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 480a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 4810b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 482f188591eSChris Mason while (1) { 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 496a826d6dcSJosef Bacik /* 497a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 498a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 499a826d6dcSJosef Bacik * any less wrong. 500a826d6dcSJosef Bacik */ 501581c1760SQu Wenruo if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags) || 502581c1760SQu Wenruo ret == -EUCLEAN) 503ea466794SJosef Bacik break; 504ea466794SJosef Bacik 5050b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 506f188591eSChris Mason eb->start, eb->len); 5074235298eSChris Mason if (num_copies == 1) 508ea466794SJosef Bacik break; 5094235298eSChris Mason 5105cf1ab56SJosef Bacik if (!failed_mirror) { 5115cf1ab56SJosef Bacik failed = 1; 5125cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 5135cf1ab56SJosef Bacik } 5145cf1ab56SJosef Bacik 515f188591eSChris Mason mirror_num++; 516ea466794SJosef Bacik if (mirror_num == failed_mirror) 517ea466794SJosef Bacik mirror_num++; 518ea466794SJosef Bacik 5194235298eSChris Mason if (mirror_num > num_copies) 520ea466794SJosef Bacik break; 521f188591eSChris Mason } 522ea466794SJosef Bacik 523c0901581SStefan Behrens if (failed && !ret && failed_mirror) 5242ff7e61eSJeff Mahoney repair_eb_io_failure(fs_info, eb, failed_mirror); 525ea466794SJosef Bacik 526ea466794SJosef Bacik return ret; 527f188591eSChris Mason } 52819c00ddcSChris Mason 529d352ac68SChris Mason /* 530d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 531d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 532d352ac68SChris Mason */ 533d397712bSChris Mason 53401d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 53519c00ddcSChris Mason { 5364eee4fa4SMiao Xie u64 start = page_offset(page); 53719c00ddcSChris Mason u64 found_start; 53819c00ddcSChris Mason struct extent_buffer *eb; 539f188591eSChris Mason 5404f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5414f2de97aSJosef Bacik if (page != eb->pages[0]) 5424f2de97aSJosef Bacik return 0; 5430f805531SAlex Lyakas 54419c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 5450f805531SAlex Lyakas /* 5460f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5470f805531SAlex Lyakas * It is useful to know what went wrong. 5480f805531SAlex Lyakas */ 5490f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5500f805531SAlex Lyakas return -EUCLEAN; 5510f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5520f805531SAlex Lyakas return -EUCLEAN; 5530f805531SAlex Lyakas 5540f805531SAlex Lyakas ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, 5550f805531SAlex Lyakas btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); 5560f805531SAlex Lyakas 5578bd98f0eSAlex Lyakas return csum_tree_block(fs_info, eb, 0); 55819c00ddcSChris Mason } 55919c00ddcSChris Mason 56001d58472SDaniel Dressler static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, 5612b82032cSYan Zheng struct extent_buffer *eb) 5622b82032cSYan Zheng { 56301d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 56444880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 5652b82032cSYan Zheng int ret = 1; 5662b82032cSYan Zheng 5670a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5682b82032cSYan Zheng while (fs_devices) { 5692b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5702b82032cSYan Zheng ret = 0; 5712b82032cSYan Zheng break; 5722b82032cSYan Zheng } 5732b82032cSYan Zheng fs_devices = fs_devices->seed; 5742b82032cSYan Zheng } 5752b82032cSYan Zheng return ret; 5762b82032cSYan Zheng } 5772b82032cSYan Zheng 578facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 579facc8a22SMiao Xie u64 phy_offset, struct page *page, 580facc8a22SMiao Xie u64 start, u64 end, int mirror) 581ce9adaa5SChris Mason { 582ce9adaa5SChris Mason u64 found_start; 583ce9adaa5SChris Mason int found_level; 584ce9adaa5SChris Mason struct extent_buffer *eb; 585ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 58602873e43SZhao Lei struct btrfs_fs_info *fs_info = root->fs_info; 587f188591eSChris Mason int ret = 0; 588727011e0SChris Mason int reads_done; 589ce9adaa5SChris Mason 590ce9adaa5SChris Mason if (!page->private) 591ce9adaa5SChris Mason goto out; 592d397712bSChris Mason 5934f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 594d397712bSChris Mason 5950b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 5960b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 5970b32f4bbSJosef Bacik */ 5980b32f4bbSJosef Bacik extent_buffer_get(eb); 5990b32f4bbSJosef Bacik 6000b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 601727011e0SChris Mason if (!reads_done) 602727011e0SChris Mason goto err; 603f188591eSChris Mason 6045cf1ab56SJosef Bacik eb->read_mirror = mirror; 605656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 606ea466794SJosef Bacik ret = -EIO; 607ea466794SJosef Bacik goto err; 608ea466794SJosef Bacik } 609ea466794SJosef Bacik 610ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 611727011e0SChris Mason if (found_start != eb->start) { 612893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 613893bf4b1SSu Yue eb->start, found_start); 614f188591eSChris Mason ret = -EIO; 615ce9adaa5SChris Mason goto err; 616ce9adaa5SChris Mason } 61702873e43SZhao Lei if (check_tree_block_fsid(fs_info, eb)) { 61802873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 61994647322SDavid Sterba eb->start); 6201259ab75SChris Mason ret = -EIO; 6211259ab75SChris Mason goto err; 6221259ab75SChris Mason } 623ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6241c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 625893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 626893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 6271c24c3ceSJosef Bacik ret = -EIO; 6281c24c3ceSJosef Bacik goto err; 6291c24c3ceSJosef Bacik } 630ce9adaa5SChris Mason 63185d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 63285d4e461SChris Mason eb, found_level); 6334008c04aSChris Mason 63402873e43SZhao Lei ret = csum_tree_block(fs_info, eb, 1); 6358bd98f0eSAlex Lyakas if (ret) 636a826d6dcSJosef Bacik goto err; 637a826d6dcSJosef Bacik 638a826d6dcSJosef Bacik /* 639a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 640a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 641a826d6dcSJosef Bacik * return -EIO. 642a826d6dcSJosef Bacik */ 6432f659546SQu Wenruo if (found_level == 0 && btrfs_check_leaf_full(fs_info, eb)) { 644a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 645a826d6dcSJosef Bacik ret = -EIO; 646a826d6dcSJosef Bacik } 647ce9adaa5SChris Mason 6482f659546SQu Wenruo if (found_level > 0 && btrfs_check_node(fs_info, eb)) 649053ab70fSLiu Bo ret = -EIO; 650053ab70fSLiu Bo 6510b32f4bbSJosef Bacik if (!ret) 6520b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 653ce9adaa5SChris Mason err: 65479fb65a1SJosef Bacik if (reads_done && 65579fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 656d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6574bb31e92SArne Jansen 65853b381b3SDavid Woodhouse if (ret) { 65953b381b3SDavid Woodhouse /* 66053b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 66153b381b3SDavid Woodhouse * again, we have to make sure it has something 66253b381b3SDavid Woodhouse * to decrement 66353b381b3SDavid Woodhouse */ 66453b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6650b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 66653b381b3SDavid Woodhouse } 6670b32f4bbSJosef Bacik free_extent_buffer(eb); 668ce9adaa5SChris Mason out: 669f188591eSChris Mason return ret; 670ce9adaa5SChris Mason } 671ce9adaa5SChris Mason 672ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6734bb31e92SArne Jansen { 6744bb31e92SArne Jansen struct extent_buffer *eb; 6754bb31e92SArne Jansen 6764f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 677656f30dbSFilipe Manana set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); 6785cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 67953b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 680ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 681d48d71aaSDavid Sterba btree_readahead_hook(eb, -EIO); 6824bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6834bb31e92SArne Jansen } 6844bb31e92SArne Jansen 6854246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 686ce9adaa5SChris Mason { 68797eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 688ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6899e0af237SLiu Bo struct btrfs_workqueue *wq; 6909e0af237SLiu Bo btrfs_work_func_t func; 691ce9adaa5SChris Mason 692ce9adaa5SChris Mason fs_info = end_io_wq->info; 6934e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 694d20f7043SChris Mason 69537226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 6969e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 6979e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 6989e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 6999e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 7009e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 7019e0af237SLiu Bo func = btrfs_freespace_write_helper; 7029e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7039e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7049e0af237SLiu Bo func = btrfs_endio_raid56_helper; 705d20f7043SChris Mason } else { 7069e0af237SLiu Bo wq = fs_info->endio_write_workers; 7079e0af237SLiu Bo func = btrfs_endio_write_helper; 708d20f7043SChris Mason } 7099e0af237SLiu Bo } else { 7108b110e39SMiao Xie if (unlikely(end_io_wq->metadata == 7118b110e39SMiao Xie BTRFS_WQ_ENDIO_DIO_REPAIR)) { 7128b110e39SMiao Xie wq = fs_info->endio_repair_workers; 7138b110e39SMiao Xie func = btrfs_endio_repair_helper; 7148b110e39SMiao Xie } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7159e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7169e0af237SLiu Bo func = btrfs_endio_raid56_helper; 7179e0af237SLiu Bo } else if (end_io_wq->metadata) { 7189e0af237SLiu Bo wq = fs_info->endio_meta_workers; 7199e0af237SLiu Bo func = btrfs_endio_meta_helper; 7209e0af237SLiu Bo } else { 7219e0af237SLiu Bo wq = fs_info->endio_workers; 7229e0af237SLiu Bo func = btrfs_endio_helper; 7239e0af237SLiu Bo } 7249e0af237SLiu Bo } 7259e0af237SLiu Bo 7269e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 7279e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 728ce9adaa5SChris Mason } 729ce9adaa5SChris Mason 7304e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 731bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7320b86a832SChris Mason { 73397eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7348b110e39SMiao Xie 73597eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 736ce9adaa5SChris Mason if (!end_io_wq) 7374e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 738ce9adaa5SChris Mason 739ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 740ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 74122c59948SChris Mason end_io_wq->info = info; 7424e4cbee9SChristoph Hellwig end_io_wq->status = 0; 743ce9adaa5SChris Mason end_io_wq->bio = bio; 74422c59948SChris Mason end_io_wq->metadata = metadata; 745ce9adaa5SChris Mason 746ce9adaa5SChris Mason bio->bi_private = end_io_wq; 747ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 74822c59948SChris Mason return 0; 74922c59948SChris Mason } 75022c59948SChris Mason 7514a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7524a69a410SChris Mason { 7534a69a410SChris Mason struct async_submit_bio *async; 7544e4cbee9SChristoph Hellwig blk_status_t ret; 7554a69a410SChris Mason 7564a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 757c6100a4bSJosef Bacik ret = async->submit_bio_start(async->private_data, async->bio, 758eaf25d93SChris Mason async->bio_offset); 75979787eaaSJeff Mahoney if (ret) 7604e4cbee9SChristoph Hellwig async->status = ret; 7614a69a410SChris Mason } 7624a69a410SChris Mason 7634a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7648b712842SChris Mason { 7658b712842SChris Mason struct async_submit_bio *async; 7668b712842SChris Mason 7678b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7684854ddd0SChris Mason 769bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7704e4cbee9SChristoph Hellwig if (async->status) { 7714e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7724246a0b6SChristoph Hellwig bio_endio(async->bio); 77379787eaaSJeff Mahoney return; 77479787eaaSJeff Mahoney } 77579787eaaSJeff Mahoney 776e288c080SDavid Sterba btrfs_submit_bio_done(async->private_data, async->bio, async->mirror_num); 7774a69a410SChris Mason } 7784a69a410SChris Mason 7794a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7804a69a410SChris Mason { 7814a69a410SChris Mason struct async_submit_bio *async; 7824a69a410SChris Mason 7834a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7848b712842SChris Mason kfree(async); 7858b712842SChris Mason } 7868b712842SChris Mason 7878c27cb35SLinus Torvalds blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 788c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 789c6100a4bSJosef Bacik u64 bio_offset, void *private_data, 790e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 79144b8bd7eSChris Mason { 79244b8bd7eSChris Mason struct async_submit_bio *async; 79344b8bd7eSChris Mason 79444b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 79544b8bd7eSChris Mason if (!async) 7964e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 79744b8bd7eSChris Mason 798c6100a4bSJosef Bacik async->private_data = private_data; 79944b8bd7eSChris Mason async->bio = bio; 80044b8bd7eSChris Mason async->mirror_num = mirror_num; 8014a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8024a69a410SChris Mason 8039e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 8045cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 8054a69a410SChris Mason 806eaf25d93SChris Mason async->bio_offset = bio_offset; 8078c8bee1dSChris Mason 8084e4cbee9SChristoph Hellwig async->status = 0; 80979787eaaSJeff Mahoney 81067f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 8115cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 812d313d7a3SChris Mason 8135cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 81444b8bd7eSChris Mason return 0; 81544b8bd7eSChris Mason } 81644b8bd7eSChris Mason 8174e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 818ce3ed71aSChris Mason { 8192c30c71bSKent Overstreet struct bio_vec *bvec; 820ce3ed71aSChris Mason struct btrfs_root *root; 8212c30c71bSKent Overstreet int i, ret = 0; 822ce3ed71aSChris Mason 823c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8242c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 825ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 82601d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 82779787eaaSJeff Mahoney if (ret) 82879787eaaSJeff Mahoney break; 829ce3ed71aSChris Mason } 8302c30c71bSKent Overstreet 8314e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 832ce3ed71aSChris Mason } 833ce3ed71aSChris Mason 834d0ee3934SDavid Sterba static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, 835eaf25d93SChris Mason u64 bio_offset) 83622c59948SChris Mason { 8378b712842SChris Mason /* 8388b712842SChris Mason * when we're called for a write, we're already in the async 8395443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8408b712842SChris Mason */ 84179787eaaSJeff Mahoney return btree_csum_one_bio(bio); 84222c59948SChris Mason } 84322c59948SChris Mason 84418fdc679SLiu Bo static int check_async_write(struct btrfs_inode *bi) 845de0022b9SJosef Bacik { 8466300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8476300463bSLiu Bo return 0; 848de0022b9SJosef Bacik #ifdef CONFIG_X86 849bc696ca0SBorislav Petkov if (static_cpu_has(X86_FEATURE_XMM4_2)) 850de0022b9SJosef Bacik return 0; 851de0022b9SJosef Bacik #endif 852de0022b9SJosef Bacik return 1; 853de0022b9SJosef Bacik } 854de0022b9SJosef Bacik 8558c27cb35SLinus Torvalds static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, 856eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 857eaf25d93SChris Mason u64 bio_offset) 85844b8bd7eSChris Mason { 859c6100a4bSJosef Bacik struct inode *inode = private_data; 8600b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 86118fdc679SLiu Bo int async = check_async_write(BTRFS_I(inode)); 8624e4cbee9SChristoph Hellwig blk_status_t ret; 863cad321adSChris Mason 86437226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 865cad321adSChris Mason /* 866cad321adSChris Mason * called for a read, do the setup so that checksum validation 867cad321adSChris Mason * can happen in the async kernel threads 868cad321adSChris Mason */ 8690b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8700b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8711d4284bdSChris Mason if (ret) 87261891923SStefan Behrens goto out_w_error; 8732ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 874de0022b9SJosef Bacik } else if (!async) { 875de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 876de0022b9SJosef Bacik if (ret) 87761891923SStefan Behrens goto out_w_error; 8782ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 87961891923SStefan Behrens } else { 880cad321adSChris Mason /* 88161891923SStefan Behrens * kthread helpers are used to submit writes so that 88261891923SStefan Behrens * checksumming can happen in parallel across all CPUs 883cad321adSChris Mason */ 884c6100a4bSJosef Bacik ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, 885c6100a4bSJosef Bacik bio_offset, private_data, 886e288c080SDavid Sterba btree_submit_bio_start); 88744b8bd7eSChris Mason } 88844b8bd7eSChris Mason 8894246a0b6SChristoph Hellwig if (ret) 8904246a0b6SChristoph Hellwig goto out_w_error; 8914246a0b6SChristoph Hellwig return 0; 8924246a0b6SChristoph Hellwig 89361891923SStefan Behrens out_w_error: 8944e4cbee9SChristoph Hellwig bio->bi_status = ret; 8954246a0b6SChristoph Hellwig bio_endio(bio); 89661891923SStefan Behrens return ret; 89761891923SStefan Behrens } 89861891923SStefan Behrens 8993dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 900784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 901a6bc32b8SMel Gorman struct page *newpage, struct page *page, 902a6bc32b8SMel Gorman enum migrate_mode mode) 903784b4e29SChris Mason { 904784b4e29SChris Mason /* 905784b4e29SChris Mason * we can't safely write a btree page from here, 906784b4e29SChris Mason * we haven't done the locking hook 907784b4e29SChris Mason */ 908784b4e29SChris Mason if (PageDirty(page)) 909784b4e29SChris Mason return -EAGAIN; 910784b4e29SChris Mason /* 911784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 912784b4e29SChris Mason * We must have no buffers or drop them. 913784b4e29SChris Mason */ 914784b4e29SChris Mason if (page_has_private(page) && 915784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 916784b4e29SChris Mason return -EAGAIN; 917a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 918784b4e29SChris Mason } 9193dd1462eSJan Beulich #endif 920784b4e29SChris Mason 9210da5468fSChris Mason 9220da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9230da5468fSChris Mason struct writeback_control *wbc) 9240da5468fSChris Mason { 925e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 926e2d84521SMiao Xie int ret; 927e2d84521SMiao Xie 928d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 929448d640bSChris Mason 930448d640bSChris Mason if (wbc->for_kupdate) 931448d640bSChris Mason return 0; 932448d640bSChris Mason 933e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 934b9473439SChris Mason /* this is a bit racy, but that's ok */ 935d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 936d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 937d814a491SEthan Lien fs_info->dirty_metadata_batch); 938e2d84521SMiao Xie if (ret < 0) 939793955bcSChris Mason return 0; 940793955bcSChris Mason } 9410b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9420da5468fSChris Mason } 9430da5468fSChris Mason 944b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9455f39d397SChris Mason { 946d1310b2eSChris Mason struct extent_io_tree *tree; 947d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9488ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 9495f39d397SChris Mason } 9505f39d397SChris Mason 95170dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9525f39d397SChris Mason { 95398509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 95498509cfcSChris Mason return 0; 9550c4e538bSDavid Sterba 956f7a52a40SDavid Sterba return try_release_extent_buffer(page); 957d98237b3SChris Mason } 958d98237b3SChris Mason 959d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 960d47992f8SLukas Czerner unsigned int length) 961d98237b3SChris Mason { 962d1310b2eSChris Mason struct extent_io_tree *tree; 963d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9645f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9655f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9669ad6b7bcSChris Mason if (PagePrivate(page)) { 967efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 968efe120a0SFrank Holton "page private not zero on page %llu", 969efe120a0SFrank Holton (unsigned long long)page_offset(page)); 9709ad6b7bcSChris Mason ClearPagePrivate(page); 9719ad6b7bcSChris Mason set_page_private(page, 0); 97209cbfeafSKirill A. Shutemov put_page(page); 9739ad6b7bcSChris Mason } 974d98237b3SChris Mason } 975d98237b3SChris Mason 9760b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9770b32f4bbSJosef Bacik { 978bb146eb2SJosef Bacik #ifdef DEBUG 9790b32f4bbSJosef Bacik struct extent_buffer *eb; 9800b32f4bbSJosef Bacik 9810b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9820b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9830b32f4bbSJosef Bacik BUG_ON(!eb); 9840b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9850b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 9860b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 987bb146eb2SJosef Bacik #endif 9880b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 9890b32f4bbSJosef Bacik } 9900b32f4bbSJosef Bacik 9917f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 992d98237b3SChris Mason .readpage = btree_readpage, 9930da5468fSChris Mason .writepages = btree_writepages, 9945f39d397SChris Mason .releasepage = btree_releasepage, 9955f39d397SChris Mason .invalidatepage = btree_invalidatepage, 9965a92bc88SChris Mason #ifdef CONFIG_MIGRATION 997784b4e29SChris Mason .migratepage = btree_migratepage, 9985a92bc88SChris Mason #endif 9990b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1000d98237b3SChris Mason }; 1001123abc88SChris Mason 10022ff7e61eSJeff Mahoney void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) 1003090d1875SChris Mason { 10045f39d397SChris Mason struct extent_buffer *buf = NULL; 10052ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1006090d1875SChris Mason 10072ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1008c871b0f2SLiu Bo if (IS_ERR(buf)) 10096197d86eSDavid Sterba return; 1010d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 10116af49dbdSDavid Sterba buf, WAIT_NONE, 0); 10125f39d397SChris Mason free_extent_buffer(buf); 1013090d1875SChris Mason } 1014090d1875SChris Mason 10152ff7e61eSJeff Mahoney int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, 1016ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1017ab0fff03SArne Jansen { 1018ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 10192ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1020ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1021ab0fff03SArne Jansen int ret; 1022ab0fff03SArne Jansen 10232ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1024c871b0f2SLiu Bo if (IS_ERR(buf)) 1025ab0fff03SArne Jansen return 0; 1026ab0fff03SArne Jansen 1027ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1028ab0fff03SArne Jansen 10298436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, 10306af49dbdSDavid Sterba mirror_num); 1031ab0fff03SArne Jansen if (ret) { 1032ab0fff03SArne Jansen free_extent_buffer(buf); 1033ab0fff03SArne Jansen return ret; 1034ab0fff03SArne Jansen } 1035ab0fff03SArne Jansen 1036ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1037ab0fff03SArne Jansen free_extent_buffer(buf); 1038ab0fff03SArne Jansen return -EIO; 10390b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1040ab0fff03SArne Jansen *eb = buf; 1041ab0fff03SArne Jansen } else { 1042ab0fff03SArne Jansen free_extent_buffer(buf); 1043ab0fff03SArne Jansen } 1044ab0fff03SArne Jansen return 0; 1045ab0fff03SArne Jansen } 1046ab0fff03SArne Jansen 10472ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10482ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10490308af44SDavid Sterba u64 bytenr) 10500999df54SChris Mason { 10510b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10520b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10530b246afaSJeff Mahoney return alloc_extent_buffer(fs_info, bytenr); 10540999df54SChris Mason } 10550999df54SChris Mason 10560999df54SChris Mason 1057e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1058e02119d5SChris Mason { 1059727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 10608aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1061e02119d5SChris Mason } 1062e02119d5SChris Mason 10633189ff77SJeff Layton void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1064e02119d5SChris Mason { 10653189ff77SJeff Layton filemap_fdatawait_range(buf->pages[0]->mapping, 10668aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1067e02119d5SChris Mason } 1068e02119d5SChris Mason 1069581c1760SQu Wenruo /* 1070581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1071581c1760SQu Wenruo * verification. 