if (verify) { 324607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 325e4204dedSChris Mason u32 val; 326e4204dedSChris Mason u32 found = 0; 327607d432dSJosef Bacik memcpy(&found, result, csum_size); 328e4204dedSChris Mason 329607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 33094647322SDavid Sterba btrfs_warn_rl(fs_info, 33194647322SDavid Sterba "%s checksum verify failed on %llu wanted %X found %X " 33294647322SDavid Sterba "level %d", 33301d58472SDaniel Dressler fs_info->sb->s_id, buf->start, 334c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 335607d432dSJosef Bacik if (result != (char *)&inline_result) 336607d432dSJosef Bacik kfree(result); 33719c00ddcSChris Mason return 1; 33819c00ddcSChris Mason } 33919c00ddcSChris Mason } else { 340607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 34119c00ddcSChris Mason } 342607d432dSJosef Bacik if (result != (char *)&inline_result) 343607d432dSJosef Bacik kfree(result); 34419c00ddcSChris Mason return 0; 34519c00ddcSChris Mason } 34619c00ddcSChris Mason 347d352ac68SChris Mason /* 348d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 349d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 350d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 351d352ac68SChris Mason * in the wrong place. 352d352ac68SChris Mason */ 3531259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 354b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 355b9fab919SChris Mason int atomic) 3561259ab75SChris Mason { 3572ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3581259ab75SChris Mason int ret; 3592755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 3601259ab75SChris Mason 3611259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3621259ab75SChris Mason return 0; 3631259ab75SChris Mason 364b9fab919SChris Mason if (atomic) 365b9fab919SChris Mason return -EAGAIN; 366b9fab919SChris Mason 367a26e8c9fSJosef Bacik if (need_lock) { 368a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 369a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 370a26e8c9fSJosef Bacik } 371a26e8c9fSJosef Bacik 3722ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 373ff13db41SDavid Sterba &cached_state); 3740b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3751259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3761259ab75SChris Mason ret = 0; 3771259ab75SChris Mason goto out; 3781259ab75SChris Mason } 37994647322SDavid Sterba btrfs_err_rl(eb->fs_info, 38094647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 38194647322SDavid Sterba eb->start, 38229549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3831259ab75SChris Mason ret = 1; 384a26e8c9fSJosef Bacik 385a26e8c9fSJosef Bacik /* 386a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 387a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 388a26e8c9fSJosef Bacik * block that has been free'd and re-allocated. So don't clear uptodate 389a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 390a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 391a26e8c9fSJosef Bacik * making everybody very sad. 392a26e8c9fSJosef Bacik */ 393a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3940b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 39533958dc6SChris Mason out: 3962ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 3972ac55d41SJosef Bacik &cached_state, GFP_NOFS); 398472b909fSJosef Bacik if (need_lock) 399a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 4001259ab75SChris Mason return ret; 4011259ab75SChris Mason } 4021259ab75SChris Mason 403d352ac68SChris Mason /* 4041104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 4051104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 4061104a885SDavid Sterba */ 4071104a885SDavid Sterba static int btrfs_check_super_csum(char *raw_disk_sb) 4081104a885SDavid Sterba { 4091104a885SDavid Sterba struct btrfs_super_block *disk_sb = 4101104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 4111104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 4121104a885SDavid Sterba int ret = 0; 4131104a885SDavid Sterba 4141104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 4151104a885SDavid Sterba u32 crc = ~(u32)0; 4161104a885SDavid Sterba const int csum_size = sizeof(crc); 4171104a885SDavid Sterba char result[csum_size]; 4181104a885SDavid Sterba 4191104a885SDavid Sterba /* 4201104a885SDavid Sterba * The super_block structure does not span the whole 4211104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 4221104a885SDavid Sterba * is filled with zeros and is included in the checkum. 4231104a885SDavid Sterba */ 4241104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 4251104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 4261104a885SDavid Sterba btrfs_csum_final(crc, result); 4271104a885SDavid Sterba 4281104a885SDavid Sterba if (memcmp(raw_disk_sb, result, csum_size)) 4291104a885SDavid Sterba ret = 1; 4301104a885SDavid Sterba } 4311104a885SDavid Sterba 4321104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 433efe120a0SFrank Holton printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", 4341104a885SDavid Sterba csum_type); 4351104a885SDavid Sterba ret = 1; 4361104a885SDavid Sterba } 4371104a885SDavid Sterba 4381104a885SDavid Sterba return ret; 4391104a885SDavid Sterba } 4401104a885SDavid Sterba 4411104a885SDavid Sterba /* 442d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 443d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 444d352ac68SChris Mason */ 445f188591eSChris Mason static int btree_read_extent_buffer_pages(struct btrfs_root *root, 446f188591eSChris Mason struct extent_buffer *eb, 447ca7a79adSChris Mason u64 start, u64 parent_transid) 448f188591eSChris Mason { 449f188591eSChris Mason struct extent_io_tree *io_tree; 450ea466794SJosef Bacik int failed = 0; 451f188591eSChris Mason int ret; 452f188591eSChris Mason int num_copies = 0; 453f188591eSChris Mason int mirror_num = 0; 454ea466794SJosef Bacik int failed_mirror = 0; 455f188591eSChris Mason 456a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 457f188591eSChris Mason io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 458f188591eSChris Mason while (1) { 459bb82ab88SArne Jansen ret = read_extent_buffer_pages(io_tree, eb, start, 460bb82ab88SArne Jansen WAIT_COMPLETE, 461f188591eSChris Mason btree_get_extent, mirror_num); 462256dd1bbSStefan Behrens if (!ret) { 463256dd1bbSStefan Behrens if (!verify_parent_transid(io_tree, eb, 464b9fab919SChris Mason parent_transid, 0)) 465ea466794SJosef Bacik break; 466256dd1bbSStefan Behrens else 467256dd1bbSStefan Behrens ret = -EIO; 468256dd1bbSStefan Behrens } 469d397712bSChris Mason 470a826d6dcSJosef Bacik /* 471a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 472a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 473a826d6dcSJosef Bacik * any less wrong. 474a826d6dcSJosef Bacik */ 475a826d6dcSJosef Bacik if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 476ea466794SJosef Bacik break; 477ea466794SJosef Bacik 4785d964051SStefan Behrens num_copies = btrfs_num_copies(root->fs_info, 479f188591eSChris Mason eb->start, eb->len); 4804235298eSChris Mason if (num_copies == 1) 481ea466794SJosef Bacik break; 4824235298eSChris Mason 4835cf1ab56SJosef Bacik if (!failed_mirror) { 4845cf1ab56SJosef Bacik failed = 1; 4855cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4865cf1ab56SJosef Bacik } 4875cf1ab56SJosef Bacik 488f188591eSChris Mason mirror_num++; 489ea466794SJosef Bacik if (mirror_num == failed_mirror) 490ea466794SJosef Bacik mirror_num++; 491ea466794SJosef Bacik 4924235298eSChris Mason if (mirror_num > num_copies) 493ea466794SJosef Bacik break; 494f188591eSChris Mason } 495ea466794SJosef Bacik 496c0901581SStefan Behrens if (failed && !ret && failed_mirror) 497ea466794SJosef Bacik repair_eb_io_failure(root, eb, failed_mirror); 498ea466794SJosef Bacik 499ea466794SJosef Bacik return ret; 500f188591eSChris Mason } 50119c00ddcSChris Mason 502d352ac68SChris Mason /* 503d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 504d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 505d352ac68SChris Mason */ 506d397712bSChris Mason 50701d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 50819c00ddcSChris Mason { 5094eee4fa4SMiao Xie u64 start = page_offset(page); 51019c00ddcSChris Mason u64 found_start; 51119c00ddcSChris Mason struct extent_buffer *eb; 512f188591eSChris Mason 5134f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5144f2de97aSJosef Bacik if (page != eb->pages[0]) 5154f2de97aSJosef Bacik return 0; 51619c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 517fae7f21cSDulshani Gunawardhana if (WARN_ON(found_start != start || !PageUptodate(page))) 5184f2de97aSJosef Bacik return 0; 51901d58472SDaniel Dressler csum_tree_block(fs_info, eb, 0); 52019c00ddcSChris Mason return 0; 52119c00ddcSChris Mason } 52219c00ddcSChris Mason 52301d58472SDaniel Dressler static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, 5242b82032cSYan Zheng struct extent_buffer *eb) 5252b82032cSYan Zheng { 52601d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 5272b82032cSYan Zheng u8 fsid[BTRFS_UUID_SIZE]; 5282b82032cSYan Zheng int ret = 1; 5292b82032cSYan Zheng 5300a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5312b82032cSYan Zheng while (fs_devices) { 5322b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5332b82032cSYan Zheng ret = 0; 5342b82032cSYan Zheng break; 5352b82032cSYan Zheng } 5362b82032cSYan Zheng fs_devices = fs_devices->seed; 5372b82032cSYan Zheng } 5382b82032cSYan Zheng return ret; 5392b82032cSYan Zheng } 5402b82032cSYan Zheng 541a826d6dcSJosef Bacik #define CORRUPT(reason, eb, root, slot) \ 542efe120a0SFrank Holton btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ 543efe120a0SFrank Holton "root=%llu, slot=%d", reason, \ 544c1c9ff7cSGeert Uytterhoeven btrfs_header_bytenr(eb), root->objectid, slot) 545a826d6dcSJosef Bacik 546a826d6dcSJosef Bacik static noinline int check_leaf(struct btrfs_root *root, 547a826d6dcSJosef Bacik struct extent_buffer *leaf) 548a826d6dcSJosef Bacik { 549a826d6dcSJosef Bacik struct btrfs_key key; 550a826d6dcSJosef Bacik struct btrfs_key leaf_key; 551a826d6dcSJosef Bacik u32 nritems = btrfs_header_nritems(leaf); 552a826d6dcSJosef Bacik int slot; 553a826d6dcSJosef Bacik 554a826d6dcSJosef Bacik if (nritems == 0) 555a826d6dcSJosef Bacik return 0; 556a826d6dcSJosef Bacik 557a826d6dcSJosef Bacik /* Check the 0 item */ 558a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 559a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 560a826d6dcSJosef Bacik CORRUPT("invalid item offset size pair", leaf, root, 0); 561a826d6dcSJosef Bacik return -EIO; 562a826d6dcSJosef Bacik } 563a826d6dcSJosef Bacik 564a826d6dcSJosef Bacik /* 565a826d6dcSJosef Bacik * Check to make sure each items keys are in the correct order and their 566a826d6dcSJosef Bacik * offsets make sense. We only have to loop through nritems-1 because 567a826d6dcSJosef Bacik * we check the current slot against the next slot, which verifies the 568a826d6dcSJosef Bacik * next slot's offset+size makes sense and that the current's slot 569a826d6dcSJosef Bacik * offset is correct. 570a826d6dcSJosef Bacik */ 571a826d6dcSJosef Bacik for (slot = 0; slot < nritems - 1; slot++) { 572a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &leaf_key, slot); 573a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &key, slot + 1); 574a826d6dcSJosef Bacik 575a826d6dcSJosef Bacik /* Make sure the keys are in the right order */ 576a826d6dcSJosef Bacik if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { 577a826d6dcSJosef Bacik CORRUPT("bad key order", leaf, root, slot); 578a826d6dcSJosef Bacik return -EIO; 579a826d6dcSJosef Bacik } 580a826d6dcSJosef Bacik 581a826d6dcSJosef Bacik /* 582a826d6dcSJosef Bacik * Make sure the offset and ends are right, remember that the 583a826d6dcSJosef Bacik * item data starts at the end of the leaf and grows towards the 584a826d6dcSJosef Bacik * front. 585a826d6dcSJosef Bacik */ 586a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, slot) != 587a826d6dcSJosef Bacik btrfs_item_end_nr(leaf, slot + 1)) { 588a826d6dcSJosef Bacik CORRUPT("slot offset bad", leaf, root, slot); 589a826d6dcSJosef Bacik return -EIO; 590a826d6dcSJosef Bacik } 591a826d6dcSJosef Bacik 592a826d6dcSJosef Bacik /* 593a826d6dcSJosef Bacik * Check to make sure that we don't point outside of the leaf, 594a826d6dcSJosef Bacik * just incase all the items are consistent to eachother, but 595a826d6dcSJosef Bacik * all point outside of the leaf. 596a826d6dcSJosef Bacik */ 597a826d6dcSJosef Bacik if (btrfs_item_end_nr(leaf, slot) > 598a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 599a826d6dcSJosef Bacik CORRUPT("slot end outside of leaf", leaf, root, slot); 600a826d6dcSJosef Bacik return -EIO; 601a826d6dcSJosef Bacik } 602a826d6dcSJosef Bacik } 603a826d6dcSJosef Bacik 604a826d6dcSJosef Bacik return 0; 605a826d6dcSJosef Bacik } 606a826d6dcSJosef Bacik 607facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 608facc8a22SMiao Xie u64 phy_offset, struct page *page, 609facc8a22SMiao Xie u64 start, u64 end, int mirror) 610ce9adaa5SChris Mason { 611ce9adaa5SChris Mason u64 found_start; 612ce9adaa5SChris Mason int found_level; 613ce9adaa5SChris Mason struct extent_buffer *eb; 614ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 615f188591eSChris Mason int ret = 0; 616727011e0SChris Mason int reads_done; 617ce9adaa5SChris Mason 618ce9adaa5SChris Mason if (!page->private) 619ce9adaa5SChris Mason goto out; 620d397712bSChris Mason 6214f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 622d397712bSChris Mason 6230b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 6240b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 6250b32f4bbSJosef Bacik */ 6260b32f4bbSJosef Bacik extent_buffer_get(eb); 6270b32f4bbSJosef Bacik 6280b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 629727011e0SChris Mason if (!reads_done) 630727011e0SChris Mason goto err; 631f188591eSChris Mason 6325cf1ab56SJosef Bacik eb->read_mirror = mirror; 633656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 634ea466794SJosef Bacik ret = -EIO; 635ea466794SJosef Bacik goto err; 636ea466794SJosef Bacik } 637ea466794SJosef Bacik 638ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 639727011e0SChris Mason if (found_start != eb->start) { 64094647322SDavid Sterba btrfs_err_rl(eb->fs_info, "bad tree block start %llu %llu", 64194647322SDavid Sterba found_start, eb->start); 642f188591eSChris Mason ret = -EIO; 643ce9adaa5SChris Mason goto err; 644ce9adaa5SChris Mason } 64501d58472SDaniel Dressler if (check_tree_block_fsid(root->fs_info, eb)) { 64694647322SDavid Sterba btrfs_err_rl(eb->fs_info, "bad fsid on block %llu", 64794647322SDavid Sterba eb->start); 6481259ab75SChris Mason ret = -EIO; 6491259ab75SChris Mason goto err; 6501259ab75SChris Mason } 651ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6521c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 65368b663d1SDavid Sterba btrfs_err(root->fs_info, "bad tree block level %d", 6541c24c3ceSJosef Bacik (int)btrfs_header_level(eb)); 6551c24c3ceSJosef Bacik ret = -EIO; 6561c24c3ceSJosef Bacik goto err; 6571c24c3ceSJosef Bacik } 658ce9adaa5SChris Mason 65985d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 66085d4e461SChris Mason eb, found_level); 6614008c04aSChris Mason 66201d58472SDaniel Dressler ret = csum_tree_block(root->fs_info, eb, 1); 663a826d6dcSJosef Bacik if (ret) { 664f188591eSChris Mason ret = -EIO; 665a826d6dcSJosef Bacik goto err; 666a826d6dcSJosef Bacik } 667a826d6dcSJosef Bacik 668a826d6dcSJosef Bacik /* 669a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 670a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 671a826d6dcSJosef Bacik * return -EIO. 672a826d6dcSJosef Bacik */ 673a826d6dcSJosef Bacik if (found_level == 0 && check_leaf(root, eb)) { 674a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 675a826d6dcSJosef Bacik ret = -EIO; 676a826d6dcSJosef Bacik } 677ce9adaa5SChris Mason 6780b32f4bbSJosef Bacik if (!ret) 6790b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 680ce9adaa5SChris Mason err: 68179fb65a1SJosef Bacik if (reads_done && 68279fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 6834bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, ret); 6844bb31e92SArne Jansen 68553b381b3SDavid Woodhouse if (ret) { 68653b381b3SDavid Woodhouse /* 68753b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 68853b381b3SDavid Woodhouse * again, we have to make sure it has something 68953b381b3SDavid Woodhouse * to decrement 69053b381b3SDavid Woodhouse */ 69153b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6920b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 69353b381b3SDavid Woodhouse } 6940b32f4bbSJosef Bacik free_extent_buffer(eb); 695ce9adaa5SChris Mason out: 696f188591eSChris Mason return ret; 697ce9adaa5SChris Mason } 698ce9adaa5SChris Mason 699ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 7004bb31e92SArne Jansen { 7014bb31e92SArne Jansen struct extent_buffer *eb; 7024bb31e92SArne Jansen struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 7034bb31e92SArne Jansen 7044f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 705656f30dbSFilipe Manana set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); 7065cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 70753b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 708ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 7094bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, -EIO); 7104bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 7114bb31e92SArne Jansen } 7124bb31e92SArne Jansen 7134246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 714ce9adaa5SChris Mason { 71597eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 716ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 7179e0af237SLiu Bo struct btrfs_workqueue *wq; 7189e0af237SLiu Bo btrfs_work_func_t func; 719ce9adaa5SChris Mason 720ce9adaa5SChris Mason fs_info = end_io_wq->info; 7214246a0b6SChristoph Hellwig end_io_wq->error = bio->bi_error; 722d20f7043SChris Mason 7237b6d91daSChristoph Hellwig if (bio->bi_rw & REQ_WRITE) { 7249e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 7259e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 7269e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 7279e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 7289e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 7299e0af237SLiu Bo func = btrfs_freespace_write_helper; 7309e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7319e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7329e0af237SLiu Bo func = btrfs_endio_raid56_helper; 733d20f7043SChris Mason } else { 7349e0af237SLiu Bo wq = fs_info->endio_write_workers; 7359e0af237SLiu Bo func = btrfs_endio_write_helper; 736d20f7043SChris Mason } 7379e0af237SLiu Bo } else { 7388b110e39SMiao Xie if (unlikely(end_io_wq->metadata == 7398b110e39SMiao Xie BTRFS_WQ_ENDIO_DIO_REPAIR)) { 7408b110e39SMiao Xie wq = fs_info->endio_repair_workers; 7418b110e39SMiao Xie func = btrfs_endio_repair_helper; 7428b110e39SMiao Xie } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7439e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7449e0af237SLiu Bo func = btrfs_endio_raid56_helper; 7459e0af237SLiu Bo } else if (end_io_wq->metadata) { 7469e0af237SLiu Bo wq = fs_info->endio_meta_workers; 7479e0af237SLiu Bo func = btrfs_endio_meta_helper; 7489e0af237SLiu Bo } else { 7499e0af237SLiu Bo wq = fs_info->endio_workers; 7509e0af237SLiu Bo func = btrfs_endio_helper; 7519e0af237SLiu Bo } 7529e0af237SLiu Bo } 7539e0af237SLiu Bo 7549e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 7559e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 756ce9adaa5SChris Mason } 757ce9adaa5SChris Mason 75822c59948SChris Mason int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 759bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7600b86a832SChris Mason { 76197eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7628b110e39SMiao Xie 76397eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 764ce9adaa5SChris Mason if (!end_io_wq) 765ce9adaa5SChris Mason return -ENOMEM; 766ce9adaa5SChris Mason 767ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 768ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 76922c59948SChris Mason end_io_wq->info = info; 770ce9adaa5SChris Mason end_io_wq->error = 0; 771ce9adaa5SChris Mason end_io_wq->bio = bio; 77222c59948SChris Mason end_io_wq->metadata = metadata; 773ce9adaa5SChris Mason 774ce9adaa5SChris Mason bio->bi_private = end_io_wq; 775ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 77622c59948SChris Mason return 0; 77722c59948SChris Mason } 77822c59948SChris Mason 779b64a2851SChris Mason unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 7804854ddd0SChris Mason { 7814854ddd0SChris Mason unsigned long limit = min_t(unsigned long, 7825cdc7ad3SQu Wenruo info->thread_pool_size, 7834854ddd0SChris Mason info->fs_devices->open_devices); 7844854ddd0SChris Mason return 256 * limit; 7854854ddd0SChris Mason } 7864854ddd0SChris Mason 7874a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7884a69a410SChris Mason { 7894a69a410SChris Mason struct async_submit_bio *async; 79079787eaaSJeff Mahoney int ret; 7914a69a410SChris Mason 7924a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 79379787eaaSJeff Mahoney ret = async->submit_bio_start(async->inode, async->rw, async->bio, 794eaf25d93SChris Mason async->mirror_num, async->bio_flags, 795eaf25d93SChris Mason async->bio_offset); 79679787eaaSJeff Mahoney if (ret) 79779787eaaSJeff Mahoney async->error = ret; 7984a69a410SChris Mason } 7994a69a410SChris Mason 8004a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 8018b712842SChris Mason { 8028b712842SChris Mason struct btrfs_fs_info *fs_info; 8038b712842SChris Mason struct async_submit_bio *async; 8044854ddd0SChris Mason int limit; 8058b712842SChris Mason 8068b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 8078b712842SChris Mason fs_info = BTRFS_I(async->inode)->root->fs_info; 8084854ddd0SChris Mason 809b64a2851SChris Mason limit = btrfs_async_submit_limit(fs_info); 8104854ddd0SChris Mason limit = limit * 2 / 3; 8114854ddd0SChris Mason 812ee863954SDavid Sterba /* 813ee863954SDavid Sterba * atomic_dec_return implies a barrier for waitqueue_active 814ee863954SDavid Sterba */ 81566657b31SJosef Bacik if (atomic_dec_return(&fs_info->nr_async_submits) < limit && 816b64a2851SChris Mason waitqueue_active(&fs_info->async_submit_wait)) 8174854ddd0SChris Mason wake_up(&fs_info->async_submit_wait); 8184854ddd0SChris Mason 81979787eaaSJeff Mahoney /* If an error occured we just want to clean up the bio and move on */ 82079787eaaSJeff Mahoney if (async->error) { 8214246a0b6SChristoph Hellwig async->bio->bi_error = async->error; 8224246a0b6SChristoph Hellwig bio_endio(async->bio); 82379787eaaSJeff Mahoney return; 82479787eaaSJeff Mahoney } 82579787eaaSJeff Mahoney 8264a69a410SChris Mason async->submit_bio_done(async->inode, async->rw, async->bio, 827eaf25d93SChris Mason async->mirror_num, async->bio_flags, 828eaf25d93SChris Mason async->bio_offset); 8294a69a410SChris Mason } 8304a69a410SChris Mason 8314a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8324a69a410SChris Mason { 8334a69a410SChris Mason struct async_submit_bio *async; 8344a69a410SChris Mason 8354a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8368b712842SChris Mason kfree(async); 8378b712842SChris Mason } 8388b712842SChris Mason 83944b8bd7eSChris Mason int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, 84044b8bd7eSChris Mason int rw, struct bio *bio, int mirror_num, 841c8b97818SChris Mason unsigned long bio_flags, 842eaf25d93SChris Mason u64 bio_offset, 8434a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start, 8444a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done) 84544b8bd7eSChris Mason { 84644b8bd7eSChris Mason struct async_submit_bio *async; 84744b8bd7eSChris Mason 84844b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 84944b8bd7eSChris Mason if (!async) 85044b8bd7eSChris Mason return -ENOMEM; 85144b8bd7eSChris Mason 85244b8bd7eSChris Mason async->inode = inode; 85344b8bd7eSChris Mason async->rw = rw; 85444b8bd7eSChris Mason async->bio = bio; 85544b8bd7eSChris Mason async->mirror_num = mirror_num; 8564a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8574a69a410SChris Mason async->submit_bio_done = submit_bio_done; 8584a69a410SChris Mason 8599e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 