1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 26cbd5570SChris Mason /* 36cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 46cbd5570SChris Mason */ 56cbd5570SChris Mason 6e20d96d6SChris Mason #include <linux/fs.h> 7d98237b3SChris Mason #include <linux/blkdev.h> 80f7d52f4SChris Mason #include <linux/radix-tree.h> 935b7e476SChris Mason #include <linux/writeback.h> 10d397712bSChris Mason #include <linux/buffer_head.h> 11ce9adaa5SChris Mason #include <linux/workqueue.h> 12a74a4b97SChris Mason #include <linux/kthread.h> 135a0e3ad6STejun Heo #include <linux/slab.h> 14784b4e29SChris Mason #include <linux/migrate.h> 157a36ddecSDavid Sterba #include <linux/ratelimit.h> 166463fe58SStefan Behrens #include <linux/uuid.h> 17803b2f54SStefan Behrens #include <linux/semaphore.h> 18540adea3SMasami Hiramatsu #include <linux/error-injection.h> 199678c543SNikolay Borisov #include <linux/crc32c.h> 207e75bf3fSDavid Sterba #include <asm/unaligned.h> 21eb60ceacSChris Mason #include "ctree.h" 22eb60ceacSChris Mason #include "disk-io.h" 23e089f05cSChris Mason #include "transaction.h" 240f7d52f4SChris Mason #include "btrfs_inode.h" 250b86a832SChris Mason #include "volumes.h" 26db94535dSChris Mason #include "print-tree.h" 27925baeddSChris Mason #include "locking.h" 28e02119d5SChris Mason #include "tree-log.h" 29fa9c0d79SChris Mason #include "free-space-cache.h" 3070f6d82eSOmar Sandoval #include "free-space-tree.h" 31581bb050SLi Zefan #include "inode-map.h" 3221adbd5cSStefan Behrens #include "check-integrity.h" 33606686eeSJosef Bacik #include "rcu-string.h" 348dabb742SStefan Behrens #include "dev-replace.h" 3553b381b3SDavid Woodhouse #include "raid56.h" 365ac1d209SJeff Mahoney #include "sysfs.h" 37fcebe456SJosef Bacik #include "qgroup.h" 38ebb8765bSAnand Jain #include "compression.h" 39557ea5ddSQu Wenruo #include "tree-checker.h" 40fd708b81SJosef Bacik #include "ref-verify.h" 41eb60ceacSChris Mason 42de0022b9SJosef Bacik #ifdef CONFIG_X86 43de0022b9SJosef Bacik #include <asm/cpufeature.h> 44de0022b9SJosef Bacik #endif 45de0022b9SJosef Bacik 46319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 47319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 48319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 49319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 50e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 51e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 52319e4d06SQu Wenruo 53e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops; 548b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 55143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 56acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 572ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 58143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 592ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 60acce952bSliubo struct extent_io_tree *dirty_pages, 61acce952bSliubo int mark); 622ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 63acce952bSliubo struct extent_io_tree *pinned_extents); 642ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 652ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 66ce9adaa5SChris Mason 67d352ac68SChris Mason /* 6897eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 6997eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 70d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 71d352ac68SChris Mason */ 7297eb6b69SDavid Sterba struct btrfs_end_io_wq { 73ce9adaa5SChris Mason struct bio *bio; 74ce9adaa5SChris Mason bio_end_io_t *end_io; 75ce9adaa5SChris Mason void *private; 76ce9adaa5SChris Mason struct btrfs_fs_info *info; 774e4cbee9SChristoph Hellwig blk_status_t status; 78bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 798b712842SChris Mason struct btrfs_work work; 80ce9adaa5SChris Mason }; 810da5468fSChris Mason 8297eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 8397eb6b69SDavid Sterba 8497eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 8597eb6b69SDavid Sterba { 8697eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 8797eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 8897eb6b69SDavid Sterba 0, 89fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 9097eb6b69SDavid Sterba NULL); 9197eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 9297eb6b69SDavid Sterba return -ENOMEM; 9397eb6b69SDavid Sterba return 0; 9497eb6b69SDavid Sterba } 9597eb6b69SDavid Sterba 96e67c718bSDavid Sterba void __cold btrfs_end_io_wq_exit(void) 9797eb6b69SDavid Sterba { 9897eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 9997eb6b69SDavid Sterba } 10097eb6b69SDavid Sterba 101d352ac68SChris Mason /* 102d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 103d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 104d352ac68SChris Mason * just before they are sent down the IO stack. 105d352ac68SChris Mason */ 10644b8bd7eSChris Mason struct async_submit_bio { 107c6100a4bSJosef Bacik void *private_data; 10844b8bd7eSChris Mason struct bio *bio; 109a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 11044b8bd7eSChris Mason int mirror_num; 111eaf25d93SChris Mason /* 112eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 113eaf25d93SChris Mason * can't tell us where in the file the bio should go 114eaf25d93SChris Mason */ 115eaf25d93SChris Mason u64 bio_offset; 1168b712842SChris Mason struct btrfs_work work; 1174e4cbee9SChristoph Hellwig blk_status_t status; 11844b8bd7eSChris Mason }; 11944b8bd7eSChris Mason 12085d4e461SChris Mason /* 12185d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12285d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 12385d4e461SChris Mason * the level the eb occupies in the tree. 1244008c04aSChris Mason * 12585d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 12685d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 12785d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 1284fd786e6SMisono Tomohiro * by btrfs_root->root_key.objectid. This ensures that all special purpose 1294fd786e6SMisono Tomohiro * roots have separate keysets. 1304008c04aSChris Mason * 13185d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13285d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 13385d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1344008c04aSChris Mason * 13585d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 13685d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 13785d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 13885d4e461SChris Mason * 13985d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14085d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14185d4e461SChris Mason * needs update as well. 1424008c04aSChris Mason */ 1434008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1444008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1454008c04aSChris Mason # error 1464008c04aSChris Mason # endif 14785d4e461SChris Mason 14885d4e461SChris Mason static struct btrfs_lockdep_keyset { 14985d4e461SChris Mason u64 id; /* root objectid */ 15085d4e461SChris Mason const char *name_stem; /* lock name stem */ 15185d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 15285d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 15385d4e461SChris Mason } btrfs_lockdep_keysets[] = { 15485d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 15585d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 15685d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 15785d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 15885d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 15985d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 16060b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 16185d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 16285d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 16385d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 16413fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1656b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 16685d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1674008c04aSChris Mason }; 16885d4e461SChris Mason 16985d4e461SChris Mason void __init btrfs_init_lockdep(void) 17085d4e461SChris Mason { 17185d4e461SChris Mason int i, j; 17285d4e461SChris Mason 17385d4e461SChris Mason /* initialize lockdep class names */ 17485d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 17585d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 17685d4e461SChris Mason 17785d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 17885d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 17985d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 18085d4e461SChris Mason } 18185d4e461SChris Mason } 18285d4e461SChris Mason 18385d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 18485d4e461SChris Mason int level) 18585d4e461SChris Mason { 18685d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 18785d4e461SChris Mason 18885d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 18985d4e461SChris Mason 19085d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 19185d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 19285d4e461SChris Mason if (ks->id == objectid) 19385d4e461SChris Mason break; 19485d4e461SChris Mason 19585d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 19685d4e461SChris Mason &ks->keys[level], ks->names[level]); 19785d4e461SChris Mason } 19885d4e461SChris Mason 1994008c04aSChris Mason #endif 2004008c04aSChris Mason 201d352ac68SChris Mason /* 202d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 203d352ac68SChris Mason * that covers the entire device 204d352ac68SChris Mason */ 2056af49dbdSDavid Sterba struct extent_map *btree_get_extent(struct btrfs_inode *inode, 206306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 2075f39d397SChris Mason int create) 2085f39d397SChris Mason { 2093ffbd68cSDavid Sterba struct btrfs_fs_info *fs_info = inode->root->fs_info; 210fc4f21b1SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 2115f39d397SChris Mason struct extent_map *em; 2125f39d397SChris Mason int ret; 2135f39d397SChris Mason 214890871beSChris Mason read_lock(&em_tree->lock); 215d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 216a061fc8dSChris Mason if (em) { 2170b246afaSJeff Mahoney em->bdev = fs_info->fs_devices->latest_bdev; 218890871beSChris Mason read_unlock(&em_tree->lock); 2195f39d397SChris Mason goto out; 220a061fc8dSChris Mason } 221890871beSChris Mason read_unlock(&em_tree->lock); 2227b13b7b1SChris Mason 223172ddd60SDavid Sterba em = alloc_extent_map(); 2245f39d397SChris Mason if (!em) { 2255f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2265f39d397SChris Mason goto out; 2275f39d397SChris Mason } 2285f39d397SChris Mason em->start = 0; 2290afbaf8cSChris Mason em->len = (u64)-1; 230c8b97818SChris Mason em->block_len = (u64)-1; 2315f39d397SChris Mason em->block_start = 0; 2320b246afaSJeff Mahoney em->bdev = fs_info->fs_devices->latest_bdev; 233d1310b2eSChris Mason 234890871beSChris Mason write_lock(&em_tree->lock); 23509a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2365f39d397SChris Mason if (ret == -EEXIST) { 2375f39d397SChris Mason free_extent_map(em); 2387b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 239b4f359abSTsutomu Itoh if (!em) 2400433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2415f39d397SChris Mason } else if (ret) { 2427b13b7b1SChris Mason free_extent_map(em); 2430433f20dSTsutomu Itoh em = ERR_PTR(ret); 2445f39d397SChris Mason } 245890871beSChris Mason write_unlock(&em_tree->lock); 2467b13b7b1SChris Mason 2475f39d397SChris Mason out: 2485f39d397SChris Mason return em; 2495f39d397SChris Mason } 2505f39d397SChris Mason 2519ed57367SDavid Sterba u32 btrfs_csum_data(const char *data, u32 seed, size_t len) 25219c00ddcSChris Mason { 2539678c543SNikolay Borisov return crc32c(seed, data, len); 25419c00ddcSChris Mason } 25519c00ddcSChris Mason 2560b5e3dafSDomagoj Tršan void btrfs_csum_final(u32 crc, u8 *result) 25719c00ddcSChris Mason { 2587e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 25919c00ddcSChris Mason } 26019c00ddcSChris Mason 261d352ac68SChris Mason /* 262d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 263d352ac68SChris Mason * into the csum field of the block. 264d352ac68SChris Mason */ 26501d58472SDaniel Dressler static int csum_tree_block(struct btrfs_fs_info *fs_info, 26601d58472SDaniel Dressler struct extent_buffer *buf, 26719c00ddcSChris Mason int verify) 26819c00ddcSChris Mason { 26901d58472SDaniel Dressler u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 27071a63551SDavid Sterba char result[BTRFS_CSUM_SIZE]; 27119c00ddcSChris Mason unsigned long len; 27219c00ddcSChris Mason unsigned long cur_len; 27319c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 27419c00ddcSChris Mason char *kaddr; 27519c00ddcSChris Mason unsigned long map_start; 27619c00ddcSChris Mason unsigned long map_len; 27719c00ddcSChris Mason int err; 27819c00ddcSChris Mason u32 crc = ~(u32)0; 27919c00ddcSChris Mason 28019c00ddcSChris Mason len = buf->len - offset; 28119c00ddcSChris Mason while (len > 0) { 28219c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 283a6591715SChris Mason &kaddr, &map_start, &map_len); 284d397712bSChris Mason if (err) 2858bd98f0eSAlex Lyakas return err; 28619c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 287b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 28819c00ddcSChris Mason crc, cur_len); 28919c00ddcSChris Mason len -= cur_len; 29019c00ddcSChris Mason offset += cur_len; 29119c00ddcSChris Mason } 29271a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 293607d432dSJosef Bacik 29419c00ddcSChris Mason btrfs_csum_final(crc, result); 29519c00ddcSChris Mason 29619c00ddcSChris Mason if (verify) { 297607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 298e4204dedSChris Mason u32 val; 299e4204dedSChris Mason u32 found = 0; 300607d432dSJosef Bacik memcpy(&found, result, csum_size); 301e4204dedSChris Mason 302607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 30394647322SDavid Sterba btrfs_warn_rl(fs_info, 3045d163e0eSJeff Mahoney "%s checksum verify failed on %llu wanted %X found %X level %d", 30501d58472SDaniel Dressler fs_info->sb->s_id, buf->start, 306c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 3078bd98f0eSAlex Lyakas return -EUCLEAN; 30819c00ddcSChris Mason } 30919c00ddcSChris Mason } else { 310607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 31119c00ddcSChris Mason } 31271a63551SDavid Sterba 31319c00ddcSChris Mason return 0; 31419c00ddcSChris Mason } 31519c00ddcSChris Mason 316d352ac68SChris Mason /* 317d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 318d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 319d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 320d352ac68SChris Mason * in the wrong place. 321d352ac68SChris Mason */ 3221259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 323b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 324b9fab919SChris Mason int atomic) 3251259ab75SChris Mason { 3262ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3271259ab75SChris Mason int ret; 3282755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 3291259ab75SChris Mason 3301259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3311259ab75SChris Mason return 0; 3321259ab75SChris Mason 333b9fab919SChris Mason if (atomic) 334b9fab919SChris Mason return -EAGAIN; 335b9fab919SChris Mason 336a26e8c9fSJosef Bacik if (need_lock) { 337a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 338a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 339a26e8c9fSJosef Bacik } 340a26e8c9fSJosef Bacik 3412ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 342ff13db41SDavid Sterba &cached_state); 3430b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3441259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3451259ab75SChris Mason ret = 0; 3461259ab75SChris Mason goto out; 3471259ab75SChris Mason } 34894647322SDavid Sterba btrfs_err_rl(eb->fs_info, 34994647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 35094647322SDavid Sterba eb->start, 35129549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3521259ab75SChris Mason ret = 1; 353a26e8c9fSJosef Bacik 354a26e8c9fSJosef Bacik /* 355a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 356a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 35701327610SNicholas D Steeves * block that has been freed and re-allocated. So don't clear uptodate 358a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 359a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 360a26e8c9fSJosef Bacik * making everybody very sad. 361a26e8c9fSJosef Bacik */ 362a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3630b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 36433958dc6SChris Mason out: 3652ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 366e43bbe5eSDavid Sterba &cached_state); 367472b909fSJosef Bacik if (need_lock) 368a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3691259ab75SChris Mason return ret; 3701259ab75SChris Mason } 3711259ab75SChris Mason 372d352ac68SChris Mason /* 3731104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3741104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3751104a885SDavid Sterba */ 376ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 377ab8d0fc4SJeff Mahoney char *raw_disk_sb) 3781104a885SDavid Sterba { 3791104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3801104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 3811104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 3821104a885SDavid Sterba int ret = 0; 3831104a885SDavid Sterba 3841104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 3851104a885SDavid Sterba u32 crc = ~(u32)0; 386776c4a7cSNikolay Borisov char result[sizeof(crc)]; 3871104a885SDavid Sterba 3881104a885SDavid Sterba /* 3891104a885SDavid Sterba * The super_block structure does not span the whole 3901104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 39101327610SNicholas D Steeves * is filled with zeros and is included in the checksum. 3921104a885SDavid Sterba */ 3931104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 3941104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3951104a885SDavid Sterba btrfs_csum_final(crc, result); 3961104a885SDavid Sterba 397776c4a7cSNikolay Borisov if (memcmp(raw_disk_sb, result, sizeof(result))) 3981104a885SDavid Sterba ret = 1; 3991104a885SDavid Sterba } 4001104a885SDavid Sterba 4011104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 402ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "unsupported checksum algorithm %u", 4031104a885SDavid Sterba csum_type); 4041104a885SDavid Sterba ret = 1; 4051104a885SDavid Sterba } 4061104a885SDavid Sterba 4071104a885SDavid Sterba return ret; 4081104a885SDavid Sterba } 4091104a885SDavid Sterba 410581c1760SQu Wenruo static int verify_level_key(struct btrfs_fs_info *fs_info, 411581c1760SQu Wenruo struct extent_buffer *eb, int level, 412ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 413581c1760SQu Wenruo { 414581c1760SQu Wenruo int found_level; 415581c1760SQu Wenruo struct btrfs_key found_key; 416581c1760SQu Wenruo int ret; 417581c1760SQu Wenruo 418581c1760SQu Wenruo found_level = btrfs_header_level(eb); 419581c1760SQu Wenruo if (found_level != level) { 420581c1760SQu Wenruo #ifdef CONFIG_BTRFS_DEBUG 421581c1760SQu Wenruo WARN_ON(1); 422581c1760SQu Wenruo btrfs_err(fs_info, 423581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 424581c1760SQu Wenruo eb->start, level, found_level); 425581c1760SQu Wenruo #endif 426581c1760SQu Wenruo return -EIO; 427581c1760SQu Wenruo } 428581c1760SQu Wenruo 429581c1760SQu Wenruo if (!first_key) 430581c1760SQu Wenruo return 0; 431581c1760SQu Wenruo 4325d41be6fSQu Wenruo /* 4335d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 4345d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 4355d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 4365d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 4375d41be6fSQu Wenruo */ 4385d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 4395d41be6fSQu Wenruo return 0; 440581c1760SQu Wenruo if (found_level) 441581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 442581c1760SQu Wenruo else 443581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 444581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 445581c1760SQu Wenruo 446581c1760SQu Wenruo #ifdef CONFIG_BTRFS_DEBUG 447581c1760SQu Wenruo if (ret) { 448581c1760SQu Wenruo WARN_ON(1); 449581c1760SQu Wenruo btrfs_err(fs_info, 450ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 451ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 452ff76a864SLiu Bo first_key->type, first_key->offset, 453ff76a864SLiu Bo found_key.objectid, found_key.type, 454ff76a864SLiu Bo found_key.offset); 455581c1760SQu Wenruo } 456581c1760SQu Wenruo #endif 457581c1760SQu Wenruo return ret; 458581c1760SQu Wenruo } 459581c1760SQu Wenruo 4601104a885SDavid Sterba /* 461d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 462d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 463581c1760SQu Wenruo * 464581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 465581c1760SQu Wenruo * @level: expected level, mandatory check 466581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 467d352ac68SChris Mason */ 4682ff7e61eSJeff Mahoney static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info, 469f188591eSChris Mason struct extent_buffer *eb, 470581c1760SQu Wenruo u64 parent_transid, int level, 471581c1760SQu Wenruo struct btrfs_key *first_key) 472f188591eSChris Mason { 473f188591eSChris Mason struct extent_io_tree *io_tree; 474ea466794SJosef Bacik int failed = 0; 475f188591eSChris Mason int ret; 476f188591eSChris Mason int num_copies = 0; 477f188591eSChris Mason int mirror_num = 0; 478ea466794SJosef Bacik int failed_mirror = 0; 479f188591eSChris Mason 4800b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 481f188591eSChris Mason while (1) { 482f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 4838436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, 4846af49dbdSDavid Sterba mirror_num); 485256dd1bbSStefan Behrens if (!ret) { 486581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 487b9fab919SChris Mason parent_transid, 0)) 488256dd1bbSStefan Behrens ret = -EIO; 489581c1760SQu Wenruo else if (verify_level_key(fs_info, eb, level, 490ff76a864SLiu Bo first_key, parent_transid)) 491581c1760SQu Wenruo ret = -EUCLEAN; 492581c1760SQu Wenruo else 493581c1760SQu Wenruo break; 494256dd1bbSStefan Behrens } 495d397712bSChris Mason 4960b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 497f188591eSChris Mason eb->start, eb->len); 4984235298eSChris Mason if (num_copies == 1) 499ea466794SJosef Bacik break; 5004235298eSChris Mason 5015cf1ab56SJosef Bacik if (!failed_mirror) { 5025cf1ab56SJosef Bacik failed = 1; 5035cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 5045cf1ab56SJosef Bacik } 5055cf1ab56SJosef Bacik 506f188591eSChris Mason mirror_num++; 507ea466794SJosef Bacik if (mirror_num == failed_mirror) 508ea466794SJosef Bacik mirror_num++; 509ea466794SJosef Bacik 5104235298eSChris Mason if (mirror_num > num_copies) 511ea466794SJosef Bacik break; 512f188591eSChris Mason } 513ea466794SJosef Bacik 514c0901581SStefan Behrens if (failed && !ret && failed_mirror) 5152ff7e61eSJeff Mahoney repair_eb_io_failure(fs_info, eb, failed_mirror); 516ea466794SJosef Bacik 517ea466794SJosef Bacik return ret; 518f188591eSChris Mason } 51919c00ddcSChris Mason 520d352ac68SChris Mason /* 521d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 522d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 523d352ac68SChris Mason */ 524d397712bSChris Mason 52501d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 52619c00ddcSChris Mason { 5274eee4fa4SMiao Xie u64 start = page_offset(page); 52819c00ddcSChris Mason u64 found_start; 52919c00ddcSChris Mason struct extent_buffer *eb; 530f188591eSChris Mason 5314f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5324f2de97aSJosef Bacik if (page != eb->pages[0]) 5334f2de97aSJosef Bacik return 0; 5340f805531SAlex Lyakas 53519c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 5360f805531SAlex Lyakas /* 5370f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5380f805531SAlex Lyakas * It is useful to know what went wrong. 