1072581c1760SQu Wenruo * 1073581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 1074581c1760SQu Wenruo * @level: expected level, mandatory check 1075581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 1076581c1760SQu Wenruo */ 10772ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 1078581c1760SQu Wenruo u64 parent_transid, int level, 1079581c1760SQu Wenruo struct btrfs_key *first_key) 1080e20d96d6SChris Mason { 10815f39d397SChris Mason struct extent_buffer *buf = NULL; 108219c00ddcSChris Mason int ret; 108319c00ddcSChris Mason 10842ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1085c871b0f2SLiu Bo if (IS_ERR(buf)) 1086c871b0f2SLiu Bo return buf; 1087e4204dedSChris Mason 1088581c1760SQu Wenruo ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid, 1089581c1760SQu Wenruo level, first_key); 10900f0fe8f7SFilipe David Borba Manana if (ret) { 10910f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 109264c043deSLiu Bo return ERR_PTR(ret); 10930f0fe8f7SFilipe David Borba Manana } 10945f39d397SChris Mason return buf; 1095ce9adaa5SChris Mason 1096eb60ceacSChris Mason } 1097eb60ceacSChris Mason 10987c302b49SDavid Sterba void clean_tree_block(struct btrfs_fs_info *fs_info, 10995f39d397SChris Mason struct extent_buffer *buf) 1100ed2ff2cbSChris Mason { 110155c69072SChris Mason if (btrfs_header_generation(buf) == 1102e2d84521SMiao Xie fs_info->running_transaction->transid) { 1103b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1104b4ce94deSChris Mason 1105b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1106104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1107e2d84521SMiao Xie -buf->len, 1108e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1109b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1110b9473439SChris Mason btrfs_set_lock_blocking(buf); 11110b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1112925baeddSChris Mason } 11135f39d397SChris Mason } 1114ed7b63ebSJosef Bacik } 11155f39d397SChris Mason 11168257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11178257b2dcSMiao Xie { 11188257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11198257b2dcSMiao Xie int ret; 11208257b2dcSMiao Xie 11218257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11228257b2dcSMiao Xie if (!writers) 11238257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11248257b2dcSMiao Xie 11258a5a916dSJeff Mahoney ret = percpu_counter_init(&writers->counter, 0, GFP_NOFS); 11268257b2dcSMiao Xie if (ret < 0) { 11278257b2dcSMiao Xie kfree(writers); 11288257b2dcSMiao Xie return ERR_PTR(ret); 11298257b2dcSMiao Xie } 11308257b2dcSMiao Xie 11318257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11328257b2dcSMiao Xie return writers; 11338257b2dcSMiao Xie } 11348257b2dcSMiao Xie 11358257b2dcSMiao Xie static void 11368257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11378257b2dcSMiao Xie { 11388257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 11398257b2dcSMiao Xie kfree(writers); 11408257b2dcSMiao Xie } 11418257b2dcSMiao Xie 1142da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1143e20d96d6SChris Mason u64 objectid) 1144d97e63b6SChris Mason { 11457c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 1146cfaa7295SChris Mason root->node = NULL; 1147a28ec197SChris Mason root->commit_root = NULL; 114827cdeb70SMiao Xie root->state = 0; 1149d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11500b86a832SChris Mason 11510f7d52f4SChris Mason root->last_trans = 0; 115213a8a7c8SYan, Zheng root->highest_objectid = 0; 1153eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1154199c2a9cSMiao Xie root->nr_ordered_extents = 0; 11556bef4d31SEric Paris root->inode_tree = RB_ROOT; 115616cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1157f0486c68SYan, Zheng root->block_rsv = NULL; 11580b86a832SChris Mason 11590b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11605d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1161eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1162eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1163199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1164199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 11652ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11662ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 11675d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1168eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1169199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1170f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11712ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11722ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 11738287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1174a2135011SChris Mason mutex_init(&root->objectid_mutex); 1175e02119d5SChris Mason mutex_init(&root->log_mutex); 117631f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1177573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 11787237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11797237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11807237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11818b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11828b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11837237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11847237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11857237f183SYan Zheng atomic_set(&root->log_writers, 0); 11862ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11870700cea7SElena Reshetova refcount_set(&root->refs, 1); 1188ea14b57fSDavid Sterba atomic_set(&root->will_be_snapshotted, 0); 11898ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 11907237f183SYan Zheng root->log_transid = 0; 1191d1433debSMiao Xie root->log_transid_committed = -1; 1192257c62e1SChris Mason root->last_log_commit = 0; 11937c0260eeSJeff Mahoney if (!dummy) 1194c6100a4bSJosef Bacik extent_io_tree_init(&root->dirty_log_pages, NULL); 1195017e5369SChris Mason 11963768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 11973768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 11986702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 11997c0260eeSJeff Mahoney if (!dummy) 12003f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 120106ea65a3SJosef Bacik else 120206ea65a3SJosef Bacik root->defrag_trans_start = 0; 12034d775673SChris Mason root->root_key.objectid = objectid; 12040ee5dc67SAl Viro root->anon_dev = 0; 12058ea05e3aSAlexander Block 12065f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 12073768f368SChris Mason } 12083768f368SChris Mason 120974e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 121074e4d827SDavid Sterba gfp_t flags) 12116f07e42eSAl Viro { 121274e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12136f07e42eSAl Viro if (root) 12146f07e42eSAl Viro root->fs_info = fs_info; 12156f07e42eSAl Viro return root; 12166f07e42eSAl Viro } 12176f07e42eSAl Viro 121806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 121906ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1220da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 122106ea65a3SJosef Bacik { 122206ea65a3SJosef Bacik struct btrfs_root *root; 122306ea65a3SJosef Bacik 12247c0260eeSJeff Mahoney if (!fs_info) 12257c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12267c0260eeSJeff Mahoney 12277c0260eeSJeff Mahoney root = btrfs_alloc_root(fs_info, GFP_KERNEL); 122806ea65a3SJosef Bacik if (!root) 122906ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1230da17066cSJeff Mahoney 1231b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1232da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 1233faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 123406ea65a3SJosef Bacik 123506ea65a3SJosef Bacik return root; 123606ea65a3SJosef Bacik } 123706ea65a3SJosef Bacik #endif 123806ea65a3SJosef Bacik 123920897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 124020897f5cSArne Jansen struct btrfs_fs_info *fs_info, 124120897f5cSArne Jansen u64 objectid) 124220897f5cSArne Jansen { 124320897f5cSArne Jansen struct extent_buffer *leaf; 124420897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 124520897f5cSArne Jansen struct btrfs_root *root; 124620897f5cSArne Jansen struct btrfs_key key; 124720897f5cSArne Jansen int ret = 0; 124833d85fdaSQu Wenruo uuid_le uuid = NULL_UUID_LE; 124920897f5cSArne Jansen 125074e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_KERNEL); 125120897f5cSArne Jansen if (!root) 125220897f5cSArne Jansen return ERR_PTR(-ENOMEM); 125320897f5cSArne Jansen 1254da17066cSJeff Mahoney __setup_root(root, fs_info, objectid); 125520897f5cSArne Jansen root->root_key.objectid = objectid; 125620897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 125720897f5cSArne Jansen root->root_key.offset = 0; 125820897f5cSArne Jansen 12594d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 126020897f5cSArne Jansen if (IS_ERR(leaf)) { 126120897f5cSArne Jansen ret = PTR_ERR(leaf); 12621dd05682STsutomu Itoh leaf = NULL; 126320897f5cSArne Jansen goto fail; 126420897f5cSArne Jansen } 126520897f5cSArne Jansen 126620897f5cSArne Jansen root->node = leaf; 126720897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 126820897f5cSArne Jansen 126920897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 127027cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 127120897f5cSArne Jansen 127220897f5cSArne Jansen root->root_item.flags = 0; 127320897f5cSArne Jansen root->root_item.byte_limit = 0; 127420897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 127520897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 127620897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 127720897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 127820897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 127920897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 128020897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 128133d85fdaSQu Wenruo if (is_fstree(objectid)) 12826463fe58SStefan Behrens uuid_le_gen(&uuid); 12836463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 128420897f5cSArne Jansen root->root_item.drop_level = 0; 128520897f5cSArne Jansen 128620897f5cSArne Jansen key.objectid = objectid; 128720897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 128820897f5cSArne Jansen key.offset = 0; 128920897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 129020897f5cSArne Jansen if (ret) 129120897f5cSArne Jansen goto fail; 129220897f5cSArne Jansen 129320897f5cSArne Jansen btrfs_tree_unlock(leaf); 129420897f5cSArne Jansen 129520897f5cSArne Jansen return root; 12961dd05682STsutomu Itoh 12971dd05682STsutomu Itoh fail: 12981dd05682STsutomu Itoh if (leaf) { 12991dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 130059885b39STsutomu Itoh free_extent_buffer(root->commit_root); 13011dd05682STsutomu Itoh free_extent_buffer(leaf); 13021dd05682STsutomu Itoh } 13031dd05682STsutomu Itoh kfree(root); 13041dd05682STsutomu Itoh 13051dd05682STsutomu Itoh return ERR_PTR(ret); 130620897f5cSArne Jansen } 130720897f5cSArne Jansen 13087237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1309e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13100f7d52f4SChris Mason { 13110f7d52f4SChris Mason struct btrfs_root *root; 13127237f183SYan Zheng struct extent_buffer *leaf; 1313e02119d5SChris Mason 131474e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1315e02119d5SChris Mason if (!root) 13167237f183SYan Zheng return ERR_PTR(-ENOMEM); 1317e02119d5SChris Mason 1318da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1319e02119d5SChris Mason 1320e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1321e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1322e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 132327cdeb70SMiao Xie 13247237f183SYan Zheng /* 132527cdeb70SMiao Xie * DON'T set REF_COWS for log trees 132627cdeb70SMiao Xie * 13277237f183SYan Zheng * log trees do not get reference counted because they go away 13287237f183SYan Zheng * before a real commit is actually done. They do store pointers 13297237f183SYan Zheng * to file data extents, and those reference counts still get 13307237f183SYan Zheng * updated (along with back refs to the log tree). 13317237f183SYan Zheng */ 1332e02119d5SChris Mason 13334d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13344d75f8a9SDavid Sterba NULL, 0, 0, 0); 13357237f183SYan Zheng if (IS_ERR(leaf)) { 13367237f183SYan Zheng kfree(root); 13377237f183SYan Zheng return ERR_CAST(leaf); 13387237f183SYan Zheng } 1339e02119d5SChris Mason 13407237f183SYan Zheng root->node = leaf; 1341e02119d5SChris Mason 1342e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1343e02119d5SChris Mason btrfs_tree_unlock(root->node); 13447237f183SYan Zheng return root; 13457237f183SYan Zheng } 13467237f183SYan Zheng 13477237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13487237f183SYan Zheng struct btrfs_fs_info *fs_info) 13497237f183SYan Zheng { 13507237f183SYan Zheng struct btrfs_root *log_root; 13517237f183SYan Zheng 13527237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13537237f183SYan Zheng if (IS_ERR(log_root)) 13547237f183SYan Zheng return PTR_ERR(log_root); 13557237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13567237f183SYan Zheng fs_info->log_root_tree = log_root; 13577237f183SYan Zheng return 0; 13587237f183SYan Zheng } 13597237f183SYan Zheng 13607237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13617237f183SYan Zheng struct btrfs_root *root) 13627237f183SYan Zheng { 13630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13647237f183SYan Zheng struct btrfs_root *log_root; 13657237f183SYan Zheng struct btrfs_inode_item *inode_item; 13667237f183SYan Zheng 13670b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13687237f183SYan Zheng if (IS_ERR(log_root)) 13697237f183SYan Zheng return PTR_ERR(log_root); 13707237f183SYan Zheng 13717237f183SYan Zheng log_root->last_trans = trans->transid; 13727237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13737237f183SYan Zheng 13747237f183SYan Zheng inode_item = &log_root->root_item.inode; 13753cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13763cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13773cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1378da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13790b246afaSJeff Mahoney fs_info->nodesize); 13803cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13817237f183SYan Zheng 13825d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13837237f183SYan Zheng 13847237f183SYan Zheng WARN_ON(root->log_root); 13857237f183SYan Zheng root->log_root = log_root; 13867237f183SYan Zheng root->log_transid = 0; 1387d1433debSMiao Xie root->log_transid_committed = -1; 1388257c62e1SChris Mason root->last_log_commit = 0; 1389e02119d5SChris Mason return 0; 1390e02119d5SChris Mason } 1391e02119d5SChris Mason 139235a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1393cb517eabSMiao Xie struct btrfs_key *key) 1394e02119d5SChris Mason { 1395e02119d5SChris Mason struct btrfs_root *root; 1396e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 13970f7d52f4SChris Mason struct btrfs_path *path; 139884234f3aSYan Zheng u64 generation; 1399cb517eabSMiao Xie int ret; 1400581c1760SQu Wenruo int level; 1401cb517eabSMiao Xie 1402cb517eabSMiao Xie path = btrfs_alloc_path(); 1403cb517eabSMiao Xie if (!path) 1404cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 14050f7d52f4SChris Mason 140674e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1407cb517eabSMiao Xie if (!root) { 1408cb517eabSMiao Xie ret = -ENOMEM; 1409cb517eabSMiao Xie goto alloc_fail; 14100f7d52f4SChris Mason } 14110f7d52f4SChris Mason 1412da17066cSJeff Mahoney __setup_root(root, fs_info, key->objectid); 14130f7d52f4SChris Mason 1414cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1415cb517eabSMiao Xie &root->root_item, &root->root_key); 14160f7d52f4SChris Mason if (ret) { 141713a8a7c8SYan, Zheng if (ret > 0) 141813a8a7c8SYan, Zheng ret = -ENOENT; 1419cb517eabSMiao Xie goto find_fail; 14200f7d52f4SChris Mason } 142113a8a7c8SYan, Zheng 142284234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1423581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 14242ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 14252ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 1426581c1760SQu Wenruo generation, level, NULL); 142764c043deSLiu Bo if (IS_ERR(root->node)) { 142864c043deSLiu Bo ret = PTR_ERR(root->node); 1429cb517eabSMiao Xie goto find_fail; 1430cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1431cb517eabSMiao Xie ret = -EIO; 143264c043deSLiu Bo free_extent_buffer(root->node); 143364c043deSLiu Bo goto find_fail; 1434416bc658SJosef Bacik } 14355d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 143613a8a7c8SYan, Zheng out: 1437cb517eabSMiao Xie btrfs_free_path(path); 1438cb517eabSMiao Xie return root; 1439cb517eabSMiao Xie 1440cb517eabSMiao Xie find_fail: 1441cb517eabSMiao Xie kfree(root); 1442cb517eabSMiao Xie alloc_fail: 1443cb517eabSMiao Xie root = ERR_PTR(ret); 1444cb517eabSMiao Xie goto out; 1445cb517eabSMiao Xie } 1446cb517eabSMiao Xie 1447cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1448cb517eabSMiao Xie struct btrfs_key *location) 1449cb517eabSMiao Xie { 1450cb517eabSMiao Xie struct btrfs_root *root; 1451cb517eabSMiao Xie 1452cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1453cb517eabSMiao Xie if (IS_ERR(root)) 1454cb517eabSMiao Xie return root; 1455cb517eabSMiao Xie 1456cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 145727cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 145808fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 145908fe4db1SLi Zefan } 146013a8a7c8SYan, Zheng 14615eda7b5eSChris Mason return root; 14625eda7b5eSChris Mason } 14635eda7b5eSChris Mason 1464cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1465cb517eabSMiao Xie { 1466cb517eabSMiao Xie int ret; 14678257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1468cb517eabSMiao Xie 1469cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1470cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1471cb517eabSMiao Xie GFP_NOFS); 1472cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1473cb517eabSMiao Xie ret = -ENOMEM; 1474cb517eabSMiao Xie goto fail; 1475cb517eabSMiao Xie } 1476cb517eabSMiao Xie 14778257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 14788257b2dcSMiao Xie if (IS_ERR(writers)) { 14798257b2dcSMiao Xie ret = PTR_ERR(writers); 14808257b2dcSMiao Xie goto fail; 14818257b2dcSMiao Xie } 14828257b2dcSMiao Xie root->subv_writers = writers; 14838257b2dcSMiao Xie 1484cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 148557cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 148657cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1487cb517eabSMiao Xie 1488cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1489cb517eabSMiao Xie if (ret) 1490876d2cf1SLiu Bo goto fail; 1491f32e48e9SChandan Rajendra 1492f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1493f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1494f32e48e9SChandan Rajendra &root->highest_objectid); 1495f32e48e9SChandan Rajendra if (ret) { 1496f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1497876d2cf1SLiu Bo goto fail; 1498f32e48e9SChandan Rajendra } 1499f32e48e9SChandan Rajendra 1500f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1501f32e48e9SChandan Rajendra 1502f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1503f32e48e9SChandan Rajendra 1504cb517eabSMiao Xie return 0; 1505cb517eabSMiao Xie fail: 150684db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1507cb517eabSMiao Xie return ret; 1508cb517eabSMiao Xie } 1509cb517eabSMiao Xie 151035bbb97fSJeff Mahoney struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1511cb517eabSMiao Xie u64 root_id) 1512cb517eabSMiao Xie { 1513cb517eabSMiao Xie struct btrfs_root *root; 1514cb517eabSMiao Xie 1515cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1516cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1517cb517eabSMiao Xie (unsigned long)root_id); 1518cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1519cb517eabSMiao Xie return root; 1520cb517eabSMiao Xie } 1521cb517eabSMiao Xie 1522cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1523cb517eabSMiao Xie struct btrfs_root *root) 1524cb517eabSMiao Xie { 1525cb517eabSMiao Xie int ret; 1526cb517eabSMiao Xie 1527e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1528cb517eabSMiao Xie if (ret) 1529cb517eabSMiao Xie return ret; 1530cb517eabSMiao Xie 1531cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1532cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1533cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1534cb517eabSMiao