16cbd5570SChris Mason /* 26cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 36cbd5570SChris Mason * 46cbd5570SChris Mason * This program is free software; you can redistribute it and/or 56cbd5570SChris Mason * modify it under the terms of the GNU General Public 66cbd5570SChris Mason * License v2 as published by the Free Software Foundation. 76cbd5570SChris Mason * 86cbd5570SChris Mason * This program is distributed in the hope that it will be useful, 96cbd5570SChris Mason * but WITHOUT ANY WARRANTY; without even the implied warranty of 106cbd5570SChris Mason * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 116cbd5570SChris Mason * General Public License for more details. 126cbd5570SChris Mason * 136cbd5570SChris Mason * You should have received a copy of the GNU General Public 146cbd5570SChris Mason * License along with this program; if not, write to the 156cbd5570SChris Mason * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 166cbd5570SChris Mason * Boston, MA 021110-1307, USA. 176cbd5570SChris Mason */ 186cbd5570SChris Mason 19e20d96d6SChris Mason #include <linux/fs.h> 20d98237b3SChris Mason #include <linux/blkdev.h> 2187cbda5cSChris Mason #include <linux/scatterlist.h> 2222b0ebdaSChris Mason #include <linux/swap.h> 230f7d52f4SChris Mason #include <linux/radix-tree.h> 2435b7e476SChris Mason #include <linux/writeback.h> 25d397712bSChris Mason #include <linux/buffer_head.h> 26ce9adaa5SChris Mason #include <linux/workqueue.h> 27a74a4b97SChris Mason #include <linux/kthread.h> 285a0e3ad6STejun Heo #include <linux/slab.h> 29784b4e29SChris Mason #include <linux/migrate.h> 307a36ddecSDavid Sterba #include <linux/ratelimit.h> 316463fe58SStefan Behrens #include <linux/uuid.h> 32803b2f54SStefan Behrens #include <linux/semaphore.h> 337e75bf3fSDavid Sterba #include <asm/unaligned.h> 34eb60ceacSChris Mason #include "ctree.h" 35eb60ceacSChris Mason #include "disk-io.h" 360b947affSFilipe David Borba Manana #include "hash.h" 37e089f05cSChris Mason #include "transaction.h" 380f7d52f4SChris Mason #include "btrfs_inode.h" 390b86a832SChris Mason #include "volumes.h" 40db94535dSChris Mason #include "print-tree.h" 41925baeddSChris Mason #include "locking.h" 42e02119d5SChris Mason #include "tree-log.h" 43fa9c0d79SChris Mason #include "free-space-cache.h" 4470f6d82eSOmar Sandoval #include "free-space-tree.h" 45581bb050SLi Zefan #include "inode-map.h" 4621adbd5cSStefan Behrens #include "check-integrity.h" 47606686eeSJosef Bacik #include "rcu-string.h" 488dabb742SStefan Behrens #include "dev-replace.h" 4953b381b3SDavid Woodhouse #include "raid56.h" 505ac1d209SJeff Mahoney #include "sysfs.h" 51fcebe456SJosef Bacik #include "qgroup.h" 52ebb8765bSAnand Jain #include "compression.h" 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 kmem_cache_destroy(btrfs_end_io_wq_cache); 11497eb6b69SDavid Sterba } 11597eb6b69SDavid Sterba 116d352ac68SChris Mason /* 117d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 118d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 119d352ac68SChris Mason * just before they are sent down the IO stack. 120d352ac68SChris Mason */ 12144b8bd7eSChris Mason struct async_submit_bio { 12244b8bd7eSChris Mason struct inode *inode; 12344b8bd7eSChris Mason struct bio *bio; 12444b8bd7eSChris Mason struct list_head list; 1254a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start; 1264a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done; 12744b8bd7eSChris Mason int rw; 12844b8bd7eSChris Mason int mirror_num; 129c8b97818SChris Mason unsigned long bio_flags; 130eaf25d93SChris Mason /* 131eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 132eaf25d93SChris Mason * can't tell us where in the file the bio should go 133eaf25d93SChris Mason */ 134eaf25d93SChris Mason u64 bio_offset; 1358b712842SChris Mason struct btrfs_work work; 13679787eaaSJeff Mahoney int error; 13744b8bd7eSChris Mason }; 13844b8bd7eSChris Mason 13985d4e461SChris Mason /* 14085d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 14185d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 14285d4e461SChris Mason * the level the eb occupies in the tree. 1434008c04aSChris Mason * 14485d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 14585d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 14685d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 14785d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 14885d4e461SChris Mason * have separate keysets. 1494008c04aSChris Mason * 15085d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 15185d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 15285d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1534008c04aSChris Mason * 15485d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 15585d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 15685d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 15785d4e461SChris Mason * 15885d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 15985d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 16085d4e461SChris Mason * needs update as well. 1614008c04aSChris Mason */ 1624008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1634008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1644008c04aSChris Mason # error 1654008c04aSChris Mason # endif 16685d4e461SChris Mason 16785d4e461SChris Mason static struct btrfs_lockdep_keyset { 16885d4e461SChris Mason u64 id; /* root objectid */ 16985d4e461SChris Mason const char *name_stem; /* lock name stem */ 17085d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 17185d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 17285d4e461SChris Mason } btrfs_lockdep_keysets[] = { 17385d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 17485d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 17585d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 17685d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 17785d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 17885d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 17960b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 18085d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 18185d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 18285d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 18313fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1846b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 18585d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1864008c04aSChris Mason }; 18785d4e461SChris Mason 18885d4e461SChris Mason void __init btrfs_init_lockdep(void) 18985d4e461SChris Mason { 19085d4e461SChris Mason int i, j; 19185d4e461SChris Mason 19285d4e461SChris Mason /* initialize lockdep class names */ 19385d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 19485d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 19585d4e461SChris Mason 19685d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 19785d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 19885d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 19985d4e461SChris Mason } 20085d4e461SChris Mason } 20185d4e461SChris Mason 20285d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 20385d4e461SChris Mason int level) 20485d4e461SChris Mason { 20585d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 20685d4e461SChris Mason 20785d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 20885d4e461SChris Mason 20985d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 21085d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 21185d4e461SChris Mason if (ks->id == objectid) 21285d4e461SChris Mason break; 21385d4e461SChris Mason 21485d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 21585d4e461SChris Mason &ks->keys[level], ks->names[level]); 21685d4e461SChris Mason } 21785d4e461SChris Mason 2184008c04aSChris Mason #endif 2194008c04aSChris Mason 220d352ac68SChris Mason /* 221d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 222d352ac68SChris Mason * that covers the entire device 223d352ac68SChris Mason */ 224b2950863SChristoph Hellwig static struct extent_map *btree_get_extent(struct inode *inode, 225306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 2265f39d397SChris Mason int create) 2275f39d397SChris Mason { 2285f39d397SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 2295f39d397SChris Mason struct extent_map *em; 2305f39d397SChris Mason int ret; 2315f39d397SChris Mason 232890871beSChris Mason read_lock(&em_tree->lock); 233d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 234a061fc8dSChris Mason if (em) { 235a061fc8dSChris Mason em->bdev = 236a061fc8dSChris Mason BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 237890871beSChris Mason read_unlock(&em_tree->lock); 2385f39d397SChris Mason goto out; 239a061fc8dSChris Mason } 240890871beSChris Mason read_unlock(&em_tree->lock); 2417b13b7b1SChris Mason 242172ddd60SDavid Sterba em = alloc_extent_map(); 2435f39d397SChris Mason if (!em) { 2445f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2455f39d397SChris Mason goto out; 2465f39d397SChris Mason } 2475f39d397SChris Mason em->start = 0; 2480afbaf8cSChris Mason em->len = (u64)-1; 249c8b97818SChris Mason em->block_len = (u64)-1; 2505f39d397SChris Mason em->block_start = 0; 251a061fc8dSChris Mason em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 252d1310b2eSChris Mason 253890871beSChris Mason write_lock(&em_tree->lock); 25409a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2555f39d397SChris Mason if (ret == -EEXIST) { 2565f39d397SChris Mason free_extent_map(em); 2577b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 258b4f359abSTsutomu Itoh if (!em) 2590433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2605f39d397SChris Mason } else if (ret) { 2617b13b7b1SChris Mason free_extent_map(em); 2620433f20dSTsutomu Itoh em = ERR_PTR(ret); 2635f39d397SChris Mason } 264890871beSChris Mason write_unlock(&em_tree->lock); 2657b13b7b1SChris Mason 2665f39d397SChris Mason out: 2675f39d397SChris Mason return em; 2685f39d397SChris Mason } 2695f39d397SChris Mason 270b0496686SLiu Bo u32 btrfs_csum_data(char *data, u32 seed, size_t len) 27119c00ddcSChris Mason { 2720b947affSFilipe David Borba Manana return btrfs_crc32c(seed, data, len); 27319c00ddcSChris Mason } 27419c00ddcSChris Mason 27519c00ddcSChris Mason void btrfs_csum_final(u32 crc, char *result) 27619c00ddcSChris Mason { 2777e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 27819c00ddcSChris Mason } 27919c00ddcSChris Mason 280d352ac68SChris Mason /* 281d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 282d352ac68SChris Mason * into the csum field of the block. 283d352ac68SChris Mason */ 28401d58472SDaniel Dressler static int csum_tree_block(struct btrfs_fs_info *fs_info, 28501d58472SDaniel Dressler struct extent_buffer *buf, 28619c00ddcSChris Mason int verify) 28719c00ddcSChris Mason { 28801d58472SDaniel Dressler u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 289607d432dSJosef Bacik char *result = NULL; 29019c00ddcSChris Mason unsigned long len; 29119c00ddcSChris Mason unsigned long cur_len; 29219c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 29319c00ddcSChris Mason char *kaddr; 29419c00ddcSChris Mason unsigned long map_start; 29519c00ddcSChris Mason unsigned long map_len; 29619c00ddcSChris Mason int err; 29719c00ddcSChris Mason u32 crc = ~(u32)0; 298607d432dSJosef Bacik unsigned long inline_result; 29919c00ddcSChris Mason 30019c00ddcSChris Mason len = buf->len - offset; 30119c00ddcSChris Mason while (len > 0) { 30219c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 303a6591715SChris Mason &kaddr, &map_start, &map_len); 304d397712bSChris Mason if (err) 3058bd98f0eSAlex Lyakas return err; 30619c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 307b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 30819c00ddcSChris Mason crc, cur_len); 30919c00ddcSChris Mason len -= cur_len; 31019c00ddcSChris Mason offset += cur_len; 31119c00ddcSChris Mason } 312607d432dSJosef Bacik if (csum_size > sizeof(inline_result)) { 31331e818feSDavid Sterba result = kzalloc(csum_size, GFP_NOFS); 314607d432dSJosef Bacik if (!result) 3158bd98f0eSAlex Lyakas return -ENOMEM; 316607d432dSJosef Bacik } else { 317607d432dSJosef Bacik result = (char *)&inline_result; 318607d432dSJosef Bacik } 319607d432dSJosef Bacik 32019c00ddcSChris Mason btrfs_csum_final(crc, result); 32119c00ddcSChris Mason 32219c00ddcSChris Mason if (verify) { 323607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 324e4204dedSChris Mason u32 val; 325e4204dedSChris Mason u32 found = 0; 326607d432dSJosef Bacik memcpy(&found, result, csum_size); 327e4204dedSChris Mason 328607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 32994647322SDavid Sterba btrfs_warn_rl(fs_info, 33094647322SDavid Sterba "%s checksum verify failed on %llu wanted %X found %X " 33194647322SDavid Sterba "level %d", 33201d58472SDaniel Dressler fs_info->sb->s_id, buf->start, 333c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 334607d432dSJosef Bacik if (result != (char *)&inline_result) 335607d432dSJosef Bacik kfree(result); 3368bd98f0eSAlex Lyakas return -EUCLEAN; 33719c00ddcSChris Mason } 33819c00ddcSChris Mason } else { 339607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 34019c00ddcSChris Mason } 341607d432dSJosef Bacik if (result != (char *)&inline_result) 342607d432dSJosef Bacik kfree(result); 34319c00ddcSChris Mason return 0; 34419c00ddcSChris Mason } 34519c00ddcSChris Mason 346d352ac68SChris Mason /* 347d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 348d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 349d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 350d352ac68SChris Mason * in the wrong place. 351d352ac68SChris Mason */ 3521259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 353b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 354b9fab919SChris Mason int atomic) 3551259ab75SChris Mason { 3562ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3571259ab75SChris Mason int ret; 3582755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 3591259ab75SChris Mason 3601259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3611259ab75SChris Mason return 0; 3621259ab75SChris Mason 363b9fab919SChris Mason if (atomic) 364b9fab919SChris Mason return -EAGAIN; 365b9fab919SChris Mason 366a26e8c9fSJosef Bacik if (need_lock) { 367a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 368a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 369a26e8c9fSJosef Bacik } 370a26e8c9fSJosef Bacik 3712ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 372ff13db41SDavid Sterba &cached_state); 3730b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3741259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3751259ab75SChris Mason ret = 0; 3761259ab75SChris Mason goto out; 3771259ab75SChris Mason } 37894647322SDavid Sterba btrfs_err_rl(eb->fs_info, 37994647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 38094647322SDavid Sterba eb->start, 38129549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3821259ab75SChris Mason ret = 1; 383a26e8c9fSJosef Bacik 384a26e8c9fSJosef Bacik /* 385a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 386a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 387a26e8c9fSJosef Bacik * block that has been free'd and re-allocated. So don't clear uptodate 388a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 389a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 390a26e8c9fSJosef Bacik * making everybody very sad. 391a26e8c9fSJosef Bacik */ 392a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3930b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 39433958dc6SChris Mason out: 3952ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 3962ac55d41SJosef Bacik &cached_state, GFP_NOFS); 397472b909fSJosef Bacik if (need_lock) 398a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3991259ab75SChris Mason return ret; 4001259ab75SChris Mason } 4011259ab75SChris Mason 402d352ac68SChris Mason /* 4031104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 4041104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 4051104a885SDavid Sterba */ 4061104a885SDavid Sterba static int btrfs_check_super_csum(char *raw_disk_sb) 4071104a885SDavid Sterba { 4081104a885SDavid Sterba struct btrfs_super_block *disk_sb = 4091104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 4101104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 4111104a885SDavid Sterba int ret = 0; 4121104a885SDavid Sterba 4131104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 4141104a885SDavid Sterba u32 crc = ~(u32)0; 4151104a885SDavid Sterba const int csum_size = sizeof(crc); 4161104a885SDavid Sterba char result[csum_size]; 4171104a885SDavid Sterba 4181104a885SDavid Sterba /* 4191104a885SDavid Sterba * The super_block structure does not span the whole 4201104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 4211104a885SDavid Sterba * is filled with zeros and is included in the checkum. 4221104a885SDavid Sterba */ 4231104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 4241104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 4251104a885SDavid Sterba btrfs_csum_final(crc, result); 4261104a885SDavid Sterba 4271104a885SDavid Sterba if (memcmp(raw_disk_sb, result, csum_size)) 4281104a885SDavid Sterba ret = 1; 4291104a885SDavid Sterba } 4301104a885SDavid Sterba 4311104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 432efe120a0SFrank Holton printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", 4331104a885SDavid Sterba csum_type); 4341104a885SDavid Sterba ret = 1; 4351104a885SDavid Sterba } 4361104a885SDavid Sterba 4371104a885SDavid Sterba return ret; 4381104a885SDavid Sterba } 4391104a885SDavid Sterba 4401104a885SDavid Sterba /* 441d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 442d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 443d352ac68SChris Mason */ 444f188591eSChris Mason static int btree_read_extent_buffer_pages(struct btrfs_root *root, 445f188591eSChris Mason struct extent_buffer *eb, 446ca7a79adSChris Mason u64 start, u64 parent_transid) 447f188591eSChris Mason { 448f188591eSChris Mason struct extent_io_tree *io_tree; 449ea466794SJosef Bacik int failed = 0; 450f188591eSChris Mason int ret; 451f188591eSChris Mason int num_copies = 0; 452f188591eSChris Mason int mirror_num = 0; 453ea466794SJosef Bacik int failed_mirror = 0; 454f188591eSChris Mason 455a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 456f188591eSChris Mason io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 457f188591eSChris Mason while (1) { 458bb82ab88SArne Jansen ret = read_extent_buffer_pages(io_tree, eb, start, 459bb82ab88SArne Jansen WAIT_COMPLETE, 460f188591eSChris Mason btree_get_extent, mirror_num); 461256dd1bbSStefan Behrens if (!ret) { 462256dd1bbSStefan Behrens if (!verify_parent_transid(io_tree, eb, 463b9fab919SChris Mason parent_transid, 0)) 464ea466794SJosef Bacik break; 465256dd1bbSStefan Behrens else 466256dd1bbSStefan Behrens ret = -EIO; 467256dd1bbSStefan Behrens } 468d397712bSChris Mason 469a826d6dcSJosef Bacik /* 470a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 471a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 472a826d6dcSJosef Bacik * any less wrong. 473a826d6dcSJosef Bacik */ 474a826d6dcSJosef Bacik if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 475ea466794SJosef Bacik break; 476ea466794SJosef Bacik 4775d964051SStefan Behrens num_copies = btrfs_num_copies(root->fs_info, 478f188591eSChris Mason eb->start, eb->len); 4794235298eSChris Mason if (num_copies == 1) 480ea466794SJosef Bacik break; 4814235298eSChris Mason 4825cf1ab56SJosef Bacik if (!failed_mirror) { 4835cf1ab56SJosef Bacik failed = 1; 4845cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4855cf1ab56SJosef Bacik } 4865cf1ab56SJosef Bacik 487f188591eSChris Mason mirror_num++; 488ea466794SJosef Bacik if (mirror_num == failed_mirror) 489ea466794SJosef Bacik mirror_num++; 490ea466794SJosef Bacik 4914235298eSChris Mason if (mirror_num > num_copies) 492ea466794SJosef Bacik break; 493f188591eSChris Mason } 494ea466794SJosef Bacik 495c0901581SStefan Behrens if (failed && !ret && failed_mirror) 496ea466794SJosef Bacik repair_eb_io_failure(root, eb, failed_mirror); 497ea466794SJosef Bacik 498ea466794SJosef Bacik return ret; 499f188591eSChris Mason } 50019c00ddcSChris Mason 501d352ac68SChris Mason /* 502d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 503d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 504d352ac68SChris Mason */ 505d397712bSChris Mason 50601d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 50719c00ddcSChris Mason { 5084eee4fa4SMiao Xie u64 start = page_offset(page); 50919c00ddcSChris Mason u64 found_start; 51019c00ddcSChris Mason struct extent_buffer *eb; 511f188591eSChris Mason 5124f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5134f2de97aSJosef Bacik if (page != eb->pages[0]) 5144f2de97aSJosef Bacik return 0; 5150f805531SAlex Lyakas 51619c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 5170f805531SAlex Lyakas /* 5180f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5190f805531SAlex Lyakas * It is useful to know what went wrong. 5200f805531SAlex Lyakas */ 5210f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5220f805531SAlex Lyakas return -EUCLEAN; 5230f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5240f805531SAlex Lyakas return -EUCLEAN; 5250f805531SAlex Lyakas 5260f805531SAlex Lyakas ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, 5270f805531SAlex Lyakas btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); 5280f805531SAlex Lyakas 5298bd98f0eSAlex Lyakas return csum_tree_block(fs_info, eb, 0); 53019c00ddcSChris Mason } 53119c00ddcSChris Mason 53201d58472SDaniel Dressler static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, 5332b82032cSYan Zheng struct extent_buffer *eb) 5342b82032cSYan Zheng { 53501d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 5362b82032cSYan Zheng u8 fsid[BTRFS_UUID_SIZE]; 5372b82032cSYan Zheng int ret = 1; 5382b82032cSYan Zheng 5390a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5402b82032cSYan Zheng while (fs_devices) { 5412b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5422b82032cSYan Zheng ret = 0; 5432b82032cSYan Zheng break; 5442b82032cSYan Zheng } 5452b82032cSYan Zheng fs_devices = fs_devices->seed; 5462b82032cSYan Zheng } 5472b82032cSYan Zheng return ret; 5482b82032cSYan Zheng } 5492b82032cSYan Zheng 550a826d6dcSJosef Bacik #define CORRUPT(reason, eb, root, slot) \ 551efe120a0SFrank Holton btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ 552efe120a0SFrank Holton "root=%llu, slot=%d", reason, \ 553c1c9ff7cSGeert Uytterhoeven btrfs_header_bytenr(eb), root->objectid, slot) 554a826d6dcSJosef Bacik 555a826d6dcSJosef Bacik static noinline int check_leaf(struct btrfs_root *root, 556a826d6dcSJosef Bacik struct extent_buffer *leaf) 557a826d6dcSJosef Bacik { 558a826d6dcSJosef Bacik struct btrfs_key key; 559a826d6dcSJosef Bacik struct btrfs_key leaf_key; 560a826d6dcSJosef Bacik u32 nritems = btrfs_header_nritems(leaf); 561a826d6dcSJosef Bacik int slot; 562a826d6dcSJosef Bacik 563a826d6dcSJosef Bacik if (nritems == 0) 564a826d6dcSJosef Bacik return 0; 565a826d6dcSJosef Bacik 566a826d6dcSJosef Bacik /* Check the 0 item */ 567a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 568a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 569a826d6dcSJosef Bacik CORRUPT("invalid item offset size pair", leaf, root, 0); 570a826d6dcSJosef Bacik return -EIO; 571a826d6dcSJosef Bacik } 572a826d6dcSJosef Bacik 573a826d6dcSJosef Bacik /* 574a826d6dcSJosef Bacik * Check to make sure each items keys are in the correct order and their 575a826d6dcSJosef Bacik * offsets make sense. We only have to loop through nritems-1 because 576a826d6dcSJosef Bacik * we check the current slot against the next slot, which verifies the 577a826d6dcSJosef Bacik * next slot's offset+size makes sense and that the current's slot 578a826d6dcSJosef Bacik * offset is correct. 579a826d6dcSJosef Bacik */ 580a826d6dcSJosef Bacik for (slot = 0; slot < nritems - 1; slot++) { 581a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &leaf_key, slot); 582a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &key, slot + 1); 583a826d6dcSJosef Bacik 584a826d6dcSJosef Bacik /* Make sure the keys are in the right order */ 585a826d6dcSJosef Bacik if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { 586a826d6dcSJosef Bacik CORRUPT("bad key order", leaf, root, slot); 587a826d6dcSJosef Bacik return -EIO; 588a826d6dcSJosef Bacik } 589a826d6dcSJosef Bacik 590a826d6dcSJosef Bacik /* 591a826d6dcSJosef Bacik * Make sure the offset and ends are right, remember that the 592a826d6dcSJosef Bacik * item data starts at the end of the leaf and grows towards the 593a826d6dcSJosef Bacik * front. 