8605cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 8614a69a410SChris Mason 862c8b97818SChris Mason async->bio_flags = bio_flags; 863eaf25d93SChris Mason async->bio_offset = bio_offset; 8648c8bee1dSChris Mason 86579787eaaSJeff Mahoney async->error = 0; 86679787eaaSJeff Mahoney 867cb03c743SChris Mason atomic_inc(&fs_info->nr_async_submits); 868d313d7a3SChris Mason 8697b6d91daSChristoph Hellwig if (rw & REQ_SYNC) 8705cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 871d313d7a3SChris Mason 8725cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 8739473f16cSChris Mason 874771ed689SChris Mason while (atomic_read(&fs_info->async_submit_draining) && 875771ed689SChris Mason atomic_read(&fs_info->nr_async_submits)) { 876771ed689SChris Mason wait_event(fs_info->async_submit_wait, 877771ed689SChris Mason (atomic_read(&fs_info->nr_async_submits) == 0)); 878771ed689SChris Mason } 879771ed689SChris Mason 88044b8bd7eSChris Mason return 0; 88144b8bd7eSChris Mason } 88244b8bd7eSChris Mason 883ce3ed71aSChris Mason static int btree_csum_one_bio(struct bio *bio) 884ce3ed71aSChris Mason { 8852c30c71bSKent Overstreet struct bio_vec *bvec; 886ce3ed71aSChris Mason struct btrfs_root *root; 8872c30c71bSKent Overstreet int i, ret = 0; 888ce3ed71aSChris Mason 8892c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 890ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 89101d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 89279787eaaSJeff Mahoney if (ret) 89379787eaaSJeff Mahoney break; 894ce3ed71aSChris Mason } 8952c30c71bSKent Overstreet 89679787eaaSJeff Mahoney return ret; 897ce3ed71aSChris Mason } 898ce3ed71aSChris Mason 8994a69a410SChris Mason static int __btree_submit_bio_start(struct inode *inode, int rw, 9004a69a410SChris Mason struct bio *bio, int mirror_num, 901eaf25d93SChris Mason unsigned long bio_flags, 902eaf25d93SChris Mason u64 bio_offset) 90322c59948SChris Mason { 9048b712842SChris Mason /* 9058b712842SChris Mason * when we're called for a write, we're already in the async 9065443be45SChris Mason * submission context. Just jump into btrfs_map_bio 9078b712842SChris Mason */ 90879787eaaSJeff Mahoney return btree_csum_one_bio(bio); 90922c59948SChris Mason } 91022c59948SChris Mason 9114a69a410SChris Mason static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 912eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 913eaf25d93SChris Mason u64 bio_offset) 9144a69a410SChris Mason { 91561891923SStefan Behrens int ret; 91661891923SStefan Behrens 9178b712842SChris Mason /* 9184a69a410SChris Mason * when we're called for a write, we're already in the async 9194a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 9208b712842SChris Mason */ 92161891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 9224246a0b6SChristoph Hellwig if (ret) { 9234246a0b6SChristoph Hellwig bio->bi_error = ret; 9244246a0b6SChristoph Hellwig bio_endio(bio); 9254246a0b6SChristoph Hellwig } 92661891923SStefan Behrens return ret; 9270b86a832SChris Mason } 9280b86a832SChris Mason 929de0022b9SJosef Bacik static int check_async_write(struct inode *inode, unsigned long bio_flags) 930de0022b9SJosef Bacik { 931de0022b9SJosef Bacik if (bio_flags & EXTENT_BIO_TREE_LOG) 932de0022b9SJosef Bacik return 0; 933de0022b9SJosef Bacik #ifdef CONFIG_X86 934362f924bSBorislav Petkov if (static_cpu_has_safe(X86_FEATURE_XMM4_2)) 935de0022b9SJosef Bacik return 0; 936de0022b9SJosef Bacik #endif 937de0022b9SJosef Bacik return 1; 938de0022b9SJosef Bacik } 939de0022b9SJosef Bacik 94044b8bd7eSChris Mason static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 941eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 942eaf25d93SChris Mason u64 bio_offset) 94344b8bd7eSChris Mason { 944de0022b9SJosef Bacik int async = check_async_write(inode, bio_flags); 9454a69a410SChris Mason int ret; 946cad321adSChris Mason 9477b6d91daSChristoph Hellwig if (!(rw & REQ_WRITE)) { 948cad321adSChris Mason /* 949cad321adSChris Mason * called for a read, do the setup so that checksum validation 950cad321adSChris Mason * can happen in the async kernel threads 951cad321adSChris Mason */ 952f3f266abSChris Mason ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 953bfebd8b5SDavid Sterba bio, BTRFS_WQ_ENDIO_METADATA); 9541d4284bdSChris Mason if (ret) 95561891923SStefan Behrens goto out_w_error; 95661891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 9576f3577bdSChris Mason mirror_num, 0); 958de0022b9SJosef Bacik } else if (!async) { 959de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 960de0022b9SJosef Bacik if (ret) 96161891923SStefan Behrens goto out_w_error; 96261891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 963de0022b9SJosef Bacik mirror_num, 0); 96461891923SStefan Behrens } else { 965cad321adSChris Mason /* 96661891923SStefan Behrens * kthread helpers are used to submit writes so that 96761891923SStefan Behrens * checksumming can happen in parallel across all CPUs 968cad321adSChris Mason */ 96961891923SStefan Behrens ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 970c8b97818SChris Mason inode, rw, bio, mirror_num, 0, 971eaf25d93SChris Mason bio_offset, 9724a69a410SChris Mason __btree_submit_bio_start, 9734a69a410SChris Mason __btree_submit_bio_done); 97444b8bd7eSChris Mason } 97544b8bd7eSChris Mason 9764246a0b6SChristoph Hellwig if (ret) 9774246a0b6SChristoph Hellwig goto out_w_error; 9784246a0b6SChristoph Hellwig return 0; 9794246a0b6SChristoph Hellwig 98061891923SStefan Behrens out_w_error: 9814246a0b6SChristoph Hellwig bio->bi_error = ret; 9824246a0b6SChristoph Hellwig bio_endio(bio); 98361891923SStefan Behrens return ret; 98461891923SStefan Behrens } 98561891923SStefan Behrens 9863dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 987784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 988a6bc32b8SMel Gorman struct page *newpage, struct page *page, 989a6bc32b8SMel Gorman enum migrate_mode mode) 990784b4e29SChris Mason { 991784b4e29SChris Mason /* 992784b4e29SChris Mason * we can't safely write a btree page from here, 993784b4e29SChris Mason * we haven't done the locking hook 994784b4e29SChris Mason */ 995784b4e29SChris Mason if (PageDirty(page)) 996784b4e29SChris Mason return -EAGAIN; 997784b4e29SChris Mason /* 998784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 999784b4e29SChris Mason * We must have no buffers or drop them. 1000784b4e29SChris Mason */ 1001784b4e29SChris Mason if (page_has_private(page) && 1002784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 1003784b4e29SChris Mason return -EAGAIN; 1004a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 1005784b4e29SChris Mason } 10063dd1462eSJan Beulich #endif 1007784b4e29SChris Mason 10080da5468fSChris Mason 10090da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 10100da5468fSChris Mason struct writeback_control *wbc) 10110da5468fSChris Mason { 1012e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 1013e2d84521SMiao Xie int ret; 1014e2d84521SMiao Xie 1015d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 1016448d640bSChris Mason 1017448d640bSChris Mason if (wbc->for_kupdate) 1018448d640bSChris Mason return 0; 1019448d640bSChris Mason 1020e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 1021b9473439SChris Mason /* this is a bit racy, but that's ok */ 1022e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 1023e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 1024e2d84521SMiao Xie if (ret < 0) 1025793955bcSChris Mason return 0; 1026793955bcSChris Mason } 10270b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 10280da5468fSChris Mason } 10290da5468fSChris Mason 1030b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 10315f39d397SChris Mason { 1032d1310b2eSChris Mason struct extent_io_tree *tree; 1033d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10348ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 10355f39d397SChris Mason } 10365f39d397SChris Mason 103770dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10385f39d397SChris Mason { 103998509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 104098509cfcSChris Mason return 0; 10410c4e538bSDavid Sterba 1042f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1043d98237b3SChris Mason } 1044d98237b3SChris Mason 1045d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1046d47992f8SLukas Czerner unsigned int length) 1047d98237b3SChris Mason { 1048d1310b2eSChris Mason struct extent_io_tree *tree; 1049d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10505f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10515f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10529ad6b7bcSChris Mason if (PagePrivate(page)) { 1053efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1054efe120a0SFrank Holton "page private not zero on page %llu", 1055efe120a0SFrank Holton (unsigned long long)page_offset(page)); 10569ad6b7bcSChris Mason ClearPagePrivate(page); 10579ad6b7bcSChris Mason set_page_private(page, 0); 10589ad6b7bcSChris Mason page_cache_release(page); 10599ad6b7bcSChris Mason } 1060d98237b3SChris Mason } 1061d98237b3SChris Mason 10620b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10630b32f4bbSJosef Bacik { 1064bb146eb2SJosef Bacik #ifdef DEBUG 10650b32f4bbSJosef Bacik struct extent_buffer *eb; 10660b32f4bbSJosef Bacik 10670b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10680b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10690b32f4bbSJosef Bacik BUG_ON(!eb); 10700b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10710b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10720b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 1073bb146eb2SJosef Bacik #endif 10740b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10750b32f4bbSJosef Bacik } 10760b32f4bbSJosef Bacik 10777f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1078d98237b3SChris Mason .readpage = btree_readpage, 10790da5468fSChris Mason .writepages = btree_writepages, 10805f39d397SChris Mason .releasepage = btree_releasepage, 10815f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10825a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1083784b4e29SChris Mason .migratepage = btree_migratepage, 10845a92bc88SChris Mason #endif 10850b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1086d98237b3SChris Mason }; 1087123abc88SChris Mason 1088d3e46feaSDavid Sterba void readahead_tree_block(struct btrfs_root *root, u64 bytenr) 1089090d1875SChris Mason { 10905f39d397SChris Mason struct extent_buffer *buf = NULL; 10915f39d397SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 1092090d1875SChris Mason 1093a83fffb7SDavid Sterba buf = btrfs_find_create_tree_block(root, bytenr); 10945f39d397SChris Mason if (!buf) 10956197d86eSDavid Sterba return; 1096d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1097bb82ab88SArne Jansen buf, 0, WAIT_NONE, btree_get_extent, 0); 10985f39d397SChris Mason free_extent_buffer(buf); 1099090d1875SChris Mason } 1100090d1875SChris Mason 1101c0dcaa4dSDavid Sterba int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, 1102ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1103ab0fff03SArne Jansen { 1104ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 1105ab0fff03SArne Jansen struct inode *btree_inode = root->fs_info->btree_inode; 1106ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1107ab0fff03SArne Jansen int ret; 1108ab0fff03SArne Jansen 1109a83fffb7SDavid Sterba buf = btrfs_find_create_tree_block(root, bytenr); 1110ab0fff03SArne Jansen if (!buf) 1111ab0fff03SArne Jansen return 0; 1112ab0fff03SArne Jansen 1113ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1114ab0fff03SArne Jansen 1115ab0fff03SArne Jansen ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1116ab0fff03SArne Jansen btree_get_extent, mirror_num); 1117ab0fff03SArne Jansen if (ret) { 1118ab0fff03SArne Jansen free_extent_buffer(buf); 1119ab0fff03SArne Jansen return ret; 1120ab0fff03SArne Jansen } 1121ab0fff03SArne Jansen 1122ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1123ab0fff03SArne Jansen free_extent_buffer(buf); 1124ab0fff03SArne Jansen return -EIO; 11250b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1126ab0fff03SArne Jansen *eb = buf; 1127ab0fff03SArne Jansen } else { 1128ab0fff03SArne Jansen free_extent_buffer(buf); 1129ab0fff03SArne Jansen } 1130ab0fff03SArne Jansen return 0; 1131ab0fff03SArne Jansen } 1132ab0fff03SArne Jansen 113301d58472SDaniel Dressler struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, 11340308af44SDavid Sterba u64 bytenr) 11350999df54SChris Mason { 113601d58472SDaniel Dressler return find_extent_buffer(fs_info, bytenr); 11370999df54SChris Mason } 11380999df54SChris Mason 11390999df54SChris Mason struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 1140a83fffb7SDavid Sterba u64 bytenr) 11410999df54SChris Mason { 1142fccb84c9SDavid Sterba if (btrfs_test_is_dummy_root(root)) 1143ce3e6984SDavid Sterba return alloc_test_extent_buffer(root->fs_info, bytenr); 1144ce3e6984SDavid Sterba return alloc_extent_buffer(root->fs_info, bytenr); 11450999df54SChris Mason } 11460999df54SChris Mason 11470999df54SChris Mason 1148e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1149e02119d5SChris Mason { 1150727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 11518aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1152e02119d5SChris Mason } 1153e02119d5SChris Mason 1154e02119d5SChris Mason int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1155e02119d5SChris Mason { 1156727011e0SChris Mason return filemap_fdatawait_range(buf->pages[0]->mapping, 11578aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1158e02119d5SChris Mason } 1159e02119d5SChris Mason 1160db94535dSChris Mason struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1161ce86cd59SDavid Sterba u64 parent_transid) 1162e20d96d6SChris Mason { 11635f39d397SChris Mason struct extent_buffer *buf = NULL; 116419c00ddcSChris Mason int ret; 116519c00ddcSChris Mason 1166a83fffb7SDavid Sterba buf = btrfs_find_create_tree_block(root, bytenr); 11675f39d397SChris Mason if (!buf) 116864c043deSLiu Bo return ERR_PTR(-ENOMEM); 1169e4204dedSChris Mason 1170ca7a79adSChris Mason ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 11710f0fe8f7SFilipe David Borba Manana if (ret) { 11720f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 117364c043deSLiu Bo return ERR_PTR(ret); 11740f0fe8f7SFilipe David Borba Manana } 11755f39d397SChris Mason return buf; 1176ce9adaa5SChris Mason 1177eb60ceacSChris Mason } 1178eb60ceacSChris Mason 117901d58472SDaniel Dressler void clean_tree_block(struct btrfs_trans_handle *trans, 118001d58472SDaniel Dressler struct btrfs_fs_info *fs_info, 11815f39d397SChris Mason struct extent_buffer *buf) 1182ed2ff2cbSChris Mason { 118355c69072SChris Mason if (btrfs_header_generation(buf) == 1184e2d84521SMiao Xie fs_info->running_transaction->transid) { 1185b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1186b4ce94deSChris Mason 1187b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1188e2d84521SMiao Xie __percpu_counter_add(&fs_info->dirty_metadata_bytes, 1189e2d84521SMiao Xie -buf->len, 1190e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1191b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1192b9473439SChris Mason btrfs_set_lock_blocking(buf); 11930b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1194925baeddSChris Mason } 11955f39d397SChris Mason } 1196ed7b63ebSJosef Bacik } 11975f39d397SChris Mason 11988257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11998257b2dcSMiao Xie { 12008257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 12018257b2dcSMiao Xie int ret; 12028257b2dcSMiao Xie 12038257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 12048257b2dcSMiao Xie if (!writers) 12058257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 12068257b2dcSMiao Xie 1207908c7f19STejun Heo ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); 12088257b2dcSMiao Xie if (ret < 0) { 12098257b2dcSMiao Xie kfree(writers); 12108257b2dcSMiao Xie return ERR_PTR(ret); 12118257b2dcSMiao Xie } 12128257b2dcSMiao Xie 12138257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 12148257b2dcSMiao Xie return writers; 12158257b2dcSMiao Xie } 12168257b2dcSMiao Xie 12178257b2dcSMiao Xie static void 12188257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 12198257b2dcSMiao Xie { 12208257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 12218257b2dcSMiao Xie kfree(writers); 12228257b2dcSMiao Xie } 12238257b2dcSMiao Xie 1224707e8a07SDavid Sterba static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, 1225707e8a07SDavid Sterba struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1226e20d96d6SChris Mason u64 objectid) 1227d97e63b6SChris Mason { 1228cfaa7295SChris Mason root->node = NULL; 1229a28ec197SChris Mason root->commit_root = NULL; 1230db94535dSChris Mason root->sectorsize = sectorsize; 1231db94535dSChris Mason root->nodesize = nodesize; 123287ee04ebSChris Mason root->stripesize = stripesize; 123327cdeb70SMiao Xie root->state = 0; 1234d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 12350b86a832SChris Mason 12360f7d52f4SChris Mason root->objectid = objectid; 12370f7d52f4SChris Mason root->last_trans = 0; 123813a8a7c8SYan, Zheng root->highest_objectid = 0; 1239eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1240199c2a9cSMiao Xie root->nr_ordered_extents = 0; 124158176a96SJosef Bacik root->name = NULL; 12426bef4d31SEric Paris root->inode_tree = RB_ROOT; 124316cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1244f0486c68SYan, Zheng root->block_rsv = NULL; 1245d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 12460b86a832SChris Mason 12470b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 12485d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1249eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1250eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1251199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1252199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 12532ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 12542ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1255d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 12565d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1257eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1258199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1259f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 12602ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 12612ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 1262a2135011SChris Mason mutex_init(&root->objectid_mutex); 1263e02119d5SChris Mason mutex_init(&root->log_mutex); 126431f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1265573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 12667237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 12677237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 12687237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 12698b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 12708b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 12717237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 12727237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 12737237f183SYan Zheng atomic_set(&root->log_writers, 0); 12742ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 12758a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 1276b0feb9d9SMiao Xie atomic_set(&root->refs, 1); 12778257b2dcSMiao Xie atomic_set(&root->will_be_snapshoted, 0); 127855eeaf05SQu Wenruo atomic_set(&root->qgroup_meta_rsv, 0); 12797237f183SYan Zheng root->log_transid = 0; 1280d1433debSMiao Xie root->log_transid_committed = -1; 1281257c62e1SChris Mason root->last_log_commit = 0; 128206ea65a3SJosef Bacik if (fs_info) 1283d0c803c4SChris Mason extent_io_tree_init(&root->dirty_log_pages, 1284f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 1285017e5369SChris Mason 12863768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12873768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12886702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 128906ea65a3SJosef Bacik if (fs_info) 12903f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 129106ea65a3SJosef Bacik else 129206ea65a3SJosef Bacik root->defrag_trans_start = 0; 12934d775673SChris Mason root->root_key.objectid = objectid; 12940ee5dc67SAl Viro root->anon_dev = 0; 12958ea05e3aSAlexander Block 12965f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 12973768f368SChris Mason } 12983768f368SChris Mason 1299f84a8bd6SAl Viro static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) 13006f07e42eSAl Viro { 13016f07e42eSAl Viro struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); 13026f07e42eSAl Viro if (root) 13036f07e42eSAl Viro root->fs_info = fs_info; 13046f07e42eSAl Viro return root; 13056f07e42eSAl Viro } 13066f07e42eSAl Viro 130706ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 130806ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 130906ea65a3SJosef Bacik struct btrfs_root *btrfs_alloc_dummy_root(void) 131006ea65a3SJosef Bacik { 131106ea65a3SJosef Bacik struct btrfs_root *root; 131206ea65a3SJosef Bacik 131306ea65a3SJosef Bacik root = btrfs_alloc_root(NULL); 131406ea65a3SJosef Bacik if (!root) 131506ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1316707e8a07SDavid Sterba __setup_root(4096, 4096, 4096, root, NULL, 1); 131727cdeb70SMiao Xie set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state); 1318faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 131906ea65a3SJosef Bacik 132006ea65a3SJosef Bacik return root; 132106ea65a3SJosef Bacik } 132206ea65a3SJosef Bacik #endif 132306ea65a3SJosef Bacik 132420897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 132520897f5cSArne Jansen struct btrfs_fs_info *fs_info, 132620897f5cSArne Jansen u64 objectid) 132720897f5cSArne Jansen { 132820897f5cSArne Jansen struct extent_buffer *leaf; 132920897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 133020897f5cSArne Jansen struct btrfs_root *root; 133120897f5cSArne Jansen struct btrfs_key key; 133220897f5cSArne Jansen int ret = 0; 13336463fe58SStefan Behrens uuid_le uuid; 133420897f5cSArne Jansen 133520897f5cSArne Jansen root = btrfs_alloc_root(fs_info); 133620897f5cSArne Jansen if (!root) 133720897f5cSArne Jansen return ERR_PTR(-ENOMEM); 133820897f5cSArne Jansen 1339707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1340707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, objectid); 134120897f5cSArne Jansen root->root_key.objectid = objectid; 134220897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 134320897f5cSArne Jansen root->root_key.offset = 0; 134420897f5cSArne Jansen 13454d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 134620897f5cSArne Jansen if (IS_ERR(leaf)) { 134720897f5cSArne Jansen ret = PTR_ERR(leaf); 13481dd05682STsutomu Itoh leaf = NULL; 134920897f5cSArne Jansen goto fail; 135020897f5cSArne Jansen } 135120897f5cSArne Jansen 135220897f5cSArne Jansen memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 135320897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 135420897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 135520897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 135620897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 135720897f5cSArne Jansen root->node = leaf; 135820897f5cSArne Jansen 13590a4e5586SRoss Kirk write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 136020897f5cSArne Jansen BTRFS_FSID_SIZE); 136120897f5cSArne Jansen write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 1362b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 136320897f5cSArne Jansen BTRFS_UUID_SIZE); 136420897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 136520897f5cSArne Jansen 136620897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 136727cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 136820897f5cSArne Jansen 136920897f5cSArne Jansen root->root_item.flags = 0; 137020897f5cSArne Jansen root->root_item.byte_limit = 0; 137120897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 137220897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 137320897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 137420897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 137520897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 137620897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 137720897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 13786463fe58SStefan Behrens uuid_le_gen(&uuid); 13796463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 138020897f5cSArne Jansen root->root_item.drop_level = 0; 138120897f5cSArne Jansen 138220897f5cSArne Jansen key.objectid = objectid; 138320897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 138420897f5cSArne Jansen key.offset = 0; 138520897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 138620897f5cSArne Jansen if (ret) 138720897f5cSArne Jansen goto fail; 138820897f5cSArne Jansen 138920897f5cSArne Jansen btrfs_tree_unlock(leaf); 139020897f5cSArne Jansen 139120897f5cSArne Jansen return root; 13921dd05682STsutomu Itoh 13931dd05682STsutomu Itoh fail: 13941dd05682STsutomu Itoh if (leaf) { 13951dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 139659885b39STsutomu Itoh free_extent_buffer(root->commit_root); 13971dd05682STsutomu Itoh free_extent_buffer(leaf); 13981dd05682STsutomu Itoh } 13991dd05682STsutomu Itoh kfree(root); 14001dd05682STsutomu Itoh 14011dd05682STsutomu Itoh return ERR_PTR(ret); 140220897f5cSArne Jansen } 140320897f5cSArne Jansen 14047237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1405e02119d5SChris Mason struct btrfs_fs_info *fs_info) 14060f7d52f4SChris Mason { 14070f7d52f4SChris Mason struct btrfs_root *root; 14080f7d52f4SChris Mason struct btrfs_root *tree_root = fs_info->tree_root; 14097237f183SYan Zheng struct extent_buffer *leaf; 1410e02119d5SChris Mason 14116f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1412e02119d5SChris Mason if (!root) 14137237f183SYan Zheng return ERR_PTR(-ENOMEM); 1414e02119d5SChris Mason 1415707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1416707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, 1417707e8a07SDavid Sterba BTRFS_TREE_LOG_OBJECTID); 1418e02119d5SChris Mason 1419e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1420e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1421e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 142227cdeb70SMiao Xie 14237237f183SYan Zheng /* 142427cdeb70SMiao Xie * DON'T set REF_COWS for log trees 142527cdeb70SMiao Xie * 14267237f183SYan Zheng * log trees do not get reference counted because they go away 14277237f183SYan Zheng * before a real commit is actually done. They do store pointers 14287237f183SYan Zheng * to file data extents, and those reference counts still get 14297237f183SYan Zheng * updated (along with back refs to the log tree). 