Xie root); 1535cb517eabSMiao Xie if (ret == 0) 153627cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1537cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1538cb517eabSMiao Xie radix_tree_preload_end(); 1539cb517eabSMiao Xie 1540cb517eabSMiao Xie return ret; 1541cb517eabSMiao Xie } 1542cb517eabSMiao Xie 1543c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1544c00869f1SMiao Xie struct btrfs_key *location, 1545c00869f1SMiao Xie bool check_ref) 15465eda7b5eSChris Mason { 15475eda7b5eSChris Mason struct btrfs_root *root; 1548381cf658SDavid Sterba struct btrfs_path *path; 15491d4c08e0SDavid Sterba struct btrfs_key key; 15505eda7b5eSChris Mason int ret; 15515eda7b5eSChris Mason 1552edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1553edbd8d4eSChris Mason return fs_info->tree_root; 1554edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1555edbd8d4eSChris Mason return fs_info->extent_root; 15568f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 15578f18cf13SChris Mason return fs_info->chunk_root; 15588f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 15598f18cf13SChris Mason return fs_info->dev_root; 15600403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 15610403e47eSYan Zheng return fs_info->csum_root; 1562bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1563bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1564bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1565f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1566f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1567f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 156870f6d82eSOmar Sandoval if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 156970f6d82eSOmar Sandoval return fs_info->free_space_root ? fs_info->free_space_root : 157070f6d82eSOmar Sandoval ERR_PTR(-ENOENT); 15714df27c4dSYan, Zheng again: 1572cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 157348475471SStefan Behrens if (root) { 1574c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 157548475471SStefan Behrens return ERR_PTR(-ENOENT); 15765eda7b5eSChris Mason return root; 157748475471SStefan Behrens } 15785eda7b5eSChris Mason 1579cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 15805eda7b5eSChris Mason if (IS_ERR(root)) 15815eda7b5eSChris Mason return root; 15823394e160SChris Mason 1583c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1584d68fc57bSYan, Zheng ret = -ENOENT; 1585d68fc57bSYan, Zheng goto fail; 1586d68fc57bSYan, Zheng } 1587d68fc57bSYan, Zheng 1588cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1589cb517eabSMiao Xie if (ret) 1590cb517eabSMiao Xie goto fail; 1591cb517eabSMiao Xie 1592381cf658SDavid Sterba path = btrfs_alloc_path(); 1593381cf658SDavid Sterba if (!path) { 1594381cf658SDavid Sterba ret = -ENOMEM; 1595381cf658SDavid Sterba goto fail; 1596381cf658SDavid Sterba } 15971d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 15981d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 15991d4c08e0SDavid Sterba key.offset = location->objectid; 16001d4c08e0SDavid Sterba 16011d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1602381cf658SDavid Sterba btrfs_free_path(path); 1603d68fc57bSYan, Zheng if (ret < 0) 1604d68fc57bSYan, Zheng goto fail; 1605d68fc57bSYan, Zheng if (ret == 0) 160627cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1607d68fc57bSYan, Zheng 1608cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16090f7d52f4SChris Mason if (ret) { 16104df27c4dSYan, Zheng if (ret == -EEXIST) { 161184db5ccfSDavid Sterba btrfs_free_fs_root(root); 16124df27c4dSYan, Zheng goto again; 16130f7d52f4SChris Mason } 16144df27c4dSYan, Zheng goto fail; 16154df27c4dSYan, Zheng } 1616edbd8d4eSChris Mason return root; 16174df27c4dSYan, Zheng fail: 161884db5ccfSDavid Sterba btrfs_free_fs_root(root); 16194df27c4dSYan, Zheng return ERR_PTR(ret); 1620edbd8d4eSChris Mason } 1621edbd8d4eSChris Mason 162204160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 162304160088SChris Mason { 162404160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 162504160088SChris Mason int ret = 0; 162604160088SChris Mason struct btrfs_device *device; 162704160088SChris Mason struct backing_dev_info *bdi; 1628b7967db7SChris Mason 16291f78160cSXiao Guangrong rcu_read_lock(); 16301f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1631dfe25020SChris Mason if (!device->bdev) 1632dfe25020SChris Mason continue; 1633efa7c9f9SJan Kara bdi = device->bdev->bd_bdi; 1634ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 163504160088SChris Mason ret = 1; 163604160088SChris Mason break; 163704160088SChris Mason } 163804160088SChris Mason } 16391f78160cSXiao Guangrong rcu_read_unlock(); 164004160088SChris Mason return ret; 164104160088SChris Mason } 164204160088SChris Mason 16438b712842SChris Mason /* 16448b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16458b712842SChris Mason * functions. This is where read checksum verification actually happens 16468b712842SChris Mason */ 16478b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1648ce9adaa5SChris Mason { 1649ce9adaa5SChris Mason struct bio *bio; 165097eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1651ce9adaa5SChris Mason 165297eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1653ce9adaa5SChris Mason bio = end_io_wq->bio; 16548b712842SChris Mason 16554e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1656ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1657ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 165897eb6b69SDavid Sterba kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 16594246a0b6SChristoph Hellwig bio_endio(bio); 1660ce9adaa5SChris Mason } 166144b8bd7eSChris Mason 1662a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1663a74a4b97SChris Mason { 1664a74a4b97SChris Mason struct btrfs_root *root = arg; 16650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1666d0278245SMiao Xie int again; 1667da288d28SFilipe Manana struct btrfs_trans_handle *trans; 1668a74a4b97SChris Mason 1669a74a4b97SChris Mason do { 1670d0278245SMiao Xie again = 0; 16719d1a2a3aSDavid Sterba 1672d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 16732ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1674d0278245SMiao Xie goto sleep; 1675d0278245SMiao Xie 167690c711abSZygo Blaxell /* 167790c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 167890c711abSZygo Blaxell * before we have finished mounting the filesystem. 167990c711abSZygo Blaxell */ 16800b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 168190c711abSZygo Blaxell goto sleep; 168290c711abSZygo Blaxell 16830b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1684d0278245SMiao Xie goto sleep; 1685d0278245SMiao Xie 1686dc7f370cSMiao Xie /* 1687dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1688dc7f370cSMiao Xie * during the above check and trylock. 1689dc7f370cSMiao Xie */ 16902ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 16910b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1692dc7f370cSMiao Xie goto sleep; 1693dc7f370cSMiao Xie } 1694dc7f370cSMiao Xie 16950b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_delayed_iput_mutex); 16962ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 16970b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); 1698c2d6cb16SFilipe Manana 16999d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 17000b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1701a74a4b97SChris Mason 1702d0278245SMiao Xie /* 170305323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 170405323cd1SMiao Xie * needn't do anything special here. 1705d0278245SMiao Xie */ 17060b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 170767c5e7d4SFilipe Manana 170867c5e7d4SFilipe Manana /* 170967c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 171067c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 171167c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 171267c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 171367c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 171467c5e7d4SFilipe Manana * unused block groups. 171567c5e7d4SFilipe Manana */ 17160b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1717d0278245SMiao Xie sleep: 1718838fe188SJiri Kosina if (!again) { 1719a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 17208929ecfaSYan, Zheng if (!kthread_should_stop()) 1721a74a4b97SChris Mason schedule(); 1722a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1723a74a4b97SChris Mason } 1724a74a4b97SChris Mason } while (!kthread_should_stop()); 1725da288d28SFilipe Manana 1726da288d28SFilipe Manana /* 1727da288d28SFilipe Manana * Transaction kthread is stopped before us and wakes us up. 1728da288d28SFilipe Manana * However we might have started a new transaction and COWed some 1729da288d28SFilipe Manana * tree blocks when deleting unused block groups for example. So 1730da288d28SFilipe Manana * make sure we commit the transaction we started to have a clean 1731da288d28SFilipe Manana * shutdown when evicting the btree inode - if it has dirty pages 1732da288d28SFilipe Manana * when we do the final iput() on it, eviction will trigger a 1733da288d28SFilipe Manana * writeback for it which will fail with null pointer dereferences 1734da288d28SFilipe Manana * since work queues and other resources were already released and 1735da288d28SFilipe Manana * destroyed by the time the iput/eviction/writeback is made. 1736da288d28SFilipe Manana */ 1737da288d28SFilipe Manana trans = btrfs_attach_transaction(root); 1738da288d28SFilipe Manana if (IS_ERR(trans)) { 1739da288d28SFilipe Manana if (PTR_ERR(trans) != -ENOENT) 17400b246afaSJeff Mahoney btrfs_err(fs_info, 1741da288d28SFilipe Manana "cleaner transaction attach returned %ld", 1742da288d28SFilipe Manana PTR_ERR(trans)); 1743da288d28SFilipe Manana } else { 1744da288d28SFilipe Manana int ret; 1745da288d28SFilipe Manana 17463a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 1747da288d28SFilipe Manana if (ret) 17480b246afaSJeff Mahoney btrfs_err(fs_info, 1749da288d28SFilipe Manana "cleaner open transaction commit returned %d", 1750da288d28SFilipe Manana ret); 1751da288d28SFilipe Manana } 1752da288d28SFilipe Manana 1753a74a4b97SChris Mason return 0; 1754a74a4b97SChris Mason } 1755a74a4b97SChris Mason 1756a74a4b97SChris Mason static int transaction_kthread(void *arg) 1757a74a4b97SChris Mason { 1758a74a4b97SChris Mason struct btrfs_root *root = arg; 17590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1760a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1761a74a4b97SChris Mason struct btrfs_transaction *cur; 17628929ecfaSYan, Zheng u64 transid; 1763a944442cSAllen Pais time64_t now; 1764a74a4b97SChris Mason unsigned long delay; 1765914b2007SJan Kara bool cannot_commit; 1766a74a4b97SChris Mason 1767a74a4b97SChris Mason do { 1768914b2007SJan Kara cannot_commit = false; 17690b246afaSJeff Mahoney delay = HZ * fs_info->commit_interval; 17700b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1771a74a4b97SChris Mason 17720b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 17730b246afaSJeff Mahoney cur = fs_info->running_transaction; 1774a74a4b97SChris Mason if (!cur) { 17750b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1776a74a4b97SChris Mason goto sleep; 1777a74a4b97SChris Mason } 177831153d81SYan Zheng 1779afd48513SArnd Bergmann now = ktime_get_seconds(); 17804a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 1781a514d638SQu Wenruo !test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) && 17828b87dc17SDavid Sterba (now < cur->start_time || 17830b246afaSJeff Mahoney now - cur->start_time < fs_info->commit_interval)) { 17840b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1785a74a4b97SChris Mason delay = HZ * 5; 1786a74a4b97SChris Mason goto sleep; 1787a74a4b97SChris Mason } 17888929ecfaSYan, Zheng transid = cur->transid; 17890b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 179056bec294SChris Mason 179179787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1792354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1793914b2007SJan Kara if (IS_ERR(trans)) { 1794354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1795914b2007SJan Kara cannot_commit = true; 179679787eaaSJeff Mahoney goto sleep; 1797914b2007SJan Kara } 17988929ecfaSYan, Zheng if (transid == trans->transid) { 17993a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 18008929ecfaSYan, Zheng } else { 18013a45bb20SJeff Mahoney btrfs_end_transaction(trans); 18028929ecfaSYan, Zheng } 1803a74a4b97SChris Mason sleep: 18040b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 18050b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1806a74a4b97SChris Mason 18074e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18080b246afaSJeff Mahoney &fs_info->fs_state))) 18092ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 18108929ecfaSYan, Zheng if (!kthread_should_stop() && 18110b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1812914b2007SJan Kara cannot_commit)) 1813bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1814a74a4b97SChris Mason } while (!kthread_should_stop()); 1815a74a4b97SChris Mason return 0; 1816a74a4b97SChris Mason } 1817a74a4b97SChris Mason 1818af31f5e5SChris Mason /* 1819af31f5e5SChris Mason * this will find the highest generation in the array of 1820af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1821af31f5e5SChris Mason * or -1 if we can't find anything. 1822af31f5e5SChris Mason * 1823af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1824af31f5e5SChris Mason * generation of the latest root in the array with the generation 1825af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1826af31f5e5SChris Mason */ 1827af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1828af31f5e5SChris Mason { 1829af31f5e5SChris Mason u64 cur; 1830af31f5e5SChris Mason int newest_index = -1; 1831af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1832af31f5e5SChris Mason int i; 1833af31f5e5SChris Mason 1834af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1835af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1836af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1837af31f5e5SChris Mason if (cur == newest_gen) 1838af31f5e5SChris Mason newest_index = i; 1839af31f5e5SChris Mason } 1840af31f5e5SChris Mason 1841af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1842af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1843af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1844af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1845af31f5e5SChris Mason if (cur == newest_gen) 1846af31f5e5SChris Mason newest_index = 0; 1847af31f5e5SChris Mason } 1848af31f5e5SChris Mason return newest_index; 1849af31f5e5SChris Mason } 1850af31f5e5SChris Mason 1851af31f5e5SChris Mason 1852af31f5e5SChris Mason /* 1853af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1854af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1855af31f5e5SChris Mason * field in the fs_info struct 1856af31f5e5SChris Mason */ 1857af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1858af31f5e5SChris Mason u64 newest_gen) 1859af31f5e5SChris Mason { 1860af31f5e5SChris Mason int newest_index = -1; 1861af31f5e5SChris Mason 1862af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1863af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1864af31f5e5SChris Mason if (newest_index == -1) { 1865af31f5e5SChris Mason info->backup_root_index = 0; 1866af31f5e5SChris Mason } else { 1867af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1868af31f5e5SChris Mason } 1869af31f5e5SChris Mason } 1870af31f5e5SChris Mason 1871af31f5e5SChris Mason /* 1872af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1873af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1874af31f5e5SChris Mason * done 1875af31f5e5SChris Mason */ 1876af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1877af31f5e5SChris Mason { 1878af31f5e5SChris Mason int next_backup; 1879af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1880af31f5e5SChris Mason int last_backup; 1881af31f5e5SChris Mason 1882af31f5e5SChris Mason next_backup = info->backup_root_index; 1883af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1884af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1885af31f5e5SChris Mason 1886af31f5e5SChris Mason /* 1887af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1888af31f5e5SChris Mason * this happens only at umount 1889af31f5e5SChris Mason */ 1890af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1891af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1892af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1893af31f5e5SChris Mason next_backup = last_backup; 1894af31f5e5SChris Mason 1895af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1896af31f5e5SChris Mason 1897af31f5e5SChris Mason /* 1898af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1899af31f5e5SChris Mason * regardless of which ones we use today 1900af31f5e5SChris Mason */ 1901af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1902af31f5e5SChris Mason 1903af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1904af31f5e5SChris Mason 1905af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1906af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1907af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1908af31f5e5SChris Mason 1909af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1910af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1911af31f5e5SChris Mason 1912af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1913af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1914af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1915af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1916af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1917af31f5e5SChris Mason 1918af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1919af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1920af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1921af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1922af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1923af31f5e5SChris Mason 19247c7e82a7SChris Mason /* 19257c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19267c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19277c7e82a7SChris Mason */ 19287c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19297c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19307c7e82a7SChris Mason info->fs_root->node->start); 1931af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1932af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1933af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1934af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19357c7e82a7SChris Mason } 1936af31f5e5SChris Mason 1937af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1938af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1939af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1940af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1941af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1942af31f5e5SChris Mason 1943af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1944af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1945af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1946af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1947af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1948af31f5e5SChris Mason 1949af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1950af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1951af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1952af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1953af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1954af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1955af31f5e5SChris Mason 1956af31f5e5SChris Mason /* 1957af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1958af31f5e5SChris Mason * for the next commit 1959af31f5e5SChris Mason */ 1960af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1961af31f5e5SChris Mason &info->super_for_commit->super_roots, 1962af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1963af31f5e5SChris Mason } 1964af31f5e5SChris Mason 1965af31f5e5SChris Mason /* 1966af31f5e5SChris Mason * this copies info out of the root backup array and back into 1967af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 1968af31f5e5SChris Mason * the array, so you send it the number of backups you've already 1969af31f5e5SChris Mason * tried and the last backup index you used. 1970af31f5e5SChris Mason * 1971af31f5e5SChris Mason * this returns -1 when it has tried all the backups 1972af31f5e5SChris Mason */ 1973af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 1974af31f5e5SChris Mason struct btrfs_super_block *super, 1975af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 1976af31f5e5SChris Mason { 1977af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1978af31f5e5SChris Mason int newest = *backup_index; 1979af31f5e5SChris Mason 1980af31f5e5SChris Mason if (*num_backups_tried == 0) { 1981af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 1982af31f5e5SChris Mason 1983af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 1984af31f5e5SChris Mason if (newest == -1) 1985af31f5e5SChris Mason return -1; 1986af31f5e5SChris Mason 1987af31f5e5SChris Mason *backup_index = newest; 1988af31f5e5SChris Mason *num_backups_tried = 1; 1989af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 1990af31f5e5SChris Mason /* we've tried all the backups, all done */ 1991af31f5e5SChris Mason return -1; 1992af31f5e5SChris Mason } else { 1993af31f5e5SChris Mason /* jump to the next oldest backup */ 1994af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 1995af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1996af31f5e5SChris Mason *backup_index = newest; 1997af31f5e5SChris Mason *num_backups_tried += 1; 1998af31f5e5SChris Mason } 1999af31f5e5SChris Mason root_backup = super->super_roots + newest; 2000af31f5e5SChris Mason 2001af31f5e5SChris Mason btrfs_set_super_generation(super, 2002af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2003af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2004af31f5e5SChris Mason btrfs_set_super_root_level(super, 2005af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2006af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2007af31f5e5SChris Mason 2008af31f5e5SChris Mason /* 2009af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2010af31f5e5SChris Mason * need a fsck 2011af31f5e5SChris Mason */ 2012af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2013af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2014af31f5e5SChris Mason return 0; 2015af31f5e5SChris Mason } 2016af31f5e5SChris Mason 20177abadb64SLiu Bo /* helper to cleanup workers */ 20187abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20197abadb64SLiu Bo { 2020dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2021afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20225cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2023fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2024fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 