5390f805531SAlex Lyakas */ 5400f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5410f805531SAlex Lyakas return -EUCLEAN; 5420f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5430f805531SAlex Lyakas return -EUCLEAN; 5440f805531SAlex Lyakas 5450f805531SAlex Lyakas ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, 5460f805531SAlex Lyakas btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); 5470f805531SAlex Lyakas 5488bd98f0eSAlex Lyakas return csum_tree_block(fs_info, eb, 0); 54919c00ddcSChris Mason } 55019c00ddcSChris Mason 55101d58472SDaniel Dressler static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, 5522b82032cSYan Zheng struct extent_buffer *eb) 5532b82032cSYan Zheng { 55401d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 55544880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 5562b82032cSYan Zheng int ret = 1; 5572b82032cSYan Zheng 5580a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5592b82032cSYan Zheng while (fs_devices) { 5602b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5612b82032cSYan Zheng ret = 0; 5622b82032cSYan Zheng break; 5632b82032cSYan Zheng } 5642b82032cSYan Zheng fs_devices = fs_devices->seed; 5652b82032cSYan Zheng } 5662b82032cSYan Zheng return ret; 5672b82032cSYan Zheng } 5682b82032cSYan Zheng 569facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 570facc8a22SMiao Xie u64 phy_offset, struct page *page, 571facc8a22SMiao Xie u64 start, u64 end, int mirror) 572ce9adaa5SChris Mason { 573ce9adaa5SChris Mason u64 found_start; 574ce9adaa5SChris Mason int found_level; 575ce9adaa5SChris Mason struct extent_buffer *eb; 576ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 57702873e43SZhao Lei struct btrfs_fs_info *fs_info = root->fs_info; 578f188591eSChris Mason int ret = 0; 579727011e0SChris Mason int reads_done; 580ce9adaa5SChris Mason 581ce9adaa5SChris Mason if (!page->private) 582ce9adaa5SChris Mason goto out; 583d397712bSChris Mason 5844f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 585d397712bSChris Mason 5860b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 5870b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 5880b32f4bbSJosef Bacik */ 5890b32f4bbSJosef Bacik extent_buffer_get(eb); 5900b32f4bbSJosef Bacik 5910b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 592727011e0SChris Mason if (!reads_done) 593727011e0SChris Mason goto err; 594f188591eSChris Mason 5955cf1ab56SJosef Bacik eb->read_mirror = mirror; 596656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 597ea466794SJosef Bacik ret = -EIO; 598ea466794SJosef Bacik goto err; 599ea466794SJosef Bacik } 600ea466794SJosef Bacik 601ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 602727011e0SChris Mason if (found_start != eb->start) { 603893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 604893bf4b1SSu Yue eb->start, found_start); 605f188591eSChris Mason ret = -EIO; 606ce9adaa5SChris Mason goto err; 607ce9adaa5SChris Mason } 60802873e43SZhao Lei if (check_tree_block_fsid(fs_info, eb)) { 60902873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 61094647322SDavid Sterba eb->start); 6111259ab75SChris Mason ret = -EIO; 6121259ab75SChris Mason goto err; 6131259ab75SChris Mason } 614ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6151c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 616893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 617893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 6181c24c3ceSJosef Bacik ret = -EIO; 6191c24c3ceSJosef Bacik goto err; 6201c24c3ceSJosef Bacik } 621ce9adaa5SChris Mason 62285d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 62385d4e461SChris Mason eb, found_level); 6244008c04aSChris Mason 62502873e43SZhao Lei ret = csum_tree_block(fs_info, eb, 1); 6268bd98f0eSAlex Lyakas if (ret) 627a826d6dcSJosef Bacik goto err; 628a826d6dcSJosef Bacik 629a826d6dcSJosef Bacik /* 630a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 631a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 632a826d6dcSJosef Bacik * return -EIO. 633a826d6dcSJosef Bacik */ 6342f659546SQu Wenruo if (found_level == 0 && btrfs_check_leaf_full(fs_info, eb)) { 635a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 636a826d6dcSJosef Bacik ret = -EIO; 637a826d6dcSJosef Bacik } 638ce9adaa5SChris Mason 6392f659546SQu Wenruo if (found_level > 0 && btrfs_check_node(fs_info, eb)) 640053ab70fSLiu Bo ret = -EIO; 641053ab70fSLiu Bo 6420b32f4bbSJosef Bacik if (!ret) 6430b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 644ce9adaa5SChris Mason err: 64579fb65a1SJosef Bacik if (reads_done && 64679fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 647d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6484bb31e92SArne Jansen 64953b381b3SDavid Woodhouse if (ret) { 65053b381b3SDavid Woodhouse /* 65153b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 65253b381b3SDavid Woodhouse * again, we have to make sure it has something 65353b381b3SDavid Woodhouse * to decrement 65453b381b3SDavid Woodhouse */ 65553b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6560b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 65753b381b3SDavid Woodhouse } 6580b32f4bbSJosef Bacik free_extent_buffer(eb); 659ce9adaa5SChris Mason out: 660f188591eSChris Mason return ret; 661ce9adaa5SChris Mason } 662ce9adaa5SChris Mason 663ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6644bb31e92SArne Jansen { 6654bb31e92SArne Jansen struct extent_buffer *eb; 6664bb31e92SArne Jansen 6674f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 668656f30dbSFilipe Manana set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); 6695cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 67053b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 671ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 672d48d71aaSDavid Sterba btree_readahead_hook(eb, -EIO); 6734bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6744bb31e92SArne Jansen } 6754bb31e92SArne Jansen 6764246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 677ce9adaa5SChris Mason { 67897eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 679ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6809e0af237SLiu Bo struct btrfs_workqueue *wq; 6819e0af237SLiu Bo btrfs_work_func_t func; 682ce9adaa5SChris Mason 683ce9adaa5SChris Mason fs_info = end_io_wq->info; 6844e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 685d20f7043SChris Mason 68637226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 6879e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 6889e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 6899e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 6909e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 6919e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 6929e0af237SLiu Bo func = btrfs_freespace_write_helper; 6939e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 6949e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 6959e0af237SLiu Bo func = btrfs_endio_raid56_helper; 696d20f7043SChris Mason } else { 6979e0af237SLiu Bo wq = fs_info->endio_write_workers; 6989e0af237SLiu Bo func = btrfs_endio_write_helper; 699d20f7043SChris Mason } 7009e0af237SLiu Bo } else { 7018b110e39SMiao Xie if (unlikely(end_io_wq->metadata == 7028b110e39SMiao Xie BTRFS_WQ_ENDIO_DIO_REPAIR)) { 7038b110e39SMiao Xie wq = fs_info->endio_repair_workers; 7048b110e39SMiao Xie func = btrfs_endio_repair_helper; 7058b110e39SMiao Xie } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7069e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7079e0af237SLiu Bo func = btrfs_endio_raid56_helper; 7089e0af237SLiu Bo } else if (end_io_wq->metadata) { 7099e0af237SLiu Bo wq = fs_info->endio_meta_workers; 7109e0af237SLiu Bo func = btrfs_endio_meta_helper; 7119e0af237SLiu Bo } else { 7129e0af237SLiu Bo wq = fs_info->endio_workers; 7139e0af237SLiu Bo func = btrfs_endio_helper; 7149e0af237SLiu Bo } 7159e0af237SLiu Bo } 7169e0af237SLiu Bo 7179e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 7189e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 719ce9adaa5SChris Mason } 720ce9adaa5SChris Mason 7214e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 722bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7230b86a832SChris Mason { 72497eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7258b110e39SMiao Xie 72697eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 727ce9adaa5SChris Mason if (!end_io_wq) 7284e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 729ce9adaa5SChris Mason 730ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 731ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 73222c59948SChris Mason end_io_wq->info = info; 7334e4cbee9SChristoph Hellwig end_io_wq->status = 0; 734ce9adaa5SChris Mason end_io_wq->bio = bio; 73522c59948SChris Mason end_io_wq->metadata = metadata; 736ce9adaa5SChris Mason 737ce9adaa5SChris Mason bio->bi_private = end_io_wq; 738ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 73922c59948SChris Mason return 0; 74022c59948SChris Mason } 74122c59948SChris Mason 7424a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7434a69a410SChris Mason { 7444a69a410SChris Mason struct async_submit_bio *async; 7454e4cbee9SChristoph Hellwig blk_status_t ret; 7464a69a410SChris Mason 7474a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 748c6100a4bSJosef Bacik ret = async->submit_bio_start(async->private_data, async->bio, 749eaf25d93SChris Mason async->bio_offset); 75079787eaaSJeff Mahoney if (ret) 7514e4cbee9SChristoph Hellwig async->status = ret; 7524a69a410SChris Mason } 7534a69a410SChris Mason 7544a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7558b712842SChris Mason { 7568b712842SChris Mason struct async_submit_bio *async; 7578b712842SChris Mason 7588b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7594854ddd0SChris Mason 760bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7614e4cbee9SChristoph Hellwig if (async->status) { 7624e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7634246a0b6SChristoph Hellwig bio_endio(async->bio); 76479787eaaSJeff Mahoney return; 76579787eaaSJeff Mahoney } 76679787eaaSJeff Mahoney 767e288c080SDavid Sterba btrfs_submit_bio_done(async->private_data, async->bio, async->mirror_num); 7684a69a410SChris Mason } 7694a69a410SChris Mason 7704a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7714a69a410SChris Mason { 7724a69a410SChris Mason struct async_submit_bio *async; 7734a69a410SChris Mason 7744a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7758b712842SChris Mason kfree(async); 7768b712842SChris Mason } 7778b712842SChris Mason 7788c27cb35SLinus Torvalds blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 779c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 780c6100a4bSJosef Bacik u64 bio_offset, void *private_data, 781e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 78244b8bd7eSChris Mason { 78344b8bd7eSChris Mason struct async_submit_bio *async; 78444b8bd7eSChris Mason 78544b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 78644b8bd7eSChris Mason if (!async) 7874e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 78844b8bd7eSChris Mason 789c6100a4bSJosef Bacik async->private_data = private_data; 79044b8bd7eSChris Mason async->bio = bio; 79144b8bd7eSChris Mason async->mirror_num = mirror_num; 7924a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7934a69a410SChris Mason 7949e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 7955cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 7964a69a410SChris Mason 797eaf25d93SChris Mason async->bio_offset = bio_offset; 7988c8bee1dSChris Mason 7994e4cbee9SChristoph Hellwig async->status = 0; 80079787eaaSJeff Mahoney 80167f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 8025cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 803d313d7a3SChris Mason 8045cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 80544b8bd7eSChris Mason return 0; 80644b8bd7eSChris Mason } 80744b8bd7eSChris Mason 8084e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 809ce3ed71aSChris Mason { 8102c30c71bSKent Overstreet struct bio_vec *bvec; 811ce3ed71aSChris Mason struct btrfs_root *root; 8122c30c71bSKent Overstreet int i, ret = 0; 813ce3ed71aSChris Mason 814c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8152c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 816ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 81701d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 81879787eaaSJeff Mahoney if (ret) 81979787eaaSJeff Mahoney break; 820ce3ed71aSChris Mason } 8212c30c71bSKent Overstreet 8224e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 823ce3ed71aSChris Mason } 824ce3ed71aSChris Mason 825d0ee3934SDavid Sterba static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, 826eaf25d93SChris Mason u64 bio_offset) 82722c59948SChris Mason { 8288b712842SChris Mason /* 8298b712842SChris Mason * when we're called for a write, we're already in the async 8305443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8318b712842SChris Mason */ 83279787eaaSJeff Mahoney return btree_csum_one_bio(bio); 83322c59948SChris Mason } 83422c59948SChris Mason 83518fdc679SLiu Bo static int check_async_write(struct btrfs_inode *bi) 836de0022b9SJosef Bacik { 8376300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8386300463bSLiu Bo return 0; 839de0022b9SJosef Bacik #ifdef CONFIG_X86 840bc696ca0SBorislav Petkov if (static_cpu_has(X86_FEATURE_XMM4_2)) 841de0022b9SJosef Bacik return 0; 842de0022b9SJosef Bacik #endif 843de0022b9SJosef Bacik return 1; 844de0022b9SJosef Bacik } 845de0022b9SJosef Bacik 8468c27cb35SLinus Torvalds static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, 847eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 848eaf25d93SChris Mason u64 bio_offset) 84944b8bd7eSChris Mason { 850c6100a4bSJosef Bacik struct inode *inode = private_data; 8510b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 85218fdc679SLiu Bo int async = check_async_write(BTRFS_I(inode)); 8534e4cbee9SChristoph Hellwig blk_status_t ret; 854cad321adSChris Mason 85537226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 856cad321adSChris Mason /* 857cad321adSChris Mason * called for a read, do the setup so that checksum validation 858cad321adSChris Mason * can happen in the async kernel threads 859cad321adSChris Mason */ 8600b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8610b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8621d4284bdSChris Mason if (ret) 86361891923SStefan Behrens goto out_w_error; 8642ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 865de0022b9SJosef Bacik } else if (!async) { 866de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 867de0022b9SJosef Bacik if (ret) 86861891923SStefan Behrens goto out_w_error; 8692ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 87061891923SStefan Behrens } else { 871cad321adSChris Mason /* 87261891923SStefan Behrens * kthread helpers are used to submit writes so that 87361891923SStefan Behrens * checksumming can happen in parallel across all CPUs 874cad321adSChris Mason */ 875c6100a4bSJosef Bacik ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, 876c6100a4bSJosef Bacik bio_offset, private_data, 877e288c080SDavid Sterba btree_submit_bio_start); 87844b8bd7eSChris Mason } 87944b8bd7eSChris Mason 8804246a0b6SChristoph Hellwig if (ret) 8814246a0b6SChristoph Hellwig goto out_w_error; 8824246a0b6SChristoph Hellwig return 0; 8834246a0b6SChristoph Hellwig 88461891923SStefan Behrens out_w_error: 8854e4cbee9SChristoph Hellwig bio->bi_status = ret; 8864246a0b6SChristoph Hellwig bio_endio(bio); 88761891923SStefan Behrens return ret; 88861891923SStefan Behrens } 88961891923SStefan Behrens 8903dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 891784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 892a6bc32b8SMel Gorman struct page *newpage, struct page *page, 893a6bc32b8SMel Gorman enum migrate_mode mode) 894784b4e29SChris Mason { 895784b4e29SChris Mason /* 896784b4e29SChris Mason * we can't safely write a btree page from here, 897784b4e29SChris Mason * we haven't done the locking hook 898784b4e29SChris Mason */ 899784b4e29SChris Mason if (PageDirty(page)) 900784b4e29SChris Mason return -EAGAIN; 901784b4e29SChris Mason /* 902784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 903784b4e29SChris Mason * We must have no buffers or drop them. 904784b4e29SChris Mason */ 905784b4e29SChris Mason if (page_has_private(page) && 906784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 907784b4e29SChris Mason return -EAGAIN; 908a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 909784b4e29SChris Mason } 9103dd1462eSJan Beulich #endif 911784b4e29SChris Mason 9120da5468fSChris Mason 9130da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9140da5468fSChris Mason struct writeback_control *wbc) 9150da5468fSChris Mason { 916e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 917e2d84521SMiao Xie int ret; 918e2d84521SMiao Xie 919d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 920448d640bSChris Mason 921448d640bSChris Mason if (wbc->for_kupdate) 922448d640bSChris Mason return 0; 923448d640bSChris Mason 924e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 925b9473439SChris Mason /* this is a bit racy, but that's ok */ 926d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 927d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 928d814a491SEthan Lien fs_info->dirty_metadata_batch); 929e2d84521SMiao Xie if (ret < 0) 930793955bcSChris Mason return 0; 931793955bcSChris Mason } 9320b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9330da5468fSChris Mason } 9340da5468fSChris Mason 935b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9365f39d397SChris Mason { 937d1310b2eSChris Mason struct extent_io_tree *tree; 938d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9398ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 9405f39d397SChris Mason } 9415f39d397SChris Mason 94270dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9435f39d397SChris Mason { 94498509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 94598509cfcSChris Mason return 0; 9460c4e538bSDavid Sterba 947f7a52a40SDavid Sterba return try_release_extent_buffer(page); 948d98237b3SChris Mason } 949d98237b3SChris Mason 950d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 951d47992f8SLukas Czerner unsigned int length) 952d98237b3SChris Mason { 953d1310b2eSChris Mason struct extent_io_tree *tree; 954d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9555f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9565f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9579ad6b7bcSChris Mason if (PagePrivate(page)) { 958efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 959efe120a0SFrank Holton "page private not zero on page %llu", 960efe120a0SFrank Holton (unsigned long long)page_offset(page)); 9619ad6b7bcSChris Mason ClearPagePrivate(page); 9629ad6b7bcSChris Mason set_page_private(page, 0); 96309cbfeafSKirill A. Shutemov put_page(page); 9649ad6b7bcSChris Mason } 965d98237b3SChris Mason } 966d98237b3SChris Mason 9670b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9680b32f4bbSJosef Bacik { 969bb146eb2SJosef Bacik #ifdef DEBUG 9700b32f4bbSJosef Bacik struct extent_buffer *eb; 9710b32f4bbSJosef Bacik 9720b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9730b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9740b32f4bbSJosef Bacik BUG_ON(!eb); 9750b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9760b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 9770b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 978bb146eb2SJosef Bacik #endif 9790b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 9800b32f4bbSJosef Bacik } 9810b32f4bbSJosef Bacik 9827f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 983d98237b3SChris Mason .readpage = btree_readpage, 9840da5468fSChris Mason .writepages = btree_writepages, 9855f39d397SChris Mason .releasepage = btree_releasepage, 9865f39d397SChris Mason .invalidatepage = btree_invalidatepage, 9875a92bc88SChris Mason #ifdef CONFIG_MIGRATION 988784b4e29SChris Mason .migratepage = btree_migratepage, 9895a92bc88SChris Mason #endif 9900b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 991d98237b3SChris Mason }; 992123abc88SChris Mason 9932ff7e61eSJeff Mahoney void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) 994090d1875SChris Mason { 9955f39d397SChris Mason struct extent_buffer *buf = NULL; 9962ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 997090d1875SChris Mason 9982ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 999c871b0f2SLiu Bo if (IS_ERR(buf)) 10006197d86eSDavid Sterba return; 1001d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 10026af49dbdSDavid Sterba buf, WAIT_NONE, 0); 10035f39d397SChris Mason free_extent_buffer(buf); 1004090d1875SChris Mason } 1005090d1875SChris Mason 10062ff7e61eSJeff Mahoney int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, 1007ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1008ab0fff03SArne Jansen { 1009ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 10102ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1011ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1012ab0fff03SArne Jansen int ret; 1013ab0fff03SArne Jansen 10142ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1015c871b0f2SLiu Bo if (IS_ERR(buf)) 1016ab0fff03SArne Jansen return 0; 1017ab0fff03SArne Jansen 1018ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1019ab0fff03SArne Jansen 10208436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, 10216af49dbdSDavid Sterba mirror_num); 1022ab0fff03SArne Jansen if (ret) { 1023ab0fff03SArne Jansen free_extent_buffer(buf); 1024ab0fff03SArne Jansen return ret; 1025ab0fff03SArne Jansen } 1026ab0fff03SArne Jansen 1027ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1028ab0fff03SArne Jansen free_extent_buffer(buf); 1029ab0fff03SArne Jansen return -EIO; 10300b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1031ab0fff03SArne Jansen *eb = buf; 1032ab0fff03SArne Jansen } else { 1033ab0fff03SArne Jansen free_extent_buffer(buf); 1034ab0fff03SArne Jansen } 1035ab0fff03SArne Jansen return 0; 1036ab0fff03SArne Jansen } 1037ab0fff03SArne Jansen 10382ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10392ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10400308af44SDavid Sterba u64 bytenr) 10410999df54SChris Mason { 10420b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10430b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10440b246afaSJeff Mahoney return alloc_extent_buffer(fs_info, bytenr); 10450999df54SChris Mason } 10460999df54SChris Mason 10470999df54SChris Mason 1048e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1049e02119d5SChris Mason { 1050727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 10518aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1052e02119d5SChris Mason } 1053e02119d5SChris Mason 10543189ff77SJeff Layton void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1055e02119d5SChris Mason { 10563189ff77SJeff Layton filemap_fdatawait_range(buf->pages[0]->mapping, 10578aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1058e02119d5SChris Mason } 1059e02119d5SChris Mason 1060581c1760SQu Wenruo /* 1061581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1062581c1760SQu Wenruo * verification. 1063581c1760SQu Wenruo * 1064581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 1065581c1760SQu Wenruo * @level: expected level, mandatory check 1066581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 1067581c1760SQu Wenruo */ 10682ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 1069581c1760SQu Wenruo u64 parent_transid, int level, 1070581c1760SQu Wenruo struct btrfs_key *first_key) 1071e20d96d6SChris Mason { 10725f39d397SChris Mason struct extent_buffer *buf = NULL; 107319c00ddcSChris Mason int ret; 107419c00ddcSChris Mason 10752ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1076c871b0f2SLiu Bo if (IS_ERR(buf)) 1077c871b0f2SLiu Bo return buf; 1078e4204dedSChris Mason 1079581c1760SQu Wenruo ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid, 1080581c1760SQu Wenruo level, first_key); 10810f0fe8f7SFilipe David Borba Manana if (ret) { 10820f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 108364c043deSLiu Bo return ERR_PTR(ret); 10840f0fe8f7SFilipe David Borba Manana } 10855f39d397SChris Mason return buf; 1086ce9adaa5SChris Mason 1087eb60ceacSChris Mason } 1088eb60ceacSChris Mason 10897c302b49SDavid Sterba void clean_tree_block(struct btrfs_fs_info *fs_info, 10905f39d397SChris Mason struct extent_buffer *buf) 1091ed2ff2cbSChris Mason { 109255c69072SChris Mason if (btrfs_header_generation(buf) == 1093e2d84521SMiao Xie fs_info->running_transaction->transid) { 1094b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1095b4ce94deSChris Mason 1096b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1097104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1098e2d84521SMiao Xie -buf->len, 1099e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1100b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1101b9473439SChris Mason btrfs_set_lock_blocking(buf); 11020b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1103925baeddSChris Mason } 11045f39d397SChris Mason } 1105ed7b63ebSJosef Bacik } 11065f39d397SChris Mason 11078257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11088257b2dcSMiao Xie { 11098257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11108257b2dcSMiao Xie int ret; 11118257b2dcSMiao Xie 11128257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11138257b2dcSMiao Xie if (!writers) 11148257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11158257b2dcSMiao Xie 11168a5a916dSJeff Mahoney ret = percpu_counter_init(&writers->counter, 0, GFP_NOFS); 11178257b2dcSMiao Xie if (ret < 0) { 11188257b2dcSMiao Xie kfree(writers); 11198257b2dcSMiao Xie return ERR_PTR(ret); 11208257b2dcSMiao Xie } 11218257b2dcSMiao Xie 11228257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11238257b2dcSMiao Xie return writers; 11248257b2dcSMiao Xie } 11258257b2dcSMiao Xie 11268257b2dcSMiao Xie static void 11278257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11288257b2dcSMiao Xie { 11298257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 11308257b2dcSMiao Xie kfree(writers); 11318257b2dcSMiao Xie } 11328257b2dcSMiao Xie 1133da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1134e20d96d6SChris Mason u64 objectid) 1135d97e63b6SChris Mason { 11367c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 1137cfaa7295SChris Mason root->node = NULL; 1138a28ec197SChris Mason root->commit_root = NULL; 113927cdeb70SMiao Xie root->state = 0; 1140d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11410b86a832SChris Mason 11420f7d52f4SChris Mason root->last_trans = 0; 114313a8a7c8SYan, Zheng root->highest_objectid = 0; 1144eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1145199c2a9cSMiao Xie root->nr_ordered_extents = 0; 11466bef4d31SEric Paris root->inode_tree = RB_ROOT; 114716cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1148f0486c68SYan, Zheng root->block_rsv = NULL; 11490b86a832SChris Mason 11500b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11515d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1152eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1153eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1154199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1155199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 