14307237f183SYan Zheng */ 1431e02119d5SChris Mason 14324d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 14334d75f8a9SDavid Sterba NULL, 0, 0, 0); 14347237f183SYan Zheng if (IS_ERR(leaf)) { 14357237f183SYan Zheng kfree(root); 14367237f183SYan Zheng return ERR_CAST(leaf); 14377237f183SYan Zheng } 1438e02119d5SChris Mason 14395d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 14405d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 14415d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 14425d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 14435d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 14447237f183SYan Zheng root->node = leaf; 1445e02119d5SChris Mason 1446e02119d5SChris Mason write_extent_buffer(root->node, root->fs_info->fsid, 14470a4e5586SRoss Kirk btrfs_header_fsid(), BTRFS_FSID_SIZE); 1448e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1449e02119d5SChris Mason btrfs_tree_unlock(root->node); 14507237f183SYan Zheng return root; 14517237f183SYan Zheng } 14527237f183SYan Zheng 14537237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 14547237f183SYan Zheng struct btrfs_fs_info *fs_info) 14557237f183SYan Zheng { 14567237f183SYan Zheng struct btrfs_root *log_root; 14577237f183SYan Zheng 14587237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 14597237f183SYan Zheng if (IS_ERR(log_root)) 14607237f183SYan Zheng return PTR_ERR(log_root); 14617237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14627237f183SYan Zheng fs_info->log_root_tree = log_root; 14637237f183SYan Zheng return 0; 14647237f183SYan Zheng } 14657237f183SYan Zheng 14667237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14677237f183SYan Zheng struct btrfs_root *root) 14687237f183SYan Zheng { 14697237f183SYan Zheng struct btrfs_root *log_root; 14707237f183SYan Zheng struct btrfs_inode_item *inode_item; 14717237f183SYan Zheng 14727237f183SYan Zheng log_root = alloc_log_tree(trans, root->fs_info); 14737237f183SYan Zheng if (IS_ERR(log_root)) 14747237f183SYan Zheng return PTR_ERR(log_root); 14757237f183SYan Zheng 14767237f183SYan Zheng log_root->last_trans = trans->transid; 14777237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14787237f183SYan Zheng 14797237f183SYan Zheng inode_item = &log_root->root_item.inode; 14803cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14813cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14823cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1483707e8a07SDavid Sterba btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 14843cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14857237f183SYan Zheng 14865d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14877237f183SYan Zheng 14887237f183SYan Zheng WARN_ON(root->log_root); 14897237f183SYan Zheng root->log_root = log_root; 14907237f183SYan Zheng root->log_transid = 0; 1491d1433debSMiao Xie root->log_transid_committed = -1; 1492257c62e1SChris Mason root->last_log_commit = 0; 1493e02119d5SChris Mason return 0; 1494e02119d5SChris Mason } 1495e02119d5SChris Mason 149635a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1497cb517eabSMiao Xie struct btrfs_key *key) 1498e02119d5SChris Mason { 1499e02119d5SChris Mason struct btrfs_root *root; 1500e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 15010f7d52f4SChris Mason struct btrfs_path *path; 150284234f3aSYan Zheng u64 generation; 1503cb517eabSMiao Xie int ret; 1504cb517eabSMiao Xie 1505cb517eabSMiao Xie path = btrfs_alloc_path(); 1506cb517eabSMiao Xie if (!path) 1507cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 15080f7d52f4SChris Mason 15096f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1510cb517eabSMiao Xie if (!root) { 1511cb517eabSMiao Xie ret = -ENOMEM; 1512cb517eabSMiao Xie goto alloc_fail; 15130f7d52f4SChris Mason } 15140f7d52f4SChris Mason 1515707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1516707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, key->objectid); 15170f7d52f4SChris Mason 1518cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1519cb517eabSMiao Xie &root->root_item, &root->root_key); 15200f7d52f4SChris Mason if (ret) { 152113a8a7c8SYan, Zheng if (ret > 0) 152213a8a7c8SYan, Zheng ret = -ENOENT; 1523cb517eabSMiao Xie goto find_fail; 15240f7d52f4SChris Mason } 152513a8a7c8SYan, Zheng 152684234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1527db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 1528ce86cd59SDavid Sterba generation); 152964c043deSLiu Bo if (IS_ERR(root->node)) { 153064c043deSLiu Bo ret = PTR_ERR(root->node); 1531cb517eabSMiao Xie goto find_fail; 1532cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1533cb517eabSMiao Xie ret = -EIO; 153464c043deSLiu Bo free_extent_buffer(root->node); 153564c043deSLiu Bo goto find_fail; 1536416bc658SJosef Bacik } 15375d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 153813a8a7c8SYan, Zheng out: 1539cb517eabSMiao Xie btrfs_free_path(path); 1540cb517eabSMiao Xie return root; 1541cb517eabSMiao Xie 1542cb517eabSMiao Xie find_fail: 1543cb517eabSMiao Xie kfree(root); 1544cb517eabSMiao Xie alloc_fail: 1545cb517eabSMiao Xie root = ERR_PTR(ret); 1546cb517eabSMiao Xie goto out; 1547cb517eabSMiao Xie } 1548cb517eabSMiao Xie 1549cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1550cb517eabSMiao Xie struct btrfs_key *location) 1551cb517eabSMiao Xie { 1552cb517eabSMiao Xie struct btrfs_root *root; 1553cb517eabSMiao Xie 1554cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1555cb517eabSMiao Xie if (IS_ERR(root)) 1556cb517eabSMiao Xie return root; 1557cb517eabSMiao Xie 1558cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 155927cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 156008fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 156108fe4db1SLi Zefan } 156213a8a7c8SYan, Zheng 15635eda7b5eSChris Mason return root; 15645eda7b5eSChris Mason } 15655eda7b5eSChris Mason 1566cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1567cb517eabSMiao Xie { 1568cb517eabSMiao Xie int ret; 15698257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1570cb517eabSMiao Xie 1571cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1572cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1573cb517eabSMiao Xie GFP_NOFS); 1574cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1575cb517eabSMiao Xie ret = -ENOMEM; 1576cb517eabSMiao Xie goto fail; 1577cb517eabSMiao Xie } 1578cb517eabSMiao Xie 15798257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 15808257b2dcSMiao Xie if (IS_ERR(writers)) { 15818257b2dcSMiao Xie ret = PTR_ERR(writers); 15828257b2dcSMiao Xie goto fail; 15838257b2dcSMiao Xie } 15848257b2dcSMiao Xie root->subv_writers = writers; 15858257b2dcSMiao Xie 1586cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 158757cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 158857cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1589cb517eabSMiao Xie 1590cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1591cb517eabSMiao Xie if (ret) 15928257b2dcSMiao Xie goto free_writers; 1593f32e48e9SChandan Rajendra 1594f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1595f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1596f32e48e9SChandan Rajendra &root->highest_objectid); 1597f32e48e9SChandan Rajendra if (ret) { 1598f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1599f32e48e9SChandan Rajendra goto free_root_dev; 1600f32e48e9SChandan Rajendra } 1601f32e48e9SChandan Rajendra 1602f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1603f32e48e9SChandan Rajendra 1604f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1605f32e48e9SChandan Rajendra 1606cb517eabSMiao Xie return 0; 16078257b2dcSMiao Xie 1608f32e48e9SChandan Rajendra free_root_dev: 1609f32e48e9SChandan Rajendra free_anon_bdev(root->anon_dev); 16108257b2dcSMiao Xie free_writers: 16118257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 1612cb517eabSMiao Xie fail: 1613cb517eabSMiao Xie kfree(root->free_ino_ctl); 1614cb517eabSMiao Xie kfree(root->free_ino_pinned); 1615cb517eabSMiao Xie return ret; 1616cb517eabSMiao Xie } 1617cb517eabSMiao Xie 1618171170c1SSergei Trofimovich static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1619cb517eabSMiao Xie u64 root_id) 1620cb517eabSMiao Xie { 1621cb517eabSMiao Xie struct btrfs_root *root; 1622cb517eabSMiao Xie 1623cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1624cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1625cb517eabSMiao Xie (unsigned long)root_id); 1626cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1627cb517eabSMiao Xie return root; 1628cb517eabSMiao Xie } 1629cb517eabSMiao Xie 1630cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1631cb517eabSMiao Xie struct btrfs_root *root) 1632cb517eabSMiao Xie { 1633cb517eabSMiao Xie int ret; 1634cb517eabSMiao Xie 1635cb517eabSMiao Xie ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 1636cb517eabSMiao Xie if (ret) 1637cb517eabSMiao Xie return ret; 1638cb517eabSMiao Xie 1639cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1640cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1641cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1642cb517eabSMiao Xie root); 1643cb517eabSMiao Xie if (ret == 0) 164427cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1645cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1646cb517eabSMiao Xie radix_tree_preload_end(); 1647cb517eabSMiao Xie 1648cb517eabSMiao Xie return ret; 1649cb517eabSMiao Xie } 1650cb517eabSMiao Xie 1651c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1652c00869f1SMiao Xie struct btrfs_key *location, 1653c00869f1SMiao Xie bool check_ref) 16545eda7b5eSChris Mason { 16555eda7b5eSChris Mason struct btrfs_root *root; 1656381cf658SDavid Sterba struct btrfs_path *path; 16571d4c08e0SDavid Sterba struct btrfs_key key; 16585eda7b5eSChris Mason int ret; 16595eda7b5eSChris Mason 1660edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1661edbd8d4eSChris Mason return fs_info->tree_root; 1662edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1663edbd8d4eSChris Mason return fs_info->extent_root; 16648f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 16658f18cf13SChris Mason return fs_info->chunk_root; 16668f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 16678f18cf13SChris Mason return fs_info->dev_root; 16680403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 16690403e47eSYan Zheng return fs_info->csum_root; 1670bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1671bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1672bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1673f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1674f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1675f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 167670f6d82eSOmar Sandoval if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 167770f6d82eSOmar Sandoval return fs_info->free_space_root ? fs_info->free_space_root : 167870f6d82eSOmar Sandoval ERR_PTR(-ENOENT); 16794df27c4dSYan, Zheng again: 1680cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 168148475471SStefan Behrens if (root) { 1682c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 168348475471SStefan Behrens return ERR_PTR(-ENOENT); 16845eda7b5eSChris Mason return root; 168548475471SStefan Behrens } 16865eda7b5eSChris Mason 1687cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 16885eda7b5eSChris Mason if (IS_ERR(root)) 16895eda7b5eSChris Mason return root; 16903394e160SChris Mason 1691c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1692d68fc57bSYan, Zheng ret = -ENOENT; 1693d68fc57bSYan, Zheng goto fail; 1694d68fc57bSYan, Zheng } 1695d68fc57bSYan, Zheng 1696cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1697cb517eabSMiao Xie if (ret) 1698cb517eabSMiao Xie goto fail; 1699cb517eabSMiao Xie 1700381cf658SDavid Sterba path = btrfs_alloc_path(); 1701381cf658SDavid Sterba if (!path) { 1702381cf658SDavid Sterba ret = -ENOMEM; 1703381cf658SDavid Sterba goto fail; 1704381cf658SDavid Sterba } 17051d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 17061d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 17071d4c08e0SDavid Sterba key.offset = location->objectid; 17081d4c08e0SDavid Sterba 17091d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1710381cf658SDavid Sterba btrfs_free_path(path); 1711d68fc57bSYan, Zheng if (ret < 0) 1712d68fc57bSYan, Zheng goto fail; 1713d68fc57bSYan, Zheng if (ret == 0) 171427cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1715d68fc57bSYan, Zheng 1716cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17170f7d52f4SChris Mason if (ret) { 17184df27c4dSYan, Zheng if (ret == -EEXIST) { 17194df27c4dSYan, Zheng free_fs_root(root); 17204df27c4dSYan, Zheng goto again; 17210f7d52f4SChris Mason } 17224df27c4dSYan, Zheng goto fail; 17234df27c4dSYan, Zheng } 1724edbd8d4eSChris Mason return root; 17254df27c4dSYan, Zheng fail: 17264df27c4dSYan, Zheng free_fs_root(root); 17274df27c4dSYan, Zheng return ERR_PTR(ret); 1728edbd8d4eSChris Mason } 1729edbd8d4eSChris Mason 173004160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 173104160088SChris Mason { 173204160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 173304160088SChris Mason int ret = 0; 173404160088SChris Mason struct btrfs_device *device; 173504160088SChris Mason struct backing_dev_info *bdi; 1736b7967db7SChris Mason 17371f78160cSXiao Guangrong rcu_read_lock(); 17381f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1739dfe25020SChris Mason if (!device->bdev) 1740dfe25020SChris Mason continue; 174104160088SChris Mason bdi = blk_get_backing_dev_info(device->bdev); 1742ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 174304160088SChris Mason ret = 1; 174404160088SChris Mason break; 174504160088SChris Mason } 174604160088SChris Mason } 17471f78160cSXiao Guangrong rcu_read_unlock(); 174804160088SChris Mason return ret; 174904160088SChris Mason } 175004160088SChris Mason 175104160088SChris Mason static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 175204160088SChris Mason { 1753ad081f14SJens Axboe int err; 1754ad081f14SJens Axboe 1755b4caecd4SChristoph Hellwig err = bdi_setup_and_register(bdi, "btrfs"); 1756ad081f14SJens Axboe if (err) 1757ad081f14SJens Axboe return err; 1758ad081f14SJens Axboe 1759df0ce26cSChristoph Hellwig bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE; 176004160088SChris Mason bdi->congested_fn = btrfs_congested_fn; 176104160088SChris Mason bdi->congested_data = info; 1762da2f0f74SChris Mason bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 176304160088SChris Mason return 0; 176404160088SChris Mason } 176504160088SChris Mason 17668b712842SChris Mason /* 17678b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 17688b712842SChris Mason * functions. This is where read checksum verification actually happens 17698b712842SChris Mason */ 17708b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1771ce9adaa5SChris Mason { 1772ce9adaa5SChris Mason struct bio *bio; 177397eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1774ce9adaa5SChris Mason 177597eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1776ce9adaa5SChris Mason bio = end_io_wq->bio; 17778b712842SChris Mason 17784246a0b6SChristoph Hellwig bio->bi_error = end_io_wq->error; 1779ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1780ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 178197eb6b69SDavid Sterba kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 17824246a0b6SChristoph Hellwig bio_endio(bio); 1783ce9adaa5SChris Mason } 178444b8bd7eSChris Mason 1785a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1786a74a4b97SChris Mason { 1787a74a4b97SChris Mason struct btrfs_root *root = arg; 1788d0278245SMiao Xie int again; 1789da288d28SFilipe Manana struct btrfs_trans_handle *trans; 1790a74a4b97SChris Mason 1791a74a4b97SChris Mason do { 1792d0278245SMiao Xie again = 0; 17939d1a2a3aSDavid Sterba 1794d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 1795babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) 1796d0278245SMiao Xie goto sleep; 1797d0278245SMiao Xie 1798d0278245SMiao Xie if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1799d0278245SMiao Xie goto sleep; 1800d0278245SMiao Xie 1801dc7f370cSMiao Xie /* 1802dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1803dc7f370cSMiao Xie * during the above check and trylock. 1804dc7f370cSMiao Xie */ 1805babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) { 1806dc7f370cSMiao Xie mutex_unlock(&root->fs_info->cleaner_mutex); 1807dc7f370cSMiao Xie goto sleep; 1808dc7f370cSMiao Xie } 1809dc7f370cSMiao Xie 1810c2d6cb16SFilipe Manana mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); 181124bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 1812c2d6cb16SFilipe Manana mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); 1813c2d6cb16SFilipe Manana 18149d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 1815a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 1816a74a4b97SChris Mason 1817d0278245SMiao Xie /* 181805323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 181905323cd1SMiao Xie * needn't do anything special here. 1820d0278245SMiao Xie */ 1821d0278245SMiao Xie btrfs_run_defrag_inodes(root->fs_info); 182267c5e7d4SFilipe Manana 182367c5e7d4SFilipe Manana /* 182467c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 182567c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 182667c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 182767c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 182867c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 182967c5e7d4SFilipe Manana * unused block groups. 183067c5e7d4SFilipe Manana */ 183167c5e7d4SFilipe Manana btrfs_delete_unused_bgs(root->fs_info); 1832d0278245SMiao Xie sleep: 18339d1a2a3aSDavid Sterba if (!try_to_freeze() && !again) { 1834a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 18358929ecfaSYan, Zheng if (!kthread_should_stop()) 1836a74a4b97SChris Mason schedule(); 1837a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1838a74a4b97SChris Mason } 1839a74a4b97SChris Mason } while (!kthread_should_stop()); 1840da288d28SFilipe Manana 1841da288d28SFilipe Manana /* 1842da288d28SFilipe Manana * Transaction kthread is stopped before us and wakes us up. 1843da288d28SFilipe Manana * However we might have started a new transaction and COWed some 1844da288d28SFilipe Manana * tree blocks when deleting unused block groups for example. So 1845da288d28SFilipe Manana * make sure we commit the transaction we started to have a clean 1846da288d28SFilipe Manana * shutdown when evicting the btree inode - if it has dirty pages 1847da288d28SFilipe Manana * when we do the final iput() on it, eviction will trigger a 1848da288d28SFilipe Manana * writeback for it which will fail with null pointer dereferences 1849da288d28SFilipe Manana * since work queues and other resources were already released and 1850da288d28SFilipe Manana * destroyed by the time the iput/eviction/writeback is made. 1851da288d28SFilipe Manana */ 1852da288d28SFilipe Manana trans = btrfs_attach_transaction(root); 1853da288d28SFilipe Manana if (IS_ERR(trans)) { 1854da288d28SFilipe Manana if (PTR_ERR(trans) != -ENOENT) 1855da288d28SFilipe Manana btrfs_err(root->fs_info, 1856da288d28SFilipe Manana "cleaner transaction attach returned %ld", 1857da288d28SFilipe Manana PTR_ERR(trans)); 1858da288d28SFilipe Manana } else { 1859da288d28SFilipe Manana int ret; 1860da288d28SFilipe Manana 1861da288d28SFilipe Manana ret = btrfs_commit_transaction(trans, root); 1862da288d28SFilipe Manana if (ret) 1863da288d28SFilipe Manana btrfs_err(root->fs_info, 1864da288d28SFilipe Manana "cleaner open transaction commit returned %d", 1865da288d28SFilipe Manana ret); 1866da288d28SFilipe Manana } 1867da288d28SFilipe Manana 1868a74a4b97SChris Mason return 0; 1869a74a4b97SChris Mason } 1870a74a4b97SChris Mason 1871a74a4b97SChris Mason static int transaction_kthread(void *arg) 1872a74a4b97SChris Mason { 1873a74a4b97SChris Mason struct btrfs_root *root = arg; 1874a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1875a74a4b97SChris Mason struct btrfs_transaction *cur; 18768929ecfaSYan, Zheng u64 transid; 1877a74a4b97SChris Mason unsigned long now; 1878a74a4b97SChris Mason unsigned long delay; 1879914b2007SJan Kara bool cannot_commit; 1880a74a4b97SChris Mason 1881a74a4b97SChris Mason do { 1882914b2007SJan Kara cannot_commit = false; 18838b87dc17SDavid Sterba delay = HZ * root->fs_info->commit_interval; 1884a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 1885a74a4b97SChris Mason 1886a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 1887a74a4b97SChris Mason cur = root->fs_info->running_transaction; 1888a74a4b97SChris Mason if (!cur) { 1889a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1890a74a4b97SChris Mason goto sleep; 1891a74a4b97SChris Mason } 189231153d81SYan Zheng 1893a74a4b97SChris Mason now = get_seconds(); 18944a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 18958b87dc17SDavid Sterba (now < cur->start_time || 18968b87dc17SDavid Sterba now - cur->start_time < root->fs_info->commit_interval)) { 1897a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1898a74a4b97SChris Mason delay = HZ * 5; 1899a74a4b97SChris Mason goto sleep; 1900a74a4b97SChris Mason } 19018929ecfaSYan, Zheng transid = cur->transid; 1902a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 190356bec294SChris Mason 190479787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1905354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1906914b2007SJan Kara if (IS_ERR(trans)) { 1907354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1908914b2007SJan Kara cannot_commit = true; 190979787eaaSJeff Mahoney goto sleep; 1910914b2007SJan Kara } 19118929ecfaSYan, Zheng if (transid == trans->transid) { 191279787eaaSJeff Mahoney btrfs_commit_transaction(trans, root); 19138929ecfaSYan, Zheng } else { 19148929ecfaSYan, Zheng btrfs_end_transaction(trans, root); 19158929ecfaSYan, Zheng } 1916a74a4b97SChris Mason sleep: 1917a74a4b97SChris Mason wake_up_process(root->fs_info->cleaner_kthread); 1918a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1919a74a4b97SChris Mason 19204e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 19214e121c06SJosef Bacik &root->fs_info->fs_state))) 19224e121c06SJosef Bacik btrfs_cleanup_transaction(root); 1923a0acae0eSTejun Heo if (!try_to_freeze()) { 1924a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 19258929ecfaSYan, Zheng if (!kthread_should_stop() && 1926914b2007SJan Kara (!btrfs_transaction_blocked(root->fs_info) || 1927914b2007SJan Kara cannot_commit)) 1928a74a4b97SChris Mason schedule_timeout(delay); 1929a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1930a74a4b97SChris Mason } 1931a74a4b97SChris Mason } while (!kthread_should_stop()); 1932a74a4b97SChris Mason return 0; 1933a74a4b97SChris Mason } 1934a74a4b97SChris Mason 1935af31f5e5SChris Mason /* 1936af31f5e5SChris Mason * this will find the highest generation in the array of 1937af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1938af31f5e5SChris Mason * or -1 if we can't find anything. 