20258b110e39SMiao Xie btrfs_destroy_workqueue(fs_info->endio_repair_workers); 2026d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2027fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2028fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2029a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 20305b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2031e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2032736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2033a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2034fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2035a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 2036a9b9477dSFilipe Manana /* 2037a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2038a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2039a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2040a9b9477dSFilipe Manana */ 2041a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2042a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 20437abadb64SLiu Bo } 20447abadb64SLiu Bo 20452e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20462e9f5954SRashika { 20472e9f5954SRashika if (root) { 20482e9f5954SRashika free_extent_buffer(root->node); 20492e9f5954SRashika free_extent_buffer(root->commit_root); 20502e9f5954SRashika root->node = NULL; 20512e9f5954SRashika root->commit_root = NULL; 20522e9f5954SRashika } 20532e9f5954SRashika } 20542e9f5954SRashika 2055af31f5e5SChris Mason /* helper to cleanup tree roots */ 2056af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2057af31f5e5SChris Mason { 20582e9f5954SRashika free_root_extent_buffers(info->tree_root); 2059655b09feSJosef Bacik 20602e9f5954SRashika free_root_extent_buffers(info->dev_root); 20612e9f5954SRashika free_root_extent_buffers(info->extent_root); 20622e9f5954SRashika free_root_extent_buffers(info->csum_root); 20632e9f5954SRashika free_root_extent_buffers(info->quota_root); 20642e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20652e9f5954SRashika if (chunk_root) 20662e9f5954SRashika free_root_extent_buffers(info->chunk_root); 206770f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2068af31f5e5SChris Mason } 2069af31f5e5SChris Mason 2070faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2071171f6537SJosef Bacik { 2072171f6537SJosef Bacik int ret; 2073171f6537SJosef Bacik struct btrfs_root *gang[8]; 2074171f6537SJosef Bacik int i; 2075171f6537SJosef Bacik 2076171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2077171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2078171f6537SJosef Bacik struct btrfs_root, root_list); 2079171f6537SJosef Bacik list_del(&gang[0]->root_list); 2080171f6537SJosef Bacik 208127cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2082cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2083171f6537SJosef Bacik } else { 2084171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2085171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2086b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2087171f6537SJosef Bacik } 2088171f6537SJosef Bacik } 2089171f6537SJosef Bacik 2090171f6537SJosef Bacik while (1) { 2091171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2092171f6537SJosef Bacik (void **)gang, 0, 2093171f6537SJosef Bacik ARRAY_SIZE(gang)); 2094171f6537SJosef Bacik if (!ret) 2095171f6537SJosef Bacik break; 2096171f6537SJosef Bacik for (i = 0; i < ret; i++) 2097cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2098171f6537SJosef Bacik } 20991a4319ccSLiu Bo 21001a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 21011a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 21022ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 21031a4319ccSLiu Bo } 2104171f6537SJosef Bacik } 2105af31f5e5SChris Mason 2106638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2107638aa7edSEric Sandeen { 2108638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2109638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2110638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2111638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2112638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2113638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2114638aa7edSEric Sandeen fs_info->scrub_workers_refcnt = 0; 2115638aa7edSEric Sandeen } 2116638aa7edSEric Sandeen 2117779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2118779a65a4SEric Sandeen { 2119779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2120779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2121779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2122779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2123779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2124779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2125779a65a4SEric Sandeen } 2126779a65a4SEric Sandeen 21276bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2128f37938e0SEric Sandeen { 21292ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 21302ff7e61eSJeff Mahoney 21312ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 21322ff7e61eSJeff Mahoney set_nlink(inode, 1); 2133f37938e0SEric Sandeen /* 2134f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2135f37938e0SEric Sandeen * the real end of the address space is determined by all of 2136f37938e0SEric Sandeen * the devices in the system 2137f37938e0SEric Sandeen */ 21382ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 21392ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2140f37938e0SEric Sandeen 21412ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 2142c6100a4bSJosef Bacik extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); 21432ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.track_uptodate = 0; 21442ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2145f37938e0SEric Sandeen 21462ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; 2147f37938e0SEric Sandeen 21482ff7e61eSJeff Mahoney BTRFS_I(inode)->root = fs_info->tree_root; 21492ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21502ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21512ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2152f37938e0SEric Sandeen } 2153f37938e0SEric Sandeen 2154ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2155ad618368SEric Sandeen { 2156ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 215773beece9SLiu Bo rwlock_init(&fs_info->dev_replace.lock); 215873beece9SLiu Bo atomic_set(&fs_info->dev_replace.blocking_readers, 0); 21597f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 216073beece9SLiu Bo init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); 2161ad618368SEric Sandeen } 2162ad618368SEric Sandeen 2163f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2164f9e92e40SEric Sandeen { 2165f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2166f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2167f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2168f9e92e40SEric Sandeen fs_info->qgroup_op_tree = RB_ROOT; 2169f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2170f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2171f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2172d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2173f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2174f9e92e40SEric Sandeen } 2175f9e92e40SEric Sandeen 21762a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 21772a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 21782a458198SEric Sandeen { 2179f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 21806f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21812a458198SEric Sandeen 21822a458198SEric Sandeen fs_info->workers = 2183cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2184cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 21852a458198SEric Sandeen 21862a458198SEric Sandeen fs_info->delalloc_workers = 2187cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2188cb001095SJeff Mahoney flags, max_active, 2); 21892a458198SEric Sandeen 21902a458198SEric Sandeen fs_info->flush_workers = 2191cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2192cb001095SJeff Mahoney flags, max_active, 0); 21932a458198SEric Sandeen 21942a458198SEric Sandeen fs_info->caching_workers = 2195cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 21962a458198SEric Sandeen 21972a458198SEric Sandeen /* 21982a458198SEric Sandeen * a higher idle thresh on the submit workers makes it much more 21992a458198SEric Sandeen * likely that bios will be send down in a sane order to the 22002a458198SEric Sandeen * devices 22012a458198SEric Sandeen */ 22022a458198SEric Sandeen fs_info->submit_workers = 2203cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "submit", flags, 22042a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22052a458198SEric Sandeen max_active), 64); 22062a458198SEric Sandeen 22072a458198SEric Sandeen fs_info->fixup_workers = 2208cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 22092a458198SEric Sandeen 22102a458198SEric Sandeen /* 22112a458198SEric Sandeen * endios are largely parallel and should have a very 22122a458198SEric Sandeen * low idle thresh 22132a458198SEric Sandeen */ 22142a458198SEric Sandeen fs_info->endio_workers = 2215cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 22162a458198SEric Sandeen fs_info->endio_meta_workers = 2217cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2218cb001095SJeff Mahoney max_active, 4); 22192a458198SEric Sandeen fs_info->endio_meta_write_workers = 2220cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2221cb001095SJeff Mahoney max_active, 2); 22222a458198SEric Sandeen fs_info->endio_raid56_workers = 2223cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2224cb001095SJeff Mahoney max_active, 4); 22252a458198SEric Sandeen fs_info->endio_repair_workers = 2226cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); 22272a458198SEric Sandeen fs_info->rmw_workers = 2228cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 22292a458198SEric Sandeen fs_info->endio_write_workers = 2230cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2231cb001095SJeff Mahoney max_active, 2); 22322a458198SEric Sandeen fs_info->endio_freespace_worker = 2233cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2234cb001095SJeff Mahoney max_active, 0); 22352a458198SEric Sandeen fs_info->delayed_workers = 2236cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2237cb001095SJeff Mahoney max_active, 0); 22382a458198SEric Sandeen fs_info->readahead_workers = 2239cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2240cb001095SJeff Mahoney max_active, 2); 22412a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2242cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 22432a458198SEric Sandeen fs_info->extent_workers = 2244cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "extent-refs", flags, 22452a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22462a458198SEric Sandeen max_active), 8); 22472a458198SEric Sandeen 22482a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 22492a458198SEric Sandeen fs_info->submit_workers && fs_info->flush_workers && 22502a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 22512a458198SEric Sandeen fs_info->endio_meta_write_workers && 22522a458198SEric Sandeen fs_info->endio_repair_workers && 22532a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 22542a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 22552a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 22562a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 22572a458198SEric Sandeen fs_info->extent_workers && 22582a458198SEric Sandeen fs_info->qgroup_rescan_workers)) { 22592a458198SEric Sandeen return -ENOMEM; 22602a458198SEric Sandeen } 22612a458198SEric Sandeen 22622a458198SEric Sandeen return 0; 22632a458198SEric Sandeen } 22642a458198SEric Sandeen 226563443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 226663443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 226763443bf5SEric Sandeen { 226863443bf5SEric Sandeen int ret; 226963443bf5SEric Sandeen struct btrfs_root *log_tree_root; 227063443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 227163443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2272581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 227363443bf5SEric Sandeen 227463443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2275f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 227663443bf5SEric Sandeen return -EIO; 227763443bf5SEric Sandeen } 227863443bf5SEric Sandeen 227974e4d827SDavid Sterba log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 228063443bf5SEric Sandeen if (!log_tree_root) 228163443bf5SEric Sandeen return -ENOMEM; 228263443bf5SEric Sandeen 2283da17066cSJeff Mahoney __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 228463443bf5SEric Sandeen 22852ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 2286581c1760SQu Wenruo fs_info->generation + 1, 2287581c1760SQu Wenruo level, NULL); 228864c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2289f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22900eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 229164c043deSLiu Bo kfree(log_tree_root); 22920eeff236SLiu Bo return ret; 229364c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2294f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 229563443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 229663443bf5SEric Sandeen kfree(log_tree_root); 229763443bf5SEric Sandeen return -EIO; 229863443bf5SEric Sandeen } 229963443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 230063443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 230163443bf5SEric Sandeen if (ret) { 23020b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 230363443bf5SEric Sandeen "Failed to recover log tree"); 230463443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 230563443bf5SEric Sandeen kfree(log_tree_root); 230663443bf5SEric Sandeen return ret; 230763443bf5SEric Sandeen } 230863443bf5SEric Sandeen 2309bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 23106bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 231163443bf5SEric Sandeen if (ret) 231263443bf5SEric Sandeen return ret; 231363443bf5SEric Sandeen } 231463443bf5SEric Sandeen 231563443bf5SEric Sandeen return 0; 231663443bf5SEric Sandeen } 231763443bf5SEric Sandeen 23186bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 23194bbcaa64SEric Sandeen { 23206bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2321a4f3d2c4SDavid Sterba struct btrfs_root *root; 23224bbcaa64SEric Sandeen struct btrfs_key location; 23234bbcaa64SEric Sandeen int ret; 23244bbcaa64SEric Sandeen 23256bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 23266bccf3abSJeff Mahoney 23274bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 23284bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 23294bbcaa64SEric Sandeen location.offset = 0; 23304bbcaa64SEric Sandeen 2331a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2332f50f4353SLiu Bo if (IS_ERR(root)) { 2333f50f4353SLiu Bo ret = PTR_ERR(root); 2334f50f4353SLiu Bo goto out; 2335f50f4353SLiu Bo } 2336a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2337a4f3d2c4SDavid Sterba fs_info->extent_root = root; 23384bbcaa64SEric Sandeen 23394bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2340a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2341f50f4353SLiu Bo if (IS_ERR(root)) { 2342f50f4353SLiu Bo ret = PTR_ERR(root); 2343f50f4353SLiu Bo goto out; 2344f50f4353SLiu Bo } 2345a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2346a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23474bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 23484bbcaa64SEric Sandeen 23494bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2350a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2351f50f4353SLiu Bo if (IS_ERR(root)) { 2352f50f4353SLiu Bo ret = PTR_ERR(root); 2353f50f4353SLiu Bo goto out; 2354f50f4353SLiu Bo } 2355a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2356a4f3d2c4SDavid Sterba fs_info->csum_root = root; 23574bbcaa64SEric Sandeen 23584bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2359a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2360a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2361a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2362afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2363a4f3d2c4SDavid Sterba fs_info->quota_root = root; 23644bbcaa64SEric Sandeen } 23654bbcaa64SEric Sandeen 23664bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2367a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2368a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2369a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 23704bbcaa64SEric Sandeen if (ret != -ENOENT) 2371f50f4353SLiu Bo goto out; 23724bbcaa64SEric Sandeen } else { 2373a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2374a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 23754bbcaa64SEric Sandeen } 23764bbcaa64SEric Sandeen 237770f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 237870f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 237970f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2380f50f4353SLiu Bo if (IS_ERR(root)) { 2381f50f4353SLiu Bo ret = PTR_ERR(root); 2382f50f4353SLiu Bo goto out; 2383f50f4353SLiu Bo } 238470f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 238570f6d82eSOmar Sandoval fs_info->free_space_root = root; 238670f6d82eSOmar Sandoval } 238770f6d82eSOmar Sandoval 23884bbcaa64SEric Sandeen return 0; 2389f50f4353SLiu Bo out: 2390f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2391f50f4353SLiu Bo location.objectid, ret); 2392f50f4353SLiu Bo return ret; 23934bbcaa64SEric Sandeen } 23944bbcaa64SEric Sandeen 2395069ec957SQu Wenruo /* 2396069ec957SQu Wenruo * Real super block validation 2397069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2398069ec957SQu Wenruo * 2399069ec957SQu Wenruo * @sb: super block to check 2400069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2401069ec957SQu Wenruo * 0 the primary (1st) sb 2402069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2403069ec957SQu Wenruo * -1 skip bytenr check 2404069ec957SQu Wenruo */ 2405069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2406069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 240721a852b0SQu Wenruo { 240821a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 240921a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 241021a852b0SQu Wenruo int ret = 0; 241121a852b0SQu Wenruo 241221a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 241321a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 241421a852b0SQu Wenruo ret = -EINVAL; 241521a852b0SQu Wenruo } 241621a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 241721a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 241821a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 241921a852b0SQu Wenruo ret = -EINVAL; 242021a852b0SQu Wenruo } 242121a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 242221a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 242321a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 242421a852b0SQu Wenruo ret = -EINVAL; 242521a852b0SQu Wenruo } 242621a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 242721a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 242821a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 242921a852b0SQu Wenruo ret = -EINVAL; 243021a852b0SQu Wenruo } 243121a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 243221a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 243321a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 243421a852b0SQu Wenruo ret = -EINVAL; 243521a852b0SQu Wenruo } 243621a852b0SQu Wenruo 243721a852b0SQu Wenruo /* 243821a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 243921a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 244021a852b0SQu Wenruo */ 244121a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 244221a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 244321a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 244421a852b0SQu Wenruo ret = -EINVAL; 244521a852b0SQu Wenruo } 244621a852b0SQu Wenruo /* Only PAGE SIZE is supported yet */ 244721a852b0SQu Wenruo if (sectorsize != PAGE_SIZE) { 244821a852b0SQu Wenruo btrfs_err(fs_info, 244921a852b0SQu Wenruo "sectorsize %llu not supported yet, only support %lu", 245021a852b0SQu Wenruo sectorsize, PAGE_SIZE); 245121a852b0SQu Wenruo ret = -EINVAL; 245221a852b0SQu Wenruo } 245321a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 245421a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 245521a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 245621a852b0SQu Wenruo ret = -EINVAL; 245721a852b0SQu Wenruo } 245821a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 245921a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 246021a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 246121a852b0SQu Wenruo ret = -EINVAL; 246221a852b0SQu Wenruo } 246321a852b0SQu Wenruo 246421a852b0SQu Wenruo /* Root alignment check */ 246521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 246621a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 246721a852b0SQu Wenruo btrfs_super_root(sb)); 246821a852b0SQu Wenruo ret = -EINVAL; 246921a852b0SQu Wenruo } 247021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 247121a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 247221a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 247321a852b0SQu Wenruo ret = -EINVAL; 247421a852b0SQu Wenruo } 247521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 247621a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 247721a852b0SQu Wenruo btrfs_super_log_root(sb)); 247821a852b0SQu Wenruo ret = -EINVAL; 247921a852b0SQu Wenruo } 248021a852b0SQu Wenruo 248121a852b0SQu Wenruo if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { 248221a852b0SQu Wenruo btrfs_err(fs_info, 248321a852b0SQu Wenruo "dev_item UUID does not match fsid: %pU != %pU", 248421a852b0SQu Wenruo fs_info->fsid, sb->dev_item.fsid); 248521a852b0SQu Wenruo ret = -EINVAL; 248621a852b0SQu Wenruo } 248721a852b0SQu Wenruo 248821a852b0SQu Wenruo /* 248921a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 249021a852b0SQu Wenruo * done later 249121a852b0SQu Wenruo */ 249221a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 249321a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 249421a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 249521a852b0SQu Wenruo ret = -EINVAL; 249621a852b0SQu Wenruo } 249721a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 249821a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 249921a852b0SQu Wenruo btrfs_super_stripesize(sb)); 250021a852b0SQu Wenruo ret = -EINVAL; 250121a852b0SQu Wenruo } 250221a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 250321a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 250421a852b0SQu Wenruo btrfs_super_num_devices(sb)); 250521a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 250621a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 250721a852b0SQu Wenruo ret = -EINVAL; 250821a852b0SQu Wenruo } 250921a852b0SQu Wenruo 2510069ec957SQu Wenruo if (mirror_num >= 0 && 2511069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 251221a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 251321a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 251421a852b0SQu Wenruo ret = -EINVAL; 251521a852b0SQu Wenruo } 251621a852b0SQu Wenruo 251721a852b0SQu Wenruo /* 251821a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 251921a852b0SQu Wenruo * and one chunk 252021a852b0SQu Wenruo */ 252121a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 252221a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 252321a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 252421a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 252521a852b0SQu Wenruo ret = -EINVAL; 252621a852b0SQu Wenruo } 252721a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 252821a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 252921a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 253021a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 253121a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 253221a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 253321a852b0SQu Wenruo ret = -EINVAL; 253421a852b0SQu Wenruo } 253521a852b0SQu Wenruo 253621a852b0SQu Wenruo /* 253721a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 253821a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 253921a852b0SQu Wenruo */ 254021a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 254121a852b0SQu Wenruo btrfs_warn(fs_info, 254221a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 254321a852b0SQu Wenruo btrfs_super_generation(sb), 254421a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 254521a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 254621a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 254721a852b0SQu Wenruo btrfs_warn(fs_info, 254821a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 254921a852b0SQu Wenruo btrfs_super_generation(sb), 255021a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 255121a852b0SQu Wenruo 255221a852b0SQu Wenruo return ret; 255321a852b0SQu Wenruo } 255421a852b0SQu Wenruo 2555069ec957SQu Wenruo /* 2556069ec957SQu Wenruo * Validation of super block at mount time. 2557069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2558069ec957SQu Wenruo * flags will be skipped. 