11562ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11572ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 11585d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1159eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1160199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1161f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11622ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11632ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 11648287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1165a2135011SChris Mason mutex_init(&root->objectid_mutex); 1166e02119d5SChris Mason mutex_init(&root->log_mutex); 116731f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1168573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 11697237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11707237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11717237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11728b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11738b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11747237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11757237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11767237f183SYan Zheng atomic_set(&root->log_writers, 0); 11772ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11780700cea7SElena Reshetova refcount_set(&root->refs, 1); 1179ea14b57fSDavid Sterba atomic_set(&root->will_be_snapshotted, 0); 11808ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1181eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 11827237f183SYan Zheng root->log_transid = 0; 1183d1433debSMiao Xie root->log_transid_committed = -1; 1184257c62e1SChris Mason root->last_log_commit = 0; 11857c0260eeSJeff Mahoney if (!dummy) 1186c6100a4bSJosef Bacik extent_io_tree_init(&root->dirty_log_pages, NULL); 1187017e5369SChris Mason 11883768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 11893768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 11906702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 11917c0260eeSJeff Mahoney if (!dummy) 11923f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 119306ea65a3SJosef Bacik else 119406ea65a3SJosef Bacik root->defrag_trans_start = 0; 11954d775673SChris Mason root->root_key.objectid = objectid; 11960ee5dc67SAl Viro root->anon_dev = 0; 11978ea05e3aSAlexander Block 11985f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 11993768f368SChris Mason } 12003768f368SChris Mason 120174e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 120274e4d827SDavid Sterba gfp_t flags) 12036f07e42eSAl Viro { 120474e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12056f07e42eSAl Viro if (root) 12066f07e42eSAl Viro root->fs_info = fs_info; 12076f07e42eSAl Viro return root; 12086f07e42eSAl Viro } 12096f07e42eSAl Viro 121006ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 121106ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1212da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 121306ea65a3SJosef Bacik { 121406ea65a3SJosef Bacik struct btrfs_root *root; 121506ea65a3SJosef Bacik 12167c0260eeSJeff Mahoney if (!fs_info) 12177c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12187c0260eeSJeff Mahoney 12197c0260eeSJeff Mahoney root = btrfs_alloc_root(fs_info, GFP_KERNEL); 122006ea65a3SJosef Bacik if (!root) 122106ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1222da17066cSJeff Mahoney 1223b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1224da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 1225faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 122606ea65a3SJosef Bacik 122706ea65a3SJosef Bacik return root; 122806ea65a3SJosef Bacik } 122906ea65a3SJosef Bacik #endif 123006ea65a3SJosef Bacik 123120897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 123220897f5cSArne Jansen struct btrfs_fs_info *fs_info, 123320897f5cSArne Jansen u64 objectid) 123420897f5cSArne Jansen { 123520897f5cSArne Jansen struct extent_buffer *leaf; 123620897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 123720897f5cSArne Jansen struct btrfs_root *root; 123820897f5cSArne Jansen struct btrfs_key key; 123920897f5cSArne Jansen int ret = 0; 124033d85fdaSQu Wenruo uuid_le uuid = NULL_UUID_LE; 124120897f5cSArne Jansen 124274e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_KERNEL); 124320897f5cSArne Jansen if (!root) 124420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 124520897f5cSArne Jansen 1246da17066cSJeff Mahoney __setup_root(root, fs_info, objectid); 124720897f5cSArne Jansen root->root_key.objectid = objectid; 124820897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 124920897f5cSArne Jansen root->root_key.offset = 0; 125020897f5cSArne Jansen 12514d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 125220897f5cSArne Jansen if (IS_ERR(leaf)) { 125320897f5cSArne Jansen ret = PTR_ERR(leaf); 12541dd05682STsutomu Itoh leaf = NULL; 125520897f5cSArne Jansen goto fail; 125620897f5cSArne Jansen } 125720897f5cSArne Jansen 125820897f5cSArne Jansen root->node = leaf; 125920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 126020897f5cSArne Jansen 126120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 126227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 126320897f5cSArne Jansen 126420897f5cSArne Jansen root->root_item.flags = 0; 126520897f5cSArne Jansen root->root_item.byte_limit = 0; 126620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 126720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 126820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 126920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 127020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 127120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 127220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 127333d85fdaSQu Wenruo if (is_fstree(objectid)) 12746463fe58SStefan Behrens uuid_le_gen(&uuid); 12756463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 127620897f5cSArne Jansen root->root_item.drop_level = 0; 127720897f5cSArne Jansen 127820897f5cSArne Jansen key.objectid = objectid; 127920897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 128020897f5cSArne Jansen key.offset = 0; 128120897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 128220897f5cSArne Jansen if (ret) 128320897f5cSArne Jansen goto fail; 128420897f5cSArne Jansen 128520897f5cSArne Jansen btrfs_tree_unlock(leaf); 128620897f5cSArne Jansen 128720897f5cSArne Jansen return root; 12881dd05682STsutomu Itoh 12891dd05682STsutomu Itoh fail: 12901dd05682STsutomu Itoh if (leaf) { 12911dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 129259885b39STsutomu Itoh free_extent_buffer(root->commit_root); 12931dd05682STsutomu Itoh free_extent_buffer(leaf); 12941dd05682STsutomu Itoh } 12951dd05682STsutomu Itoh kfree(root); 12961dd05682STsutomu Itoh 12971dd05682STsutomu Itoh return ERR_PTR(ret); 129820897f5cSArne Jansen } 129920897f5cSArne Jansen 13007237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1301e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13020f7d52f4SChris Mason { 13030f7d52f4SChris Mason struct btrfs_root *root; 13047237f183SYan Zheng struct extent_buffer *leaf; 1305e02119d5SChris Mason 130674e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1307e02119d5SChris Mason if (!root) 13087237f183SYan Zheng return ERR_PTR(-ENOMEM); 1309e02119d5SChris Mason 1310da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1311e02119d5SChris Mason 1312e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1313e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1314e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 131527cdeb70SMiao Xie 13167237f183SYan Zheng /* 131727cdeb70SMiao Xie * DON'T set REF_COWS for log trees 131827cdeb70SMiao Xie * 13197237f183SYan Zheng * log trees do not get reference counted because they go away 13207237f183SYan Zheng * before a real commit is actually done. They do store pointers 13217237f183SYan Zheng * to file data extents, and those reference counts still get 13227237f183SYan Zheng * updated (along with back refs to the log tree). 13237237f183SYan Zheng */ 1324e02119d5SChris Mason 13254d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13264d75f8a9SDavid Sterba NULL, 0, 0, 0); 13277237f183SYan Zheng if (IS_ERR(leaf)) { 13287237f183SYan Zheng kfree(root); 13297237f183SYan Zheng return ERR_CAST(leaf); 13307237f183SYan Zheng } 1331e02119d5SChris Mason 13327237f183SYan Zheng root->node = leaf; 1333e02119d5SChris Mason 1334e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1335e02119d5SChris Mason btrfs_tree_unlock(root->node); 13367237f183SYan Zheng return root; 13377237f183SYan Zheng } 13387237f183SYan Zheng 13397237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13407237f183SYan Zheng struct btrfs_fs_info *fs_info) 13417237f183SYan Zheng { 13427237f183SYan Zheng struct btrfs_root *log_root; 13437237f183SYan Zheng 13447237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13457237f183SYan Zheng if (IS_ERR(log_root)) 13467237f183SYan Zheng return PTR_ERR(log_root); 13477237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13487237f183SYan Zheng fs_info->log_root_tree = log_root; 13497237f183SYan Zheng return 0; 13507237f183SYan Zheng } 13517237f183SYan Zheng 13527237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13537237f183SYan Zheng struct btrfs_root *root) 13547237f183SYan Zheng { 13550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13567237f183SYan Zheng struct btrfs_root *log_root; 13577237f183SYan Zheng struct btrfs_inode_item *inode_item; 13587237f183SYan Zheng 13590b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13607237f183SYan Zheng if (IS_ERR(log_root)) 13617237f183SYan Zheng return PTR_ERR(log_root); 13627237f183SYan Zheng 13637237f183SYan Zheng log_root->last_trans = trans->transid; 13647237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13657237f183SYan Zheng 13667237f183SYan Zheng inode_item = &log_root->root_item.inode; 13673cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13683cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13693cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1370da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13710b246afaSJeff Mahoney fs_info->nodesize); 13723cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13737237f183SYan Zheng 13745d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13757237f183SYan Zheng 13767237f183SYan Zheng WARN_ON(root->log_root); 13777237f183SYan Zheng root->log_root = log_root; 13787237f183SYan Zheng root->log_transid = 0; 1379d1433debSMiao Xie root->log_transid_committed = -1; 1380257c62e1SChris Mason root->last_log_commit = 0; 1381e02119d5SChris Mason return 0; 1382e02119d5SChris Mason } 1383e02119d5SChris Mason 138435a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1385cb517eabSMiao Xie struct btrfs_key *key) 1386e02119d5SChris Mason { 1387e02119d5SChris Mason struct btrfs_root *root; 1388e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 13890f7d52f4SChris Mason struct btrfs_path *path; 139084234f3aSYan Zheng u64 generation; 1391cb517eabSMiao Xie int ret; 1392581c1760SQu Wenruo int level; 1393cb517eabSMiao Xie 1394cb517eabSMiao Xie path = btrfs_alloc_path(); 1395cb517eabSMiao Xie if (!path) 1396cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 13970f7d52f4SChris Mason 139874e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1399cb517eabSMiao Xie if (!root) { 1400cb517eabSMiao Xie ret = -ENOMEM; 1401cb517eabSMiao Xie goto alloc_fail; 14020f7d52f4SChris Mason } 14030f7d52f4SChris Mason 1404da17066cSJeff Mahoney __setup_root(root, fs_info, key->objectid); 14050f7d52f4SChris Mason 1406cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1407cb517eabSMiao Xie &root->root_item, &root->root_key); 14080f7d52f4SChris Mason if (ret) { 140913a8a7c8SYan, Zheng if (ret > 0) 141013a8a7c8SYan, Zheng ret = -ENOENT; 1411cb517eabSMiao Xie goto find_fail; 14120f7d52f4SChris Mason } 141313a8a7c8SYan, Zheng 141484234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1415581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 14162ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 14172ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 1418581c1760SQu Wenruo generation, level, NULL); 141964c043deSLiu Bo if (IS_ERR(root->node)) { 142064c043deSLiu Bo ret = PTR_ERR(root->node); 1421cb517eabSMiao Xie goto find_fail; 1422cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1423cb517eabSMiao Xie ret = -EIO; 142464c043deSLiu Bo free_extent_buffer(root->node); 142564c043deSLiu Bo goto find_fail; 1426416bc658SJosef Bacik } 14275d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 142813a8a7c8SYan, Zheng out: 1429cb517eabSMiao Xie btrfs_free_path(path); 1430cb517eabSMiao Xie return root; 1431cb517eabSMiao Xie 1432cb517eabSMiao Xie find_fail: 1433cb517eabSMiao Xie kfree(root); 1434cb517eabSMiao Xie alloc_fail: 1435cb517eabSMiao Xie root = ERR_PTR(ret); 1436cb517eabSMiao Xie goto out; 1437cb517eabSMiao Xie } 1438cb517eabSMiao Xie 1439cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1440cb517eabSMiao Xie struct btrfs_key *location) 1441cb517eabSMiao Xie { 1442cb517eabSMiao Xie struct btrfs_root *root; 1443cb517eabSMiao Xie 1444cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1445cb517eabSMiao Xie if (IS_ERR(root)) 1446cb517eabSMiao Xie return root; 1447cb517eabSMiao Xie 1448cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 144927cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 145008fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 145108fe4db1SLi Zefan } 145213a8a7c8SYan, Zheng 14535eda7b5eSChris Mason return root; 14545eda7b5eSChris Mason } 14555eda7b5eSChris Mason 1456cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1457cb517eabSMiao Xie { 1458cb517eabSMiao Xie int ret; 14598257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1460cb517eabSMiao Xie 1461cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1462cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1463cb517eabSMiao Xie GFP_NOFS); 1464cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1465cb517eabSMiao Xie ret = -ENOMEM; 1466cb517eabSMiao Xie goto fail; 1467cb517eabSMiao Xie } 1468cb517eabSMiao Xie 14698257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 14708257b2dcSMiao Xie if (IS_ERR(writers)) { 14718257b2dcSMiao Xie ret = PTR_ERR(writers); 14728257b2dcSMiao Xie goto fail; 14738257b2dcSMiao Xie } 14748257b2dcSMiao Xie root->subv_writers = writers; 14758257b2dcSMiao Xie 1476cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 147757cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 147857cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1479cb517eabSMiao Xie 1480cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1481cb517eabSMiao Xie if (ret) 1482876d2cf1SLiu Bo goto fail; 1483f32e48e9SChandan Rajendra 1484f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1485f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1486f32e48e9SChandan Rajendra &root->highest_objectid); 1487f32e48e9SChandan Rajendra if (ret) { 1488f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1489876d2cf1SLiu Bo goto fail; 1490f32e48e9SChandan Rajendra } 1491f32e48e9SChandan Rajendra 1492f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1493f32e48e9SChandan Rajendra 1494f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1495f32e48e9SChandan Rajendra 1496cb517eabSMiao Xie return 0; 1497cb517eabSMiao Xie fail: 149884db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1499cb517eabSMiao Xie return ret; 1500cb517eabSMiao Xie } 1501cb517eabSMiao Xie 150235bbb97fSJeff Mahoney struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1503cb517eabSMiao Xie u64 root_id) 1504cb517eabSMiao Xie { 1505cb517eabSMiao Xie struct btrfs_root *root; 1506cb517eabSMiao Xie 1507cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1508cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1509cb517eabSMiao Xie (unsigned long)root_id); 1510cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1511cb517eabSMiao Xie return root; 1512cb517eabSMiao Xie } 1513cb517eabSMiao Xie 1514cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1515cb517eabSMiao Xie struct btrfs_root *root) 1516cb517eabSMiao Xie { 1517cb517eabSMiao Xie int ret; 1518cb517eabSMiao Xie 1519e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1520cb517eabSMiao Xie if (ret) 1521cb517eabSMiao Xie return ret; 1522cb517eabSMiao Xie 1523cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1524cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1525cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1526cb517eabSMiao Xie root); 1527cb517eabSMiao Xie if (ret == 0) 152827cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1529cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1530cb517eabSMiao Xie radix_tree_preload_end(); 1531cb517eabSMiao Xie 1532cb517eabSMiao Xie return ret; 1533cb517eabSMiao Xie } 1534cb517eabSMiao Xie 1535c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1536c00869f1SMiao Xie struct btrfs_key *location, 1537c00869f1SMiao Xie bool check_ref) 15385eda7b5eSChris Mason { 15395eda7b5eSChris Mason struct btrfs_root *root; 1540381cf658SDavid Sterba struct btrfs_path *path; 15411d4c08e0SDavid Sterba struct btrfs_key key; 15425eda7b5eSChris Mason int ret; 15435eda7b5eSChris Mason 1544edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1545edbd8d4eSChris Mason return fs_info->tree_root; 1546edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1547edbd8d4eSChris Mason return fs_info->extent_root; 15488f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 15498f18cf13SChris Mason return fs_info->chunk_root; 15508f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 15518f18cf13SChris Mason return fs_info->dev_root; 15520403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 15530403e47eSYan Zheng return fs_info->csum_root; 1554bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1555bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1556bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1557f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1558f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1559f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 156070f6d82eSOmar Sandoval if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 156170f6d82eSOmar Sandoval return fs_info->free_space_root ? fs_info->free_space_root : 156270f6d82eSOmar Sandoval ERR_PTR(-ENOENT); 15634df27c4dSYan, Zheng again: 1564cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 156548475471SStefan Behrens if (root) { 1566c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 156748475471SStefan Behrens return ERR_PTR(-ENOENT); 15685eda7b5eSChris Mason return root; 156948475471SStefan Behrens } 15705eda7b5eSChris Mason 1571cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 15725eda7b5eSChris Mason if (IS_ERR(root)) 15735eda7b5eSChris Mason return root; 15743394e160SChris Mason 1575c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1576d68fc57bSYan, Zheng ret = -ENOENT; 1577d68fc57bSYan, Zheng goto fail; 1578d68fc57bSYan, Zheng } 1579d68fc57bSYan, Zheng 1580cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1581cb517eabSMiao Xie if (ret) 1582cb517eabSMiao Xie goto fail; 1583cb517eabSMiao Xie 1584381cf658SDavid Sterba path = btrfs_alloc_path(); 1585381cf658SDavid Sterba if (!path) { 1586381cf658SDavid Sterba ret = -ENOMEM; 1587381cf658SDavid Sterba goto fail; 1588381cf658SDavid Sterba } 15891d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 15901d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 15911d4c08e0SDavid Sterba key.offset = location->objectid; 15921d4c08e0SDavid Sterba 15931d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1594381cf658SDavid Sterba btrfs_free_path(path); 1595d68fc57bSYan, Zheng if (ret < 0) 1596d68fc57bSYan, Zheng goto fail; 1597d68fc57bSYan, Zheng if (ret == 0) 159827cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1599d68fc57bSYan, Zheng 1600cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16010f7d52f4SChris Mason if (ret) { 16024df27c4dSYan, Zheng if (ret == -EEXIST) { 160384db5ccfSDavid Sterba btrfs_free_fs_root(root); 16044df27c4dSYan, Zheng goto again; 16050f7d52f4SChris Mason } 16064df27c4dSYan, Zheng goto fail; 16074df27c4dSYan, Zheng } 1608edbd8d4eSChris Mason return root; 16094df27c4dSYan, Zheng fail: 161084db5ccfSDavid Sterba btrfs_free_fs_root(root); 16114df27c4dSYan, Zheng return ERR_PTR(ret); 1612edbd8d4eSChris Mason } 1613edbd8d4eSChris Mason 161404160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 161504160088SChris Mason { 161604160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 161704160088SChris Mason int ret = 0; 161804160088SChris Mason struct btrfs_device *device; 161904160088SChris Mason struct backing_dev_info *bdi; 1620b7967db7SChris Mason 16211f78160cSXiao Guangrong rcu_read_lock(); 16221f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1623dfe25020SChris Mason if (!device->bdev) 1624dfe25020SChris Mason continue; 1625efa7c9f9SJan Kara bdi = device->bdev->bd_bdi; 1626ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 162704160088SChris Mason ret = 1; 162804160088SChris Mason break; 162904160088SChris Mason } 163004160088SChris Mason } 16311f78160cSXiao Guangrong rcu_read_unlock(); 163204160088SChris Mason return ret; 163304160088SChris Mason } 163404160088SChris Mason 16358b712842SChris Mason /* 16368b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16378b712842SChris Mason * functions. This is where read checksum verification actually happens 16388b712842SChris Mason */ 16398b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1640ce9adaa5SChris Mason { 1641ce9adaa5SChris Mason struct bio *bio; 164297eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1643ce9adaa5SChris Mason 164497eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1645ce9adaa5SChris Mason bio = end_io_wq->bio; 16468b712842SChris Mason 16474e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1648ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1649ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 165097eb6b69SDavid Sterba kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 16514246a0b6SChristoph Hellwig bio_endio(bio); 1652ce9adaa5SChris Mason } 165344b8bd7eSChris Mason 1654a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1655a74a4b97SChris Mason { 1656a74a4b97SChris Mason struct btrfs_root *root = arg; 16570b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1658d0278245SMiao Xie int again; 1659a74a4b97SChris Mason 1660d6fd0ae2SOmar Sandoval while (1) { 1661d0278245SMiao Xie again = 0; 16629d1a2a3aSDavid Sterba 1663d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 16642ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1665d0278245SMiao Xie goto sleep; 1666d0278245SMiao Xie 166790c711abSZygo Blaxell /* 166890c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 166990c711abSZygo Blaxell * before we have finished mounting the filesystem. 167090c711abSZygo Blaxell */ 16710b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 167290c711abSZygo Blaxell goto sleep; 167390c711abSZygo Blaxell 16740b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1675d0278245SMiao Xie goto sleep; 1676d0278245SMiao Xie 1677dc7f370cSMiao Xie /* 1678dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1679dc7f370cSMiao Xie * during the above check and trylock. 1680dc7f370cSMiao Xie */ 16812ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 16820b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1683dc7f370cSMiao Xie goto sleep; 1684dc7f370cSMiao Xie } 1685dc7f370cSMiao Xie 16860b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_delayed_iput_mutex); 16872ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 16880b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); 1689c2d6cb16SFilipe Manana 16909d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 16910b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1692a74a4b97SChris Mason 1693d0278245SMiao Xie /* 169405323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 169505323cd1SMiao Xie * needn't do anything special here. 1696d0278245SMiao Xie */ 16970b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 169867c5e7d4SFilipe Manana 169967c5e7d4SFilipe Manana /* 170067c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 170167c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 170267c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 170367c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 170467c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 170567c5e7d4SFilipe Manana * unused block groups. 