1939af31f5e5SChris Mason * 1940af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1941af31f5e5SChris Mason * generation of the latest root in the array with the generation 1942af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1943af31f5e5SChris Mason */ 1944af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1945af31f5e5SChris Mason { 1946af31f5e5SChris Mason u64 cur; 1947af31f5e5SChris Mason int newest_index = -1; 1948af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1949af31f5e5SChris Mason int i; 1950af31f5e5SChris Mason 1951af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1952af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1953af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1954af31f5e5SChris Mason if (cur == newest_gen) 1955af31f5e5SChris Mason newest_index = i; 1956af31f5e5SChris Mason } 1957af31f5e5SChris Mason 1958af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1959af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1960af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1961af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1962af31f5e5SChris Mason if (cur == newest_gen) 1963af31f5e5SChris Mason newest_index = 0; 1964af31f5e5SChris Mason } 1965af31f5e5SChris Mason return newest_index; 1966af31f5e5SChris Mason } 1967af31f5e5SChris Mason 1968af31f5e5SChris Mason 1969af31f5e5SChris Mason /* 1970af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1971af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1972af31f5e5SChris Mason * field in the fs_info struct 1973af31f5e5SChris Mason */ 1974af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1975af31f5e5SChris Mason u64 newest_gen) 1976af31f5e5SChris Mason { 1977af31f5e5SChris Mason int newest_index = -1; 1978af31f5e5SChris Mason 1979af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1980af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1981af31f5e5SChris Mason if (newest_index == -1) { 1982af31f5e5SChris Mason info->backup_root_index = 0; 1983af31f5e5SChris Mason } else { 1984af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1985af31f5e5SChris Mason } 1986af31f5e5SChris Mason } 1987af31f5e5SChris Mason 1988af31f5e5SChris Mason /* 1989af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1990af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1991af31f5e5SChris Mason * done 1992af31f5e5SChris Mason */ 1993af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1994af31f5e5SChris Mason { 1995af31f5e5SChris Mason int next_backup; 1996af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1997af31f5e5SChris Mason int last_backup; 1998af31f5e5SChris Mason 1999af31f5e5SChris Mason next_backup = info->backup_root_index; 2000af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 2001af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2002af31f5e5SChris Mason 2003af31f5e5SChris Mason /* 2004af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 2005af31f5e5SChris Mason * this happens only at umount 2006af31f5e5SChris Mason */ 2007af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 2008af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 2009af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 2010af31f5e5SChris Mason next_backup = last_backup; 2011af31f5e5SChris Mason 2012af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 2013af31f5e5SChris Mason 2014af31f5e5SChris Mason /* 2015af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 2016af31f5e5SChris Mason * regardless of which ones we use today 2017af31f5e5SChris Mason */ 2018af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 2019af31f5e5SChris Mason 2020af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 2021af31f5e5SChris Mason 2022af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 2023af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 2024af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 2025af31f5e5SChris Mason 2026af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 2027af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 2028af31f5e5SChris Mason 2029af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 2030af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 2031af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 2032af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 2033af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 2034af31f5e5SChris Mason 2035af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 2036af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 2037af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 2038af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 2039af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 2040af31f5e5SChris Mason 20417c7e82a7SChris Mason /* 20427c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20437c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20447c7e82a7SChris Mason */ 20457c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20467c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20477c7e82a7SChris Mason info->fs_root->node->start); 2048af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2049af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2050af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2051af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20527c7e82a7SChris Mason } 2053af31f5e5SChris Mason 2054af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2055af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2056af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2057af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2058af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2059af31f5e5SChris Mason 2060af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 2061af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 2062af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 2063af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 2064af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 2065af31f5e5SChris Mason 2066af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2067af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2068af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2069af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2070af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2071af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2072af31f5e5SChris Mason 2073af31f5e5SChris Mason /* 2074af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2075af31f5e5SChris Mason * for the next commit 2076af31f5e5SChris Mason */ 2077af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2078af31f5e5SChris Mason &info->super_for_commit->super_roots, 2079af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2080af31f5e5SChris Mason } 2081af31f5e5SChris Mason 2082af31f5e5SChris Mason /* 2083af31f5e5SChris Mason * this copies info out of the root backup array and back into 2084af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 2085af31f5e5SChris Mason * the array, so you send it the number of backups you've already 2086af31f5e5SChris Mason * tried and the last backup index you used. 2087af31f5e5SChris Mason * 2088af31f5e5SChris Mason * this returns -1 when it has tried all the backups 2089af31f5e5SChris Mason */ 2090af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 2091af31f5e5SChris Mason struct btrfs_super_block *super, 2092af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 2093af31f5e5SChris Mason { 2094af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2095af31f5e5SChris Mason int newest = *backup_index; 2096af31f5e5SChris Mason 2097af31f5e5SChris Mason if (*num_backups_tried == 0) { 2098af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 2099af31f5e5SChris Mason 2100af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 2101af31f5e5SChris Mason if (newest == -1) 2102af31f5e5SChris Mason return -1; 2103af31f5e5SChris Mason 2104af31f5e5SChris Mason *backup_index = newest; 2105af31f5e5SChris Mason *num_backups_tried = 1; 2106af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 2107af31f5e5SChris Mason /* we've tried all the backups, all done */ 2108af31f5e5SChris Mason return -1; 2109af31f5e5SChris Mason } else { 2110af31f5e5SChris Mason /* jump to the next oldest backup */ 2111af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 2112af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2113af31f5e5SChris Mason *backup_index = newest; 2114af31f5e5SChris Mason *num_backups_tried += 1; 2115af31f5e5SChris Mason } 2116af31f5e5SChris Mason root_backup = super->super_roots + newest; 2117af31f5e5SChris Mason 2118af31f5e5SChris Mason btrfs_set_super_generation(super, 2119af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2120af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2121af31f5e5SChris Mason btrfs_set_super_root_level(super, 2122af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2123af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2124af31f5e5SChris Mason 2125af31f5e5SChris Mason /* 2126af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2127af31f5e5SChris Mason * need a fsck 2128af31f5e5SChris Mason */ 2129af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2130af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2131af31f5e5SChris Mason return 0; 2132af31f5e5SChris Mason } 2133af31f5e5SChris Mason 21347abadb64SLiu Bo /* helper to cleanup workers */ 21357abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 21367abadb64SLiu Bo { 2137dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2138afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 21395cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2140fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2141fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2142fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 21438b110e39SMiao Xie btrfs_destroy_workqueue(fs_info->endio_repair_workers); 2144d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2145fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 2146fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2147fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2148a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 21495b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2150e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2151736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2152a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2153fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2154a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 21557abadb64SLiu Bo } 21567abadb64SLiu Bo 21572e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21582e9f5954SRashika { 21592e9f5954SRashika if (root) { 21602e9f5954SRashika free_extent_buffer(root->node); 21612e9f5954SRashika free_extent_buffer(root->commit_root); 21622e9f5954SRashika root->node = NULL; 21632e9f5954SRashika root->commit_root = NULL; 21642e9f5954SRashika } 21652e9f5954SRashika } 21662e9f5954SRashika 2167af31f5e5SChris Mason /* helper to cleanup tree roots */ 2168af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2169af31f5e5SChris Mason { 21702e9f5954SRashika free_root_extent_buffers(info->tree_root); 2171655b09feSJosef Bacik 21722e9f5954SRashika free_root_extent_buffers(info->dev_root); 21732e9f5954SRashika free_root_extent_buffers(info->extent_root); 21742e9f5954SRashika free_root_extent_buffers(info->csum_root); 21752e9f5954SRashika free_root_extent_buffers(info->quota_root); 21762e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21772e9f5954SRashika if (chunk_root) 21782e9f5954SRashika free_root_extent_buffers(info->chunk_root); 217970f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2180af31f5e5SChris Mason } 2181af31f5e5SChris Mason 2182faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2183171f6537SJosef Bacik { 2184171f6537SJosef Bacik int ret; 2185171f6537SJosef Bacik struct btrfs_root *gang[8]; 2186171f6537SJosef Bacik int i; 2187171f6537SJosef Bacik 2188171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2189171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2190171f6537SJosef Bacik struct btrfs_root, root_list); 2191171f6537SJosef Bacik list_del(&gang[0]->root_list); 2192171f6537SJosef Bacik 219327cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2194cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2195171f6537SJosef Bacik } else { 2196171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2197171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2198b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2199171f6537SJosef Bacik } 2200171f6537SJosef Bacik } 2201171f6537SJosef Bacik 2202171f6537SJosef Bacik while (1) { 2203171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2204171f6537SJosef Bacik (void **)gang, 0, 2205171f6537SJosef Bacik ARRAY_SIZE(gang)); 2206171f6537SJosef Bacik if (!ret) 2207171f6537SJosef Bacik break; 2208171f6537SJosef Bacik for (i = 0; i < ret; i++) 2209cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2210171f6537SJosef Bacik } 22111a4319ccSLiu Bo 22121a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 22131a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 22141a4319ccSLiu Bo btrfs_destroy_pinned_extent(fs_info->tree_root, 22151a4319ccSLiu Bo fs_info->pinned_extents); 22161a4319ccSLiu Bo } 2217171f6537SJosef Bacik } 2218af31f5e5SChris Mason 2219638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2220638aa7edSEric Sandeen { 2221638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2222638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2223638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2224638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2225638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2226638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2227638aa7edSEric Sandeen fs_info->scrub_workers_refcnt = 0; 2228638aa7edSEric Sandeen } 2229638aa7edSEric Sandeen 2230779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2231779a65a4SEric Sandeen { 2232779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2233779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2234779a65a4SEric Sandeen atomic_set(&fs_info->balance_running, 0); 2235779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2236779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2237779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2238779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2239779a65a4SEric Sandeen } 2240779a65a4SEric Sandeen 2241f37938e0SEric Sandeen static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, 2242f37938e0SEric Sandeen struct btrfs_root *tree_root) 2243f37938e0SEric Sandeen { 2244f37938e0SEric Sandeen fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2245f37938e0SEric Sandeen set_nlink(fs_info->btree_inode, 1); 2246f37938e0SEric Sandeen /* 2247f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2248f37938e0SEric Sandeen * the real end of the address space is determined by all of 2249f37938e0SEric Sandeen * the devices in the system 2250f37938e0SEric Sandeen */ 2251f37938e0SEric Sandeen fs_info->btree_inode->i_size = OFFSET_MAX; 2252f37938e0SEric Sandeen fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 2253f37938e0SEric Sandeen 2254f37938e0SEric Sandeen RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2255f37938e0SEric Sandeen extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2256f37938e0SEric Sandeen fs_info->btree_inode->i_mapping); 2257f37938e0SEric Sandeen BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2258f37938e0SEric Sandeen extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 2259f37938e0SEric Sandeen 2260f37938e0SEric Sandeen BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2261f37938e0SEric Sandeen 2262f37938e0SEric Sandeen BTRFS_I(fs_info->btree_inode)->root = tree_root; 2263f37938e0SEric Sandeen memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 2264f37938e0SEric Sandeen sizeof(struct btrfs_key)); 2265f37938e0SEric Sandeen set_bit(BTRFS_INODE_DUMMY, 2266f37938e0SEric Sandeen &BTRFS_I(fs_info->btree_inode)->runtime_flags); 2267f37938e0SEric Sandeen btrfs_insert_inode_hash(fs_info->btree_inode); 2268f37938e0SEric Sandeen } 2269f37938e0SEric Sandeen 2270ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2271ad618368SEric Sandeen { 2272ad618368SEric Sandeen fs_info->dev_replace.lock_owner = 0; 2273ad618368SEric Sandeen atomic_set(&fs_info->dev_replace.nesting_level, 0); 2274ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2275ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_management_lock); 2276ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock); 2277ad618368SEric Sandeen init_waitqueue_head(&fs_info->replace_wait); 2278ad618368SEric Sandeen } 2279ad618368SEric Sandeen 2280f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2281f9e92e40SEric Sandeen { 2282f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2283f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2284f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2285f9e92e40SEric Sandeen fs_info->qgroup_op_tree = RB_ROOT; 2286f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2287f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2288f9e92e40SEric Sandeen fs_info->quota_enabled = 0; 2289f9e92e40SEric Sandeen fs_info->pending_quota_state = 0; 2290f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2291f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2292f9e92e40SEric Sandeen } 2293f9e92e40SEric Sandeen 22942a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 22952a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 22962a458198SEric Sandeen { 22972a458198SEric Sandeen int max_active = fs_info->thread_pool_size; 22986f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22992a458198SEric Sandeen 23002a458198SEric Sandeen fs_info->workers = 23012a458198SEric Sandeen btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, 23022a458198SEric Sandeen max_active, 16); 23032a458198SEric Sandeen 23042a458198SEric Sandeen fs_info->delalloc_workers = 23052a458198SEric Sandeen btrfs_alloc_workqueue("delalloc", flags, max_active, 2); 23062a458198SEric Sandeen 23072a458198SEric Sandeen fs_info->flush_workers = 23082a458198SEric Sandeen btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); 23092a458198SEric Sandeen 23102a458198SEric Sandeen fs_info->caching_workers = 23112a458198SEric Sandeen btrfs_alloc_workqueue("cache", flags, max_active, 0); 23122a458198SEric Sandeen 23132a458198SEric Sandeen /* 23142a458198SEric Sandeen * a higher idle thresh on the submit workers makes it much more 23152a458198SEric Sandeen * likely that bios will be send down in a sane order to the 23162a458198SEric Sandeen * devices 23172a458198SEric Sandeen */ 23182a458198SEric Sandeen fs_info->submit_workers = 23192a458198SEric Sandeen btrfs_alloc_workqueue("submit", flags, 23202a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 23212a458198SEric Sandeen max_active), 64); 23222a458198SEric Sandeen 23232a458198SEric Sandeen fs_info->fixup_workers = 23242a458198SEric Sandeen btrfs_alloc_workqueue("fixup", flags, 1, 0); 23252a458198SEric Sandeen 23262a458198SEric Sandeen /* 23272a458198SEric Sandeen * endios are largely parallel and should have a very 23282a458198SEric Sandeen * low idle thresh 23292a458198SEric Sandeen */ 23302a458198SEric Sandeen fs_info->endio_workers = 23312a458198SEric Sandeen btrfs_alloc_workqueue("endio", flags, max_active, 4); 23322a458198SEric Sandeen fs_info->endio_meta_workers = 23332a458198SEric Sandeen btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); 23342a458198SEric Sandeen fs_info->endio_meta_write_workers = 23352a458198SEric Sandeen btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); 23362a458198SEric Sandeen fs_info->endio_raid56_workers = 23372a458198SEric Sandeen btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); 23382a458198SEric Sandeen fs_info->endio_repair_workers = 23392a458198SEric Sandeen btrfs_alloc_workqueue("endio-repair", flags, 1, 0); 23402a458198SEric Sandeen fs_info->rmw_workers = 23412a458198SEric Sandeen btrfs_alloc_workqueue("rmw", flags, max_active, 2); 23422a458198SEric Sandeen fs_info->endio_write_workers = 23432a458198SEric Sandeen btrfs_alloc_workqueue("endio-write", flags, max_active, 2); 23442a458198SEric Sandeen fs_info->endio_freespace_worker = 23452a458198SEric Sandeen btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); 23462a458198SEric Sandeen fs_info->delayed_workers = 23472a458198SEric Sandeen btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); 23482a458198SEric Sandeen fs_info->readahead_workers = 23492a458198SEric Sandeen btrfs_alloc_workqueue("readahead", flags, max_active, 2); 23502a458198SEric Sandeen fs_info->qgroup_rescan_workers = 23512a458198SEric Sandeen btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); 23522a458198SEric Sandeen fs_info->extent_workers = 23532a458198SEric Sandeen btrfs_alloc_workqueue("extent-refs", flags, 23542a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 23552a458198SEric Sandeen max_active), 8); 23562a458198SEric Sandeen 23572a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 23582a458198SEric Sandeen fs_info->submit_workers && fs_info->flush_workers && 23592a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 23602a458198SEric Sandeen fs_info->endio_meta_write_workers && 23612a458198SEric Sandeen fs_info->endio_repair_workers && 23622a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23632a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 23642a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 23652a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 23662a458198SEric Sandeen fs_info->extent_workers && 23672a458198SEric Sandeen fs_info->qgroup_rescan_workers)) { 23682a458198SEric Sandeen return -ENOMEM; 23692a458198SEric Sandeen } 23702a458198SEric Sandeen 23712a458198SEric Sandeen return 0; 23722a458198SEric Sandeen } 23732a458198SEric Sandeen 237463443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 237563443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 237663443bf5SEric Sandeen { 237763443bf5SEric Sandeen int ret; 237863443bf5SEric Sandeen struct btrfs_root *tree_root = fs_info->tree_root; 237963443bf5SEric Sandeen struct btrfs_root *log_tree_root; 238063443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 238163443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 238263443bf5SEric Sandeen 238363443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2384f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 238563443bf5SEric Sandeen return -EIO; 238663443bf5SEric Sandeen } 238763443bf5SEric Sandeen 238863443bf5SEric Sandeen log_tree_root = btrfs_alloc_root(fs_info); 238963443bf5SEric Sandeen if (!log_tree_root) 239063443bf5SEric Sandeen return -ENOMEM; 239163443bf5SEric Sandeen 239263443bf5SEric Sandeen __setup_root(tree_root->nodesize, tree_root->sectorsize, 239363443bf5SEric Sandeen tree_root->stripesize, log_tree_root, fs_info, 239463443bf5SEric Sandeen BTRFS_TREE_LOG_OBJECTID); 239563443bf5SEric Sandeen 239663443bf5SEric Sandeen log_tree_root->node = read_tree_block(tree_root, bytenr, 239763443bf5SEric Sandeen fs_info->generation + 1); 239864c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2399f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 24000eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 240164c043deSLiu Bo kfree(log_tree_root); 24020eeff236SLiu Bo return ret; 240364c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2404f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 240563443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 240663443bf5SEric Sandeen kfree(log_tree_root); 