2559069ec957SQu Wenruo */ 2560069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2561069ec957SQu Wenruo { 2562069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2563069ec957SQu Wenruo } 2564069ec957SQu Wenruo 256575cb857dSQu Wenruo /* 256675cb857dSQu Wenruo * Validation of super block at write time. 256775cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 256875cb857dSQu Wenruo * overwritten soon. 256975cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 257075cb857dSQu Wenruo */ 257175cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 257275cb857dSQu Wenruo struct btrfs_super_block *sb) 257375cb857dSQu Wenruo { 257475cb857dSQu Wenruo int ret; 257575cb857dSQu Wenruo 257675cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 257775cb857dSQu Wenruo if (ret < 0) 257875cb857dSQu Wenruo goto out; 257975cb857dSQu Wenruo if (btrfs_super_csum_type(sb) != BTRFS_CSUM_TYPE_CRC32) { 258075cb857dSQu Wenruo ret = -EUCLEAN; 258175cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 258275cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 258375cb857dSQu Wenruo goto out; 258475cb857dSQu Wenruo } 258575cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 258675cb857dSQu Wenruo ret = -EUCLEAN; 258775cb857dSQu Wenruo btrfs_err(fs_info, 258875cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 258975cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 259075cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 259175cb857dSQu Wenruo goto out; 259275cb857dSQu Wenruo } 259375cb857dSQu Wenruo out: 259475cb857dSQu Wenruo if (ret < 0) 259575cb857dSQu Wenruo btrfs_err(fs_info, 259675cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 259775cb857dSQu Wenruo return ret; 259875cb857dSQu Wenruo } 259975cb857dSQu Wenruo 2600ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2601dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2602dfe25020SChris Mason char *options) 2603eb60ceacSChris Mason { 2604db94535dSChris Mason u32 sectorsize; 2605db94535dSChris Mason u32 nodesize; 260687ee04ebSChris Mason u32 stripesize; 260784234f3aSYan Zheng u64 generation; 2608f2b636e8SJosef Bacik u64 features; 26093de4586cSChris Mason struct btrfs_key location; 2610a061fc8dSChris Mason struct buffer_head *bh; 26114d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2612815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2613f84a8bd6SAl Viro struct btrfs_root *tree_root; 26144d34b278SIlya Dryomov struct btrfs_root *chunk_root; 2615eb60ceacSChris Mason int ret; 2616e58ca020SYan int err = -EINVAL; 2617af31f5e5SChris Mason int num_backups_tried = 0; 2618af31f5e5SChris Mason int backup_index = 0; 26196675df31SOmar Sandoval int clear_free_space_tree = 0; 2620581c1760SQu Wenruo int level; 26214543df7eSChris Mason 262274e4d827SDavid Sterba tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 262374e4d827SDavid Sterba chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 2624cb517eabSMiao Xie if (!tree_root || !chunk_root) { 262539279cc3SChris Mason err = -ENOMEM; 262639279cc3SChris Mason goto fail; 262739279cc3SChris Mason } 262876dda93cSYan, Zheng 262976dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 263076dda93cSYan, Zheng if (ret) { 263176dda93cSYan, Zheng err = ret; 263276dda93cSYan, Zheng goto fail; 263376dda93cSYan, Zheng } 263476dda93cSYan, Zheng 2635908c7f19STejun Heo ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2636e2d84521SMiao Xie if (ret) { 2637e2d84521SMiao Xie err = ret; 26389e11ceeeSJan Kara goto fail_srcu; 2639e2d84521SMiao Xie } 264009cbfeafSKirill A. Shutemov fs_info->dirty_metadata_batch = PAGE_SIZE * 2641e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2642e2d84521SMiao Xie 2643908c7f19STejun Heo ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2644963d678bSMiao Xie if (ret) { 2645963d678bSMiao Xie err = ret; 2646963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2647963d678bSMiao Xie } 2648963d678bSMiao Xie 26497f8d236aSDavid Sterba ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 26507f8d236aSDavid Sterba GFP_KERNEL); 2651c404e0dcSMiao Xie if (ret) { 2652c404e0dcSMiao Xie err = ret; 2653c404e0dcSMiao Xie goto fail_delalloc_bytes; 2654c404e0dcSMiao Xie } 2655c404e0dcSMiao Xie 265676dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2657f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26588fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2659facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 266024bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2661eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 266211833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 266375cb379dSJeff Mahoney INIT_LIST_HEAD(&fs_info->pending_raid_kobjs); 266475cb379dSJeff Mahoney spin_lock_init(&fs_info->pending_raid_kobjs_lock); 2665eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2666a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 266776dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 266824bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26694cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2670f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2671ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2672fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2673f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 267447ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2675f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2676d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 267767c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 26787585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2679573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2680c2d6cb16SFilipe Manana mutex_init(&fs_info->cleaner_delayed_iput_mutex); 2681de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 268219c00ddcSChris Mason 26830b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 26846324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2685f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 268647ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 26870b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 268866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 268966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 269066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 269166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 269266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 269366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 269466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2695771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 26964cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2697fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 26982fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2699fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2700e20d96d6SChris Mason fs_info->sb = sb; 270195ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27029ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27034cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2704a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2705f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27068b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2707f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 270890519d66SArne Jansen /* readahead state */ 2709d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 271090519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2711fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2712c8b97818SChris Mason 2713b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2714b34b086cSChris Mason num_online_cpus() + 2, 8); 27150afbaf8cSChris Mason 2716199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2717199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 271869fe2d75SJosef Bacik 271969fe2d75SJosef Bacik fs_info->btree_inode = new_inode(sb); 272069fe2d75SJosef Bacik if (!fs_info->btree_inode) { 272169fe2d75SJosef Bacik err = -ENOMEM; 272269fe2d75SJosef Bacik goto fail_bio_counter; 272369fe2d75SJosef Bacik } 272469fe2d75SJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 272569fe2d75SJosef Bacik 272616cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 272774e4d827SDavid Sterba GFP_KERNEL); 272816cdcec7SMiao Xie if (!fs_info->delayed_root) { 272916cdcec7SMiao Xie err = -ENOMEM; 273016cdcec7SMiao Xie goto fail_iput; 273116cdcec7SMiao Xie } 273216cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 27333eaa2885SChris Mason 2734638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 273521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 273621adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 273721adbd5cSStefan Behrens #endif 2738779a65a4SEric Sandeen btrfs_init_balance(fs_info); 273921c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2740a2de733cSArne Jansen 27419f6d2510SDavid Sterba sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 27429f6d2510SDavid Sterba sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 2743a061fc8dSChris Mason 27446bccf3abSJeff Mahoney btrfs_init_btree_inode(fs_info); 274539279cc3SChris Mason 2746e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 27476bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2748a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2749925baeddSChris Mason 2750c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[0], NULL); 2751c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[1], NULL); 275211833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2753afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 2754509659cdSChris Mason 27555a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 275679154b1bSChris Mason mutex_init(&fs_info->tree_log_mutex); 2757925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2758a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2759a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27601bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27619e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2762c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 276376dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2764803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2765fa9c0d79SChris Mason 2766ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2767f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2768416ac51dSArne Jansen 2769fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2770fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2771fa9c0d79SChris Mason 2772e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2773f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2774bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27754854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 27763768f368SChris Mason 277704216820SFilipe Manana INIT_LIST_HEAD(&fs_info->pinned_chunks); 277804216820SFilipe Manana 2779da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2780da17066cSJeff Mahoney fs_info->nodesize = 4096; 2781da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2782da17066cSJeff Mahoney fs_info->stripesize = 4096; 2783da17066cSJeff Mahoney 278453b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 278553b381b3SDavid Woodhouse if (ret) { 278683c8266aSDavid Sterba err = ret; 278753b381b3SDavid Woodhouse goto fail_alloc; 278853b381b3SDavid Woodhouse } 278953b381b3SDavid Woodhouse 2790da17066cSJeff Mahoney __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 27917eccb903SChris Mason 27923c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 27931104a885SDavid Sterba 27941104a885SDavid Sterba /* 27951104a885SDavid Sterba * Read super block and check the signature bytes only 27961104a885SDavid Sterba */ 2797a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 279892fc03fbSAnand Jain if (IS_ERR(bh)) { 279992fc03fbSAnand Jain err = PTR_ERR(bh); 280016cdcec7SMiao Xie goto fail_alloc; 280120b45077SDave Young } 280239279cc3SChris Mason 28031104a885SDavid Sterba /* 28041104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 28051104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 28061104a885SDavid Sterba */ 2807ab8d0fc4SJeff Mahoney if (btrfs_check_super_csum(fs_info, bh->b_data)) { 280805135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 28091104a885SDavid Sterba err = -EINVAL; 2810b2acdddfSAnand Jain brelse(bh); 28111104a885SDavid Sterba goto fail_alloc; 28121104a885SDavid Sterba } 28131104a885SDavid Sterba 28141104a885SDavid Sterba /* 28151104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 28161104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 28171104a885SDavid Sterba * the whole block of INFO_SIZE 28181104a885SDavid Sterba */ 28196c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 28206c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 28216c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2822a061fc8dSChris Mason brelse(bh); 28235f39d397SChris Mason 28246c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 28250b86a832SChris Mason 2826069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 28271104a885SDavid Sterba if (ret) { 282805135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 28291104a885SDavid Sterba err = -EINVAL; 28301104a885SDavid Sterba goto fail_alloc; 28311104a885SDavid Sterba } 28321104a885SDavid Sterba 28336c41761fSDavid Sterba disk_super = fs_info->super_copy; 28340f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 283516cdcec7SMiao Xie goto fail_alloc; 28360f7d52f4SChris Mason 2837acce952bSliubo /* check FS state, whether FS is broken. */ 283887533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 283987533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2840acce952bSliubo 284175e7cb7fSLiu Bo /* 2842af31f5e5SChris Mason * run through our array of backup supers and setup 2843af31f5e5SChris Mason * our ring pointer to the oldest one 2844af31f5e5SChris Mason */ 2845af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2846af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2847af31f5e5SChris Mason 2848af31f5e5SChris Mason /* 284975e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 285075e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 285175e7cb7fSLiu Bo */ 285275e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 285375e7cb7fSLiu Bo 28542ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 28552b82032cSYan Zheng if (ret) { 28562b82032cSYan Zheng err = ret; 285716cdcec7SMiao Xie goto fail_alloc; 28582b82032cSYan Zheng } 2859dfe25020SChris Mason 2860f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2861f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2862f2b636e8SJosef Bacik if (features) { 286305135f59SDavid Sterba btrfs_err(fs_info, 286405135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 2865c1c9ff7cSGeert Uytterhoeven features); 2866f2b636e8SJosef Bacik err = -EINVAL; 286716cdcec7SMiao Xie goto fail_alloc; 2868f2b636e8SJosef Bacik } 2869f2b636e8SJosef Bacik 28705d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 28715d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 28720b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 2873a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 28745c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 28755c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 2876727011e0SChris Mason 28773173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 287805135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 28793173a18fSJosef Bacik 2880727011e0SChris Mason /* 2881727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2882727011e0SChris Mason * they are bigger than the page size 2883727011e0SChris Mason */ 288409cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 2885727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 288605135f59SDavid Sterba btrfs_info(fs_info, 288705135f59SDavid Sterba "flagging fs with big metadata feature"); 2888727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2889727011e0SChris Mason } 2890727011e0SChris Mason 2891bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2892bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2893b7f67055SChandan Rajendra stripesize = sectorsize; 2894707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2895963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2896bc3f116fSChris Mason 2897da17066cSJeff Mahoney /* Cache block sizes */ 2898da17066cSJeff Mahoney fs_info->nodesize = nodesize; 2899da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 2900da17066cSJeff Mahoney fs_info->stripesize = stripesize; 2901da17066cSJeff Mahoney 2902bc3f116fSChris Mason /* 2903bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2904bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2905bc3f116fSChris Mason */ 2906bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2907707e8a07SDavid Sterba (sectorsize != nodesize)) { 290805135f59SDavid Sterba btrfs_err(fs_info, 290905135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 291005135f59SDavid Sterba nodesize, sectorsize); 2911bc3f116fSChris Mason goto fail_alloc; 2912bc3f116fSChris Mason } 2913bc3f116fSChris Mason 2914ceda0864SMiao Xie /* 2915ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2916ceda0864SMiao Xie * update the flag. 2917ceda0864SMiao Xie */ 29185d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 29195d4f98a2SYan Zheng 2920f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2921f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2922bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 292305135f59SDavid Sterba btrfs_err(fs_info, 292405135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 2925c1c9ff7cSGeert Uytterhoeven features); 2926f2b636e8SJosef Bacik err = -EINVAL; 292716cdcec7SMiao Xie goto fail_alloc; 2928f2b636e8SJosef Bacik } 292961d92c32SChris Mason 29302a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 29312a458198SEric Sandeen if (ret) { 29322a458198SEric Sandeen err = ret; 29330dc3b84aSJosef Bacik goto fail_sb_buffer; 29340dc3b84aSJosef Bacik } 29354543df7eSChris Mason 29369e11ceeeSJan Kara sb->s_bdi->congested_fn = btrfs_congested_fn; 29379e11ceeeSJan Kara sb->s_bdi->congested_data = fs_info; 29389e11ceeeSJan Kara sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 2939d4417e22SNikolay Borisov sb->s_bdi->ra_pages = VM_MAX_READAHEAD * SZ_1K / PAGE_SIZE; 29409e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 29419e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 29424575c9ccSChris Mason 2943a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2944a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2945ee87cf5eSAnand Jain memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE); 2946db94535dSChris Mason 2947925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 29486bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 2949925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 295084eed90fSChris Mason if (ret) { 295105135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 29525d4f98a2SYan Zheng goto fail_sb_buffer; 295384eed90fSChris Mason } 29540b86a832SChris Mason 295584234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 2956581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 29570b86a832SChris Mason 2958da17066cSJeff Mahoney __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 29590b86a832SChris Mason 29602ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 29610b86a832SChris Mason btrfs_super_chunk_root(disk_super), 2962581c1760SQu Wenruo generation, level, NULL); 296364c043deSLiu Bo if (IS_ERR(chunk_root->node) || 296464c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 296505135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 2966e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 2967e5fffbacSchandan free_extent_buffer(chunk_root->node); 296895ab1f64SZhao Lei chunk_root->node = NULL; 2969af31f5e5SChris Mason goto fail_tree_roots; 297083121942SDavid Woodhouse } 29715d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 29725d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 29730b86a832SChris Mason 2974e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2975b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2976e17cade2SChris Mason 29775b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 29782b82032cSYan Zheng if (ret) { 297905135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 2980af31f5e5SChris Mason goto fail_tree_roots; 29812b82032cSYan Zheng } 29820b86a832SChris Mason 29838dabb742SStefan Behrens /* 29849b99b115SAnand Jain * Keep the devid that is marked to be the target device for the 29859b99b115SAnand Jain * device replace procedure 29868dabb742SStefan Behrens */ 29879b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 0); 2988dfe25020SChris Mason 2989a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 299005135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 2991a6b0d5c8SChris Mason goto fail_tree_roots; 