170667c5e7d4SFilipe Manana */ 17070b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1708d0278245SMiao Xie sleep: 1709d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1710d6fd0ae2SOmar Sandoval kthread_parkme(); 1711d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1712d6fd0ae2SOmar Sandoval return 0; 1713838fe188SJiri Kosina if (!again) { 1714a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1715a74a4b97SChris Mason schedule(); 1716a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1717a74a4b97SChris Mason } 1718da288d28SFilipe Manana } 1719a74a4b97SChris Mason } 1720a74a4b97SChris Mason 1721a74a4b97SChris Mason static int transaction_kthread(void *arg) 1722a74a4b97SChris Mason { 1723a74a4b97SChris Mason struct btrfs_root *root = arg; 17240b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1725a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1726a74a4b97SChris Mason struct btrfs_transaction *cur; 17278929ecfaSYan, Zheng u64 transid; 1728a944442cSAllen Pais time64_t now; 1729a74a4b97SChris Mason unsigned long delay; 1730914b2007SJan Kara bool cannot_commit; 1731a74a4b97SChris Mason 1732a74a4b97SChris Mason do { 1733914b2007SJan Kara cannot_commit = false; 17340b246afaSJeff Mahoney delay = HZ * fs_info->commit_interval; 17350b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1736a74a4b97SChris Mason 17370b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 17380b246afaSJeff Mahoney cur = fs_info->running_transaction; 1739a74a4b97SChris Mason if (!cur) { 17400b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1741a74a4b97SChris Mason goto sleep; 1742a74a4b97SChris Mason } 174331153d81SYan Zheng 1744afd48513SArnd Bergmann now = ktime_get_seconds(); 17454a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 1746a514d638SQu Wenruo !test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) && 17478b87dc17SDavid Sterba (now < cur->start_time || 17480b246afaSJeff Mahoney now - cur->start_time < fs_info->commit_interval)) { 17490b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1750a74a4b97SChris Mason delay = HZ * 5; 1751a74a4b97SChris Mason goto sleep; 1752a74a4b97SChris Mason } 17538929ecfaSYan, Zheng transid = cur->transid; 17540b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 175556bec294SChris Mason 175679787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1757354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1758914b2007SJan Kara if (IS_ERR(trans)) { 1759354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1760914b2007SJan Kara cannot_commit = true; 176179787eaaSJeff Mahoney goto sleep; 1762914b2007SJan Kara } 17638929ecfaSYan, Zheng if (transid == trans->transid) { 17643a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 17658929ecfaSYan, Zheng } else { 17663a45bb20SJeff Mahoney btrfs_end_transaction(trans); 17678929ecfaSYan, Zheng } 1768a74a4b97SChris Mason sleep: 17690b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 17700b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1771a74a4b97SChris Mason 17724e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 17730b246afaSJeff Mahoney &fs_info->fs_state))) 17742ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 17758929ecfaSYan, Zheng if (!kthread_should_stop() && 17760b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1777914b2007SJan Kara cannot_commit)) 1778bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1779a74a4b97SChris Mason } while (!kthread_should_stop()); 1780a74a4b97SChris Mason return 0; 1781a74a4b97SChris Mason } 1782a74a4b97SChris Mason 1783af31f5e5SChris Mason /* 1784af31f5e5SChris Mason * this will find the highest generation in the array of 1785af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1786af31f5e5SChris Mason * or -1 if we can't find anything. 1787af31f5e5SChris Mason * 1788af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1789af31f5e5SChris Mason * generation of the latest root in the array with the generation 1790af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1791af31f5e5SChris Mason */ 1792af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1793af31f5e5SChris Mason { 1794af31f5e5SChris Mason u64 cur; 1795af31f5e5SChris Mason int newest_index = -1; 1796af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1797af31f5e5SChris Mason int i; 1798af31f5e5SChris Mason 1799af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1800af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1801af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1802af31f5e5SChris Mason if (cur == newest_gen) 1803af31f5e5SChris Mason newest_index = i; 1804af31f5e5SChris Mason } 1805af31f5e5SChris Mason 1806af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1807af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1808af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1809af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1810af31f5e5SChris Mason if (cur == newest_gen) 1811af31f5e5SChris Mason newest_index = 0; 1812af31f5e5SChris Mason } 1813af31f5e5SChris Mason return newest_index; 1814af31f5e5SChris Mason } 1815af31f5e5SChris Mason 1816af31f5e5SChris Mason 1817af31f5e5SChris Mason /* 1818af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1819af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1820af31f5e5SChris Mason * field in the fs_info struct 1821af31f5e5SChris Mason */ 1822af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1823af31f5e5SChris Mason u64 newest_gen) 1824af31f5e5SChris Mason { 1825af31f5e5SChris Mason int newest_index = -1; 1826af31f5e5SChris Mason 1827af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1828af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1829af31f5e5SChris Mason if (newest_index == -1) { 1830af31f5e5SChris Mason info->backup_root_index = 0; 1831af31f5e5SChris Mason } else { 1832af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1833af31f5e5SChris Mason } 1834af31f5e5SChris Mason } 1835af31f5e5SChris Mason 1836af31f5e5SChris Mason /* 1837af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1838af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1839af31f5e5SChris Mason * done 1840af31f5e5SChris Mason */ 1841af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1842af31f5e5SChris Mason { 1843af31f5e5SChris Mason int next_backup; 1844af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1845af31f5e5SChris Mason int last_backup; 1846af31f5e5SChris Mason 1847af31f5e5SChris Mason next_backup = info->backup_root_index; 1848af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1849af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1850af31f5e5SChris Mason 1851af31f5e5SChris Mason /* 1852af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1853af31f5e5SChris Mason * this happens only at umount 1854af31f5e5SChris Mason */ 1855af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1856af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1857af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1858af31f5e5SChris Mason next_backup = last_backup; 1859af31f5e5SChris Mason 1860af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1861af31f5e5SChris Mason 1862af31f5e5SChris Mason /* 1863af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1864af31f5e5SChris Mason * regardless of which ones we use today 1865af31f5e5SChris Mason */ 1866af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1867af31f5e5SChris Mason 1868af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1869af31f5e5SChris Mason 1870af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1871af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1872af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1873af31f5e5SChris Mason 1874af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1875af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1876af31f5e5SChris Mason 1877af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1878af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1879af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1880af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1881af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1882af31f5e5SChris Mason 1883af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1884af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1885af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1886af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1887af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1888af31f5e5SChris Mason 18897c7e82a7SChris Mason /* 18907c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 18917c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 18927c7e82a7SChris Mason */ 18937c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 18947c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 18957c7e82a7SChris Mason info->fs_root->node->start); 1896af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1897af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1898af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1899af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19007c7e82a7SChris Mason } 1901af31f5e5SChris Mason 1902af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1903af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1904af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1905af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1906af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1907af31f5e5SChris Mason 1908af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1909af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1910af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1911af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1912af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1913af31f5e5SChris Mason 1914af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1915af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1916af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1917af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1918af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1919af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1920af31f5e5SChris Mason 1921af31f5e5SChris Mason /* 1922af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1923af31f5e5SChris Mason * for the next commit 1924af31f5e5SChris Mason */ 1925af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1926af31f5e5SChris Mason &info->super_for_commit->super_roots, 1927af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1928af31f5e5SChris Mason } 1929af31f5e5SChris Mason 1930af31f5e5SChris Mason /* 1931af31f5e5SChris Mason * this copies info out of the root backup array and back into 1932af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 1933af31f5e5SChris Mason * the array, so you send it the number of backups you've already 1934af31f5e5SChris Mason * tried and the last backup index you used. 1935af31f5e5SChris Mason * 1936af31f5e5SChris Mason * this returns -1 when it has tried all the backups 1937af31f5e5SChris Mason */ 1938af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 1939af31f5e5SChris Mason struct btrfs_super_block *super, 1940af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 1941af31f5e5SChris Mason { 1942af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1943af31f5e5SChris Mason int newest = *backup_index; 1944af31f5e5SChris Mason 1945af31f5e5SChris Mason if (*num_backups_tried == 0) { 1946af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 1947af31f5e5SChris Mason 1948af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 1949af31f5e5SChris Mason if (newest == -1) 1950af31f5e5SChris Mason return -1; 1951af31f5e5SChris Mason 1952af31f5e5SChris Mason *backup_index = newest; 1953af31f5e5SChris Mason *num_backups_tried = 1; 1954af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 1955af31f5e5SChris Mason /* we've tried all the backups, all done */ 1956af31f5e5SChris Mason return -1; 1957af31f5e5SChris Mason } else { 1958af31f5e5SChris Mason /* jump to the next oldest backup */ 1959af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 1960af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1961af31f5e5SChris Mason *backup_index = newest; 1962af31f5e5SChris Mason *num_backups_tried += 1; 1963af31f5e5SChris Mason } 1964af31f5e5SChris Mason root_backup = super->super_roots + newest; 1965af31f5e5SChris Mason 1966af31f5e5SChris Mason btrfs_set_super_generation(super, 1967af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 1968af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1969af31f5e5SChris Mason btrfs_set_super_root_level(super, 1970af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 1971af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1972af31f5e5SChris Mason 1973af31f5e5SChris Mason /* 1974af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 1975af31f5e5SChris Mason * need a fsck 1976af31f5e5SChris Mason */ 1977af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1978af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1979af31f5e5SChris Mason return 0; 1980af31f5e5SChris Mason } 1981af31f5e5SChris Mason 19827abadb64SLiu Bo /* helper to cleanup workers */ 19837abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 19847abadb64SLiu Bo { 1985dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1986afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 19875cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1988fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 1989fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 19908b110e39SMiao Xie btrfs_destroy_workqueue(fs_info->endio_repair_workers); 1991d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 1992fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1993fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 1994a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 19955b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1996e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1997736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 1998a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1999fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2000a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 2001a9b9477dSFilipe Manana /* 2002a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2003a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2004a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2005a9b9477dSFilipe Manana */ 2006a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2007a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 20087abadb64SLiu Bo } 20097abadb64SLiu Bo 20102e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20112e9f5954SRashika { 20122e9f5954SRashika if (root) { 20132e9f5954SRashika free_extent_buffer(root->node); 20142e9f5954SRashika free_extent_buffer(root->commit_root); 20152e9f5954SRashika root->node = NULL; 20162e9f5954SRashika root->commit_root = NULL; 20172e9f5954SRashika } 20182e9f5954SRashika } 20192e9f5954SRashika 2020af31f5e5SChris Mason /* helper to cleanup tree roots */ 2021af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2022af31f5e5SChris Mason { 20232e9f5954SRashika free_root_extent_buffers(info->tree_root); 2024655b09feSJosef Bacik 20252e9f5954SRashika free_root_extent_buffers(info->dev_root); 20262e9f5954SRashika free_root_extent_buffers(info->extent_root); 20272e9f5954SRashika free_root_extent_buffers(info->csum_root); 20282e9f5954SRashika free_root_extent_buffers(info->quota_root); 20292e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20302e9f5954SRashika if (chunk_root) 20312e9f5954SRashika free_root_extent_buffers(info->chunk_root); 203270f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2033af31f5e5SChris Mason } 2034af31f5e5SChris Mason 2035faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2036171f6537SJosef Bacik { 2037171f6537SJosef Bacik int ret; 2038171f6537SJosef Bacik struct btrfs_root *gang[8]; 2039171f6537SJosef Bacik int i; 2040171f6537SJosef Bacik 2041171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2042171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2043171f6537SJosef Bacik struct btrfs_root, root_list); 2044171f6537SJosef Bacik list_del(&gang[0]->root_list); 2045171f6537SJosef Bacik 204627cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2047cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2048171f6537SJosef Bacik } else { 2049171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2050171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2051b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2052171f6537SJosef Bacik } 2053171f6537SJosef Bacik } 2054171f6537SJosef Bacik 2055171f6537SJosef Bacik while (1) { 2056171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2057171f6537SJosef Bacik (void **)gang, 0, 2058171f6537SJosef Bacik ARRAY_SIZE(gang)); 2059171f6537SJosef Bacik if (!ret) 2060171f6537SJosef Bacik break; 2061171f6537SJosef Bacik for (i = 0; i < ret; i++) 2062cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2063171f6537SJosef Bacik } 20641a4319ccSLiu Bo 20651a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 20661a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 20672ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 20681a4319ccSLiu Bo } 2069171f6537SJosef Bacik } 2070af31f5e5SChris Mason 2071638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2072638aa7edSEric Sandeen { 2073638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2074638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2075638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2076638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2077638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2078638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2079638aa7edSEric Sandeen fs_info->scrub_workers_refcnt = 0; 2080638aa7edSEric Sandeen } 2081638aa7edSEric Sandeen 2082779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2083779a65a4SEric Sandeen { 2084779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2085779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2086779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2087779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2088779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2089779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2090779a65a4SEric Sandeen } 2091779a65a4SEric Sandeen 20926bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2093f37938e0SEric Sandeen { 20942ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 20952ff7e61eSJeff Mahoney 20962ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 20972ff7e61eSJeff Mahoney set_nlink(inode, 1); 2098f37938e0SEric Sandeen /* 2099f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2100f37938e0SEric Sandeen * the real end of the address space is determined by all of 2101f37938e0SEric Sandeen * the devices in the system 2102f37938e0SEric Sandeen */ 21032ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 21042ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2105f37938e0SEric Sandeen 21062ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 2107c6100a4bSJosef Bacik extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); 21082ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.track_uptodate = 0; 21092ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2110f37938e0SEric Sandeen 21112ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; 2112f37938e0SEric Sandeen 21132ff7e61eSJeff Mahoney BTRFS_I(inode)->root = fs_info->tree_root; 21142ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21152ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21162ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2117f37938e0SEric Sandeen } 2118f37938e0SEric Sandeen 2119ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2120ad618368SEric Sandeen { 2121ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 212273beece9SLiu Bo rwlock_init(&fs_info->dev_replace.lock); 212373beece9SLiu Bo atomic_set(&fs_info->dev_replace.blocking_readers, 0); 21247f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 212573beece9SLiu Bo init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); 2126ad618368SEric Sandeen } 2127ad618368SEric Sandeen 2128f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2129f9e92e40SEric Sandeen { 2130f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2131f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2132f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2133f9e92e40SEric Sandeen fs_info->qgroup_op_tree = RB_ROOT; 2134f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2135f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2136f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2137d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2138f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2139f9e92e40SEric Sandeen } 2140f9e92e40SEric Sandeen 21412a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 21422a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 21432a458198SEric Sandeen { 2144f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 21456f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21462a458198SEric Sandeen 21472a458198SEric Sandeen fs_info->workers = 2148cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2149cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 21502a458198SEric Sandeen 21512a458198SEric Sandeen fs_info->delalloc_workers = 2152cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2153cb001095SJeff Mahoney flags, max_active, 2); 21542a458198SEric Sandeen 21552a458198SEric Sandeen fs_info->flush_workers = 2156cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2157cb001095SJeff Mahoney flags, max_active, 0); 21582a458198SEric Sandeen 21592a458198SEric Sandeen fs_info->caching_workers = 2160cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 21612a458198SEric Sandeen 21622a458198SEric Sandeen /* 21632a458198SEric Sandeen * a higher idle thresh on the submit workers makes it much more 21642a458198SEric Sandeen * likely that bios will be send down in a sane order to the 21652a458198SEric Sandeen * devices 21662a458198SEric Sandeen */ 21672a458198SEric Sandeen fs_info->submit_workers = 2168cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "submit", flags, 21692a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 21702a458198SEric Sandeen max_active), 64); 21712a458198SEric Sandeen 21722a458198SEric Sandeen fs_info->fixup_workers = 2173cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 21742a458198SEric Sandeen 21752a458198SEric Sandeen /* 21762a458198SEric Sandeen * endios are largely parallel and should have a very 21772a458198SEric Sandeen * low idle thresh 21782a458198SEric Sandeen */ 21792a458198SEric Sandeen fs_info->endio_workers = 2180cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 21812a458198SEric Sandeen fs_info->endio_meta_workers = 2182cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2183cb001095SJeff Mahoney max_active, 4); 21842a458198SEric Sandeen fs_info->endio_meta_write_workers = 2185cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2186cb001095SJeff Mahoney max_active, 2); 21872a458198SEric Sandeen fs_info->endio_raid56_workers = 2188cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2189cb001095SJeff Mahoney max_active, 4); 21902a458198SEric Sandeen fs_info->endio_repair_workers = 2191cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); 21922a458198SEric Sandeen fs_info->rmw_workers = 2193cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 21942a458198SEric Sandeen fs_info->endio_write_workers = 2195cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2196cb001095SJeff Mahoney max_active, 2); 21972a458198SEric Sandeen fs_info->endio_freespace_worker = 2198cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2199cb001095SJeff Mahoney max_active, 0); 22002a458198SEric Sandeen fs_info->delayed_workers = 2201cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2202cb001095SJeff Mahoney max_active, 0); 22032a458198SEric Sandeen fs_info->readahead_workers = 2204cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2205cb001095SJeff Mahoney max_active, 2); 22062a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2207cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 22082a458198SEric Sandeen fs_info->extent_workers = 2209cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "extent-refs", flags, 22102a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22112a458198SEric Sandeen max_active), 8); 22122a458198SEric Sandeen 22132a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 22142a458198SEric Sandeen fs_info->submit_workers && fs_info->flush_workers && 22152a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 22162a458198SEric Sandeen fs_info->endio_meta_write_workers && 22172a458198SEric Sandeen fs_info->endio_repair_workers && 22182a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 22192a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 22202a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 22212a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 22222a458198SEric Sandeen fs_info->extent_workers && 22232a458198SEric Sandeen fs_info->qgroup_rescan_workers)) { 22242a458198SEric Sandeen return -ENOMEM; 22252a458198SEric Sandeen } 22262a458198SEric Sandeen 22272a458198SEric Sandeen return 0; 22282a458198SEric Sandeen } 22292a458198SEric Sandeen 223063443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 223163443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 223263443bf5SEric Sandeen { 223363443bf5SEric Sandeen int ret; 223463443bf5SEric Sandeen struct btrfs_root *log_tree_root; 223563443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 223663443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2237581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 223863443bf5SEric Sandeen 223963443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2240f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 224163443bf5SEric Sandeen return -EIO; 224263443bf5SEric Sandeen } 224363443bf5SEric Sandeen 224474e4d827SDavid Sterba log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 224563443bf5SEric Sandeen if (!log_tree_root) 224663443bf5SEric Sandeen return -ENOMEM; 224763443bf5SEric Sandeen 2248da17066cSJeff Mahoney __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 224963443bf5SEric Sandeen 22502ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 2251581c1760SQu Wenruo fs_info->generation + 1, 2252581c1760SQu Wenruo level, NULL); 225364c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2254f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22550eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 225664c043deSLiu Bo kfree(log_tree_root); 22570eeff236SLiu Bo return ret; 225864c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2259f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 226063443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 226163443bf5SEric Sandeen kfree(log_tree_root); 