594a826d6dcSJosef Bacik */ 595a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, slot) != 596a826d6dcSJosef Bacik btrfs_item_end_nr(leaf, slot + 1)) { 597a826d6dcSJosef Bacik CORRUPT("slot offset bad", leaf, root, slot); 598a826d6dcSJosef Bacik return -EIO; 599a826d6dcSJosef Bacik } 600a826d6dcSJosef Bacik 601a826d6dcSJosef Bacik /* 602a826d6dcSJosef Bacik * Check to make sure that we don't point outside of the leaf, 603a826d6dcSJosef Bacik * just incase all the items are consistent to eachother, but 604a826d6dcSJosef Bacik * all point outside of the leaf. 605a826d6dcSJosef Bacik */ 606a826d6dcSJosef Bacik if (btrfs_item_end_nr(leaf, slot) > 607a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 608a826d6dcSJosef Bacik CORRUPT("slot end outside of leaf", leaf, root, slot); 609a826d6dcSJosef Bacik return -EIO; 610a826d6dcSJosef Bacik } 611a826d6dcSJosef Bacik } 612a826d6dcSJosef Bacik 613a826d6dcSJosef Bacik return 0; 614a826d6dcSJosef Bacik } 615a826d6dcSJosef Bacik 616facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 617facc8a22SMiao Xie u64 phy_offset, struct page *page, 618facc8a22SMiao Xie u64 start, u64 end, int mirror) 619ce9adaa5SChris Mason { 620ce9adaa5SChris Mason u64 found_start; 621ce9adaa5SChris Mason int found_level; 622ce9adaa5SChris Mason struct extent_buffer *eb; 623ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 62402873e43SZhao Lei struct btrfs_fs_info *fs_info = root->fs_info; 625f188591eSChris Mason int ret = 0; 626727011e0SChris Mason int reads_done; 627ce9adaa5SChris Mason 628ce9adaa5SChris Mason if (!page->private) 629ce9adaa5SChris Mason goto out; 630d397712bSChris Mason 6314f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 632d397712bSChris Mason 6330b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 6340b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 6350b32f4bbSJosef Bacik */ 6360b32f4bbSJosef Bacik extent_buffer_get(eb); 6370b32f4bbSJosef Bacik 6380b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 639727011e0SChris Mason if (!reads_done) 640727011e0SChris Mason goto err; 641f188591eSChris Mason 6425cf1ab56SJosef Bacik eb->read_mirror = mirror; 643656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 644ea466794SJosef Bacik ret = -EIO; 645ea466794SJosef Bacik goto err; 646ea466794SJosef Bacik } 647ea466794SJosef Bacik 648ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 649727011e0SChris Mason if (found_start != eb->start) { 65002873e43SZhao Lei btrfs_err_rl(fs_info, "bad tree block start %llu %llu", 65194647322SDavid Sterba found_start, eb->start); 652f188591eSChris Mason ret = -EIO; 653ce9adaa5SChris Mason goto err; 654ce9adaa5SChris Mason } 65502873e43SZhao Lei if (check_tree_block_fsid(fs_info, eb)) { 65602873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 65794647322SDavid Sterba eb->start); 6581259ab75SChris Mason ret = -EIO; 6591259ab75SChris Mason goto err; 6601259ab75SChris Mason } 661ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6621c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 66302873e43SZhao Lei btrfs_err(fs_info, "bad tree block level %d", 6641c24c3ceSJosef Bacik (int)btrfs_header_level(eb)); 6651c24c3ceSJosef Bacik ret = -EIO; 6661c24c3ceSJosef Bacik goto err; 6671c24c3ceSJosef Bacik } 668ce9adaa5SChris Mason 66985d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 67085d4e461SChris Mason eb, found_level); 6714008c04aSChris Mason 67202873e43SZhao Lei ret = csum_tree_block(fs_info, eb, 1); 6738bd98f0eSAlex Lyakas if (ret) 674a826d6dcSJosef Bacik goto err; 675a826d6dcSJosef Bacik 676a826d6dcSJosef Bacik /* 677a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 678a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 679a826d6dcSJosef Bacik * return -EIO. 680a826d6dcSJosef Bacik */ 681a826d6dcSJosef Bacik if (found_level == 0 && check_leaf(root, eb)) { 682a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 683a826d6dcSJosef Bacik ret = -EIO; 684a826d6dcSJosef Bacik } 685ce9adaa5SChris Mason 6860b32f4bbSJosef Bacik if (!ret) 6870b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 688ce9adaa5SChris Mason err: 68979fb65a1SJosef Bacik if (reads_done && 69079fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 69102873e43SZhao Lei btree_readahead_hook(fs_info, eb, eb->start, ret); 6924bb31e92SArne Jansen 69353b381b3SDavid Woodhouse if (ret) { 69453b381b3SDavid Woodhouse /* 69553b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 69653b381b3SDavid Woodhouse * again, we have to make sure it has something 69753b381b3SDavid Woodhouse * to decrement 69853b381b3SDavid Woodhouse */ 69953b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 7000b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 70153b381b3SDavid Woodhouse } 7020b32f4bbSJosef Bacik free_extent_buffer(eb); 703ce9adaa5SChris Mason out: 704f188591eSChris Mason return ret; 705ce9adaa5SChris Mason } 706ce9adaa5SChris Mason 707ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 7084bb31e92SArne Jansen { 7094bb31e92SArne Jansen struct extent_buffer *eb; 7104bb31e92SArne Jansen 7114f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 712656f30dbSFilipe Manana set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); 7135cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 71453b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 715ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 71602873e43SZhao Lei btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO); 7174bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 7184bb31e92SArne Jansen } 7194bb31e92SArne Jansen 7204246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 721ce9adaa5SChris Mason { 72297eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 723ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 7249e0af237SLiu Bo struct btrfs_workqueue *wq; 7259e0af237SLiu Bo btrfs_work_func_t func; 726ce9adaa5SChris Mason 727ce9adaa5SChris Mason fs_info = end_io_wq->info; 7284246a0b6SChristoph Hellwig end_io_wq->error = bio->bi_error; 729d20f7043SChris Mason 7307b6d91daSChristoph Hellwig if (bio->bi_rw & REQ_WRITE) { 7319e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 7329e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 7339e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 7349e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 7359e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 7369e0af237SLiu Bo func = btrfs_freespace_write_helper; 7379e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7389e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7399e0af237SLiu Bo func = btrfs_endio_raid56_helper; 740d20f7043SChris Mason } else { 7419e0af237SLiu Bo wq = fs_info->endio_write_workers; 7429e0af237SLiu Bo func = btrfs_endio_write_helper; 743d20f7043SChris Mason } 7449e0af237SLiu Bo } else { 7458b110e39SMiao Xie if (unlikely(end_io_wq->metadata == 7468b110e39SMiao Xie BTRFS_WQ_ENDIO_DIO_REPAIR)) { 7478b110e39SMiao Xie wq = fs_info->endio_repair_workers; 7488b110e39SMiao Xie func = btrfs_endio_repair_helper; 7498b110e39SMiao Xie } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7509e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7519e0af237SLiu Bo func = btrfs_endio_raid56_helper; 7529e0af237SLiu Bo } else if (end_io_wq->metadata) { 7539e0af237SLiu Bo wq = fs_info->endio_meta_workers; 7549e0af237SLiu Bo func = btrfs_endio_meta_helper; 7559e0af237SLiu Bo } else { 7569e0af237SLiu Bo wq = fs_info->endio_workers; 7579e0af237SLiu Bo func = btrfs_endio_helper; 7589e0af237SLiu Bo } 7599e0af237SLiu Bo } 7609e0af237SLiu Bo 7619e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 7629e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 763ce9adaa5SChris Mason } 764ce9adaa5SChris Mason 76522c59948SChris Mason int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 766bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7670b86a832SChris Mason { 76897eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7698b110e39SMiao Xie 77097eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 771ce9adaa5SChris Mason if (!end_io_wq) 772ce9adaa5SChris Mason return -ENOMEM; 773ce9adaa5SChris Mason 774ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 775ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 77622c59948SChris Mason end_io_wq->info = info; 777ce9adaa5SChris Mason end_io_wq->error = 0; 778ce9adaa5SChris Mason end_io_wq->bio = bio; 77922c59948SChris Mason end_io_wq->metadata = metadata; 780ce9adaa5SChris Mason 781ce9adaa5SChris Mason bio->bi_private = end_io_wq; 782ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 78322c59948SChris Mason return 0; 78422c59948SChris Mason } 78522c59948SChris Mason 786b64a2851SChris Mason unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 7874854ddd0SChris Mason { 7884854ddd0SChris Mason unsigned long limit = min_t(unsigned long, 7895cdc7ad3SQu Wenruo info->thread_pool_size, 7904854ddd0SChris Mason info->fs_devices->open_devices); 7914854ddd0SChris Mason return 256 * limit; 7924854ddd0SChris Mason } 7934854ddd0SChris Mason 7944a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7954a69a410SChris Mason { 7964a69a410SChris Mason struct async_submit_bio *async; 79779787eaaSJeff Mahoney int ret; 7984a69a410SChris Mason 7994a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 80079787eaaSJeff Mahoney ret = async->submit_bio_start(async->inode, async->rw, async->bio, 801eaf25d93SChris Mason async->mirror_num, async->bio_flags, 802eaf25d93SChris Mason async->bio_offset); 80379787eaaSJeff Mahoney if (ret) 80479787eaaSJeff Mahoney async->error = ret; 8054a69a410SChris Mason } 8064a69a410SChris Mason 8074a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 8088b712842SChris Mason { 8098b712842SChris Mason struct btrfs_fs_info *fs_info; 8108b712842SChris Mason struct async_submit_bio *async; 8114854ddd0SChris Mason int limit; 8128b712842SChris Mason 8138b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 8148b712842SChris Mason fs_info = BTRFS_I(async->inode)->root->fs_info; 8154854ddd0SChris Mason 816b64a2851SChris Mason limit = btrfs_async_submit_limit(fs_info); 8174854ddd0SChris Mason limit = limit * 2 / 3; 8184854ddd0SChris Mason 819ee863954SDavid Sterba /* 820ee863954SDavid Sterba * atomic_dec_return implies a barrier for waitqueue_active 821ee863954SDavid Sterba */ 82266657b31SJosef Bacik if (atomic_dec_return(&fs_info->nr_async_submits) < limit && 823b64a2851SChris Mason waitqueue_active(&fs_info->async_submit_wait)) 8244854ddd0SChris Mason wake_up(&fs_info->async_submit_wait); 8254854ddd0SChris Mason 826bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 82779787eaaSJeff Mahoney if (async->error) { 8284246a0b6SChristoph Hellwig async->bio->bi_error = async->error; 8294246a0b6SChristoph Hellwig bio_endio(async->bio); 83079787eaaSJeff Mahoney return; 83179787eaaSJeff Mahoney } 83279787eaaSJeff Mahoney 8334a69a410SChris Mason async->submit_bio_done(async->inode, async->rw, async->bio, 834eaf25d93SChris Mason async->mirror_num, async->bio_flags, 835eaf25d93SChris Mason async->bio_offset); 8364a69a410SChris Mason } 8374a69a410SChris Mason 8384a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8394a69a410SChris Mason { 8404a69a410SChris Mason struct async_submit_bio *async; 8414a69a410SChris Mason 8424a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8438b712842SChris Mason kfree(async); 8448b712842SChris Mason } 8458b712842SChris Mason 84644b8bd7eSChris Mason int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, 84744b8bd7eSChris Mason int rw, struct bio *bio, int mirror_num, 848c8b97818SChris Mason unsigned long bio_flags, 849eaf25d93SChris Mason u64 bio_offset, 8504a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start, 8514a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done) 85244b8bd7eSChris Mason { 85344b8bd7eSChris Mason struct async_submit_bio *async; 85444b8bd7eSChris Mason 85544b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 85644b8bd7eSChris Mason if (!async) 85744b8bd7eSChris Mason return -ENOMEM; 85844b8bd7eSChris Mason 85944b8bd7eSChris Mason async->inode = inode; 86044b8bd7eSChris Mason async->rw = rw; 86144b8bd7eSChris Mason async->bio = bio; 86244b8bd7eSChris Mason async->mirror_num = mirror_num; 8634a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8644a69a410SChris Mason async->submit_bio_done = submit_bio_done; 8654a69a410SChris Mason 8669e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 8675cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 8684a69a410SChris Mason 869c8b97818SChris Mason async->bio_flags = bio_flags; 870eaf25d93SChris Mason async->bio_offset = bio_offset; 8718c8bee1dSChris Mason 87279787eaaSJeff Mahoney async->error = 0; 87379787eaaSJeff Mahoney 874cb03c743SChris Mason atomic_inc(&fs_info->nr_async_submits); 875d313d7a3SChris Mason 8767b6d91daSChristoph Hellwig if (rw & REQ_SYNC) 8775cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 878d313d7a3SChris Mason 8795cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 8809473f16cSChris Mason 881771ed689SChris Mason while (atomic_read(&fs_info->async_submit_draining) && 882771ed689SChris Mason atomic_read(&fs_info->nr_async_submits)) { 883771ed689SChris Mason wait_event(fs_info->async_submit_wait, 884771ed689SChris Mason (atomic_read(&fs_info->nr_async_submits) == 0)); 885771ed689SChris Mason } 886771ed689SChris Mason 88744b8bd7eSChris Mason return 0; 88844b8bd7eSChris Mason } 88944b8bd7eSChris Mason 890ce3ed71aSChris Mason static int btree_csum_one_bio(struct bio *bio) 891ce3ed71aSChris Mason { 8922c30c71bSKent Overstreet struct bio_vec *bvec; 893ce3ed71aSChris Mason struct btrfs_root *root; 8942c30c71bSKent Overstreet int i, ret = 0; 895ce3ed71aSChris Mason 8962c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 897ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 89801d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 89979787eaaSJeff Mahoney if (ret) 90079787eaaSJeff Mahoney break; 901ce3ed71aSChris Mason } 9022c30c71bSKent Overstreet 90379787eaaSJeff Mahoney return ret; 904ce3ed71aSChris Mason } 905ce3ed71aSChris Mason 9064a69a410SChris Mason static int __btree_submit_bio_start(struct inode *inode, int rw, 9074a69a410SChris Mason struct bio *bio, int mirror_num, 908eaf25d93SChris Mason unsigned long bio_flags, 909eaf25d93SChris Mason u64 bio_offset) 91022c59948SChris Mason { 9118b712842SChris Mason /* 9128b712842SChris Mason * when we're called for a write, we're already in the async 9135443be45SChris Mason * submission context. Just jump into btrfs_map_bio 9148b712842SChris Mason */ 91579787eaaSJeff Mahoney return btree_csum_one_bio(bio); 91622c59948SChris Mason } 91722c59948SChris Mason 9184a69a410SChris Mason static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 919eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 920eaf25d93SChris Mason u64 bio_offset) 9214a69a410SChris Mason { 92261891923SStefan Behrens int ret; 92361891923SStefan Behrens 9248b712842SChris Mason /* 9254a69a410SChris Mason * when we're called for a write, we're already in the async 9264a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 9278b712842SChris Mason */ 92861891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 9294246a0b6SChristoph Hellwig if (ret) { 9304246a0b6SChristoph Hellwig bio->bi_error = ret; 9314246a0b6SChristoph Hellwig bio_endio(bio); 9324246a0b6SChristoph Hellwig } 93361891923SStefan Behrens return ret; 9340b86a832SChris Mason } 9350b86a832SChris Mason 936de0022b9SJosef Bacik static int check_async_write(struct inode *inode, unsigned long bio_flags) 937de0022b9SJosef Bacik { 938de0022b9SJosef Bacik if (bio_flags & EXTENT_BIO_TREE_LOG) 939de0022b9SJosef Bacik return 0; 940de0022b9SJosef Bacik #ifdef CONFIG_X86 941bc696ca0SBorislav Petkov if (static_cpu_has(X86_FEATURE_XMM4_2)) 942de0022b9SJosef Bacik return 0; 943de0022b9SJosef Bacik #endif 944de0022b9SJosef Bacik return 1; 945de0022b9SJosef Bacik } 946de0022b9SJosef Bacik 94744b8bd7eSChris Mason static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 948eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 949eaf25d93SChris Mason u64 bio_offset) 95044b8bd7eSChris Mason { 951de0022b9SJosef Bacik int async = check_async_write(inode, bio_flags); 9524a69a410SChris Mason int ret; 953cad321adSChris Mason 9547b6d91daSChristoph Hellwig if (!(rw & REQ_WRITE)) { 955cad321adSChris Mason /* 956cad321adSChris Mason * called for a read, do the setup so that checksum validation 957cad321adSChris Mason * can happen in the async kernel threads 958cad321adSChris Mason */ 959f3f266abSChris Mason ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 960bfebd8b5SDavid Sterba bio, BTRFS_WQ_ENDIO_METADATA); 9611d4284bdSChris Mason if (ret) 96261891923SStefan Behrens goto out_w_error; 96361891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 9646f3577bdSChris Mason mirror_num, 0); 965de0022b9SJosef Bacik } else if (!async) { 966de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 967de0022b9SJosef Bacik if (ret) 96861891923SStefan Behrens goto out_w_error; 96961891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 970de0022b9SJosef Bacik mirror_num, 0); 97161891923SStefan Behrens } else { 972cad321adSChris Mason /* 97361891923SStefan Behrens * kthread helpers are used to submit writes so that 97461891923SStefan Behrens * checksumming can happen in parallel across all CPUs 975cad321adSChris Mason */ 97661891923SStefan Behrens ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 977c8b97818SChris Mason inode, rw, bio, mirror_num, 0, 978eaf25d93SChris Mason bio_offset, 9794a69a410SChris Mason __btree_submit_bio_start, 9804a69a410SChris Mason __btree_submit_bio_done); 98144b8bd7eSChris Mason } 98244b8bd7eSChris Mason 9834246a0b6SChristoph Hellwig if (ret) 9844246a0b6SChristoph Hellwig goto out_w_error; 9854246a0b6SChristoph Hellwig return 0; 9864246a0b6SChristoph Hellwig 98761891923SStefan Behrens out_w_error: 9884246a0b6SChristoph Hellwig bio->bi_error = ret; 9894246a0b6SChristoph Hellwig bio_endio(bio); 99061891923SStefan Behrens return ret; 99161891923SStefan Behrens } 99261891923SStefan Behrens 9933dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 994784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 995a6bc32b8SMel Gorman struct page *newpage, struct page *page, 996a6bc32b8SMel Gorman enum migrate_mode mode) 997784b4e29SChris Mason { 998784b4e29SChris Mason /* 999784b4e29SChris Mason * we can't safely write a btree page from here, 1000784b4e29SChris Mason * we haven't done the locking hook 1001784b4e29SChris Mason */ 1002784b4e29SChris Mason if (PageDirty(page)) 1003784b4e29SChris Mason return -EAGAIN; 1004784b4e29SChris Mason /* 1005784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 1006784b4e29SChris Mason * We must have no buffers or drop them. 1007784b4e29SChris Mason */ 1008784b4e29SChris Mason if (page_has_private(page) && 1009784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 1010784b4e29SChris Mason return -EAGAIN; 1011a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 1012784b4e29SChris Mason } 10133dd1462eSJan Beulich #endif 1014784b4e29SChris Mason 10150da5468fSChris Mason 10160da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 10170da5468fSChris Mason struct writeback_control *wbc) 10180da5468fSChris Mason { 1019e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 1020e2d84521SMiao Xie int ret; 1021e2d84521SMiao Xie 1022d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 1023448d640bSChris Mason 1024448d640bSChris Mason if (wbc->for_kupdate) 1025448d640bSChris Mason return 0; 1026448d640bSChris Mason 1027e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 1028b9473439SChris Mason /* this is a bit racy, but that's ok */ 1029e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 1030e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 1031e2d84521SMiao Xie if (ret < 0) 1032793955bcSChris Mason return 0; 1033793955bcSChris Mason } 10340b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 10350da5468fSChris Mason } 10360da5468fSChris Mason 1037b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 10385f39d397SChris Mason { 1039d1310b2eSChris Mason struct extent_io_tree *tree; 1040d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10418ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 10425f39d397SChris Mason } 10435f39d397SChris Mason 104470dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10455f39d397SChris Mason { 104698509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 104798509cfcSChris Mason return 0; 10480c4e538bSDavid Sterba 1049f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1050d98237b3SChris Mason } 1051d98237b3SChris Mason 1052d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1053d47992f8SLukas Czerner unsigned int length) 1054d98237b3SChris Mason { 1055d1310b2eSChris Mason struct extent_io_tree *tree; 1056d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10575f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10585f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10599ad6b7bcSChris Mason if (PagePrivate(page)) { 1060efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1061efe120a0SFrank Holton "page private not zero on page %llu", 1062efe120a0SFrank Holton (unsigned long long)page_offset(page)); 10639ad6b7bcSChris Mason ClearPagePrivate(page); 10649ad6b7bcSChris Mason set_page_private(page, 0); 106509cbfeafSKirill A. Shutemov put_page(page); 10669ad6b7bcSChris Mason } 1067d98237b3SChris Mason } 1068d98237b3SChris Mason 10690b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10700b32f4bbSJosef Bacik { 1071bb146eb2SJosef Bacik #ifdef DEBUG 10720b32f4bbSJosef Bacik struct extent_buffer *eb; 10730b32f4bbSJosef Bacik 10740b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10750b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10760b32f4bbSJosef Bacik BUG_ON(!eb); 10770b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10780b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10790b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 1080bb146eb2SJosef Bacik #endif 10810b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10820b32f4bbSJosef Bacik } 10830b32f4bbSJosef Bacik 10847f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1085d98237b3SChris Mason .readpage = btree_readpage, 10860da5468fSChris Mason .writepages = btree_writepages, 10875f39d397SChris Mason .releasepage = btree_releasepage, 10885f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10895a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1090784b4e29SChris Mason .migratepage = btree_migratepage, 10915a92bc88SChris Mason #endif 10920b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1093d98237b3SChris Mason }; 1094123abc88SChris Mason 1095d3e46feaSDavid Sterba void readahead_tree_block(struct btrfs_root *root, u64 bytenr) 1096090d1875SChris Mason { 10975f39d397SChris Mason struct extent_buffer *buf = NULL; 10985f39d397SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 1099090d1875SChris Mason 1100a83fffb7SDavid Sterba buf = btrfs_find_create_tree_block(root, bytenr); 11015f39d397SChris Mason if (!buf) 11026197d86eSDavid Sterba return; 1103d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1104bb82ab88SArne Jansen buf, 0, WAIT_NONE, btree_get_extent, 0); 11055f39d397SChris Mason free_extent_buffer(buf); 1106090d1875SChris Mason } 1107090d1875SChris Mason 1108c0dcaa4dSDavid Sterba int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, 1109ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1110ab0fff03SArne Jansen { 1111ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 1112ab0fff03SArne Jansen struct inode *btree_inode = root->fs_info->btree_inode; 1113ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1114ab0fff03SArne Jansen int ret; 1115ab0fff03SArne Jansen 1116a83fffb7SDavid Sterba buf = btrfs_find_create_tree_block(root, bytenr); 1117ab0fff03SArne Jansen if (!buf) 1118ab0fff03SArne Jansen return 0; 1119ab0fff03SArne Jansen 1120ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1121ab0fff03SArne Jansen 1122ab0fff03SArne Jansen ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1123ab0fff03SArne Jansen btree_get_extent, mirror_num); 1124ab0fff03SArne Jansen if (ret) { 1125ab0fff03SArne Jansen free_extent_buffer(buf); 1126ab0fff03SArne Jansen return ret; 1127ab0fff03SArne Jansen } 1128ab0fff03SArne Jansen 1129ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1130ab0fff03SArne Jansen free_extent_buffer(buf); 1131ab0fff03SArne Jansen return -EIO; 11320b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1133ab0fff03SArne Jansen *eb = buf; 1134ab0fff03SArne Jansen } else { 1135ab0fff03SArne Jansen free_extent_buffer(buf); 1136ab0fff03SArne Jansen } 1137ab0fff03SArne Jansen return 0; 1138ab0fff03SArne Jansen } 1139ab0fff03SArne Jansen 114001d58472SDaniel Dressler struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, 11410308af44SDavid Sterba u64 bytenr) 11420999df54SChris Mason { 114301d58472SDaniel Dressler return find_extent_buffer(fs_info, bytenr); 11440999df54SChris Mason } 11450999df54SChris Mason 11460999df54SChris Mason struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 1147a83fffb7SDavid Sterba u64 bytenr) 11480999df54SChris Mason { 1149fccb84c9SDavid Sterba if (btrfs_test_is_dummy_root(root)) 1150ce3e6984SDavid Sterba return alloc_test_extent_buffer(root->fs_info, bytenr); 1151ce3e6984SDavid Sterba return alloc_extent_buffer(root->fs_info, bytenr); 11520999df54SChris Mason } 11530999df54SChris Mason 11540999df54SChris Mason 1155e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1156e02119d5SChris Mason { 1157727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 11588aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1159e02119d5SChris Mason } 1160e02119d5SChris Mason 1161e02119d5SChris Mason int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1162e02119d5SChris Mason { 1163727011e0SChris Mason return filemap_fdatawait_range(buf->pages[0]->mapping, 11648aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1165e02119d5SChris Mason } 1166e02119d5SChris Mason 1167db94535dSChris Mason struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1168ce86cd59SDavid Sterba u64 parent_transid) 1169e20d96d6SChris Mason { 11705f39d397SChris Mason struct extent_buffer *buf = NULL; 117119c00ddcSChris Mason int ret; 117219c00ddcSChris Mason 1173a83fffb7SDavid Sterba buf = btrfs_find_create_tree_block(root, bytenr); 11745f39d397SChris Mason if (!buf) 117564c043deSLiu Bo return ERR_PTR(-ENOMEM); 1176e4204dedSChris Mason 1177ca7a79adSChris Mason ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 11780f0fe8f7SFilipe David Borba Manana if (ret) { 11790f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 118064c043deSLiu Bo return ERR_PTR(ret); 11810f0fe8f7SFilipe David Borba Manana } 11825f39d397SChris Mason return buf; 1183ce9adaa5SChris Mason 1184eb60ceacSChris Mason } 1185eb60ceacSChris Mason 118601d58472SDaniel Dressler void clean_tree_block(struct btrfs_trans_handle *trans, 118701d58472SDaniel Dressler struct btrfs_fs_info *fs_info, 11885f39d397SChris Mason struct extent_buffer *buf) 1189ed2ff2cbSChris Mason { 119055c69072SChris Mason if (btrfs_header_generation(buf) == 1191e2d84521SMiao Xie fs_info->running_transaction->transid) { 1192b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1193b4ce94deSChris Mason 1194b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1195e2d84521SMiao Xie __percpu_counter_add(&fs_info->dirty_metadata_bytes, 