2992a6b0d5c8SChris Mason } 2993a6b0d5c8SChris Mason 2994af31f5e5SChris Mason retry_root_backup: 299584234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 2996581c1760SQu Wenruo level = btrfs_super_root_level(disk_super); 29970b86a832SChris Mason 29982ff7e61eSJeff Mahoney tree_root->node = read_tree_block(fs_info, 2999db94535dSChris Mason btrfs_super_root(disk_super), 3000581c1760SQu Wenruo generation, level, NULL); 300164c043deSLiu Bo if (IS_ERR(tree_root->node) || 300264c043deSLiu Bo !extent_buffer_uptodate(tree_root->node)) { 300305135f59SDavid Sterba btrfs_warn(fs_info, "failed to read tree root"); 3004e5fffbacSchandan if (!IS_ERR(tree_root->node)) 3005e5fffbacSchandan free_extent_buffer(tree_root->node); 300695ab1f64SZhao Lei tree_root->node = NULL; 3007af31f5e5SChris Mason goto recovery_tree_root; 300883121942SDavid Woodhouse } 3009af31f5e5SChris Mason 30105d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 30115d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 301269e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 3013db94535dSChris Mason 3014f32e48e9SChandan Rajendra mutex_lock(&tree_root->objectid_mutex); 3015f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(tree_root, 3016f32e48e9SChandan Rajendra &tree_root->highest_objectid); 3017f32e48e9SChandan Rajendra if (ret) { 3018f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 3019f32e48e9SChandan Rajendra goto recovery_tree_root; 3020f32e48e9SChandan Rajendra } 3021f32e48e9SChandan Rajendra 3022f32e48e9SChandan Rajendra ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 3023f32e48e9SChandan Rajendra 3024f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 3025f32e48e9SChandan Rajendra 30266bccf3abSJeff Mahoney ret = btrfs_read_roots(fs_info); 30274bbcaa64SEric Sandeen if (ret) 3028af31f5e5SChris Mason goto recovery_tree_root; 3029f7a81ea4SStefan Behrens 30308929ecfaSYan, Zheng fs_info->generation = generation; 30318929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 30328929ecfaSYan, Zheng 3033cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3034cf90d884SQu Wenruo if (ret) { 3035cf90d884SQu Wenruo btrfs_err(fs_info, 3036cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3037cf90d884SQu Wenruo ret); 3038cf90d884SQu Wenruo goto fail_block_groups; 3039cf90d884SQu Wenruo } 304068310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 304168310a5eSIlya Dryomov if (ret) { 304205135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 304368310a5eSIlya Dryomov goto fail_block_groups; 304468310a5eSIlya Dryomov } 304568310a5eSIlya Dryomov 3046733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3047733f4fbbSStefan Behrens if (ret) { 304805135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3049733f4fbbSStefan Behrens goto fail_block_groups; 3050733f4fbbSStefan Behrens } 3051733f4fbbSStefan Behrens 30528dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 30538dabb742SStefan Behrens if (ret) { 305405135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 30558dabb742SStefan Behrens goto fail_block_groups; 30568dabb742SStefan Behrens } 30578dabb742SStefan Behrens 30589b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 1); 30598dabb742SStefan Behrens 3060b7c35e81SAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices, NULL); 3061b7c35e81SAnand Jain if (ret) { 306205135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 306305135f59SDavid Sterba ret); 3064b7c35e81SAnand Jain goto fail_block_groups; 3065b7c35e81SAnand Jain } 3066b7c35e81SAnand Jain 3067b7c35e81SAnand Jain ret = btrfs_sysfs_add_device(fs_devices); 3068b7c35e81SAnand Jain if (ret) { 306905135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs device interface: %d", 307005135f59SDavid Sterba ret); 3071b7c35e81SAnand Jain goto fail_fsdev_sysfs; 3072b7c35e81SAnand Jain } 3073b7c35e81SAnand Jain 307496f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 30755ac1d209SJeff Mahoney if (ret) { 307605135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3077b7c35e81SAnand Jain goto fail_fsdev_sysfs; 30785ac1d209SJeff Mahoney } 30795ac1d209SJeff Mahoney 3080c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3081c59021f8Sliubo if (ret) { 308205135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 30832365dd3cSAnand Jain goto fail_sysfs; 3084c59021f8Sliubo } 3085c59021f8Sliubo 30865b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 30871b1d1f66SJosef Bacik if (ret) { 308805135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 30892365dd3cSAnand Jain goto fail_sysfs; 30901b1d1f66SJosef Bacik } 30914330e183SQu Wenruo 30926528b99dSAnand Jain if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { 309305135f59SDavid Sterba btrfs_warn(fs_info, 30944330e183SQu Wenruo "writeable mount is not allowed due to too many missing devices"); 30952365dd3cSAnand Jain goto fail_sysfs; 3096292fd7fcSStefan Behrens } 30979078a3e1SChris Mason 3098a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3099a74a4b97SChris Mason "btrfs-cleaner"); 310057506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 31012365dd3cSAnand Jain goto fail_sysfs; 3102a74a4b97SChris Mason 3103a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3104a74a4b97SChris Mason tree_root, 3105a74a4b97SChris Mason "btrfs-transaction"); 310657506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 31073f157a2fSChris Mason goto fail_cleaner; 3108a74a4b97SChris Mason 3109583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3110c289811cSChris Mason !fs_info->fs_devices->rotating) { 3111583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3112c289811cSChris Mason } 3113c289811cSChris Mason 3114572d9ab7SDavid Sterba /* 311501327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3116572d9ab7SDavid Sterba * not have to wait for transaction commit 3117572d9ab7SDavid Sterba */ 3118572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 31193818aea2SQu Wenruo 312021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 31210b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 31222ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 31230b246afaSJeff Mahoney btrfs_test_opt(fs_info, 312421adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 312521adbd5cSStefan Behrens 1 : 0, 312621adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 312721adbd5cSStefan Behrens if (ret) 312805135f59SDavid Sterba btrfs_warn(fs_info, 312905135f59SDavid Sterba "failed to initialize integrity check module: %d", 313005135f59SDavid Sterba ret); 313121adbd5cSStefan Behrens } 313221adbd5cSStefan Behrens #endif 3133bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3134bcef60f2SArne Jansen if (ret) 3135bcef60f2SArne Jansen goto fail_trans_kthread; 313621adbd5cSStefan Behrens 3137fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3138fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3139fd708b81SJosef Bacik 314096da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 314196da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 31420b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 314363443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 314479787eaaSJeff Mahoney if (ret) { 314563443bf5SEric Sandeen err = ret; 314628c16cbbSWang Shilong goto fail_qgroup; 3147e556ce2cSYan Zheng } 3148e02119d5SChris Mason } 31491a40e23bSZheng Yan 31506bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 315179787eaaSJeff Mahoney if (ret) 315228c16cbbSWang Shilong goto fail_qgroup; 315376dda93cSYan, Zheng 3154bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 3155d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 315644c44af2SIlya Dryomov if (ret) 315728c16cbbSWang Shilong goto fail_qgroup; 315890c711abSZygo Blaxell 315990c711abSZygo Blaxell mutex_lock(&fs_info->cleaner_mutex); 31605d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 316190c711abSZygo Blaxell mutex_unlock(&fs_info->cleaner_mutex); 3162d7ce5843SMiao Xie if (ret < 0) { 316305135f59SDavid Sterba btrfs_warn(fs_info, "failed to recover relocation: %d", 316405135f59SDavid Sterba ret); 3165d7ce5843SMiao Xie err = -EINVAL; 3166bcef60f2SArne Jansen goto fail_qgroup; 3167d7ce5843SMiao Xie } 31687c2ca468SChris Mason } 31691a40e23bSZheng Yan 31703de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 31713de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 3172cb517eabSMiao Xie location.offset = 0; 31733de4586cSChris Mason 31743de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 31753140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 31763140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3177f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3178bcef60f2SArne Jansen goto fail_qgroup; 31793140c9a3SDan Carpenter } 3180c289811cSChris Mason 3181bc98a42cSDavid Howells if (sb_rdonly(sb)) 31822b6ba629SIlya Dryomov return 0; 31832b6ba629SIlya Dryomov 3184f8d468a1SOmar Sandoval if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 3185f8d468a1SOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 31866675df31SOmar Sandoval clear_free_space_tree = 1; 31876675df31SOmar Sandoval } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 31886675df31SOmar Sandoval !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 31896675df31SOmar Sandoval btrfs_warn(fs_info, "free space tree is invalid"); 31906675df31SOmar Sandoval clear_free_space_tree = 1; 31916675df31SOmar Sandoval } 31926675df31SOmar Sandoval 31936675df31SOmar Sandoval if (clear_free_space_tree) { 3194f8d468a1SOmar Sandoval btrfs_info(fs_info, "clearing free space tree"); 3195f8d468a1SOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 3196f8d468a1SOmar Sandoval if (ret) { 3197f8d468a1SOmar Sandoval btrfs_warn(fs_info, 3198f8d468a1SOmar Sandoval "failed to clear free space tree: %d", ret); 31996bccf3abSJeff Mahoney close_ctree(fs_info); 3200f8d468a1SOmar Sandoval return ret; 3201f8d468a1SOmar Sandoval } 3202f8d468a1SOmar Sandoval } 3203f8d468a1SOmar Sandoval 32040b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 3205511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 320605135f59SDavid Sterba btrfs_info(fs_info, "creating free space tree"); 3207511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3208511711afSChris Mason if (ret) { 320905135f59SDavid Sterba btrfs_warn(fs_info, 321005135f59SDavid Sterba "failed to create free space tree: %d", ret); 32116bccf3abSJeff Mahoney close_ctree(fs_info); 3212511711afSChris Mason return ret; 3213511711afSChris Mason } 3214511711afSChris Mason } 3215511711afSChris Mason 3216e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 32172b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 32182b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 32192b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 32206bccf3abSJeff Mahoney close_ctree(fs_info); 32212b6ba629SIlya Dryomov return ret; 32222b6ba629SIlya Dryomov } 3223e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 322459641015SIlya Dryomov 32252b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 32262b6ba629SIlya Dryomov if (ret) { 322705135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume balance: %d", ret); 32286bccf3abSJeff Mahoney close_ctree(fs_info); 32292b6ba629SIlya Dryomov return ret; 3230e3acc2a6SJosef Bacik } 3231e3acc2a6SJosef Bacik 32328dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 32338dabb742SStefan Behrens if (ret) { 323405135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 32356bccf3abSJeff Mahoney close_ctree(fs_info); 32368dabb742SStefan Behrens return ret; 32378dabb742SStefan Behrens } 32388dabb742SStefan Behrens 3239b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3240b382a324SJan Schmidt 32414bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 324205135f59SDavid Sterba btrfs_info(fs_info, "creating UUID tree"); 3243f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3244f7a81ea4SStefan Behrens if (ret) { 324505135f59SDavid Sterba btrfs_warn(fs_info, 324605135f59SDavid Sterba "failed to create the UUID tree: %d", ret); 32476bccf3abSJeff Mahoney close_ctree(fs_info); 3248f7a81ea4SStefan Behrens return ret; 3249f7a81ea4SStefan Behrens } 32500b246afaSJeff Mahoney } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 32514bbcaa64SEric Sandeen fs_info->generation != 32524bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 325305135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 325470f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 325570f80175SStefan Behrens if (ret) { 325605135f59SDavid Sterba btrfs_warn(fs_info, 325705135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 32586bccf3abSJeff Mahoney close_ctree(fs_info); 325970f80175SStefan Behrens return ret; 326070f80175SStefan Behrens } 326170f80175SStefan Behrens } else { 3262afcdd129SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 3263f7a81ea4SStefan Behrens } 3264afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 326547ab2a6cSJosef Bacik 32668dcddfa0SQu Wenruo /* 32678dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 32688dcddfa0SQu Wenruo * no need to keep the flag 32698dcddfa0SQu Wenruo */ 32708dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 32718dcddfa0SQu Wenruo 3272ad2b2c80SAl Viro return 0; 327339279cc3SChris Mason 3274bcef60f2SArne Jansen fail_qgroup: 3275bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 32767c2ca468SChris Mason fail_trans_kthread: 32777c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 32782ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3279faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 32803f157a2fSChris Mason fail_cleaner: 3281a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 32827c2ca468SChris Mason 32837c2ca468SChris Mason /* 32847c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 32857c2ca468SChris Mason * kthreads 32867c2ca468SChris Mason */ 32877c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 32887c2ca468SChris Mason 32892365dd3cSAnand Jain fail_sysfs: 32906618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 32912365dd3cSAnand Jain 3292b7c35e81SAnand Jain fail_fsdev_sysfs: 3293b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3294b7c35e81SAnand Jain 32951b1d1f66SJosef Bacik fail_block_groups: 329654067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3297af31f5e5SChris Mason 3298af31f5e5SChris Mason fail_tree_roots: 3299af31f5e5SChris Mason free_root_pointers(fs_info, 1); 33002b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3301af31f5e5SChris Mason 330239279cc3SChris Mason fail_sb_buffer: 33037abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 33045cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 330516cdcec7SMiao Xie fail_alloc: 33064543df7eSChris Mason fail_iput: 3307586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3308586e46e2SIlya Dryomov 33094543df7eSChris Mason iput(fs_info->btree_inode); 3310c404e0dcSMiao Xie fail_bio_counter: 33117f8d236aSDavid Sterba percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 3312963d678bSMiao Xie fail_delalloc_bytes: 3313963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3314e2d84521SMiao Xie fail_dirty_metadata_bytes: 3315e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 331676dda93cSYan, Zheng fail_srcu: 331776dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 33187e662854SQinghuang Feng fail: 331953b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3320586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3321ad2b2c80SAl Viro return err; 3322af31f5e5SChris Mason 3323af31f5e5SChris Mason recovery_tree_root: 33240b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 3325af31f5e5SChris Mason goto fail_tree_roots; 3326af31f5e5SChris Mason 3327af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3328af31f5e5SChris Mason 3329af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3330af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3331af31f5e5SChris Mason 3332af31f5e5SChris Mason /* we can't trust the free space cache either */ 3333af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3334af31f5e5SChris Mason 3335af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3336af31f5e5SChris Mason &num_backups_tried, &backup_index); 3337af31f5e5SChris Mason if (ret == -1) 3338af31f5e5SChris Mason goto fail_block_groups; 3339af31f5e5SChris Mason goto retry_root_backup; 3340eb60ceacSChris Mason } 3341663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3342eb60ceacSChris Mason 3343f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3344f2984462SChris Mason { 3345f2984462SChris Mason if (uptodate) { 3346f2984462SChris Mason set_buffer_uptodate(bh); 3347f2984462SChris Mason } else { 3348442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3349442a4f63SStefan Behrens bh->b_private; 3350442a4f63SStefan Behrens 3351fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3352b14af3b4SDavid Sterba "lost page write due to IO error on %s", 3353606686eeSJosef Bacik rcu_str_deref(device->name)); 335401327610SNicholas D Steeves /* note, we don't set_buffer_write_io_error because we have 33551259ab75SChris Mason * our own ways of dealing with the IO errors 33561259ab75SChris Mason */ 3357f2984462SChris Mason clear_buffer_uptodate(bh); 3358442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3359f2984462SChris Mason } 3360f2984462SChris Mason unlock_buffer(bh); 3361f2984462SChris Mason put_bh(bh); 3362f2984462SChris Mason } 3363f2984462SChris Mason 336429c36d72SAnand Jain int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, 336529c36d72SAnand Jain struct buffer_head **bh_ret) 336629c36d72SAnand Jain { 336729c36d72SAnand Jain struct buffer_head *bh; 336829c36d72SAnand Jain struct btrfs_super_block *super; 336929c36d72SAnand Jain u64 bytenr; 337029c36d72SAnand Jain 337129c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 337229c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 337329c36d72SAnand Jain return -EINVAL; 337429c36d72SAnand Jain 33759f6d2510SDavid Sterba bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); 337629c36d72SAnand Jain /* 337729c36d72SAnand Jain * If we fail to read from the underlying devices, as of now 337829c36d72SAnand Jain * the best option we have is to mark it EIO. 337929c36d72SAnand Jain */ 338029c36d72SAnand Jain if (!bh) 338129c36d72SAnand Jain return -EIO; 338229c36d72SAnand Jain 338329c36d72SAnand Jain super = (struct btrfs_super_block *)bh->b_data; 338429c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 338529c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 338629c36d72SAnand Jain brelse(bh); 338729c36d72SAnand Jain return -EINVAL; 338829c36d72SAnand Jain } 338929c36d72SAnand Jain 339029c36d72SAnand Jain *bh_ret = bh; 339129c36d72SAnand Jain return 0; 339229c36d72SAnand Jain } 339329c36d72SAnand Jain 339429c36d72SAnand Jain 3395a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3396a512bbf8SYan Zheng { 3397a512bbf8SYan Zheng struct buffer_head *bh; 3398a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3399a512bbf8SYan Zheng struct btrfs_super_block *super; 3400a512bbf8SYan Zheng int i; 3401a512bbf8SYan Zheng u64 transid = 0; 340292fc03fbSAnand Jain int ret = -EINVAL; 3403a512bbf8SYan Zheng 3404a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3405a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3406a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3407a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3408a512bbf8SYan Zheng */ 3409a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 341029c36d72SAnand Jain ret = btrfs_read_dev_one_super(bdev, i, &bh); 341129c36d72SAnand Jain if (ret) 3412a512bbf8SYan Zheng continue; 3413a512bbf8SYan Zheng 3414a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3415a512bbf8SYan Zheng 3416a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3417a512bbf8SYan Zheng brelse(latest); 3418a512bbf8SYan Zheng latest = bh; 3419a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3420a512bbf8SYan Zheng } else { 3421a512bbf8SYan Zheng brelse(bh); 3422a512bbf8SYan Zheng } 3423a512bbf8SYan Zheng } 342492fc03fbSAnand Jain 342592fc03fbSAnand Jain if (!latest) 342692fc03fbSAnand Jain return ERR_PTR(ret); 342792fc03fbSAnand Jain 3428a512bbf8SYan Zheng return latest; 3429a512bbf8SYan Zheng } 3430a512bbf8SYan Zheng 34314eedeb75SHisashi Hifumi /* 3432abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3433abbb3b8eSDavid Sterba * buffer heads we write are pinned. 34344eedeb75SHisashi Hifumi * 3435abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3436abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3437abbb3b8eSDavid Sterba * same for write and wait phases. 34384eedeb75SHisashi Hifumi * 3439abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or submission fails. 34404eedeb75SHisashi Hifumi */ 3441a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3442abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3443a512bbf8SYan Zheng { 3444a512bbf8SYan Zheng struct buffer_head *bh; 3445a512bbf8SYan Zheng int i; 3446a512bbf8SYan Zheng int ret; 3447a512bbf8SYan Zheng int errors = 0; 3448a512bbf8SYan Zheng u32 crc; 3449a512bbf8SYan Zheng u64 bytenr; 34501b9e619cSOmar Sandoval int op_flags; 3451a512bbf8SYan Zheng 3452a512bbf8SYan Zheng if (max_mirrors == 0) 3453a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3454a512bbf8SYan Zheng 3455a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3456a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3457935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3458935e5cc9SMiao Xie device->commit_total_bytes) 3459a512bbf8SYan Zheng break; 3460a512bbf8SYan Zheng 3461a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3462a512bbf8SYan Zheng 3463a512bbf8SYan Zheng crc = ~(u32)0; 3464abbb3b8eSDavid Sterba crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, 3465abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3466a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3467a512bbf8SYan Zheng 3468abbb3b8eSDavid Sterba /* One reference for us, and we leave it for the caller */ 34699f6d2510SDavid Sterba bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, 3470a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3471634554dcSJosef Bacik if (!bh) { 3472fb456252SJeff Mahoney btrfs_err(device->fs_info, 3473f14d104dSDavid Sterba "couldn't get super buffer head for bytenr %llu", 3474f14d104dSDavid Sterba bytenr); 3475634554dcSJosef Bacik errors++; 3476634554dcSJosef Bacik continue; 3477634554dcSJosef Bacik } 3478634554dcSJosef Bacik 3479a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3480a512bbf8SYan Zheng 34814eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3482a512bbf8SYan Zheng get_bh(bh); 34834eedeb75SHisashi Hifumi 34844eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3485a512bbf8SYan Zheng lock_buffer(bh); 3486a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3487442a4f63SStefan Behrens bh->b_private = device; 3488a512bbf8SYan Zheng 3489387125fcSChris Mason /* 3490387125fcSChris Mason * we fua the first super. The others we allow 3491387125fcSChris Mason * to go down lazy. 3492387125fcSChris Mason */ 34931b9e619cSOmar Sandoval op_flags = REQ_SYNC | REQ_META | REQ_PRIO; 34941b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 34951b9e619cSOmar Sandoval op_flags |= REQ_FUA; 34961b9e619cSOmar Sandoval ret = btrfsic_submit_bh(REQ_OP_WRITE, op_flags, bh); 34974eedeb75SHisashi Hifumi if (ret) 3498a512bbf8SYan Zheng errors++; 3499a512bbf8SYan Zheng } 3500a512bbf8SYan Zheng return errors < i ? 