226263443bf5SEric Sandeen return -EIO; 226363443bf5SEric Sandeen } 226463443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 226563443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 226663443bf5SEric Sandeen if (ret) { 22670b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 226863443bf5SEric Sandeen "Failed to recover log tree"); 226963443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 227063443bf5SEric Sandeen kfree(log_tree_root); 227163443bf5SEric Sandeen return ret; 227263443bf5SEric Sandeen } 227363443bf5SEric Sandeen 2274bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 22756bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 227663443bf5SEric Sandeen if (ret) 227763443bf5SEric Sandeen return ret; 227863443bf5SEric Sandeen } 227963443bf5SEric Sandeen 228063443bf5SEric Sandeen return 0; 228163443bf5SEric Sandeen } 228263443bf5SEric Sandeen 22836bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 22844bbcaa64SEric Sandeen { 22856bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2286a4f3d2c4SDavid Sterba struct btrfs_root *root; 22874bbcaa64SEric Sandeen struct btrfs_key location; 22884bbcaa64SEric Sandeen int ret; 22894bbcaa64SEric Sandeen 22906bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22916bccf3abSJeff Mahoney 22924bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 22934bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22944bbcaa64SEric Sandeen location.offset = 0; 22954bbcaa64SEric Sandeen 2296a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2297f50f4353SLiu Bo if (IS_ERR(root)) { 2298f50f4353SLiu Bo ret = PTR_ERR(root); 2299f50f4353SLiu Bo goto out; 2300f50f4353SLiu Bo } 2301a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2302a4f3d2c4SDavid Sterba fs_info->extent_root = root; 23034bbcaa64SEric Sandeen 23044bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2305a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2306f50f4353SLiu Bo if (IS_ERR(root)) { 2307f50f4353SLiu Bo ret = PTR_ERR(root); 2308f50f4353SLiu Bo goto out; 2309f50f4353SLiu Bo } 2310a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2311a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23124bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 23134bbcaa64SEric Sandeen 23144bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2315a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2316f50f4353SLiu Bo if (IS_ERR(root)) { 2317f50f4353SLiu Bo ret = PTR_ERR(root); 2318f50f4353SLiu Bo goto out; 2319f50f4353SLiu Bo } 2320a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2321a4f3d2c4SDavid Sterba fs_info->csum_root = root; 23224bbcaa64SEric Sandeen 23234bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2324a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2325a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2326a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2327afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2328a4f3d2c4SDavid Sterba fs_info->quota_root = root; 23294bbcaa64SEric Sandeen } 23304bbcaa64SEric Sandeen 23314bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2332a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2333a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2334a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 23354bbcaa64SEric Sandeen if (ret != -ENOENT) 2336f50f4353SLiu Bo goto out; 23374bbcaa64SEric Sandeen } else { 2338a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2339a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 23404bbcaa64SEric Sandeen } 23414bbcaa64SEric Sandeen 234270f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 234370f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 234470f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2345f50f4353SLiu Bo if (IS_ERR(root)) { 2346f50f4353SLiu Bo ret = PTR_ERR(root); 2347f50f4353SLiu Bo goto out; 2348f50f4353SLiu Bo } 234970f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 235070f6d82eSOmar Sandoval fs_info->free_space_root = root; 235170f6d82eSOmar Sandoval } 235270f6d82eSOmar Sandoval 23534bbcaa64SEric Sandeen return 0; 2354f50f4353SLiu Bo out: 2355f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2356f50f4353SLiu Bo location.objectid, ret); 2357f50f4353SLiu Bo return ret; 23584bbcaa64SEric Sandeen } 23594bbcaa64SEric Sandeen 2360069ec957SQu Wenruo /* 2361069ec957SQu Wenruo * Real super block validation 2362069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2363069ec957SQu Wenruo * 2364069ec957SQu Wenruo * @sb: super block to check 2365069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2366069ec957SQu Wenruo * 0 the primary (1st) sb 2367069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2368069ec957SQu Wenruo * -1 skip bytenr check 2369069ec957SQu Wenruo */ 2370069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2371069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 237221a852b0SQu Wenruo { 237321a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 237421a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 237521a852b0SQu Wenruo int ret = 0; 237621a852b0SQu Wenruo 237721a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 237821a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 237921a852b0SQu Wenruo ret = -EINVAL; 238021a852b0SQu Wenruo } 238121a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 238221a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 238321a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 238421a852b0SQu Wenruo ret = -EINVAL; 238521a852b0SQu Wenruo } 238621a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 238721a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 238821a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 238921a852b0SQu Wenruo ret = -EINVAL; 239021a852b0SQu Wenruo } 239121a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 239221a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 239321a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 239421a852b0SQu Wenruo ret = -EINVAL; 239521a852b0SQu Wenruo } 239621a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 239721a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 239821a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 239921a852b0SQu Wenruo ret = -EINVAL; 240021a852b0SQu Wenruo } 240121a852b0SQu Wenruo 240221a852b0SQu Wenruo /* 240321a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 240421a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 240521a852b0SQu Wenruo */ 240621a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 240721a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 240821a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 240921a852b0SQu Wenruo ret = -EINVAL; 241021a852b0SQu Wenruo } 241121a852b0SQu Wenruo /* Only PAGE SIZE is supported yet */ 241221a852b0SQu Wenruo if (sectorsize != PAGE_SIZE) { 241321a852b0SQu Wenruo btrfs_err(fs_info, 241421a852b0SQu Wenruo "sectorsize %llu not supported yet, only support %lu", 241521a852b0SQu Wenruo sectorsize, PAGE_SIZE); 241621a852b0SQu Wenruo ret = -EINVAL; 241721a852b0SQu Wenruo } 241821a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 241921a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 242021a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 242121a852b0SQu Wenruo ret = -EINVAL; 242221a852b0SQu Wenruo } 242321a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 242421a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 242521a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 242621a852b0SQu Wenruo ret = -EINVAL; 242721a852b0SQu Wenruo } 242821a852b0SQu Wenruo 242921a852b0SQu Wenruo /* Root alignment check */ 243021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 243121a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 243221a852b0SQu Wenruo btrfs_super_root(sb)); 243321a852b0SQu Wenruo ret = -EINVAL; 243421a852b0SQu Wenruo } 243521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 243621a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 243721a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 243821a852b0SQu Wenruo ret = -EINVAL; 243921a852b0SQu Wenruo } 244021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 244121a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 244221a852b0SQu Wenruo btrfs_super_log_root(sb)); 244321a852b0SQu Wenruo ret = -EINVAL; 244421a852b0SQu Wenruo } 244521a852b0SQu Wenruo 244621a852b0SQu Wenruo if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { 244721a852b0SQu Wenruo btrfs_err(fs_info, 244821a852b0SQu Wenruo "dev_item UUID does not match fsid: %pU != %pU", 244921a852b0SQu Wenruo fs_info->fsid, sb->dev_item.fsid); 245021a852b0SQu Wenruo ret = -EINVAL; 245121a852b0SQu Wenruo } 245221a852b0SQu Wenruo 245321a852b0SQu Wenruo /* 245421a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 245521a852b0SQu Wenruo * done later 245621a852b0SQu Wenruo */ 245721a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 245821a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 245921a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 246021a852b0SQu Wenruo ret = -EINVAL; 246121a852b0SQu Wenruo } 246221a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 246321a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 246421a852b0SQu Wenruo btrfs_super_stripesize(sb)); 246521a852b0SQu Wenruo ret = -EINVAL; 246621a852b0SQu Wenruo } 246721a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 246821a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 246921a852b0SQu Wenruo btrfs_super_num_devices(sb)); 247021a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 247121a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 247221a852b0SQu Wenruo ret = -EINVAL; 247321a852b0SQu Wenruo } 247421a852b0SQu Wenruo 2475069ec957SQu Wenruo if (mirror_num >= 0 && 2476069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 247721a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 247821a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 247921a852b0SQu Wenruo ret = -EINVAL; 248021a852b0SQu Wenruo } 248121a852b0SQu Wenruo 248221a852b0SQu Wenruo /* 248321a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 248421a852b0SQu Wenruo * and one chunk 248521a852b0SQu Wenruo */ 248621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 248721a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 248821a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 248921a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 249021a852b0SQu Wenruo ret = -EINVAL; 249121a852b0SQu Wenruo } 249221a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 249321a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 249421a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 249521a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 249621a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 249721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 249821a852b0SQu Wenruo ret = -EINVAL; 249921a852b0SQu Wenruo } 250021a852b0SQu Wenruo 250121a852b0SQu Wenruo /* 250221a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 250321a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 250421a852b0SQu Wenruo */ 250521a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 250621a852b0SQu Wenruo btrfs_warn(fs_info, 250721a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 250821a852b0SQu Wenruo btrfs_super_generation(sb), 250921a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 251021a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 251121a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 251221a852b0SQu Wenruo btrfs_warn(fs_info, 251321a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 251421a852b0SQu Wenruo btrfs_super_generation(sb), 251521a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 251621a852b0SQu Wenruo 251721a852b0SQu Wenruo return ret; 251821a852b0SQu Wenruo } 251921a852b0SQu Wenruo 2520069ec957SQu Wenruo /* 2521069ec957SQu Wenruo * Validation of super block at mount time. 2522069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2523069ec957SQu Wenruo * flags will be skipped. 2524069ec957SQu Wenruo */ 2525069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2526069ec957SQu Wenruo { 2527069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2528069ec957SQu Wenruo } 2529069ec957SQu Wenruo 253075cb857dSQu Wenruo /* 253175cb857dSQu Wenruo * Validation of super block at write time. 253275cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 253375cb857dSQu Wenruo * overwritten soon. 253475cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 253575cb857dSQu Wenruo */ 253675cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 253775cb857dSQu Wenruo struct btrfs_super_block *sb) 253875cb857dSQu Wenruo { 253975cb857dSQu Wenruo int ret; 254075cb857dSQu Wenruo 254175cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 254275cb857dSQu Wenruo if (ret < 0) 254375cb857dSQu Wenruo goto out; 254475cb857dSQu Wenruo if (btrfs_super_csum_type(sb) != BTRFS_CSUM_TYPE_CRC32) { 254575cb857dSQu Wenruo ret = -EUCLEAN; 254675cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 254775cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 254875cb857dSQu Wenruo goto out; 254975cb857dSQu Wenruo } 255075cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 255175cb857dSQu Wenruo ret = -EUCLEAN; 255275cb857dSQu Wenruo btrfs_err(fs_info, 255375cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 255475cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 255575cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 255675cb857dSQu Wenruo goto out; 255775cb857dSQu Wenruo } 255875cb857dSQu Wenruo out: 255975cb857dSQu Wenruo if (ret < 0) 256075cb857dSQu Wenruo btrfs_err(fs_info, 256175cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 256275cb857dSQu Wenruo return ret; 256375cb857dSQu Wenruo } 256475cb857dSQu Wenruo 2565ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2566dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2567dfe25020SChris Mason char *options) 2568eb60ceacSChris Mason { 2569db94535dSChris Mason u32 sectorsize; 2570db94535dSChris Mason u32 nodesize; 257187ee04ebSChris Mason u32 stripesize; 257284234f3aSYan Zheng u64 generation; 2573f2b636e8SJosef Bacik u64 features; 25743de4586cSChris Mason struct btrfs_key location; 2575a061fc8dSChris Mason struct buffer_head *bh; 25764d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2577815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2578f84a8bd6SAl Viro struct btrfs_root *tree_root; 25794d34b278SIlya Dryomov struct btrfs_root *chunk_root; 2580eb60ceacSChris Mason int ret; 2581e58ca020SYan int err = -EINVAL; 2582af31f5e5SChris Mason int num_backups_tried = 0; 2583af31f5e5SChris Mason int backup_index = 0; 25846675df31SOmar Sandoval int clear_free_space_tree = 0; 2585581c1760SQu Wenruo int level; 25864543df7eSChris Mason 258774e4d827SDavid Sterba tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 258874e4d827SDavid Sterba chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 2589cb517eabSMiao Xie if (!tree_root || !chunk_root) { 259039279cc3SChris Mason err = -ENOMEM; 259139279cc3SChris Mason goto fail; 259239279cc3SChris Mason } 259376dda93cSYan, Zheng 259476dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 259576dda93cSYan, Zheng if (ret) { 259676dda93cSYan, Zheng err = ret; 259776dda93cSYan, Zheng goto fail; 259876dda93cSYan, Zheng } 259976dda93cSYan, Zheng 2600908c7f19STejun Heo ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2601e2d84521SMiao Xie if (ret) { 2602e2d84521SMiao Xie err = ret; 26039e11ceeeSJan Kara goto fail_srcu; 2604e2d84521SMiao Xie } 260509cbfeafSKirill A. Shutemov fs_info->dirty_metadata_batch = PAGE_SIZE * 2606e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2607e2d84521SMiao Xie 2608908c7f19STejun Heo ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2609963d678bSMiao Xie if (ret) { 2610963d678bSMiao Xie err = ret; 2611963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2612963d678bSMiao Xie } 2613963d678bSMiao Xie 26147f8d236aSDavid Sterba ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 26157f8d236aSDavid Sterba GFP_KERNEL); 2616c404e0dcSMiao Xie if (ret) { 2617c404e0dcSMiao Xie err = ret; 2618c404e0dcSMiao Xie goto fail_delalloc_bytes; 2619c404e0dcSMiao Xie } 2620c404e0dcSMiao Xie 262176dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2622f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26238fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2624facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 262524bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2626eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 262711833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 262875cb379dSJeff Mahoney INIT_LIST_HEAD(&fs_info->pending_raid_kobjs); 262975cb379dSJeff Mahoney spin_lock_init(&fs_info->pending_raid_kobjs_lock); 2630eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2631a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 263276dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 263324bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26344cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2635f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2636ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2637fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2638f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 263947ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2640f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2641d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 264267c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 26437585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2644573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2645c2d6cb16SFilipe Manana mutex_init(&fs_info->cleaner_delayed_iput_mutex); 2646de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 264719c00ddcSChris Mason 26480b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 26496324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2650f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 265147ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 26520b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 265366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 265466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 265566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 265666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 265766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 265866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 265966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2660771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 26614cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2662fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 26632fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2664fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2665e20d96d6SChris Mason fs_info->sb = sb; 266695ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 26679ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 26684cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2669a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2670f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 26718b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2672f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 267390519d66SArne Jansen /* readahead state */ 2674d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 267590519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2676fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2677c8b97818SChris Mason 2678b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2679b34b086cSChris Mason num_online_cpus() + 2, 8); 26800afbaf8cSChris Mason 2681199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2682199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 268369fe2d75SJosef Bacik 268469fe2d75SJosef Bacik fs_info->btree_inode = new_inode(sb); 268569fe2d75SJosef Bacik if (!fs_info->btree_inode) { 268669fe2d75SJosef Bacik err = -ENOMEM; 268769fe2d75SJosef Bacik goto fail_bio_counter; 268869fe2d75SJosef Bacik } 268969fe2d75SJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 269069fe2d75SJosef Bacik 269116cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 269274e4d827SDavid Sterba GFP_KERNEL); 269316cdcec7SMiao Xie if (!fs_info->delayed_root) { 269416cdcec7SMiao Xie err = -ENOMEM; 269516cdcec7SMiao Xie goto fail_iput; 269616cdcec7SMiao Xie } 269716cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 26983eaa2885SChris Mason 2699638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 270021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 270121adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 270221adbd5cSStefan Behrens #endif 2703779a65a4SEric Sandeen btrfs_init_balance(fs_info); 270421c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2705a2de733cSArne Jansen 27069f6d2510SDavid Sterba sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 27079f6d2510SDavid Sterba sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 2708a061fc8dSChris Mason 27096bccf3abSJeff Mahoney btrfs_init_btree_inode(fs_info); 271039279cc3SChris Mason 2711e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 27126bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2713a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2714925baeddSChris Mason 2715c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[0], NULL); 2716c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[1], NULL); 271711833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2718afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 2719509659cdSChris Mason 27205a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 272179154b1bSChris Mason mutex_init(&fs_info->tree_log_mutex); 2722925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2723a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2724a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27251bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27269e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2727c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 272876dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2729803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2730fa9c0d79SChris Mason 2731ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2732f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2733416ac51dSArne Jansen 2734fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2735fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2736fa9c0d79SChris Mason 2737e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2738f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2739bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27404854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 27413768f368SChris Mason 274204216820SFilipe Manana INIT_LIST_HEAD(&fs_info->pinned_chunks); 274304216820SFilipe Manana 2744da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2745da17066cSJeff Mahoney fs_info->nodesize = 4096; 2746da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2747da17066cSJeff Mahoney fs_info->stripesize = 4096; 2748da17066cSJeff Mahoney 2749eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2750eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2751eede2bf3SOmar Sandoval 275253b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 275353b381b3SDavid Woodhouse if (ret) { 275483c8266aSDavid Sterba err = ret; 275553b381b3SDavid Woodhouse goto fail_alloc; 275653b381b3SDavid Woodhouse } 275753b381b3SDavid Woodhouse 2758da17066cSJeff Mahoney __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 27597eccb903SChris Mason 27603c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 27611104a885SDavid Sterba 27621104a885SDavid Sterba /* 27631104a885SDavid Sterba * Read super block and check the signature bytes only 27641104a885SDavid Sterba */ 2765a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 276692fc03fbSAnand Jain if (IS_ERR(bh)) { 276792fc03fbSAnand Jain err = PTR_ERR(bh); 276816cdcec7SMiao Xie goto fail_alloc; 276920b45077SDave Young } 277039279cc3SChris Mason 27711104a885SDavid Sterba /* 27721104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 27731104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 27741104a885SDavid Sterba */ 2775ab8d0fc4SJeff Mahoney if (btrfs_check_super_csum(fs_info, bh->b_data)) { 277605135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 27771104a885SDavid Sterba err = -EINVAL; 2778b2acdddfSAnand Jain brelse(bh); 27791104a885SDavid Sterba goto fail_alloc; 27801104a885SDavid Sterba } 27811104a885SDavid Sterba 27821104a885SDavid Sterba /* 27831104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 27841104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 27851104a885SDavid Sterba * the whole block of INFO_SIZE 27861104a885SDavid Sterba */ 27876c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 27886c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 27896c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2790a061fc8dSChris Mason brelse(bh); 27915f39d397SChris Mason 27926c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 27930b86a832SChris Mason 2794069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 27951104a885SDavid Sterba if (ret) { 279605135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 27971104a885SDavid Sterba err = -EINVAL; 27981104a885SDavid Sterba goto fail_alloc; 27991104a885SDavid Sterba } 28001104a885SDavid Sterba 28016c41761fSDavid Sterba disk_super = fs_info->super_copy; 28020f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 280316cdcec7SMiao Xie goto fail_alloc; 28040f7d52f4SChris Mason 2805acce952bSliubo /* check FS state, whether FS is broken. */ 280687533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 280787533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2808acce952bSliubo 