1196e2d84521SMiao Xie -buf->len, 1197e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1198b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1199b9473439SChris Mason btrfs_set_lock_blocking(buf); 12000b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1201925baeddSChris Mason } 12025f39d397SChris Mason } 1203ed7b63ebSJosef Bacik } 12045f39d397SChris Mason 12058257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 12068257b2dcSMiao Xie { 12078257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 12088257b2dcSMiao Xie int ret; 12098257b2dcSMiao Xie 12108257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 12118257b2dcSMiao Xie if (!writers) 12128257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 12138257b2dcSMiao Xie 1214908c7f19STejun Heo ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); 12158257b2dcSMiao Xie if (ret < 0) { 12168257b2dcSMiao Xie kfree(writers); 12178257b2dcSMiao Xie return ERR_PTR(ret); 12188257b2dcSMiao Xie } 12198257b2dcSMiao Xie 12208257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 12218257b2dcSMiao Xie return writers; 12228257b2dcSMiao Xie } 12238257b2dcSMiao Xie 12248257b2dcSMiao Xie static void 12258257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 12268257b2dcSMiao Xie { 12278257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 12288257b2dcSMiao Xie kfree(writers); 12298257b2dcSMiao Xie } 12308257b2dcSMiao Xie 1231707e8a07SDavid Sterba static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, 1232707e8a07SDavid Sterba struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1233e20d96d6SChris Mason u64 objectid) 1234d97e63b6SChris Mason { 1235cfaa7295SChris Mason root->node = NULL; 1236a28ec197SChris Mason root->commit_root = NULL; 1237db94535dSChris Mason root->sectorsize = sectorsize; 1238db94535dSChris Mason root->nodesize = nodesize; 123987ee04ebSChris Mason root->stripesize = stripesize; 124027cdeb70SMiao Xie root->state = 0; 1241d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 12420b86a832SChris Mason 12430f7d52f4SChris Mason root->objectid = objectid; 12440f7d52f4SChris Mason root->last_trans = 0; 124513a8a7c8SYan, Zheng root->highest_objectid = 0; 1246eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1247199c2a9cSMiao Xie root->nr_ordered_extents = 0; 124858176a96SJosef Bacik root->name = NULL; 12496bef4d31SEric Paris root->inode_tree = RB_ROOT; 125016cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1251f0486c68SYan, Zheng root->block_rsv = NULL; 1252d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 12530b86a832SChris Mason 12540b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 12555d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1256eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1257eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1258199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1259199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 12602ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 12612ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1262d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 12635d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1264eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1265199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1266f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 12672ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 12682ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 1269a2135011SChris Mason mutex_init(&root->objectid_mutex); 1270e02119d5SChris Mason mutex_init(&root->log_mutex); 127131f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1272573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 12737237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 12747237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 12757237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 12768b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 12778b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 12787237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 12797237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 12807237f183SYan Zheng atomic_set(&root->log_writers, 0); 12812ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 12828a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 1283b0feb9d9SMiao Xie atomic_set(&root->refs, 1); 12848257b2dcSMiao Xie atomic_set(&root->will_be_snapshoted, 0); 128555eeaf05SQu Wenruo atomic_set(&root->qgroup_meta_rsv, 0); 12867237f183SYan Zheng root->log_transid = 0; 1287d1433debSMiao Xie root->log_transid_committed = -1; 1288257c62e1SChris Mason root->last_log_commit = 0; 128906ea65a3SJosef Bacik if (fs_info) 1290d0c803c4SChris Mason extent_io_tree_init(&root->dirty_log_pages, 1291f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 1292017e5369SChris Mason 12933768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12943768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12956702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 129606ea65a3SJosef Bacik if (fs_info) 12973f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 129806ea65a3SJosef Bacik else 129906ea65a3SJosef Bacik root->defrag_trans_start = 0; 13004d775673SChris Mason root->root_key.objectid = objectid; 13010ee5dc67SAl Viro root->anon_dev = 0; 13028ea05e3aSAlexander Block 13035f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 13043768f368SChris Mason } 13053768f368SChris Mason 130674e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 130774e4d827SDavid Sterba gfp_t flags) 13086f07e42eSAl Viro { 130974e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 13106f07e42eSAl Viro if (root) 13116f07e42eSAl Viro root->fs_info = fs_info; 13126f07e42eSAl Viro return root; 13136f07e42eSAl Viro } 13146f07e42eSAl Viro 131506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 131606ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 131706ea65a3SJosef Bacik struct btrfs_root *btrfs_alloc_dummy_root(void) 131806ea65a3SJosef Bacik { 131906ea65a3SJosef Bacik struct btrfs_root *root; 132006ea65a3SJosef Bacik 132174e4d827SDavid Sterba root = btrfs_alloc_root(NULL, GFP_KERNEL); 132206ea65a3SJosef Bacik if (!root) 132306ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1324707e8a07SDavid Sterba __setup_root(4096, 4096, 4096, root, NULL, 1); 132527cdeb70SMiao Xie set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state); 1326faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 132706ea65a3SJosef Bacik 132806ea65a3SJosef Bacik return root; 132906ea65a3SJosef Bacik } 133006ea65a3SJosef Bacik #endif 133106ea65a3SJosef Bacik 133220897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 133320897f5cSArne Jansen struct btrfs_fs_info *fs_info, 133420897f5cSArne Jansen u64 objectid) 133520897f5cSArne Jansen { 133620897f5cSArne Jansen struct extent_buffer *leaf; 133720897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 133820897f5cSArne Jansen struct btrfs_root *root; 133920897f5cSArne Jansen struct btrfs_key key; 134020897f5cSArne Jansen int ret = 0; 13416463fe58SStefan Behrens uuid_le uuid; 134220897f5cSArne Jansen 134374e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_KERNEL); 134420897f5cSArne Jansen if (!root) 134520897f5cSArne Jansen return ERR_PTR(-ENOMEM); 134620897f5cSArne Jansen 1347707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1348707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, objectid); 134920897f5cSArne Jansen root->root_key.objectid = objectid; 135020897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 135120897f5cSArne Jansen root->root_key.offset = 0; 135220897f5cSArne Jansen 13534d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 135420897f5cSArne Jansen if (IS_ERR(leaf)) { 135520897f5cSArne Jansen ret = PTR_ERR(leaf); 13561dd05682STsutomu Itoh leaf = NULL; 135720897f5cSArne Jansen goto fail; 135820897f5cSArne Jansen } 135920897f5cSArne Jansen 136020897f5cSArne Jansen memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 136120897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 136220897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 136320897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 136420897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 136520897f5cSArne Jansen root->node = leaf; 136620897f5cSArne Jansen 13670a4e5586SRoss Kirk write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 136820897f5cSArne Jansen BTRFS_FSID_SIZE); 136920897f5cSArne Jansen write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 1370b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 137120897f5cSArne Jansen BTRFS_UUID_SIZE); 137220897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 137320897f5cSArne Jansen 137420897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 137527cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 137620897f5cSArne Jansen 137720897f5cSArne Jansen root->root_item.flags = 0; 137820897f5cSArne Jansen root->root_item.byte_limit = 0; 137920897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 138020897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 138120897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 138220897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 138320897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 138420897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 138520897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 13866463fe58SStefan Behrens uuid_le_gen(&uuid); 13876463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 138820897f5cSArne Jansen root->root_item.drop_level = 0; 138920897f5cSArne Jansen 139020897f5cSArne Jansen key.objectid = objectid; 139120897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 139220897f5cSArne Jansen key.offset = 0; 139320897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 139420897f5cSArne Jansen if (ret) 139520897f5cSArne Jansen goto fail; 139620897f5cSArne Jansen 139720897f5cSArne Jansen btrfs_tree_unlock(leaf); 139820897f5cSArne Jansen 139920897f5cSArne Jansen return root; 14001dd05682STsutomu Itoh 14011dd05682STsutomu Itoh fail: 14021dd05682STsutomu Itoh if (leaf) { 14031dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 140459885b39STsutomu Itoh free_extent_buffer(root->commit_root); 14051dd05682STsutomu Itoh free_extent_buffer(leaf); 14061dd05682STsutomu Itoh } 14071dd05682STsutomu Itoh kfree(root); 14081dd05682STsutomu Itoh 14091dd05682STsutomu Itoh return ERR_PTR(ret); 141020897f5cSArne Jansen } 141120897f5cSArne Jansen 14127237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1413e02119d5SChris Mason struct btrfs_fs_info *fs_info) 14140f7d52f4SChris Mason { 14150f7d52f4SChris Mason struct btrfs_root *root; 14160f7d52f4SChris Mason struct btrfs_root *tree_root = fs_info->tree_root; 14177237f183SYan Zheng struct extent_buffer *leaf; 1418e02119d5SChris Mason 141974e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1420e02119d5SChris Mason if (!root) 14217237f183SYan Zheng return ERR_PTR(-ENOMEM); 1422e02119d5SChris Mason 1423707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1424707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, 1425707e8a07SDavid Sterba BTRFS_TREE_LOG_OBJECTID); 1426e02119d5SChris Mason 1427e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1428e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1429e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 143027cdeb70SMiao Xie 14317237f183SYan Zheng /* 143227cdeb70SMiao Xie * DON'T set REF_COWS for log trees 143327cdeb70SMiao Xie * 14347237f183SYan Zheng * log trees do not get reference counted because they go away 14357237f183SYan Zheng * before a real commit is actually done. They do store pointers 14367237f183SYan Zheng * to file data extents, and those reference counts still get 14377237f183SYan Zheng * updated (along with back refs to the log tree). 14387237f183SYan Zheng */ 1439e02119d5SChris Mason 14404d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 14414d75f8a9SDavid Sterba NULL, 0, 0, 0); 14427237f183SYan Zheng if (IS_ERR(leaf)) { 14437237f183SYan Zheng kfree(root); 14447237f183SYan Zheng return ERR_CAST(leaf); 14457237f183SYan Zheng } 1446e02119d5SChris Mason 14475d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 14485d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 14495d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 14505d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 14515d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 14527237f183SYan Zheng root->node = leaf; 1453e02119d5SChris Mason 1454e02119d5SChris Mason write_extent_buffer(root->node, root->fs_info->fsid, 14550a4e5586SRoss Kirk btrfs_header_fsid(), BTRFS_FSID_SIZE); 1456e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1457e02119d5SChris Mason btrfs_tree_unlock(root->node); 14587237f183SYan Zheng return root; 14597237f183SYan Zheng } 14607237f183SYan Zheng 14617237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 14627237f183SYan Zheng struct btrfs_fs_info *fs_info) 14637237f183SYan Zheng { 14647237f183SYan Zheng struct btrfs_root *log_root; 14657237f183SYan Zheng 14667237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 14677237f183SYan Zheng if (IS_ERR(log_root)) 14687237f183SYan Zheng return PTR_ERR(log_root); 14697237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14707237f183SYan Zheng fs_info->log_root_tree = log_root; 14717237f183SYan Zheng return 0; 14727237f183SYan Zheng } 14737237f183SYan Zheng 14747237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14757237f183SYan Zheng struct btrfs_root *root) 14767237f183SYan Zheng { 14777237f183SYan Zheng struct btrfs_root *log_root; 14787237f183SYan Zheng struct btrfs_inode_item *inode_item; 14797237f183SYan Zheng 14807237f183SYan Zheng log_root = alloc_log_tree(trans, root->fs_info); 14817237f183SYan Zheng if (IS_ERR(log_root)) 14827237f183SYan Zheng return PTR_ERR(log_root); 14837237f183SYan Zheng 14847237f183SYan Zheng log_root->last_trans = trans->transid; 14857237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14867237f183SYan Zheng 14877237f183SYan Zheng inode_item = &log_root->root_item.inode; 14883cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14893cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14903cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1491707e8a07SDavid Sterba btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 14923cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14937237f183SYan Zheng 14945d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14957237f183SYan Zheng 14967237f183SYan Zheng WARN_ON(root->log_root); 14977237f183SYan Zheng root->log_root = log_root; 14987237f183SYan Zheng root->log_transid = 0; 1499d1433debSMiao Xie root->log_transid_committed = -1; 1500257c62e1SChris Mason root->last_log_commit = 0; 1501e02119d5SChris Mason return 0; 1502e02119d5SChris Mason } 1503e02119d5SChris Mason 150435a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1505cb517eabSMiao Xie struct btrfs_key *key) 1506e02119d5SChris Mason { 1507e02119d5SChris Mason struct btrfs_root *root; 1508e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 15090f7d52f4SChris Mason struct btrfs_path *path; 151084234f3aSYan Zheng u64 generation; 1511cb517eabSMiao Xie int ret; 1512cb517eabSMiao Xie 1513cb517eabSMiao Xie path = btrfs_alloc_path(); 1514cb517eabSMiao Xie if (!path) 1515cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 15160f7d52f4SChris Mason 151774e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1518cb517eabSMiao Xie if (!root) { 1519cb517eabSMiao Xie ret = -ENOMEM; 1520cb517eabSMiao Xie goto alloc_fail; 15210f7d52f4SChris Mason } 15220f7d52f4SChris Mason 1523707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1524707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, key->objectid); 15250f7d52f4SChris Mason 1526cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1527cb517eabSMiao Xie &root->root_item, &root->root_key); 15280f7d52f4SChris Mason if (ret) { 152913a8a7c8SYan, Zheng if (ret > 0) 153013a8a7c8SYan, Zheng ret = -ENOENT; 1531cb517eabSMiao Xie goto find_fail; 15320f7d52f4SChris Mason } 153313a8a7c8SYan, Zheng 153484234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1535db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 1536ce86cd59SDavid Sterba generation); 153764c043deSLiu Bo if (IS_ERR(root->node)) { 153864c043deSLiu Bo ret = PTR_ERR(root->node); 1539cb517eabSMiao Xie goto find_fail; 1540cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1541cb517eabSMiao Xie ret = -EIO; 154264c043deSLiu Bo free_extent_buffer(root->node); 154364c043deSLiu Bo goto find_fail; 1544416bc658SJosef Bacik } 15455d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 154613a8a7c8SYan, Zheng out: 1547cb517eabSMiao Xie btrfs_free_path(path); 1548cb517eabSMiao Xie return root; 1549cb517eabSMiao Xie 1550cb517eabSMiao Xie find_fail: 1551cb517eabSMiao Xie kfree(root); 1552cb517eabSMiao Xie alloc_fail: 1553cb517eabSMiao Xie root = ERR_PTR(ret); 1554cb517eabSMiao Xie goto out; 1555cb517eabSMiao Xie } 1556cb517eabSMiao Xie 1557cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1558cb517eabSMiao Xie struct btrfs_key *location) 1559cb517eabSMiao Xie { 1560cb517eabSMiao Xie struct btrfs_root *root; 1561cb517eabSMiao Xie 1562cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1563cb517eabSMiao Xie if (IS_ERR(root)) 1564cb517eabSMiao Xie return root; 1565cb517eabSMiao Xie 1566cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 156727cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 156808fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 156908fe4db1SLi Zefan } 157013a8a7c8SYan, Zheng 15715eda7b5eSChris Mason return root; 15725eda7b5eSChris Mason } 15735eda7b5eSChris Mason 1574cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1575cb517eabSMiao Xie { 1576cb517eabSMiao Xie int ret; 15778257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1578cb517eabSMiao Xie 1579cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1580cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1581cb517eabSMiao Xie GFP_NOFS); 1582cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1583cb517eabSMiao Xie ret = -ENOMEM; 1584cb517eabSMiao Xie goto fail; 1585cb517eabSMiao Xie } 1586cb517eabSMiao Xie 15878257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 15888257b2dcSMiao Xie if (IS_ERR(writers)) { 15898257b2dcSMiao Xie ret = PTR_ERR(writers); 15908257b2dcSMiao Xie goto fail; 15918257b2dcSMiao Xie } 15928257b2dcSMiao Xie root->subv_writers = writers; 15938257b2dcSMiao Xie 1594cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 159557cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 159657cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1597cb517eabSMiao Xie 1598cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1599cb517eabSMiao Xie if (ret) 16008257b2dcSMiao Xie goto free_writers; 1601f32e48e9SChandan Rajendra 1602f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1603f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1604f32e48e9SChandan Rajendra &root->highest_objectid); 1605f32e48e9SChandan Rajendra if (ret) { 1606f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1607f32e48e9SChandan Rajendra goto free_root_dev; 1608f32e48e9SChandan Rajendra } 1609f32e48e9SChandan Rajendra 1610f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1611f32e48e9SChandan Rajendra 1612f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1613f32e48e9SChandan Rajendra 1614cb517eabSMiao Xie return 0; 16158257b2dcSMiao Xie 1616f32e48e9SChandan Rajendra free_root_dev: 1617f32e48e9SChandan Rajendra free_anon_bdev(root->anon_dev); 16188257b2dcSMiao Xie free_writers: 16198257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 1620cb517eabSMiao Xie fail: 1621cb517eabSMiao Xie kfree(root->free_ino_ctl); 1622cb517eabSMiao Xie kfree(root->free_ino_pinned); 1623cb517eabSMiao Xie return ret; 1624cb517eabSMiao Xie } 1625cb517eabSMiao Xie 1626171170c1SSergei Trofimovich static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1627cb517eabSMiao Xie u64 root_id) 1628cb517eabSMiao Xie { 1629cb517eabSMiao Xie struct btrfs_root *root; 1630cb517eabSMiao Xie 1631cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1632cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1633cb517eabSMiao Xie (unsigned long)root_id); 1634cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1635cb517eabSMiao Xie return root; 1636cb517eabSMiao Xie } 1637cb517eabSMiao Xie 1638cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1639cb517eabSMiao Xie struct btrfs_root *root) 1640cb517eabSMiao Xie { 1641cb517eabSMiao Xie int ret; 1642cb517eabSMiao Xie 1643cb517eabSMiao Xie ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 1644cb517eabSMiao Xie if (ret) 1645cb517eabSMiao Xie return ret; 1646cb517eabSMiao Xie 1647cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1648cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1649cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1650cb517eabSMiao Xie root); 1651cb517eabSMiao Xie if (ret == 0) 165227cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1653cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1654cb517eabSMiao Xie radix_tree_preload_end(); 1655cb517eabSMiao Xie 1656cb517eabSMiao Xie return ret; 1657cb517eabSMiao Xie } 1658cb517eabSMiao Xie 1659c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1660c00869f1SMiao Xie struct btrfs_key *location, 1661c00869f1SMiao Xie bool check_ref) 16625eda7b5eSChris Mason { 16635eda7b5eSChris Mason struct btrfs_root *root; 1664381cf658SDavid Sterba struct btrfs_path *path; 16651d4c08e0SDavid Sterba struct btrfs_key key; 16665eda7b5eSChris Mason int ret; 16675eda7b5eSChris Mason 1668edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1669edbd8d4eSChris Mason return fs_info->tree_root; 1670edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1671edbd8d4eSChris Mason return fs_info->extent_root; 16728f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 16738f18cf13SChris Mason return fs_info->chunk_root; 16748f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 16758f18cf13SChris Mason return fs_info->dev_root; 16760403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 16770403e47eSYan Zheng return fs_info->csum_root; 1678bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1679bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1680bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1681f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1682f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1683f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 168470f6d82eSOmar Sandoval if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 168570f6d82eSOmar Sandoval return fs_info->free_space_root ? fs_info->free_space_root : 168670f6d82eSOmar Sandoval ERR_PTR(-ENOENT); 16874df27c4dSYan, Zheng again: 1688cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 168948475471SStefan Behrens if (root) { 1690c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 169148475471SStefan Behrens return ERR_PTR(-ENOENT); 16925eda7b5eSChris Mason return root; 169348475471SStefan Behrens } 16945eda7b5eSChris Mason 1695cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 16965eda7b5eSChris Mason if (IS_ERR(root)) 16975eda7b5eSChris Mason return root; 16983394e160SChris Mason 1699c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1700d68fc57bSYan, Zheng ret = -ENOENT; 1701d68fc57bSYan, Zheng goto fail; 1702d68fc57bSYan, Zheng } 1703d68fc57bSYan, Zheng 1704cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1705cb517eabSMiao Xie if (ret) 1706cb517eabSMiao Xie goto fail; 1707cb517eabSMiao Xie 1708381cf658SDavid Sterba path = btrfs_alloc_path(); 1709381cf658SDavid Sterba if (!path) { 1710381cf658SDavid Sterba ret = -ENOMEM; 1711381cf658SDavid Sterba goto fail; 1712381cf658SDavid Sterba } 17131d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 17141d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 17151d4c08e0SDavid Sterba key.offset = location->objectid; 17161d4c08e0SDavid Sterba 17171d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1718381cf658SDavid Sterba btrfs_free_path(path); 1719d68fc57bSYan, Zheng if (ret < 0) 1720d68fc57bSYan, Zheng goto fail; 1721d68fc57bSYan, Zheng if (ret == 0) 172227cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1723d68fc57bSYan, Zheng 1724cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17250f7d52f4SChris Mason if (ret) { 17264df27c4dSYan, Zheng if (ret == -EEXIST) { 17274df27c4dSYan, Zheng free_fs_root(root); 17284df27c4dSYan, Zheng goto again; 17290f7d52f4SChris Mason } 17304df27c4dSYan, Zheng goto fail; 17314df27c4dSYan, Zheng } 1732edbd8d4eSChris Mason return root; 17334df27c4dSYan, Zheng fail: 17344df27c4dSYan, Zheng free_fs_root(root); 17354df27c4dSYan, Zheng return ERR_PTR(ret); 1736edbd8d4eSChris Mason } 1737edbd8d4eSChris Mason 173804160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 173904160088SChris Mason { 174004160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 174104160088SChris Mason int ret = 0; 174204160088SChris Mason struct btrfs_device *device; 174304160088SChris Mason struct backing_dev_info *bdi; 1744b7967db7SChris Mason 17451f78160cSXiao Guangrong rcu_read_lock(); 17461f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1747dfe25020SChris Mason if (!device->bdev) 1748dfe25020SChris Mason continue; 174904160088SChris Mason bdi = blk_get_backing_dev_info(device->bdev); 1750ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 175104160088SChris Mason ret = 1; 175204160088SChris Mason break; 175304160088SChris Mason } 175404160088SChris Mason } 17551f78160cSXiao Guangrong rcu_read_unlock(); 175604160088SChris Mason return ret; 175704160088SChris Mason } 175804160088SChris Mason 175904160088SChris Mason static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 176004160088SChris Mason { 1761ad081f14SJens Axboe int err; 1762ad081f14SJens Axboe 1763b4caecd4SChristoph Hellwig err = bdi_setup_and_register(bdi, "btrfs"); 1764ad081f14SJens Axboe if (err) 1765ad081f14SJens Axboe return err; 1766ad081f14SJens Axboe 176709cbfeafSKirill A. Shutemov bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE; 176804160088SChris Mason bdi->congested_fn = btrfs_congested_fn; 176904160088SChris Mason bdi->congested_data = info; 1770da2f0f74SChris Mason bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 177104160088SChris Mason return 0; 177204160088SChris Mason } 177304160088SChris Mason 17748b712842SChris Mason /* 17758b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 17768b712842SChris Mason * functions. This is where read checksum verification actually happens 17778b712842SChris Mason */ 17788b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1779ce9adaa5SChris Mason { 1780ce9adaa5SChris Mason struct bio *bio; 178197eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1782ce9adaa5SChris Mason 178397eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1784ce9adaa5SChris Mason bio = end_io_wq->bio; 17858b712842SChris Mason 17864246a0b6SChristoph Hellwig bio->bi_error = end_io_wq->error; 1787ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1788ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 178997eb6b69SDavid Sterba kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 17904246a0b6SChristoph Hellwig bio_endio(bio); 1791ce9adaa5SChris Mason } 179244b8bd7eSChris Mason 1793a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1794a74a4b97SChris Mason { 1795a74a4b97SChris Mason struct btrfs_root *root = arg; 1796d0278245SMiao Xie int again; 1797da288d28SFilipe Manana struct btrfs_trans_handle *trans; 1798a74a4b97SChris Mason 1799a74a4b97SChris Mason do { 1800d0278245SMiao Xie again = 0; 18019d1a2a3aSDavid Sterba 1802d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 1803babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) 1804d0278245SMiao Xie goto sleep; 1805d0278245SMiao Xie 1806d0278245SMiao Xie if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1807d0278245SMiao Xie goto sleep; 1808d0278245SMiao Xie 1809dc7f370cSMiao Xie /* 1810dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1811dc7f370cSMiao Xie * during the above check and trylock. 