0 : -1; 3501a512bbf8SYan Zheng } 3502a512bbf8SYan Zheng 3503387125fcSChris Mason /* 3504abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3505abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3506abbb3b8eSDavid Sterba * 3507abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or not marked up to 3508abbb3b8eSDavid Sterba * date. 3509abbb3b8eSDavid Sterba */ 3510abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3511abbb3b8eSDavid Sterba { 3512abbb3b8eSDavid Sterba struct buffer_head *bh; 3513abbb3b8eSDavid Sterba int i; 3514abbb3b8eSDavid Sterba int errors = 0; 3515b6a535faSHoward McLauchlan bool primary_failed = false; 3516abbb3b8eSDavid Sterba u64 bytenr; 3517abbb3b8eSDavid Sterba 3518abbb3b8eSDavid Sterba if (max_mirrors == 0) 3519abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3520abbb3b8eSDavid Sterba 3521abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3522abbb3b8eSDavid Sterba bytenr = btrfs_sb_offset(i); 3523abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3524abbb3b8eSDavid Sterba device->commit_total_bytes) 3525abbb3b8eSDavid Sterba break; 3526abbb3b8eSDavid Sterba 35279f6d2510SDavid Sterba bh = __find_get_block(device->bdev, 35289f6d2510SDavid Sterba bytenr / BTRFS_BDEV_BLOCKSIZE, 3529abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE); 3530abbb3b8eSDavid Sterba if (!bh) { 3531abbb3b8eSDavid Sterba errors++; 3532b6a535faSHoward McLauchlan if (i == 0) 3533b6a535faSHoward McLauchlan primary_failed = true; 3534abbb3b8eSDavid Sterba continue; 3535abbb3b8eSDavid Sterba } 3536abbb3b8eSDavid Sterba wait_on_buffer(bh); 3537b6a535faSHoward McLauchlan if (!buffer_uptodate(bh)) { 3538abbb3b8eSDavid Sterba errors++; 3539b6a535faSHoward McLauchlan if (i == 0) 3540b6a535faSHoward McLauchlan primary_failed = true; 3541b6a535faSHoward McLauchlan } 3542abbb3b8eSDavid Sterba 3543abbb3b8eSDavid Sterba /* drop our reference */ 3544abbb3b8eSDavid Sterba brelse(bh); 3545abbb3b8eSDavid Sterba 3546abbb3b8eSDavid Sterba /* drop the reference from the writing run */ 3547abbb3b8eSDavid Sterba brelse(bh); 3548abbb3b8eSDavid Sterba } 3549abbb3b8eSDavid Sterba 3550b6a535faSHoward McLauchlan /* log error, force error return */ 3551b6a535faSHoward McLauchlan if (primary_failed) { 3552b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3553b6a535faSHoward McLauchlan device->devid); 3554b6a535faSHoward McLauchlan return -1; 3555b6a535faSHoward McLauchlan } 3556b6a535faSHoward McLauchlan 3557abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3558abbb3b8eSDavid Sterba } 3559abbb3b8eSDavid Sterba 3560abbb3b8eSDavid Sterba /* 3561387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3562387125fcSChris Mason * for the barrier when it is done 3563387125fcSChris Mason */ 35644246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3565387125fcSChris Mason { 3566387125fcSChris Mason complete(bio->bi_private); 3567387125fcSChris Mason } 3568387125fcSChris Mason 3569387125fcSChris Mason /* 35704fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 35714fc6441aSAnand Jain * done in the waiting counterpart. 3572387125fcSChris Mason */ 35734fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3574387125fcSChris Mason { 3575c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3576e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3577387125fcSChris Mason 3578c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 35794fc6441aSAnand Jain return; 3580387125fcSChris Mason 3581e0ae9994SDavid Sterba bio_reset(bio); 3582387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 358374d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 35848d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3585387125fcSChris Mason init_completion(&device->flush_wait); 3586387125fcSChris Mason bio->bi_private = &device->flush_wait; 3587387125fcSChris Mason 358843a01111SLu Fengqi btrfsic_submit_bio(bio); 35891c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 35904fc6441aSAnand Jain } 3591387125fcSChris Mason 35924fc6441aSAnand Jain /* 35934fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 35944fc6441aSAnand Jain */ 35958c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 35964fc6441aSAnand Jain { 35974fc6441aSAnand Jain struct bio *bio = device->flush_bio; 35984fc6441aSAnand Jain 35991c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 360058efbc9fSOmar Sandoval return BLK_STS_OK; 36014fc6441aSAnand Jain 36021c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 36032980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 36044fc6441aSAnand Jain 36058c27cb35SLinus Torvalds return bio->bi_status; 3606387125fcSChris Mason } 3607387125fcSChris Mason 3608d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3609401b41e5SAnand Jain { 36106528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 3611401b41e5SAnand Jain return -EIO; 3612387125fcSChris Mason return 0; 3613387125fcSChris Mason } 3614387125fcSChris Mason 3615387125fcSChris Mason /* 3616387125fcSChris Mason * send an empty flush down to each device in parallel, 3617387125fcSChris Mason * then wait for them 3618387125fcSChris Mason */ 3619387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3620387125fcSChris Mason { 3621387125fcSChris Mason struct list_head *head; 3622387125fcSChris Mason struct btrfs_device *dev; 36235af3e8ccSStefan Behrens int errors_wait = 0; 36244e4cbee9SChristoph Hellwig blk_status_t ret; 3625387125fcSChris Mason 36261538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3627387125fcSChris Mason /* send down all the barriers */ 3628387125fcSChris Mason head = &info->fs_devices->devices; 36291538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3630e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3631f88ba6a2SHidetoshi Seto continue; 3632cea7c8bfSAnand Jain if (!dev->bdev) 3633387125fcSChris Mason continue; 3634e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3635ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3636387125fcSChris Mason continue; 3637387125fcSChris Mason 36384fc6441aSAnand Jain write_dev_flush(dev); 363958efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3640387125fcSChris Mason } 3641387125fcSChris Mason 3642387125fcSChris Mason /* wait for all the barriers */ 36431538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3644e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3645f88ba6a2SHidetoshi Seto continue; 3646387125fcSChris Mason if (!dev->bdev) { 36475af3e8ccSStefan Behrens errors_wait++; 3648387125fcSChris Mason continue; 3649387125fcSChris Mason } 3650e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3651ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3652387125fcSChris Mason continue; 3653387125fcSChris Mason 36544fc6441aSAnand Jain ret = wait_dev_flush(dev); 3655401b41e5SAnand Jain if (ret) { 3656401b41e5SAnand Jain dev->last_flush_error = ret; 365766b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 365866b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 36595af3e8ccSStefan Behrens errors_wait++; 3660387125fcSChris Mason } 3661401b41e5SAnand Jain } 3662401b41e5SAnand Jain 3663cea7c8bfSAnand Jain if (errors_wait) { 3664401b41e5SAnand Jain /* 3665401b41e5SAnand Jain * At some point we need the status of all disks 3666401b41e5SAnand Jain * to arrive at the volume status. So error checking 3667401b41e5SAnand Jain * is being pushed to a separate loop. 3668401b41e5SAnand Jain */ 3669d10b82feSQu Wenruo return check_barrier_error(info); 3670401b41e5SAnand Jain } 3671387125fcSChris Mason return 0; 3672387125fcSChris Mason } 3673387125fcSChris Mason 3674943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3675943c6e99SZhao Lei { 36768789f4feSZhao Lei int raid_type; 36778789f4feSZhao Lei int min_tolerated = INT_MAX; 3678943c6e99SZhao Lei 36798789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 36808789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 36818789f4feSZhao Lei min_tolerated = min(min_tolerated, 36828789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 36838789f4feSZhao Lei tolerated_failures); 3684943c6e99SZhao Lei 36858789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 36868789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 36878789f4feSZhao Lei continue; 368841a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 36898789f4feSZhao Lei continue; 36908789f4feSZhao Lei min_tolerated = min(min_tolerated, 36918789f4feSZhao Lei btrfs_raid_array[raid_type]. 36928789f4feSZhao Lei tolerated_failures); 36938789f4feSZhao Lei } 3694943c6e99SZhao Lei 36958789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3696ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 36978789f4feSZhao Lei min_tolerated = 0; 36988789f4feSZhao Lei } 36998789f4feSZhao Lei 37008789f4feSZhao Lei return min_tolerated; 3701943c6e99SZhao Lei } 3702943c6e99SZhao Lei 3703eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3704f2984462SChris Mason { 3705e5e9a520SChris Mason struct list_head *head; 3706f2984462SChris Mason struct btrfs_device *dev; 3707a061fc8dSChris Mason struct btrfs_super_block *sb; 3708f2984462SChris Mason struct btrfs_dev_item *dev_item; 3709f2984462SChris Mason int ret; 3710f2984462SChris Mason int do_barriers; 3711a236aed1SChris Mason int max_errors; 3712a236aed1SChris Mason int total_errors = 0; 3713a061fc8dSChris Mason u64 flags; 3714f2984462SChris Mason 37150b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3716fed3b381SLiu Bo 3717fed3b381SLiu Bo /* 3718fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3719fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3720fed3b381SLiu Bo * been consistent on disk. 3721fed3b381SLiu Bo */ 3722fed3b381SLiu Bo if (max_mirrors == 0) 37230b246afaSJeff Mahoney backup_super_roots(fs_info); 3724f2984462SChris Mason 37250b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3726a061fc8dSChris Mason dev_item = &sb->dev_item; 3727e5e9a520SChris Mason 37280b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 37290b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 37300b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3731387125fcSChris Mason 37325af3e8ccSStefan Behrens if (do_barriers) { 37330b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 37345af3e8ccSStefan Behrens if (ret) { 37355af3e8ccSStefan Behrens mutex_unlock( 37360b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 37370b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 37385af3e8ccSStefan Behrens "errors while submitting device barriers."); 37395af3e8ccSStefan Behrens return ret; 37405af3e8ccSStefan Behrens } 37415af3e8ccSStefan Behrens } 3742387125fcSChris Mason 37431538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3744dfe25020SChris Mason if (!dev->bdev) { 3745dfe25020SChris Mason total_errors++; 3746dfe25020SChris Mason continue; 3747dfe25020SChris Mason } 3748e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3749ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3750dfe25020SChris Mason continue; 3751dfe25020SChris Mason 37522b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3753a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3754a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 37557df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3756935e5cc9SMiao Xie dev->commit_total_bytes); 3757ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3758ce7213c7SMiao Xie dev->commit_bytes_used); 3759a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3760a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3761a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3762a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 376344880fdcSAnand Jain memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); 3764a512bbf8SYan Zheng 3765a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3766a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3767f2984462SChris Mason 376875cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 376975cb857dSQu Wenruo if (ret < 0) { 377075cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 377175cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 377275cb857dSQu Wenruo "unexpected superblock corruption detected"); 377375cb857dSQu Wenruo return -EUCLEAN; 377475cb857dSQu Wenruo } 377575cb857dSQu Wenruo 3776abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3777a236aed1SChris Mason if (ret) 3778a236aed1SChris Mason total_errors++; 3779f2984462SChris Mason } 3780a236aed1SChris Mason if (total_errors > max_errors) { 37810b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3782d397712bSChris Mason total_errors); 37830b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 378479787eaaSJeff Mahoney 37859d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 37860b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 37870b246afaSJeff Mahoney "%d errors while writing supers", 37880b246afaSJeff Mahoney total_errors); 37899d565ba4SStefan Behrens return -EIO; 3790a236aed1SChris Mason } 3791f2984462SChris Mason 3792a512bbf8SYan Zheng total_errors = 0; 37931538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3794dfe25020SChris Mason if (!dev->bdev) 3795dfe25020SChris Mason continue; 3796e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3797ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3798dfe25020SChris Mason continue; 3799dfe25020SChris Mason 3800abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3801a512bbf8SYan Zheng if (ret) 38021259ab75SChris Mason total_errors++; 3803f2984462SChris Mason } 38040b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3805a236aed1SChris Mason if (total_errors > max_errors) { 38060b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 38070b246afaSJeff Mahoney "%d errors while writing supers", 38080b246afaSJeff Mahoney total_errors); 380979787eaaSJeff Mahoney return -EIO; 3810a236aed1SChris Mason } 3811f2984462SChris Mason return 0; 3812f2984462SChris Mason } 3813f2984462SChris Mason 3814cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3815cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3816cb517eabSMiao Xie struct btrfs_root *root) 38172619ba1fSChris Mason { 38184df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 38192619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 38202619ba1fSChris Mason (unsigned long)root->root_key.objectid); 38214df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 382276dda93cSYan, Zheng 382376dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 382476dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 382576dda93cSYan, Zheng 38261c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 38273321719eSLiu Bo btrfs_free_log(NULL, root); 38281c1ea4f7SLiu Bo if (root->reloc_root) { 38291c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->node); 38301c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->commit_root); 38311c1ea4f7SLiu Bo btrfs_put_fs_root(root->reloc_root); 38321c1ea4f7SLiu Bo root->reloc_root = NULL; 38331c1ea4f7SLiu Bo } 38341c1ea4f7SLiu Bo } 38353321719eSLiu Bo 3836faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3837581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3838faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3839581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 384084db5ccfSDavid Sterba btrfs_free_fs_root(root); 38414df27c4dSYan, Zheng } 38424df27c4dSYan, Zheng 384384db5ccfSDavid Sterba void btrfs_free_fs_root(struct btrfs_root *root) 38444df27c4dSYan, Zheng { 384557cdc8dbSDavid Sterba iput(root->ino_cache_inode); 38464df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 38470ee5dc67SAl Viro if (root->anon_dev) 38480ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 38498257b2dcSMiao Xie if (root->subv_writers) 38508257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 38515f39d397SChris Mason free_extent_buffer(root->node); 38525f39d397SChris Mason free_extent_buffer(root->commit_root); 3853581bb050SLi Zefan kfree(root->free_ino_ctl); 3854581bb050SLi Zefan kfree(root->free_ino_pinned); 3855b0feb9d9SMiao Xie btrfs_put_fs_root(root); 38562619ba1fSChris Mason } 38572619ba1fSChris Mason 3858c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3859c146afadSYan Zheng { 3860c146afadSYan Zheng u64 root_objectid = 0; 3861c146afadSYan Zheng struct btrfs_root *gang[8]; 386265d33fd7SQu Wenruo int i = 0; 386365d33fd7SQu Wenruo int err = 0; 386465d33fd7SQu Wenruo unsigned int ret = 0; 386565d33fd7SQu Wenruo int index; 3866c146afadSYan Zheng 3867c146afadSYan Zheng while (1) { 386865d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3869c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3870c146afadSYan Zheng (void **)gang, root_objectid, 3871c146afadSYan Zheng ARRAY_SIZE(gang)); 387265d33fd7SQu Wenruo if (!ret) { 387365d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3874c146afadSYan Zheng break; 387565d33fd7SQu Wenruo } 38765d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 387766b4ffd1SJosef Bacik 387865d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 387965d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 388065d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 388165d33fd7SQu Wenruo gang[i] = NULL; 388265d33fd7SQu Wenruo continue; 388365d33fd7SQu Wenruo } 388465d33fd7SQu Wenruo /* grab all the search result for later use */ 388565d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 388665d33fd7SQu Wenruo } 388765d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 388865d33fd7SQu Wenruo 388965d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 389065d33fd7SQu Wenruo if (!gang[i]) 389165d33fd7SQu Wenruo continue; 3892c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 389366b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 389466b4ffd1SJosef Bacik if (err) 389565d33fd7SQu Wenruo break; 389665d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3897c146afadSYan Zheng } 3898c146afadSYan Zheng root_objectid++; 3899c146afadSYan Zheng } 390065d33fd7SQu Wenruo 390165d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 390265d33fd7SQu Wenruo for (; i < ret; i++) { 390365d33fd7SQu Wenruo if (gang[i]) 390465d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 390565d33fd7SQu Wenruo } 390665d33fd7SQu Wenruo return err; 3907c146afadSYan Zheng } 3908c146afadSYan Zheng 39096bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 3910c146afadSYan Zheng { 39116bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 3912c146afadSYan Zheng struct btrfs_trans_handle *trans; 3913c146afadSYan Zheng 39140b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 39152ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 39160b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 39170b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 3918c71bf099SYan, Zheng 3919c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 39200b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 39210b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 3922c71bf099SYan, Zheng 39237a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 39243612b495STsutomu Itoh if (IS_ERR(trans)) 39253612b495STsutomu Itoh return PTR_ERR(trans); 39263a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 3927c146afadSYan Zheng } 3928c146afadSYan Zheng 39296bccf3abSJeff Mahoney void close_ctree(struct btrfs_fs_info *fs_info) 3930eb60ceacSChris Mason { 3931c146afadSYan Zheng int ret; 3932e089f05cSChris Mason 3933afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 3934a2135011SChris Mason 39357343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 3936d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 39377343dd61SJustin Maggard 3938803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3939803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3940803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3941803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3942803b2f54SStefan Behrens 3943837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3944aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3945837d5b6eSIlya Dryomov 39468dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 39478dabb742SStefan Behrens 3948aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 39494cb5300bSChris Mason 39504cb5300bSChris Mason /* wait for any defraggers to finish */ 39514cb5300bSChris Mason wait_event(fs_info->transaction_wait, 39524cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 39534cb5300bSChris Mason 39544cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 395526176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 39564cb5300bSChris Mason 395721c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 395821c7e756SMiao Xie 3959bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 3960e44163e1SJeff Mahoney /* 3961e44163e1SJeff Mahoney * If the cleaner thread is stopped and there are 3962e44163e1SJeff Mahoney * block groups queued for removal, the deletion will be 3963e44163e1SJeff Mahoney * skipped when we quit the cleaner thread. 3964e44163e1SJeff Mahoney */ 39650b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 3966e44163e1SJeff Mahoney 39676bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 3968d397712bSChris Mason if (ret) 396904892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 3970c146afadSYan Zheng } 3971ed2ff2cbSChris Mason 3972af722733SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || 3973af722733SLiu Bo test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) 39742ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 3975acce952bSliubo 3976e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3977e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 39788929ecfaSYan, Zheng 3979e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 3980afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 3981f25784b3SYan Zheng 398204892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 3983fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 3984bcef60f2SArne Jansen 3985963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 398604892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 3987963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3988b0c68f8bSChris Mason } 398931153d81SYan Zheng 39906618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 3991b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 39925ac1d209SJeff Mahoney 3993faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3994d10c5f31SChris Mason 39951a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 39961a4319ccSLiu Bo 3997de348ee0SWang Shilong /* 3998de348ee0SWang Shilong * we must make sure there is not any read request to 3999de348ee0SWang Shilong * submit after we stopping all workers. 