280975e7cb7fSLiu Bo /* 2810af31f5e5SChris Mason * run through our array of backup supers and setup 2811af31f5e5SChris Mason * our ring pointer to the oldest one 2812af31f5e5SChris Mason */ 2813af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2814af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2815af31f5e5SChris Mason 2816af31f5e5SChris Mason /* 281775e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 281875e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 281975e7cb7fSLiu Bo */ 282075e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 282175e7cb7fSLiu Bo 28222ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 28232b82032cSYan Zheng if (ret) { 28242b82032cSYan Zheng err = ret; 282516cdcec7SMiao Xie goto fail_alloc; 28262b82032cSYan Zheng } 2827dfe25020SChris Mason 2828f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2829f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2830f2b636e8SJosef Bacik if (features) { 283105135f59SDavid Sterba btrfs_err(fs_info, 283205135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 2833c1c9ff7cSGeert Uytterhoeven features); 2834f2b636e8SJosef Bacik err = -EINVAL; 283516cdcec7SMiao Xie goto fail_alloc; 2836f2b636e8SJosef Bacik } 2837f2b636e8SJosef Bacik 28385d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 28395d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 28400b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 2841a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 28425c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 28435c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 2844727011e0SChris Mason 28453173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 284605135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 28473173a18fSJosef Bacik 2848727011e0SChris Mason /* 2849727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2850727011e0SChris Mason * they are bigger than the page size 2851727011e0SChris Mason */ 285209cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 2853727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 285405135f59SDavid Sterba btrfs_info(fs_info, 285505135f59SDavid Sterba "flagging fs with big metadata feature"); 2856727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2857727011e0SChris Mason } 2858727011e0SChris Mason 2859bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2860bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2861b7f67055SChandan Rajendra stripesize = sectorsize; 2862707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2863963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2864bc3f116fSChris Mason 2865da17066cSJeff Mahoney /* Cache block sizes */ 2866da17066cSJeff Mahoney fs_info->nodesize = nodesize; 2867da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 2868da17066cSJeff Mahoney fs_info->stripesize = stripesize; 2869da17066cSJeff Mahoney 2870bc3f116fSChris Mason /* 2871bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2872bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2873bc3f116fSChris Mason */ 2874bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2875707e8a07SDavid Sterba (sectorsize != nodesize)) { 287605135f59SDavid Sterba btrfs_err(fs_info, 287705135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 287805135f59SDavid Sterba nodesize, sectorsize); 2879bc3f116fSChris Mason goto fail_alloc; 2880bc3f116fSChris Mason } 2881bc3f116fSChris Mason 2882ceda0864SMiao Xie /* 2883ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2884ceda0864SMiao Xie * update the flag. 2885ceda0864SMiao Xie */ 28865d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 28875d4f98a2SYan Zheng 2888f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2889f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2890bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 289105135f59SDavid Sterba btrfs_err(fs_info, 289205135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 2893c1c9ff7cSGeert Uytterhoeven features); 2894f2b636e8SJosef Bacik err = -EINVAL; 289516cdcec7SMiao Xie goto fail_alloc; 2896f2b636e8SJosef Bacik } 289761d92c32SChris Mason 28982a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 28992a458198SEric Sandeen if (ret) { 29002a458198SEric Sandeen err = ret; 29010dc3b84aSJosef Bacik goto fail_sb_buffer; 29020dc3b84aSJosef Bacik } 29034543df7eSChris Mason 29049e11ceeeSJan Kara sb->s_bdi->congested_fn = btrfs_congested_fn; 29059e11ceeeSJan Kara sb->s_bdi->congested_data = fs_info; 29069e11ceeeSJan Kara sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 2907d4417e22SNikolay Borisov sb->s_bdi->ra_pages = VM_MAX_READAHEAD * SZ_1K / PAGE_SIZE; 29089e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 29099e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 29104575c9ccSChris Mason 2911a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2912a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2913ee87cf5eSAnand Jain memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE); 2914db94535dSChris Mason 2915925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 29166bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 2917925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 291884eed90fSChris Mason if (ret) { 291905135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 29205d4f98a2SYan Zheng goto fail_sb_buffer; 292184eed90fSChris Mason } 29220b86a832SChris Mason 292384234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 2924581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 29250b86a832SChris Mason 2926da17066cSJeff Mahoney __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 29270b86a832SChris Mason 29282ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 29290b86a832SChris Mason btrfs_super_chunk_root(disk_super), 2930581c1760SQu Wenruo generation, level, NULL); 293164c043deSLiu Bo if (IS_ERR(chunk_root->node) || 293264c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 293305135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 2934e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 2935e5fffbacSchandan free_extent_buffer(chunk_root->node); 293695ab1f64SZhao Lei chunk_root->node = NULL; 2937af31f5e5SChris Mason goto fail_tree_roots; 293883121942SDavid Woodhouse } 29395d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 29405d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 29410b86a832SChris Mason 2942e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2943b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2944e17cade2SChris Mason 29455b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 29462b82032cSYan Zheng if (ret) { 294705135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 2948af31f5e5SChris Mason goto fail_tree_roots; 29492b82032cSYan Zheng } 29500b86a832SChris Mason 29518dabb742SStefan Behrens /* 29529b99b115SAnand Jain * Keep the devid that is marked to be the target device for the 29539b99b115SAnand Jain * device replace procedure 29548dabb742SStefan Behrens */ 29559b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 0); 2956dfe25020SChris Mason 2957a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 295805135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 2959a6b0d5c8SChris Mason goto fail_tree_roots; 2960a6b0d5c8SChris Mason } 2961a6b0d5c8SChris Mason 2962af31f5e5SChris Mason retry_root_backup: 296384234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 2964581c1760SQu Wenruo level = btrfs_super_root_level(disk_super); 29650b86a832SChris Mason 29662ff7e61eSJeff Mahoney tree_root->node = read_tree_block(fs_info, 2967db94535dSChris Mason btrfs_super_root(disk_super), 2968581c1760SQu Wenruo generation, level, NULL); 296964c043deSLiu Bo if (IS_ERR(tree_root->node) || 297064c043deSLiu Bo !extent_buffer_uptodate(tree_root->node)) { 297105135f59SDavid Sterba btrfs_warn(fs_info, "failed to read tree root"); 2972e5fffbacSchandan if (!IS_ERR(tree_root->node)) 2973e5fffbacSchandan free_extent_buffer(tree_root->node); 297495ab1f64SZhao Lei tree_root->node = NULL; 2975af31f5e5SChris Mason goto recovery_tree_root; 297683121942SDavid Woodhouse } 2977af31f5e5SChris Mason 29785d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 29795d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 298069e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2981db94535dSChris Mason 2982f32e48e9SChandan Rajendra mutex_lock(&tree_root->objectid_mutex); 2983f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(tree_root, 2984f32e48e9SChandan Rajendra &tree_root->highest_objectid); 2985f32e48e9SChandan Rajendra if (ret) { 2986f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2987f32e48e9SChandan Rajendra goto recovery_tree_root; 2988f32e48e9SChandan Rajendra } 2989f32e48e9SChandan Rajendra 2990f32e48e9SChandan Rajendra ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2991f32e48e9SChandan Rajendra 2992f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2993f32e48e9SChandan Rajendra 29946bccf3abSJeff Mahoney ret = btrfs_read_roots(fs_info); 29954bbcaa64SEric Sandeen if (ret) 2996af31f5e5SChris Mason goto recovery_tree_root; 2997f7a81ea4SStefan Behrens 29988929ecfaSYan, Zheng fs_info->generation = generation; 29998929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 30008929ecfaSYan, Zheng 3001cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3002cf90d884SQu Wenruo if (ret) { 3003cf90d884SQu Wenruo btrfs_err(fs_info, 3004cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3005cf90d884SQu Wenruo ret); 3006cf90d884SQu Wenruo goto fail_block_groups; 3007cf90d884SQu Wenruo } 300868310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 300968310a5eSIlya Dryomov if (ret) { 301005135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 301168310a5eSIlya Dryomov goto fail_block_groups; 301268310a5eSIlya Dryomov } 301368310a5eSIlya Dryomov 3014733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3015733f4fbbSStefan Behrens if (ret) { 301605135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3017733f4fbbSStefan Behrens goto fail_block_groups; 3018733f4fbbSStefan Behrens } 3019733f4fbbSStefan Behrens 30208dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 30218dabb742SStefan Behrens if (ret) { 302205135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 30238dabb742SStefan Behrens goto fail_block_groups; 30248dabb742SStefan Behrens } 30258dabb742SStefan Behrens 30269b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 1); 30278dabb742SStefan Behrens 3028b7c35e81SAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices, NULL); 3029b7c35e81SAnand Jain if (ret) { 303005135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 303105135f59SDavid Sterba ret); 3032b7c35e81SAnand Jain goto fail_block_groups; 3033b7c35e81SAnand Jain } 3034b7c35e81SAnand Jain 3035b7c35e81SAnand Jain ret = btrfs_sysfs_add_device(fs_devices); 3036b7c35e81SAnand Jain if (ret) { 303705135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs device interface: %d", 303805135f59SDavid Sterba ret); 3039b7c35e81SAnand Jain goto fail_fsdev_sysfs; 3040b7c35e81SAnand Jain } 3041b7c35e81SAnand Jain 304296f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 30435ac1d209SJeff Mahoney if (ret) { 304405135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3045b7c35e81SAnand Jain goto fail_fsdev_sysfs; 30465ac1d209SJeff Mahoney } 30475ac1d209SJeff Mahoney 3048c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3049c59021f8Sliubo if (ret) { 305005135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 30512365dd3cSAnand Jain goto fail_sysfs; 3052c59021f8Sliubo } 3053c59021f8Sliubo 30545b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 30551b1d1f66SJosef Bacik if (ret) { 305605135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 30572365dd3cSAnand Jain goto fail_sysfs; 30581b1d1f66SJosef Bacik } 30594330e183SQu Wenruo 30606528b99dSAnand Jain if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { 306105135f59SDavid Sterba btrfs_warn(fs_info, 30624330e183SQu Wenruo "writeable mount is not allowed due to too many missing devices"); 30632365dd3cSAnand Jain goto fail_sysfs; 3064292fd7fcSStefan Behrens } 30659078a3e1SChris Mason 3066a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3067a74a4b97SChris Mason "btrfs-cleaner"); 306857506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 30692365dd3cSAnand Jain goto fail_sysfs; 3070a74a4b97SChris Mason 3071a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3072a74a4b97SChris Mason tree_root, 3073a74a4b97SChris Mason "btrfs-transaction"); 307457506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 30753f157a2fSChris Mason goto fail_cleaner; 3076a74a4b97SChris Mason 3077583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3078c289811cSChris Mason !fs_info->fs_devices->rotating) { 3079583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3080c289811cSChris Mason } 3081c289811cSChris Mason 3082572d9ab7SDavid Sterba /* 308301327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3084572d9ab7SDavid Sterba * not have to wait for transaction commit 3085572d9ab7SDavid Sterba */ 3086572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 30873818aea2SQu Wenruo 308821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 30890b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 30902ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 30910b246afaSJeff Mahoney btrfs_test_opt(fs_info, 309221adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 309321adbd5cSStefan Behrens 1 : 0, 309421adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 309521adbd5cSStefan Behrens if (ret) 309605135f59SDavid Sterba btrfs_warn(fs_info, 309705135f59SDavid Sterba "failed to initialize integrity check module: %d", 309805135f59SDavid Sterba ret); 309921adbd5cSStefan Behrens } 310021adbd5cSStefan Behrens #endif 3101bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3102bcef60f2SArne Jansen if (ret) 3103bcef60f2SArne Jansen goto fail_trans_kthread; 310421adbd5cSStefan Behrens 3105fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3106fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3107fd708b81SJosef Bacik 310896da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 310996da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 31100b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 311163443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 311279787eaaSJeff Mahoney if (ret) { 311363443bf5SEric Sandeen err = ret; 311428c16cbbSWang Shilong goto fail_qgroup; 3115e556ce2cSYan Zheng } 3116e02119d5SChris Mason } 31171a40e23bSZheng Yan 31186bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 311979787eaaSJeff Mahoney if (ret) 312028c16cbbSWang Shilong goto fail_qgroup; 312176dda93cSYan, Zheng 3122bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 3123d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 312444c44af2SIlya Dryomov if (ret) 312528c16cbbSWang Shilong goto fail_qgroup; 312690c711abSZygo Blaxell 312790c711abSZygo Blaxell mutex_lock(&fs_info->cleaner_mutex); 31285d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 312990c711abSZygo Blaxell mutex_unlock(&fs_info->cleaner_mutex); 3130d7ce5843SMiao Xie if (ret < 0) { 313105135f59SDavid Sterba btrfs_warn(fs_info, "failed to recover relocation: %d", 313205135f59SDavid Sterba ret); 3133d7ce5843SMiao Xie err = -EINVAL; 3134bcef60f2SArne Jansen goto fail_qgroup; 3135d7ce5843SMiao Xie } 31367c2ca468SChris Mason } 31371a40e23bSZheng Yan 31383de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 31393de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 3140cb517eabSMiao Xie location.offset = 0; 31413de4586cSChris Mason 31423de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 31433140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 31443140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3145f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3146bcef60f2SArne Jansen goto fail_qgroup; 31473140c9a3SDan Carpenter } 3148c289811cSChris Mason 3149bc98a42cSDavid Howells if (sb_rdonly(sb)) 31502b6ba629SIlya Dryomov return 0; 31512b6ba629SIlya Dryomov 3152f8d468a1SOmar Sandoval if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 3153f8d468a1SOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 31546675df31SOmar Sandoval clear_free_space_tree = 1; 31556675df31SOmar Sandoval } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 31566675df31SOmar Sandoval !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 31576675df31SOmar Sandoval btrfs_warn(fs_info, "free space tree is invalid"); 31586675df31SOmar Sandoval clear_free_space_tree = 1; 31596675df31SOmar Sandoval } 31606675df31SOmar Sandoval 31616675df31SOmar Sandoval if (clear_free_space_tree) { 3162f8d468a1SOmar Sandoval btrfs_info(fs_info, "clearing free space tree"); 3163f8d468a1SOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 3164f8d468a1SOmar Sandoval if (ret) { 3165f8d468a1SOmar Sandoval btrfs_warn(fs_info, 3166f8d468a1SOmar Sandoval "failed to clear free space tree: %d", ret); 31676bccf3abSJeff Mahoney close_ctree(fs_info); 3168f8d468a1SOmar Sandoval return ret; 3169f8d468a1SOmar Sandoval } 3170f8d468a1SOmar Sandoval } 3171f8d468a1SOmar Sandoval 31720b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 3173511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 317405135f59SDavid Sterba btrfs_info(fs_info, "creating free space tree"); 3175511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3176511711afSChris Mason if (ret) { 317705135f59SDavid Sterba btrfs_warn(fs_info, 317805135f59SDavid Sterba "failed to create free space tree: %d", ret); 31796bccf3abSJeff Mahoney close_ctree(fs_info); 3180511711afSChris Mason return ret; 3181511711afSChris Mason } 3182511711afSChris Mason } 3183511711afSChris Mason 3184e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 31852b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 31862b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 31872b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 31886bccf3abSJeff Mahoney close_ctree(fs_info); 31892b6ba629SIlya Dryomov return ret; 31902b6ba629SIlya Dryomov } 3191e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 319259641015SIlya Dryomov 31932b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 31942b6ba629SIlya Dryomov if (ret) { 319505135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume balance: %d", ret); 31966bccf3abSJeff Mahoney close_ctree(fs_info); 31972b6ba629SIlya Dryomov return ret; 3198e3acc2a6SJosef Bacik } 3199e3acc2a6SJosef Bacik 32008dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 32018dabb742SStefan Behrens if (ret) { 320205135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 32036bccf3abSJeff Mahoney close_ctree(fs_info); 32048dabb742SStefan Behrens return ret; 32058dabb742SStefan Behrens } 32068dabb742SStefan Behrens 3207b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3208b382a324SJan Schmidt 32094bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 321005135f59SDavid Sterba btrfs_info(fs_info, "creating UUID tree"); 3211f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3212f7a81ea4SStefan Behrens if (ret) { 321305135f59SDavid Sterba btrfs_warn(fs_info, 321405135f59SDavid Sterba "failed to create the UUID tree: %d", ret); 32156bccf3abSJeff Mahoney close_ctree(fs_info); 3216f7a81ea4SStefan Behrens return ret; 3217f7a81ea4SStefan Behrens } 32180b246afaSJeff Mahoney } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 32194bbcaa64SEric Sandeen fs_info->generation != 32204bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 322105135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 322270f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 322370f80175SStefan Behrens if (ret) { 322405135f59SDavid Sterba btrfs_warn(fs_info, 322505135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 32266bccf3abSJeff Mahoney close_ctree(fs_info); 322770f80175SStefan Behrens return ret; 322870f80175SStefan Behrens } 322970f80175SStefan Behrens } else { 3230afcdd129SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 3231f7a81ea4SStefan Behrens } 3232afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 323347ab2a6cSJosef Bacik 32348dcddfa0SQu Wenruo /* 32358dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 32368dcddfa0SQu Wenruo * no need to keep the flag 32378dcddfa0SQu Wenruo */ 32388dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 32398dcddfa0SQu Wenruo 3240ad2b2c80SAl Viro return 0; 324139279cc3SChris Mason 3242bcef60f2SArne Jansen fail_qgroup: 3243bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 32447c2ca468SChris Mason fail_trans_kthread: 32457c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 32462ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3247faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 32483f157a2fSChris Mason fail_cleaner: 3249a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 32507c2ca468SChris Mason 32517c2ca468SChris Mason /* 32527c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 32537c2ca468SChris Mason * kthreads 32547c2ca468SChris Mason */ 32557c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 32567c2ca468SChris Mason 32572365dd3cSAnand Jain fail_sysfs: 32586618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 32592365dd3cSAnand Jain 3260b7c35e81SAnand Jain fail_fsdev_sysfs: 3261b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3262b7c35e81SAnand Jain 32631b1d1f66SJosef Bacik fail_block_groups: 326454067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3265af31f5e5SChris Mason 3266af31f5e5SChris Mason fail_tree_roots: 3267af31f5e5SChris Mason free_root_pointers(fs_info, 1); 32682b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3269af31f5e5SChris Mason 327039279cc3SChris Mason fail_sb_buffer: 32717abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 32725cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 327316cdcec7SMiao Xie fail_alloc: 32744543df7eSChris Mason fail_iput: 3275586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3276586e46e2SIlya Dryomov 32774543df7eSChris Mason iput(fs_info->btree_inode); 3278c404e0dcSMiao Xie fail_bio_counter: 32797f8d236aSDavid Sterba percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 3280963d678bSMiao Xie fail_delalloc_bytes: 3281963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3282e2d84521SMiao Xie fail_dirty_metadata_bytes: 3283e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 328476dda93cSYan, Zheng fail_srcu: 328576dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 32867e662854SQinghuang Feng fail: 328753b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3288586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3289ad2b2c80SAl Viro return err; 3290af31f5e5SChris Mason 3291af31f5e5SChris Mason recovery_tree_root: 32920b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 3293af31f5e5SChris Mason goto fail_tree_roots; 3294af31f5e5SChris Mason 3295af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3296af31f5e5SChris Mason 3297af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3298af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3299af31f5e5SChris Mason 3300af31f5e5SChris Mason /* we can't trust the free space cache either */ 3301af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3302af31f5e5SChris Mason 3303af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3304af31f5e5SChris Mason &num_backups_tried, &backup_index); 3305af31f5e5SChris Mason if (ret == -1) 3306af31f5e5SChris Mason goto fail_block_groups; 3307af31f5e5SChris Mason goto retry_root_backup; 3308eb60ceacSChris Mason } 3309663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3310eb60ceacSChris Mason 3311f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3312f2984462SChris Mason { 3313f2984462SChris Mason if (uptodate) { 3314f2984462SChris Mason set_buffer_uptodate(bh); 3315f2984462SChris Mason } else { 3316442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3317442a4f63SStefan Behrens bh->b_private; 3318442a4f63SStefan Behrens 3319fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3320b14af3b4SDavid Sterba "lost page write due to IO error on %s", 3321606686eeSJosef Bacik rcu_str_deref(device->name)); 332201327610SNicholas D Steeves /* note, we don't set_buffer_write_io_error because we have 33231259ab75SChris Mason * our own ways of dealing with the IO errors 33241259ab75SChris Mason */ 3325f2984462SChris Mason clear_buffer_uptodate(bh); 3326442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3327f2984462SChris Mason } 3328f2984462SChris Mason unlock_buffer(bh); 3329f2984462SChris Mason put_bh(bh); 3330f2984462SChris Mason } 3331f2984462SChris Mason 333229c36d72SAnand Jain int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, 333329c36d72SAnand Jain struct buffer_head **bh_ret) 333429c36d72SAnand Jain { 333529c36d72SAnand Jain struct buffer_head *bh; 333629c36d72SAnand Jain struct btrfs_super_block *super; 333729c36d72SAnand Jain u64 bytenr; 333829c36d72SAnand Jain 333929c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 334029c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 334129c36d72SAnand Jain return -EINVAL; 334229c36d72SAnand Jain 33439f6d2510SDavid Sterba bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); 334429c36d72SAnand Jain /* 334529c36d72SAnand Jain * If we fail to read from the underlying devices, as of now 334629c36d72SAnand Jain * the best option we have is to mark it EIO. 334729c36d72SAnand Jain */ 334829c36d72SAnand Jain if (!bh) 334929c36d72SAnand Jain return -EIO; 335029c36d72SAnand Jain 335129c36d72SAnand Jain super = (struct btrfs_super_block *)bh->b_data; 335229c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 335329c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 335429c36d72SAnand Jain brelse(bh); 335529c36d72SAnand Jain return -EINVAL; 335629c36d72SAnand Jain } 335729c36d72SAnand Jain 335829c36d72SAnand Jain *bh_ret = bh; 335929c36d72SAnand Jain return 0; 336029c36d72SAnand Jain } 336129c36d72SAnand Jain 336229c36d72SAnand Jain 3363a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3364a512bbf8SYan Zheng { 3365a512bbf8SYan Zheng struct buffer_head *bh; 3366a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3367a512bbf8SYan Zheng struct btrfs_super_block *super; 3368a512bbf8SYan Zheng int i; 3369a512bbf8SYan Zheng u64 transid = 0; 337092fc03fbSAnand Jain int ret = -EINVAL; 3371a512bbf8SYan Zheng 3372a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3373a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3374a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3375a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3376a512bbf8SYan Zheng */ 3377a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 337829c36d72SAnand Jain ret = btrfs_read_dev_one_super(bdev, i, &bh); 337929c36d72SAnand Jain if (ret) 3380a512bbf8SYan Zheng continue; 3381a512bbf8SYan Zheng 3382a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3383a512bbf8SYan Zheng 3384a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3385a512bbf8SYan Zheng brelse(latest); 3386a512bbf8SYan Zheng latest = bh; 3387a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3388a512bbf8SYan Zheng } else { 3389a512bbf8SYan Zheng brelse(bh); 3390a512bbf8SYan Zheng } 3391a512bbf8SYan Zheng } 339292fc03fbSAnand Jain 339392fc03fbSAnand Jain if (!latest) 339492fc03fbSAnand Jain return ERR_PTR(ret); 339592fc03fbSAnand Jain 3396a512bbf8SYan Zheng return latest; 3397a512bbf8SYan Zheng } 3398a512bbf8SYan Zheng 33994eedeb75SHisashi Hifumi /* 3400abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3401abbb3b8eSDavid Sterba * buffer heads we write are pinned. 