1812dc7f370cSMiao Xie */ 1813babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) { 1814dc7f370cSMiao Xie mutex_unlock(&root->fs_info->cleaner_mutex); 1815dc7f370cSMiao Xie goto sleep; 1816dc7f370cSMiao Xie } 1817dc7f370cSMiao Xie 1818c2d6cb16SFilipe Manana mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); 181924bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 1820c2d6cb16SFilipe Manana mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); 1821c2d6cb16SFilipe Manana 18229d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 1823a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 1824a74a4b97SChris Mason 1825d0278245SMiao Xie /* 182605323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 182705323cd1SMiao Xie * needn't do anything special here. 1828d0278245SMiao Xie */ 1829d0278245SMiao Xie btrfs_run_defrag_inodes(root->fs_info); 183067c5e7d4SFilipe Manana 183167c5e7d4SFilipe Manana /* 183267c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 183367c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 183467c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 183567c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 183667c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 183767c5e7d4SFilipe Manana * unused block groups. 183867c5e7d4SFilipe Manana */ 183967c5e7d4SFilipe Manana btrfs_delete_unused_bgs(root->fs_info); 1840d0278245SMiao Xie sleep: 1841838fe188SJiri Kosina if (!again) { 1842a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 18438929ecfaSYan, Zheng if (!kthread_should_stop()) 1844a74a4b97SChris Mason schedule(); 1845a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1846a74a4b97SChris Mason } 1847a74a4b97SChris Mason } while (!kthread_should_stop()); 1848da288d28SFilipe Manana 1849da288d28SFilipe Manana /* 1850da288d28SFilipe Manana * Transaction kthread is stopped before us and wakes us up. 1851da288d28SFilipe Manana * However we might have started a new transaction and COWed some 1852da288d28SFilipe Manana * tree blocks when deleting unused block groups for example. So 1853da288d28SFilipe Manana * make sure we commit the transaction we started to have a clean 1854da288d28SFilipe Manana * shutdown when evicting the btree inode - if it has dirty pages 1855da288d28SFilipe Manana * when we do the final iput() on it, eviction will trigger a 1856da288d28SFilipe Manana * writeback for it which will fail with null pointer dereferences 1857da288d28SFilipe Manana * since work queues and other resources were already released and 1858da288d28SFilipe Manana * destroyed by the time the iput/eviction/writeback is made. 1859da288d28SFilipe Manana */ 1860da288d28SFilipe Manana trans = btrfs_attach_transaction(root); 1861da288d28SFilipe Manana if (IS_ERR(trans)) { 1862da288d28SFilipe Manana if (PTR_ERR(trans) != -ENOENT) 1863da288d28SFilipe Manana btrfs_err(root->fs_info, 1864da288d28SFilipe Manana "cleaner transaction attach returned %ld", 1865da288d28SFilipe Manana PTR_ERR(trans)); 1866da288d28SFilipe Manana } else { 1867da288d28SFilipe Manana int ret; 1868da288d28SFilipe Manana 1869da288d28SFilipe Manana ret = btrfs_commit_transaction(trans, root); 1870da288d28SFilipe Manana if (ret) 1871da288d28SFilipe Manana btrfs_err(root->fs_info, 1872da288d28SFilipe Manana "cleaner open transaction commit returned %d", 1873da288d28SFilipe Manana ret); 1874da288d28SFilipe Manana } 1875da288d28SFilipe Manana 1876a74a4b97SChris Mason return 0; 1877a74a4b97SChris Mason } 1878a74a4b97SChris Mason 1879a74a4b97SChris Mason static int transaction_kthread(void *arg) 1880a74a4b97SChris Mason { 1881a74a4b97SChris Mason struct btrfs_root *root = arg; 1882a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1883a74a4b97SChris Mason struct btrfs_transaction *cur; 18848929ecfaSYan, Zheng u64 transid; 1885a74a4b97SChris Mason unsigned long now; 1886a74a4b97SChris Mason unsigned long delay; 1887914b2007SJan Kara bool cannot_commit; 1888a74a4b97SChris Mason 1889a74a4b97SChris Mason do { 1890914b2007SJan Kara cannot_commit = false; 18918b87dc17SDavid Sterba delay = HZ * root->fs_info->commit_interval; 1892a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 1893a74a4b97SChris Mason 1894a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 1895a74a4b97SChris Mason cur = root->fs_info->running_transaction; 1896a74a4b97SChris Mason if (!cur) { 1897a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1898a74a4b97SChris Mason goto sleep; 1899a74a4b97SChris Mason } 190031153d81SYan Zheng 1901a74a4b97SChris Mason now = get_seconds(); 19024a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 19038b87dc17SDavid Sterba (now < cur->start_time || 19048b87dc17SDavid Sterba now - cur->start_time < root->fs_info->commit_interval)) { 1905a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1906a74a4b97SChris Mason delay = HZ * 5; 1907a74a4b97SChris Mason goto sleep; 1908a74a4b97SChris Mason } 19098929ecfaSYan, Zheng transid = cur->transid; 1910a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 191156bec294SChris Mason 191279787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1913354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1914914b2007SJan Kara if (IS_ERR(trans)) { 1915354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1916914b2007SJan Kara cannot_commit = true; 191779787eaaSJeff Mahoney goto sleep; 1918914b2007SJan Kara } 19198929ecfaSYan, Zheng if (transid == trans->transid) { 192079787eaaSJeff Mahoney btrfs_commit_transaction(trans, root); 19218929ecfaSYan, Zheng } else { 19228929ecfaSYan, Zheng btrfs_end_transaction(trans, root); 19238929ecfaSYan, Zheng } 1924a74a4b97SChris Mason sleep: 1925a74a4b97SChris Mason wake_up_process(root->fs_info->cleaner_kthread); 1926a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1927a74a4b97SChris Mason 19284e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 19294e121c06SJosef Bacik &root->fs_info->fs_state))) 19304e121c06SJosef Bacik btrfs_cleanup_transaction(root); 1931a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 19328929ecfaSYan, Zheng if (!kthread_should_stop() && 1933914b2007SJan Kara (!btrfs_transaction_blocked(root->fs_info) || 1934914b2007SJan Kara cannot_commit)) 1935a74a4b97SChris Mason schedule_timeout(delay); 1936a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1937a74a4b97SChris Mason } while (!kthread_should_stop()); 1938a74a4b97SChris Mason return 0; 1939a74a4b97SChris Mason } 1940a74a4b97SChris Mason 1941af31f5e5SChris Mason /* 1942af31f5e5SChris Mason * this will find the highest generation in the array of 1943af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1944af31f5e5SChris Mason * or -1 if we can't find anything. 1945af31f5e5SChris Mason * 1946af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1947af31f5e5SChris Mason * generation of the latest root in the array with the generation 1948af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1949af31f5e5SChris Mason */ 1950af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1951af31f5e5SChris Mason { 1952af31f5e5SChris Mason u64 cur; 1953af31f5e5SChris Mason int newest_index = -1; 1954af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1955af31f5e5SChris Mason int i; 1956af31f5e5SChris Mason 1957af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1958af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1959af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1960af31f5e5SChris Mason if (cur == newest_gen) 1961af31f5e5SChris Mason newest_index = i; 1962af31f5e5SChris Mason } 1963af31f5e5SChris Mason 1964af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1965af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1966af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1967af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1968af31f5e5SChris Mason if (cur == newest_gen) 1969af31f5e5SChris Mason newest_index = 0; 1970af31f5e5SChris Mason } 1971af31f5e5SChris Mason return newest_index; 1972af31f5e5SChris Mason } 1973af31f5e5SChris Mason 1974af31f5e5SChris Mason 1975af31f5e5SChris Mason /* 1976af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1977af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1978af31f5e5SChris Mason * field in the fs_info struct 1979af31f5e5SChris Mason */ 1980af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1981af31f5e5SChris Mason u64 newest_gen) 1982af31f5e5SChris Mason { 1983af31f5e5SChris Mason int newest_index = -1; 1984af31f5e5SChris Mason 1985af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1986af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1987af31f5e5SChris Mason if (newest_index == -1) { 1988af31f5e5SChris Mason info->backup_root_index = 0; 1989af31f5e5SChris Mason } else { 1990af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1991af31f5e5SChris Mason } 1992af31f5e5SChris Mason } 1993af31f5e5SChris Mason 1994af31f5e5SChris Mason /* 1995af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1996af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1997af31f5e5SChris Mason * done 1998af31f5e5SChris Mason */ 1999af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 2000af31f5e5SChris Mason { 2001af31f5e5SChris Mason int next_backup; 2002af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2003af31f5e5SChris Mason int last_backup; 2004af31f5e5SChris Mason 2005af31f5e5SChris Mason next_backup = info->backup_root_index; 2006af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 2007af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2008af31f5e5SChris Mason 2009af31f5e5SChris Mason /* 2010af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 2011af31f5e5SChris Mason * this happens only at umount 2012af31f5e5SChris Mason */ 2013af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 2014af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 2015af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 2016af31f5e5SChris Mason next_backup = last_backup; 2017af31f5e5SChris Mason 2018af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 2019af31f5e5SChris Mason 2020af31f5e5SChris Mason /* 2021af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 2022af31f5e5SChris Mason * regardless of which ones we use today 2023af31f5e5SChris Mason */ 2024af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 2025af31f5e5SChris Mason 2026af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 2027af31f5e5SChris Mason 2028af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 2029af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 2030af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 2031af31f5e5SChris Mason 2032af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 2033af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 2034af31f5e5SChris Mason 2035af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 2036af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 2037af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 2038af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 2039af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 2040af31f5e5SChris Mason 2041af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 2042af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 2043af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 2044af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 2045af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 2046af31f5e5SChris Mason 20477c7e82a7SChris Mason /* 20487c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20497c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20507c7e82a7SChris Mason */ 20517c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20527c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20537c7e82a7SChris Mason info->fs_root->node->start); 2054af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2055af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2056af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2057af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20587c7e82a7SChris Mason } 2059af31f5e5SChris Mason 2060af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2061af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2062af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2063af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2064af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2065af31f5e5SChris Mason 2066af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 2067af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 2068af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 2069af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 2070af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 2071af31f5e5SChris Mason 2072af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2073af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2074af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2075af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2076af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2077af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2078af31f5e5SChris Mason 2079af31f5e5SChris Mason /* 2080af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2081af31f5e5SChris Mason * for the next commit 2082af31f5e5SChris Mason */ 2083af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2084af31f5e5SChris Mason &info->super_for_commit->super_roots, 2085af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2086af31f5e5SChris Mason } 2087af31f5e5SChris Mason 2088af31f5e5SChris Mason /* 2089af31f5e5SChris Mason * this copies info out of the root backup array and back into 2090af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 2091af31f5e5SChris Mason * the array, so you send it the number of backups you've already 2092af31f5e5SChris Mason * tried and the last backup index you used. 2093af31f5e5SChris Mason * 2094af31f5e5SChris Mason * this returns -1 when it has tried all the backups 2095af31f5e5SChris Mason */ 2096af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 2097af31f5e5SChris Mason struct btrfs_super_block *super, 2098af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 2099af31f5e5SChris Mason { 2100af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2101af31f5e5SChris Mason int newest = *backup_index; 2102af31f5e5SChris Mason 2103af31f5e5SChris Mason if (*num_backups_tried == 0) { 2104af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 2105af31f5e5SChris Mason 2106af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 2107af31f5e5SChris Mason if (newest == -1) 2108af31f5e5SChris Mason return -1; 2109af31f5e5SChris Mason 2110af31f5e5SChris Mason *backup_index = newest; 2111af31f5e5SChris Mason *num_backups_tried = 1; 2112af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 2113af31f5e5SChris Mason /* we've tried all the backups, all done */ 2114af31f5e5SChris Mason return -1; 2115af31f5e5SChris Mason } else { 2116af31f5e5SChris Mason /* jump to the next oldest backup */ 2117af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 2118af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2119af31f5e5SChris Mason *backup_index = newest; 2120af31f5e5SChris Mason *num_backups_tried += 1; 2121af31f5e5SChris Mason } 2122af31f5e5SChris Mason root_backup = super->super_roots + newest; 2123af31f5e5SChris Mason 2124af31f5e5SChris Mason btrfs_set_super_generation(super, 2125af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2126af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2127af31f5e5SChris Mason btrfs_set_super_root_level(super, 2128af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2129af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2130af31f5e5SChris Mason 2131af31f5e5SChris Mason /* 2132af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2133af31f5e5SChris Mason * need a fsck 2134af31f5e5SChris Mason */ 2135af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2136af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2137af31f5e5SChris Mason return 0; 2138af31f5e5SChris Mason } 2139af31f5e5SChris Mason 21407abadb64SLiu Bo /* helper to cleanup workers */ 21417abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 21427abadb64SLiu Bo { 2143dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2144afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 21455cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2146fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2147fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2148fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 21498b110e39SMiao Xie btrfs_destroy_workqueue(fs_info->endio_repair_workers); 2150d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2151fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 2152fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2153fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2154a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 21555b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2156e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2157736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2158a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2159fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2160a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 21617abadb64SLiu Bo } 21627abadb64SLiu Bo 21632e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21642e9f5954SRashika { 21652e9f5954SRashika if (root) { 21662e9f5954SRashika free_extent_buffer(root->node); 21672e9f5954SRashika free_extent_buffer(root->commit_root); 21682e9f5954SRashika root->node = NULL; 21692e9f5954SRashika root->commit_root = NULL; 21702e9f5954SRashika } 21712e9f5954SRashika } 21722e9f5954SRashika 2173af31f5e5SChris Mason /* helper to cleanup tree roots */ 2174af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2175af31f5e5SChris Mason { 21762e9f5954SRashika free_root_extent_buffers(info->tree_root); 2177655b09feSJosef Bacik 21782e9f5954SRashika free_root_extent_buffers(info->dev_root); 21792e9f5954SRashika free_root_extent_buffers(info->extent_root); 21802e9f5954SRashika free_root_extent_buffers(info->csum_root); 21812e9f5954SRashika free_root_extent_buffers(info->quota_root); 21822e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21832e9f5954SRashika if (chunk_root) 21842e9f5954SRashika free_root_extent_buffers(info->chunk_root); 218570f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2186af31f5e5SChris Mason } 2187af31f5e5SChris Mason 2188faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2189171f6537SJosef Bacik { 2190171f6537SJosef Bacik int ret; 2191171f6537SJosef Bacik struct btrfs_root *gang[8]; 2192171f6537SJosef Bacik int i; 2193171f6537SJosef Bacik 2194171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2195171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2196171f6537SJosef Bacik struct btrfs_root, root_list); 2197171f6537SJosef Bacik list_del(&gang[0]->root_list); 2198171f6537SJosef Bacik 219927cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2200cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2201171f6537SJosef Bacik } else { 2202171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2203171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2204b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2205171f6537SJosef Bacik } 2206171f6537SJosef Bacik } 2207171f6537SJosef Bacik 2208171f6537SJosef Bacik while (1) { 2209171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2210171f6537SJosef Bacik (void **)gang, 0, 2211171f6537SJosef Bacik ARRAY_SIZE(gang)); 2212171f6537SJosef Bacik if (!ret) 2213171f6537SJosef Bacik break; 2214171f6537SJosef Bacik for (i = 0; i < ret; i++) 2215cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2216171f6537SJosef Bacik } 22171a4319ccSLiu Bo 22181a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 22191a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 22201a4319ccSLiu Bo btrfs_destroy_pinned_extent(fs_info->tree_root, 22211a4319ccSLiu Bo fs_info->pinned_extents); 22221a4319ccSLiu Bo } 2223171f6537SJosef Bacik } 2224af31f5e5SChris Mason 2225638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2226638aa7edSEric Sandeen { 2227638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2228638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2229638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2230638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2231638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2232638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2233638aa7edSEric Sandeen fs_info->scrub_workers_refcnt = 0; 2234638aa7edSEric Sandeen } 2235638aa7edSEric Sandeen 2236779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2237779a65a4SEric Sandeen { 2238779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2239779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2240779a65a4SEric Sandeen atomic_set(&fs_info->balance_running, 0); 2241779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2242779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2243779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2244779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2245779a65a4SEric Sandeen } 2246779a65a4SEric Sandeen 2247f37938e0SEric Sandeen static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, 2248f37938e0SEric Sandeen struct btrfs_root *tree_root) 2249f37938e0SEric Sandeen { 2250f37938e0SEric Sandeen fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2251f37938e0SEric Sandeen set_nlink(fs_info->btree_inode, 1); 2252f37938e0SEric Sandeen /* 2253f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2254f37938e0SEric Sandeen * the real end of the address space is determined by all of 2255f37938e0SEric Sandeen * the devices in the system 2256f37938e0SEric Sandeen */ 2257f37938e0SEric Sandeen fs_info->btree_inode->i_size = OFFSET_MAX; 2258f37938e0SEric Sandeen fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 2259f37938e0SEric Sandeen 2260f37938e0SEric Sandeen RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2261f37938e0SEric Sandeen extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2262f37938e0SEric Sandeen fs_info->btree_inode->i_mapping); 2263f37938e0SEric Sandeen BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2264f37938e0SEric Sandeen extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 2265f37938e0SEric Sandeen 2266f37938e0SEric Sandeen BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2267f37938e0SEric Sandeen 2268f37938e0SEric Sandeen BTRFS_I(fs_info->btree_inode)->root = tree_root; 2269f37938e0SEric Sandeen memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 2270f37938e0SEric Sandeen sizeof(struct btrfs_key)); 2271f37938e0SEric Sandeen set_bit(BTRFS_INODE_DUMMY, 2272f37938e0SEric Sandeen &BTRFS_I(fs_info->btree_inode)->runtime_flags); 2273f37938e0SEric Sandeen btrfs_insert_inode_hash(fs_info->btree_inode); 2274f37938e0SEric Sandeen } 2275f37938e0SEric Sandeen 2276ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2277ad618368SEric Sandeen { 2278ad618368SEric Sandeen fs_info->dev_replace.lock_owner = 0; 2279ad618368SEric Sandeen atomic_set(&fs_info->dev_replace.nesting_level, 0); 2280ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 228173beece9SLiu Bo rwlock_init(&fs_info->dev_replace.lock); 228273beece9SLiu Bo atomic_set(&fs_info->dev_replace.read_locks, 0); 228373beece9SLiu Bo atomic_set(&fs_info->dev_replace.blocking_readers, 0); 2284ad618368SEric Sandeen init_waitqueue_head(&fs_info->replace_wait); 228573beece9SLiu Bo init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); 2286ad618368SEric Sandeen } 2287ad618368SEric Sandeen 2288f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2289f9e92e40SEric Sandeen { 2290f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2291f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2292f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2293f9e92e40SEric Sandeen fs_info->qgroup_op_tree = RB_ROOT; 2294f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2295f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2296f9e92e40SEric Sandeen fs_info->quota_enabled = 0; 2297f9e92e40SEric Sandeen fs_info->pending_quota_state = 0; 2298f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2299f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2300f9e92e40SEric Sandeen } 2301f9e92e40SEric Sandeen 23022a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 23032a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 23042a458198SEric Sandeen { 23052a458198SEric Sandeen int max_active = fs_info->thread_pool_size; 23066f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 23072a458198SEric Sandeen 23082a458198SEric Sandeen fs_info->workers = 23092a458198SEric Sandeen btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, 23102a458198SEric Sandeen max_active, 16); 23112a458198SEric Sandeen 23122a458198SEric Sandeen fs_info->delalloc_workers = 23132a458198SEric Sandeen btrfs_alloc_workqueue("delalloc", flags, max_active, 2); 23142a458198SEric Sandeen 23152a458198SEric Sandeen fs_info->flush_workers = 23162a458198SEric Sandeen btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); 23172a458198SEric Sandeen 23182a458198SEric Sandeen fs_info->caching_workers = 23192a458198SEric Sandeen btrfs_alloc_workqueue("cache", flags, max_active, 0); 23202a458198SEric Sandeen 23212a458198SEric Sandeen /* 23222a458198SEric Sandeen * a higher idle thresh on the submit workers makes it much more 23232a458198SEric Sandeen * likely that bios will be send down in a sane order to the 23242a458198SEric Sandeen * devices 23252a458198SEric Sandeen */ 23262a458198SEric Sandeen fs_info->submit_workers = 23272a458198SEric Sandeen btrfs_alloc_workqueue("submit", flags, 23282a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 23292a458198SEric Sandeen max_active), 64); 23302a458198SEric Sandeen 23312a458198SEric Sandeen fs_info->fixup_workers = 23322a458198SEric Sandeen btrfs_alloc_workqueue("fixup", flags, 1, 0); 23332a458198SEric Sandeen 23342a458198SEric Sandeen /* 23352a458198SEric Sandeen * endios are largely parallel and should have a very 23362a458198SEric Sandeen * low idle thresh 23372a458198SEric Sandeen */ 23382a458198SEric Sandeen fs_info->endio_workers = 23392a458198SEric Sandeen btrfs_alloc_workqueue("endio", flags, max_active, 4); 23402a458198SEric Sandeen fs_info->endio_meta_workers = 23412a458198SEric Sandeen btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); 23422a458198SEric Sandeen fs_info->endio_meta_write_workers = 23432a458198SEric Sandeen btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); 23442a458198SEric Sandeen fs_info->endio_raid56_workers = 23452a458198SEric Sandeen btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); 23462a458198SEric Sandeen fs_info->endio_repair_workers = 23472a458198SEric Sandeen btrfs_alloc_workqueue("endio-repair", flags, 1, 0); 23482a458198SEric Sandeen fs_info->rmw_workers = 23492a458198SEric Sandeen btrfs_alloc_workqueue("rmw", flags, max_active, 2); 23502a458198SEric Sandeen fs_info->endio_write_workers = 23512a458198SEric Sandeen btrfs_alloc_workqueue("endio-write", flags, max_active, 2); 23522a458198SEric Sandeen fs_info->endio_freespace_worker = 23532a458198SEric Sandeen btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); 23542a458198SEric Sandeen fs_info->delayed_workers = 23552a458198SEric Sandeen btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); 23562a458198SEric Sandeen fs_info->readahead_workers = 23572a458198SEric Sandeen btrfs_alloc_workqueue("readahead", flags, max_active, 2); 23582a458198SEric Sandeen fs_info->qgroup_rescan_workers = 23592a458198SEric Sandeen btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); 23602a458198SEric Sandeen fs_info->extent_workers = 23612a458198SEric Sandeen btrfs_alloc_workqueue("extent-refs", flags, 23622a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 23632a458198SEric Sandeen max_active), 8); 23642a458198SEric Sandeen 23652a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 23662a458198SEric Sandeen fs_info->submit_workers && fs_info->flush_workers && 23672a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 23682a458198SEric Sandeen fs_info->endio_meta_write_workers && 23692a458198SEric Sandeen fs_info->endio_repair_workers && 23702a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23712a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 23722a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 23732a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 23742a458198SEric Sandeen fs_info->extent_workers && 23752a458198SEric Sandeen fs_info->qgroup_rescan_workers)) { 23762a458198SEric Sandeen return -ENOMEM; 23772a458198SEric Sandeen } 23782a458198SEric Sandeen 23792a458198SEric Sandeen return 0; 23802a458198SEric Sandeen } 23812a458198SEric Sandeen 238263443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 238363443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 238463443bf5SEric Sandeen { 238563443bf5SEric Sandeen int ret; 238663443bf5SEric Sandeen struct btrfs_root *tree_root = fs_info->tree_root; 238763443bf5SEric Sandeen struct btrfs_root *log_tree_root; 238863443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 238963443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 239063443bf5SEric Sandeen 239163443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2392f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 239363443bf5SEric Sandeen return -EIO; 239463443bf5SEric Sandeen } 239563443bf5SEric Sandeen 239674e4d827SDavid Sterba log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 239763443bf5SEric Sandeen if (!log_tree_root) 239863443bf5SEric Sandeen return -ENOMEM; 239963443bf5SEric Sandeen 240063443bf5SEric Sandeen __setup_root(tree_root->nodesize, tree_root->sectorsize, 240163443bf5SEric Sandeen tree_root->stripesize, log_tree_root, fs_info, 240263443bf5SEric Sandeen BTRFS_TREE_LOG_OBJECTID); 240363443bf5SEric Sandeen 240463443bf5SEric Sandeen log_tree_root->node = read_tree_block(tree_root, bytenr, 240563443bf5SEric Sandeen fs_info->generation + 1); 240664c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2407f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 24080eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 240964c043deSLiu Bo kfree(log_tree_root); 24100eeff236SLiu Bo return ret; 241164c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2412f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 241363443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 241463443bf5SEric Sandeen kfree(log_tree_root); 241563443bf5SEric Sandeen return -EIO; 241663443bf5SEric Sandeen } 241763443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 241863443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 241963443bf5SEric Sandeen if (ret) { 2420a4553fefSAnand Jain btrfs_std_error(tree_root->fs_info, ret, 242163443bf5SEric Sandeen "Failed to recover log tree"); 242263443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 242363443bf5SEric Sandeen kfree(log_tree_root); 242463443bf5SEric Sandeen return ret; 242563443bf5SEric Sandeen } 242663443bf5SEric Sandeen 242763443bf5SEric Sandeen if (fs_info->sb->s_flags & MS_RDONLY) { 242863443bf5SEric Sandeen ret = btrfs_commit_super(tree_root); 242963443bf5SEric Sandeen if (ret) 243063443bf5SEric Sandeen return ret; 243163443bf5SEric Sandeen } 243263443bf5SEric Sandeen 243363443bf5SEric Sandeen return 0; 243463443bf5SEric Sandeen } 243563443bf5SEric Sandeen 24364bbcaa64SEric Sandeen static int btrfs_read_roots(struct btrfs_fs_info *fs_info, 24374bbcaa64SEric Sandeen struct btrfs_root *tree_root) 24384bbcaa64SEric Sandeen { 2439a4f3d2c4SDavid Sterba struct btrfs_root *root; 24404bbcaa64SEric Sandeen struct btrfs_key location; 24414bbcaa64SEric Sandeen int ret; 24424bbcaa64SEric Sandeen 24434bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 24444bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 24454bbcaa64SEric Sandeen location.offset = 0; 24464bbcaa64SEric Sandeen 2447a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2448a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2449a4f3d2c4SDavid Sterba return PTR_ERR(root); 2450a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2451a4f3d2c4SDavid Sterba fs_info->extent_root = root; 24524bbcaa64SEric Sandeen 24534bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_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->dev_root = root; 24594bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 24604bbcaa64SEric Sandeen 24614bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2462a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2463a4f3d2c4SDavid Sterba if (IS_ERR(root)) 2464a4f3d2c4SDavid Sterba return PTR_ERR(root); 2465a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2466a4f3d2c4SDavid Sterba fs_info->csum_root = root; 24674bbcaa64SEric Sandeen 24684bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2469a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2470a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2471a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 24724bbcaa64SEric Sandeen fs_info->quota_enabled = 1; 24734bbcaa64SEric Sandeen fs_info->pending_quota_state = 1; 2474a4f3d2c4SDavid Sterba fs_info->quota_root = root; 24754bbcaa64SEric Sandeen } 24764bbcaa64SEric Sandeen 24774bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2478a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2479a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2480a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 24814bbcaa64SEric Sandeen if (ret != -ENOENT) 24824bbcaa64SEric Sandeen return ret; 24834bbcaa64SEric Sandeen } else { 2484a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2485a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 24864bbcaa64SEric Sandeen } 24874bbcaa64SEric Sandeen 248870f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 248970f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 249070f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 249170f6d82eSOmar Sandoval if (IS_ERR(root)) 249270f6d82eSOmar Sandoval return PTR_ERR(root); 249370f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 249470f6d82eSOmar Sandoval fs_info->free_space_root = root; 249570f6d82eSOmar Sandoval } 249670f6d82eSOmar Sandoval 24974bbcaa64SEric Sandeen return 0; 24984bbcaa64SEric Sandeen } 24994bbcaa64SEric Sandeen 2500ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2501dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2502dfe25020SChris Mason char *options) 2503eb60ceacSChris Mason { 2504db94535dSChris Mason u32 sectorsize; 2505db94535dSChris Mason u32 nodesize; 250687ee04ebSChris Mason u32 stripesize; 250784234f3aSYan Zheng u64 generation; 2508f2b636e8SJosef Bacik u64 features; 25093de4586cSChris Mason struct btrfs_key location; 2510a061fc8dSChris Mason struct buffer_head *bh; 25114d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2512815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2513f84a8bd6SAl Viro struct btrfs_root *tree_root; 25144d34b278SIlya Dryomov struct btrfs_root *chunk_root; 2515eb60ceacSChris Mason int ret; 2516e58ca020SYan int err = -EINVAL; 2517af31f5e5SChris Mason int num_backups_tried = 0; 2518af31f5e5SChris Mason int backup_index = 0; 25195cdc7ad3SQu Wenruo int max_active; 25204543df7eSChris Mason 252174e4d827SDavid Sterba tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 252274e4d827SDavid Sterba chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 2523cb517eabSMiao Xie if (!tree_root || !chunk_root) { 252439279cc3SChris Mason err = -ENOMEM; 252539279cc3SChris Mason goto fail; 252639279cc3SChris Mason } 252776dda93cSYan, Zheng 252876dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 252976dda93cSYan, Zheng if (ret) { 253076dda93cSYan, Zheng err = ret; 253176dda93cSYan, Zheng goto fail; 253276dda93cSYan, Zheng } 253376dda93cSYan, Zheng 253476dda93cSYan, Zheng ret = setup_bdi(fs_info, &fs_info->bdi); 253576dda93cSYan, Zheng if (ret) { 253676dda93cSYan, Zheng err = ret; 253776dda93cSYan, Zheng goto fail_srcu; 253876dda93cSYan, Zheng } 253976dda93cSYan, Zheng 2540908c7f19STejun Heo ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2541e2d84521SMiao Xie if (ret) { 2542e2d84521SMiao Xie err = ret; 2543e2d84521SMiao Xie goto fail_bdi; 2544e2d84521SMiao Xie } 254509cbfeafSKirill A. Shutemov fs_info->dirty_metadata_batch = PAGE_SIZE * 2546e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2547e2d84521SMiao Xie 2548908c7f19STejun Heo ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2549963d678bSMiao Xie if (ret) { 2550963d678bSMiao Xie err = ret; 2551963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2552963d678bSMiao Xie } 2553963d678bSMiao Xie 2554908c7f19STejun Heo ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); 2555c404e0dcSMiao Xie if (ret) { 2556c404e0dcSMiao Xie err = ret; 2557c404e0dcSMiao Xie goto fail_delalloc_bytes; 2558c404e0dcSMiao Xie } 2559c404e0dcSMiao Xie 256076dda93cSYan, Zheng fs_info->btree_inode = new_inode(sb); 256176dda93cSYan, Zheng if (!fs_info->btree_inode) { 256276dda93cSYan, Zheng err = -ENOMEM; 2563c404e0dcSMiao Xie goto fail_bio_counter; 256476dda93cSYan, Zheng } 256576dda93cSYan, Zheng 2566a6591715SChris Mason mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 25671561dedaSMiao Xie 256876dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2569f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 25708fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2571facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 257224bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2573eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 257411833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2575eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2576a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 257776dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 257824bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 25794cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 25802bf64758SJosef Bacik spin_lock_init(&fs_info->free_chunk_lock); 2581f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2582ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2583fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2584f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 258547ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2586f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2587d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 258867c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 25897585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2590573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2591c2d6cb16SFilipe Manana mutex_init(&fs_info->cleaner_delayed_iput_mutex); 2592de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 259319c00ddcSChris Mason 25940b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 25956324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2596f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 259747ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 25980b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 259966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 260066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 260166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 260266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELALLOC); 260366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 260466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 260566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 260666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 260766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2608cb03c743SChris Mason atomic_set(&fs_info->nr_async_submits, 0); 2609771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 26108c8bee1dSChris Mason atomic_set(&fs_info->async_submit_draining, 0); 26110986fe9eSChris Mason atomic_set(&fs_info->nr_async_bios, 0); 26124cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2613fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 26142fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2615fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2616e20d96d6SChris Mason fs_info->sb = sb; 261795ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 26189ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 26194cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 26202bf64758SJosef Bacik fs_info->free_chunk_space = 0; 2621f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 26228b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2623f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 262490519d66SArne Jansen /* readahead state */ 2625d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 262690519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2627c8b97818SChris Mason 2628b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2629b34b086cSChris Mason num_online_cpus() + 2, 8); 26300afbaf8cSChris Mason 2631199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2632199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 263316cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 263474e4d827SDavid Sterba GFP_KERNEL); 263516cdcec7SMiao Xie if (!fs_info->delayed_root) { 263616cdcec7SMiao Xie err = -ENOMEM; 263716cdcec7SMiao Xie goto fail_iput; 263816cdcec7SMiao Xie } 263916cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 26403eaa2885SChris Mason 2641638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 264221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 264321adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 264421adbd5cSStefan Behrens #endif 2645779a65a4SEric Sandeen btrfs_init_balance(fs_info); 264621c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2647a2de733cSArne Jansen 2648a061fc8dSChris Mason sb->s_blocksize = 4096; 2649a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(4096); 265032a88aa1SJens Axboe sb->s_bdi = &fs_info->bdi; 2651a061fc8dSChris Mason 2652f37938e0SEric Sandeen btrfs_init_btree_inode(fs_info, tree_root); 265339279cc3SChris Mason 2654e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 26556bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2656a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2657925baeddSChris Mason 265811833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[0], 2659f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 266011833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[1], 2661f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 266211833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2663e6dcd2dcSChris Mason fs_info->do_barriers = 1; 2664f9295749SChris Mason 2665d98237b3SChris Mason 26665a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 266730ae8467SChris Mason mutex_init(&fs_info->tree_log_mutex); 2668925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2669925baeddSChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2670a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 2671a74a4b97SChris Mason mutex_init(&fs_info->volume_mutex); 26721bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 26739e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2674c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 267576dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2676803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2677fa9c0d79SChris Mason 2678ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2679f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2680416ac51dSArne Jansen 2681fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2682fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2683fa9c0d79SChris Mason 26847d9eb12cSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 26853768f368SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2686bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 26874854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 26889a8dd150SChris Mason 268904216820SFilipe Manana INIT_LIST_HEAD(&fs_info->pinned_chunks); 269004216820SFilipe Manana 269153b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 269253b381b3SDavid Woodhouse if (ret) { 269383c8266aSDavid Sterba err = ret; 269453b381b3SDavid Woodhouse goto fail_alloc; 269553b381b3SDavid Woodhouse } 269653b381b3SDavid Woodhouse 2697707e8a07SDavid Sterba __setup_root(4096, 4096, 4096, tree_root, 26982c90e5d6SChris Mason fs_info, BTRFS_ROOT_TREE_OBJECTID); 26997eccb903SChris Mason 27003c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 27011104a885SDavid Sterba 27021104a885SDavid Sterba /* 27031104a885SDavid Sterba * Read super block and check the signature bytes only 27041104a885SDavid Sterba */ 2705a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 270692fc03fbSAnand Jain if (IS_ERR(bh)) { 270792fc03fbSAnand Jain err = PTR_ERR(bh); 270816cdcec7SMiao Xie goto fail_alloc; 270920b45077SDave Young } 271039279cc3SChris Mason 27111104a885SDavid Sterba /* 27121104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 27131104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 27141104a885SDavid Sterba */ 27151104a885SDavid Sterba if (btrfs_check_super_csum(bh->b_data)) { 2716efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); 27171104a885SDavid Sterba err = -EINVAL; 2718b2acdddfSAnand Jain brelse(bh); 27191104a885SDavid Sterba goto fail_alloc; 27201104a885SDavid Sterba } 27211104a885SDavid Sterba 27221104a885SDavid Sterba /* 27231104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 27241104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 27251104a885SDavid Sterba * the whole block of INFO_SIZE 27261104a885SDavid Sterba */ 27276c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 27286c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 27296c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2730a061fc8dSChris Mason brelse(bh); 27315f39d397SChris Mason 27326c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 27330b86a832SChris Mason 27341104a885SDavid Sterba ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 27351104a885SDavid Sterba if (ret) { 2736efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); 27371104a885SDavid Sterba err = -EINVAL; 27381104a885SDavid Sterba goto fail_alloc; 27391104a885SDavid Sterba } 27401104a885SDavid Sterba 27416c41761fSDavid Sterba disk_super = fs_info->super_copy; 27420f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 274316cdcec7SMiao Xie goto fail_alloc; 27440f7d52f4SChris Mason 2745acce952bSliubo /* check FS state, whether FS is broken. */ 274687533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 274787533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2748acce952bSliubo 274975e7cb7fSLiu Bo /* 2750af31f5e5SChris Mason * run through our array of backup supers and setup 2751af31f5e5SChris Mason * our ring pointer to the oldest one 2752af31f5e5SChris Mason */ 2753af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2754af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2755af31f5e5SChris Mason 2756af31f5e5SChris Mason /* 275775e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 275875e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 275975e7cb7fSLiu Bo */ 276075e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 276175e7cb7fSLiu Bo 276296da0919SQu Wenruo ret = btrfs_parse_options(tree_root, options, sb->s_flags); 27632b82032cSYan Zheng if (ret) { 27642b82032cSYan Zheng err = ret; 276516cdcec7SMiao Xie goto fail_alloc; 27662b82032cSYan Zheng } 2767dfe25020SChris Mason 2768f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2769f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2770f2b636e8SJosef Bacik if (features) { 2771f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount because of " 2772f2b636e8SJosef Bacik "unsupported optional features (%Lx).\n", 2773c1c9ff7cSGeert Uytterhoeven features); 2774f2b636e8SJosef Bacik err = -EINVAL; 277516cdcec7SMiao Xie goto fail_alloc; 2776f2b636e8SJosef Bacik } 2777f2b636e8SJosef Bacik 27785d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 27795d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 278069e380d1SLi Zefan if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2781a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2782727011e0SChris Mason 27833173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 27845efa0490SDavid Sterba printk(KERN_INFO "BTRFS: has skinny extents\n"); 27853173a18fSJosef Bacik 2786727011e0SChris Mason /* 2787727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2788727011e0SChris Mason * they are bigger than the page size 2789727011e0SChris Mason */ 279009cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 2791727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2792efe120a0SFrank Holton printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); 2793727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2794727011e0SChris Mason } 2795727011e0SChris Mason 2796bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2797bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2798bc3f116fSChris Mason stripesize = btrfs_super_stripesize(disk_super); 2799707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2800963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2801bc3f116fSChris Mason 2802bc3f116fSChris Mason /* 2803bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2804bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2805bc3f116fSChris Mason */ 2806bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2807707e8a07SDavid Sterba (sectorsize != nodesize)) { 2808aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes " 2809bc3f116fSChris Mason "are not allowed for mixed block groups on %s\n", 2810bc3f116fSChris Mason sb->s_id); 2811bc3f116fSChris Mason goto fail_alloc; 2812bc3f116fSChris Mason } 2813bc3f116fSChris Mason 2814ceda0864SMiao Xie /* 2815ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2816ceda0864SMiao Xie * update the flag. 2817ceda0864SMiao Xie */ 28185d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 28195d4f98a2SYan Zheng 2820f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2821f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2822f2b636e8SJosef Bacik if (!