240763443bf5SEric Sandeen return -EIO; 240863443bf5SEric Sandeen } 240963443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 241063443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 241163443bf5SEric Sandeen if (ret) { 2412a4553fefSAnand Jain btrfs_std_error(tree_root->fs_info, ret, 241363443bf5SEric Sandeen "Failed to recover log tree"); 241463443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 241563443bf5SEric Sandeen kfree(log_tree_root); 241663443bf5SEric Sandeen return ret; 241763443bf5SEric Sandeen } 241863443bf5SEric Sandeen 241963443bf5SEric Sandeen if (fs_info->sb->s_flags & MS_RDONLY) { 242063443bf5SEric Sandeen ret = btrfs_commit_super(tree_root); 242163443bf5SEric Sandeen if (ret) 242263443bf5SEric Sandeen return ret; 242363443bf5SEric Sandeen } 242463443bf5SEric Sandeen 242563443bf5SEric Sandeen return 0; 242663443bf5SEric Sandeen } 242763443bf5SEric Sandeen 24284bbcaa64SEric Sandeen static int btrfs_read_roots(struct btrfs_fs_info *fs_info, 24294bbcaa64SEric Sandeen struct btrfs_root *tree_root) 24304bbcaa64SEric Sandeen { 2431a4f3d2c4SDavid Sterba struct btrfs_root *root; 24324bbcaa64SEric Sandeen struct btrfs_key location; 24334bbcaa64SEric Sandeen int ret; 24344bbcaa64SEric Sandeen 24354bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 24364bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 24374bbcaa64SEric Sandeen location.offset = 0; 24384bbcaa64SEric Sandeen 2439a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2440a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2441a4f3d2c4SDavid Sterba return PTR_ERR(root); 2442a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2443a4f3d2c4SDavid Sterba fs_info->extent_root = root; 24444bbcaa64SEric Sandeen 24454bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2446a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2447a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2448a4f3d2c4SDavid Sterba return PTR_ERR(root); 2449a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2450a4f3d2c4SDavid Sterba fs_info->dev_root = root; 24514bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 24524bbcaa64SEric Sandeen 24534bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2454a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2455a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2456a4f3d2c4SDavid Sterba return PTR_ERR(root); 2457a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2458a4f3d2c4SDavid Sterba fs_info->csum_root = root; 24594bbcaa64SEric Sandeen 24604bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2461a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2462a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2463a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 24644bbcaa64SEric Sandeen fs_info->quota_enabled = 1; 24654bbcaa64SEric Sandeen fs_info->pending_quota_state = 1; 2466a4f3d2c4SDavid Sterba fs_info->quota_root = root; 24674bbcaa64SEric Sandeen } 24684bbcaa64SEric Sandeen 24694bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2470a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2471a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2472a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 24734bbcaa64SEric Sandeen if (ret != -ENOENT) 24744bbcaa64SEric Sandeen return ret; 24754bbcaa64SEric Sandeen } else { 2476a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2477a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 24784bbcaa64SEric Sandeen } 24794bbcaa64SEric Sandeen 248070f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 248170f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 248270f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 248370f6d82eSOmar Sandoval if (IS_ERR(root)) 248470f6d82eSOmar Sandoval return PTR_ERR(root); 248570f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 248670f6d82eSOmar Sandoval fs_info->free_space_root = root; 248770f6d82eSOmar Sandoval } 248870f6d82eSOmar Sandoval 24894bbcaa64SEric Sandeen return 0; 24904bbcaa64SEric Sandeen } 24914bbcaa64SEric Sandeen 2492ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2493dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2494dfe25020SChris Mason char *options) 2495eb60ceacSChris Mason { 2496db94535dSChris Mason u32 sectorsize; 2497db94535dSChris Mason u32 nodesize; 249887ee04ebSChris Mason u32 stripesize; 249984234f3aSYan Zheng u64 generation; 2500f2b636e8SJosef Bacik u64 features; 25013de4586cSChris Mason struct btrfs_key location; 2502a061fc8dSChris Mason struct buffer_head *bh; 25034d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2504815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2505f84a8bd6SAl Viro struct btrfs_root *tree_root; 25064d34b278SIlya Dryomov struct btrfs_root *chunk_root; 2507eb60ceacSChris Mason int ret; 2508e58ca020SYan int err = -EINVAL; 2509af31f5e5SChris Mason int num_backups_tried = 0; 2510af31f5e5SChris Mason int backup_index = 0; 25115cdc7ad3SQu Wenruo int max_active; 25124543df7eSChris Mason 2513f84a8bd6SAl Viro tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); 25146f07e42eSAl Viro chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); 2515cb517eabSMiao Xie if (!tree_root || !chunk_root) { 251639279cc3SChris Mason err = -ENOMEM; 251739279cc3SChris Mason goto fail; 251839279cc3SChris Mason } 251976dda93cSYan, Zheng 252076dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 252176dda93cSYan, Zheng if (ret) { 252276dda93cSYan, Zheng err = ret; 252376dda93cSYan, Zheng goto fail; 252476dda93cSYan, Zheng } 252576dda93cSYan, Zheng 252676dda93cSYan, Zheng ret = setup_bdi(fs_info, &fs_info->bdi); 252776dda93cSYan, Zheng if (ret) { 252876dda93cSYan, Zheng err = ret; 252976dda93cSYan, Zheng goto fail_srcu; 253076dda93cSYan, Zheng } 253176dda93cSYan, Zheng 2532908c7f19STejun Heo ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2533e2d84521SMiao Xie if (ret) { 2534e2d84521SMiao Xie err = ret; 2535e2d84521SMiao Xie goto fail_bdi; 2536e2d84521SMiao Xie } 2537e2d84521SMiao Xie fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * 2538e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2539e2d84521SMiao Xie 2540908c7f19STejun Heo ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2541963d678bSMiao Xie if (ret) { 2542963d678bSMiao Xie err = ret; 2543963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2544963d678bSMiao Xie } 2545963d678bSMiao Xie 2546908c7f19STejun Heo ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); 2547c404e0dcSMiao Xie if (ret) { 2548c404e0dcSMiao Xie err = ret; 2549c404e0dcSMiao Xie goto fail_delalloc_bytes; 2550c404e0dcSMiao Xie } 2551c404e0dcSMiao Xie 255276dda93cSYan, Zheng fs_info->btree_inode = new_inode(sb); 255376dda93cSYan, Zheng if (!fs_info->btree_inode) { 255476dda93cSYan, Zheng err = -ENOMEM; 2555c404e0dcSMiao Xie goto fail_bio_counter; 255676dda93cSYan, Zheng } 255776dda93cSYan, Zheng 2558a6591715SChris Mason mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 25591561dedaSMiao Xie 256076dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2561f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 25628fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2563facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 256424bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2565eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 256611833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2567eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2568a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 256976dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 257024bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 25714cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 25722bf64758SJosef Bacik spin_lock_init(&fs_info->free_chunk_lock); 2573f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2574ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2575fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2576f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 257747ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2578f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2579d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 258067c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 25817585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2582573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2583c2d6cb16SFilipe Manana mutex_init(&fs_info->cleaner_delayed_iput_mutex); 2584de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 258519c00ddcSChris Mason 25860b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 25876324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2588f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 258947ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 25900b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 259166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 259266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 259366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 259466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELALLOC); 259566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 259666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 259766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 259866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 259966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2600cb03c743SChris Mason atomic_set(&fs_info->nr_async_submits, 0); 2601771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 26028c8bee1dSChris Mason atomic_set(&fs_info->async_submit_draining, 0); 26030986fe9eSChris Mason atomic_set(&fs_info->nr_async_bios, 0); 26044cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2605fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 2606fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2607e20d96d6SChris Mason fs_info->sb = sb; 260895ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 26099ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 26104cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 26112bf64758SJosef Bacik fs_info->free_chunk_space = 0; 2612f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 26138b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2614f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 261590519d66SArne Jansen /* readahead state */ 2616d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 261790519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2618c8b97818SChris Mason 2619b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2620b34b086cSChris Mason num_online_cpus() + 2, 8); 26210afbaf8cSChris Mason 2622199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2623199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 262416cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 262516cdcec7SMiao Xie GFP_NOFS); 262616cdcec7SMiao Xie if (!fs_info->delayed_root) { 262716cdcec7SMiao Xie err = -ENOMEM; 262816cdcec7SMiao Xie goto fail_iput; 262916cdcec7SMiao Xie } 263016cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 26313eaa2885SChris Mason 2632638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 263321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 263421adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 263521adbd5cSStefan Behrens #endif 2636779a65a4SEric Sandeen btrfs_init_balance(fs_info); 263721c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2638a2de733cSArne Jansen 2639a061fc8dSChris Mason sb->s_blocksize = 4096; 2640a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(4096); 264132a88aa1SJens Axboe sb->s_bdi = &fs_info->bdi; 2642a061fc8dSChris Mason 2643f37938e0SEric Sandeen btrfs_init_btree_inode(fs_info, tree_root); 264439279cc3SChris Mason 2645e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 26466bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2647a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2648925baeddSChris Mason 264911833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[0], 2650f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 265111833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[1], 2652f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 265311833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2654e6dcd2dcSChris Mason fs_info->do_barriers = 1; 2655f9295749SChris Mason 2656d98237b3SChris Mason 26575a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 265830ae8467SChris Mason mutex_init(&fs_info->tree_log_mutex); 2659925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2660925baeddSChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2661a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 2662a74a4b97SChris Mason mutex_init(&fs_info->volume_mutex); 26631bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 26649e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2665c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 266676dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2667803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2668fa9c0d79SChris Mason 2669ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2670f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2671416ac51dSArne Jansen 2672fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2673fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2674fa9c0d79SChris Mason 26757d9eb12cSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 26763768f368SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2677bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 26784854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 26799a8dd150SChris Mason 268004216820SFilipe Manana INIT_LIST_HEAD(&fs_info->pinned_chunks); 268104216820SFilipe Manana 268253b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 268353b381b3SDavid Woodhouse if (ret) { 268483c8266aSDavid Sterba err = ret; 268553b381b3SDavid Woodhouse goto fail_alloc; 268653b381b3SDavid Woodhouse } 268753b381b3SDavid Woodhouse 2688707e8a07SDavid Sterba __setup_root(4096, 4096, 4096, tree_root, 26892c90e5d6SChris Mason fs_info, BTRFS_ROOT_TREE_OBJECTID); 26907eccb903SChris Mason 26913c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 26921104a885SDavid Sterba 26931104a885SDavid Sterba /* 26941104a885SDavid Sterba * Read super block and check the signature bytes only 26951104a885SDavid Sterba */ 2696a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 269792fc03fbSAnand Jain if (IS_ERR(bh)) { 269892fc03fbSAnand Jain err = PTR_ERR(bh); 269916cdcec7SMiao Xie goto fail_alloc; 270020b45077SDave Young } 270139279cc3SChris Mason 27021104a885SDavid Sterba /* 27031104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 27041104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 27051104a885SDavid Sterba */ 27061104a885SDavid Sterba if (btrfs_check_super_csum(bh->b_data)) { 2707efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); 27081104a885SDavid Sterba err = -EINVAL; 2709b2acdddfSAnand Jain brelse(bh); 27101104a885SDavid Sterba goto fail_alloc; 27111104a885SDavid Sterba } 27121104a885SDavid Sterba 27131104a885SDavid Sterba /* 27141104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 27151104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 27161104a885SDavid Sterba * the whole block of INFO_SIZE 27171104a885SDavid Sterba */ 27186c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 27196c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 27206c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2721a061fc8dSChris Mason brelse(bh); 27225f39d397SChris Mason 27236c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 27240b86a832SChris Mason 27251104a885SDavid Sterba ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 27261104a885SDavid Sterba if (ret) { 2727efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); 27281104a885SDavid Sterba err = -EINVAL; 27291104a885SDavid Sterba goto fail_alloc; 27301104a885SDavid Sterba } 27311104a885SDavid Sterba 27326c41761fSDavid Sterba disk_super = fs_info->super_copy; 27330f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 273416cdcec7SMiao Xie goto fail_alloc; 27350f7d52f4SChris Mason 2736acce952bSliubo /* check FS state, whether FS is broken. */ 273787533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 273887533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2739acce952bSliubo 274075e7cb7fSLiu Bo /* 2741af31f5e5SChris Mason * run through our array of backup supers and setup 2742af31f5e5SChris Mason * our ring pointer to the oldest one 2743af31f5e5SChris Mason */ 2744af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2745af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2746af31f5e5SChris Mason 2747af31f5e5SChris Mason /* 274875e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 274975e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 275075e7cb7fSLiu Bo */ 275175e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 275275e7cb7fSLiu Bo 275396da0919SQu Wenruo ret = btrfs_parse_options(tree_root, options, sb->s_flags); 27542b82032cSYan Zheng if (ret) { 27552b82032cSYan Zheng err = ret; 275616cdcec7SMiao Xie goto fail_alloc; 27572b82032cSYan Zheng } 2758dfe25020SChris Mason 2759f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2760f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2761f2b636e8SJosef Bacik if (features) { 2762f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount because of " 2763f2b636e8SJosef Bacik "unsupported optional features (%Lx).\n", 2764c1c9ff7cSGeert Uytterhoeven features); 2765f2b636e8SJosef Bacik err = -EINVAL; 276616cdcec7SMiao Xie goto fail_alloc; 2767f2b636e8SJosef Bacik } 2768f2b636e8SJosef Bacik 27695d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 27705d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 277169e380d1SLi Zefan if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2772a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2773727011e0SChris Mason 27743173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 27755efa0490SDavid Sterba printk(KERN_INFO "BTRFS: has skinny extents\n"); 27763173a18fSJosef Bacik 2777727011e0SChris Mason /* 2778727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2779727011e0SChris Mason * they are bigger than the page size 2780727011e0SChris Mason */ 2781707e8a07SDavid Sterba if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) { 2782727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2783efe120a0SFrank Holton printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); 2784727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2785727011e0SChris Mason } 2786727011e0SChris Mason 2787bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2788bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2789bc3f116fSChris Mason stripesize = btrfs_super_stripesize(disk_super); 2790707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2791963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2792bc3f116fSChris Mason 2793bc3f116fSChris Mason /* 2794bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2795bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2796bc3f116fSChris Mason */ 2797bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2798707e8a07SDavid Sterba (sectorsize != nodesize)) { 2799aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes " 2800bc3f116fSChris Mason "are not allowed for mixed block groups on %s\n", 2801bc3f116fSChris Mason sb->s_id); 2802bc3f116fSChris Mason goto fail_alloc; 2803bc3f116fSChris Mason } 2804bc3f116fSChris Mason 2805ceda0864SMiao Xie /* 2806ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2807ceda0864SMiao Xie * update the flag. 2808ceda0864SMiao Xie */ 28095d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 28105d4f98a2SYan Zheng 2811f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2812f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2813f2b636e8SJosef Bacik if (!(sb->s_flags & MS_RDONLY) && features) { 2814f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2815f2b636e8SJosef Bacik "unsupported option features (%Lx).\n", 2816c1c9ff7cSGeert Uytterhoeven features); 2817f2b636e8SJosef Bacik err = -EINVAL; 281816cdcec7SMiao Xie goto fail_alloc; 2819f2b636e8SJosef Bacik } 282061d92c32SChris Mason 28215cdc7ad3SQu Wenruo max_active = fs_info->thread_pool_size; 282261d92c32SChris Mason 28232a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 28242a458198SEric Sandeen if (ret) { 28252a458198SEric Sandeen err = ret; 28260dc3b84aSJosef Bacik goto fail_sb_buffer; 28270dc3b84aSJosef Bacik } 28284543df7eSChris Mason 28294575c9ccSChris Mason fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2830c8b97818SChris Mason fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2831ee22184bSByongho Lee SZ_4M / PAGE_CACHE_SIZE); 28324575c9ccSChris Mason 2833db94535dSChris Mason tree_root->nodesize = nodesize; 2834db94535dSChris Mason tree_root->sectorsize = sectorsize; 283587ee04ebSChris Mason tree_root->stripesize = stripesize; 2836a061fc8dSChris Mason 2837a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2838a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2839db94535dSChris Mason 2840925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 2841e4404d6eSYan Zheng ret = btrfs_read_sys_array(tree_root); 2842925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 284384eed90fSChris Mason if (ret) { 2844aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read the system " 2845d397712bSChris Mason "array on %s\n", sb->s_id); 28465d4f98a2SYan Zheng goto fail_sb_buffer; 284784eed90fSChris Mason } 28480b86a832SChris Mason 284984234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 28500b86a832SChris Mason 2851707e8a07SDavid Sterba __setup_root(nodesize, sectorsize, stripesize, chunk_root, 2852707e8a07SDavid Sterba fs_info, BTRFS_CHUNK_TREE_OBJECTID); 28530b86a832SChris Mason 28540b86a832SChris Mason chunk_root->node = read_tree_block(chunk_root, 28550b86a832SChris Mason btrfs_super_chunk_root(disk_super), 2856ce86cd59SDavid Sterba generation); 285764c043deSLiu Bo if (IS_ERR(chunk_root->node) || 285864c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 2859aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n", 286083121942SDavid Woodhouse sb->s_id); 2861e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 2862e5fffbacSchandan free_extent_buffer(chunk_root->node); 286395ab1f64SZhao Lei chunk_root->node = NULL; 2864af31f5e5SChris Mason goto fail_tree_roots; 286583121942SDavid Woodhouse } 28665d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 28675d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 28680b86a832SChris Mason 2869e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2870b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2871e17cade2SChris Mason 28720b86a832SChris Mason ret = btrfs_read_chunk_tree(chunk_root); 28732b82032cSYan Zheng if (ret) { 2874aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n", 2875d397712bSChris Mason sb->s_id); 2876af31f5e5SChris Mason goto fail_tree_roots; 28772b82032cSYan Zheng } 28780b86a832SChris Mason 28798dabb742SStefan Behrens /* 28808dabb742SStefan Behrens * keep the device that is marked to be the target device for the 28818dabb742SStefan Behrens * dev_replace procedure 28828dabb742SStefan Behrens */ 28839eaed21eSEric Sandeen btrfs_close_extra_devices(fs_devices, 0); 2884dfe25020SChris Mason 2885a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 2886aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read devices on %s\n", 2887a6b0d5c8SChris Mason sb->s_id); 2888a6b0d5c8SChris Mason goto fail_tree_roots; 2889a6b0d5c8SChris Mason } 2890a6b0d5c8SChris Mason 2891af31f5e5SChris Mason retry_root_backup: 289284234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 28930b86a832SChris Mason 2894e20d96d6SChris Mason tree_root->node = read_tree_block(tree_root, 2895db94535dSChris Mason btrfs_super_root(disk_super), 2896ce86cd59SDavid Sterba generation); 289764c043deSLiu Bo if (IS_ERR(tree_root->node) || 289864c043deSLiu Bo !extent_buffer_uptodate(tree_root->node)) { 2899efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", 290083121942SDavid Woodhouse sb->s_id); 2901e5fffbacSchandan if (!IS_ERR(tree_root->node)) 2902e5fffbacSchandan free_extent_buffer(tree_root->node); 290395ab1f64SZhao Lei tree_root->node = NULL; 2904af31f5e5SChris Mason goto recovery_tree_root; 290583121942SDavid Woodhouse } 2906af31f5e5SChris Mason 29075d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 29085d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 290969e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2910db94535dSChris Mason 2911f32e48e9SChandan Rajendra mutex_lock(&tree_root->objectid_mutex); 2912f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(tree_root, 2913f32e48e9SChandan Rajendra &tree_root->highest_objectid); 2914f32e48e9SChandan Rajendra if (ret) { 2915f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2916f32e48e9SChandan Rajendra goto recovery_tree_root; 2917f32e48e9SChandan Rajendra } 2918f32e48e9SChandan Rajendra 2919f32e48e9SChandan Rajendra ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2920f32e48e9SChandan Rajendra 2921f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2922f32e48e9SChandan Rajendra 29234bbcaa64SEric Sandeen ret = btrfs_read_roots(fs_info, tree_root); 29244bbcaa64SEric Sandeen if (ret) 2925af31f5e5SChris Mason goto recovery_tree_root; 2926f7a81ea4SStefan Behrens 29278929ecfaSYan, Zheng fs_info->generation = generation; 29288929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 29298929ecfaSYan, Zheng 293068310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 293168310a5eSIlya Dryomov if (ret) { 2932aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to recover balance\n"); 293368310a5eSIlya Dryomov goto fail_block_groups; 293468310a5eSIlya Dryomov } 293568310a5eSIlya Dryomov 2936733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2937733f4fbbSStefan Behrens if (ret) { 2938efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", 2939733f4fbbSStefan Behrens ret); 2940733f4fbbSStefan Behrens goto fail_block_groups; 2941733f4fbbSStefan Behrens } 2942733f4fbbSStefan Behrens 29438dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 29448dabb742SStefan Behrens if (ret) { 2945efe120a0SFrank Holton pr_err("BTRFS: failed to init dev_replace: %d\n", ret); 29468dabb742SStefan Behrens goto fail_block_groups; 29478dabb742SStefan Behrens } 29488dabb742SStefan Behrens 29499eaed21eSEric Sandeen btrfs_close_extra_devices(fs_devices, 1); 29508dabb742SStefan Behrens 2951b7c35e81SAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices, NULL); 2952b7c35e81SAnand Jain if (ret) { 2953b7c35e81SAnand Jain pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret); 2954b7c35e81SAnand Jain goto fail_block_groups; 2955b7c35e81SAnand Jain } 2956b7c35e81SAnand Jain 2957b7c35e81SAnand Jain ret = btrfs_sysfs_add_device(fs_devices); 2958b7c35e81SAnand Jain if (ret) { 2959b7c35e81SAnand Jain pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret); 2960b7c35e81SAnand Jain goto fail_fsdev_sysfs; 2961b7c35e81SAnand Jain } 2962b7c35e81SAnand Jain 296396f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 29645ac1d209SJeff Mahoney if (ret) { 2965efe120a0SFrank Holton pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); 2966b7c35e81SAnand Jain goto fail_fsdev_sysfs; 29675ac1d209SJeff Mahoney } 29685ac1d209SJeff Mahoney 2969c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2970c59021f8Sliubo if (ret) { 2971efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); 29722365dd3cSAnand Jain goto fail_sysfs; 2973c59021f8Sliubo } 2974c59021f8Sliubo 29754bbcaa64SEric Sandeen ret = btrfs_read_block_groups(fs_info->extent_root); 29761b1d1f66SJosef Bacik if (ret) { 2977efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); 29782365dd3cSAnand Jain goto fail_sysfs; 29791b1d1f66SJosef Bacik } 29805af3e8ccSStefan Behrens fs_info->num_tolerated_disk_barrier_failures = 29815af3e8ccSStefan Behrens btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2982292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 2983292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 2984292fd7fcSStefan Behrens !