34024eedeb75SHisashi Hifumi * 3403abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3404abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3405abbb3b8eSDavid Sterba * same for write and wait phases. 34064eedeb75SHisashi Hifumi * 3407abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or submission fails. 34084eedeb75SHisashi Hifumi */ 3409a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3410abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3411a512bbf8SYan Zheng { 3412a512bbf8SYan Zheng struct buffer_head *bh; 3413a512bbf8SYan Zheng int i; 3414a512bbf8SYan Zheng int ret; 3415a512bbf8SYan Zheng int errors = 0; 3416a512bbf8SYan Zheng u32 crc; 3417a512bbf8SYan Zheng u64 bytenr; 34181b9e619cSOmar Sandoval int op_flags; 3419a512bbf8SYan Zheng 3420a512bbf8SYan Zheng if (max_mirrors == 0) 3421a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3422a512bbf8SYan Zheng 3423a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3424a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3425935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3426935e5cc9SMiao Xie device->commit_total_bytes) 3427a512bbf8SYan Zheng break; 3428a512bbf8SYan Zheng 3429a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3430a512bbf8SYan Zheng 3431a512bbf8SYan Zheng crc = ~(u32)0; 3432abbb3b8eSDavid Sterba crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, 3433abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3434a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3435a512bbf8SYan Zheng 3436abbb3b8eSDavid Sterba /* One reference for us, and we leave it for the caller */ 34379f6d2510SDavid Sterba bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, 3438a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3439634554dcSJosef Bacik if (!bh) { 3440fb456252SJeff Mahoney btrfs_err(device->fs_info, 3441f14d104dSDavid Sterba "couldn't get super buffer head for bytenr %llu", 3442f14d104dSDavid Sterba bytenr); 3443634554dcSJosef Bacik errors++; 3444634554dcSJosef Bacik continue; 3445634554dcSJosef Bacik } 3446634554dcSJosef Bacik 3447a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3448a512bbf8SYan Zheng 34494eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3450a512bbf8SYan Zheng get_bh(bh); 34514eedeb75SHisashi Hifumi 34524eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3453a512bbf8SYan Zheng lock_buffer(bh); 3454a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3455442a4f63SStefan Behrens bh->b_private = device; 3456a512bbf8SYan Zheng 3457387125fcSChris Mason /* 3458387125fcSChris Mason * we fua the first super. The others we allow 3459387125fcSChris Mason * to go down lazy. 3460387125fcSChris Mason */ 34611b9e619cSOmar Sandoval op_flags = REQ_SYNC | REQ_META | REQ_PRIO; 34621b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 34631b9e619cSOmar Sandoval op_flags |= REQ_FUA; 34641b9e619cSOmar Sandoval ret = btrfsic_submit_bh(REQ_OP_WRITE, op_flags, bh); 34654eedeb75SHisashi Hifumi if (ret) 3466a512bbf8SYan Zheng errors++; 3467a512bbf8SYan Zheng } 3468a512bbf8SYan Zheng return errors < i ? 0 : -1; 3469a512bbf8SYan Zheng } 3470a512bbf8SYan Zheng 3471387125fcSChris Mason /* 3472abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3473abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3474abbb3b8eSDavid Sterba * 3475abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or not marked up to 3476abbb3b8eSDavid Sterba * date. 3477abbb3b8eSDavid Sterba */ 3478abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3479abbb3b8eSDavid Sterba { 3480abbb3b8eSDavid Sterba struct buffer_head *bh; 3481abbb3b8eSDavid Sterba int i; 3482abbb3b8eSDavid Sterba int errors = 0; 3483b6a535faSHoward McLauchlan bool primary_failed = false; 3484abbb3b8eSDavid Sterba u64 bytenr; 3485abbb3b8eSDavid Sterba 3486abbb3b8eSDavid Sterba if (max_mirrors == 0) 3487abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3488abbb3b8eSDavid Sterba 3489abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3490abbb3b8eSDavid Sterba bytenr = btrfs_sb_offset(i); 3491abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3492abbb3b8eSDavid Sterba device->commit_total_bytes) 3493abbb3b8eSDavid Sterba break; 3494abbb3b8eSDavid Sterba 34959f6d2510SDavid Sterba bh = __find_get_block(device->bdev, 34969f6d2510SDavid Sterba bytenr / BTRFS_BDEV_BLOCKSIZE, 3497abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE); 3498abbb3b8eSDavid Sterba if (!bh) { 3499abbb3b8eSDavid Sterba errors++; 3500b6a535faSHoward McLauchlan if (i == 0) 3501b6a535faSHoward McLauchlan primary_failed = true; 3502abbb3b8eSDavid Sterba continue; 3503abbb3b8eSDavid Sterba } 3504abbb3b8eSDavid Sterba wait_on_buffer(bh); 3505b6a535faSHoward McLauchlan if (!buffer_uptodate(bh)) { 3506abbb3b8eSDavid Sterba errors++; 3507b6a535faSHoward McLauchlan if (i == 0) 3508b6a535faSHoward McLauchlan primary_failed = true; 3509b6a535faSHoward McLauchlan } 3510abbb3b8eSDavid Sterba 3511abbb3b8eSDavid Sterba /* drop our reference */ 3512abbb3b8eSDavid Sterba brelse(bh); 3513abbb3b8eSDavid Sterba 3514abbb3b8eSDavid Sterba /* drop the reference from the writing run */ 3515abbb3b8eSDavid Sterba brelse(bh); 3516abbb3b8eSDavid Sterba } 3517abbb3b8eSDavid Sterba 3518b6a535faSHoward McLauchlan /* log error, force error return */ 3519b6a535faSHoward McLauchlan if (primary_failed) { 3520b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3521b6a535faSHoward McLauchlan device->devid); 3522b6a535faSHoward McLauchlan return -1; 3523b6a535faSHoward McLauchlan } 3524b6a535faSHoward McLauchlan 3525abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3526abbb3b8eSDavid Sterba } 3527abbb3b8eSDavid Sterba 3528abbb3b8eSDavid Sterba /* 3529387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3530387125fcSChris Mason * for the barrier when it is done 3531387125fcSChris Mason */ 35324246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3533387125fcSChris Mason { 3534387125fcSChris Mason complete(bio->bi_private); 3535387125fcSChris Mason } 3536387125fcSChris Mason 3537387125fcSChris Mason /* 35384fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 35394fc6441aSAnand Jain * done in the waiting counterpart. 3540387125fcSChris Mason */ 35414fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3542387125fcSChris Mason { 3543c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3544e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3545387125fcSChris Mason 3546c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 35474fc6441aSAnand Jain return; 3548387125fcSChris Mason 3549e0ae9994SDavid Sterba bio_reset(bio); 3550387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 355174d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 35528d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3553387125fcSChris Mason init_completion(&device->flush_wait); 3554387125fcSChris Mason bio->bi_private = &device->flush_wait; 3555387125fcSChris Mason 355643a01111SLu Fengqi btrfsic_submit_bio(bio); 35571c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 35584fc6441aSAnand Jain } 3559387125fcSChris Mason 35604fc6441aSAnand Jain /* 35614fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 35624fc6441aSAnand Jain */ 35638c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 35644fc6441aSAnand Jain { 35654fc6441aSAnand Jain struct bio *bio = device->flush_bio; 35664fc6441aSAnand Jain 35671c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 356858efbc9fSOmar Sandoval return BLK_STS_OK; 35694fc6441aSAnand Jain 35701c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 35712980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 35724fc6441aSAnand Jain 35738c27cb35SLinus Torvalds return bio->bi_status; 3574387125fcSChris Mason } 3575387125fcSChris Mason 3576d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3577401b41e5SAnand Jain { 35786528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 3579401b41e5SAnand Jain return -EIO; 3580387125fcSChris Mason return 0; 3581387125fcSChris Mason } 3582387125fcSChris Mason 3583387125fcSChris Mason /* 3584387125fcSChris Mason * send an empty flush down to each device in parallel, 3585387125fcSChris Mason * then wait for them 3586387125fcSChris Mason */ 3587387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3588387125fcSChris Mason { 3589387125fcSChris Mason struct list_head *head; 3590387125fcSChris Mason struct btrfs_device *dev; 35915af3e8ccSStefan Behrens int errors_wait = 0; 35924e4cbee9SChristoph Hellwig blk_status_t ret; 3593387125fcSChris Mason 35941538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3595387125fcSChris Mason /* send down all the barriers */ 3596387125fcSChris Mason head = &info->fs_devices->devices; 35971538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3598e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3599f88ba6a2SHidetoshi Seto continue; 3600cea7c8bfSAnand Jain if (!dev->bdev) 3601387125fcSChris Mason continue; 3602e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3603ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3604387125fcSChris Mason continue; 3605387125fcSChris Mason 36064fc6441aSAnand Jain write_dev_flush(dev); 360758efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3608387125fcSChris Mason } 3609387125fcSChris Mason 3610387125fcSChris Mason /* wait for all the barriers */ 36111538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3612e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3613f88ba6a2SHidetoshi Seto continue; 3614387125fcSChris Mason if (!dev->bdev) { 36155af3e8ccSStefan Behrens errors_wait++; 3616387125fcSChris Mason continue; 3617387125fcSChris Mason } 3618e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3619ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3620387125fcSChris Mason continue; 3621387125fcSChris Mason 36224fc6441aSAnand Jain ret = wait_dev_flush(dev); 3623401b41e5SAnand Jain if (ret) { 3624401b41e5SAnand Jain dev->last_flush_error = ret; 362566b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 362666b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 36275af3e8ccSStefan Behrens errors_wait++; 3628387125fcSChris Mason } 3629401b41e5SAnand Jain } 3630401b41e5SAnand Jain 3631cea7c8bfSAnand Jain if (errors_wait) { 3632401b41e5SAnand Jain /* 3633401b41e5SAnand Jain * At some point we need the status of all disks 3634401b41e5SAnand Jain * to arrive at the volume status. So error checking 3635401b41e5SAnand Jain * is being pushed to a separate loop. 3636401b41e5SAnand Jain */ 3637d10b82feSQu Wenruo return check_barrier_error(info); 3638401b41e5SAnand Jain } 3639387125fcSChris Mason return 0; 3640387125fcSChris Mason } 3641387125fcSChris Mason 3642943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3643943c6e99SZhao Lei { 36448789f4feSZhao Lei int raid_type; 36458789f4feSZhao Lei int min_tolerated = INT_MAX; 3646943c6e99SZhao Lei 36478789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 36488789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 36498789f4feSZhao Lei min_tolerated = min(min_tolerated, 36508789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 36518789f4feSZhao Lei tolerated_failures); 3652943c6e99SZhao Lei 36538789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 36548789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 36558789f4feSZhao Lei continue; 365641a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 36578789f4feSZhao Lei continue; 36588789f4feSZhao Lei min_tolerated = min(min_tolerated, 36598789f4feSZhao Lei btrfs_raid_array[raid_type]. 36608789f4feSZhao Lei tolerated_failures); 36618789f4feSZhao Lei } 3662943c6e99SZhao Lei 36638789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3664ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 36658789f4feSZhao Lei min_tolerated = 0; 36668789f4feSZhao Lei } 36678789f4feSZhao Lei 36688789f4feSZhao Lei return min_tolerated; 3669943c6e99SZhao Lei } 3670943c6e99SZhao Lei 3671eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3672f2984462SChris Mason { 3673e5e9a520SChris Mason struct list_head *head; 3674f2984462SChris Mason struct btrfs_device *dev; 3675a061fc8dSChris Mason struct btrfs_super_block *sb; 3676f2984462SChris Mason struct btrfs_dev_item *dev_item; 3677f2984462SChris Mason int ret; 3678f2984462SChris Mason int do_barriers; 3679a236aed1SChris Mason int max_errors; 3680a236aed1SChris Mason int total_errors = 0; 3681a061fc8dSChris Mason u64 flags; 3682f2984462SChris Mason 36830b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3684fed3b381SLiu Bo 3685fed3b381SLiu Bo /* 3686fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3687fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3688fed3b381SLiu Bo * been consistent on disk. 3689fed3b381SLiu Bo */ 3690fed3b381SLiu Bo if (max_mirrors == 0) 36910b246afaSJeff Mahoney backup_super_roots(fs_info); 3692f2984462SChris Mason 36930b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3694a061fc8dSChris Mason dev_item = &sb->dev_item; 3695e5e9a520SChris Mason 36960b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 36970b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 36980b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3699387125fcSChris Mason 37005af3e8ccSStefan Behrens if (do_barriers) { 37010b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 37025af3e8ccSStefan Behrens if (ret) { 37035af3e8ccSStefan Behrens mutex_unlock( 37040b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 37050b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 37065af3e8ccSStefan Behrens "errors while submitting device barriers."); 37075af3e8ccSStefan Behrens return ret; 37085af3e8ccSStefan Behrens } 37095af3e8ccSStefan Behrens } 3710387125fcSChris Mason 37111538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3712dfe25020SChris Mason if (!dev->bdev) { 3713dfe25020SChris Mason total_errors++; 3714dfe25020SChris Mason continue; 3715dfe25020SChris Mason } 3716e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3717ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3718dfe25020SChris Mason continue; 3719dfe25020SChris Mason 37202b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3721a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3722a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 37237df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3724935e5cc9SMiao Xie dev->commit_total_bytes); 3725ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3726ce7213c7SMiao Xie dev->commit_bytes_used); 3727a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3728a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3729a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3730a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 373144880fdcSAnand Jain memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); 3732a512bbf8SYan Zheng 3733a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3734a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3735f2984462SChris Mason 373675cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 373775cb857dSQu Wenruo if (ret < 0) { 373875cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 373975cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 374075cb857dSQu Wenruo "unexpected superblock corruption detected"); 374175cb857dSQu Wenruo return -EUCLEAN; 374275cb857dSQu Wenruo } 374375cb857dSQu Wenruo 3744abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3745a236aed1SChris Mason if (ret) 3746a236aed1SChris Mason total_errors++; 3747f2984462SChris Mason } 3748a236aed1SChris Mason if (total_errors > max_errors) { 37490b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3750d397712bSChris Mason total_errors); 37510b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 375279787eaaSJeff Mahoney 37539d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 37540b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 37550b246afaSJeff Mahoney "%d errors while writing supers", 37560b246afaSJeff Mahoney total_errors); 37579d565ba4SStefan Behrens return -EIO; 3758a236aed1SChris Mason } 3759f2984462SChris Mason 3760a512bbf8SYan Zheng total_errors = 0; 37611538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3762dfe25020SChris Mason if (!dev->bdev) 3763dfe25020SChris Mason continue; 3764e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3765ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3766dfe25020SChris Mason continue; 3767dfe25020SChris Mason 3768abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3769a512bbf8SYan Zheng if (ret) 37701259ab75SChris Mason total_errors++; 3771f2984462SChris Mason } 37720b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3773a236aed1SChris Mason if (total_errors > max_errors) { 37740b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 37750b246afaSJeff Mahoney "%d errors while writing supers", 37760b246afaSJeff Mahoney total_errors); 377779787eaaSJeff Mahoney return -EIO; 3778a236aed1SChris Mason } 3779f2984462SChris Mason return 0; 3780f2984462SChris Mason } 3781f2984462SChris Mason 3782cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3783cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3784cb517eabSMiao Xie struct btrfs_root *root) 37852619ba1fSChris Mason { 37864df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 37872619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 37882619ba1fSChris Mason (unsigned long)root->root_key.objectid); 37894df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 379076dda93cSYan, Zheng 379176dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 379276dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 379376dda93cSYan, Zheng 37941c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 37953321719eSLiu Bo btrfs_free_log(NULL, root); 37961c1ea4f7SLiu Bo if (root->reloc_root) { 37971c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->node); 37981c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->commit_root); 37991c1ea4f7SLiu Bo btrfs_put_fs_root(root->reloc_root); 38001c1ea4f7SLiu Bo root->reloc_root = NULL; 38011c1ea4f7SLiu Bo } 38021c1ea4f7SLiu Bo } 38033321719eSLiu Bo 3804faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3805581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3806faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3807581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 380884db5ccfSDavid Sterba btrfs_free_fs_root(root); 38094df27c4dSYan, Zheng } 38104df27c4dSYan, Zheng 381184db5ccfSDavid Sterba void btrfs_free_fs_root(struct btrfs_root *root) 38124df27c4dSYan, Zheng { 381357cdc8dbSDavid Sterba iput(root->ino_cache_inode); 38144df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 38150ee5dc67SAl Viro if (root->anon_dev) 38160ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 38178257b2dcSMiao Xie if (root->subv_writers) 38188257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 38195f39d397SChris Mason free_extent_buffer(root->node); 38205f39d397SChris Mason free_extent_buffer(root->commit_root); 3821581bb050SLi Zefan kfree(root->free_ino_ctl); 3822581bb050SLi Zefan kfree(root->free_ino_pinned); 3823b0feb9d9SMiao Xie btrfs_put_fs_root(root); 38242619ba1fSChris Mason } 38252619ba1fSChris Mason 3826c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3827c146afadSYan Zheng { 3828c146afadSYan Zheng u64 root_objectid = 0; 3829c146afadSYan Zheng struct btrfs_root *gang[8]; 383065d33fd7SQu Wenruo int i = 0; 383165d33fd7SQu Wenruo int err = 0; 383265d33fd7SQu Wenruo unsigned int ret = 0; 383365d33fd7SQu Wenruo int index; 3834c146afadSYan Zheng 3835c146afadSYan Zheng while (1) { 383665d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3837c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3838c146afadSYan Zheng (void **)gang, root_objectid, 3839c146afadSYan Zheng ARRAY_SIZE(gang)); 384065d33fd7SQu Wenruo if (!ret) { 384165d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3842c146afadSYan Zheng break; 384365d33fd7SQu Wenruo } 38445d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 384566b4ffd1SJosef Bacik 384665d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 384765d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 384865d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 384965d33fd7SQu Wenruo gang[i] = NULL; 385065d33fd7SQu Wenruo continue; 385165d33fd7SQu Wenruo } 385265d33fd7SQu Wenruo /* grab all the search result for later use */ 385365d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 385465d33fd7SQu Wenruo } 385565d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 385665d33fd7SQu Wenruo 385765d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 385865d33fd7SQu Wenruo if (!gang[i]) 385965d33fd7SQu Wenruo continue; 3860c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 386166b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 386266b4ffd1SJosef Bacik if (err) 386365d33fd7SQu Wenruo break; 386465d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3865c146afadSYan Zheng } 3866c146afadSYan Zheng root_objectid++; 3867c146afadSYan Zheng } 386865d33fd7SQu Wenruo 386965d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 387065d33fd7SQu Wenruo for (; i < ret; i++) { 387165d33fd7SQu Wenruo if (gang[i]) 387265d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 387365d33fd7SQu Wenruo } 387465d33fd7SQu Wenruo return err; 3875c146afadSYan Zheng } 3876c146afadSYan Zheng 38776bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 3878c146afadSYan Zheng { 38796bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 3880c146afadSYan Zheng struct btrfs_trans_handle *trans; 3881c146afadSYan Zheng 38820b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 38832ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 38840b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 38850b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 3886c71bf099SYan, Zheng 3887c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 38880b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 38890b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 3890c71bf099SYan, Zheng 38917a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 38923612b495STsutomu Itoh if (IS_ERR(trans)) 38933612b495STsutomu Itoh return PTR_ERR(trans); 38943a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 3895c146afadSYan Zheng } 3896c146afadSYan Zheng 38976bccf3abSJeff Mahoney void close_ctree(struct btrfs_fs_info *fs_info) 3898eb60ceacSChris Mason { 3899c146afadSYan Zheng int ret; 3900e089f05cSChris Mason 3901afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 3902d6fd0ae2SOmar Sandoval /* 3903d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 3904d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 3905d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 3906d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 3907d6fd0ae2SOmar Sandoval */ 3908d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 3909a2135011SChris Mason 39107343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 3911d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 39127343dd61SJustin Maggard 3913803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3914803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3915803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3916803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3917803b2f54SStefan Behrens 3918837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3919aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3920837d5b6eSIlya Dryomov 39218dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 39228dabb742SStefan Behrens 3923aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 39244cb5300bSChris Mason 39254cb5300bSChris Mason /* wait for any defraggers to finish */ 39264cb5300bSChris Mason wait_event(fs_info->transaction_wait, 39274cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 39284cb5300bSChris Mason 39294cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 393026176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 39314cb5300bSChris Mason 393221c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 393321c7e756SMiao Xie 3934bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 3935e44163e1SJeff Mahoney /* 3936d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 3937d6fd0ae2SOmar Sandoval * unused block groups. 