4000de348ee0SWang Shilong */ 4001de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 400296192499SJosef Bacik btrfs_stop_all_workers(fs_info); 400396192499SJosef Bacik 40045cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 40055cdd7db6SFilipe Manana 4006afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 400713e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 40089ad6b7bcSChris Mason 400913e6c37bSJosef Bacik iput(fs_info->btree_inode); 4010d6bfde87SChris Mason 401121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 40120b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 40132ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 401421adbd5cSStefan Behrens #endif 401521adbd5cSStefan Behrens 4016dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 40170b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 4018b248a415SChris Mason 4019e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 4020963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 40217f8d236aSDavid Sterba percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 402276dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 40230b86a832SChris Mason 402453b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 4025fd708b81SJosef Bacik btrfs_free_ref_cache(fs_info); 402653b381b3SDavid Woodhouse 402704216820SFilipe Manana while (!list_empty(&fs_info->pinned_chunks)) { 402804216820SFilipe Manana struct extent_map *em; 402904216820SFilipe Manana 403004216820SFilipe Manana em = list_first_entry(&fs_info->pinned_chunks, 403104216820SFilipe Manana struct extent_map, list); 403204216820SFilipe Manana list_del_init(&em->list); 403304216820SFilipe Manana free_extent_map(em); 403404216820SFilipe Manana } 4035eb60ceacSChris Mason } 4036eb60ceacSChris Mason 4037b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4038b9fab919SChris Mason int atomic) 4039ccd467d6SChris Mason { 40401259ab75SChris Mason int ret; 4041727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 40421259ab75SChris Mason 40430b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 40441259ab75SChris Mason if (!ret) 40451259ab75SChris Mason return ret; 40461259ab75SChris Mason 40471259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4048b9fab919SChris Mason parent_transid, atomic); 4049b9fab919SChris Mason if (ret == -EAGAIN) 4050b9fab919SChris Mason return ret; 40511259ab75SChris Mason return !ret; 40525f39d397SChris Mason } 40536702ed49SChris Mason 40545f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 40555f39d397SChris Mason { 40560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info; 405706ea65a3SJosef Bacik struct btrfs_root *root; 40585f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4059b9473439SChris Mason int was_dirty; 4060b4ce94deSChris Mason 406106ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 406206ea65a3SJosef Bacik /* 406306ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 4064b0132a3bSNikolay Borisov * enabled. Normal people shouldn't be using umapped buffers as dirty 406506ea65a3SJosef Bacik * outside of the sanity tests. 406606ea65a3SJosef Bacik */ 4067b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 406806ea65a3SJosef Bacik return; 406906ea65a3SJosef Bacik #endif 407006ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 40710b246afaSJeff Mahoney fs_info = root->fs_info; 4072b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 40730b246afaSJeff Mahoney if (transid != fs_info->generation) 40745d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 40750b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 40760b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4077e2d84521SMiao Xie if (!was_dirty) 4078104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4079e2d84521SMiao Xie buf->len, 40800b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 40811f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 408269fc6cbbSQu Wenruo /* 408369fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 408469fc6cbbSQu Wenruo * but item data not updated. 408569fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 408669fc6cbbSQu Wenruo */ 408769fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 40882f659546SQu Wenruo btrfs_check_leaf_relaxed(fs_info, buf)) { 4089a4f78750SDavid Sterba btrfs_print_leaf(buf); 40901f21ef0aSFilipe Manana ASSERT(0); 40911f21ef0aSFilipe Manana } 40921f21ef0aSFilipe Manana #endif 4093eb60ceacSChris Mason } 4094eb60ceacSChris Mason 40952ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4096b53d3f5dSLiu Bo int flush_delayed) 409735b7e476SChris Mason { 4098188de649SChris Mason /* 4099188de649SChris Mason * looks as though older kernels can get into trouble with 4100188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4101188de649SChris Mason */ 4102e2d84521SMiao Xie int ret; 4103d6bfde87SChris Mason 41046933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4105d6bfde87SChris Mason return; 4106d6bfde87SChris Mason 4107b53d3f5dSLiu Bo if (flush_delayed) 41082ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 410916cdcec7SMiao Xie 4110d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4111d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4112d814a491SEthan Lien fs_info->dirty_metadata_batch); 4113e2d84521SMiao Xie if (ret > 0) { 41140b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 411516cdcec7SMiao Xie } 411616cdcec7SMiao Xie } 411716cdcec7SMiao Xie 41182ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 411916cdcec7SMiao Xie { 41202ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 412135b7e476SChris Mason } 4122b53d3f5dSLiu Bo 41232ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4124b53d3f5dSLiu Bo { 41252ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4126d6bfde87SChris Mason } 41276b80053dSChris Mason 4128581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4129581c1760SQu Wenruo struct btrfs_key *first_key) 41306b80053dSChris Mason { 4131727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 41322ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 41332ff7e61eSJeff Mahoney 4134581c1760SQu Wenruo return btree_read_extent_buffer_pages(fs_info, buf, parent_transid, 4135581c1760SQu Wenruo level, first_key); 41366b80053dSChris Mason } 41370da5468fSChris Mason 41382ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4139acce952bSliubo { 4140fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4141fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4142fe816d0fSNikolay Borisov 41430b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41442ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41450b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4146acce952bSliubo 41470b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41480b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4149acce952bSliubo } 4150acce952bSliubo 4151143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4152acce952bSliubo { 4153acce952bSliubo struct btrfs_ordered_extent *ordered; 4154acce952bSliubo 4155199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4156779880efSJosef Bacik /* 4157779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4158779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4159779880efSJosef Bacik */ 4160199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4161779880efSJosef Bacik root_extent_list) 4162779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4163199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4164199c2a9cSMiao Xie } 4165199c2a9cSMiao Xie 4166199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4167199c2a9cSMiao Xie { 4168199c2a9cSMiao Xie struct btrfs_root *root; 4169199c2a9cSMiao Xie struct list_head splice; 4170199c2a9cSMiao Xie 4171199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4172199c2a9cSMiao Xie 4173199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4174199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4175199c2a9cSMiao Xie while (!list_empty(&splice)) { 4176199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4177199c2a9cSMiao Xie ordered_root); 41781de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 41791de2cfdeSJosef Bacik &fs_info->ordered_roots); 4180199c2a9cSMiao Xie 41812a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4182199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4183199c2a9cSMiao Xie 41842a85d9caSLiu Bo cond_resched(); 41852a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4186199c2a9cSMiao Xie } 4187199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 4188acce952bSliubo } 4189acce952bSliubo 419035a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 41912ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4192acce952bSliubo { 4193acce952bSliubo struct rb_node *node; 4194acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4195acce952bSliubo struct btrfs_delayed_ref_node *ref; 4196acce952bSliubo int ret = 0; 4197acce952bSliubo 4198acce952bSliubo delayed_refs = &trans->delayed_refs; 4199acce952bSliubo 4200acce952bSliubo spin_lock(&delayed_refs->lock); 4201d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4202cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 42030b246afaSJeff Mahoney btrfs_info(fs_info, "delayed_refs has NO entry"); 4204acce952bSliubo return ret; 4205acce952bSliubo } 4206acce952bSliubo 42075c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4208d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 42090e0adbcfSJosef Bacik struct rb_node *n; 4210e78417d1SJosef Bacik bool pin_bytes = false; 4211acce952bSliubo 4212d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4213d7df2c79SJosef Bacik href_node); 4214b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 4215d278850eSJosef Bacik refcount_inc(&head->refs); 421679787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 4217b939d1abSJosef Bacik 4218acce952bSliubo mutex_lock(&head->mutex); 4219b939d1abSJosef Bacik mutex_unlock(&head->mutex); 4220d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4221e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 4222b939d1abSJosef Bacik continue; 4223b939d1abSJosef Bacik } 4224d7df2c79SJosef Bacik spin_lock(&head->lock); 4225e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 42260e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 42270e0adbcfSJosef Bacik ref_node); 4228d7df2c79SJosef Bacik ref->in_tree = 0; 4229e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 42300e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 42311d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 42321d57ee94SWang Xiaoguang list_del(&ref->add_list); 4233d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4234d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4235d7df2c79SJosef Bacik } 423654067ae9SJosef Bacik if (head->must_insert_reserved) 4237e78417d1SJosef Bacik pin_bytes = true; 423878a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4239acce952bSliubo delayed_refs->num_heads--; 4240d7df2c79SJosef Bacik if (head->processing == 0) 4241acce952bSliubo delayed_refs->num_heads_ready--; 4242d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 42435c9d028bSLiu Bo rb_erase_cached(&head->href_node, &delayed_refs->href_root); 4244d278850eSJosef Bacik RB_CLEAR_NODE(&head->href_node); 4245d7df2c79SJosef Bacik spin_unlock(&head->lock); 4246acce952bSliubo spin_unlock(&delayed_refs->lock); 4247e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4248acce952bSliubo 4249d7df2c79SJosef Bacik if (pin_bytes) 4250d278850eSJosef Bacik btrfs_pin_extent(fs_info, head->bytenr, 4251d278850eSJosef Bacik head->num_bytes, 1); 4252d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4253acce952bSliubo cond_resched(); 4254acce952bSliubo spin_lock(&delayed_refs->lock); 4255acce952bSliubo } 4256acce952bSliubo 4257acce952bSliubo spin_unlock(&delayed_refs->lock); 4258acce952bSliubo 4259acce952bSliubo return ret; 4260acce952bSliubo } 4261acce952bSliubo 4262143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4263acce952bSliubo { 4264acce952bSliubo struct btrfs_inode *btrfs_inode; 4265acce952bSliubo struct list_head splice; 4266acce952bSliubo 4267acce952bSliubo INIT_LIST_HEAD(&splice); 4268acce952bSliubo 4269eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4270eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4271acce952bSliubo 4272acce952bSliubo while (!list_empty(&splice)) { 4273fe816d0fSNikolay Borisov struct inode *inode = NULL; 4274eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4275acce952bSliubo delalloc_inodes); 4276fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4277eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4278acce952bSliubo 4279fe816d0fSNikolay Borisov /* 4280fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4281fe816d0fSNikolay Borisov * meanwhile. 4282fe816d0fSNikolay Borisov */ 4283fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4284fe816d0fSNikolay Borisov if (inode) { 4285fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4286fe816d0fSNikolay Borisov iput(inode); 4287fe816d0fSNikolay Borisov } 4288eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4289acce952bSliubo } 4290eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4291eb73c1b7SMiao Xie } 4292eb73c1b7SMiao Xie 4293eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4294eb73c1b7SMiao Xie { 4295eb73c1b7SMiao Xie struct btrfs_root *root; 4296eb73c1b7SMiao Xie struct list_head splice; 4297eb73c1b7SMiao Xie 4298eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4299eb73c1b7SMiao Xie 4300eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4301eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4302eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4303eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4304eb73c1b7SMiao Xie delalloc_root); 4305eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 4306eb73c1b7SMiao Xie BUG_ON(!root); 4307eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4308eb73c1b7SMiao Xie 4309eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 4310eb73c1b7SMiao Xie btrfs_put_fs_root(root); 4311eb73c1b7SMiao Xie 4312eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4313eb73c1b7SMiao Xie } 4314eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4315acce952bSliubo } 4316acce952bSliubo 43172ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4318acce952bSliubo struct extent_io_tree *dirty_pages, 4319acce952bSliubo int mark) 4320acce952bSliubo { 4321acce952bSliubo int ret; 4322acce952bSliubo struct extent_buffer *eb; 4323acce952bSliubo u64 start = 0; 4324acce952bSliubo u64 end; 4325acce952bSliubo 4326acce952bSliubo while (1) { 4327acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4328e6138876SJosef Bacik mark, NULL); 4329acce952bSliubo if (ret) 4330acce952bSliubo break; 4331acce952bSliubo 433291166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4333acce952bSliubo while (start <= end) { 43340b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 43350b246afaSJeff Mahoney start += fs_info->nodesize; 4336fd8b2b61SJosef Bacik if (!eb) 4337acce952bSliubo continue; 4338fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4339acce952bSliubo 4340fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4341fd8b2b61SJosef Bacik &eb->bflags)) 4342fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4343fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4344acce952bSliubo } 4345acce952bSliubo } 4346acce952bSliubo 4347acce952bSliubo return ret; 4348acce952bSliubo } 4349acce952bSliubo 43502ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4351acce952bSliubo struct extent_io_tree *pinned_extents) 4352acce952bSliubo { 4353acce952bSliubo struct extent_io_tree *unpin; 4354acce952bSliubo u64 start; 4355acce952bSliubo u64 end; 4356acce952bSliubo int ret; 4357ed0eaa14SLiu Bo bool loop = true; 4358acce952bSliubo 4359acce952bSliubo unpin = pinned_extents; 4360ed0eaa14SLiu Bo again: 4361acce952bSliubo while (1) { 4362acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4363e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4364acce952bSliubo if (ret) 4365acce952bSliubo break; 4366acce952bSliubo 4367af6f8f60SDavid Sterba clear_extent_dirty(unpin, start, end); 43682ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4369acce952bSliubo cond_resched(); 4370acce952bSliubo } 4371acce952bSliubo 4372ed0eaa14SLiu Bo if (loop) { 43730b246afaSJeff Mahoney if (unpin == &fs_info->freed_extents[0]) 43740b246afaSJeff Mahoney unpin = &fs_info->freed_extents[1]; 4375ed0eaa14SLiu Bo else 43760b246afaSJeff Mahoney unpin = &fs_info->freed_extents[0]; 4377ed0eaa14SLiu Bo loop = false; 4378ed0eaa14SLiu Bo goto again; 4379ed0eaa14SLiu Bo } 4380ed0eaa14SLiu Bo 4381acce952bSliubo return 0; 4382acce952bSliubo } 4383acce952bSliubo 4384c79a1751SLiu Bo static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) 4385c79a1751SLiu Bo { 4386c79a1751SLiu Bo struct inode *inode; 4387c79a1751SLiu Bo 4388c79a1751SLiu Bo inode = cache->io_ctl.inode; 4389c79a1751SLiu Bo if (inode) { 4390c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4391c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4392c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4393c79a1751SLiu Bo iput(inode); 4394c79a1751SLiu Bo } 4395c79a1751SLiu Bo btrfs_put_block_group(cache); 4396c79a1751SLiu Bo } 4397c79a1751SLiu Bo 4398c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 43992ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4400c79a1751SLiu Bo { 4401c79a1751SLiu Bo struct btrfs_block_group_cache *cache; 4402c79a1751SLiu Bo 4403c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4404c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4405c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 4406c79a1751SLiu Bo struct btrfs_block_group_cache, 4407c79a1751SLiu Bo dirty_list); 4408c79a1751SLiu Bo 4409c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4410c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4411c79a1751SLiu Bo list_del_init(&cache->io_list); 4412c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4413c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4414c79a1751SLiu Bo } 4415c79a1751SLiu Bo 4416c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4417c79a1751SLiu Bo spin_lock(&cache->lock); 4418c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4419c79a1751SLiu Bo spin_unlock(&cache->lock); 4420c79a1751SLiu Bo 4421c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4422c79a1751SLiu Bo btrfs_put_block_group(cache); 4423c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4424c79a1751SLiu Bo } 4425c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4426c79a1751SLiu Bo 442745ae2c18SNikolay Borisov /* 442845ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 442945ae2c18SNikolay Borisov * to use it without any locking 443045ae2c18SNikolay Borisov */ 4431c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4432c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 4433c79a1751SLiu Bo struct btrfs_block_group_cache, 4434c79a1751SLiu Bo io_list); 4435c79a1751SLiu Bo 4436c79a1751SLiu Bo list_del_init(&cache->io_list); 4437c79a1751SLiu Bo spin_lock(&cache->lock); 4438c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4439c79a1751SLiu Bo spin_unlock(&cache->lock); 4440c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4441c79a1751SLiu Bo } 4442c79a1751SLiu Bo } 4443c79a1751SLiu Bo 444449b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 44452ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 444649b25e05SJeff Mahoney { 44472ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4448c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4449c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4450c79a1751SLiu Bo 44512ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 445249b25e05SJeff Mahoney 44534a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 44540b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 445549b25e05SJeff Mahoney 44564a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 44570b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 445849b25e05SJeff Mahoney 4459ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4460ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 446167cde344SMiao Xie 44622ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 446349b25e05SJeff Mahoney EXTENT_DIRTY); 44642ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, 44650b246afaSJeff Mahoney fs_info->pinned_extents); 446649b25e05SJeff Mahoney 44674a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 44684a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 446949b25e05SJeff Mahoney } 447049b25e05SJeff Mahoney 44712ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4472acce952bSliubo { 4473acce952bSliubo struct btrfs_transaction *t; 4474acce952bSliubo 44750b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4476acce952bSliubo 44770b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 44780b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 44790b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4480724e2315SJosef Bacik struct btrfs_transaction, list); 4481724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 44829b64f57dSElena Reshetova refcount_inc(&t->use_count); 44830b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 44842ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4485724e2315SJosef Bacik btrfs_put_transaction(t); 44860b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4487724e2315SJosef Bacik continue; 4488724e2315SJosef Bacik } 44890b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4490724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 44910b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4492724e2315SJosef Bacik /* 4493724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4494724e2315SJosef Bacik * handle open currently for this transaction. 4495724e2315SJosef Bacik */ 4496724e2315SJosef Bacik wait_event(t->writer_wait, 4497724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4498724e2315SJosef Bacik } else { 44990b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4500724e2315SJosef Bacik } 45012ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4502724e2315SJosef Bacik 45030b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 45040b246afaSJeff Mahoney if (t == fs_info->running_transaction) 45050b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4506724e2315SJosef Bacik list_del_init(&t->list); 45070b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4508a4abeea4SJosef Bacik 4509724e2315SJosef Bacik btrfs_put_transaction(t); 45102ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 45110b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4512724e2315SJosef Bacik } 45130b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 45140b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4515ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4516ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 45172ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 45180b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 45190b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4520acce952bSliubo 4521acce952bSliubo return 0; 4522acce952bSliubo } 4523acce952bSliubo 4524e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 45254d53dddbSDavid Sterba /* mandatory callbacks */ 45260b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 45274d53dddbSDavid Sterba .readpage_end_io_hook = btree_readpage_end_io_hook, 452820a7db8aSDavid Sterba .readpage_io_failed_hook = btree_io_failed_hook, 45294d53dddbSDavid Sterba 45304d53dddbSDavid Sterba /* optional callbacks */ 45310da5468fSChris Mason }; 4532