(sb->s_flags & MS_RDONLY)) { 298578fa1770SZhao Lei pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n", 298678fa1770SZhao Lei fs_info->fs_devices->missing_devices, 298778fa1770SZhao Lei fs_info->num_tolerated_disk_barrier_failures); 29882365dd3cSAnand Jain goto fail_sysfs; 2989292fd7fcSStefan Behrens } 29909078a3e1SChris Mason 2991a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2992a74a4b97SChris Mason "btrfs-cleaner"); 299357506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 29942365dd3cSAnand Jain goto fail_sysfs; 2995a74a4b97SChris Mason 2996a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2997a74a4b97SChris Mason tree_root, 2998a74a4b97SChris Mason "btrfs-transaction"); 299957506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 30003f157a2fSChris Mason goto fail_cleaner; 3001a74a4b97SChris Mason 3002c289811cSChris Mason if (!btrfs_test_opt(tree_root, SSD) && 3003c289811cSChris Mason !btrfs_test_opt(tree_root, NOSSD) && 3004c289811cSChris Mason !fs_info->fs_devices->rotating) { 3005efe120a0SFrank Holton printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " 3006c289811cSChris Mason "mode\n"); 3007c289811cSChris Mason btrfs_set_opt(fs_info->mount_opt, SSD); 3008c289811cSChris Mason } 3009c289811cSChris Mason 3010572d9ab7SDavid Sterba /* 3011572d9ab7SDavid Sterba * Mount does not set all options immediatelly, we can do it now and do 3012572d9ab7SDavid Sterba * not have to wait for transaction commit 3013572d9ab7SDavid Sterba */ 3014572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 30153818aea2SQu Wenruo 301621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 301721adbd5cSStefan Behrens if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 301821adbd5cSStefan Behrens ret = btrfsic_mount(tree_root, fs_devices, 301921adbd5cSStefan Behrens btrfs_test_opt(tree_root, 302021adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 302121adbd5cSStefan Behrens 1 : 0, 302221adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 302321adbd5cSStefan Behrens if (ret) 3024efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to initialize" 302521adbd5cSStefan Behrens " integrity check module %s\n", sb->s_id); 302621adbd5cSStefan Behrens } 302721adbd5cSStefan Behrens #endif 3028bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3029bcef60f2SArne Jansen if (ret) 3030bcef60f2SArne Jansen goto fail_trans_kthread; 303121adbd5cSStefan Behrens 303296da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 303396da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 303496da0919SQu Wenruo !btrfs_test_opt(tree_root, NOLOGREPLAY)) { 303563443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 303679787eaaSJeff Mahoney if (ret) { 303763443bf5SEric Sandeen err = ret; 303828c16cbbSWang Shilong goto fail_qgroup; 3039e556ce2cSYan Zheng } 3040e02119d5SChris Mason } 30411a40e23bSZheng Yan 304276dda93cSYan, Zheng ret = btrfs_find_orphan_roots(tree_root); 304379787eaaSJeff Mahoney if (ret) 304428c16cbbSWang Shilong goto fail_qgroup; 304576dda93cSYan, Zheng 30467c2ca468SChris Mason if (!(sb->s_flags & MS_RDONLY)) { 3047d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 304844c44af2SIlya Dryomov if (ret) 304928c16cbbSWang Shilong goto fail_qgroup; 3050d68fc57bSYan, Zheng 30515f316481SWang Shilong mutex_lock(&fs_info->cleaner_mutex); 30525d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 30535f316481SWang Shilong mutex_unlock(&fs_info->cleaner_mutex); 3054d7ce5843SMiao Xie if (ret < 0) { 3055d7ce5843SMiao Xie printk(KERN_WARNING 3056efe120a0SFrank Holton "BTRFS: failed to recover relocation\n"); 3057d7ce5843SMiao Xie err = -EINVAL; 3058bcef60f2SArne Jansen goto fail_qgroup; 3059d7ce5843SMiao Xie } 30607c2ca468SChris Mason } 30611a40e23bSZheng Yan 30623de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 30633de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 3064cb517eabSMiao Xie location.offset = 0; 30653de4586cSChris Mason 30663de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 30673140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 30683140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3069bcef60f2SArne Jansen goto fail_qgroup; 30703140c9a3SDan Carpenter } 3071c289811cSChris Mason 30722b6ba629SIlya Dryomov if (sb->s_flags & MS_RDONLY) 30732b6ba629SIlya Dryomov return 0; 30742b6ba629SIlya Dryomov 3075511711afSChris Mason if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) && 3076511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 3077511711afSChris Mason pr_info("BTRFS: creating free space tree\n"); 3078511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3079511711afSChris Mason if (ret) { 3080511711afSChris Mason pr_warn("BTRFS: failed to create free space tree %d\n", 3081511711afSChris Mason ret); 3082511711afSChris Mason close_ctree(tree_root); 3083511711afSChris Mason return ret; 3084511711afSChris Mason } 3085511711afSChris Mason } 3086511711afSChris Mason 3087e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 30882b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 30892b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 30902b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 30912b6ba629SIlya Dryomov close_ctree(tree_root); 30922b6ba629SIlya Dryomov return ret; 30932b6ba629SIlya Dryomov } 3094e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 309559641015SIlya Dryomov 30962b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 30972b6ba629SIlya Dryomov if (ret) { 3098efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to resume balance\n"); 309966b4ffd1SJosef Bacik close_ctree(tree_root); 31002b6ba629SIlya Dryomov return ret; 3101e3acc2a6SJosef Bacik } 3102e3acc2a6SJosef Bacik 31038dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 31048dabb742SStefan Behrens if (ret) { 3105efe120a0SFrank Holton pr_warn("BTRFS: failed to resume dev_replace\n"); 31068dabb742SStefan Behrens close_ctree(tree_root); 31078dabb742SStefan Behrens return ret; 31088dabb742SStefan Behrens } 31098dabb742SStefan Behrens 3110b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3111b382a324SJan Schmidt 311270f6d82eSOmar Sandoval if (btrfs_test_opt(tree_root, CLEAR_CACHE) && 311370f6d82eSOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 311470f6d82eSOmar Sandoval pr_info("BTRFS: clearing free space tree\n"); 311570f6d82eSOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 311670f6d82eSOmar Sandoval if (ret) { 311770f6d82eSOmar Sandoval pr_warn("BTRFS: failed to clear free space tree %d\n", 311870f6d82eSOmar Sandoval ret); 311970f6d82eSOmar Sandoval close_ctree(tree_root); 312070f6d82eSOmar Sandoval return ret; 312170f6d82eSOmar Sandoval } 312270f6d82eSOmar Sandoval } 312370f6d82eSOmar Sandoval 31244bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 3125efe120a0SFrank Holton pr_info("BTRFS: creating UUID tree\n"); 3126f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3127f7a81ea4SStefan Behrens if (ret) { 3128efe120a0SFrank Holton pr_warn("BTRFS: failed to create the UUID tree %d\n", 3129f7a81ea4SStefan Behrens ret); 3130f7a81ea4SStefan Behrens close_ctree(tree_root); 3131f7a81ea4SStefan Behrens return ret; 3132f7a81ea4SStefan Behrens } 31334bbcaa64SEric Sandeen } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) || 31344bbcaa64SEric Sandeen fs_info->generation != 31354bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 3136efe120a0SFrank Holton pr_info("BTRFS: checking UUID tree\n"); 313770f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 313870f80175SStefan Behrens if (ret) { 3139efe120a0SFrank Holton pr_warn("BTRFS: failed to check the UUID tree %d\n", 314070f80175SStefan Behrens ret); 314170f80175SStefan Behrens close_ctree(tree_root); 314270f80175SStefan Behrens return ret; 314370f80175SStefan Behrens } 314470f80175SStefan Behrens } else { 314570f80175SStefan Behrens fs_info->update_uuid_tree_gen = 1; 3146f7a81ea4SStefan Behrens } 3147f7a81ea4SStefan Behrens 314847ab2a6cSJosef Bacik fs_info->open = 1; 314947ab2a6cSJosef Bacik 31508dcddfa0SQu Wenruo /* 31518dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 31528dcddfa0SQu Wenruo * no need to keep the flag 31538dcddfa0SQu Wenruo */ 31548dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 31558dcddfa0SQu Wenruo 3156ad2b2c80SAl Viro return 0; 315739279cc3SChris Mason 3158bcef60f2SArne Jansen fail_qgroup: 3159bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 31607c2ca468SChris Mason fail_trans_kthread: 31617c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 316254067ae9SJosef Bacik btrfs_cleanup_transaction(fs_info->tree_root); 3163faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 31643f157a2fSChris Mason fail_cleaner: 3165a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 31667c2ca468SChris Mason 31677c2ca468SChris Mason /* 31687c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 31697c2ca468SChris Mason * kthreads 31707c2ca468SChris Mason */ 31717c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 31727c2ca468SChris Mason 31732365dd3cSAnand Jain fail_sysfs: 31746618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 31752365dd3cSAnand Jain 3176b7c35e81SAnand Jain fail_fsdev_sysfs: 3177b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3178b7c35e81SAnand Jain 31791b1d1f66SJosef Bacik fail_block_groups: 318054067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 31811b1d1f66SJosef Bacik btrfs_free_block_groups(fs_info); 3182af31f5e5SChris Mason 3183af31f5e5SChris Mason fail_tree_roots: 3184af31f5e5SChris Mason free_root_pointers(fs_info, 1); 31852b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3186af31f5e5SChris Mason 318739279cc3SChris Mason fail_sb_buffer: 31887abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 318916cdcec7SMiao Xie fail_alloc: 31904543df7eSChris Mason fail_iput: 3191586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3192586e46e2SIlya Dryomov 31934543df7eSChris Mason iput(fs_info->btree_inode); 3194c404e0dcSMiao Xie fail_bio_counter: 3195c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 3196963d678bSMiao Xie fail_delalloc_bytes: 3197963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3198e2d84521SMiao Xie fail_dirty_metadata_bytes: 3199e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3200ad081f14SJens Axboe fail_bdi: 32017e662854SQinghuang Feng bdi_destroy(&fs_info->bdi); 320276dda93cSYan, Zheng fail_srcu: 320376dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 32047e662854SQinghuang Feng fail: 320553b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3206586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3207ad2b2c80SAl Viro return err; 3208af31f5e5SChris Mason 3209af31f5e5SChris Mason recovery_tree_root: 32108dcddfa0SQu Wenruo if (!btrfs_test_opt(tree_root, USEBACKUPROOT)) 3211af31f5e5SChris Mason goto fail_tree_roots; 3212af31f5e5SChris Mason 3213af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3214af31f5e5SChris Mason 3215af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3216af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3217af31f5e5SChris Mason 3218af31f5e5SChris Mason /* we can't trust the free space cache either */ 3219af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3220af31f5e5SChris Mason 3221af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3222af31f5e5SChris Mason &num_backups_tried, &backup_index); 3223af31f5e5SChris Mason if (ret == -1) 3224af31f5e5SChris Mason goto fail_block_groups; 3225af31f5e5SChris Mason goto retry_root_backup; 3226eb60ceacSChris Mason } 3227eb60ceacSChris Mason 3228f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3229f2984462SChris Mason { 3230f2984462SChris Mason if (uptodate) { 3231f2984462SChris Mason set_buffer_uptodate(bh); 3232f2984462SChris Mason } else { 3233442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3234442a4f63SStefan Behrens bh->b_private; 3235442a4f63SStefan Behrens 3236b14af3b4SDavid Sterba btrfs_warn_rl_in_rcu(device->dev_root->fs_info, 3237b14af3b4SDavid Sterba "lost page write due to IO error on %s", 3238606686eeSJosef Bacik rcu_str_deref(device->name)); 32391259ab75SChris Mason /* note, we dont' set_buffer_write_io_error because we have 32401259ab75SChris Mason * our own ways of dealing with the IO errors 32411259ab75SChris Mason */ 3242f2984462SChris Mason clear_buffer_uptodate(bh); 3243442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3244f2984462SChris Mason } 3245f2984462SChris Mason unlock_buffer(bh); 3246f2984462SChris Mason put_bh(bh); 3247f2984462SChris Mason } 3248f2984462SChris Mason 324929c36d72SAnand Jain int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, 325029c36d72SAnand Jain struct buffer_head **bh_ret) 325129c36d72SAnand Jain { 325229c36d72SAnand Jain struct buffer_head *bh; 325329c36d72SAnand Jain struct btrfs_super_block *super; 325429c36d72SAnand Jain u64 bytenr; 325529c36d72SAnand Jain 325629c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 325729c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 325829c36d72SAnand Jain return -EINVAL; 325929c36d72SAnand Jain 326029c36d72SAnand Jain bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE); 326129c36d72SAnand Jain /* 326229c36d72SAnand Jain * If we fail to read from the underlying devices, as of now 326329c36d72SAnand Jain * the best option we have is to mark it EIO. 326429c36d72SAnand Jain */ 326529c36d72SAnand Jain if (!bh) 326629c36d72SAnand Jain return -EIO; 326729c36d72SAnand Jain 326829c36d72SAnand Jain super = (struct btrfs_super_block *)bh->b_data; 326929c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 327029c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 327129c36d72SAnand Jain brelse(bh); 327229c36d72SAnand Jain return -EINVAL; 327329c36d72SAnand Jain } 327429c36d72SAnand Jain 327529c36d72SAnand Jain *bh_ret = bh; 327629c36d72SAnand Jain return 0; 327729c36d72SAnand Jain } 327829c36d72SAnand Jain 327929c36d72SAnand Jain 3280a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3281a512bbf8SYan Zheng { 3282a512bbf8SYan Zheng struct buffer_head *bh; 3283a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3284a512bbf8SYan Zheng struct btrfs_super_block *super; 3285a512bbf8SYan Zheng int i; 3286a512bbf8SYan Zheng u64 transid = 0; 328792fc03fbSAnand Jain int ret = -EINVAL; 3288a512bbf8SYan Zheng 3289a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3290a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3291a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3292a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3293a512bbf8SYan Zheng */ 3294a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 329529c36d72SAnand Jain ret = btrfs_read_dev_one_super(bdev, i, &bh); 329629c36d72SAnand Jain if (ret) 3297a512bbf8SYan Zheng continue; 3298a512bbf8SYan Zheng 3299a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3300a512bbf8SYan Zheng 3301a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3302a512bbf8SYan Zheng brelse(latest); 3303a512bbf8SYan Zheng latest = bh; 3304a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3305a512bbf8SYan Zheng } else { 3306a512bbf8SYan Zheng brelse(bh); 3307a512bbf8SYan Zheng } 3308a512bbf8SYan Zheng } 330992fc03fbSAnand Jain 331092fc03fbSAnand Jain if (!latest) 331192fc03fbSAnand Jain return ERR_PTR(ret); 331292fc03fbSAnand Jain 3313a512bbf8SYan Zheng return latest; 3314a512bbf8SYan Zheng } 3315a512bbf8SYan Zheng 33164eedeb75SHisashi Hifumi /* 33174eedeb75SHisashi Hifumi * this should be called twice, once with wait == 0 and 33184eedeb75SHisashi Hifumi * once with wait == 1. When wait == 0 is done, all the buffer heads 33194eedeb75SHisashi Hifumi * we write are pinned. 33204eedeb75SHisashi Hifumi * 33214eedeb75SHisashi Hifumi * They are released when wait == 1 is done. 33224eedeb75SHisashi Hifumi * max_mirrors must be the same for both runs, and it indicates how 33234eedeb75SHisashi Hifumi * many supers on this one device should be written. 33244eedeb75SHisashi Hifumi * 33254eedeb75SHisashi Hifumi * max_mirrors == 0 means to write them all. 33264eedeb75SHisashi Hifumi */ 3327a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3328a512bbf8SYan Zheng struct btrfs_super_block *sb, 3329a512bbf8SYan Zheng int do_barriers, int wait, int max_mirrors) 3330a512bbf8SYan Zheng { 3331a512bbf8SYan Zheng struct buffer_head *bh; 3332a512bbf8SYan Zheng int i; 3333a512bbf8SYan Zheng int ret; 3334a512bbf8SYan Zheng int errors = 0; 3335a512bbf8SYan Zheng u32 crc; 3336a512bbf8SYan Zheng u64 bytenr; 3337a512bbf8SYan Zheng 3338a512bbf8SYan Zheng if (max_mirrors == 0) 3339a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3340a512bbf8SYan Zheng 3341a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3342a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3343935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3344935e5cc9SMiao Xie device->commit_total_bytes) 3345a512bbf8SYan Zheng break; 3346a512bbf8SYan Zheng 3347a512bbf8SYan Zheng if (wait) { 3348a512bbf8SYan Zheng bh = __find_get_block(device->bdev, bytenr / 4096, 3349a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3350634554dcSJosef Bacik if (!bh) { 3351634554dcSJosef Bacik errors++; 3352634554dcSJosef Bacik continue; 3353634554dcSJosef Bacik } 3354a512bbf8SYan Zheng wait_on_buffer(bh); 33554eedeb75SHisashi Hifumi if (!buffer_uptodate(bh)) 33564eedeb75SHisashi Hifumi errors++; 33574eedeb75SHisashi Hifumi 33584eedeb75SHisashi Hifumi /* drop our reference */ 33594eedeb75SHisashi Hifumi brelse(bh); 33604eedeb75SHisashi Hifumi 33614eedeb75SHisashi Hifumi /* drop the reference from the wait == 0 run */ 3362a512bbf8SYan Zheng brelse(bh); 3363a512bbf8SYan Zheng continue; 3364a512bbf8SYan Zheng } else { 3365a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3366a512bbf8SYan Zheng 3367a512bbf8SYan Zheng crc = ~(u32)0; 3368b0496686SLiu Bo crc = btrfs_csum_data((char *)sb + 3369a512bbf8SYan Zheng BTRFS_CSUM_SIZE, crc, 3370a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE - 3371a512bbf8SYan Zheng BTRFS_CSUM_SIZE); 3372a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3373a512bbf8SYan Zheng 33744eedeb75SHisashi Hifumi /* 33754eedeb75SHisashi Hifumi * one reference for us, and we leave it for the 33764eedeb75SHisashi Hifumi * caller 33774eedeb75SHisashi Hifumi */ 3378a512bbf8SYan Zheng bh = __getblk(device->bdev, bytenr / 4096, 3379a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3380634554dcSJosef Bacik if (!bh) { 3381f14d104dSDavid Sterba btrfs_err(device->dev_root->fs_info, 3382f14d104dSDavid Sterba "couldn't get super buffer head for bytenr %llu", 3383f14d104dSDavid Sterba bytenr); 3384634554dcSJosef Bacik errors++; 3385634554dcSJosef Bacik continue; 3386634554dcSJosef Bacik } 3387634554dcSJosef Bacik 3388a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3389a512bbf8SYan Zheng 33904eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3391a512bbf8SYan Zheng get_bh(bh); 33924eedeb75SHisashi Hifumi 33934eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3394a512bbf8SYan Zheng lock_buffer(bh); 3395a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3396442a4f63SStefan Behrens bh->b_private = device; 3397a512bbf8SYan Zheng } 3398a512bbf8SYan Zheng 3399387125fcSChris Mason /* 3400387125fcSChris Mason * we fua the first super. The others we allow 3401387125fcSChris Mason * to go down lazy. 3402387125fcSChris Mason */ 3403e8117c26SWang Shilong if (i == 0) 340421adbd5cSStefan Behrens ret = btrfsic_submit_bh(WRITE_FUA, bh); 3405e8117c26SWang Shilong else 3406e8117c26SWang Shilong ret = btrfsic_submit_bh(WRITE_SYNC, bh); 34074eedeb75SHisashi Hifumi if (ret) 3408a512bbf8SYan Zheng errors++; 3409a512bbf8SYan Zheng } 3410a512bbf8SYan Zheng return errors < i ? 0 : -1; 3411a512bbf8SYan Zheng } 3412a512bbf8SYan Zheng 3413387125fcSChris Mason /* 3414387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3415387125fcSChris Mason * for the barrier when it is done 3416387125fcSChris Mason */ 34174246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3418387125fcSChris Mason { 3419387125fcSChris Mason if (bio->bi_private) 3420387125fcSChris Mason complete(bio->bi_private); 3421387125fcSChris Mason bio_put(bio); 3422387125fcSChris Mason } 3423387125fcSChris Mason 3424387125fcSChris Mason /* 3425387125fcSChris Mason * trigger flushes for one the devices. If you pass wait == 0, the flushes are 3426387125fcSChris Mason * sent down. With wait == 1, it waits for the previous flush. 