3938e44163e1SJeff Mahoney */ 39390b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 3940e44163e1SJeff Mahoney 39416bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 3942d397712bSChris Mason if (ret) 394304892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 3944c146afadSYan Zheng } 3945ed2ff2cbSChris Mason 3946af722733SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || 3947af722733SLiu Bo test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) 39482ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 3949acce952bSliubo 3950e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3951e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 39528929ecfaSYan, Zheng 3953e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 3954afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 3955f25784b3SYan Zheng 395604892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 3957fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 3958bcef60f2SArne Jansen 3959963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 396004892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 3961963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3962b0c68f8bSChris Mason } 396331153d81SYan Zheng 39646618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 3965b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 39665ac1d209SJeff Mahoney 3967faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3968d10c5f31SChris Mason 39691a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 39701a4319ccSLiu Bo 3971de348ee0SWang Shilong /* 3972de348ee0SWang Shilong * we must make sure there is not any read request to 3973de348ee0SWang Shilong * submit after we stopping all workers. 3974de348ee0SWang Shilong */ 3975de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 397696192499SJosef Bacik btrfs_stop_all_workers(fs_info); 397796192499SJosef Bacik 39785cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 39795cdd7db6SFilipe Manana 3980afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 398113e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 39829ad6b7bcSChris Mason 398313e6c37bSJosef Bacik iput(fs_info->btree_inode); 3984d6bfde87SChris Mason 398521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 39860b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 39872ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 398821adbd5cSStefan Behrens #endif 398921adbd5cSStefan Behrens 3990dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 39910b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3992b248a415SChris Mason 3993e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3994963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 39957f8d236aSDavid Sterba percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 399676dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 39970b86a832SChris Mason 399853b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3999fd708b81SJosef Bacik btrfs_free_ref_cache(fs_info); 400053b381b3SDavid Woodhouse 400104216820SFilipe Manana while (!list_empty(&fs_info->pinned_chunks)) { 400204216820SFilipe Manana struct extent_map *em; 400304216820SFilipe Manana 400404216820SFilipe Manana em = list_first_entry(&fs_info->pinned_chunks, 400504216820SFilipe Manana struct extent_map, list); 400604216820SFilipe Manana list_del_init(&em->list); 400704216820SFilipe Manana free_extent_map(em); 400804216820SFilipe Manana } 4009eb60ceacSChris Mason } 4010eb60ceacSChris Mason 4011b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4012b9fab919SChris Mason int atomic) 4013ccd467d6SChris Mason { 40141259ab75SChris Mason int ret; 4015727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 40161259ab75SChris Mason 40170b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 40181259ab75SChris Mason if (!ret) 40191259ab75SChris Mason return ret; 40201259ab75SChris Mason 40211259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4022b9fab919SChris Mason parent_transid, atomic); 4023b9fab919SChris Mason if (ret == -EAGAIN) 4024b9fab919SChris Mason return ret; 40251259ab75SChris Mason return !ret; 40265f39d397SChris Mason } 40276702ed49SChris Mason 40285f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 40295f39d397SChris Mason { 40300b246afaSJeff Mahoney struct btrfs_fs_info *fs_info; 403106ea65a3SJosef Bacik struct btrfs_root *root; 40325f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4033b9473439SChris Mason int was_dirty; 4034b4ce94deSChris Mason 403506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 403606ea65a3SJosef Bacik /* 403706ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 4038b0132a3bSNikolay Borisov * enabled. Normal people shouldn't be using umapped buffers as dirty 403906ea65a3SJosef Bacik * outside of the sanity tests. 404006ea65a3SJosef Bacik */ 4041b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 404206ea65a3SJosef Bacik return; 404306ea65a3SJosef Bacik #endif 404406ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 40450b246afaSJeff Mahoney fs_info = root->fs_info; 4046b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 40470b246afaSJeff Mahoney if (transid != fs_info->generation) 40485d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 40490b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 40500b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4051e2d84521SMiao Xie if (!was_dirty) 4052104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4053e2d84521SMiao Xie buf->len, 40540b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 40551f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 405669fc6cbbSQu Wenruo /* 405769fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 405869fc6cbbSQu Wenruo * but item data not updated. 405969fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 406069fc6cbbSQu Wenruo */ 406169fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 40622f659546SQu Wenruo btrfs_check_leaf_relaxed(fs_info, buf)) { 4063a4f78750SDavid Sterba btrfs_print_leaf(buf); 40641f21ef0aSFilipe Manana ASSERT(0); 40651f21ef0aSFilipe Manana } 40661f21ef0aSFilipe Manana #endif 4067eb60ceacSChris Mason } 4068eb60ceacSChris Mason 40692ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4070b53d3f5dSLiu Bo int flush_delayed) 407135b7e476SChris Mason { 4072188de649SChris Mason /* 4073188de649SChris Mason * looks as though older kernels can get into trouble with 4074188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4075188de649SChris Mason */ 4076e2d84521SMiao Xie int ret; 4077d6bfde87SChris Mason 40786933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4079d6bfde87SChris Mason return; 4080d6bfde87SChris Mason 4081b53d3f5dSLiu Bo if (flush_delayed) 40822ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 408316cdcec7SMiao Xie 4084d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4085d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4086d814a491SEthan Lien fs_info->dirty_metadata_batch); 4087e2d84521SMiao Xie if (ret > 0) { 40880b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 408916cdcec7SMiao Xie } 409016cdcec7SMiao Xie } 409116cdcec7SMiao Xie 40922ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 409316cdcec7SMiao Xie { 40942ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 409535b7e476SChris Mason } 4096b53d3f5dSLiu Bo 40972ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4098b53d3f5dSLiu Bo { 40992ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4100d6bfde87SChris Mason } 41016b80053dSChris Mason 4102581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4103581c1760SQu Wenruo struct btrfs_key *first_key) 41046b80053dSChris Mason { 4105727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 41062ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 41072ff7e61eSJeff Mahoney 4108581c1760SQu Wenruo return btree_read_extent_buffer_pages(fs_info, buf, parent_transid, 4109581c1760SQu Wenruo level, first_key); 41106b80053dSChris Mason } 41110da5468fSChris Mason 41122ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4113acce952bSliubo { 4114fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4115fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4116fe816d0fSNikolay Borisov 41170b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41182ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41190b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4120acce952bSliubo 41210b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41220b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4123acce952bSliubo } 4124acce952bSliubo 4125143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4126acce952bSliubo { 4127acce952bSliubo struct btrfs_ordered_extent *ordered; 4128acce952bSliubo 4129199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4130779880efSJosef Bacik /* 4131779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4132779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4133779880efSJosef Bacik */ 4134199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4135779880efSJosef Bacik root_extent_list) 4136779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4137199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4138199c2a9cSMiao Xie } 4139199c2a9cSMiao Xie 4140199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4141199c2a9cSMiao Xie { 4142199c2a9cSMiao Xie struct btrfs_root *root; 4143199c2a9cSMiao Xie struct list_head splice; 4144199c2a9cSMiao Xie 4145199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4146199c2a9cSMiao Xie 4147199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4148199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4149199c2a9cSMiao Xie while (!list_empty(&splice)) { 4150199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4151199c2a9cSMiao Xie ordered_root); 41521de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 41531de2cfdeSJosef Bacik &fs_info->ordered_roots); 4154199c2a9cSMiao Xie 41552a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4156199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4157199c2a9cSMiao Xie 41582a85d9caSLiu Bo cond_resched(); 41592a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4160199c2a9cSMiao Xie } 4161199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 4162acce952bSliubo } 4163acce952bSliubo 416435a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 41652ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4166acce952bSliubo { 4167acce952bSliubo struct rb_node *node; 4168acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4169acce952bSliubo struct btrfs_delayed_ref_node *ref; 4170acce952bSliubo int ret = 0; 4171acce952bSliubo 4172acce952bSliubo delayed_refs = &trans->delayed_refs; 4173acce952bSliubo 4174acce952bSliubo spin_lock(&delayed_refs->lock); 4175d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4176cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 41770b246afaSJeff Mahoney btrfs_info(fs_info, "delayed_refs has NO entry"); 4178acce952bSliubo return ret; 4179acce952bSliubo } 4180acce952bSliubo 41815c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4182d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 41830e0adbcfSJosef Bacik struct rb_node *n; 4184e78417d1SJosef Bacik bool pin_bytes = false; 4185acce952bSliubo 4186d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4187d7df2c79SJosef Bacik href_node); 4188b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 4189d278850eSJosef Bacik refcount_inc(&head->refs); 419079787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 4191b939d1abSJosef Bacik 4192acce952bSliubo mutex_lock(&head->mutex); 4193b939d1abSJosef Bacik mutex_unlock(&head->mutex); 4194d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4195e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 4196b939d1abSJosef Bacik continue; 4197b939d1abSJosef Bacik } 4198d7df2c79SJosef Bacik spin_lock(&head->lock); 4199e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 42000e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 42010e0adbcfSJosef Bacik ref_node); 4202d7df2c79SJosef Bacik ref->in_tree = 0; 4203e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 42040e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 42051d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 42061d57ee94SWang Xiaoguang list_del(&ref->add_list); 4207d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4208d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4209d7df2c79SJosef Bacik } 421054067ae9SJosef Bacik if (head->must_insert_reserved) 4211e78417d1SJosef Bacik pin_bytes = true; 421278a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4213acce952bSliubo delayed_refs->num_heads--; 4214d7df2c79SJosef Bacik if (head->processing == 0) 4215acce952bSliubo delayed_refs->num_heads_ready--; 4216d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 42175c9d028bSLiu Bo rb_erase_cached(&head->href_node, &delayed_refs->href_root); 4218d278850eSJosef Bacik RB_CLEAR_NODE(&head->href_node); 4219d7df2c79SJosef Bacik spin_unlock(&head->lock); 4220acce952bSliubo spin_unlock(&delayed_refs->lock); 4221e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4222acce952bSliubo 4223d7df2c79SJosef Bacik if (pin_bytes) 4224d278850eSJosef Bacik btrfs_pin_extent(fs_info, head->bytenr, 4225d278850eSJosef Bacik head->num_bytes, 1); 4226d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4227acce952bSliubo cond_resched(); 4228acce952bSliubo spin_lock(&delayed_refs->lock); 4229acce952bSliubo } 4230acce952bSliubo 4231acce952bSliubo spin_unlock(&delayed_refs->lock); 4232acce952bSliubo 4233acce952bSliubo return ret; 4234acce952bSliubo } 4235acce952bSliubo 4236143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4237acce952bSliubo { 4238acce952bSliubo struct btrfs_inode *btrfs_inode; 4239acce952bSliubo struct list_head splice; 4240acce952bSliubo 4241acce952bSliubo INIT_LIST_HEAD(&splice); 4242acce952bSliubo 4243eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4244eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4245acce952bSliubo 4246acce952bSliubo while (!list_empty(&splice)) { 4247fe816d0fSNikolay Borisov struct inode *inode = NULL; 4248eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4249acce952bSliubo delalloc_inodes); 4250fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4251eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4252acce952bSliubo 4253fe816d0fSNikolay Borisov /* 4254fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4255fe816d0fSNikolay Borisov * meanwhile. 4256fe816d0fSNikolay Borisov */ 4257fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4258fe816d0fSNikolay Borisov if (inode) { 4259fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4260fe816d0fSNikolay Borisov iput(inode); 4261fe816d0fSNikolay Borisov } 4262eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4263acce952bSliubo } 4264eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4265eb73c1b7SMiao Xie } 4266eb73c1b7SMiao Xie 4267eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4268eb73c1b7SMiao Xie { 4269eb73c1b7SMiao Xie struct btrfs_root *root; 4270eb73c1b7SMiao Xie struct list_head splice; 4271eb73c1b7SMiao Xie 4272eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4273eb73c1b7SMiao Xie 4274eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4275eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4276eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4277eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4278eb73c1b7SMiao Xie delalloc_root); 4279eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 4280eb73c1b7SMiao Xie BUG_ON(!root); 4281eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4282eb73c1b7SMiao Xie 4283eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 4284eb73c1b7SMiao Xie btrfs_put_fs_root(root); 4285eb73c1b7SMiao Xie 4286eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4287eb73c1b7SMiao Xie } 4288eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4289acce952bSliubo } 4290acce952bSliubo 42912ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4292acce952bSliubo struct extent_io_tree *dirty_pages, 4293acce952bSliubo int mark) 4294acce952bSliubo { 4295acce952bSliubo int ret; 4296acce952bSliubo struct extent_buffer *eb; 4297acce952bSliubo u64 start = 0; 4298acce952bSliubo u64 end; 4299acce952bSliubo 4300acce952bSliubo while (1) { 4301acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4302e6138876SJosef Bacik mark, NULL); 4303acce952bSliubo if (ret) 4304acce952bSliubo break; 4305acce952bSliubo 430691166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4307acce952bSliubo while (start <= end) { 43080b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 43090b246afaSJeff Mahoney start += fs_info->nodesize; 4310fd8b2b61SJosef Bacik if (!eb) 4311acce952bSliubo continue; 4312fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4313acce952bSliubo 4314fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4315fd8b2b61SJosef Bacik &eb->bflags)) 4316fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4317fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4318acce952bSliubo } 4319acce952bSliubo } 4320acce952bSliubo 4321acce952bSliubo return ret; 4322acce952bSliubo } 4323acce952bSliubo 43242ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4325acce952bSliubo struct extent_io_tree *pinned_extents) 4326acce952bSliubo { 4327acce952bSliubo struct extent_io_tree *unpin; 4328acce952bSliubo u64 start; 4329acce952bSliubo u64 end; 4330acce952bSliubo int ret; 4331ed0eaa14SLiu Bo bool loop = true; 4332acce952bSliubo 4333acce952bSliubo unpin = pinned_extents; 4334ed0eaa14SLiu Bo again: 4335acce952bSliubo while (1) { 4336fcd5e742SLu Fengqi /* 4337fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4338fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4339fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4340fcd5e742SLu Fengqi * the same extent range. 4341fcd5e742SLu Fengqi */ 4342fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4343acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4344e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4345fcd5e742SLu Fengqi if (ret) { 4346fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4347acce952bSliubo break; 4348fcd5e742SLu Fengqi } 4349acce952bSliubo 4350af6f8f60SDavid Sterba clear_extent_dirty(unpin, start, end); 43512ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4352fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4353acce952bSliubo cond_resched(); 4354acce952bSliubo } 4355acce952bSliubo 4356ed0eaa14SLiu Bo if (loop) { 43570b246afaSJeff Mahoney if (unpin == &fs_info->freed_extents[0]) 43580b246afaSJeff Mahoney unpin = &fs_info->freed_extents[1]; 4359ed0eaa14SLiu Bo else 43600b246afaSJeff Mahoney unpin = &fs_info->freed_extents[0]; 4361ed0eaa14SLiu Bo loop = false; 4362ed0eaa14SLiu Bo goto again; 4363ed0eaa14SLiu Bo } 4364ed0eaa14SLiu Bo 4365acce952bSliubo return 0; 4366acce952bSliubo } 4367acce952bSliubo 4368c79a1751SLiu Bo static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) 4369c79a1751SLiu Bo { 4370c79a1751SLiu Bo struct inode *inode; 4371c79a1751SLiu Bo 4372c79a1751SLiu Bo inode = cache->io_ctl.inode; 4373c79a1751SLiu Bo if (inode) { 4374c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4375c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4376c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4377c79a1751SLiu Bo iput(inode); 4378c79a1751SLiu Bo } 4379c79a1751SLiu Bo btrfs_put_block_group(cache); 4380c79a1751SLiu Bo } 4381c79a1751SLiu Bo 4382c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 43832ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4384c79a1751SLiu Bo { 4385c79a1751SLiu Bo struct btrfs_block_group_cache *cache; 4386c79a1751SLiu Bo 4387c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4388c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4389c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 4390c79a1751SLiu Bo struct btrfs_block_group_cache, 4391c79a1751SLiu Bo dirty_list); 4392c79a1751SLiu Bo 4393c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4394c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4395c79a1751SLiu Bo list_del_init(&cache->io_list); 4396c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4397c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4398c79a1751SLiu Bo } 4399c79a1751SLiu Bo 4400c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4401c79a1751SLiu Bo spin_lock(&cache->lock); 4402c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4403c79a1751SLiu Bo spin_unlock(&cache->lock); 4404c79a1751SLiu Bo 4405c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4406c79a1751SLiu Bo btrfs_put_block_group(cache); 4407c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4408c79a1751SLiu Bo } 4409c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4410c79a1751SLiu Bo 441145ae2c18SNikolay Borisov /* 441245ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 441345ae2c18SNikolay Borisov * to use it without any locking 441445ae2c18SNikolay Borisov */ 4415c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4416c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 4417c79a1751SLiu Bo struct btrfs_block_group_cache, 4418c79a1751SLiu Bo io_list); 4419c79a1751SLiu Bo 4420c79a1751SLiu Bo list_del_init(&cache->io_list); 4421c79a1751SLiu Bo spin_lock(&cache->lock); 4422c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4423c79a1751SLiu Bo spin_unlock(&cache->lock); 4424c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4425c79a1751SLiu Bo } 4426c79a1751SLiu Bo } 4427c79a1751SLiu Bo 442849b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 44292ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 443049b25e05SJeff Mahoney { 44312ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4432c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4433c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4434c79a1751SLiu Bo 44352ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 443649b25e05SJeff Mahoney 44374a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 44380b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 443949b25e05SJeff Mahoney 44404a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 44410b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 444249b25e05SJeff Mahoney 4443ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4444ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 444567cde344SMiao Xie 44462ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 444749b25e05SJeff Mahoney EXTENT_DIRTY); 44482ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, 44490b246afaSJeff Mahoney fs_info->pinned_extents); 445049b25e05SJeff Mahoney 44514a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 44524a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 445349b25e05SJeff Mahoney } 445449b25e05SJeff Mahoney 44552ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4456acce952bSliubo { 4457acce952bSliubo struct btrfs_transaction *t; 4458acce952bSliubo 44590b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4460acce952bSliubo 44610b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 44620b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 44630b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4464724e2315SJosef Bacik struct btrfs_transaction, list); 4465724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 44669b64f57dSElena Reshetova refcount_inc(&t->use_count); 44670b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 44682ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4469724e2315SJosef Bacik btrfs_put_transaction(t); 44700b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4471724e2315SJosef Bacik continue; 4472724e2315SJosef Bacik } 44730b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4474724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 44750b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4476724e2315SJosef Bacik /* 4477724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4478724e2315SJosef Bacik * handle open currently for this transaction. 4479724e2315SJosef Bacik */ 4480724e2315SJosef Bacik wait_event(t->writer_wait, 4481724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4482724e2315SJosef Bacik } else { 44830b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4484724e2315SJosef Bacik } 44852ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4486724e2315SJosef Bacik 44870b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 44880b246afaSJeff Mahoney if (t == fs_info->running_transaction) 44890b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4490724e2315SJosef Bacik list_del_init(&t->list); 44910b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4492a4abeea4SJosef Bacik 4493724e2315SJosef Bacik btrfs_put_transaction(t); 44942ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 44950b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4496724e2315SJosef Bacik } 44970b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 44980b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4499ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4500ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 45012ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 45020b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 45030b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4504acce952bSliubo 4505acce952bSliubo return 0; 4506acce952bSliubo } 4507acce952bSliubo 4508e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 45094d53dddbSDavid Sterba /* mandatory callbacks */ 45100b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 45114d53dddbSDavid Sterba .readpage_end_io_hook = btree_readpage_end_io_hook, 451220a7db8aSDavid Sterba .readpage_io_failed_hook = btree_io_failed_hook, 45134d53dddbSDavid Sterba 45144d53dddbSDavid Sterba /* optional callbacks */ 45150da5468fSChris Mason }; 4516