(sb->s_flags & MS_RDONLY) && features) { 2823f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2824f2b636e8SJosef Bacik "unsupported option features (%Lx).\n", 2825c1c9ff7cSGeert Uytterhoeven features); 2826f2b636e8SJosef Bacik err = -EINVAL; 282716cdcec7SMiao Xie goto fail_alloc; 2828f2b636e8SJosef Bacik } 282961d92c32SChris Mason 28305cdc7ad3SQu Wenruo max_active = fs_info->thread_pool_size; 283161d92c32SChris Mason 28322a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 28332a458198SEric Sandeen if (ret) { 28342a458198SEric Sandeen err = ret; 28350dc3b84aSJosef Bacik goto fail_sb_buffer; 28360dc3b84aSJosef Bacik } 28374543df7eSChris Mason 28384575c9ccSChris Mason fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2839c8b97818SChris Mason fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 284009cbfeafSKirill A. Shutemov SZ_4M / PAGE_SIZE); 28414575c9ccSChris Mason 2842db94535dSChris Mason tree_root->nodesize = nodesize; 2843db94535dSChris Mason tree_root->sectorsize = sectorsize; 284487ee04ebSChris Mason tree_root->stripesize = stripesize; 2845a061fc8dSChris Mason 2846a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2847a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2848db94535dSChris Mason 2849925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 2850e4404d6eSYan Zheng ret = btrfs_read_sys_array(tree_root); 2851925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 285284eed90fSChris Mason if (ret) { 2853aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read the system " 2854d397712bSChris Mason "array on %s\n", sb->s_id); 28555d4f98a2SYan Zheng goto fail_sb_buffer; 285684eed90fSChris Mason } 28570b86a832SChris Mason 285884234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 28590b86a832SChris Mason 2860707e8a07SDavid Sterba __setup_root(nodesize, sectorsize, stripesize, chunk_root, 2861707e8a07SDavid Sterba fs_info, BTRFS_CHUNK_TREE_OBJECTID); 28620b86a832SChris Mason 28630b86a832SChris Mason chunk_root->node = read_tree_block(chunk_root, 28640b86a832SChris Mason btrfs_super_chunk_root(disk_super), 2865ce86cd59SDavid Sterba generation); 286664c043deSLiu Bo if (IS_ERR(chunk_root->node) || 286764c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 2868aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n", 286983121942SDavid Woodhouse sb->s_id); 2870e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 2871e5fffbacSchandan free_extent_buffer(chunk_root->node); 287295ab1f64SZhao Lei chunk_root->node = NULL; 2873af31f5e5SChris Mason goto fail_tree_roots; 287483121942SDavid Woodhouse } 28755d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 28765d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 28770b86a832SChris Mason 2878e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2879b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2880e17cade2SChris Mason 28810b86a832SChris Mason ret = btrfs_read_chunk_tree(chunk_root); 28822b82032cSYan Zheng if (ret) { 2883aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n", 2884d397712bSChris Mason sb->s_id); 2885af31f5e5SChris Mason goto fail_tree_roots; 28862b82032cSYan Zheng } 28870b86a832SChris Mason 28888dabb742SStefan Behrens /* 28898dabb742SStefan Behrens * keep the device that is marked to be the target device for the 28908dabb742SStefan Behrens * dev_replace procedure 28918dabb742SStefan Behrens */ 28929eaed21eSEric Sandeen btrfs_close_extra_devices(fs_devices, 0); 2893dfe25020SChris Mason 2894a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 2895aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to read devices on %s\n", 2896a6b0d5c8SChris Mason sb->s_id); 2897a6b0d5c8SChris Mason goto fail_tree_roots; 2898a6b0d5c8SChris Mason } 2899a6b0d5c8SChris Mason 2900af31f5e5SChris Mason retry_root_backup: 290184234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 29020b86a832SChris Mason 2903e20d96d6SChris Mason tree_root->node = read_tree_block(tree_root, 2904db94535dSChris Mason btrfs_super_root(disk_super), 2905ce86cd59SDavid Sterba generation); 290664c043deSLiu Bo if (IS_ERR(tree_root->node) || 290764c043deSLiu Bo !extent_buffer_uptodate(tree_root->node)) { 2908efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", 290983121942SDavid Woodhouse sb->s_id); 2910e5fffbacSchandan if (!IS_ERR(tree_root->node)) 2911e5fffbacSchandan free_extent_buffer(tree_root->node); 291295ab1f64SZhao Lei tree_root->node = NULL; 2913af31f5e5SChris Mason goto recovery_tree_root; 291483121942SDavid Woodhouse } 2915af31f5e5SChris Mason 29165d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 29175d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 291869e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2919db94535dSChris Mason 2920f32e48e9SChandan Rajendra mutex_lock(&tree_root->objectid_mutex); 2921f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(tree_root, 2922f32e48e9SChandan Rajendra &tree_root->highest_objectid); 2923f32e48e9SChandan Rajendra if (ret) { 2924f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2925f32e48e9SChandan Rajendra goto recovery_tree_root; 2926f32e48e9SChandan Rajendra } 2927f32e48e9SChandan Rajendra 2928f32e48e9SChandan Rajendra ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2929f32e48e9SChandan Rajendra 2930f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2931f32e48e9SChandan Rajendra 29324bbcaa64SEric Sandeen ret = btrfs_read_roots(fs_info, tree_root); 29334bbcaa64SEric Sandeen if (ret) 2934af31f5e5SChris Mason goto recovery_tree_root; 2935f7a81ea4SStefan Behrens 29368929ecfaSYan, Zheng fs_info->generation = generation; 29378929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 29388929ecfaSYan, Zheng 293968310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 294068310a5eSIlya Dryomov if (ret) { 2941aa8ee312SDavid Sterba printk(KERN_ERR "BTRFS: failed to recover balance\n"); 294268310a5eSIlya Dryomov goto fail_block_groups; 294368310a5eSIlya Dryomov } 294468310a5eSIlya Dryomov 2945733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2946733f4fbbSStefan Behrens if (ret) { 2947efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", 2948733f4fbbSStefan Behrens ret); 2949733f4fbbSStefan Behrens goto fail_block_groups; 2950733f4fbbSStefan Behrens } 2951733f4fbbSStefan Behrens 29528dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 29538dabb742SStefan Behrens if (ret) { 2954efe120a0SFrank Holton pr_err("BTRFS: failed to init dev_replace: %d\n", ret); 29558dabb742SStefan Behrens goto fail_block_groups; 29568dabb742SStefan Behrens } 29578dabb742SStefan Behrens 29589eaed21eSEric Sandeen btrfs_close_extra_devices(fs_devices, 1); 29598dabb742SStefan Behrens 2960b7c35e81SAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices, NULL); 2961b7c35e81SAnand Jain if (ret) { 2962b7c35e81SAnand Jain pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret); 2963b7c35e81SAnand Jain goto fail_block_groups; 2964b7c35e81SAnand Jain } 2965b7c35e81SAnand Jain 2966b7c35e81SAnand Jain ret = btrfs_sysfs_add_device(fs_devices); 2967b7c35e81SAnand Jain if (ret) { 2968b7c35e81SAnand Jain pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret); 2969b7c35e81SAnand Jain goto fail_fsdev_sysfs; 2970b7c35e81SAnand Jain } 2971b7c35e81SAnand Jain 297296f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 29735ac1d209SJeff Mahoney if (ret) { 2974efe120a0SFrank Holton pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); 2975b7c35e81SAnand Jain goto fail_fsdev_sysfs; 29765ac1d209SJeff Mahoney } 29775ac1d209SJeff Mahoney 2978c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2979c59021f8Sliubo if (ret) { 2980efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); 29812365dd3cSAnand Jain goto fail_sysfs; 2982c59021f8Sliubo } 2983c59021f8Sliubo 29844bbcaa64SEric Sandeen ret = btrfs_read_block_groups(fs_info->extent_root); 29851b1d1f66SJosef Bacik if (ret) { 2986efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); 29872365dd3cSAnand Jain goto fail_sysfs; 29881b1d1f66SJosef Bacik } 29895af3e8ccSStefan Behrens fs_info->num_tolerated_disk_barrier_failures = 29905af3e8ccSStefan Behrens btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2991292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 2992292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 2993292fd7fcSStefan Behrens !(sb->s_flags & MS_RDONLY)) { 299478fa1770SZhao Lei pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n", 299578fa1770SZhao Lei fs_info->fs_devices->missing_devices, 299678fa1770SZhao Lei fs_info->num_tolerated_disk_barrier_failures); 29972365dd3cSAnand Jain goto fail_sysfs; 2998292fd7fcSStefan Behrens } 29999078a3e1SChris Mason 3000a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3001a74a4b97SChris Mason "btrfs-cleaner"); 300257506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 30032365dd3cSAnand Jain goto fail_sysfs; 3004a74a4b97SChris Mason 3005a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3006a74a4b97SChris Mason tree_root, 3007a74a4b97SChris Mason "btrfs-transaction"); 300857506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 30093f157a2fSChris Mason goto fail_cleaner; 3010a74a4b97SChris Mason 3011c289811cSChris Mason if (!btrfs_test_opt(tree_root, SSD) && 3012c289811cSChris Mason !btrfs_test_opt(tree_root, NOSSD) && 3013c289811cSChris Mason !fs_info->fs_devices->rotating) { 3014efe120a0SFrank Holton printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " 3015c289811cSChris Mason "mode\n"); 3016c289811cSChris Mason btrfs_set_opt(fs_info->mount_opt, SSD); 3017c289811cSChris Mason } 3018c289811cSChris Mason 3019572d9ab7SDavid Sterba /* 3020572d9ab7SDavid Sterba * Mount does not set all options immediatelly, we can do it now and do 3021572d9ab7SDavid Sterba * not have to wait for transaction commit 3022572d9ab7SDavid Sterba */ 3023572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 30243818aea2SQu Wenruo 302521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 302621adbd5cSStefan Behrens if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 302721adbd5cSStefan Behrens ret = btrfsic_mount(tree_root, fs_devices, 302821adbd5cSStefan Behrens btrfs_test_opt(tree_root, 302921adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 303021adbd5cSStefan Behrens 1 : 0, 303121adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 303221adbd5cSStefan Behrens if (ret) 3033efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to initialize" 303421adbd5cSStefan Behrens " integrity check module %s\n", sb->s_id); 303521adbd5cSStefan Behrens } 303621adbd5cSStefan Behrens #endif 3037bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3038bcef60f2SArne Jansen if (ret) 3039bcef60f2SArne Jansen goto fail_trans_kthread; 304021adbd5cSStefan Behrens 304196da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 304296da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 304396da0919SQu Wenruo !btrfs_test_opt(tree_root, NOLOGREPLAY)) { 304463443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 304579787eaaSJeff Mahoney if (ret) { 304663443bf5SEric Sandeen err = ret; 304728c16cbbSWang Shilong goto fail_qgroup; 3048e556ce2cSYan Zheng } 3049e02119d5SChris Mason } 30501a40e23bSZheng Yan 305176dda93cSYan, Zheng ret = btrfs_find_orphan_roots(tree_root); 305279787eaaSJeff Mahoney if (ret) 305328c16cbbSWang Shilong goto fail_qgroup; 305476dda93cSYan, Zheng 30557c2ca468SChris Mason if (!(sb->s_flags & MS_RDONLY)) { 3056d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 305744c44af2SIlya Dryomov if (ret) 305828c16cbbSWang Shilong goto fail_qgroup; 3059d68fc57bSYan, Zheng 30605f316481SWang Shilong mutex_lock(&fs_info->cleaner_mutex); 30615d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 30625f316481SWang Shilong mutex_unlock(&fs_info->cleaner_mutex); 3063d7ce5843SMiao Xie if (ret < 0) { 3064d7ce5843SMiao Xie printk(KERN_WARNING 3065efe120a0SFrank Holton "BTRFS: failed to recover relocation\n"); 3066d7ce5843SMiao Xie err = -EINVAL; 3067bcef60f2SArne Jansen goto fail_qgroup; 3068d7ce5843SMiao Xie } 30697c2ca468SChris Mason } 30701a40e23bSZheng Yan 30713de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 30723de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 3073cb517eabSMiao Xie location.offset = 0; 30743de4586cSChris Mason 30753de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 30763140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 30773140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3078bcef60f2SArne Jansen goto fail_qgroup; 30793140c9a3SDan Carpenter } 3080c289811cSChris Mason 30812b6ba629SIlya Dryomov if (sb->s_flags & MS_RDONLY) 30822b6ba629SIlya Dryomov return 0; 30832b6ba629SIlya Dryomov 3084511711afSChris Mason if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) && 3085511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 3086511711afSChris Mason pr_info("BTRFS: creating free space tree\n"); 3087511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3088511711afSChris Mason if (ret) { 3089511711afSChris Mason pr_warn("BTRFS: failed to create free space tree %d\n", 3090511711afSChris Mason ret); 3091511711afSChris Mason close_ctree(tree_root); 3092511711afSChris Mason return ret; 3093511711afSChris Mason } 3094511711afSChris Mason } 3095511711afSChris Mason 3096e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 30972b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 30982b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 30992b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 31002b6ba629SIlya Dryomov close_ctree(tree_root); 31012b6ba629SIlya Dryomov return ret; 31022b6ba629SIlya Dryomov } 3103e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 310459641015SIlya Dryomov 31052b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 31062b6ba629SIlya Dryomov if (ret) { 3107efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to resume balance\n"); 310866b4ffd1SJosef Bacik close_ctree(tree_root); 31092b6ba629SIlya Dryomov return ret; 3110e3acc2a6SJosef Bacik } 3111e3acc2a6SJosef Bacik 31128dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 31138dabb742SStefan Behrens if (ret) { 3114efe120a0SFrank Holton pr_warn("BTRFS: failed to resume dev_replace\n"); 31158dabb742SStefan Behrens close_ctree(tree_root); 31168dabb742SStefan Behrens return ret; 31178dabb742SStefan Behrens } 31188dabb742SStefan Behrens 3119b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3120b382a324SJan Schmidt 312170f6d82eSOmar Sandoval if (btrfs_test_opt(tree_root, CLEAR_CACHE) && 312270f6d82eSOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 312370f6d82eSOmar Sandoval pr_info("BTRFS: clearing free space tree\n"); 312470f6d82eSOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 312570f6d82eSOmar Sandoval if (ret) { 312670f6d82eSOmar Sandoval pr_warn("BTRFS: failed to clear free space tree %d\n", 312770f6d82eSOmar Sandoval ret); 312870f6d82eSOmar Sandoval close_ctree(tree_root); 312970f6d82eSOmar Sandoval return ret; 313070f6d82eSOmar Sandoval } 313170f6d82eSOmar Sandoval } 313270f6d82eSOmar Sandoval 31334bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 3134efe120a0SFrank Holton pr_info("BTRFS: creating UUID tree\n"); 3135f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3136f7a81ea4SStefan Behrens if (ret) { 3137efe120a0SFrank Holton pr_warn("BTRFS: failed to create the UUID tree %d\n", 3138f7a81ea4SStefan Behrens ret); 3139f7a81ea4SStefan Behrens close_ctree(tree_root); 3140f7a81ea4SStefan Behrens return ret; 3141f7a81ea4SStefan Behrens } 31424bbcaa64SEric Sandeen } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) || 31434bbcaa64SEric Sandeen fs_info->generation != 31444bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 3145efe120a0SFrank Holton pr_info("BTRFS: checking UUID tree\n"); 314670f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 314770f80175SStefan Behrens if (ret) { 3148efe120a0SFrank Holton pr_warn("BTRFS: failed to check the UUID tree %d\n", 314970f80175SStefan Behrens ret); 315070f80175SStefan Behrens close_ctree(tree_root); 315170f80175SStefan Behrens return ret; 315270f80175SStefan Behrens } 315370f80175SStefan Behrens } else { 315470f80175SStefan Behrens fs_info->update_uuid_tree_gen = 1; 3155f7a81ea4SStefan Behrens } 3156f7a81ea4SStefan Behrens 315747ab2a6cSJosef Bacik fs_info->open = 1; 315847ab2a6cSJosef Bacik 31598dcddfa0SQu Wenruo /* 31608dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 31618dcddfa0SQu Wenruo * no need to keep the flag 31628dcddfa0SQu Wenruo */ 31638dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 31648dcddfa0SQu Wenruo 3165ad2b2c80SAl Viro return 0; 316639279cc3SChris Mason 3167bcef60f2SArne Jansen fail_qgroup: 3168bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 31697c2ca468SChris Mason fail_trans_kthread: 31707c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 317154067ae9SJosef Bacik btrfs_cleanup_transaction(fs_info->tree_root); 3172faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 31733f157a2fSChris Mason fail_cleaner: 3174a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 31757c2ca468SChris Mason 31767c2ca468SChris Mason /* 31777c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 31787c2ca468SChris Mason * kthreads 31797c2ca468SChris Mason */ 31807c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 31817c2ca468SChris Mason 31822365dd3cSAnand Jain fail_sysfs: 31836618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 31842365dd3cSAnand Jain 3185b7c35e81SAnand Jain fail_fsdev_sysfs: 3186b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3187b7c35e81SAnand Jain 31881b1d1f66SJosef Bacik fail_block_groups: 318954067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 31901b1d1f66SJosef Bacik btrfs_free_block_groups(fs_info); 3191af31f5e5SChris Mason 3192af31f5e5SChris Mason fail_tree_roots: 3193af31f5e5SChris Mason free_root_pointers(fs_info, 1); 31942b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3195af31f5e5SChris Mason 319639279cc3SChris Mason fail_sb_buffer: 31977abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 319816cdcec7SMiao Xie fail_alloc: 31994543df7eSChris Mason fail_iput: 3200586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3201586e46e2SIlya Dryomov 32024543df7eSChris Mason iput(fs_info->btree_inode); 3203c404e0dcSMiao Xie fail_bio_counter: 3204c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 3205963d678bSMiao Xie fail_delalloc_bytes: 3206963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3207e2d84521SMiao Xie fail_dirty_metadata_bytes: 3208e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3209ad081f14SJens Axboe fail_bdi: 32107e662854SQinghuang Feng bdi_destroy(&fs_info->bdi); 321176dda93cSYan, Zheng fail_srcu: 321276dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 32137e662854SQinghuang Feng fail: 321453b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3215586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3216ad2b2c80SAl Viro return err; 3217af31f5e5SChris Mason 3218af31f5e5SChris Mason recovery_tree_root: 32198dcddfa0SQu Wenruo if (!btrfs_test_opt(tree_root, USEBACKUPROOT)) 3220af31f5e5SChris Mason goto fail_tree_roots; 3221af31f5e5SChris Mason 3222af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3223af31f5e5SChris Mason 3224af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3225af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3226af31f5e5SChris Mason 3227af31f5e5SChris Mason /* we can't trust the free space cache either */ 3228af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3229af31f5e5SChris Mason 3230af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3231af31f5e5SChris Mason &num_backups_tried, &backup_index); 3232af31f5e5SChris Mason if (ret == -1) 3233af31f5e5SChris Mason goto fail_block_groups; 3234af31f5e5SChris Mason goto retry_root_backup; 3235eb60ceacSChris Mason } 3236eb60ceacSChris Mason 3237f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3238f2984462SChris Mason { 3239f2984462SChris Mason if (uptodate) { 3240f2984462SChris Mason set_buffer_uptodate(bh); 3241f2984462SChris Mason } else { 3242442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3243442a4f63SStefan Behrens bh->b_private; 3244442a4f63SStefan Behrens 3245b14af3b4SDavid Sterba btrfs_warn_rl_in_rcu(device->dev_root->fs_info, 3246b14af3b4SDavid Sterba "lost page write due to IO error on %s", 3247606686eeSJosef Bacik rcu_str_deref(device->name)); 32481259ab75SChris Mason /* note, we dont' set_buffer_write_io_error because we have 32491259ab75SChris Mason * our own ways of dealing with the IO errors 32501259ab75SChris Mason */ 3251f2984462SChris Mason clear_buffer_uptodate(bh); 3252442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3253f2984462SChris Mason } 3254f2984462SChris Mason unlock_buffer(bh); 3255f2984462SChris Mason put_bh(bh); 3256f2984462SChris Mason } 3257f2984462SChris Mason 325829c36d72SAnand Jain int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, 325929c36d72SAnand Jain struct buffer_head **bh_ret) 326029c36d72SAnand Jain { 326129c36d72SAnand Jain struct buffer_head *bh; 326229c36d72SAnand Jain struct btrfs_super_block *super; 326329c36d72SAnand Jain u64 bytenr; 326429c36d72SAnand Jain 326529c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 326629c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 326729c36d72SAnand Jain return -EINVAL; 326829c36d72SAnand Jain 326929c36d72SAnand Jain bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE); 327029c36d72SAnand Jain /* 327129c36d72SAnand Jain * If we fail to read from the underlying devices, as of now 327229c36d72SAnand Jain * the best option we have is to mark it EIO. 327329c36d72SAnand Jain */ 327429c36d72SAnand Jain if (!bh) 327529c36d72SAnand Jain return -EIO; 327629c36d72SAnand Jain 327729c36d72SAnand Jain super = (struct btrfs_super_block *)bh->b_data; 327829c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 327929c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 328029c36d72SAnand Jain brelse(bh); 328129c36d72SAnand Jain return -EINVAL; 328229c36d72SAnand Jain } 328329c36d72SAnand Jain 328429c36d72SAnand Jain *bh_ret = bh; 328529c36d72SAnand Jain return 0; 328629c36d72SAnand Jain } 328729c36d72SAnand Jain 328829c36d72SAnand Jain 3289a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3290a512bbf8SYan Zheng { 3291a512bbf8SYan Zheng struct buffer_head *bh; 3292a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3293a512bbf8SYan Zheng struct btrfs_super_block *super; 3294a512bbf8SYan Zheng int i; 3295a512bbf8SYan Zheng u64 transid = 0; 329692fc03fbSAnand Jain int ret = -EINVAL; 3297a512bbf8SYan Zheng 3298a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3299a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3300a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3301a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3302a512bbf8SYan Zheng */ 3303a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 330429c36d72SAnand Jain ret = btrfs_read_dev_one_super(bdev, i, &bh); 330529c36d72SAnand Jain if (ret) 3306a512bbf8SYan Zheng continue; 3307a512bbf8SYan Zheng 3308a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3309a512bbf8SYan Zheng 3310a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3311a512bbf8SYan Zheng brelse(latest); 3312a512bbf8SYan Zheng latest = bh; 3313a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3314a512bbf8SYan Zheng } else { 3315a512bbf8SYan Zheng brelse(bh); 3316a512bbf8SYan Zheng } 3317a512bbf8SYan Zheng } 331892fc03fbSAnand Jain 331992fc03fbSAnand Jain if (!latest) 332092fc03fbSAnand Jain return ERR_PTR(ret); 332192fc03fbSAnand Jain 3322a512bbf8SYan Zheng return latest; 3323a512bbf8SYan Zheng } 3324a512bbf8SYan Zheng 33254eedeb75SHisashi Hifumi /* 33264eedeb75SHisashi Hifumi * this should be called twice, once with wait == 0 and 33274eedeb75SHisashi Hifumi * once with wait == 1. When wait == 0 is done, all the buffer heads 33284eedeb75SHisashi Hifumi * we write are pinned. 33294eedeb75SHisashi Hifumi * 33304eedeb75SHisashi Hifumi * They are released when wait == 1 is done. 33314eedeb75SHisashi Hifumi * max_mirrors must be the same for both runs, and it indicates how 33324eedeb75SHisashi Hifumi * many supers on this one device should be written. 33334eedeb75SHisashi Hifumi * 33344eedeb75SHisashi Hifumi * max_mirrors == 0 means to write them all. 33354eedeb75SHisashi Hifumi */ 3336a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3337a512bbf8SYan Zheng struct btrfs_super_block *sb, 3338a512bbf8SYan Zheng int do_barriers, int wait, int max_mirrors) 3339a512bbf8SYan Zheng { 3340a512bbf8SYan Zheng struct buffer_head *bh; 3341a512bbf8SYan Zheng int i; 3342a512bbf8SYan Zheng int ret; 3343a512bbf8SYan Zheng int errors = 0; 3344a512bbf8SYan Zheng u32 crc; 3345a512bbf8SYan Zheng u64 bytenr; 3346a512bbf8SYan Zheng 3347a512bbf8SYan Zheng if (max_mirrors == 0) 3348a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3349a512bbf8SYan Zheng 3350a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3351a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3352935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3353935e5cc9SMiao Xie device->commit_total_bytes) 3354a512bbf8SYan Zheng break; 3355a512bbf8SYan Zheng 3356a512bbf8SYan Zheng if (wait) { 3357a512bbf8SYan Zheng bh = __find_get_block(device->bdev, bytenr / 4096, 3358a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3359634554dcSJosef Bacik if (!bh) { 3360634554dcSJosef Bacik errors++; 3361634554dcSJosef Bacik continue; 3362634554dcSJosef Bacik } 3363a512bbf8SYan Zheng wait_on_buffer(bh); 33644eedeb75SHisashi Hifumi if (!buffer_uptodate(bh)) 33654eedeb75SHisashi Hifumi errors++; 33664eedeb75SHisashi Hifumi 33674eedeb75SHisashi Hifumi /* drop our reference */ 33684eedeb75SHisashi Hifumi brelse(bh); 33694eedeb75SHisashi Hifumi 33704eedeb75SHisashi Hifumi /* drop the reference from the wait == 0 run */ 3371a512bbf8SYan Zheng brelse(bh); 3372a512bbf8SYan Zheng continue; 3373a512bbf8SYan Zheng } else { 3374a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3375a512bbf8SYan Zheng 3376a512bbf8SYan Zheng crc = ~(u32)0; 3377b0496686SLiu Bo crc = btrfs_csum_data((char *)sb + 3378a512bbf8SYan Zheng BTRFS_CSUM_SIZE, crc, 3379a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE - 3380a512bbf8SYan Zheng BTRFS_CSUM_SIZE); 3381a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3382a512bbf8SYan Zheng 33834eedeb75SHisashi Hifumi /* 33844eedeb75SHisashi Hifumi * one reference for us, and we leave it for the 33854eedeb75SHisashi Hifumi * caller 33864eedeb75SHisashi Hifumi */ 3387a512bbf8SYan Zheng bh = __getblk(device->bdev, bytenr / 4096, 3388a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3389634554dcSJosef Bacik if (!bh) { 3390f14d104dSDavid Sterba btrfs_err(device->dev_root->fs_info, 3391f14d104dSDavid Sterba "couldn't get super buffer head for bytenr %llu", 3392f14d104dSDavid Sterba bytenr); 3393634554dcSJosef Bacik errors++; 3394634554dcSJosef Bacik continue; 3395634554dcSJosef Bacik } 3396634554dcSJosef Bacik 3397a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3398a512bbf8SYan Zheng 33994eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3400a512bbf8SYan Zheng get_bh(bh); 34014eedeb75SHisashi Hifumi 34024eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3403a512bbf8SYan Zheng lock_buffer(bh); 3404a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3405442a4f63SStefan Behrens bh->b_private = device; 3406a512bbf8SYan Zheng } 3407a512bbf8SYan Zheng 3408387125fcSChris Mason /* 3409387125fcSChris Mason * we fua the first super. The others we allow 3410387125fcSChris Mason * to go down lazy. 3411387125fcSChris Mason */ 3412e8117c26SWang Shilong if (i == 0) 341321adbd5cSStefan Behrens ret = btrfsic_submit_bh(WRITE_FUA, bh); 3414e8117c26SWang Shilong else 3415e8117c26SWang Shilong ret = btrfsic_submit_bh(WRITE_SYNC, bh); 34164eedeb75SHisashi Hifumi if (ret) 3417a512bbf8SYan Zheng errors++; 3418a512bbf8SYan Zheng } 3419a512bbf8SYan Zheng return errors < i ? 