3427387125fcSChris Mason * 3428387125fcSChris Mason * any device where the flush fails with eopnotsupp are flagged as not-barrier 3429387125fcSChris Mason * capable 3430387125fcSChris Mason */ 3431387125fcSChris Mason static int write_dev_flush(struct btrfs_device *device, int wait) 3432387125fcSChris Mason { 3433387125fcSChris Mason struct bio *bio; 3434387125fcSChris Mason int ret = 0; 3435387125fcSChris Mason 3436387125fcSChris Mason if (device->nobarriers) 3437387125fcSChris Mason return 0; 3438387125fcSChris Mason 3439387125fcSChris Mason if (wait) { 3440387125fcSChris Mason bio = device->flush_bio; 3441387125fcSChris Mason if (!bio) 3442387125fcSChris Mason return 0; 3443387125fcSChris Mason 3444387125fcSChris Mason wait_for_completion(&device->flush_wait); 3445387125fcSChris Mason 34464246a0b6SChristoph Hellwig if (bio->bi_error) { 34474246a0b6SChristoph Hellwig ret = bio->bi_error; 3448442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, 3449442a4f63SStefan Behrens BTRFS_DEV_STAT_FLUSH_ERRS); 3450387125fcSChris Mason } 3451387125fcSChris Mason 3452387125fcSChris Mason /* drop the reference from the wait == 0 run */ 3453387125fcSChris Mason bio_put(bio); 3454387125fcSChris Mason device->flush_bio = NULL; 3455387125fcSChris Mason 3456387125fcSChris Mason return ret; 3457387125fcSChris Mason } 3458387125fcSChris Mason 3459387125fcSChris Mason /* 3460387125fcSChris Mason * one reference for us, and we leave it for the 3461387125fcSChris Mason * caller 3462387125fcSChris Mason */ 34639c017abcSJesper Juhl device->flush_bio = NULL; 34649be3395bSChris Mason bio = btrfs_io_bio_alloc(GFP_NOFS, 0); 3465387125fcSChris Mason if (!bio) 3466387125fcSChris Mason return -ENOMEM; 3467387125fcSChris Mason 3468387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3469387125fcSChris Mason bio->bi_bdev = device->bdev; 3470387125fcSChris Mason init_completion(&device->flush_wait); 3471387125fcSChris Mason bio->bi_private = &device->flush_wait; 3472387125fcSChris Mason device->flush_bio = bio; 3473387125fcSChris Mason 3474387125fcSChris Mason bio_get(bio); 347521adbd5cSStefan Behrens btrfsic_submit_bio(WRITE_FLUSH, bio); 3476387125fcSChris Mason 3477387125fcSChris Mason return 0; 3478387125fcSChris Mason } 3479387125fcSChris Mason 3480387125fcSChris Mason /* 3481387125fcSChris Mason * send an empty flush down to each device in parallel, 3482387125fcSChris Mason * then wait for them 3483387125fcSChris Mason */ 3484387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3485387125fcSChris Mason { 3486387125fcSChris Mason struct list_head *head; 3487387125fcSChris Mason struct btrfs_device *dev; 34885af3e8ccSStefan Behrens int errors_send = 0; 34895af3e8ccSStefan Behrens int errors_wait = 0; 3490387125fcSChris Mason int ret; 3491387125fcSChris Mason 3492387125fcSChris Mason /* send down all the barriers */ 3493387125fcSChris Mason head = &info->fs_devices->devices; 3494387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3495f88ba6a2SHidetoshi Seto if (dev->missing) 3496f88ba6a2SHidetoshi Seto continue; 3497387125fcSChris Mason if (!dev->bdev) { 34985af3e8ccSStefan Behrens errors_send++; 3499387125fcSChris Mason continue; 3500387125fcSChris Mason } 3501387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3502387125fcSChris Mason continue; 3503387125fcSChris Mason 3504387125fcSChris Mason ret = write_dev_flush(dev, 0); 3505387125fcSChris Mason if (ret) 35065af3e8ccSStefan Behrens errors_send++; 3507387125fcSChris Mason } 3508387125fcSChris Mason 3509387125fcSChris Mason /* wait for all the barriers */ 3510387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3511f88ba6a2SHidetoshi Seto if (dev->missing) 3512f88ba6a2SHidetoshi Seto continue; 3513387125fcSChris Mason if (!dev->bdev) { 35145af3e8ccSStefan Behrens errors_wait++; 3515387125fcSChris Mason continue; 3516387125fcSChris Mason } 3517387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3518387125fcSChris Mason continue; 3519387125fcSChris Mason 3520387125fcSChris Mason ret = write_dev_flush(dev, 1); 3521387125fcSChris Mason if (ret) 35225af3e8ccSStefan Behrens errors_wait++; 3523387125fcSChris Mason } 35245af3e8ccSStefan Behrens if (errors_send > info->num_tolerated_disk_barrier_failures || 35255af3e8ccSStefan Behrens errors_wait > info->num_tolerated_disk_barrier_failures) 3526387125fcSChris Mason return -EIO; 3527387125fcSChris Mason return 0; 3528387125fcSChris Mason } 3529387125fcSChris Mason 3530943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3531943c6e99SZhao Lei { 35328789f4feSZhao Lei int raid_type; 35338789f4feSZhao Lei int min_tolerated = INT_MAX; 3534943c6e99SZhao Lei 35358789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 35368789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 35378789f4feSZhao Lei min_tolerated = min(min_tolerated, 35388789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 35398789f4feSZhao Lei tolerated_failures); 3540943c6e99SZhao Lei 35418789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 35428789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 35438789f4feSZhao Lei continue; 35448789f4feSZhao Lei if (!(flags & btrfs_raid_group[raid_type])) 35458789f4feSZhao Lei continue; 35468789f4feSZhao Lei min_tolerated = min(min_tolerated, 35478789f4feSZhao Lei btrfs_raid_array[raid_type]. 35488789f4feSZhao Lei tolerated_failures); 35498789f4feSZhao Lei } 3550943c6e99SZhao Lei 35518789f4feSZhao Lei if (min_tolerated == INT_MAX) { 35528789f4feSZhao Lei pr_warn("BTRFS: unknown raid flag: %llu\n", flags); 35538789f4feSZhao Lei min_tolerated = 0; 35548789f4feSZhao Lei } 35558789f4feSZhao Lei 35568789f4feSZhao Lei return min_tolerated; 3557943c6e99SZhao Lei } 3558943c6e99SZhao Lei 35595af3e8ccSStefan Behrens int btrfs_calc_num_tolerated_disk_barrier_failures( 35605af3e8ccSStefan Behrens struct btrfs_fs_info *fs_info) 35615af3e8ccSStefan Behrens { 35625af3e8ccSStefan Behrens struct btrfs_ioctl_space_info space; 35635af3e8ccSStefan Behrens struct btrfs_space_info *sinfo; 35645af3e8ccSStefan Behrens u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 35655af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_SYSTEM, 35665af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_METADATA, 35675af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 35685af3e8ccSStefan Behrens int i; 35695af3e8ccSStefan Behrens int c; 35705af3e8ccSStefan Behrens int num_tolerated_disk_barrier_failures = 35715af3e8ccSStefan Behrens (int)fs_info->fs_devices->num_devices; 35725af3e8ccSStefan Behrens 35732c458045SZhao Lei for (i = 0; i < ARRAY_SIZE(types); i++) { 35745af3e8ccSStefan Behrens struct btrfs_space_info *tmp; 35755af3e8ccSStefan Behrens 35765af3e8ccSStefan Behrens sinfo = NULL; 35775af3e8ccSStefan Behrens rcu_read_lock(); 35785af3e8ccSStefan Behrens list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 35795af3e8ccSStefan Behrens if (tmp->flags == types[i]) { 35805af3e8ccSStefan Behrens sinfo = tmp; 35815af3e8ccSStefan Behrens break; 35825af3e8ccSStefan Behrens } 35835af3e8ccSStefan Behrens } 35845af3e8ccSStefan Behrens rcu_read_unlock(); 35855af3e8ccSStefan Behrens 35865af3e8ccSStefan Behrens if (!sinfo) 35875af3e8ccSStefan Behrens continue; 35885af3e8ccSStefan Behrens 35895af3e8ccSStefan Behrens down_read(&sinfo->groups_sem); 35905af3e8ccSStefan Behrens for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 35915af3e8ccSStefan Behrens u64 flags; 35925af3e8ccSStefan Behrens 35932c458045SZhao Lei if (list_empty(&sinfo->block_groups[c])) 35942c458045SZhao Lei continue; 35952c458045SZhao Lei 35962c458045SZhao Lei btrfs_get_block_group_info(&sinfo->block_groups[c], 35972c458045SZhao Lei &space); 35982c458045SZhao Lei if (space.total_bytes == 0 || space.used_bytes == 0) 35995af3e8ccSStefan Behrens continue; 36005af3e8ccSStefan Behrens flags = space.flags; 3601943c6e99SZhao Lei 3602943c6e99SZhao Lei num_tolerated_disk_barrier_failures = min( 3603943c6e99SZhao Lei num_tolerated_disk_barrier_failures, 3604943c6e99SZhao Lei btrfs_get_num_tolerated_disk_barrier_failures( 3605943c6e99SZhao Lei flags)); 36065af3e8ccSStefan Behrens } 36075af3e8ccSStefan Behrens up_read(&sinfo->groups_sem); 36085af3e8ccSStefan Behrens } 36095af3e8ccSStefan Behrens 36105af3e8ccSStefan Behrens return num_tolerated_disk_barrier_failures; 36115af3e8ccSStefan Behrens } 36125af3e8ccSStefan Behrens 361348a3b636SEric Sandeen static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3614f2984462SChris Mason { 3615e5e9a520SChris Mason struct list_head *head; 3616f2984462SChris Mason struct btrfs_device *dev; 3617a061fc8dSChris Mason struct btrfs_super_block *sb; 3618f2984462SChris Mason struct btrfs_dev_item *dev_item; 3619f2984462SChris Mason int ret; 3620f2984462SChris Mason int do_barriers; 3621a236aed1SChris Mason int max_errors; 3622a236aed1SChris Mason int total_errors = 0; 3623a061fc8dSChris Mason u64 flags; 3624f2984462SChris Mason 3625f2984462SChris Mason do_barriers = !btrfs_test_opt(root, NOBARRIER); 3626af31f5e5SChris Mason backup_super_roots(root->fs_info); 3627f2984462SChris Mason 36286c41761fSDavid Sterba sb = root->fs_info->super_for_commit; 3629a061fc8dSChris Mason dev_item = &sb->dev_item; 3630e5e9a520SChris Mason 3631174ba509SChris Mason mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3632e5e9a520SChris Mason head = &root->fs_info->fs_devices->devices; 3633d7306801SFilipe David Borba Manana max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3634387125fcSChris Mason 36355af3e8ccSStefan Behrens if (do_barriers) { 36365af3e8ccSStefan Behrens ret = barrier_all_devices(root->fs_info); 36375af3e8ccSStefan Behrens if (ret) { 36385af3e8ccSStefan Behrens mutex_unlock( 36395af3e8ccSStefan Behrens &root->fs_info->fs_devices->device_list_mutex); 3640a4553fefSAnand Jain btrfs_std_error(root->fs_info, ret, 36415af3e8ccSStefan Behrens "errors while submitting device barriers."); 36425af3e8ccSStefan Behrens return ret; 36435af3e8ccSStefan Behrens } 36445af3e8ccSStefan Behrens } 3645387125fcSChris Mason 36461f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3647dfe25020SChris Mason if (!dev->bdev) { 3648dfe25020SChris Mason total_errors++; 3649dfe25020SChris Mason continue; 3650dfe25020SChris Mason } 36512b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3652dfe25020SChris Mason continue; 3653dfe25020SChris Mason 36542b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3655a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3656a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 36577df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3658935e5cc9SMiao Xie dev->commit_total_bytes); 3659ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3660ce7213c7SMiao Xie dev->commit_bytes_used); 3661a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3662a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3663a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3664a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 36652b82032cSYan Zheng memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3666a512bbf8SYan Zheng 3667a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3668a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3669f2984462SChris Mason 3670a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3671a236aed1SChris Mason if (ret) 3672a236aed1SChris Mason total_errors++; 3673f2984462SChris Mason } 3674a236aed1SChris Mason if (total_errors > max_errors) { 3675efe120a0SFrank Holton btrfs_err(root->fs_info, "%d errors while writing supers", 3676d397712bSChris Mason total_errors); 3677a724b436SStefan Behrens mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 367879787eaaSJeff Mahoney 36799d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 3680a4553fefSAnand Jain btrfs_std_error(root->fs_info, -EIO, 36819d565ba4SStefan Behrens "%d errors while writing supers", total_errors); 36829d565ba4SStefan Behrens return -EIO; 3683a236aed1SChris Mason } 3684f2984462SChris Mason 3685a512bbf8SYan Zheng total_errors = 0; 36861f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3687dfe25020SChris Mason if (!dev->bdev) 3688dfe25020SChris Mason continue; 36892b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3690dfe25020SChris Mason continue; 3691dfe25020SChris Mason 3692a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3693a512bbf8SYan Zheng if (ret) 36941259ab75SChris Mason total_errors++; 3695f2984462SChris Mason } 3696174ba509SChris Mason mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3697a236aed1SChris Mason if (total_errors > max_errors) { 3698a4553fefSAnand Jain btrfs_std_error(root->fs_info, -EIO, 369979787eaaSJeff Mahoney "%d errors while writing supers", total_errors); 370079787eaaSJeff Mahoney return -EIO; 3701a236aed1SChris Mason } 3702f2984462SChris Mason return 0; 3703f2984462SChris Mason } 3704f2984462SChris Mason 3705a512bbf8SYan Zheng int write_ctree_super(struct btrfs_trans_handle *trans, 3706a512bbf8SYan Zheng struct btrfs_root *root, int max_mirrors) 3707cfaa7295SChris Mason { 3708f570e757SRashika return write_all_supers(root, max_mirrors); 3709cfaa7295SChris Mason } 3710cfaa7295SChris Mason 3711cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3712cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3713cb517eabSMiao Xie struct btrfs_root *root) 37142619ba1fSChris Mason { 37154df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 37162619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 37172619ba1fSChris Mason (unsigned long)root->root_key.objectid); 37184df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 371976dda93cSYan, Zheng 372076dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 372176dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 372276dda93cSYan, Zheng 37231a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 37243321719eSLiu Bo btrfs_free_log(NULL, root); 37253321719eSLiu Bo 3726faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3727581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3728faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3729581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 37304df27c4dSYan, Zheng free_fs_root(root); 37314df27c4dSYan, Zheng } 37324df27c4dSYan, Zheng 37334df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 37344df27c4dSYan, Zheng { 373557cdc8dbSDavid Sterba iput(root->ino_cache_inode); 37364df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 37371cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 37381cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 37390ee5dc67SAl Viro if (root->anon_dev) 37400ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 37418257b2dcSMiao Xie if (root->subv_writers) 37428257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 37435f39d397SChris Mason free_extent_buffer(root->node); 37445f39d397SChris Mason free_extent_buffer(root->commit_root); 3745581bb050SLi Zefan kfree(root->free_ino_ctl); 3746581bb050SLi Zefan kfree(root->free_ino_pinned); 374758176a96SJosef Bacik kfree(root->name); 3748b0feb9d9SMiao Xie btrfs_put_fs_root(root); 37492619ba1fSChris Mason } 37502619ba1fSChris Mason 3751cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3752cb517eabSMiao Xie { 3753cb517eabSMiao Xie free_fs_root(root); 3754eb60ceacSChris Mason } 3755eb60ceacSChris Mason 3756c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3757c146afadSYan Zheng { 3758c146afadSYan Zheng u64 root_objectid = 0; 3759c146afadSYan Zheng struct btrfs_root *gang[8]; 376065d33fd7SQu Wenruo int i = 0; 376165d33fd7SQu Wenruo int err = 0; 376265d33fd7SQu Wenruo unsigned int ret = 0; 376365d33fd7SQu Wenruo int index; 3764c146afadSYan Zheng 3765c146afadSYan Zheng while (1) { 376665d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3767c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3768c146afadSYan Zheng (void **)gang, root_objectid, 3769c146afadSYan Zheng ARRAY_SIZE(gang)); 377065d33fd7SQu Wenruo if (!ret) { 377165d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3772c146afadSYan Zheng break; 377365d33fd7SQu Wenruo } 37745d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 377566b4ffd1SJosef Bacik 377665d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 377765d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 377865d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 377965d33fd7SQu Wenruo gang[i] = NULL; 378065d33fd7SQu Wenruo continue; 378165d33fd7SQu Wenruo } 378265d33fd7SQu Wenruo /* grab all the search result for later use */ 378365d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 378465d33fd7SQu Wenruo } 378565d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 378665d33fd7SQu Wenruo 378765d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 378865d33fd7SQu Wenruo if (!gang[i]) 378965d33fd7SQu Wenruo continue; 3790c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 379166b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 379266b4ffd1SJosef Bacik if (err) 379365d33fd7SQu Wenruo break; 379465d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3795c146afadSYan Zheng } 3796c146afadSYan Zheng root_objectid++; 3797c146afadSYan Zheng } 379865d33fd7SQu Wenruo 379965d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 380065d33fd7SQu Wenruo for (; i < ret; i++) { 380165d33fd7SQu Wenruo if (gang[i]) 380265d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 380365d33fd7SQu Wenruo } 380465d33fd7SQu Wenruo return err; 3805c146afadSYan Zheng } 3806c146afadSYan Zheng 3807c146afadSYan Zheng int btrfs_commit_super(struct btrfs_root *root) 3808c146afadSYan Zheng { 3809c146afadSYan Zheng struct btrfs_trans_handle *trans; 3810c146afadSYan Zheng 3811c146afadSYan Zheng mutex_lock(&root->fs_info->cleaner_mutex); 381224bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 3813c146afadSYan Zheng mutex_unlock(&root->fs_info->cleaner_mutex); 38149d1a2a3aSDavid Sterba wake_up_process(root->fs_info->cleaner_kthread); 3815c71bf099SYan, Zheng 3816c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 3817c71bf099SYan, Zheng down_write(&root->fs_info->cleanup_work_sem); 3818c71bf099SYan, Zheng up_write(&root->fs_info->cleanup_work_sem); 3819c71bf099SYan, Zheng 38207a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 38213612b495STsutomu Itoh if (IS_ERR(trans)) 38223612b495STsutomu Itoh return PTR_ERR(trans); 3823d52c1bccSLiu Bo return btrfs_commit_transaction(trans, root); 3824c146afadSYan Zheng } 3825c146afadSYan Zheng 38263abdbd78SDavid Sterba void close_ctree(struct btrfs_root *root) 3827eb60ceacSChris Mason { 38280f7d52f4SChris Mason struct btrfs_fs_info *fs_info = root->fs_info; 3829c146afadSYan Zheng int ret; 3830e089f05cSChris Mason 3831facda1e7SChris Mason fs_info->closing = 1; 3832a2135011SChris Mason smp_mb(); 3833a2135011SChris Mason 38347343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 38357343dd61SJustin Maggard btrfs_qgroup_wait_for_completion(fs_info); 38367343dd61SJustin Maggard 3837803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3838803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3839803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3840803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3841803b2f54SStefan Behrens 3842837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3843aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3844837d5b6eSIlya Dryomov 38458dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 38468dabb742SStefan Behrens 3847aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 38484cb5300bSChris Mason 38494cb5300bSChris Mason /* wait for any defraggers to finish */ 38504cb5300bSChris Mason wait_event(fs_info->transaction_wait, 38514cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 38524cb5300bSChris Mason 38534cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 385426176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 38554cb5300bSChris Mason 385621c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 385721c7e756SMiao Xie 3858c146afadSYan Zheng if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3859e44163e1SJeff Mahoney /* 3860e44163e1SJeff Mahoney * If the cleaner thread is stopped and there are 3861e44163e1SJeff Mahoney * block groups queued for removal, the deletion will be 3862e44163e1SJeff Mahoney * skipped when we quit the cleaner thread. 3863e44163e1SJeff Mahoney */ 3864e44163e1SJeff Mahoney btrfs_delete_unused_bgs(root->fs_info); 3865e44163e1SJeff Mahoney 3866c146afadSYan Zheng ret = btrfs_commit_super(root); 3867d397712bSChris Mason if (ret) 386804892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 3869c146afadSYan Zheng } 3870ed2ff2cbSChris Mason 387187533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 387268ce9682SStefan Behrens btrfs_error_commit_super(root); 3873acce952bSliubo 3874e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3875e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 38768929ecfaSYan, Zheng 3877f25784b3SYan Zheng fs_info->closing = 2; 3878f25784b3SYan Zheng smp_mb(); 3879f25784b3SYan Zheng 388004892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 3881bcef60f2SArne Jansen 3882963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 388304892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 3884963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3885b0c68f8bSChris Mason } 388631153d81SYan Zheng 38876618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 3888b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 38895ac1d209SJeff Mahoney 3890faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3891d10c5f31SChris Mason 38921a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 38931a4319ccSLiu Bo 38942b1360daSJosef Bacik btrfs_free_block_groups(fs_info); 38952b1360daSJosef Bacik 3896de348ee0SWang Shilong /* 3897de348ee0SWang Shilong * we must make sure there is not any read request to 3898de348ee0SWang Shilong * submit after we stopping all workers. 3899de348ee0SWang Shilong */ 3900de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 390196192499SJosef Bacik btrfs_stop_all_workers(fs_info); 390296192499SJosef Bacik 390347ab2a6cSJosef Bacik fs_info->open = 0; 390413e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 39059ad6b7bcSChris Mason 390613e6c37bSJosef Bacik iput(fs_info->btree_inode); 3907d6bfde87SChris Mason 390821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 390921adbd5cSStefan Behrens if (btrfs_test_opt(root, CHECK_INTEGRITY)) 391021adbd5cSStefan Behrens btrfsic_unmount(root, fs_info->fs_devices); 391121adbd5cSStefan Behrens #endif 391221adbd5cSStefan Behrens 3913dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 39140b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3915b248a415SChris Mason 3916e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3917963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3918c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 391904160088SChris Mason bdi_destroy(&fs_info->bdi); 392076dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 39210b86a832SChris Mason 392253b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 392353b381b3SDavid Woodhouse 3924cdfb080eSChris Mason __btrfs_free_block_rsv(root->orphan_block_rsv); 39251cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 392604216820SFilipe Manana 392704216820SFilipe Manana lock_chunks(root); 392804216820SFilipe Manana while (!list_empty(&fs_info->pinned_chunks)) { 392904216820SFilipe Manana struct extent_map *em; 393004216820SFilipe Manana 393104216820SFilipe Manana em = list_first_entry(&fs_info->pinned_chunks, 393204216820SFilipe Manana struct extent_map, list); 393304216820SFilipe Manana list_del_init(&em->list); 393404216820SFilipe Manana free_extent_map(em); 393504216820SFilipe Manana } 393604216820SFilipe Manana unlock_chunks(root); 3937eb60ceacSChris Mason } 3938eb60ceacSChris Mason 3939b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3940b9fab919SChris Mason int atomic) 3941ccd467d6SChris Mason { 39421259ab75SChris Mason int ret; 3943727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 39441259ab75SChris Mason 39450b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 39461259ab75SChris Mason if (!ret) 39471259ab75SChris Mason return ret; 39481259ab75SChris Mason 39491259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3950b9fab919SChris Mason parent_transid, atomic); 3951b9fab919SChris Mason if (ret == -EAGAIN) 3952b9fab919SChris Mason return ret; 39531259ab75SChris Mason return !ret; 39545f39d397SChris Mason } 39556702ed49SChris Mason 39565f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 39575f39d397SChris Mason { 395806ea65a3SJosef Bacik struct btrfs_root *root; 39595f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3960b9473439SChris Mason int was_dirty; 3961b4ce94deSChris Mason 396206ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 396306ea65a3SJosef Bacik /* 396406ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 396506ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 396606ea65a3SJosef Bacik * outside of the sanity tests. 