0 : -1; 3420a512bbf8SYan Zheng } 3421a512bbf8SYan Zheng 3422387125fcSChris Mason /* 3423387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3424387125fcSChris Mason * for the barrier when it is done 3425387125fcSChris Mason */ 34264246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3427387125fcSChris Mason { 3428387125fcSChris Mason if (bio->bi_private) 3429387125fcSChris Mason complete(bio->bi_private); 3430387125fcSChris Mason bio_put(bio); 3431387125fcSChris Mason } 3432387125fcSChris Mason 3433387125fcSChris Mason /* 3434387125fcSChris Mason * trigger flushes for one the devices. If you pass wait == 0, the flushes are 3435387125fcSChris Mason * sent down. With wait == 1, it waits for the previous flush. 3436387125fcSChris Mason * 3437387125fcSChris Mason * any device where the flush fails with eopnotsupp are flagged as not-barrier 3438387125fcSChris Mason * capable 3439387125fcSChris Mason */ 3440387125fcSChris Mason static int write_dev_flush(struct btrfs_device *device, int wait) 3441387125fcSChris Mason { 3442387125fcSChris Mason struct bio *bio; 3443387125fcSChris Mason int ret = 0; 3444387125fcSChris Mason 3445387125fcSChris Mason if (device->nobarriers) 3446387125fcSChris Mason return 0; 3447387125fcSChris Mason 3448387125fcSChris Mason if (wait) { 3449387125fcSChris Mason bio = device->flush_bio; 3450387125fcSChris Mason if (!bio) 3451387125fcSChris Mason return 0; 3452387125fcSChris Mason 3453387125fcSChris Mason wait_for_completion(&device->flush_wait); 3454387125fcSChris Mason 34554246a0b6SChristoph Hellwig if (bio->bi_error) { 34564246a0b6SChristoph Hellwig ret = bio->bi_error; 3457442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, 3458442a4f63SStefan Behrens BTRFS_DEV_STAT_FLUSH_ERRS); 3459387125fcSChris Mason } 3460387125fcSChris Mason 3461387125fcSChris Mason /* drop the reference from the wait == 0 run */ 3462387125fcSChris Mason bio_put(bio); 3463387125fcSChris Mason device->flush_bio = NULL; 3464387125fcSChris Mason 3465387125fcSChris Mason return ret; 3466387125fcSChris Mason } 3467387125fcSChris Mason 3468387125fcSChris Mason /* 3469387125fcSChris Mason * one reference for us, and we leave it for the 3470387125fcSChris Mason * caller 3471387125fcSChris Mason */ 34729c017abcSJesper Juhl device->flush_bio = NULL; 34739be3395bSChris Mason bio = btrfs_io_bio_alloc(GFP_NOFS, 0); 3474387125fcSChris Mason if (!bio) 3475387125fcSChris Mason return -ENOMEM; 3476387125fcSChris Mason 3477387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3478387125fcSChris Mason bio->bi_bdev = device->bdev; 3479387125fcSChris Mason init_completion(&device->flush_wait); 3480387125fcSChris Mason bio->bi_private = &device->flush_wait; 3481387125fcSChris Mason device->flush_bio = bio; 3482387125fcSChris Mason 3483387125fcSChris Mason bio_get(bio); 348421adbd5cSStefan Behrens btrfsic_submit_bio(WRITE_FLUSH, bio); 3485387125fcSChris Mason 3486387125fcSChris Mason return 0; 3487387125fcSChris Mason } 3488387125fcSChris Mason 3489387125fcSChris Mason /* 3490387125fcSChris Mason * send an empty flush down to each device in parallel, 3491387125fcSChris Mason * then wait for them 3492387125fcSChris Mason */ 3493387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3494387125fcSChris Mason { 3495387125fcSChris Mason struct list_head *head; 3496387125fcSChris Mason struct btrfs_device *dev; 34975af3e8ccSStefan Behrens int errors_send = 0; 34985af3e8ccSStefan Behrens int errors_wait = 0; 3499387125fcSChris Mason int ret; 3500387125fcSChris Mason 3501387125fcSChris Mason /* send down all the barriers */ 3502387125fcSChris Mason head = &info->fs_devices->devices; 3503387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3504f88ba6a2SHidetoshi Seto if (dev->missing) 3505f88ba6a2SHidetoshi Seto continue; 3506387125fcSChris Mason if (!dev->bdev) { 35075af3e8ccSStefan Behrens errors_send++; 3508387125fcSChris Mason continue; 3509387125fcSChris Mason } 3510387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3511387125fcSChris Mason continue; 3512387125fcSChris Mason 3513387125fcSChris Mason ret = write_dev_flush(dev, 0); 3514387125fcSChris Mason if (ret) 35155af3e8ccSStefan Behrens errors_send++; 3516387125fcSChris Mason } 3517387125fcSChris Mason 3518387125fcSChris Mason /* wait for all the barriers */ 3519387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3520f88ba6a2SHidetoshi Seto if (dev->missing) 3521f88ba6a2SHidetoshi Seto continue; 3522387125fcSChris Mason if (!dev->bdev) { 35235af3e8ccSStefan Behrens errors_wait++; 3524387125fcSChris Mason continue; 3525387125fcSChris Mason } 3526387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3527387125fcSChris Mason continue; 3528387125fcSChris Mason 3529387125fcSChris Mason ret = write_dev_flush(dev, 1); 3530387125fcSChris Mason if (ret) 35315af3e8ccSStefan Behrens errors_wait++; 3532387125fcSChris Mason } 35335af3e8ccSStefan Behrens if (errors_send > info->num_tolerated_disk_barrier_failures || 35345af3e8ccSStefan Behrens errors_wait > info->num_tolerated_disk_barrier_failures) 3535387125fcSChris Mason return -EIO; 3536387125fcSChris Mason return 0; 3537387125fcSChris Mason } 3538387125fcSChris Mason 3539943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3540943c6e99SZhao Lei { 35418789f4feSZhao Lei int raid_type; 35428789f4feSZhao Lei int min_tolerated = INT_MAX; 3543943c6e99SZhao Lei 35448789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 35458789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 35468789f4feSZhao Lei min_tolerated = min(min_tolerated, 35478789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 35488789f4feSZhao Lei tolerated_failures); 3549943c6e99SZhao Lei 35508789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 35518789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 35528789f4feSZhao Lei continue; 35538789f4feSZhao Lei if (!(flags & btrfs_raid_group[raid_type])) 35548789f4feSZhao Lei continue; 35558789f4feSZhao Lei min_tolerated = min(min_tolerated, 35568789f4feSZhao Lei btrfs_raid_array[raid_type]. 35578789f4feSZhao Lei tolerated_failures); 35588789f4feSZhao Lei } 3559943c6e99SZhao Lei 35608789f4feSZhao Lei if (min_tolerated == INT_MAX) { 35618789f4feSZhao Lei pr_warn("BTRFS: unknown raid flag: %llu\n", flags); 35628789f4feSZhao Lei min_tolerated = 0; 35638789f4feSZhao Lei } 35648789f4feSZhao Lei 35658789f4feSZhao Lei return min_tolerated; 3566943c6e99SZhao Lei } 3567943c6e99SZhao Lei 35685af3e8ccSStefan Behrens int btrfs_calc_num_tolerated_disk_barrier_failures( 35695af3e8ccSStefan Behrens struct btrfs_fs_info *fs_info) 35705af3e8ccSStefan Behrens { 35715af3e8ccSStefan Behrens struct btrfs_ioctl_space_info space; 35725af3e8ccSStefan Behrens struct btrfs_space_info *sinfo; 35735af3e8ccSStefan Behrens u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 35745af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_SYSTEM, 35755af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_METADATA, 35765af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 35775af3e8ccSStefan Behrens int i; 35785af3e8ccSStefan Behrens int c; 35795af3e8ccSStefan Behrens int num_tolerated_disk_barrier_failures = 35805af3e8ccSStefan Behrens (int)fs_info->fs_devices->num_devices; 35815af3e8ccSStefan Behrens 35822c458045SZhao Lei for (i = 0; i < ARRAY_SIZE(types); i++) { 35835af3e8ccSStefan Behrens struct btrfs_space_info *tmp; 35845af3e8ccSStefan Behrens 35855af3e8ccSStefan Behrens sinfo = NULL; 35865af3e8ccSStefan Behrens rcu_read_lock(); 35875af3e8ccSStefan Behrens list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 35885af3e8ccSStefan Behrens if (tmp->flags == types[i]) { 35895af3e8ccSStefan Behrens sinfo = tmp; 35905af3e8ccSStefan Behrens break; 35915af3e8ccSStefan Behrens } 35925af3e8ccSStefan Behrens } 35935af3e8ccSStefan Behrens rcu_read_unlock(); 35945af3e8ccSStefan Behrens 35955af3e8ccSStefan Behrens if (!sinfo) 35965af3e8ccSStefan Behrens continue; 35975af3e8ccSStefan Behrens 35985af3e8ccSStefan Behrens down_read(&sinfo->groups_sem); 35995af3e8ccSStefan Behrens for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 36005af3e8ccSStefan Behrens u64 flags; 36015af3e8ccSStefan Behrens 36022c458045SZhao Lei if (list_empty(&sinfo->block_groups[c])) 36032c458045SZhao Lei continue; 36042c458045SZhao Lei 36052c458045SZhao Lei btrfs_get_block_group_info(&sinfo->block_groups[c], 36062c458045SZhao Lei &space); 36072c458045SZhao Lei if (space.total_bytes == 0 || space.used_bytes == 0) 36085af3e8ccSStefan Behrens continue; 36095af3e8ccSStefan Behrens flags = space.flags; 3610943c6e99SZhao Lei 3611943c6e99SZhao Lei num_tolerated_disk_barrier_failures = min( 3612943c6e99SZhao Lei num_tolerated_disk_barrier_failures, 3613943c6e99SZhao Lei btrfs_get_num_tolerated_disk_barrier_failures( 3614943c6e99SZhao Lei flags)); 36155af3e8ccSStefan Behrens } 36165af3e8ccSStefan Behrens up_read(&sinfo->groups_sem); 36175af3e8ccSStefan Behrens } 36185af3e8ccSStefan Behrens 36195af3e8ccSStefan Behrens return num_tolerated_disk_barrier_failures; 36205af3e8ccSStefan Behrens } 36215af3e8ccSStefan Behrens 362248a3b636SEric Sandeen static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3623f2984462SChris Mason { 3624e5e9a520SChris Mason struct list_head *head; 3625f2984462SChris Mason struct btrfs_device *dev; 3626a061fc8dSChris Mason struct btrfs_super_block *sb; 3627f2984462SChris Mason struct btrfs_dev_item *dev_item; 3628f2984462SChris Mason int ret; 3629f2984462SChris Mason int do_barriers; 3630a236aed1SChris Mason int max_errors; 3631a236aed1SChris Mason int total_errors = 0; 3632a061fc8dSChris Mason u64 flags; 3633f2984462SChris Mason 3634f2984462SChris Mason do_barriers = !btrfs_test_opt(root, NOBARRIER); 3635af31f5e5SChris Mason backup_super_roots(root->fs_info); 3636f2984462SChris Mason 36376c41761fSDavid Sterba sb = root->fs_info->super_for_commit; 3638a061fc8dSChris Mason dev_item = &sb->dev_item; 3639e5e9a520SChris Mason 3640174ba509SChris Mason mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3641e5e9a520SChris Mason head = &root->fs_info->fs_devices->devices; 3642d7306801SFilipe David Borba Manana max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3643387125fcSChris Mason 36445af3e8ccSStefan Behrens if (do_barriers) { 36455af3e8ccSStefan Behrens ret = barrier_all_devices(root->fs_info); 36465af3e8ccSStefan Behrens if (ret) { 36475af3e8ccSStefan Behrens mutex_unlock( 36485af3e8ccSStefan Behrens &root->fs_info->fs_devices->device_list_mutex); 3649a4553fefSAnand Jain btrfs_std_error(root->fs_info, ret, 36505af3e8ccSStefan Behrens "errors while submitting device barriers."); 36515af3e8ccSStefan Behrens return ret; 36525af3e8ccSStefan Behrens } 36535af3e8ccSStefan Behrens } 3654387125fcSChris Mason 36551f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3656dfe25020SChris Mason if (!dev->bdev) { 3657dfe25020SChris Mason total_errors++; 3658dfe25020SChris Mason continue; 3659dfe25020SChris Mason } 36602b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3661dfe25020SChris Mason continue; 3662dfe25020SChris Mason 36632b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3664a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3665a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 36667df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3667935e5cc9SMiao Xie dev->commit_total_bytes); 3668ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3669ce7213c7SMiao Xie dev->commit_bytes_used); 3670a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3671a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3672a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3673a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 36742b82032cSYan Zheng memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3675a512bbf8SYan Zheng 3676a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3677a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3678f2984462SChris Mason 3679a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3680a236aed1SChris Mason if (ret) 3681a236aed1SChris Mason total_errors++; 3682f2984462SChris Mason } 3683a236aed1SChris Mason if (total_errors > max_errors) { 3684efe120a0SFrank Holton btrfs_err(root->fs_info, "%d errors while writing supers", 3685d397712bSChris Mason total_errors); 3686a724b436SStefan Behrens mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 368779787eaaSJeff Mahoney 36889d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 3689a4553fefSAnand Jain btrfs_std_error(root->fs_info, -EIO, 36909d565ba4SStefan Behrens "%d errors while writing supers", total_errors); 36919d565ba4SStefan Behrens return -EIO; 3692a236aed1SChris Mason } 3693f2984462SChris Mason 3694a512bbf8SYan Zheng total_errors = 0; 36951f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3696dfe25020SChris Mason if (!dev->bdev) 3697dfe25020SChris Mason continue; 36982b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3699dfe25020SChris Mason continue; 3700dfe25020SChris Mason 3701a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3702a512bbf8SYan Zheng if (ret) 37031259ab75SChris Mason total_errors++; 3704f2984462SChris Mason } 3705174ba509SChris Mason mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3706a236aed1SChris Mason if (total_errors > max_errors) { 3707a4553fefSAnand Jain btrfs_std_error(root->fs_info, -EIO, 370879787eaaSJeff Mahoney "%d errors while writing supers", total_errors); 370979787eaaSJeff Mahoney return -EIO; 3710a236aed1SChris Mason } 3711f2984462SChris Mason return 0; 3712f2984462SChris Mason } 3713f2984462SChris Mason 3714a512bbf8SYan Zheng int write_ctree_super(struct btrfs_trans_handle *trans, 3715a512bbf8SYan Zheng struct btrfs_root *root, int max_mirrors) 3716cfaa7295SChris Mason { 3717f570e757SRashika return write_all_supers(root, max_mirrors); 3718cfaa7295SChris Mason } 3719cfaa7295SChris Mason 3720cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3721cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3722cb517eabSMiao Xie struct btrfs_root *root) 37232619ba1fSChris Mason { 37244df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 37252619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 37262619ba1fSChris Mason (unsigned long)root->root_key.objectid); 37274df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 372876dda93cSYan, Zheng 372976dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 373076dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 373176dda93cSYan, Zheng 37321a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 37333321719eSLiu Bo btrfs_free_log(NULL, root); 37343321719eSLiu Bo 3735faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3736581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3737faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3738581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 37394df27c4dSYan, Zheng free_fs_root(root); 37404df27c4dSYan, Zheng } 37414df27c4dSYan, Zheng 37424df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 37434df27c4dSYan, Zheng { 374457cdc8dbSDavid Sterba iput(root->ino_cache_inode); 37454df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 37461cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 37471cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 37480ee5dc67SAl Viro if (root->anon_dev) 37490ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 37508257b2dcSMiao Xie if (root->subv_writers) 37518257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 37525f39d397SChris Mason free_extent_buffer(root->node); 37535f39d397SChris Mason free_extent_buffer(root->commit_root); 3754581bb050SLi Zefan kfree(root->free_ino_ctl); 3755581bb050SLi Zefan kfree(root->free_ino_pinned); 375658176a96SJosef Bacik kfree(root->name); 3757b0feb9d9SMiao Xie btrfs_put_fs_root(root); 37582619ba1fSChris Mason } 37592619ba1fSChris Mason 3760cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3761cb517eabSMiao Xie { 3762cb517eabSMiao Xie free_fs_root(root); 3763eb60ceacSChris Mason } 3764eb60ceacSChris Mason 3765c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3766c146afadSYan Zheng { 3767c146afadSYan Zheng u64 root_objectid = 0; 3768c146afadSYan Zheng struct btrfs_root *gang[8]; 376965d33fd7SQu Wenruo int i = 0; 377065d33fd7SQu Wenruo int err = 0; 377165d33fd7SQu Wenruo unsigned int ret = 0; 377265d33fd7SQu Wenruo int index; 3773c146afadSYan Zheng 3774c146afadSYan Zheng while (1) { 377565d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3776c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3777c146afadSYan Zheng (void **)gang, root_objectid, 3778c146afadSYan Zheng ARRAY_SIZE(gang)); 377965d33fd7SQu Wenruo if (!ret) { 378065d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3781c146afadSYan Zheng break; 378265d33fd7SQu Wenruo } 37835d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 378466b4ffd1SJosef Bacik 378565d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 378665d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 378765d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 378865d33fd7SQu Wenruo gang[i] = NULL; 378965d33fd7SQu Wenruo continue; 379065d33fd7SQu Wenruo } 379165d33fd7SQu Wenruo /* grab all the search result for later use */ 379265d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 379365d33fd7SQu Wenruo } 379465d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 379565d33fd7SQu Wenruo 379665d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 379765d33fd7SQu Wenruo if (!gang[i]) 379865d33fd7SQu Wenruo continue; 3799c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 380066b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 380166b4ffd1SJosef Bacik if (err) 380265d33fd7SQu Wenruo break; 380365d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3804c146afadSYan Zheng } 3805c146afadSYan Zheng root_objectid++; 3806c146afadSYan Zheng } 380765d33fd7SQu Wenruo 380865d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 380965d33fd7SQu Wenruo for (; i < ret; i++) { 381065d33fd7SQu Wenruo if (gang[i]) 381165d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 381265d33fd7SQu Wenruo } 381365d33fd7SQu Wenruo return err; 3814c146afadSYan Zheng } 3815c146afadSYan Zheng 3816c146afadSYan Zheng int btrfs_commit_super(struct btrfs_root *root) 3817c146afadSYan Zheng { 3818c146afadSYan Zheng struct btrfs_trans_handle *trans; 3819c146afadSYan Zheng 3820c146afadSYan Zheng mutex_lock(&root->fs_info->cleaner_mutex); 382124bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 3822c146afadSYan Zheng mutex_unlock(&root->fs_info->cleaner_mutex); 38239d1a2a3aSDavid Sterba wake_up_process(root->fs_info->cleaner_kthread); 3824c71bf099SYan, Zheng 3825c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 3826c71bf099SYan, Zheng down_write(&root->fs_info->cleanup_work_sem); 3827c71bf099SYan, Zheng up_write(&root->fs_info->cleanup_work_sem); 3828c71bf099SYan, Zheng 38297a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 38303612b495STsutomu Itoh if (IS_ERR(trans)) 38313612b495STsutomu Itoh return PTR_ERR(trans); 3832d52c1bccSLiu Bo return btrfs_commit_transaction(trans, root); 3833c146afadSYan Zheng } 3834c146afadSYan Zheng 38353abdbd78SDavid Sterba void close_ctree(struct btrfs_root *root) 3836eb60ceacSChris Mason { 38370f7d52f4SChris Mason struct btrfs_fs_info *fs_info = root->fs_info; 3838c146afadSYan Zheng int ret; 3839e089f05cSChris Mason 3840facda1e7SChris Mason fs_info->closing = 1; 3841a2135011SChris Mason smp_mb(); 3842a2135011SChris Mason 38437343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 38447343dd61SJustin Maggard btrfs_qgroup_wait_for_completion(fs_info); 38457343dd61SJustin Maggard 3846803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3847803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3848803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3849803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3850803b2f54SStefan Behrens 3851837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3852aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3853837d5b6eSIlya Dryomov 38548dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 38558dabb742SStefan Behrens 3856aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 38574cb5300bSChris Mason 38584cb5300bSChris Mason /* wait for any defraggers to finish */ 38594cb5300bSChris Mason wait_event(fs_info->transaction_wait, 38604cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 38614cb5300bSChris Mason 38624cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 386326176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 38644cb5300bSChris Mason 386521c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 386621c7e756SMiao Xie 3867c146afadSYan Zheng if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3868e44163e1SJeff Mahoney /* 3869e44163e1SJeff Mahoney * If the cleaner thread is stopped and there are 3870e44163e1SJeff Mahoney * block groups queued for removal, the deletion will be 3871e44163e1SJeff Mahoney * skipped when we quit the cleaner thread. 3872e44163e1SJeff Mahoney */ 3873e44163e1SJeff Mahoney btrfs_delete_unused_bgs(root->fs_info); 3874e44163e1SJeff Mahoney 3875c146afadSYan Zheng ret = btrfs_commit_super(root); 3876d397712bSChris Mason if (ret) 387704892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 3878c146afadSYan Zheng } 3879ed2ff2cbSChris Mason 388087533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 388168ce9682SStefan Behrens btrfs_error_commit_super(root); 3882acce952bSliubo 3883e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3884e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 38858929ecfaSYan, Zheng 3886f25784b3SYan Zheng fs_info->closing = 2; 3887f25784b3SYan Zheng smp_mb(); 3888f25784b3SYan Zheng 388904892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 3890bcef60f2SArne Jansen 3891963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 389204892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 3893963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3894b0c68f8bSChris Mason } 389531153d81SYan Zheng 38966618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 3897b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 38985ac1d209SJeff Mahoney 3899faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3900d10c5f31SChris Mason 39011a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 39021a4319ccSLiu Bo 39032b1360daSJosef Bacik btrfs_free_block_groups(fs_info); 39042b1360daSJosef Bacik 3905de348ee0SWang Shilong /* 3906de348ee0SWang Shilong * we must make sure there is not any read request to 3907de348ee0SWang Shilong * submit after we stopping all workers. 3908de348ee0SWang Shilong */ 3909de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 391096192499SJosef Bacik btrfs_stop_all_workers(fs_info); 391196192499SJosef Bacik 391247ab2a6cSJosef Bacik fs_info->open = 0; 391313e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 39149ad6b7bcSChris Mason 391513e6c37bSJosef Bacik iput(fs_info->btree_inode); 3916d6bfde87SChris Mason 391721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 391821adbd5cSStefan Behrens if (btrfs_test_opt(root, CHECK_INTEGRITY)) 391921adbd5cSStefan Behrens btrfsic_unmount(root, fs_info->fs_devices); 392021adbd5cSStefan Behrens #endif 392121adbd5cSStefan Behrens 3922dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 39230b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3924b248a415SChris Mason 3925e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3926963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3927c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 392804160088SChris Mason bdi_destroy(&fs_info->bdi); 392976dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 39300b86a832SChris Mason 393153b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 393253b381b3SDavid Woodhouse 3933cdfb080eSChris Mason __btrfs_free_block_rsv(root->orphan_block_rsv); 39341cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 393504216820SFilipe Manana 393604216820SFilipe Manana lock_chunks(root); 393704216820SFilipe Manana while (!list_empty(&fs_info->pinned_chunks)) { 393804216820SFilipe Manana struct extent_map *em; 393904216820SFilipe Manana 394004216820SFilipe Manana em = list_first_entry(&fs_info->pinned_chunks, 394104216820SFilipe Manana struct extent_map, list); 394204216820SFilipe Manana list_del_init(&em->list); 394304216820SFilipe Manana free_extent_map(em); 394404216820SFilipe Manana } 394504216820SFilipe Manana unlock_chunks(root); 3946eb60ceacSChris Mason } 3947eb60ceacSChris Mason 3948b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3949b9fab919SChris Mason int atomic) 3950ccd467d6SChris Mason { 39511259ab75SChris Mason int ret; 3952727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 39531259ab75SChris Mason 39540b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 39551259ab75SChris Mason if (!ret) 39561259ab75SChris Mason return ret; 39571259ab75SChris Mason 39581259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3959b9fab919SChris Mason parent_transid, atomic); 3960b9fab919SChris Mason if (ret == -EAGAIN) 3961b9fab919SChris Mason return ret; 39621259ab75SChris Mason return !ret; 39635f39d397SChris Mason } 39646702ed49SChris Mason 39655f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 39665f39d397SChris Mason { 396706ea65a3SJosef Bacik struct btrfs_root *root; 39685f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3969b9473439SChris Mason int was_dirty; 3970b4ce94deSChris Mason 397106ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 397206ea65a3SJosef Bacik /* 397306ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 397406ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 397506ea65a3SJosef Bacik * outside of the sanity tests. 397606ea65a3SJosef Bacik */ 397706ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 397806ea65a3SJosef Bacik return; 397906ea65a3SJosef Bacik #endif 398006ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3981b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 398231b1a2bdSJulia Lawall if (transid != root->fs_info->generation) 398331b1a2bdSJulia Lawall WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3984d397712bSChris Mason "found %llu running %llu\n", 3985c1c9ff7cSGeert Uytterhoeven buf->start, transid, root->fs_info->generation); 39860b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3987e2d84521SMiao Xie if (!was_dirty) 3988e2d84521SMiao Xie __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 3989e2d84521SMiao Xie buf->len, 3990e2d84521SMiao Xie root->fs_info->dirty_metadata_batch); 39911f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 39921f21ef0aSFilipe Manana if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { 39931f21ef0aSFilipe Manana btrfs_print_leaf(root, buf); 39941f21ef0aSFilipe Manana ASSERT(0); 39951f21ef0aSFilipe Manana } 39961f21ef0aSFilipe Manana #endif 3997eb60ceacSChris Mason } 3998eb60ceacSChris Mason 3999b53d3f5dSLiu Bo static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 4000b53d3f5dSLiu Bo int flush_delayed) 400135b7e476SChris Mason { 4002188de649SChris Mason /* 4003188de649SChris Mason * looks as though older kernels can get into trouble with 4004188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4005188de649SChris Mason */ 4006e2d84521SMiao Xie int ret; 4007d6bfde87SChris Mason 40086933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4009d6bfde87SChris Mason return; 4010d6bfde87SChris Mason 4011b53d3f5dSLiu Bo if (flush_delayed) 401216cdcec7SMiao Xie btrfs_balance_delayed_items(root); 401316cdcec7SMiao Xie 4014e2d84521SMiao Xie ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 4015e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 4016e2d84521SMiao Xie if (ret > 0) { 4017d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited( 4018d0e1d66bSNamjae Jeon root->fs_info->btree_inode->i_mapping); 401916cdcec7SMiao Xie } 402016cdcec7SMiao Xie } 402116cdcec7SMiao Xie 4022b53d3f5dSLiu Bo void btrfs_btree_balance_dirty(struct btrfs_root *root) 402316cdcec7SMiao Xie { 4024b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 1); 402535b7e476SChris Mason } 4026b53d3f5dSLiu Bo 4027b53d3f5dSLiu Bo void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 4028b53d3f5dSLiu Bo { 4029b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 0); 4030d6bfde87SChris Mason } 40316b80053dSChris Mason 4032ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 40336b80053dSChris Mason { 4034727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 40350b32f4bbSJosef Bacik return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 40366b80053dSChris Mason } 40370da5468fSChris Mason 4038fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 4039acce952bSliubo int read_only) 4040acce952bSliubo { 4041c926093eSDavid Sterba struct btrfs_super_block *sb = fs_info->super_copy; 4042319e4d06SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 4043319e4d06SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 4044c926093eSDavid Sterba int ret = 0; 4045c926093eSDavid Sterba 4046319e4d06SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 4047319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: no valid FS found\n"); 4048319e4d06SQu Wenruo ret = -EINVAL; 4049319e4d06SQu Wenruo } 4050319e4d06SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) 4051319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n", 4052319e4d06SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 405321e7626bSDavid Sterba if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 405421e7626bSDavid Sterba printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n", 405521e7626bSDavid Sterba btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 4056c926093eSDavid Sterba ret = -EINVAL; 4057c926093eSDavid Sterba } 405821e7626bSDavid Sterba if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 405921e7626bSDavid Sterba printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n", 406021e7626bSDavid Sterba btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 4061c926093eSDavid Sterba ret = -EINVAL; 4062c926093eSDavid Sterba } 406321e7626bSDavid Sterba if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 406421e7626bSDavid Sterba printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n", 406521e7626bSDavid Sterba btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 4066c926093eSDavid Sterba ret = -EINVAL; 4067c926093eSDavid Sterba } 4068c926093eSDavid Sterba 40691104a885SDavid Sterba /* 4070319e4d06SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 4071319e4d06SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 40721104a885SDavid Sterba */ 4073319e4d06SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 4074319e4d06SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 4075319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize); 407675d6ad38SDavid Sterba ret = -EINVAL; 407775d6ad38SDavid Sterba } 4078319e4d06SQu Wenruo /* Only PAGE SIZE is supported yet */ 407909cbfeafSKirill A. Shutemov if (sectorsize != PAGE_SIZE) { 4080319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n", 408109cbfeafSKirill A. Shutemov sectorsize, PAGE_SIZE); 4082319e4d06SQu Wenruo ret = -EINVAL; 4083319e4d06SQu Wenruo } 4084319e4d06SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 4085319e4d06SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 4086319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize); 4087319e4d06SQu Wenruo ret = -EINVAL; 4088319e4d06SQu Wenruo } 4089319e4d06SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 4090319e4d06SQu Wenruo printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n", 4091319e4d06SQu Wenruo le32_to_cpu(sb->__unused_leafsize), 4092319e4d06SQu Wenruo nodesize); 4093319e4d06SQu Wenruo ret = -EINVAL; 4094319e4d06SQu Wenruo } 4095319e4d06SQu Wenruo 4096319e4d06SQu Wenruo /* Root alignment check */ 4097319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 4098319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n", 4099319e4d06SQu Wenruo btrfs_super_root(sb)); 4100319e4d06SQu Wenruo ret = -EINVAL; 4101319e4d06SQu Wenruo } 4102319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 4103319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n", 4104319e4d06SQu Wenruo btrfs_super_chunk_root(sb)); 4105319e4d06SQu Wenruo ret = -EINVAL; 4106319e4d06SQu Wenruo } 4107319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 4108319e4d06SQu Wenruo printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n", 4109319e4d06SQu Wenruo btrfs_super_log_root(sb)); 411075d6ad38SDavid Sterba ret = -EINVAL; 411175d6ad38SDavid Sterba } 411275d6ad38SDavid Sterba 4113c926093eSDavid Sterba if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) { 4114c926093eSDavid Sterba printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n", 4115c926093eSDavid Sterba fs_info->fsid, sb->dev_item.fsid); 4116c926093eSDavid Sterba ret = -EINVAL; 4117c926093eSDavid Sterba } 4118c926093eSDavid Sterba 4119c926093eSDavid Sterba /* 4120c926093eSDavid Sterba * Hint to catch really bogus numbers, bitflips or so, more exact checks are 4121c926093eSDavid Sterba * done later 4122c926093eSDavid Sterba */ 412321e7626bSDavid Sterba if (btrfs_super_num_devices(sb) > (1UL << 31)) 4124c926093eSDavid Sterba printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n", 412521e7626bSDavid Sterba btrfs_super_num_devices(sb)); 412675d6ad38SDavid Sterba if (btrfs_super_num_devices(sb) == 0) { 412775d6ad38SDavid Sterba printk(KERN_ERR "BTRFS: number of devices is 0\n"); 412875d6ad38SDavid Sterba ret = -EINVAL; 412975d6ad38SDavid Sterba } 4130c926093eSDavid Sterba 413121e7626bSDavid Sterba if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { 4132c926093eSDavid Sterba printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n", 413321e7626bSDavid Sterba btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 4134c926093eSDavid Sterba ret = -EINVAL; 4135c926093eSDavid Sterba } 4136c926093eSDavid Sterba 4137c926093eSDavid Sterba /* 4138ce7fca5fSDavid Sterba * Obvious sys_chunk_array corruptions, it must hold at least one key 4139ce7fca5fSDavid Sterba * and one chunk 4140ce7fca5fSDavid Sterba */ 4141ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 4142ce7fca5fSDavid Sterba printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n", 4143ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4144ce7fca5fSDavid Sterba BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 4145ce7fca5fSDavid Sterba ret = -EINVAL; 4146ce7fca5fSDavid Sterba } 4147ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 4148ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)) { 4149d2207129SFabian Frederick printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n", 4150ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4151ce7fca5fSDavid Sterba sizeof(struct btrfs_disk_key) 4152ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)); 4153ce7fca5fSDavid Sterba ret = -EINVAL; 4154ce7fca5fSDavid Sterba } 4155ce7fca5fSDavid Sterba 4156ce7fca5fSDavid Sterba /* 4157c926093eSDavid Sterba * The generation is a global counter, we'll trust it more than the others 4158c926093eSDavid Sterba * but it's still possible that it's the one that's wrong. 4159c926093eSDavid Sterba */ 416021e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 4161c926093eSDavid Sterba printk(KERN_WARNING 4162c926093eSDavid Sterba "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n", 416321e7626bSDavid Sterba btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb)); 416421e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 416521e7626bSDavid Sterba && btrfs_super_cache_generation(sb) != (u64)-1) 4166c926093eSDavid Sterba printk(KERN_WARNING 4167c926093eSDavid Sterba "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n", 416821e7626bSDavid Sterba btrfs_super_generation(sb), btrfs_super_cache_generation(sb)); 4169c926093eSDavid Sterba 4170c926093eSDavid Sterba return ret; 4171fcd1f065SDavid Sterba } 4172fcd1f065SDavid Sterba 417348a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root) 4174acce952bSliubo { 4175acce952bSliubo mutex_lock(&root->fs_info->cleaner_mutex); 4176acce952bSliubo btrfs_run_delayed_iputs(root); 4177acce952bSliubo mutex_unlock(&root->fs_info->cleaner_mutex); 4178acce952bSliubo 4179acce952bSliubo down_write(&root->fs_info->cleanup_work_sem); 4180acce952bSliubo up_write(&root->fs_info->cleanup_work_sem); 4181acce952bSliubo 4182acce952bSliubo /* cleanup FS via transaction */ 4183acce952bSliubo btrfs_cleanup_transaction(root); 4184acce952bSliubo } 4185acce952bSliubo 4186143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4187acce952bSliubo { 4188acce952bSliubo struct btrfs_ordered_extent *ordered; 4189acce952bSliubo 4190199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4191779880efSJosef Bacik /* 4192779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4193779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4194779880efSJosef Bacik */ 4195199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4196779880efSJosef Bacik root_extent_list) 4197779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4198199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4199199c2a9cSMiao Xie } 4200199c2a9cSMiao Xie 4201199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4202199c2a9cSMiao Xie { 4203199c2a9cSMiao Xie struct btrfs_root *root; 4204199c2a9cSMiao Xie struct list_head splice; 4205199c2a9cSMiao Xie 4206199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4207199c2a9cSMiao Xie 4208199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4209199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4210199c2a9cSMiao Xie while (!list_empty(&splice)) { 4211199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4212199c2a9cSMiao Xie ordered_root); 42131de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 42141de2cfdeSJosef Bacik &fs_info->ordered_roots); 4215199c2a9cSMiao Xie 42162a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4217199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4218199c2a9cSMiao Xie 42192a85d9caSLiu Bo cond_resched(); 42202a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4221199c2a9cSMiao Xie } 4222199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 4223acce952bSliubo } 4224acce952bSliubo 422535a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 4226acce952bSliubo struct btrfs_root *root) 4227acce952bSliubo { 4228acce952bSliubo struct rb_node *node; 4229acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4230acce952bSliubo struct btrfs_delayed_ref_node *ref; 4231acce952bSliubo int ret = 0; 4232acce952bSliubo 4233acce952bSliubo delayed_refs = &trans->delayed_refs; 4234acce952bSliubo 4235acce952bSliubo spin_lock(&delayed_refs->lock); 4236d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4237cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4238efe120a0SFrank Holton btrfs_info(root->fs_info, "delayed_refs has NO entry"); 4239acce952bSliubo return ret; 4240acce952bSliubo } 4241acce952bSliubo 4242d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 4243d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 4244c6fc2454SQu Wenruo struct btrfs_delayed_ref_node *tmp; 4245e78417d1SJosef Bacik bool pin_bytes = false; 4246acce952bSliubo 4247d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4248d7df2c79SJosef Bacik href_node); 4249b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 4250d7df2c79SJosef Bacik atomic_inc(&head->node.refs); 425179787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 4252b939d1abSJosef Bacik 4253acce952bSliubo mutex_lock(&head->mutex); 4254b939d1abSJosef Bacik mutex_unlock(&head->mutex); 4255d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 4256e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 4257b939d1abSJosef Bacik continue; 4258b939d1abSJosef Bacik } 4259d7df2c79SJosef Bacik spin_lock(&head->lock); 4260c6fc2454SQu Wenruo list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, 4261c6fc2454SQu Wenruo list) { 4262d7df2c79SJosef Bacik ref->in_tree = 0; 4263c6fc2454SQu Wenruo list_del(&ref->list); 4264d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4265d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4266d7df2c79SJosef Bacik } 426754067ae9SJosef Bacik if (head->must_insert_reserved) 4268e78417d1SJosef Bacik pin_bytes = true; 426978a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4270acce952bSliubo delayed_refs->num_heads--; 4271d7df2c79SJosef Bacik if (head->processing == 0) 4272acce952bSliubo delayed_refs->num_heads_ready--; 4273d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4274d7df2c79SJosef Bacik head->node.in_tree = 0; 4275c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 4276d7df2c79SJosef Bacik spin_unlock(&head->lock); 4277acce952bSliubo spin_unlock(&delayed_refs->lock); 4278e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4279acce952bSliubo 4280d7df2c79SJosef Bacik if (pin_bytes) 4281d7df2c79SJosef Bacik btrfs_pin_extent(root, head->node.bytenr, 4282d7df2c79SJosef Bacik head->node.num_bytes, 1); 4283d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 4284acce952bSliubo cond_resched(); 4285acce952bSliubo spin_lock(&delayed_refs->lock); 4286acce952bSliubo } 4287acce952bSliubo 4288acce952bSliubo spin_unlock(&delayed_refs->lock); 4289acce952bSliubo 4290acce952bSliubo return ret; 4291acce952bSliubo } 4292acce952bSliubo 4293143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4294acce952bSliubo { 4295acce952bSliubo struct btrfs_inode *btrfs_inode; 4296acce952bSliubo struct list_head splice; 4297acce952bSliubo 4298acce952bSliubo INIT_LIST_HEAD(&splice); 4299acce952bSliubo 4300eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4301eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4302acce952bSliubo 4303acce952bSliubo while (!list_empty(&splice)) { 4304eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4305acce952bSliubo delalloc_inodes); 4306acce952bSliubo 4307acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 4308df0af1a5SMiao Xie clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 4309df0af1a5SMiao Xie &btrfs_inode->runtime_flags); 4310eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4311acce952bSliubo 4312acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 4313b216cbfbSMiao Xie 4314eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4315acce952bSliubo } 4316acce952bSliubo 4317eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4318eb73c1b7SMiao Xie } 4319eb73c1b7SMiao Xie 4320eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4321eb73c1b7SMiao Xie { 4322eb73c1b7SMiao Xie struct btrfs_root *root; 4323eb73c1b7SMiao Xie struct list_head splice; 4324eb73c1b7SMiao Xie 4325eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4326eb73c1b7SMiao Xie 4327eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4328eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4329eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4330eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4331eb73c1b7SMiao Xie delalloc_root); 4332eb73c1b7SMiao Xie list_del_init(&root->delalloc_root); 4333eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 4334eb73c1b7SMiao Xie BUG_ON(!root); 4335eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4336eb73c1b7SMiao Xie 4337eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 4338eb73c1b7SMiao Xie btrfs_put_fs_root(root); 4339eb73c1b7SMiao Xie 4340eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4341eb73c1b7SMiao Xie } 4342eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4343acce952bSliubo } 4344acce952bSliubo 4345acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 4346acce952bSliubo struct extent_io_tree *dirty_pages, 4347acce952bSliubo int mark) 4348acce952bSliubo { 4349acce952bSliubo int ret; 4350acce952bSliubo struct extent_buffer *eb; 4351acce952bSliubo u64 start = 0; 4352acce952bSliubo u64 end; 4353acce952bSliubo 4354acce952bSliubo while (1) { 4355acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4356e6138876SJosef Bacik mark, NULL); 4357acce952bSliubo if (ret) 4358acce952bSliubo break; 4359acce952bSliubo 436091166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4361acce952bSliubo while (start <= end) { 436201d58472SDaniel Dressler eb = btrfs_find_tree_block(root->fs_info, start); 4363707e8a07SDavid Sterba start += root->nodesize; 4364fd8b2b61SJosef Bacik if (!eb) 4365acce952bSliubo continue; 4366fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4367acce952bSliubo 4368fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4369fd8b2b61SJosef Bacik &eb->bflags)) 4370fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4371fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4372acce952bSliubo } 4373acce952bSliubo } 4374acce952bSliubo 4375acce952bSliubo return ret; 4376acce952bSliubo } 4377acce952bSliubo 4378acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 4379acce952bSliubo struct extent_io_tree *pinned_extents) 4380acce952bSliubo { 4381acce952bSliubo struct extent_io_tree *unpin; 4382acce952bSliubo u64 start; 4383acce952bSliubo u64 end; 4384acce952bSliubo int ret; 4385ed0eaa14SLiu Bo bool loop = true; 4386acce952bSliubo 4387acce952bSliubo unpin = pinned_extents; 4388ed0eaa14SLiu Bo again: 4389acce952bSliubo while (1) { 4390acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4391e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4392acce952bSliubo if (ret) 4393acce952bSliubo break; 4394acce952bSliubo 4395acce952bSliubo clear_extent_dirty(unpin, start, end, GFP_NOFS); 4396acce952bSliubo btrfs_error_unpin_extent_range(root, start, end); 4397acce952bSliubo cond_resched(); 4398acce952bSliubo } 4399acce952bSliubo 4400ed0eaa14SLiu Bo if (loop) { 4401ed0eaa14SLiu Bo if (unpin == &root->fs_info->freed_extents[0]) 4402ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[1]; 4403ed0eaa14SLiu Bo else 4404ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[0]; 4405ed0eaa14SLiu Bo loop = false; 4406ed0eaa14SLiu Bo goto again; 4407ed0eaa14SLiu Bo } 4408ed0eaa14SLiu Bo 4409acce952bSliubo return 0; 4410acce952bSliubo } 4411acce952bSliubo 441249b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 441349b25e05SJeff Mahoney struct btrfs_root *root) 441449b25e05SJeff Mahoney { 441549b25e05SJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, root); 441649b25e05SJeff Mahoney 44174a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 441849b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_blocked_wait); 441949b25e05SJeff Mahoney 44204a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 442149b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_wait); 442249b25e05SJeff Mahoney 442367cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 442467cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 442567cde344SMiao Xie 442649b25e05SJeff Mahoney btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 442749b25e05SJeff Mahoney EXTENT_DIRTY); 44286e841e32SLiu Bo btrfs_destroy_pinned_extent(root, 44296e841e32SLiu Bo root->fs_info->pinned_extents); 443049b25e05SJeff Mahoney 44314a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 44324a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 44334a9d8bdeSMiao Xie 443449b25e05SJeff Mahoney /* 443549b25e05SJeff Mahoney memset(cur_trans, 0, sizeof(*cur_trans)); 443649b25e05SJeff Mahoney kmem_cache_free(btrfs_transaction_cachep, cur_trans); 443749b25e05SJeff Mahoney */ 443849b25e05SJeff Mahoney } 443949b25e05SJeff Mahoney 444048a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root) 4441acce952bSliubo { 4442acce952bSliubo struct btrfs_transaction *t; 4443acce952bSliubo 4444acce952bSliubo mutex_lock(&root->fs_info->transaction_kthread_mutex); 4445acce952bSliubo 4446a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4447724e2315SJosef Bacik while (!list_empty(&root->fs_info->trans_list)) { 4448724e2315SJosef Bacik t = list_first_entry(&root->fs_info->trans_list, 4449724e2315SJosef Bacik struct btrfs_transaction, list); 4450724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 4451724e2315SJosef Bacik atomic_inc(&t->use_count); 4452724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4453724e2315SJosef Bacik btrfs_wait_for_commit(root, t->transid); 4454724e2315SJosef Bacik btrfs_put_transaction(t); 4455724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4456724e2315SJosef Bacik continue; 4457724e2315SJosef Bacik } 4458724e2315SJosef Bacik if (t == root->fs_info->running_transaction) { 4459724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 4460724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4461724e2315SJosef Bacik /* 4462724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4463724e2315SJosef Bacik * handle open currently for this transaction. 4464724e2315SJosef Bacik */ 4465724e2315SJosef Bacik wait_event(t->writer_wait, 4466724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4467724e2315SJosef Bacik } else { 4468724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4469724e2315SJosef Bacik } 4470724e2315SJosef Bacik btrfs_cleanup_one_transaction(t, root); 4471724e2315SJosef Bacik 4472724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4473724e2315SJosef Bacik if (t == root->fs_info->running_transaction) 4474ac673879SMiao Xie root->fs_info->running_transaction = NULL; 4475724e2315SJosef Bacik list_del_init(&t->list); 4476a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4477a4abeea4SJosef Bacik 4478724e2315SJosef Bacik btrfs_put_transaction(t); 4479724e2315SJosef Bacik trace_btrfs_transaction_commit(root); 4480724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4481724e2315SJosef Bacik } 4482724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4483199c2a9cSMiao Xie btrfs_destroy_all_ordered_extents(root->fs_info); 448467cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 448567cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 4486724e2315SJosef Bacik btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4487eb73c1b7SMiao Xie btrfs_destroy_all_delalloc_inodes(root->fs_info); 4488acce952bSliubo mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4489acce952bSliubo 4490acce952bSliubo return 0; 4491acce952bSliubo } 4492acce952bSliubo 4493e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 4494ce9adaa5SChris Mason .readpage_end_io_hook = btree_readpage_end_io_hook, 44954bb31e92SArne Jansen .readpage_io_failed_hook = btree_io_failed_hook, 44960b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 4497239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4498239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 44990da5468fSChris Mason }; 4500