396706ea65a3SJosef Bacik */ 396806ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 396906ea65a3SJosef Bacik return; 397006ea65a3SJosef Bacik #endif 397106ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3972b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 397331b1a2bdSJulia Lawall if (transid != root->fs_info->generation) 397431b1a2bdSJulia Lawall WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3975d397712bSChris Mason "found %llu running %llu\n", 3976c1c9ff7cSGeert Uytterhoeven buf->start, transid, root->fs_info->generation); 39770b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3978e2d84521SMiao Xie if (!was_dirty) 3979e2d84521SMiao Xie __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 3980e2d84521SMiao Xie buf->len, 3981e2d84521SMiao Xie root->fs_info->dirty_metadata_batch); 39821f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 39831f21ef0aSFilipe Manana if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { 39841f21ef0aSFilipe Manana btrfs_print_leaf(root, buf); 39851f21ef0aSFilipe Manana ASSERT(0); 39861f21ef0aSFilipe Manana } 39871f21ef0aSFilipe Manana #endif 3988eb60ceacSChris Mason } 3989eb60ceacSChris Mason 3990b53d3f5dSLiu Bo static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 3991b53d3f5dSLiu Bo int flush_delayed) 399235b7e476SChris Mason { 3993188de649SChris Mason /* 3994188de649SChris Mason * looks as though older kernels can get into trouble with 3995188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3996188de649SChris Mason */ 3997e2d84521SMiao Xie int ret; 3998d6bfde87SChris Mason 39996933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4000d6bfde87SChris Mason return; 4001d6bfde87SChris Mason 4002b53d3f5dSLiu Bo if (flush_delayed) 400316cdcec7SMiao Xie btrfs_balance_delayed_items(root); 400416cdcec7SMiao Xie 4005e2d84521SMiao Xie ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 4006e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 4007e2d84521SMiao Xie if (ret > 0) { 4008d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited( 4009d0e1d66bSNamjae Jeon root->fs_info->btree_inode->i_mapping); 401016cdcec7SMiao Xie } 401116cdcec7SMiao Xie } 401216cdcec7SMiao Xie 4013b53d3f5dSLiu Bo void btrfs_btree_balance_dirty(struct btrfs_root *root) 401416cdcec7SMiao Xie { 4015b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 1); 401635b7e476SChris Mason } 4017b53d3f5dSLiu Bo 4018b53d3f5dSLiu Bo void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 4019b53d3f5dSLiu Bo { 4020b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 0); 4021d6bfde87SChris Mason } 40226b80053dSChris Mason 4023ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 40246b80053dSChris Mason { 4025727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 40260b32f4bbSJosef Bacik return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 40276b80053dSChris Mason } 40280da5468fSChris Mason 4029fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 4030acce952bSliubo int read_only) 4031acce952bSliubo { 4032c926093eSDavid Sterba struct btrfs_super_block *sb = fs_info->super_copy; 4033319e4d06SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 4034319e4d06SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 4035c926093eSDavid Sterba int ret = 0; 4036c926093eSDavid Sterba 4037319e4d06SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 4038319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: no valid FS found\n"); 4039319e4d06SQu Wenruo ret = -EINVAL; 4040319e4d06SQu Wenruo } 4041319e4d06SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) 4042319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n", 4043319e4d06SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 404421e7626bSDavid Sterba if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 404521e7626bSDavid Sterba printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n", 404621e7626bSDavid Sterba btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 4047c926093eSDavid Sterba ret = -EINVAL; 4048c926093eSDavid Sterba } 404921e7626bSDavid Sterba if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 405021e7626bSDavid Sterba printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n", 405121e7626bSDavid Sterba btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 4052c926093eSDavid Sterba ret = -EINVAL; 4053c926093eSDavid Sterba } 405421e7626bSDavid Sterba if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 405521e7626bSDavid Sterba printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n", 405621e7626bSDavid Sterba btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 4057c926093eSDavid Sterba ret = -EINVAL; 4058c926093eSDavid Sterba } 4059c926093eSDavid Sterba 40601104a885SDavid Sterba /* 4061319e4d06SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 4062319e4d06SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 40631104a885SDavid Sterba */ 4064319e4d06SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 4065319e4d06SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 4066319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize); 406775d6ad38SDavid Sterba ret = -EINVAL; 406875d6ad38SDavid Sterba } 4069319e4d06SQu Wenruo /* Only PAGE SIZE is supported yet */ 4070319e4d06SQu Wenruo if (sectorsize != PAGE_CACHE_SIZE) { 4071319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n", 4072319e4d06SQu Wenruo sectorsize, PAGE_CACHE_SIZE); 4073319e4d06SQu Wenruo ret = -EINVAL; 4074319e4d06SQu Wenruo } 4075319e4d06SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 4076319e4d06SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 4077319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize); 4078319e4d06SQu Wenruo ret = -EINVAL; 4079319e4d06SQu Wenruo } 4080319e4d06SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 4081319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n", 4082319e4d06SQu Wenruo le32_to_cpu(sb->__unused_leafsize), 4083319e4d06SQu Wenruo nodesize); 4084319e4d06SQu Wenruo ret = -EINVAL; 4085319e4d06SQu Wenruo } 4086319e4d06SQu Wenruo 4087319e4d06SQu Wenruo /* Root alignment check */ 4088319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 4089319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n", 4090319e4d06SQu Wenruo btrfs_super_root(sb)); 4091319e4d06SQu Wenruo ret = -EINVAL; 4092319e4d06SQu Wenruo } 4093319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 4094319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n", 4095319e4d06SQu Wenruo btrfs_super_chunk_root(sb)); 4096319e4d06SQu Wenruo ret = -EINVAL; 4097319e4d06SQu Wenruo } 4098319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 4099319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n", 4100319e4d06SQu Wenruo btrfs_super_log_root(sb)); 410175d6ad38SDavid Sterba ret = -EINVAL; 410275d6ad38SDavid Sterba } 410375d6ad38SDavid Sterba 4104c926093eSDavid Sterba if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) { 4105c926093eSDavid Sterba printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n", 4106c926093eSDavid Sterba fs_info->fsid, sb->dev_item.fsid); 4107c926093eSDavid Sterba ret = -EINVAL; 4108c926093eSDavid Sterba } 4109c926093eSDavid Sterba 4110c926093eSDavid Sterba /* 4111c926093eSDavid Sterba * Hint to catch really bogus numbers, bitflips or so, more exact checks are 4112c926093eSDavid Sterba * done later 4113c926093eSDavid Sterba */ 411421e7626bSDavid Sterba if (btrfs_super_num_devices(sb) > (1UL << 31)) 4115c926093eSDavid Sterba printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n", 411621e7626bSDavid Sterba btrfs_super_num_devices(sb)); 411775d6ad38SDavid Sterba if (btrfs_super_num_devices(sb) == 0) { 411875d6ad38SDavid Sterba printk(KERN_ERR "BTRFS: number of devices is 0\n"); 411975d6ad38SDavid Sterba ret = -EINVAL; 412075d6ad38SDavid Sterba } 4121c926093eSDavid Sterba 412221e7626bSDavid Sterba if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { 4123c926093eSDavid Sterba printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n", 412421e7626bSDavid Sterba btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 4125c926093eSDavid Sterba ret = -EINVAL; 4126c926093eSDavid Sterba } 4127c926093eSDavid Sterba 4128c926093eSDavid Sterba /* 4129ce7fca5fSDavid Sterba * Obvious sys_chunk_array corruptions, it must hold at least one key 4130ce7fca5fSDavid Sterba * and one chunk 4131ce7fca5fSDavid Sterba */ 4132ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 4133ce7fca5fSDavid Sterba printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n", 4134ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4135ce7fca5fSDavid Sterba BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 4136ce7fca5fSDavid Sterba ret = -EINVAL; 4137ce7fca5fSDavid Sterba } 4138ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 4139ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)) { 4140d2207129SFabian Frederick printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n", 4141ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4142ce7fca5fSDavid Sterba sizeof(struct btrfs_disk_key) 4143ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)); 4144ce7fca5fSDavid Sterba ret = -EINVAL; 4145ce7fca5fSDavid Sterba } 4146ce7fca5fSDavid Sterba 4147ce7fca5fSDavid Sterba /* 4148c926093eSDavid Sterba * The generation is a global counter, we'll trust it more than the others 4149c926093eSDavid Sterba * but it's still possible that it's the one that's wrong. 4150c926093eSDavid Sterba */ 415121e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 4152c926093eSDavid Sterba printk(KERN_WARNING 4153c926093eSDavid Sterba "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n", 415421e7626bSDavid Sterba btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb)); 415521e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 415621e7626bSDavid Sterba && btrfs_super_cache_generation(sb) != (u64)-1) 4157c926093eSDavid Sterba printk(KERN_WARNING 4158c926093eSDavid Sterba "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n", 415921e7626bSDavid Sterba btrfs_super_generation(sb), btrfs_super_cache_generation(sb)); 4160c926093eSDavid Sterba 4161c926093eSDavid Sterba return ret; 4162fcd1f065SDavid Sterba } 4163fcd1f065SDavid Sterba 416448a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root) 4165acce952bSliubo { 4166acce952bSliubo mutex_lock(&root->fs_info->cleaner_mutex); 4167acce952bSliubo btrfs_run_delayed_iputs(root); 4168acce952bSliubo mutex_unlock(&root->fs_info->cleaner_mutex); 4169acce952bSliubo 4170acce952bSliubo down_write(&root->fs_info->cleanup_work_sem); 4171acce952bSliubo up_write(&root->fs_info->cleanup_work_sem); 4172acce952bSliubo 4173acce952bSliubo /* cleanup FS via transaction */ 4174acce952bSliubo btrfs_cleanup_transaction(root); 4175acce952bSliubo } 4176acce952bSliubo 4177143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4178acce952bSliubo { 4179acce952bSliubo struct btrfs_ordered_extent *ordered; 4180acce952bSliubo 4181199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4182779880efSJosef Bacik /* 4183779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4184779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4185779880efSJosef Bacik */ 4186199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4187779880efSJosef Bacik root_extent_list) 4188779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4189199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4190199c2a9cSMiao Xie } 4191199c2a9cSMiao Xie 4192199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4193199c2a9cSMiao Xie { 4194199c2a9cSMiao Xie struct btrfs_root *root; 4195199c2a9cSMiao Xie struct list_head splice; 4196199c2a9cSMiao Xie 4197199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4198199c2a9cSMiao Xie 4199199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4200199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4201199c2a9cSMiao Xie while (!list_empty(&splice)) { 4202199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4203199c2a9cSMiao Xie ordered_root); 42041de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 42051de2cfdeSJosef Bacik &fs_info->ordered_roots); 4206199c2a9cSMiao Xie 42072a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4208199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4209199c2a9cSMiao Xie 42102a85d9caSLiu Bo cond_resched(); 42112a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4212199c2a9cSMiao Xie } 4213199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 4214acce952bSliubo } 4215acce952bSliubo 421635a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 4217acce952bSliubo struct btrfs_root *root) 4218acce952bSliubo { 4219acce952bSliubo struct rb_node *node; 4220acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4221acce952bSliubo struct btrfs_delayed_ref_node *ref; 4222acce952bSliubo int ret = 0; 4223acce952bSliubo 4224acce952bSliubo delayed_refs = &trans->delayed_refs; 4225acce952bSliubo 4226acce952bSliubo spin_lock(&delayed_refs->lock); 4227d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4228cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4229efe120a0SFrank Holton btrfs_info(root->fs_info, "delayed_refs has NO entry"); 4230acce952bSliubo return ret; 4231acce952bSliubo } 4232acce952bSliubo 4233d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 4234d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 4235c6fc2454SQu Wenruo struct btrfs_delayed_ref_node *tmp; 4236e78417d1SJosef Bacik bool pin_bytes = false; 4237acce952bSliubo 4238d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4239d7df2c79SJosef Bacik href_node); 4240b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 4241d7df2c79SJosef Bacik atomic_inc(&head->node.refs); 424279787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 4243b939d1abSJosef Bacik 4244acce952bSliubo mutex_lock(&head->mutex); 4245b939d1abSJosef Bacik mutex_unlock(&head->mutex); 4246d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 4247e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 4248b939d1abSJosef Bacik continue; 4249b939d1abSJosef Bacik } 4250d7df2c79SJosef Bacik spin_lock(&head->lock); 4251c6fc2454SQu Wenruo list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, 4252c6fc2454SQu Wenruo list) { 4253d7df2c79SJosef Bacik ref->in_tree = 0; 4254c6fc2454SQu Wenruo list_del(&ref->list); 4255d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4256d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4257d7df2c79SJosef Bacik } 425854067ae9SJosef Bacik if (head->must_insert_reserved) 4259e78417d1SJosef Bacik pin_bytes = true; 426078a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4261acce952bSliubo delayed_refs->num_heads--; 4262d7df2c79SJosef Bacik if (head->processing == 0) 4263acce952bSliubo delayed_refs->num_heads_ready--; 4264d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4265d7df2c79SJosef Bacik head->node.in_tree = 0; 4266c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 4267d7df2c79SJosef Bacik spin_unlock(&head->lock); 4268acce952bSliubo spin_unlock(&delayed_refs->lock); 4269e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4270acce952bSliubo 4271d7df2c79SJosef Bacik if (pin_bytes) 4272d7df2c79SJosef Bacik btrfs_pin_extent(root, head->node.bytenr, 4273d7df2c79SJosef Bacik head->node.num_bytes, 1); 4274d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 4275acce952bSliubo cond_resched(); 4276acce952bSliubo spin_lock(&delayed_refs->lock); 4277acce952bSliubo } 4278acce952bSliubo 4279acce952bSliubo spin_unlock(&delayed_refs->lock); 4280acce952bSliubo 4281acce952bSliubo return ret; 4282acce952bSliubo } 4283acce952bSliubo 4284143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4285acce952bSliubo { 4286acce952bSliubo struct btrfs_inode *btrfs_inode; 4287acce952bSliubo struct list_head splice; 4288acce952bSliubo 4289acce952bSliubo INIT_LIST_HEAD(&splice); 4290acce952bSliubo 4291eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4292eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4293acce952bSliubo 4294acce952bSliubo while (!list_empty(&splice)) { 4295eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4296acce952bSliubo delalloc_inodes); 4297acce952bSliubo 4298acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 4299df0af1a5SMiao Xie clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 4300df0af1a5SMiao Xie &btrfs_inode->runtime_flags); 4301eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4302acce952bSliubo 4303acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 4304b216cbfbSMiao Xie 4305eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4306acce952bSliubo } 4307acce952bSliubo 4308eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4309eb73c1b7SMiao Xie } 4310eb73c1b7SMiao Xie 4311eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4312eb73c1b7SMiao Xie { 4313eb73c1b7SMiao Xie struct btrfs_root *root; 4314eb73c1b7SMiao Xie struct list_head splice; 4315eb73c1b7SMiao Xie 4316eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4317eb73c1b7SMiao Xie 4318eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4319eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4320eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4321eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4322eb73c1b7SMiao Xie delalloc_root); 4323eb73c1b7SMiao Xie list_del_init(&root->delalloc_root); 4324eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 4325eb73c1b7SMiao Xie BUG_ON(!root); 4326eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4327eb73c1b7SMiao Xie 4328eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 4329eb73c1b7SMiao Xie btrfs_put_fs_root(root); 4330eb73c1b7SMiao Xie 4331eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4332eb73c1b7SMiao Xie } 4333eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4334acce952bSliubo } 4335acce952bSliubo 4336acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 4337acce952bSliubo struct extent_io_tree *dirty_pages, 4338acce952bSliubo int mark) 4339acce952bSliubo { 4340acce952bSliubo int ret; 4341acce952bSliubo struct extent_buffer *eb; 4342acce952bSliubo u64 start = 0; 4343acce952bSliubo u64 end; 4344acce952bSliubo 4345acce952bSliubo while (1) { 4346acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4347e6138876SJosef Bacik mark, NULL); 4348acce952bSliubo if (ret) 4349acce952bSliubo break; 4350acce952bSliubo 4351acce952bSliubo clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); 4352acce952bSliubo while (start <= end) { 435301d58472SDaniel Dressler eb = btrfs_find_tree_block(root->fs_info, start); 4354707e8a07SDavid Sterba start += root->nodesize; 4355fd8b2b61SJosef Bacik if (!eb) 4356acce952bSliubo continue; 4357fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4358acce952bSliubo 4359fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4360fd8b2b61SJosef Bacik &eb->bflags)) 4361fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4362fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4363acce952bSliubo } 4364acce952bSliubo } 4365acce952bSliubo 4366acce952bSliubo return ret; 4367acce952bSliubo } 4368acce952bSliubo 4369acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 4370acce952bSliubo struct extent_io_tree *pinned_extents) 4371acce952bSliubo { 4372acce952bSliubo struct extent_io_tree *unpin; 4373acce952bSliubo u64 start; 4374acce952bSliubo u64 end; 4375acce952bSliubo int ret; 4376ed0eaa14SLiu Bo bool loop = true; 4377acce952bSliubo 4378acce952bSliubo unpin = pinned_extents; 4379ed0eaa14SLiu Bo again: 4380acce952bSliubo while (1) { 4381acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4382e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4383acce952bSliubo if (ret) 4384acce952bSliubo break; 4385acce952bSliubo 4386acce952bSliubo clear_extent_dirty(unpin, start, end, GFP_NOFS); 4387acce952bSliubo btrfs_error_unpin_extent_range(root, start, end); 4388acce952bSliubo cond_resched(); 4389acce952bSliubo } 4390acce952bSliubo 4391ed0eaa14SLiu Bo if (loop) { 4392ed0eaa14SLiu Bo if (unpin == &root->fs_info->freed_extents[0]) 4393ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[1]; 4394ed0eaa14SLiu Bo else 4395ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[0]; 4396ed0eaa14SLiu Bo loop = false; 4397ed0eaa14SLiu Bo goto again; 4398ed0eaa14SLiu Bo } 4399ed0eaa14SLiu Bo 4400acce952bSliubo return 0; 4401acce952bSliubo } 4402acce952bSliubo 440349b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 440449b25e05SJeff Mahoney struct btrfs_root *root) 440549b25e05SJeff Mahoney { 440649b25e05SJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, root); 440749b25e05SJeff Mahoney 44084a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 440949b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_blocked_wait); 441049b25e05SJeff Mahoney 44114a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 441249b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_wait); 441349b25e05SJeff Mahoney 441467cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 441567cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 441667cde344SMiao Xie 441749b25e05SJeff Mahoney btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 441849b25e05SJeff Mahoney EXTENT_DIRTY); 44196e841e32SLiu Bo btrfs_destroy_pinned_extent(root, 44206e841e32SLiu Bo root->fs_info->pinned_extents); 442149b25e05SJeff Mahoney 44224a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 44234a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 44244a9d8bdeSMiao Xie 442549b25e05SJeff Mahoney /* 442649b25e05SJeff Mahoney memset(cur_trans, 0, sizeof(*cur_trans)); 442749b25e05SJeff Mahoney kmem_cache_free(btrfs_transaction_cachep, cur_trans); 442849b25e05SJeff Mahoney */ 442949b25e05SJeff Mahoney } 443049b25e05SJeff Mahoney 443148a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root) 4432acce952bSliubo { 4433acce952bSliubo struct btrfs_transaction *t; 4434acce952bSliubo 4435acce952bSliubo mutex_lock(&root->fs_info->transaction_kthread_mutex); 4436acce952bSliubo 4437a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4438724e2315SJosef Bacik while (!list_empty(&root->fs_info->trans_list)) { 4439724e2315SJosef Bacik t = list_first_entry(&root->fs_info->trans_list, 4440724e2315SJosef Bacik struct btrfs_transaction, list); 4441724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 4442724e2315SJosef Bacik atomic_inc(&t->use_count); 4443724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4444724e2315SJosef Bacik btrfs_wait_for_commit(root, t->transid); 4445724e2315SJosef Bacik btrfs_put_transaction(t); 4446724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4447724e2315SJosef Bacik continue; 4448724e2315SJosef Bacik } 4449724e2315SJosef Bacik if (t == root->fs_info->running_transaction) { 4450724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 4451724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4452724e2315SJosef Bacik /* 4453724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4454724e2315SJosef Bacik * handle open currently for this transaction. 4455724e2315SJosef Bacik */ 4456724e2315SJosef Bacik wait_event(t->writer_wait, 4457724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4458724e2315SJosef Bacik } else { 4459724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4460724e2315SJosef Bacik } 4461724e2315SJosef Bacik btrfs_cleanup_one_transaction(t, root); 4462724e2315SJosef Bacik 4463724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4464724e2315SJosef Bacik if (t == root->fs_info->running_transaction) 4465ac673879SMiao Xie root->fs_info->running_transaction = NULL; 4466724e2315SJosef Bacik list_del_init(&t->list); 4467a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4468a4abeea4SJosef Bacik 4469724e2315SJosef Bacik btrfs_put_transaction(t); 4470724e2315SJosef Bacik trace_btrfs_transaction_commit(root); 4471724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4472724e2315SJosef Bacik } 4473724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4474199c2a9cSMiao Xie btrfs_destroy_all_ordered_extents(root->fs_info); 447567cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 447667cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 4477724e2315SJosef Bacik btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4478eb73c1b7SMiao Xie btrfs_destroy_all_delalloc_inodes(root->fs_info); 4479acce952bSliubo mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4480acce952bSliubo 4481acce952bSliubo return 0; 4482acce952bSliubo } 4483acce952bSliubo 4484e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 4485ce9adaa5SChris Mason .readpage_end_io_hook = btree_readpage_end_io_hook, 44864bb31e92SArne Jansen .readpage_io_failed_hook = btree_io_failed_hook, 44870b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 4488239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4489239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 44900da5468fSChris Mason }; 4491